5518 en

European Centre for the Development
of Vocational Training
RESEARCH PAPER
No 18
Building on skills
forecasts — Comparing
methods and
applications
Conference proceedings
Luxembourg:
Publications Office of the European Union, 2012
Building on skills forecasts —
Comparing methods and
applications
Luxembourg: Publications Office of the European Union, 2012

A great deal of additional information on the European Union is available on the
Internet.
It can be accessed through the Europa server (http://europa.eu).
Cataloguing data can be found at the end of this publication.
Luxembourg: Publications Office of the European Union, 2012
ISBN 978-92-896-0892-3
ISSN 1831-5860
doi: 10.2801/36961
© European Centre for the Development of Vocational Training, 2012
All rights reserved.
The European Centre for the Development
of Vocational Training (Cedefop) is the European Union’s
reference centre for vocational education and training.
We provide information on and analyses of vocational education and
training systems, policies, research and practice.
Cedefop was established in 1975
by Council Regulation (EEC) No 337/75.
Europe 123, 570 01 Thessaloniki (Pylea), GREECE

PO Box 22427, 551 02 Thessaloniki, GREECE
Tel. +30 2310490111, Fax +30 2310490020
E-mail: info@cedefop.europa.eu
www.cedefop.europa.eu
Christian F. Lettmayr, Acting Director

Hermann Nehls, Chair of the Governing Board
Building on skills forecasts — Comparing methods and applications
Foreword
In recent years, Cedefop has provided important evidence on future skill needs.
Cedefop’s 2010 skill forecasts and its 2011 update provide the most recent
indications of the future patterns of skill supply and demand needs in the
European labour market, examining closely the impact of the recession. To
increase the quality of the new skill demand and supply forecast, which will be
published in 2012, Cedefop continues working on further developments of the
modelling framework.
This publication discusses how to best forecast skill supply and demand and
how to improve the Cedefop approach. The publication examplifies Cedefop’s
commitment not only to provide evidence for policy-making but also to support
active exchange of views and ideas among experts. It is based on the
contributions from the international expert conference organised by Cedefop in
February 2011, which brought together more than 60 researchers and experts in
labour-market forecasting. This event aimed at gathering insights on how
Cedefop’s forecast is used at national level, and what other (innovative)
forecasting activities are taking place in the individual Member States. It provides
sound evidence and suggests innovative ways of skills forecasting. The findings
support improvements of Cedefop’s forecasting methodology.
During the conference, experts touched upon all the main elements of the
basic forecasting approach and conveyed several important messages. First, it
was confirmed that even if the scope and methods of the forecasting exercise
differ across countries, the results produced are in line with Cedefop’s forecast.
Better quality data sets and better insight into national policies may however
influence the precision of the trends obtained and thus their direct relevance for
policy. Second, the forecasting results are important for further research (for
instance, to assess the contribution of skills to productivity and economic growth)
but they are especially valuable as additional labour-market intelligence (i.e. in
public employment services and for guidance and counselling) and provide
evidence for policy decisions. Third, all participants agreed that continuous
improvements of methods and tools for skills forecasting is important
Cooperation with stakeholders is necessary to achieve progress and to
ensure that these objectives can be realised. Key challenges identified for future
forecasting include refined methods for the assessment of the consequences of
migration (scenario); a more detailed interpretation of qualification and skills
levels; a more exact definition of imbalances between supply and demand as well
as innovative ways of reflecting the differences on regional labour markets.
Cedefop is committed to address these challenges.
Christian F. Lettmayr
Cedefop Acting Director
1
Acknowledgements
This publication results from proceedings of the international expert conference

Building on the skills forecasts: comparing methods and applications organised
by Cedefop in February 2011. Thanks go to Vladimir Kvetan, Alena Zukersteinova,
Antonio Ranieri and Konstantinos Pouliakas who compiled and edited the
chapters and Pascaline Descy who supervised the publication. Cedefop would
also like to acknowledge all those who contributed to the conference and this
publication.
2
Note from the editors
An agenda for new skills and jobs, one of the flagship initiatives of the Europe
2020 strategy, aims at modernising labour markets and empowering people by
developing their skills throughout the lifecycle with a view to increase labour-
market participation and to match better labour supply with demand. Following
the Council conclusions of 7 June 2010 on New skills for new jobs: the way
forward, Cedefop is mandated to provide a pan-European skill supply and
demand forecast every two years.
Cedefop started its forecasting exercise with two pilot projects focused on
the skills demand in 2008 and labour-market skills supply in 2009. The outcome
published by Cedefop in 2011 combines and presents the first pan-European
projection of skills supply and demand for 27 Member States of the European
Union, plus Norway and Switzerland.
Forecasting is a dynamic exercise and needs to be continuously developed.
To obtain feedback from international experts and scientific community Cedefop
organised a technical conference, Building on the skills forecasts: comparing
methods and applications, in February 2011 in Thessaloniki. The conference
brought together an international forum of researchers and experts in anticipating
labour-market as well as skill demand and supply. It gathered the latest insights
into use of the Cedefop forecast and on other (innovative) processes in different
Member States and helped to promote skills supply and demand among different
experts in the field.
This publication, collecting selected papers presented at the conference,
helps to identify the challenges and to offer solutions to skills and labour-market
forecasting in general and Cedefop’s in particular; it also discusses various
national approaches, methods and results and their comparisons to Cedefop
forecasts. We hope that this publication will be a step forward in increasing
transparency and sharing expertise in skills anticipation. We wish to see,
however, many more such steps helping researchers and practitioners to
exchange their views. Cedefop will continue to contribute to this process.
Vladimir Kvetan
Konstantinos Pouliakas
Antonio Ranieri
Alena Zukersteinova
3
Table of contents
Foreword ............................................................................................................... 1
Acknowledgements ............................................................................................... 2
Note from editors................................................................................................... 3
1. Cedefop’s skills supply and demand forecast: 2011 update and
reflections on the approach
Vladimir Kvetan, Rob Wilson and Alena Zukersteinova ............................... 11
2. Lessons learned: how to improve Cedefop’s forecast
Ben Kriechel and Hector Pollitt ..................................................................... 26
PART 1: BUILDING ON CEDEFOP’S FORECASTING RESULTS .................................. 36

3. Trends, drivers and qualitative scenarios in Europe 2020: for education
and training policy-makers engaged with meeting skills needs
Tom Leney ............................................................................................. 37
4. Polarisation or segmentation? Occupational change and
the implications for the future EU labour market
Antonio Ranieri and Marco Serafini .............................................................. 49
5. Use of forecast results in skills transferability project
Jiří Braňka ............................................................................................. 66
PART 2: SHARING OTHER EUROPEAN FORECASTING EXPERIENCES ......................... 81

6. The AMS-skills barometer: a web-based labour-market tool. Making
use of forecast results by stakeholders
Stefan Humpl and Daniel Bacher ................................................................. 82
7. Forecasting skills demand and labour market dynamics in the Baltic
States
Žilvinas Martinaitis ........................................................................................ 97
8. Demand and supply of labour by education in Norway towards 2030
― Linking demographic and macroeconomic models
Roger Bjørnstad, Marit L. Gjelsvik, Anna Godøy, Inger Holm and
Nils Martin Stølen ....................................................................................... 111
9. Methods and results of skills demand and supply forecasting —
the case of Germany
Tobias Maier ........................................................................................... 129
10. The FGB-LM model: structure and recent forecasts of the Italian
labour-market stocks and flows
Giuseppe Ciccarone and Massimiliano Tancioni ........................................ 147
4
11. The forecast of occupational structure of employment in Poland

Artur Gajdos ........................................................................................... 169
12. Forecasting the development of the educational level in Switzerland
Jacques Babel ........................................................................................... 181
13. Education — training complementarity in Finland
Erkki Laukkanen ......................................................................................... 195
14. How to anticipate the evolution of qualifications and training needs:
a regional point of view
Marie-Béatrice Rochard .............................................................................. 209
15. Matching formal skills in Slovakia — LFS-based evidence
Marek Radvanský ....................................................................................... 224
16. Internet job search data as a possible source of information on skills
demand (with results for Slovak university graduates)
Miroslav Štefánik ........................................................................................ 246
17. The protective effect of field-specific and general skills against
overeducation under different conditions of labour supply and demand
Martin Humburg, Andries de Grip and Rolf van der Velden ....................... 261
List of Contributors ............................................................................................ 275
5
List of tables and figures

Tables
1:1 Main impacts on employment in the different scenarios ............................ 21
4:1 Sectoral trends and polarisation, EU-27*, 2000-08.................................... 55
5:1 Definition of major skills categories (a) ....................................................... 70
5:2 Analysis of forecast approaches and their application ............................... 73
5:3 An example of analysis of highly transferable skills across sectors .......... 73
5:4 The largest occupational groups in retail trade sector ............................... 76
5:5 Growth/decline rates .................................................................................. 77
5:6 Growth/decline rates in occupations .......................................................... 77
5:7 Example of additional outputs .................................................................... 79
7:1 Forecasted and actual number of employed persons .............................. 104
7:2 Forecasted and actual number of employed persons by occupational
groups .................................................................................................... 105
7:3 Forecasted and actual number of employed persons in selected
sectors .................................................................................................... 106
7:4 Forecasted and actual number of employed persons by qualification ..... 107
8:1 Classification by main group of education and corresponding
numbers of employed in 2006 (1 000 persons) ....................................... 114
9:1 Structural information for demand and supply forecasting....................... 132
9:2 Demand and supply of labour (in millions) ............................................... 135
9:3 Results of qualification projections (in millions) ....................................... 136
9:4 Occupational mobility (in %) .................................................................... 139
9:5 Comparison of results between BIBB-IAB and Cedefop forecasts:
demand side ............................................................................................ 141
9:6 Comparison of results between BIBB-IAB and Cedefop forecast:
supply side ............................................................................................... 143
10:1 Predicted annual growth rates for the major macroeconomic
aggregates ............................................................................................... 159
10:2 Predicted outflows and hiring (thousands), 2010-14 ............................... 160
10:3 Employment stock by qualification (thousands), 2010-14........................ 162
10:4 Hiring by qualification (thousands), 2010-14 ........................................... 163
10:5 Employment stock by occupation (thousands), 2010-14 ......................... 165
10:6 Hiring by occupation (thousands), 2010-14 ............................................. 165
11:1 Location quotients in 2008 ....................................................................... 177
11:2 Shift-share analysis in 2004-08 (major ISCO groups) ............................. 177
13:1 The data ................................................................................................... 201
13:2 Marginal probabilities to positive number of training days (%)................. 202
13:3 Determinants of training days (%)............................................................ 205
14:1 Example of ‘structural’ professions in need ............................................. 219
14:2 Example of ‘cyclical’ professions in need ................................................ 219
15:1 Conversion of occupation classification (ISCO) according to required
formal skills level (education) ................................................................... 226
6
15:2 Conversion of classification of education classes (ISCED) ..................... 227

16:1 Shares of occupational groups of tertiary educated in employment,
advertisements and CVs .......................................................................... 250
16:2 Shares of economic sectors on employment of tertiary educated,
advertisements and CVs .......................................................................... 251
16:3 Selected combinations of occupation and economic sector and their
share in the total population ..................................................................... 252
16:4 Skills frequently mentioned in advertisements looking for technicians
in public services ..................................................................................... 254
16:5 Skills ranked according to the Wald Chi-Square in the generalised
linear model applied to technicians in public services ............................. 255
16:6 Skills frequently mentioned in advertisements looking for
professionals in construction ................................................................... 256
16:7 Skills ranked according to the Wald Chi-Square in the generalised
linear model applied to professionals in construction .............................. 257
17:1 Overall unemployment rates and unemployment rates in the
occupational domain of particular fields of study per country .................. 267
17:2 Linear probability model of being overeducated five years after
graduation ................................................................................................ 269
Figures
1:1 Conceptual framework of modelling the demand for and supply of
skills ...................................................................................................... 15
1:2 Impact of the recession on employment, EU-27+ ...................................... 16
1:3 Past and forecast employment changes by economic sector, EU-27+ ..... 17
1:4 Future job opportunities by occupation, EU-27+........................................ 18
1:5 Future job opportunities by broad level of qualifications, EU-27+.............. 19
1:6 Labour supply by level of qualifications (EU-27+) ...................................... 20
2:1 Overview of (Cedefop) existing approach .................................................. 31
2:2 Incorporation of supply side? ..................................................................... 34
4:1 Projected change in share in occupational structure, EU-27, 2010-20 ...... 50
4:2 Share % of highly skilled and elementary in total occupation, EU-27
(1970=100) ................................................................................................ 54
4:3 Trends in the share of occupations, in EU-27, 2000-08 ............................ 56
4:4 Shift-share groups: structural and occupational-specific growth
components ............................................................................................... 57
4:5 Main socioeconomic characteristics in elementary occupations, in-
differences analysis between ISCO 9 and total occupations EU-27 .......... 59
4:6 Change % in employment by occupation and nationality, EU- 27,
2000-08 ...................................................................................................... 60
4:7 Share of medium and high qualifications in total employment,
EU-27, 2000-08.......................................................................................... 61
7
4:8 Concentration index of non-national workers in elementary

occupations by qualification, EU-27, 2008 ................................................. 62
5:1 Groups of sectors analysed in the project (employment in thousands
within the EU)............................................................................................. 68
5:2 Possible tools for identification of future skills and their transferability ...... 80
6:1 Basic structure of occupations and skills in the AMS-skills barometer ...... 85
6:2 Occupational area ‘information technology’ ............................................... 86
6:3 Occupational field software technology and programming ........................ 87
6:4 Occupational field software technology and programming ........................ 89
7:1 Real GDP growth in the Baltics................................................................ 102
8:1 Main structure of the model system ......................................................... 113
8:2 Employment by sector, share of total employment .................................. 117
8:3 Employment by level of education as a percentage of total
employment, 1986-2030 .......................................................................... 120
8:4 Employment by education at upper secondary school level,
1 000 persons .......................................................................................... 121
8:5 Employment by education at tertiary level with a lower degree,
1 000 persons .......................................................................................... 122
8:6 Employment by education at tertiary level with higher degree,
1 000 persons .......................................................................................... 124
8:7 Population by level of education by 1 000 persons.................................. 126
8:8 Labour supply by education from Modag compared to adjusted
labour supply from Mosart ....................................................................... 127
9:1 Total demand in MOF and supply of trained persons (ISCED 3b-6) ....... 137
9:2 Demand and supply including occupational mobility (ISCED 0-6) ........... 140
10:1 Macroeconomic environment. Forecast period 2010:3-2015:3................ 158
10:2 Labour market stocks and unemployment rate (2010:3-2015:3) ............. 160
10:3 Employment and hiring by age class, 2010 and 2014 ............................. 161
10:4 Employment and hiring by qualification, 2014 ......................................... 164
11:1 LDFS — Structure.................................................................................... 171
11:2 LDFS — Database ................................................................................... 171
11:3 LDFS — Forecasting ............................................................................... 172
11:4 Forecast errors in 2010 (major ISCO groups).......................................... 172
11:5 Cedefop and LDFS — Forecast comparison ........................................... 174
11:6 Sigma-convergence, 2000-08 .................................................................. 175
11:7 Major occupational groups — Spatial concentration in 2008 ................... 176
11:8 Forecast for major occupational groups — Lublin region ........................ 178
11:9 Forecast for major occupational groups — Wielkopolska region ............. 179
12:1 2010-60 scenarios for the population development: level of education
of population aged 25 to 64 in Switzerland .............................................. 185
12:2 Percentage of high qualification in the Swiss population, by age ............ 186
12:3 Number of degrees awarded in the upper secondary programmes
(ISCED 3-4, medium qualification)........................................................... 188
12:4 Number of first degrees awarded in the Swiss universities (ISCED
5A-6, high qualifications) ......................................................................... 190
8
12:5 Number of first degrees from professional education and training

programmes (ISCED 5B, high qualification) ............................................ 191
12:6 Level of education of the population of Switzerland as a function of
the data source used for the flow ............................................................. 192
13:1 Days in vocational adult education and training by skill level during
the last 12 months ................................................................................... 197
13:2 Education — training and training — training complementarity over
time, as measured by the effect of education years and training
courses on training days (%) ................................................................... 206
14:1 Employment projections by 2020 ............................................................. 215
14:2 Net job creations and retirements by 2020 .............................................. 216
14:3 Recruitment needs (retirement 62/65) ..................................................... 218
15:1 Development of employees by the classification of occupation
(ISCO) .................................................................................................... 227
15:2 Development of economically active population classes according to
classification of education (ISCED) ......................................................... 228
15:3 Scheme of basic matching skills analyses by level of education ............. 229
15:4 Scheme of matching analysis of required skills and obtained
education level ......................................................................................... 230
15:5 General overview of formal skills matching (2009) as % of total
employment ............................................................................................. 231
15:6 Share of employees with skills matched according to obtained and
required level of education and age group .............................................. 232
15:7 Share of employees with lower level of education than required by
the actual position by age group — undereducated ................................ 233
15:8 Share of employees with higher level of education than required by
the actual position by age group — overeducated................................... 233
15:9 Share of highly educated in the economically active population
disaggregated by region .......................................................................... 234
15:10 Share of jobs according to standard classification of occupations and
share of employees according to classification of education at Slovak
NUTS III regions (2009) ........................................................................... 235
15:11 Share of employees according to classification of occupation and
classification of education according to selected sectors by SK NACE
classification (2009) ................................................................................. 238
15:12 Formal skills match at mining and quarrying, manufacturing,
electricity, gas and water supply sector (C, D, E), 2009 .......................... 239
15:13 Formal skills match in construction sector (F), 2009 ............................... 240
15:14 Formal skills match public administration (L-O), 2009 ............................. 241
15:15 Share of employees with formal skills match by selected SK NACE
sector ............................................................................................ 241
15:16 Share of employees with lower education level than required in
occupation at selected SK NACE sector — undereducated .................... 242
15:17 Share of employees with higher education level than required in
occupation at selected SK NACE sector — overeducated ...................... 242
9
15:18 Share of overqualified with high education, 2009 .................................... 243

17:1 Incidence of overeducation among college graduates five years after
graduation (%) ......................................................................................... 265
17:2 Coefficient of general skills and overall unemployment rate (%) ............. 267
17:3 Coefficient of field-specific skills and unemployment rate in
occupational domain of field of study ....................................................... 268
Boxes
3:1 Summary of likely trends in labour market skills demand and supply
(2010 to 2020) ........................................................................................... 40
4:1 Possible drivers of change in occupational structure ................................. 52
4:2 Shift-share groups interpretation ............................................................... 58
14:1 The centre region, near the Paris region, between industry and
rural life .................................................................................................... 213
10
CHAPTER 1.
Cedefop’s skills supply and demand
forecast: 2011 update and reflections on the
approach
Vladimir Kvetan, Rob Wilson and Alena Zukersteinova
The Cedefop skills supply and demand forecast is one of the important
contributions towards the coordinated anticipation of skills needs at
European level. The forecasting approach used by Cedefop has been
developed over the past few years and enabled the research team to
publish in 2010 a first pan-European cross-country comparable projection
of Skills supply and demand in Europe (Cedefop, 2010). This chapter
presents the overview of current methodological framework and approach
as well as the results of the skills demand and supply forecast released
early 2011 (1). Description of four different alternative scenarios of possible
future development of the European labour market which were developed
alongside the main baseline scenario can be considered as a unique
contribution to the discussion on the paths to recovery. The concluding part
discusses the way forward and different constraints faced.
1.1. Introduction
The European labour market is fast becoming a reality. This requires the
identification of occupations, skills, competences and qualifications, which will be
in demand in the future at pan-European level. This information is now even
more urgent for setting up the strategy to speed up the recovery process after the
end of the global economic crisis. Finding ways to obtain consistent and
comprehensive information on future skill demand, as well as supply, in Europe
— including joint European action — is a priority stated in key European strategic
documents.
In 2005, Cedefop started to explore the feasibility of establishing a European
skill needs and supply forecasting system. The main rationale was to fill the gap
in information on future skill needs at pan-European level. Today Europe faces a
situation where, despite increasing unemployment rates, there are still significant
numbers of vacancies not easy to fill. Forecasting skill needs involves estimating
(1) This article was sent for publication in spring 2011. In the meantime, Cedefop has
produced a new set of forecasting results. The new forecast confirms previous
general trends on development, though some detailed numbers may differ.
11
the expected future number of jobs available in an economy and their particular
skill or qualification requirements. Skills needs forecasts are complemented by
forecasts of the number of people (supply) with particular skills. The comparison
of demand and supply can indicate potential imbalances or skill mismatches in
future labour markets in the EU, as well as in particular Member States.
The first pan-European forecast of skill needs providing consistent and
comprehensive medium-term projections of employment and skill demand across
Europe up to 2015 was published by Cedefop in 2008 (Cedefop, 2008). In 2009,
this forecast was complemented by the first medium-term projection of skill
supply in Europe up to 2020 (Cedefop, 2009). At the beginning of 2010, the first
comprehensive forecast of both skills supply and demand was published
(Cedefop, 2010). Cedefop is continuing to work on the system developed there,
to produce regular forecasts, integrating skills supply and demand and providing
indications of skill imbalances. The 2011 update presented in this chapter
provides latest results of skills demand and supply forecasts. The projections
were carried out in 2010 and published early 2011 (Cedefop, 2011a).
Section 1.2 describes the main features of Cedefop’s methodology. It
presents the general platform and focuses on key characteristics of different
models, modules and methods used. The main outcomes and results from the
2011 update are detailed in Section 1.3. This provides information on the key
trends in labour supply by qualification and in labour demand by occupation,
sector and qualification. Conclusions and general recommendations for policy-
makers are presented in Section 1.4.
When interpreting the forecasting results it is important to keep in mind the
‘golden rule’: the forecasts do not provide a ‘crystal ball’ showing precise
information about the future. Rather they serve as a source for informed decision-
making, and an early warning for various labour-market actors on what might
happen in the future.
1.2. Methodological framework and approach

The conceptual/methodological framework (Figure 1:1) for the development of
skills supply and demand forecasts is based on a modular approach, which
enables continuous improvements and further development of particular
elements. It also allows the input data and key assumptions, to be changed
easily for updating or developing different alternative scenarios.
12
The first module is a multi-sectoral macroeconomic model (E3ME) which is

used to drive the forecast. E3ME is a pan-European multi-sectoral model
developed by Cambridge Econometrics and others (Pollitt et al., 2010) (2). This
model provides a set of consistent employment forecasts by sector and country.
The module combines elements of an annual econometric model and a medium-
term sectoral model. The model produces labour demand forecasts for 42
economic sectors (compatible with ESA 95 and NACE). Labour supply is
disaggregated by gender and five-year age groups. Eurostat (national accounts)
data, as well as some OECD data from the STAN database are used to obtain
consistent and commonly comparable forecasts for EU-27 Member Sates plus
Norway and Switzerland (EU-27+). The forecast is calibrated to match the latest
forecast from the Directorate General for Economic and Financial Affairs (DG
EcFin) of the European Commission.
Expansion demand by occupation is calculated in the second module
(EDMOD), using harmonised data from EU labour force survey. This set of data
includes consistent multi-sectoral information which is comparable across
countries. The data enable the construction of sector by occupation employment
matrices. Connecting these matrices and the outputs from E3ME provides results
for labour demand by occupation. Demand is proxied by the number of jobs. This
module is based on econometric methods which enable the introduction of
dynamics to the different elements of the matrix (the detailed description of this
module can be found in Livanos and Wilson, 2010a).
Expansion demand by qualifications (QMOD) is based on similar data and
principles to EDMOD. Qualifications can be regarded as individual characteristics
to fill jobs, as well as criteria used in the selection process. Information about
qualification requirements is an important input from the educational and VET
policies point of view. Due to data limitations the current level of disaggregation is
quite restricted to three broad formal qualification levels (low, medium and high).
The design of this module enables, in principle, further disaggregation of different
types of educational levels, skills and qualifications. In practice, the key limitation
remains the availability, quality and range of the data (detailed discussion is
provided in Livanos and Wilson, 2010a).
Replacement demand (RDMOD) represents a key element in skill demand,
concerning the need to replace those who leave labour force for different reasons
(retirement, migration, etc.). To estimate replacement demand is not a simple
(2) More detailed information is available from Internet:

http://www.camecon.com/ModellingTraining/suite_economic_models/E3ME.aspx
[cited 27.7.2011].
13
task and is crucially dependant on quality of input data. In an ideal situation this
module should be based on information on the outflows from the labour market
including inter-occupational mobility. Such data are not generally available at
pan-European level, therefore the current estimates are based on information
available on the supply of the labour force by age cohort (and inter-occupational
mobility is largely neglected) (more details of the methodology are described in
Kriechel and Sauerman, 2010).
The supply of skills is measured by the highest formal qualification achieved
by individuals. Consistent pan-European forecasts using existing data
disaggregated by age, gender and formal qualification (3) are produced. This
depends on several modules.
An augmented version of E3ME* model delivers projections of overall labour
supply as the function of economic activity, real wages, unemployment rates and
benefit rates. At present, the model parameters are estimated for different
countries, age groups and gender. This is also of key importance for modelling
educational participation and attainment since these are known to be gender and
age specific. This expanded model framework is then used to create a detailed
set of baseline projections for labour supply over a 10 to 15 year period (for more
details see Pollitt et al., 2010).
A full stock-flow model would represent an ideal tool for forecasting labour
supply by different educational types. Such a model would enable a record of
qualification development during the whole productive period of individuals to be
followed. This ideal state is not easy to achieve due to restricted availability of
data. Initial pilot analysis for a few countries suggests that it may be possible to
develop such models using data from the EU labour force survey. At present, the
main projections rely only on the overall trends in stocks in StockMod. This
module forecasts the qualification structure of labour supply (focusing on the
highest formal qualification achieved). Earlier work did focus on flows. FlowMod
analysed educational participation of young people and the initial acquisition of
qualifications, but this was based on partial and incomplete information (detailed
characteristics of both modules are described in Livanos and Wilson, 2010b).
(3) The forecast distinguishes three broad levels of qualifications based on ISCED
classification: low for ISCED 1 and 2; medium for ISCED 3 and 4 and high for ISCED
5 and 6.
14
Figure 1:1 Conceptual framework of modelling the demand for and supply of skills
Source: Cedefop (2010).
The interaction between labour supply and demand by qualifications is a

complex matter. Many jobs are filled by individuals with qualifications different
from the ‘norm’. The model that confronts demand and supply by considering this
issue is BalMod. BalMod distributes the available labour supply holding certain
qualifications into jobs, based on assumptions on trends in employment patterns
at detailed level, plus assumptions about unemployment rates by broad level of
qualifications. Final adjustments are made considering double jobbing;
differences between the place of residence and place of work; participation in
training; different definitions of unemployment; and statistical discrepancies (for
more information see Kriechel and Wilson, 2010).
15
1.3. Results: 2011 update on future skills supply and

demand for EU-27+
1.3.1. Baseline results

The Cedefop forecast of skills demand and supply in Europe published in the
2011 update (Cedefop, 2011a) suggests that the initial assessment of the impact
of the recession on employment in the 2010 report was reasonably accurate
(Figure 1:2). The crisis has significantly reduced the number of new job
opportunities. There are probably around 10 million fewer jobs across Europe
than if there had been no crisis.
Figure 1:2 Impact of the recession on employment, EU-27+
No crisis scenario
2010 baseline
Millions jobs 2011 baseline
245
240
235
230
225
220
215
210
205
2000 2005 2010 2015 2020 2025
Source: Cedefop (2011a).
At sectoral level, the recession appears to have continued the shift in jobs
from primary and basic manufacturing to services (Figure 1:3). Productivity gains
are reducing costs and improving competitiveness, but can also cut the numbers
of workers required to produce the same or even higher levels of output. The
2011 update shows that primary industries are projected to lose slightly fewer
jobs than expected in the previous forecast. Many will be in agriculture, which is
still a significant source of jobs in some countries and regions. Manufacturing
was hit less hard by the crisis than expected. This points to the continuing
importance of manufacturing to Europe’s economy and indicates that some jobs
are being saved, or relocated within rather than outside the EU. Services are still
expected to provide most of the job growth between now and 2020. However, the
16
2011 update is less optimistic about the number of jobs the service sector will
create by 2020. This is partly due to austerity measures and cutbacks in public
and private spending and investment.
Figure 1:3 Past and forecast employment changes by economic sector, EU-27+
2000-10 (historical data)
2010-20 (2010 baseline)
2010-20 (2011 baseline)
All industries
Primary sector & utilities
Manufacturing
Construction
Distribution & transport
Business & other services
Non-marketed services
-6 -4 -2 0 2 4 6 8 10 12
Millions jobs
Over the next 10 years (to 2020), some 81 million job opportunities are
projected. This includes around eight million new jobs (expansion demand). The
other job opportunities arise due to the necessity to replace those leaving the
labour force for various reasons (retirement, migration, etc.).
The ‘skill intensity’ of jobs has increased recently and this is expected to be
a continuing feature. Most of the projected net increases in employment levels up
to 2020 are concentrated in higher level occupations such as management,
professional and associate professional jobs (around eight million in total
between 2010 and 2020; Figure 1.4). However, significant employment increases
are also expected for some lower level occupations involved in service activities
such as sales, security, cleaning, catering and caring (around two million in total).
Job losses are projected for some skilled manual workers and clerks (over five
million in total). There are continued indications of possible polarisation in the
demand for skills, with significant growth in employment in both high level
occupations such as managers and professional as well as many less skilled
occupations in the service sector. These trends might have significant
17
implications for issues of social exclusion, equality and job quality (for more on
job polarisation see Cedefop, 2011b).
The net changes in employment by occupation are only part of the story.
When replacement needs due to retirement or other reasons are considered they
reinforce the projected net change. Replacement needs are much more
significant in terms of numbers than new job opportunities. As a result, total job
openings are positive for all major occupational groups (as older workers retire
and need to be replaced). Even for the skilled agricultural and fishery workers
group, where the most negative employment prospects are expected, there will
be some job openings between 2010 and 2020 following the need to replace at
least some of the current labour force. For many other occupations the numbers
are even larger, notably for managerial, professional and associate professional
and technical groups (with increases of between 10 and 15 million jobs for each
group). There are also significant increases projected for some of the least skilled
occupational groups, including the service and sales workers group, craft and
related workers and even in elementary occupations (each with around 10 million
job openings; Figure 1.4).
Figure 1:4 Future job opportunities by occupation, EU-27+
Net change
Replacement demand
Total requirement
Legislators, senior officials and managers
Professionals
Technicians and associate professionals
Clerks
Service, shop and market sales workers
Skilled agricultural and fishery workers
Craft and related trades workers
Plant and machine operators and assemblers
Elementary occupations
-5 0 5 10 15
Million jobs
Many future jobs will require a better and more highly skilled workforce. The
proportion of jobs employing highly qualified people is projected to increase to
over a third. In contrast, the proportion of those requiring low or no formal
18
qualifications is projected to decrease. Those jobs requiring intermediate level

qualifications will continue to constitute around half of all jobs.
This increase in average qualification levels is in part a supply phenomenon,
driven by strong cohort effects and the interaction between supply and demand.
The need for better qualifications is also in line with the need to ensure
better productivity and use of modern technology.
Figure 1:5 Future job opportunities by broad level of qualifications, EU-27+
Net change
Replacement demand
Total requirement
High qualification
Medium qualification
Low qualification
-20 -10 0 10 20 30 40
Millions of jobs
Future labour supply will be influenced by two important facts:

(a) the increase in the average age of the working population: future labour
force is expected to grow only in the age group 45+;
(b) higher educational attainment of younger age cohorts: younger people tend
to stay longer in education, they should acquire better qualifications raising
the average qualification levels of those in employment as they replace more
experienced workers leaving the labour market due to retirement.
Figure 1.6 presents the apparently inexorable rise in the proportions of the
workforce with medium and high formal qualification, with especially large
increases at the highest level, and sharp reductions in the shares of the
workforce with no or low formal qualifications.
There is evidence that in many countries the recession has encouraged
participation in further training and higher education in the short term. The long-
term sustainability of these trends is highly dependant on, for example, the state
of public finances, and the individual’s perceptions of the prospects of receiving a
positive return on such investments. Nevertheless, most countries appear on
19
track to meet the targets for improvements in the qualifications of their workforces
set at European level.
Figure 1:6 Labour supply by level of qualifications (EU-27+)
High
Medium
Milions Low
300
250
21 % 28 % 34 %
200
150 48 %
50 %
50 %
100
50 31 %
22 %
16 %
0
2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020
1.3.2. Alternative scenarios

One of the valuable contributions of the forecasting exercise is that it allows the
development of different scenarios describing the possible impact of certain
policy measures on the labour market. In 2011, alongside the base line forecast,
four alternative scenarios of the impact of possible policy measures taken were
calculated:
(a) the fiscal austerity scenario investigated the possible impact of announced
government fiscal policies and the impacts that they are likely to have on the
economy and employment. It included both changes to the level and pattern
of government spending and tax rates;
(b) the economic uncertainty scenario can be characterised by two exogenous
changes: a reduction in investment to reflect the uncertain nature of returns;
and a fall in the overall value of the euro;
(c) the retirement age scenario looked at the impact of higher participation rates
among the older age groups as a result of ageing populations and
government policies to raise official retirement ages;
20
(d) the ageing, health and pensions scenario investigated the impacts of higher
government spending on health care (including higher employment in the
sector), as well as higher government spending on pensions, as a result of
the ageing population trends.
It is important to separate the short-term and long-term effects. In the short

term, the scenarios can result in changes in aggregate employment rates. This is
particularly the case in the scenarios where confidence collapses, for example,
due to the austerity measures expected to be introduced in some countries. Even
so, changes in employment often lag the changes in economic output that result
from this.
Employment in the sectors that produce investment goods is most
vulnerable to general economic shocks, as output in these sectors tends to be
more volatile (investment is itself one of the most volatile components of GDP).
This has particular skills implications as there are often specific skills associated
with these sectors (e.g. engineering, motor vehicles and construction).
Table 1:1 Main impacts on employment in the different scenarios

Long-term
Short term Decreases
Increases
Fiscal austerity Overall reduction Private sector Public sector
Most sectors except Sectors that produce
Economic uncertainty Overall reduction
investment goods investment goods
Retirement age None Most sectors None
Ageing, health and
Increase in health Health sectors Most other sectors
pensions
Throughout the financial and economic crisis public sector employment has
been largely protected across much (but not all) of Europe. However, it now looks
much more vulnerable to cutbacks in spending in the future as governments try
to rebalance their economies and balance sheets.
The main exception to this is that, especially in the long term, employment in
the health sector is likely to increase by more than average, because of the
continued growth in demand for health services (associated with the increase in
the average age of the population).
In all the scenarios, in the non-public and non-investment goods sectors
there are employment losses. This leads to workers that will either withdraw from
the labour market or need to be reallocated. For example, in the scenario where
healthcare employment increases, this largely comes from attracting workers
from other sectors.
21
In general, the scenarios do not have much impact on the size of the labour
force, with one exception: if State retirement ages increase there will be more
older workers. Again this has both short and long-term implications; in the short
term it is likely to lead to higher unemployment, but if labour markets are allowed
to adjust, the long-term economic effects will be more beneficial.
The main implication of these results for policy is that, in the short term,
there could continue to be quite large reductions in employment levels compared
to the pre-crisis trend. Lagged effects mean that the impacts of the recession on
labour markets may not yet have fully worked through. In addition, impacts from
the possible policy responses that are now beginning to be implemented (such as
significant cuts in public expenditure) are yet to be felt. Any negative impacts are
likely to put further pressure on State budgets and welfare systems.
In the longer term, the results suggest that welfare will be maximised by
implementing policies that allow the transition of employment between sectors to
take place as efficiently as possible. For this, there is a need to allow wage rates
to adjust flexibly and to ensure that workers have the necessary skills to move
between sectors. The assumption in the scenarios is that workers will generally
have the necessary skills to make the transition. If this is not the case, and if
suitable retraining is not available, the macroeconomic outcomes could be
significantly worse.
1.4. Conclusions and the way forward

The work done to date proves the general feasibility of skill supply and demand
forecasting at European level, but many features can still be improved or
developed further. There are continuous requests for greater detail in the
forecasting results which leads back to the recurring issue of data problems and
constraints. The present forecasts use mainly data from the EU labour force
survey which while strong in terms of their availability and (in principle at least)
comparability across Europe, still have limitations. They were not designed for
modelling purposes. They are, in practice, often still inconsistent, both over time
and across countries and they generally provide only rather short time series.
The other key limitations are linked to the sample sizes, which are insufficient in
many countries to provide enough details by skills. There is only limited
information by occupation and qualification as a proxy for skills and no data at all
on other aspects of skills, (e.g. generic skills, etc.). Pushing the data to their limits
demonstrates the feasibility of providing some more detailed breakdowns but
further progress is likely to require more substantial investment in data at pan-
European level.
22
Many technical alternatives to the current modelling approach have been

examined and others are still in the pipeline. As an example, the desirable stock-
flow approach for the supply side shows some promise. This addresses the
cohort inconsistency problem but may be problematic for small countries. On the
demand side, the skills upgrading approach tells a broadly consistent story to the
simpler approach used in the QMOD module but is much more demanding in
terms of data requirements and cannot really substitute for the simpler methods
currently in use as it is difficult to generalise and apply it within the full, detailed
modelling framework. It is planned to explore more theoretical approaches
(similar to those used in computable general equilibrium [CGE] models), as well
as the scope for greater use of expert judgement to extend and enhance the
analysis.
The replacement demand module currently uses the cohort component
method, which is well established but still open to criticism. Alternative
approaches using panel data, which allow identification of individual decisions
and the different causes of replacement demand could be easily implemented in
countries where data are available. However, the estimates are very similar to
those from the cohort component approach, and moreover, such data are
currently not generally available across Europe. Therefore, it has been decided to
continue using the cohort component method until better data from Eurostat are
available.
When measuring potential future imbalances, simple comparison of demand
and supply is problematic for various reasons, including the difficulty in capturing
robust indicators of interaction. The results suggest that governments may need
to stimulate demand from employers and to find ways to increase the utilisation
of skills but more focused information is needed. The following indicators are
being developed to link occupational demand to educational supply: qualification
employment shares in occupations, indicators of constraint and change based on
the iterative procedure used to reconcile the supply and demand measures.
These will be further refined in the next forecasting round, with more detailed
distinction by field of education (depending on data availability).
More generally it may be possible to improve the results for individual
countries by making better use of country specific information rather than relying
solely on the pan-European data (using the same basic definitions, sources and
methods, but moving beyond the limitations imposed by reliance on the pan-
European labour force survey, using country specific information which may be
more reliable and more detailed).
The development of occupational skill profiles aims to bring the forecast
results closer to the actual skill demand by linking the results to characteristics
23
essential to a given occupation: the level of education and training required (and
hence the complexity of the occupation); the field of education and training
required; main and supplementary requirements concerning knowledge, skills,
personal abilities, attitudes and values. Occupational skill profiles for specific
occupations can be aggregated into those for occupational groups, further into
sectoral ones, then into those for national economies, and finally up to pan-
European level.
The Council conclusions of 7 June 2010 on New skills for new jobs: the way
forward (Council of the European Union, 2010), mandate Cedefop to update its
skill supply and demand forecast every two years. Though European level
forecasts are now available, there is still a lot to be done to develop various parts
of the model framework to address adequately important policy issues such as
imbalances and mismatches or migration. Therefore, the key plans for the future
include a further review and refinement of the data as well as (where possible)
additional detail for skill demand and supply. A full update of both the supply and
demand projections is planned every two years, including a thorough analysis of
mismatches and imbalances as well as development of more scenarios and their
interpretation to policy-makers. The next full update is now underway and will be
published in early 2012. In addition to the general improvements set out above, it
will include further development of models and scenarios to explore the possible
impact of climate change and related mitigation measures, exploration of
implications for productivity and growth, and further analysis of sensitivity of the
results to assumptions about migration. The work programme also includes
validation and further development of the occupational skills profiles and other
methodological developments. More ideas will arise from the various expert
contributions in this publication. In combination these elements should ensure the
provision of even better information to policy-makers and others about the
prospects for skill demand and supply in Europe in the future.
References
Cedefop (2008). Future skills needs in Europe: medium-term forecast: synthesis
report. Luxembourg: Publications Office of the European Union.
Cedefop (2009). Future skills supply in Europe: medium-term forecast up to
2020: synthesis report. Luxembourg: Publications Office of the European
Union.
Cedefop (2010). Skills supply and demand in Europe: medium-term forecast up
to 2020. Luxembourg: Publications Office of the European Union.
24
Cedefop (2011a). What next for skills on the European labour market? Luxembourg:
Publications Office of the European Union (Briefing note).
Cedefop (2011b). Labour market polarisation and elementary occupation in Europe:
blip or long-term trend? Luxembourg: Publications Office of the European
Union (Research paper No 9).
Council of the European Union (2010). Council conclusions on ‘New skills for
new jobs: the way forward. 3019th employment, social policy, health and
consumer affairs Council meeting, Brussels, 7 June 2010’. Brussels:
Consilium (Press).
Kriechel, B.; Sauerman, J. (2010). Replacement demand: methods and results.
Cedefop project on forecasting skill supply and demand in Europe, technical
report 005. Warwick: IER, Institute of employment research.
Kriechel, B; Wilson, R.A. (2010). Imbalance and mismatches: reconciling the
demand for and supply of skills. Cedefop project on forecasting skill supply
and demand in Europe. Technical report 008. Warwick: IER, Institute of
employment research.
Livanos, I.; Wilson, R.A. (2010a). Modelling the demand of skills. Cedefop project
on forecasting skill supply and demand in Europe. Technical report 002.
Warwick: IER, Institute of employment research.
Livanos, I.; Wilson, R.A. (2010b). Modelling of supply of skills. Cedefop project on
forecasting skill supply and demand in Europe. Technical report 003.
Warwick: IER, Institute of employment research.
Pollitt, H. et al. (2010). Projections of labour supply and demand from the E3ME
model. Cedefop project on forecasting skill supply and demand in Europe.
Technical report 001. Warwick: IER, Institute of employment research.
25
CHAPTER 2.
Lessons learned: how to improve Cedefop’s
forecast
Ben Kriechel and Hector Pollitt
As part of the new Europe 2020 strategy, Cedefop is mandated to produce

every two years a skill demand and supply forecast. Cedefop work is based
on the cooperation with important European research institutions and
universities. Setting up the broad platform for discussion is one of the key
priorities of Cedefop skill forecasting team. To share the knowledge and
obtain inspiration for further work, Cedefop organised the international
conference ‘Building on skills forecast: comparing methods and results’ in
Thessaloniki in February 2011. This chapter summarises key findings and
tries to indicate new insights and ideas for the pan-European model.
2.1. Introduction
As part of the new Europe 2020 strategy (European Commission, 2010), aims to
modernise the European labour market included the wish to gain insights into the
developments of skills supply and demand in the future. The influence of some of
the more important trends, such as changes in the development of work and the
greying of the workforce, are evaluated.
Cedefop started to play an active role in the process of development of tools
to improve information on future labour-market trends. The two pilot projects on
development of forecast for skills demand and skills supply (Cedefop, 2008;
2009) led to the publishing of first pan-European skills supply and demand
forecast for EU-27 Member States plus Norway and Switzerland (Cedefop,
2010). Use of single methodology and databases provides a consistent and
commonly comparable forecast across countries.
The development of such forecast is a continual process which is never
completed. The international conference on skills demand and supply forecasting
organised by Cedefop in Thessaloniki on 17 and 18 February 2011, summarised
in this chapter, had the aim to bring new insights and ideas into the pan-
European model, by discussing with international experts their views and
approaches to forecasting at national or international level.
26
This chapter provides an initial overview of the presentations (4), which are
also developed into independent chapters in this publication, by discussing
several aspects of their relation to the Cedefop pan-European model. First,
national forecasts presented at the conference are discussed. How they are
developed, which data they use and what goals they pursue can teach us both
about approaches to skill forecasting, and about the use of their results at
national level. Section 2.3 provides a summary of studies that have compared
results of the pan-European study with other sources, mainly national forecasts
or outcomes. They usually focus on the problems of the pan-European forecast
in catching national developments. Section 2.4 discusses extensions that are, be
it implicitly, suggested. Section 2.5 deals with the presentation of the results,
which is closely linked to the usability of results. Finally, some technical points
are raised in Sections 2.6-2.8.
2.2. National forecasts

In the group of ‘national forecasts’ the presentations on France (even though it is
a regional model), Germany, Italy, Norway, Poland and Switzerland are
summarised. Rather than summarise the contributions, the aim is to point to
aspects that either highlight a certain noteworthy angle of the forecasting
process, or can teach us how to (differently) approach the modelling process.
Nils Martin Stølen described the Norwegian forecasting model, a labour-
market projection model that exists in Norway since 1993. The model builds upon
rich administrative data sources that are typical for the Scandinavian countries.
Traditionally, the focus of the model is on the education side of the labour market.
Problems such a model has with migration are also noteworthy. Most information
comes from administrative data on schooling and labour-market participation, the
model had to be adjusted to fit in migrants that lack part of their education and
labour market history, as they were not in Norway at that point of time.
Tobias Maier presented the German forecasting, another national model
encompassing various modules determining supply and demand. It is a typical
example of how a national model can use the more detailed, national data, and
adapt the model to the institutional setting of the country. One of the current
challenges is the modelling process of the mismatch between supply and
demand. They use an approach to determine a ‘flexibility matrix’ that allows to
determine the historical relation between education and choice of occupation.
(4) Available from Internet:

http://events.cedefop.europa.eu/skills2011/presentations.html [cited 5.10.2011].
27
The Italian model, presented by Massimiliano Tancioni, is a macro (dynamic

stochastic general equilibrium, DSGE) simulation model that allows for the
breakdown to sector, age, profession and skill. This theory-driven model allows
with the use of the breakdown to evaluate the impact of policies enabling their
optimisation. It is a theory-based way to translate (policy) changes into economic
outcomes at sectoral, occupation or skill levels.
The Swiss educational planning model was presented by Jacques Babel.
The model includes detailed flows in the Swiss education and training system.
With a time horizon of 10 years, it is used to forecast future flows of graduates,
but also to evaluate the impact of enacted or planned policies. It is based on a
hierarchical multistate model which includes the, for Swiss important, flow of
foreign migrants.
Artur Gajdos introduced the Polish model, which emphasised the breakdown
of outcomes to the regional level. He showed the differences in sectoral and
occupational structure on the regional level, and how they are included in the
Polish model. The main model is built on a forecast based on macro model that is
further refined by a linked occupational structure forecast.
Marie-Béatrice Rochard, presented the French purely regional model of
occupation and skills monitoring. Using a macro regional model and scenario’s
data for regional economic policy are generated that allow to improve not only the
decision-making process but also form the basis for discussions between
different stakeholders in the region.
2.3. Comparison and evaluation studies

Many presentations compared, implicitly or explicitly, the outcomes of their own
models and those of the Cedefop project for their country. For the Baltic States,
Žilvinas Martinaitis, compared the Cedefop forecast by sector, occupation, and
education with the realisations published by Eurostat. Even though the group of
countries fits perfectly in the category of countries that are difficult to forecast for
as they are small, in a transition process, and prone to in- and out-migration, he
concludes that overall the forecast was close to the realisation. Many national
studies mentioned in the previous section explicitly or implicitly make a
comparison to at least a part of the national outcome of the EU-wide Cedefop
forecast. While there are differences, many times the conclusion is that the
outcomes are in line with the national model, be it with less level of detail or lower
accuracy. The Swiss presentation made a detailed comparison concluding that
there are significant differences in the age distribution (by education), while the
overall outcome of the forecast of the national and the Cedefop model are
28
similar. It also shows the importance of migration for the national labour market at
a level of detail that is not feasible for a EU wide study, given the current data.
2.4. Extensions
The extensions presented were mainly along two lines. First, there were a series
of projects and suggestions to investigate the nexus of skills and occupations
more deeply. And second, the use of alternative or additional data or modelling
environments.
For the first line of extensions, the treatment of skills, Jiří Braňka presented a
EU wide project on the transferability of skills. The projects goal is to determine
the skill composition and transferability of and between sectors and occupations.
By using several in-depth sector studies, the team tried to identify soft and
(generic) hard skills for lower and intermediary occupations. Understanding the
skill composition would help foster policies that target an increase in
employability of the lower and intermediate skill group, which is likely to come
under more and more pressure.
Other countries have tried, on the national level, to consider similar issues.
The aforementioned ‘flexibility matrix’, for example, tries to model something
similar, be it less explicitly. For Austria, Rainer Humpl showed the use of skill
contents extracted from vacancies to characterise occupations and the skills
needed. Pekka Tiainen’s question is related: how can we connect literature on
the sources of economic growth with outcomes of our skills forecast, or maybe
more ambitious how could we include feedback loops into the model that
includes the outcome of skill mismatches in the future growth expectations of a
country? He proposes an index-based growth accounting framework to this end.
The second branch of extensions was by using alternative data sources. The
above-mentioned project could also be seen in this line of extensions as both the
project by Jiří Braňka and the one by Stefan Humpl use, for example, employer
surveys to identify changes in the skill composition of employment. Furthermore,
the explicit inclusion or extension of results to the regional level was emphasised
by several contributors.
29
2.5. Presentation of skill supply and demand forecasts

A forecast is useful only if the end-users understand its outcomes and apply them
to the correct questions. The presentation of the results is of crucial importance
to achieve this. Tom May presented the qualitative component of the Cedefop
forecast, among other things it also allows to package the results into stories that
can be evaluated on the qualitative level. As such, it uses the quantitative outputs
as input to be placed in a context.
Another aspect of presentation is the collection of various results by the
variable of interest to the end-user. A good example is the AMS-skills barometer
as a tool of presentation. It collects and presents data from various, quite
detailed, sources to characterise, for example, occupations by several aspects
such as skills content, change in demand over time, income, and so on. A user-
friendly interface invites end-users such as high school students to investigate
the possibilities various educational and occupational choices might lead to.
A further element of presentation is often more implicit in the building of a
model. Many models developed out of a demand for a specific question.
Regional models were developed to present information on regional level, it is
both a methodological choice and a level of presentation. Other models are
extended to include regional components because a national presentation of
results would not otherwise be in line with an important component in the labour
market.
2.6. Technical responses to the presentations and

discussions
Discussion touched on all the main elements of the basic forecasting approach.
Figure 2:1 shows a generalised version of the modelling structure that is used to
produce the Cedefop forecasts (although this could be applied for summarising
any ex ante modelling task).
2.6.1. Historical data

There was much discussion about data issues during the conference. Although
this is highly relevant to quantitative assessment and the modelling team, in
terms of lessons learned this is largely outside our control.
30
Figure 2:1 Overview of (Cedefop) existing approach
Forward-looking
Historical data
inputs
Structure and
Model
assumptions
Expert judgment
Final results
Source: Authors.
However, one recurring theme was the trade-off between the data
requirements and the degree of complexity involved; with better quality (i.e. more
detailed and more reliable) data it is possible to apply more sophisticated
modelling techniques. This is particularly relevant when considering geographical
boundaries, as data at national level tend to offer a higher level of detail than the
pan-European data sets used in the Cedefop forecasts. The question of how to
make best use of the data available still remains largely open.
2.6.2. Forward-looking inputs

The main forward-looking inputs to the forecasting exercise are the baseline
assumptions for population growth and economic development. Both of these
were discussed at various points during the conference, particularly in relation to
the crisis (e.g. Chapter 7).
The issue of migration is important on the supply side. The current approach
uses Eurostat projections as exogenous inputs, even though post-crisis they are
now dated. The current programme of work aims to improve on this, with some
model runs based on alternative migration assumptions.
2.6.3. Expert judgment

Many of the Skillsnet team members who provided the feedback on the
projections were present at the conference. Some of them presented the
differences between the final versions of the Cedefop projections and the
national ones, or the actual outcomes (mainly discussed in the session on
national projections). This is a very valuable input to the overall exercise, as the
31
pan-European approach is not able to make use of local information, expected

developments and more specialised data. In the coming year the team will be
focusing on how to make the most use of the information provided.
The possibility of incorporating sectoral and national feedback was another
point raised in the discussion. The previous forecast made a comparison against
the DG Employment sectoral studies (5), but there are limitations on the
quantitative inputs they can provide.
2.7. Limitations of the modelling

In the introductory session to the conference, Rob Wilson noted that modelling
cannot provide an answer to everything, it can address some questions and
others not.
Modelling represents a simplification of reality, based on a series of
assumptions about how the world can be represented in a computer-based
format. A recent Cambridge Econometrics publication (Poliitt et al., 2010)
commenting on the current state of the art in modelling, four major limiting factors
to current modelling approaches were highlighted. During the conference
presentations and subsequent discussion, all of these were raised in the context
of labour-market forecasting. They are described below.
2.7.1. Degree of exogeneity

The boundaries of the model provide an important and necessary constraint on
the analysis. As described above, the main exogenous inputs in the modelling
are the population and migration projections and economic projections. Most
other inputs are assumed to remain broadly constant with historical patterns.
Developments in the rest of the world is another noted exogenous input.
2.7.2. Parameterisation and non-linear responses

Several of the presentations compared the econometric approach used in
Cedefop’s forecasting and other modelling approaches including:
(a) continuation of previous trends;
(b) use of computable general equilibrium(CGE) models or methods;
(c) use of dynamic stochastic general equilibrium (DSGE models, Chapter 10).
Each method favours a different means of deriving behavioural parameters.

Equilibrium-based approaches favour a more theoretical than empirical
(5) See http://ec.europa.eu/social/main.jsp?catId=784&langId=en.
32
methodology. Each method has its own data requirements, with the current
methodology used having quite stringent requirements (for all EU Member
States).
While it is important to compare results from the different approaches, at the
moment there is only limited scope for changing the methodology used to
produce the Cedefop projections.
The use of linear responses in the modelling approach was also implicitly
called into question. The model parameters are generally appropriate for
marginal changes but it is not necessarily true that these can be extrapolated in a
linear fashion for larger impacts. For example, many of the forecasting results,
including the Cedefop projections, slightly overestimated the employment
impacts of the recent crisis.
2.7.3. Technology
The treatment of technology in forward-looking modelling approaches is always
problematic for the basic reason that it is not possible to predict the future
direction of technologies. Currently the most discussed aspect of technological
change is the response to challenges posed by greening the economy. This is
hampered by problems with data classifications and the definitions of what is
actually ‘green’. There is ongoing work at both Cedefop and the European
Commission looking into labour-market impacts, and the research team will aim
to ensure consistency in approach over the coming year.
2.7.4. Treatment of uncertainty

Another area of modelling that is growing in importance is the treatment of
uncertainty. There are many sources of uncertainty, including many of the issues
discussed at the conference and in this chapter (e.g. data, model parameters,
exogenous inputs, policy responses).
The skills forecasts will never be ‘correct’. The modelling results present a
single set of point results and, although these are supported by additional sets of
positive and negative outcomes, this still does not address the full range and
likelihoods of possible developments.
In the previous year’s work, additional scenarios were developed to look at
some of the most important policy developments that might take place before
2020. This helped to focus on some of the key issues. However, it can be
concluded that the modelling results must be taken in context and should be
considered as part of a wider analysis. This should include qualitative elements,
telling the story and providing interpretation, as described by Tom May (in
Chapter 3).
33
2.8. Incorporation of the supply side?

One important development to the existing framework discussed at the
conference was the possibility of including the link from skills to the wider
economy. The presentation by Pekka Tiainen presented one such approach,
based on the growth-accounting method.
This link (Figure 2:2) has been put forward as one of the options to be
explored in the ongoing Cedefop project but there are methodological differences
between the supply-side approach described and the E3ME model, which is
more focused on demand. Nevertheless, it is worth further consideration.
Completing this loop would open up other possibilities in the modelling, such
as considering the transfer of skills between sectors, which was discussed by Jiří
Braňka and Tobias Maier among others.
This will be considered further in the next year.
Figure 2:2 Incorporation of supply side?
Forward-looking
Historical data
inputs
Structure and
Model
assumptions
Expert judgment
Final results
Source: Authors.
2.8.1. Impacts of the crisis

The effects of the recent financial and economic crisis were mentioned in several
of the presentations. The crisis raises important methodological questions for
different model approaches:
(a) if using trends to predict future growth, what period should be used;
(b) is it appropriate to use empirically derived parameters for forecasting post-
crisis;
(c) what is the most appropriate base year for a computable general equilibrium
(CGE) model to use.
34
The opening to the conference described the existing imbalances (with both
high levels of unemployment and skills shortages in certain sectors) which partly
result from the sharp negative impacts of the crisis.
This is something that the research team will keep in mind in the coming
year. However, it is still too early to tell what the lasting effects will be; at the
moment the last year of data is 2009 so it is not possible to make pre- and post-
crisis comparisons. The expert judgment will therefore play a particularly
important role in the upcoming projections.
2.8.2. The wider policy perspective

The European policy landscape is now shaped by the Europe 2020
strategy (European Commission, 2010), which has a target for employment rates
and also for reducing early school leavers. However, some of the other fixed
targets (e.g. the environmental targets) may also have considerable effects on
labour markets.
Cedefop forecasts are not expected to meet the employment targets (as
they are largely based on current policy only) but will offer a comparison. It is
clearly important that the wider policy perspective be considered, as well as the
links between policy areas.
Policy developments are also of major importance at national level as, for
example, many of the policies required to meet the 2020 targets will be
implemented by the Member States. In many cases this will require local
knowledge, so it is another area in which the role of the country experts will be
important.
References
Cedefop (2008). Future skills needs in Europe: medium-term forecast: synthesis
Cedefop (2009). Future skills supply in Europe: medium-term forecast up to
2020: synthesis report. Luxembourg: Publications Office of the European
Union.
European Commission (2010). Europe 2020: a European strategy for smart,
sustainable and inclusive growth (Communication from the Commission).
Luxembourg: Publications Office of the European Union (COM(2010) 2020).
Pollitt, H. et al. (2010). A scoping study on the macroeconomic view of
sustainability. Final report for the European Commission, DG Environment.
Cambridge: Cambridge econometrics.
35
Part 1:
Building on Cedefop’s
forecasting results
Qualitative scenarios for policy-makers

Tom Leney
Polarisation or segmentation?
Occupational change and the implications for the future EU labour market
Antonio Ranieri
Use of forecast results in a skills transferability project

Jiří Braňka
CHAPTER 3.
Trends, drivers and qualitative scenarios in
Europe 2020: for education and training
policy-makers engaged with meeting skills
needs
Tom Leney (6)
This chapter sets out some qualitative scenarios that emphasise skills
supply and demand in 2020. It aims to help education and training policy-
makers to contribute to meeting labour-market skills demands. The
research is based on the results of recent analyses of European labour-
market demand and supply, and education and training analysis. Using a
10-step methodology devised for the Cedefop project, the chapter
describes wide-ranging trends and identifies powerful drivers and
uncertainties that will impact on education and training policies and
provision.
The chapter then elaborates three scenarios that are based on economic
trends and focus on identifying and meeting skills needs. These are
expressed as 2020-snapshots. Each scenario has two variants (scenarios
1A/1B; 2A/2B; 3A/3B). Finally, it evaluates the usefulness of the research
conducted, and suggests steps to take in 2011.
3.1. Introduction
This chapter develops an overview of trends and drivers and builds a set of
scenarios. This is intended as a helpful tool for education and training policy-
makers who — among other priorities — have to be as sure as possible that their
policies when implemented are capable of making a strong contribution to
meeting the various demands for labour-market skills. The work forms a
qualitative part of the Cedefop project ‘Forecasting skill supply and demand in
Europe’, whose main goal is to develop an econometric model that can measure
and anticipate skills needs and supply at the European, national and sectoral
levels.
The methodology for preparing and presenting the qualitative scenarios was
developed in 2010 specifically for the project. This involves following a series of
10 steps. The chapter takes the process through to Step 8. It encompasses
(6) The paper was presented at the conference by Tom May.
37
identifying data for a range of trends, identifying powerful drivers and also
uncertainties that can be expected to impact on education and training
policies (7). As a result of the analysis, 100 vital trends are specified, and these
impact on skills development aspects of European systems of education and
training. Next, some of the key areas of likely uncertainty for education and
training planners and policy-makers are suggested and 9 factors that are driving
change in education and training planning (the drivers) are identified. Each of
these relates strongly, but not exclusively, to questions of skills demand and
supply as concerns policy development in education and training. Three
scenarios are constructed, and subjected to some preliminary consultation and
validation. They are expressed as 2020-snapshots. Each scenario has two
variants (scenarios 1A/1B; 2A/2B; 3A/3B). The scenarios are:
Scenario 1: strong economic growth: lessons learnt;
Scenario 2: gradual recovery: all to play for;
Scenario 3: double dip and down: a bumpy ride.
3.2. The 10-step methodology

A detailed paper proposing a methodology (Leney, 2010a) for developing skills
related scenarios for education and training policy-makers resulted in the
adoption of a method that the research has followed carefully as far as Step 8
(for more detail see Leney, 2010a). In outline, these are the 10 steps of the
procedure for developing qualitative scenarios in this case:
Step 1 define the problem;
Step 2 identify the contexts to be explored for trends;
Step 3 gather data and identify the key drivers;
Step 4 identify the main trends;
Step 5 decide on the most important trends whose outcome is uncertain;
Step 6 use the data to construct a grid or matrix for a set of scenarios;
Step 7 develop the provisional scenario statements;
Step 8 check out the scenario statements for plausibility;
Step 9 modify and polish, then present the scenarios. Test with prospective
users;
(7) This has entailed gathering data on a range of trends and issues in relation to skills
demand and skills supply in the European labour markets (Cedefop, 2010; 18 of the
DG Employment sectoral skills and scenarios studies) and in relation to
developments in European systems and policies for education and training (Council
of the EU, 2009). These sources contained a lack of trends concerning migration as
a factor, so recent OECD data was used to fill this gap.
38
Step 10 introduce the scenarios into policy-making. Keep under review.

The next sections describe the results of following the procedure.
3.3. Trends analysis

A careful analysis of the important trends that are likely to have a strong impact
on education and training policy in Europe was needed to develop a clearer
understanding of the challenges policy-makers face as ensure that their policies
and reforms help to meet labour market skills needs. We identified 100 important
trends from the sources consulted, using these categories:
(a) trends that lie outside the labour markets — population; natural resources
and technology; globalisation and economic factors; social and political
factors (31 trends);
(b) trends in labour-market skills demand (14);
(c) trends in labour-market skills supply (12);
(d) trends in education and training outcomes, systems and policies (43, further
subdivided).
To give an indication of the trends identified, the table that follows
summarises the labour-market demand and supply trends. Mostly, these are
identified from the Cedefop project and from the EU sector scenarios work (8).
3.4. Drivers of change and key uncertainties

From the analysis undertaken, we have been able to identify some key drivers
that are impacting on the decisions that education and training policy-makers are
responsible for (9). The most important underlying factor is the economic
trajectory beyond the recent crisis. Nine drivers are identified:
(a) global/economic shifts including economic recovery and limitations on public
funding;
(b) demand and supply pressures for skills in the labour market;
(c) organisations reviewing their structures continuously;
(d) technological development;
(e) unresolved and new challenges facing education and training systems;
(8) For a full account of the 100 trends identified see Leney (2010b, p. 7-15).
(9) In particular from the views expressed in the European Union sectoral scenario
studies, from the evidence of the Cedefop anticipating skills supply and demand
project and from the EU reports progress towards the Lisbon goals for education and
training. See reference list.
39
(f) pressures on companies to innovate and perform;

(g) the effectiveness/ineffectiveness of European, regional, national policy
regimes and networks;
(h) the effectiveness of localised links between education and training providers,
companies and government;
(i) demographic change.
Several factors associated with these drivers and the trends summarised
above create uncertainty for education planners, as they work with stakeholders
to develop their education and training systems to meet anticipated skills needs.
These can be expressed as a series of questions. Will the Member States be
successful or otherwise in emerging from the recession? Will there be sufficient
numbers of well trained teachers and trainers? Will countries be able to develop
and sustain models for increasing access to continuing training in the workplace
(continuing vocational training)? How will long-term priorities be viewed in the
light of short-term expediency — in terms of trade offs and achieving ‘more for
less’? While education systems face growing demands for medium and high
skills, will they be successful in combating school dropout and inequalities?
Higher levels of formal qualifications are expected, but what kinds of
knowledge/skills, attitudes, and competences does this imply? How can the
education systems meet both national and local or niche demands? What will be
the impacts of IT on both education structures and wider business models? Will
job polarisation be limited or extensive?
Box 3:1 Summary of likely trends in labour market skills demand and supply
(2010 to 2020)
Labour-market skills demand (2020):

• limited convergence: levels and types of skills demand continue to vary from
Member State to Member State, across the EU regions, and within sectors;
• European employment growth is gradual. Emergence from the recession is
modest and varies by country. In 2020 employment is higher than in 2010 but
below the peak of 2008 (a);
• knowledge- and skills-intensive jobs continue to take an increasing share of all
employment. Managers, professionals and technicians are increasing, while
agricultural, many craft- and some clerical-skills jobs decline;
• there is some polarisation/hollowing out of the labour market, and concern about
quality of many jobs, particularly in countries shifting to service economies;
• the number of job openings will total around 80 million (as a baseline) over the
decade, made up of replacement and expansion demand;
40
• expansion demand is positive (job creation) in knowledge and skills intensive

areas, and in sales, security, catering and caring. Losses are likely in clerical and
skilled manual jobs;
• replacement demand is a substantial amount of all job openings, as ageing EU
citizens retire. Even where there are job losses, many viable options will remain;
• overall there is increasing demand for people with high and medium qualifications,
and decreasing demand for low skills. Some jobs previously calling for low skills
call for higher skills levels;
• skills profiles are changing. Many increasingly need combinations of transversal,
core and specialist skills, as well as new technical skills and more multi-skilling;
• almost (but not quite) everywhere, there seems to be a growth in the demand for
managers;
• manufacturing: significant job losses over the decade, particularly in skilled
occupations. Yet increased specialisation and value-added activities result in more
high-skilled jobs;
• some manufacturing industries will continue to expand. Chemicals and some niche
industries are just such an example, and this has implications for the demand for
skills;
• services: significant job growth has continued. Unlike much manufacturing, there is
a steady increase in the need for low-skilled and high-skilled jobs in the services
sectors;
• a shift in skill structures in Europe is still taking place. Central and eastern Europe
specialise more in production and assembly activities, while western Europe
concentrates on the research and development, and high-value segments.
Skills supply (2020):

• limitations to convergence in qualifications and skills supply is a continuing feature
of the European map, reflecting differing demands and differing education
traditions;
• the supply of people with medium and high level qualifications has continued to
increase, particularly university or equivalent qualifications. Medium, mainly
vocational qualifications still form half of the EU labour force;
• the supply of people in the EU labour force with low-level qualifications falls
sharply;
• in individual Member States, the trends in high- and low-qualified people in the
labour force is similar to the European trend. The trend for medium-level
qualifications is more complex: shares are rising in some countries, falling in
others;
• on average, women are formally more highly qualified than men, except at medium
qualification level. More (and better qualified) women to be active in the labour
market;
• age cohort trends: the qualification structure of those aged 55 is changing (more
highly qualified). Labour-market participation rates among some older age groups
are increasing, but the impact varies.
41
• trends in skills supply within a sector are complex — with major differences
between industries and countries, in terms of expansion and contraction and
market shares;
• matching high skills demand and supply is difficult in many cases: some sectors
risk remaining short of appropriate skills. Some sectors remain unattractive to job
seekers. Often, it is more difficult for medium and high skilled people to find the
jobs they like; even so, they still have relatively better chances;
• the development/supply of innovative and effective leadership is a clear trend in
some sectors/labour markets;
• increasingly, skills that are developed in one industry can be used in another. The
situation for transfer of skills flexible: mobility, dual labour markets, generic/soft
skills;
• until 2010, free movement migration was generally for lesser skilled jobs in
expanding sectors: it declined rapidly through the economic crisis. Labour
migration has declined more slowly for higher skilled jobs. But in OECD countries
inflows remain greater than outflows;
• informal and formal skills supply: in some sectors family members/contacts are an
informal source of the skills supply. The extent varies between EU Member States
and sectors.
—————————————————————————————————————
—
a
() Where figures and percentages are included in these trends, they reflect the result of the Cedefop
project projections for labour-market skills demand and supply to 2020, emanating from a ‘baseline
scenario’ of slow but fairly steady growth.
3.5. Draft scenarios 2020: to assist European

education and training policy-makers to contribute
to meeting labour market skills needs
Education and training policy-makers have several main tasks, one of which is to
try to ensure that education and training policies, systems and reforms make as
full a contribution as possible to achieving an adequate and efficient supply of
well-skilled people in the labour market, whether as new entrants, people
continuing to expand their knowledge and skills in the workplace or as job
seekers. The context in which policy-makers operate is often one of both
uncertainty in terms of outcomes, and complexity in terms of the range of
considerations that they must consider. Scenarios are seen as a tool that can
help policy-makers as they deal with uncertainty and complexity. Also, by using
evidence to construct alternative yet plausible views of the future, scenarios can
help policy-makers to achieve a broader view of the available policy alternatives
and their likely impacts.
42
In the qualitative scenarios project, we have identified three main

components around which to construct the scenarios, taking 2020 as the vantage
point. These components are the mode and trajectory of economic recovery after
the recession, the impact of the recovery on skills demand and supply, and the
overall strategy adopted by government. These components are best expressed
as questions:
(a) to what extent have Europe and Member States recovered since 2010 from
the recession of the late 2000s, and what is the main economic trajectory;
(b) what is the impact of this recovery, or otherwise, on skills demand, and on
the supply of skills;
(c) how has government strategy in the broadest sense geared its response to
the wider economic and social developments.
At this stage, we have developed three qualitative scenarios:

Scenario 1: strong economic growth ― lessons learned?
EU is holding its own in terms of BRIC competition. Specialist
manufacturing holds up and finance and services sectors
thrive. Increased demand for managers’ skills, but limited job
polarisation. Matching labour-market supply and demand is a
challenge. Education and training providers are tasked with
equipping graduates with appropriate combinations of
specialist and generic skills.
Scenario 2: gradual growth― all to play for?
Drop in manufacturing. Increase in job polarisation. Gender
trends in labour-market transitions can be harnessed to
improve skills supply, but obstinate barriers remain if
continuing vocational training is to improve. Disparities of
labour-market demand/supply. Education and training policy-
makers have to provide employability skills, working under
funding constraints of ‘more for less’ and shifting responsibility
to the users — employers and learners.
Scenario 3: double dip and down ― a bumpy ride?
Relative collapse of manufactured goods and services,
particularly those lined to basic manufacturing industries. Job
polarisation. Little optimism, little up-scaling of innovation, yet
an increase in medium and highly qualified entrants to the
labour market. Lack of funding limits possibilities for
government intervention in education and training reforms,
while shorter, more vocational qualifications become more
attractive to learners.
43
The title of each scenario indicates the broad direction of recovery, and what
this may mean for education and training planners.
Each of these scenarios has two variants (scenarios 1A/1B; 2A/2B; 3A/3B).
These vary according to the role that government assumes or pursues. In variant
A the government seeks to have a strong and continuing role in economic
development and, not least, in developing or protecting a social model. In variant
B the government seeks (or is constrained) to lessen its role and to let the
markets find the ways forward. This is more of a laissez-faire, less interventionist
approach. For example, by 2020 we can identify two variants of the Scenario 2
the two variants are described as follows:
(a) in variant 2A governments have attempted to maintain direct influence over
education and training. Budgets may have reduced and the respective
burden of ‘who pays?’ and prioritisation of spending and reforms remains
open to debate within each Member State. The government remains the
main player in leading economic and social development. Depending on the
context, this tends to mean one variant or another of the pre-crisis or
reformed European model. Additionally, regulation has directly created new
skills demands through environmental, health and safety, labour-market
licensing and other requirements. Good planning on the part of national and
local government, effective partnerships and the translation of policies into
good practice make this the ‘more for less’ scenario: there is often less
spending, but it is better targeted;
(b) in variant 2B governments have chosen, or felt constrained, to steer
education and training policy via the markets, and thus loosen the control
and influence of the State. Liberalisation has continued to promote new
markets particularly in utilities and service industries such as healthcare and
education. Public sector employment has decreased significantly as a
consequence of government policy, although this has tended to be offset by
increases in the private tertiary sector. Individuals who found their jobs
moving from the public to private sector are now competing through cost and
or skills levels and this is a further incentive for them to both refresh and to
extend their training, often at their own expense. In terms of planning, ‘more
for less’ in this variant means giving a freer hand to the market place in
education and training and finding ways to reduce the role (and expenditure)
of the State. In contrast to variant 2A, the European social model here has
become less interventionist and more of a minimal safety net (Leney, 2010b,
p. 26-27).
At the end of 2010 we were able to consult on and test out in a limited way
both the draft scenarios and the range of statements (trends, drivers,
44
uncertainties) that lay behind the scenario construction (10). Throughout, we had
been able to liaise closely with the main, econometric results of the Cedefop
labour-market skills anticipation project.
3.6. Results of the (limited) consultation

The European experts, including Cedefop and Skillsnet experts we consulted,
expressed broad agreement as to the strengths of the scenarios exercise
undertaken:
(a) the scenarios are coherent and fairly easy to understand;
(b) almost all respondents described the scenarios as useful, or somewhat
useful, for a general understanding and planning of future skills demand and
supply;
(c) the trends and drivers provide a solid base for the development of the
scenarios;
(d) the optimistic, slow but steady growth and pessimistic scenarios are viewed
as coherent alternatives, although not all are equally plausible when applied
to particular countries. Country situations vary, and in some aspects the
scenarios can coexist;
(e) variant A and B in each scenario is helpful.
The consultation also helped to establish aspects that respondents consider

to be in need of strengthening. This includes:
(a) bringing out more clearly regional disparities in EU economic developments,
and thus more nuanced labour-market demand and supply developments;
(b) similarly, accentuating sectoral trajectories in a more differentiated way;
(c) making sure that the results of the main project are best incorporated (and
well flagged);
(d) Scenario 1 looks a bit simplistic. In discussion it was recognised that
Scenario 1 is implicitly linked to Europe 2020 (11). In this case, the
(10) Consultation on the methodology and on the draft scenarios was undertaken at the
Cedefop Skillsnet seminars in summer and autumn 2010, then through a limited
survey of experts in policy aspects of European education and training developments
(November 2010) and at a specially convened seminar of experts working at
Cedefop in Thessaloniki (December 2010). This made it possible to gather views on
the methodology and scenarios, and to prepare a section of the report that reflects
on the outcomes, methodology and potential usefulness of the approach (Leney,
2010a; 2010b).
(11) See, in particular, European Commission, 2010.
45
achievement/non-achievement of the objectives of Europe 2020 can be

treated in more balanced and refined way;
(e) migration could be an important driver and uncertainty, and calls for more
attention;
(f) the education and training trends look a bit too institutional — it is important
to find ways of reflecting the importance of on-the-job learning and other
informal and non-institutionalised aspects;
(g) the strategic role of the State could be better exemplified, or in more detail,
as this provides such an important context for education and training
policies.
In the last weeks of 2010 it was possible to consider some of these

comments. Other aspects remain on the agenda for further work in 2011. An
important and interesting challenge will be to capture the likely diversity of
directions of travel in different Member States, since it is clear that a single
European scenario outcome is unlikely.
3.7. Conclusion
As this chapter shows, the first year’s work on the qualitative scenarios has
achieved several results. We have:
(a) developed a 10-step methodology, which seems to be reliable and robust
and also to be capable of adaptation to local circumstances;
(b) linked the introduction of a new qualitative approach to on-going
econometric approach of the forecasting skill supply and demand in Europe
project;
(c) identified 100 key trends across a wide range of relevant social, economic,
labour market and educational fields;
(d) linked the key trends to several important aspects of uncertainty and to
some of the major drivers of change;
(e) established a conceptual framework for three qualitative scenarios, on the
above basis;
(f) drafted more detailed provisional scenario statements, including two variants
of each;
(g) tested and modified the scenarios in a limited way, through three
consultation exercises with appropriate groups of experts;
(h) amended the scenarios statements to consider some of the commentaries.
The work is being taken further in 2011. More work can be done to
incorporate the detailed commentaries provided by respondents and outlined
46
above. For this to happen, it will be necessary to complete the complex task of
taking the methodology through to Step 10 (Section 3:2). The scenarios also still
have to be refined.
Specifically, we are now attempting to analyse the evidence gathered using
statistical or other systematic approaches to the analysis of text, rather than
relying on the expertise of the researcher to extrapolate the key data and their
significance.
We will also use the developed statements of the trends, drivers,
uncertainties and scenarios to engage more fully with the education and training
policy community at European level. The purpose now is to work out how the
scenarios can be used to engage with the policy options and dilemmas that
planners face.
The strongest and most pertinent challenge is now to understand how the
scenarios may apply at local level — particularly as several Member States
currently tend to correspond to different scenarios, not to one single and
converging scenario. It is therefore helpful that some countries have indicated,
through the Skillsnet network and through Cedefop’s work (12), that they are
already using an approach similar to this as one of their planning tools. Several
Member States have indicated their interest in developing such an approach at
national level, they are therefore interested to see how the approach set out in
this chapter can be adapted to their national circumstances.
This sets out a formative set of tasks for the next stage of the qualitative
aspect of the Forecasting skill supply and demand in Europe project and the
possibility to build upon the solid basis already achieved.
References
European Commission (2010). Europe 2020: a strategy for smart, sustainable
and inclusive growth (Communication from the Commission). Luxembourg:
Publications Office of the European Union (COM(2010) 2020).
Council of the EU (2009). Council conclusions of 12 May 2009 on a strategic
framework for European cooperation in education and training (‘ET 2020’).
Official Journal of the European Union, C119/2, p. 2-10.
(12) Cedefop organised a conference on forecasting skill supply and demand in Europe
in Thessaloniki (February 2011).
47
Leney, T. (2010a).Trends, drivers and qualitative scenarios in Europe 2020: for

education and training policy makers engaged with meeting skills needs.
Task 04 Paper 1: Concept paper. Cedefop: Report prepared for Warwick
University/Cedefop skills forecasting project.
Leney, T. (2010b). Trends, drivers and qualitative scenarios in Europe 2020: for
education and training policy makers engaged with meeting skills needs.
Task 04 Paper 2: Trends, drivers and qualitative scenarios. Cedefop: Report
prepared for Warwick University/Cedefop skills forecasting project.
Additional websites consulted

Cedefop publications — http://www.cedefop.europa.eu/EN/publications.aspx
European Commission, employment social affairs and inclusions, Progress

programme — http://ec.europa.eu/social/main.jsp?catId=327&langId=en
European Commission, employment social affairs and inclusions, Sectoral level

analysis — http://ec.europa.eu/social/main.jsp?catId=784&langId=en
48
CHAPTER 4.
Polarisation or segmentation?
Occupational change and the implications
for the future EU labour market
Antonio Ranieri and Marco Serafini
Between 1998 and 2008, occupational polarisation emerged for the first
time in Europe not only in terms of low- and high-paid jobs: labour demand
rose for both high-skilled and elementary occupations. Using a shift-share
approach, our analysis shows that much of this phenomenon can be traced
to macroeconomic trends and structural changes between sectors. The
rising supply of non-national workers is also a factor in this, together with
some institutional characteristics of the labour markets, while technological
and task content changes seem to play a lesser role than expected.
In light of the continuous shift towards a tertiary-based economy and
predictable changes in consumption models and lifestyles of European
societies, it is likely that a relative increase in elementary occupations will
persist also in the near future. However specific policies, including
vocational education and training, could potentially favour a process of
‘occupational upgrading’ in the future, weakening the polarisation trend.
4.1. Introduction
European challenges are essentially driven by technological change,
globalisation pressures and sociodemographic developments. During the 2000s,
such processes took place in relatively favourable conditions. In 2008, when the
recession interrupted a long period of increasing employment, there were about
20 million new employees in Europe compared to the previous decade. The
international crisis dramatically changed the picture. Millions of jobs have been
lost during the last two years; while net employment growth will probably remain
low in the next decade as a result of the economic downturn. Against this
background, a better understanding of developments and changes in
occupational and labour skill structure is perhaps more important than in the past.
Over the last decades some consensus has been achieved in literature that,
besides a general trend of expansion of highly-qualified employment, continuing
polarisation is affecting labour markets in most developed economies. The
findings supporting this hypothesis are based mainly on a wage structure
analysis, highlighting that some countries have increased the shares of both
high-paid and low-paid jobs, while ‘intermediate’ occupations tend to decline. For
49
the first time, at least since 1970 and arguably even before, during the decade
1998-2008 occupational polarisation also emerged across Europe, with a
concomitant rising demand at the upper and lower ends of the occupational skill
distribution. Looking forward to the next decade, the results of Cedefop’s skills
forecast (Cedefop, 2010a) seem to confirm that polarisation is bound to continue
in the near future (Figure 4:1), with a possible simultaneous growth in the
demand for both high-skilled and elementary occupations, while most
intermediate occupations — especially manual ones — are on the decline.
Figure 4:1 Projected change in share in occupational structure (a), EU-27*,

2010-20
3.0%
2.0% 2.4%
1.0%
0.5%
0.0%
-0.3%
-1.0%
-2.0% -2.6%
-3.0%
High Skilled Skilled Elementary
skilled non-manual manual
a
( ) High-skilled occupations include major groups 1, 2 and 3 (managers legislators, senior officials and
managers; professionals; technicians and associate professionals). Skilled non-manual include major
groups 4 and 5 (clerks; service workers and shop and market sales workers). Skilled manual include major
groups 6, 7 and 8 (skilled agricultural and fishery workers; craft and related trades workers; plant and
machine operators and assemblers). Elementary includes major group 9.
Source: Cedefop (IER estimates based on E3ME, EDMOD and RDMOD).
The dimension and the intensity of these trends are expected to differ
between countries, sectors and even between specific occupations within the
major group of elementary occupations. Interpretation is complicated by the fact
that results also suggest increasing deployment of higher- and medium-qualified
people in low-ranked occupations. Several important and policy-relevant
questions emerge: how should we interpret this phenomenon; why are people
with higher qualifications increasingly employed in elementary occupations; are
there particular groups which are more affected than others; should we attribute
these trends to technological and/or task content changes occurring within
sectors?
50
4.2. Theoretical background

The possible drivers of changes in occupational structure are many and varied
(Box 4:1). Alongside the general trends of increasing return entailed by economic
growth (diversified by sector and/or occupation), the main drivers considered in
the literature can be traced back to competitiveness and international trade,
technological, organisational or task-content changes, and socio-institutional
aspects.
However, since the technological shift brought about by ICT spread through
the economy, interest in the role of technological change in explaining labour
demand composition has strongly increased. The basic premise is that new
technologies lead to higher productivity, but the impact on employment is
diversified between highly-skilled workers (who complement new technologies)
and lower-skilled workers. This is why, despite the large and constant increase in
highly-qualified labour supply, in many countries more skilled and educated
workers have increased both their relative employment rate and wages (13). The
skill-biased technological change (SBTC) hypothesis has found support in
several theoretical and empirical studies.
In line with the SBTC hypothesis, data clearly show that jobs at the upper
end of the occupational skill distribution — which usually require also a higher
level of formal qualification — tend to absorb an increasing share of labour
demand. However, in most developed countries increasing demand seems also
to emerge in a growing number of lower level jobs. The findings that support this
hypothesis are based mainly on a wage structure analysis providing evidence of
an overall increase in demand for low- and high-paid jobs, while people working
in the middle of the wage spectrum tend to decline (14).
Without contradicting the skill-biased technological change, the task-biased
technological change (TBTC) hypothesis provides a good explanation for this
trend. By lowering the opportunity-cost of capital, technological progress tends to
substitute human labour performing routine tasks, which generally occupy the
middle of the wage ranking, while many of the worst paid jobs are non-routine in
nature and, therefore, relatively unaffected by technological change. This
(13) See Katz and Murphy (1992), Davis (1992), Juhn et al. (1993), Katz et al. (1995),
Goldin and Katz (1996, 2008), Berman et al. (1998), Murphy et al. (1998), Machin
and Van Reenen (1998), Acemoglu (1998, 1999), Haskel and Slaughter (1999),
Caroli and Van Reenen (2001), De Santis (2002), Fitzenberger and Kohn (2006),
Autor et al. (2006, 2008), Atkinson (2008), Carneiro and Lee (2009).
(14) See Autor et al. (2003), Goos and Manning (2007), Spitz (2003), Spitz-Oener (2006),
Maurin and Thesmar (2005).
51
‘passive way’ to job polarisation has found support in several studies carried out
for the US, the UK, and for some European countries (15).
Box 4:1 Possible drivers of change in occupational structure
From the point of view of the individual worker or employer, occupational choice/change is not
only a consequence of exogenous macro phenomena. At micro level, the probability that a
person decides to seek/accept a specific occupation is the result of a complex decision process
that involves: the predicted returns, and expected unemployment risk, associated with
alternative occupations; the worker's family income; the inter-sectoral/inter-area mobility costs;
individual preferences and expectations. Similarly, the probability that an employer decides to
offer a job with a particular occupation involves: the expected productivity; the cost incidence of
on the job training; the actual transferability of skills between occupations; individual intuitive
perception about market perspectives. In this framework, the individual choice of workers and
employers in terms of occupations might not coincide (and they generally do not), generating an
occupational mismatch of which skill mismatch is only one component.
However, most empirical studies show that labour-market mobility is, above all, an individual
behavioural response to medium- or long-term macroeconomic changes. Four main structural
drivers of change in labour demand by occupation can be identified.
Main structural determinants of occupational changes at macro level

Demand drivers Supply drivers
Growth and  Engel’s Law (or similar)  increasing returns diversified by sectors
distribution  wage inequality/ income and/or occupations (Kaldor-Verdoon
distribution effect effect), changes in the inter-sectoral
effects
links
Competitiveness  extension of needs and  internationalisation/ globalisation
effects intensive diffusion of (Stolper-Samuelson effect, offshoring,
consumption patterns IDL and specialisation in general)
 increasing movement of goods
and people
Technological  changes in models of  skill-biased technological change,
and purchasing (retail/ delivery), routinisation hypothesis, etc.
organisational
reduction of transaction times,  job specific productivity, outsourcing,
consolidation effect, quality new inputs sectors
effects effect, innovation and
creation/substitution of needs
Socio-  increasing participation and  changes in substitution relations
demographic employment rates (especially between occupations due to flexible
women labour market institutions, black
macro trends and
 changes in family/social economy, migration flows, etc.
institutional structures and changes in  educational and social objective
effects lifestyle in general) functions (policies)
(15) For the US see Autor et al. (2003, 2006, 2008). For the UK see Goos and Manning
(2007). For France see Maurin and Thesmar (2005). For Germany, Spitz-Oener
(2006). Although affected by strong data limitations, similar results have been
obtained by Goos et al. (2010) considering the period 1993-2006 and 16 European
countries. They also report that the task dimension is a significant predictor for
employment change, even after the model is controlled for the formal qualification
level.
52
The task-biased technological change focuses on the task-dimension of

employment that is related to, but not completely explained by, the qualification-
skill level. In fact, technical and organisational advances tend to cause non-
neutral changes in productivity among job tasks rather than high- and low-skilled
jobs. However, following Autor (2007), the division of productive tasks into
abstract-complex tasks, routine and manual tasks is to be considered. Since
most non-routine tasks (both abstract and manual tasks) lie at the opposite end
of the occupational skill distribution, the task-biased technological change entails
a progressive polarisation of the labour market also in occupational terms, with
an increasing number of workers at the extremes of the spectrum and fewer in
the middle.
In this chapter, the starting point is the acceptance, in general terms, of the
skill and the task-biased technological change hypotheses. However, we focus
on the ‘occupational’ polarisation that is particularly relevant when implementing
forecast models and labour market scenario analyses as a tool for policy-making
in education and training.
4.3. Recent structural change in occupations

Looking at the very long-term picture, while the share of knowledge- and skill-
intensive occupations increased almost constantly between 1970 and 2000,
occupational polarisation has clearly emerged across Europe only since the end
of the 1990s (Figure 4:2). Even though the recent economic crisis partially
changed the picture (16), the number of workers in elementary occupations rose
by almost 3.9 million between 2000 and 2008, contributing 20% to the overall
growth in employment, and showing the highest growth rate together with high-
skilled occupations. As a consequence, the proportion of elementary occupations
in total employment increased from 8.7% to 9.6%, unlike all the other groups
except highly-skilled occupations that increased from 34.6% to 38.6%.
(16) The share of elementary occupations was steady in 2008 and declined in 2009.
53
Figure 4:2 Share % of highly skilled and elementary in total occupation, EU-27*
(1970=100)
Elementary High skilled Other occupations

150
140
130
120
110
100
90
80
70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08
Average share % of elementary in total occupation EU-27*

1970-1980 1980-1990 1990-2000 2000-2008
8.5% 8.4% 8.5% 9.3%
Source: Cedefop (IER estimates based on E3ME, EDMOD and RDMOD).
The recent growth in share of elementary occupations in Europe differs

across countries, sectors and among specific occupations within elementary
occupations. The simple analysis of level and change in share of elementary
occupation across European countries does not allow us to identify a general
trend or explanation for the variability of the phenomenon (Cedefop, 2011). Not
only level but, especially, change in share of elementary occupations in total
employment varies widely among countries; although in almost half of them, the
share of elementary occupations decreases between 2000 and 2008 and no
convergence trend can be identified.
From a sectoral point of view the picture is quite different (Table 4:2). While
a polarisation trend involves most sectors, tangible increases in elementary
occupation are concentrated in a few sectors and, in particular, in ‘private
household’, ‘agriculture’ and, to a more limited extent, ‘transport, storage and
communication’ and ‘wholesale and retail trade’. Further, ‘private household’
emerges as the unique growing sector where polarisation occurred during the
decade. Activities which are more likely to be influenced by social and
demographic structural changes (female participation in the labour-market,
54
changes in lifestyle, flexibility of the labour market, migration, etc.) rather than
technological or organisational change.
Table 4:1 Sectoral trends and polarisation, EU-27*, 2000-08

Polarisation Incidence
(changes in elementary
Overall
share ISCO (change in
growth
1+2+3+9) share ISCO 9)
(%)
(percentage (percentage
points) points)
Private household 42.3 10.3 9.2
Real estate renting and business
GROWING SECTORS
31.4 3.1 -0.2

activities
Hotels and restaurants 16.8 1.5 0.1
Health and social work 12.5 -2.0 0.0
Construction 6.0 0.5 -0.1
Other community social and personal
5.8 0.8 -1.1
service
Extra-territorial organisations 4.0 10.9 -0.3
Education -0.1 -0.3 -1.0
Financial intermediation -1.0 6.9 -0.2

Wholesale and retail trade -1.5 6.0 1.3
FALLING SECTORS
Public administration -2.8 2.5 0.6

Transport storage and communication -3.3 4.4 2.5
Electricity gas and water supply -15.1 9.3 -0.3
Manufacturing -15.9 4.6 -0.1
Fishing -26.4 -0.4 -0.1
Mining and quarrying -41.0 7.7 0.0
Agriculture -42.5 10.5 7.2
Consistent with these trends, ‘sales and service’ is the most substantial and
dynamic submajor group (Figure 4:3). This set of professions is usually
considered to require limited education and training: street or door-to-door sales
or services, cleaning services, property watching and caretaking, delivering
goods and messages or carrying luggage.
4.4. Structural changes versus occupational-specific

changes in employment
To assess to what extent the observed occupational trends can be potentially
attributed to technological or task content changes, using a shift-share approach
(Cedefop, 2011) we decompose cross-sectoral differences in occupational
structure within sector (structural effect) and between sectors (occupational skill
effect). Figure 4:4 gives a synthetic idea of the importance of the two components
55
in the overall growth of individual occupations. Looking at the clusters identified

(see also Box 4:2), the main conclusions are:
(a) although elementary occupations show positive effects for both sectoral
composition and occupational intensity, only a minor part (one third) of the
overall growth of elementary occupations can be ascribed to changes within
individual sectors (attributable to occupational-specific explanations);
(b) alongside elementary occupations, professionals, and technicians
(consistent with the polarisation hypothesis), also some intermediate groups
show positive occupational-intensity effects (skilled agricultural, and plant
and machine operators);
(c) more broadly, the analysis shows that the polarisation has been largely
determined by the sectoral trends which occurred during 2000-08, while the
potential role of technological and/or work content changes within sectors
has generally been weaker and more uncertain.
Figure 4:3 Trends in the share of occupations, EU-27*, 2000-08
Legislators, senior officials and managers 0.5%

skilled
High
Professionals 1.6%
Technician 1.9%
Clerks
manual
-1.1%
Skilled
non-
Service workers and sales 0.7%
Skilled agricultural and fishery -2.4%

manual
Skilled
Craft and related trades -1.6%
Plant and machine -0.5%
Sales and services 0.7%

Elementary
Agricultural and fishery 0.1%
Mining, construction, manifacturing and transport 0.1%
Other 0.0%
Source: Eurostat, LFS – Reference age: 15+.
56
Figure 4:4 Shift-share groups: structural and occupational-specific growth

components
50.0%
1. Faster than av erage with
f av ourable industry mix <
f av ourable specif ic f actors
30.0%
Occupational-specific component
6. Slower than av erage with 2. Faster than av erage with

unf av ourable industry mix > f av ourable industry mix >
f av ourable specif ic f actors Technician f av ourable specif ic f actors
10.0%
Plant and Elementary
Skilled agricultural
machine Prof essionals
and f ishery
Craf t and Serv ice workers

related trades and sales
Legislators, senior
5. Slower than av erage with of f icials and managers
-10.0% unf av ourable industry mix >
3. Faster than av erage with
unf av ourable specif ic
f av ourable industry mix >
f actors
unf av ourable specif ic
Clerks
-30.0%
4. Slower than av erage with

f av ourable industry mix <
unf av ourable specif ic f actors
-50.0%
-50% -30% -10% 10% 30% 50%
Structural-sectoral component
Source: Author, based on Eurostat LFS.
4.5. Potential labour-supply drivers of occupational

changes
Although these trends do not exclude a role for a ‘technological-task’ explanation
to polarisation, we are not aware of literature developing a comprehensive
approach to disentangling the effects and to assigning a specific magnitude to
the different causes. The wage-structure approach is difficult to apply to
European countries for theoretical reasons and scarcity of data, particularly when
approaching polarisation in terms of occupations rather than low- and high-paid
jobs.
57
Box 4:2 Shift-share groups interpretation
In this context, the shift-share analysis approach has been adapted to evaluate to what
extent the observed changes in occupations can be attributed to an increase of
‘occupational intensity’ within sectors, or if they are a direct consequence of changes in
sectoral composition.
Figure 4:4 must be read in four dimensions: (1) the oblique 135° line divides the
occupations between fast-growing and slow-growing occupations (faster or slower than the
average employment growth); (2) along the X axis, the relevance of the structural-sectoral
effect is measured (positive or negative changes between sectors); (3) along the Y axis the
relevance of the occupational-specific effect is measured (positive or negative changes
within sectors); (4) finally, the 45° line distinguishes the occupations where the
occupational-specific effect is more favourable (or less unfavourable) than the other.
On this basis, six relevant clusters of occupations can be identified.
Structural- Occupational
Occupations Growth in share sectoral -specific
component component
1  technician and associate faster than average positive < positive
2  professionals faster than average positive > positive
 elementary
3  service workers and shop/ faster than average positive > negative
market sales
 legislators, senior officials and
managers
4  clerks slower than average positive < negative
5  craft and related trades slower than average negative > negative
6  plants and machines slower than average negative > positive
 skilled agricultural and fishery
For high-level occupations, the skill and the task-biased technological

change hypotheses are confirmed by the dramatic increase both in numbers and
in share which occurred in recent decades, and partly corroborated by the wage
premium observed in some occupations at the upper end of income distribution.
If technological development and/or changes in work content are also in favour of
elementary occupations, an increase in the relative wage should be observed at
the lower end of the occupational structure as well. Apparently, this is not what
happened.
This leads to the question of imperfect mobility and labour-market
segmentation. When specific groups are trapped in low-paid segments of the
labour market, irrespective of their qualification and skill level, one consequence
is the alteration of labour cost differentials (considered in a broad sense), so that
employees can be induced to overuse some kind of occupations and/or worker
groups. Given the observed trend towards a higher qualification shown by
58
elementary occupations, to what extent can labour market segmentation be

considered a cause of polarisation? An analysis of elementary workers in terms
of sociodemographic characteristics provides some first evidence in this
direction. Figure 4:5 presents an analysis of differences in share (by age, gender,
education and nationality), where the levels and changes in share by
socioeconomic characteristics are calculated in differences from total
employment. In this way it is possible to note not simply the changes which
occurred during 2000-08, but also the shifts in relative labour demand for
elementary occupations among different groups of workers.
Figure 4:5 Main socioeconomic characteristics in elementary occupations, in-

differences analysis between ISCO 9 and total occupations EU-27*
30.0%
High level High level
low change Edu low high change
20.0%
Share % in differences (ISCO9 - Total) 2008
Part-time
10.0%
Immigrants
Female
Age 50-64
Age 15-24 Age 65+
0.0%
Age 35-49
Age 25-34
Edu medium
-10.0%
-20.0%
Low level Edu high Low level

low change high change
-30.0%
-8% -6% -4% -2% 0% 2% 4% 6% 8%
Change in share in differences (ISCO9 - Total) 2000-08
Figure 4:5 shows that the sharpest increase occurred in the incidence of
immigrants and medium-qualified workers, greatly exceeding all the other
components. Further, both higher- and lower-educated workers are losing ground
on the average occupation trends. These results require a deeper analysis of the
role of immigration and qualification trends in polarisation.
59
Figure 4:6 shows the strong increase in immigrant workforce in almost all
occupations (an average of +35.9% compared with +8.2% of the national
workforce). However, the rate of increase is extremely high for elementary
occupations (+85.5%) and ‘service workers and shop and market sales’
(+53.6%). Although relative change in percentage does not reflect the absolute
dimension of the trends, the impact of immigration on elementary occupations is
quite relevant. At the end of the period, the share of immigrants in elementary
occupations reaches 16.6%, much higher than in the other occupations (5.8% on
average).
Figure 4:6 Change % in employment by occupation and nationality, EU- 27*,

2000-08
National Non-national
15.9%
Legislators, senior officials and managers 31.6%
10.8%
Professionals 30.6%
24.0%
Technicians 29.7%
-1.6%
Clerks 19.2%
13.1%
Service workers and sales 53.6%
-29.5%
Skilled agricultural and fishery -34.6%
-3.8%
Craft and related trades 23.5%
3.4%
Plant and machine 3.0%
13.7%
Elementary 85.5%
8.2%
Total 35.9%
It is not a surprising result. Several studies highlighted some relevant supply-

side effects of labour related to immigration (Biffl, 1996). Lower reservation wage,
lower substitutability, dampening effect on wages, are all factors that offer a key
to interpretation for a significant part of the overall phenomenon. If we also
consider the peculiar socioeconomic characteristics of immigrants and the
institutional factors that tend to segregate part of the non-national workforce, we
60
can assume that a supply-side effect related to migration can play a significant
role in polarisation.
Data on qualification by nationality also support the idea that, in the specific
case of elementary occupations, labour-market segmentation and consequent
misallocation of labour across occupations could also play a relevant role.
Figure 4:7 shows that observed trends are significantly differentiated between the
national and non-national workforce. While the increase in medium and high
qualifications in national elementary workers is quite close to the average of total
employment, for immigrants the share of medium and high qualification increases
at almost double the average pace.
Figure 4:7 Share of medium and high qualifications in total employment, EU-27*,
2000-08
Share 2000 Change in share 2000-08
ELEMENTARY ELEMENTARY
National 39.4% National 8.9%
Non-national 32.8% Non-national 16.3%
TOTAL OCC. TOTAL OCC.
National 70.0% National 7.1%
Non-national 57.2% Non-national 8.8%
Source: Author, based on Eurostat LFS.
Further evidence of immigrants’ difficulties to find jobs matching their

qualifications emerges when considering the workforce level of qualification by
occupation. Figure 4:8 shows the ratio between non-national and national
workers by qualification levels, comparing elementary occupations with the
overall employment average. Not only are non-national workers concentrated in
elementary occupations (2.4 times more so than national workers), but also the
index rapidly increases with the level of education, peaking at 5.3 for the highly-
educated non-national workers.
Various explanations can be given to account for this phenomenon. There is
some evidence, for example, that employers tend to value education and
experience in the host country (EIU, 2009). Also, what is referred to as over-
education actually can be the result of differences in skills and competences
61
among individuals with the same qualification level. However, it is also

reasonable to come to a diagnosis of underutilisation rather than overqualification
of the workforce. The term ‘underutilisation’ is used here in opposition to ‘over-
qualification’ to underlie that, in this case, the causes of educational mismatch
should be traced back to segmentation of the labour market rather than to an
excess supply of skilled labour. It is worth underlining that in 2008 about 33% of
the tertiary-educated workers in elementary occupations were immigrants
compared to 6% on average for other occupations.
Figure 4:8 Concentration index of non-national workers in elementary

occupations by qualification, EU-27*, 2008
2000 2008
5.3
3.3
2.6 2.4
1.9 2.0
1.5 1.7
Low Medium High Total (average)
4.6. Conclusion
What can the analysis of past trends say about the potential role of elementary
occupations in the future European labour market? Much occupational
polarisation in Europe can be traced back to macroeconomic and structural
changes (between sectors), to demand-driven increase in specific service
activities (e.g. private households), and to the increasing labour supply of non-
national workers. In light of the continuous shift towards a tertiary-based
economy, together with the predictable changes in consumption models and
lifestyle of European societies, a relative increase of elementary occupations will
likely persist.
However, technological progress and/or work content changes seem to play
a minor role in this. On one side, there are many service sectors where
‘qualitative’ innovations have little to do with technology. On the other, although
the routinisation hypothesis is potentially significant, when approaching
62
occupation polarisation focusing on elementary occupations, other factors —

whether institutional or sociodemographical in nature — appear to offset and/or
assume a greater explanatory value. Migration regulation, flexible working,
undeclared work, direct public support, and role of self-employment, are all
factors that can modify the opportunity-cost of labour, such as the probability that
a person decides to accept a specific occupation. It is possible that labour market
segmentation and/or segregation tend to characterise many elementary
occupations described as ‘lousy’ jobs (low paying, few opportunities and no
benefits) even when it should not be.
It is likely that better and more inclusive labour market and migration policies
would reduce the ‘low-cost’ workforce which has no other option than to work in
simple jobs. This would induce employers to switch to production techniques that
use more higher-skilled workers and/or to rely on more capital-intensive
production methods. As a consequence, the share of elementary occupations
within some sectors — such as household services — would decrease. In this
context, specific policies, including vocational education and training, could drive
a process of occupational upgrading in the future, and counteract the polarisation
trend. In the absence of new policies and institutional changes, elementary
occupations would probably increase and become even more unattractive to the
local population. In this scenario, the share of immigrants in elementary
occupations is bound to increase.
References
Acemoglu, D. (1998). Why do new technologies complement skills? Directed
technical change and wage inequality. Quarterly journal of economics, Vol. 113,
No 4, p. 1055-1090.
Acemoglu, D. (1999). Changes in unemployment and wage inequality: an
alternative theory and some evidence. American economic review, Vol. 89,
Issue 5, p. 1259-1278.
Atkinson, A.B. (2008). The changing distribution of earnings in OECD countries.
New York: Oxford University Press (Rodolfo De Benedetti Lecture series).
Autor, D. (2007). The polarisation of the U.S. labour market: evidence, explanations
and implications for higher education. Cambridge, MA: MIT Press.
Autor, D. et al. (2003). The skill content of recent technological change: an empirical
exploration. Quarterly journal of economics, Vol. 118, No 4, p. 1279-1333.
Autor, D. et al. (2006). The polarisation of the U.S. labour market’. American
economic review, Vol. 96, Issue 2, p. 189-194.
Autor, D. et al. (2008). Trends in U.S. wage inequality: re-assessing the revisionists.
Review of economics and statistics, Vol. 90, Issue 2, p. 300-323.
63
Berman, E. et al. (1998). Implications of skill-biased technological change: international

evidence. Quarterly journal of economics, Vol. 113, Issue 4, p. 1245-1279.
Biffl, G. (1996). Immigrant labour integration. In: Schmid, G. et al. (eds) International
handbook of labour market policy and evaluation. Cheltenham: Edward Elgar.
Carneiro, P.; Lee, S. (2009). Trends in quality-adjusted skill premia in the United
States, 1960-2000. London: CEMMAP (Working paper, CWP02/09).
Caroli, E.; Van Reenen, J. (2001). Skilled biased technological change? Evidence
from a panel of British and French establishments. Quarterly journal of
economics, Vol. 116, No 4, p. 1449-1492.
Cedefop (2010a). Skills supply and demand in Europe: medium-term forecast up
Cedefop (2011). Labour market polarisation and elementary occupations in
Europe. Blip or long-term trend? Luxembourg: Publications Office of the
European Union (Cedefop research paper, No 9).
Davis, S.J. (1992). Cross-country patterns of change in relative wages. In: Stanley, F.;
Blanchard, O.J. (eds) NBER macroeconomics annual 1992, Volume 7.
Cambridge, MA: MIT Press, p. 239-300.
De Santis, R.A. (2002). Wage inequality between and within groups: trade-induced
or skill-biased technical change? Alternative age models for the UK. Economic
modelling, Vol. 19, Issue 5, p. 725-746.
EIU – Economist Intelligence Unit (2009). Estimating potential labour shortage and
supply in the European economic area. London: EIU, country and economic
research.
Fitzenberger, B.; Kohn, K. (2006). Skill wage premia, employment, and cohort
effects: are workers in Germany all of the same type? University of Freiburg
(Working paper).
Goldin, C.; Katz, L. (1996). The origins of technology-skill complementarity.
Cambridge, MA: NBER (Working paper No 565).
Goldin, C.; Katz, L. (2008). The race between education and technology.
Cambridge, MA: Harvard University Press.
Goos, M.; et al. (2010). Explaining job polarisation in Europe: the roles of
technology, globalisation and institutions. London: LSE (CEP Discussion
paper, No 1026).
Goos, M.; Manning, A. (2007). Lousy and lovely jobs: the rising polarisation of work
in Britain. Review of economics and statistics, Vol. 89, Issue 1, p. 118-133.
Haskel, J.; Slaughter, M.J. (1999). Trade, technology and UK wage inequality.
Cambridge, MA: NBER (Working papers, 6978).
Juhn, C. et al. (1993). Wage inequality and the rise in returns to skill. Journal of
political economy, Vol. 101, No 3, p. 410-442.
Katz, F.K.; Murphy, K.M. (1992). Changes in relative wages, 1963-87: supply and
demand factors. Quarterly journal of economics, Vol. 107, No 1, p. 35-78.
64
Katz, L.F. et al. (1995). A comparison of changes in the structure of wages in four
OECD countries. In: Freeman, R.B.; Katz, L.F. (eds) Differences and
changes in wage structures. University of Chicago Press, p. 25-66.
Machin, S.; Van Reenen, J. (1998). Technology and changes in skill structure:
evidence from seven OECD countries. Quarterly journal of economics,
Vol. 113, Issue 4, p. 1215-1244.
Maurin, E.; Thesmar, D. (2005). Changes in the functional structure and the
demand for skills. Journal of labor economics, Vol. 22, No 3, p. 639-664.
Murphy, K.M. et al. (1998). Wages, skills and technology in the United States and
Canada. In: Helpman, E. (ed.). General purpose technologies. Cambridge,
MA: MIT Press.
Spitz, A. (2003). IT capital, job content and educational attainment. Centre for
European economic research (Discussion Paper No 03-04).
Spitz-Oener, A. (2006). Technical change, job tasks and rising educational
demands: looking outside the wage structure. Journal of labour economics,
Vol. 24, No 2, p. 235-270.
65
CHAPTER 5.
Use of forecast results in skills transferability
project
Jiří Braňka
The aim of the skills transferability project was to identify transferable skills
and their role and importance with regard to the employability, adaptability
and occupational mobility of people on the labour market. One of the
project tasks was to analyse transferability of skills within and among
20 sectors and 204 occupations up to the year 2020. Project outputs
included a catalogue of skills for 219 occupations in 20 sectors of the
economy, matrixes of skills transferability, analysis of the EU survey and in-
depth interviews, examples of good practice and recommendations for
further development of skills transferability.
The project also examined in detail potential use of the Cedefop projection
for the purposes of determining future transferable skills. The research
team has come to the conclusion that the projection provides a highly
valuable and comprehensive source of information, nevertheless certain
weak points require to add further tools that are mostly still in development.
Combination of these tools will compensate for weaknesses of each
approach and provide detailed and reliable information.
5.1. Context of the forecast use

During 2010, the European Commission, through a consortium of Czech and
German partners, carried out the project ‘transferable skills across economic
sectors: role and importance for employment at European level’ (17). The aim of
the project was to identify transferable skills and their role and importance with
regard to the employability, adaptability and occupational mobility of people on
the labour market. The project consisted of these major parts:
(a) the analysis of the role of transferable skills in occupational pathways and
the fluidity of the labour market;
(b) the analysis of skills transferable across economic sectors;
(c) the analysis of the role, importance and involvement of stakeholders active
in development of transferable skills;
(d) the analysis of the instruments, tools and methods used to enhance mobility
on the labour market through transferable skills (RPIC-ViP et al., 2010).
(17) Called hereafter ‘skills transferability project’.
66
The research consortium consisted of five companies/institutions, four of

them seated in the Czech Republic (RPIC-ViP s.r.o.; Trexima spol. s r.o.;
National Training Fund; and Research Institute for Labour and Social Affairs) and
one in Germany (Institut für Wirtschaft, Arbeit und Kultur at Goethe University in
Frankfurt-am-Main).
Main outputs of the project included:
(a) skills catalogue containing a total of 292 skills sorted in three main
categories (soft skills, generic hard skills and specific hard skills);
(b) occupational profiles for 219 occupations in 20 sectors;
(c) matrixes of skills transferability within particular sectors, between particular
sectors and occupations and the evaluation of their development by the year
2020;
(d) outcomes of the EU survey (a total of 451 questionnaires and 66 in-depth
interviews among employers and representatives of the public and
administration service sector and the education sector in a total of
21 European countries (18));
(e) examples of good practice related to the development of skills transferability;
(f) recommendations aimed at further development of skills transferability within
the EU.
5.2. Sectors, occupations and skills
5.2.1. Sectors
The scope of the project encompassed 20 economic sectors (or rather groups of
sectors) which had to cover the whole EU economy. Three largest sectors had
more than 20 million employees (telecommunications, management, public and
administration service; electromechanical engineering and related services; and
health and social care) while the smallest one, ICT, had just 2.5 million
employees (Figure 5:1).
(18) Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Greece,
Hungary, Ireland, Italy, Lithuania, the Netherlands, Poland, Portugal, Romania,
Slovenia, Spain, Sweden, Switzerland and the United Kingdom.
67
Figure 5:1 Groups of sectors analysed in the project (employment in thousands

within the EU)
0 5000 10000 15000 20000 25000 30000 35000
Telecom., management, public and administration service
Manufacture of metals, electronic equip. & transport vehicles
Health and social care activities
Retail trade
Civil engineering and constructing
Education
Specialised services, postal and librarian services
Agriculture, forestry and fishing
Wholesale, warehousing and rental
Accommodation and food and beverage service activities
Transport, sewerage, security
Activities of households & other personal service activities

Manufacture of chemical and pharmaceutical products;
supply of electricity, gas, steam and water; repair of computers
Manufacture of paper, rubber & plastics products;
other manufacturing
Manufacture of food products, beverages and tobacco
Mining & engineering
Media
Manufacture of textile and leather
Manufacture of wood and furniture
ICT
Source: Eurostat, labour force survey, annual averages for 2009, own calculations.
These sectors were put together on the basis of occupational cluster

analysis. The analysis examined occupational structure in particular sectors at 2-
digit level of NACE and 4-digit level of ISCO-88 (19) (based on the labour force
survey, annual averages for 2009). Created sectors were composed of industries
that were, according to ISCO-88, most similar to one another in terms of
occupational structure.
5.2.2. Occupations
On the basis of the labour force survey analysis, key occupations were selected
in each of the created sectors. The analysis works with 204 different occupations
at 4-digit level of ISCO-88. Mid-qualified and low-qualified occupations were
(19) International standard classification of occupations.
68
preferred in the selection (20). These occupations (at 1-digit level of ISCO-88,
belonging to the major categories 3-9) are usually more vulnerable to cyclical
economic developments and restructuring processes and therefore the research
of transferable skills and their relevance in terms of employability is of particular
importance. The presence of some occupations in more sectors led to an
increase in the number of occupations to 219.
The occupational profile (consisting of a set of skills relevant for each
occupation) was defined on the basis of two main sources: the integrated system
of typal (standard) positions (21) and O*Net (22).
5.2.3. Skills
Skills catalogue, created on the basis of occupational profiles, included a total of
292 skills. The skills were divided — with regard to their fundamental
transferability characteristics — into three major categories: soft skills, generic
hard skills and specific hard skills. (Table 5:1).
5.3. Current transferability of skills

The definition of the skills categories implies that soft skills and generic hard skills
are skills of generally much higher transferability than specific hard skills. The
occupational profiles of 204 analysed occupations proved that the most frequent
20 skills represent more than a half (56%) of all employers’ requirements in terms
of skills and they are demanded by more than one third of occupations. With only
one exception (23), these skills always belong among generic hard skills or soft
skills.
(20) The selection, however, includes also a number of high-qualified occupations as

some sectors (such as ICT, education) represent a significant share in total
employment.
(21) Ministry of Labour and Social Affairs of the Czech Republic: http://ktp.istp.cz (in
Czech only). The system provides information on occupations in the Czech Republic
related to level and field of education and set of skills required, average wages and
other relevant information. It is used by Czech sector councils as background for the
Czech national occupations framework.
(22) O*Net Online: http://onet.onetcenter.org.
(23) Administration of information and documentation records.
69
Table 5:1 Definition of major skills categories (a)

Hard skills Soft skills
Range of technical, job- Not job-specific skills, which
specific abilities which require are related to individual ability
training and instruction for a to operate effectively in the
worker to become proficient or workplace either alone or with
skilled within a particular job others
reference
General skills
Skills applicable in most firms, Generic hard skills (6)
occupations and sectors
Specific skills Soft skills (22)

Skills applicable in very narrow
number of firms, occupations Specific hard skills (264)
and sectors (or even only in
one firm)
NB: Language knowledge, basic ICT skills, basic knowledge of science and technology (like mathematics)
can be described as examples of generic hard skills. Specific hard skills are applicable in very narrow
number of firms, occupations and sectors and describe special attributes of performing the occupation in
practice (e.g. flame cutting, technical drawing or engineering in machinery production).
Source: Research team (skills transferability project).
From the sectoral perspective, the most interesting aspect of the analysis is
its part focused on ‘skills intensity’ — an average number of skills required by an
occupation. The outputs show that performance of an occupation (one of the
analysed occupations) within the ICT sector requires 22 skills as this sector is
one of the most demanding in this respect, while the manufacture of textile and
leather sector requires only 13 skills and, therefore, ranks at the end of the group
of sectors. The analysis does not consider a relative complexity of acquiring a
particular skill (this might disadvantage some sectors, e.g. health care). The
output of the analysis has been also determined by the fact, that within the
particular sectors, only the top 10-15 key occupations (in terms of number of
employees) have been considered.
The skills intensity analysis offers a reliable basis to determine whether the
given sectors are internally homogeneous in terms of key occupations’
requirements and whether it is easy or complicated for a worker performing an
occupation to be retrained for the performance of a different one. ‘Skills
transferability’ represents a result of the analysis and it has been defined as
follows: number of skills transferable for all pairs of occupations within each
sector was quantified and divided by the number of cross-possible combinations
among occupations (e.g. there could be 45 possible cross-combinations for a
sector with 10 key occupations, etc.). The resulting number was called skills
transferability. It, for example, says that ICT sector is proving highest skills
70
transferability, because one can find on average almost 17 skills that are
transferable between two key occupations within the sector.
Both indicators might be further used to calculate the ‘transferability index’
quantifying the ratio of the skills that are transferable between two key
occupations within the sector and an average number of skills per key
occupation.
Skills transferability
Transferability index:
Skills intensity
The result of this calculation for example says: the education sector has the
highest skills transferability index, because an average key occupation within this
sector has 18 skills and on average 15 of them (which is 80%) are shared with
some other occupation.
This proves that in some sectors (those with a higher skills transferability
index) it is easier to switch from one job to another, while in other sectors it is
much more complicated. This has some impact on sectoral labour market
mobility.
Although identification of transferable skills between occupations can be
considered as very interesting output, its practical usage is limited by the fact that
only occupations in the same group of sectors are compared with one another.
Limitations arise when choosing a new occupation outside the given sectoral
group, which can be caused by both individual preferences and economic
downturn affecting the whole sectoral group and leading to a lower number of job
opportunities.
5.4. Future transferability of skills ― General

approach
One of the project tasks was to analyse transferability of skills within and among
sectors and occupations up to the year 2020.
The identification of future skills needs can be generally based on
quantitative or qualitative approaches. The first one is represented by quantitative
projections of future supply of and demand for skills, for example Cedefop
projections (Cedefop, 2009a; 2009b), while the second one is focused more on
71
qualitative description of future skills needs without any quantifications of the

skills needs, for example 18 EU sector studies published in 2009 (24).
Speaking of the EU level, both approaches have been in development
during recent years. The research team consisting of the Institute for Economic
Research at Warwick University, Cambridge Econometrics, Research Centre for
Education, the Labour Market (ROA) at Maastricht University and Alphametrics
Ltd develops medium- to long-term skills projections at EU level, broken down by
country, sector, education level and occupational group for Cedefop. The EU
sector studies provide an outlook on key skills to be required in selected sectors
of the economy in respective occupational groups based on different scenarios.
Since both approaches have their indisputable advantages (but also
weaknesses) and there is a lot of high-quality results based on both qualitative
and quantitative methods, the research team decided to build on existing
research work related to forecasting and matching skills and concentrate on their
usage bringing new added value rather than developing and testing a completely
new methodology. The approach of the research team was to identify, analyse
and use results of available future-oriented studies and projections and adjust
them to acquire a relevant and reliable base for additional analyses. Table 5:2
summarises specification of future skills needs analysis developed in this study,
and the specification of main information sources used for it, i.e. Cedefop
projections and 18 EU sector studies.
The analysis of EU sector studies focused on future knowledge and skills
needs recognised in 18 EU sector studies published in 2009 and identified skills
that were pivotal in key occupations within these sectors. Studies, however, do
not describe all necessary skills, but focus on those most important in relation to
main drivers of future development in each sector (Rodrigues, 2007).
Information on future skills needs development varies — in terms of
occupations ― by the level of detail (it is provided on 1-3-digit levels of ISCO-88).
In case of these occupational groups, the research team analysed for example
the most transferable skills among particular sectors. The analysis is based on
counting the share of occupational groups or sectors or both, where separate
knowledge and skills will be applicable (example in Table 5:3).
(24) European Commission, Employment, Social Affairs and Inclusion, sectoral level
analysis web pages. Available from Internet:
http://ec.europa.eu/social/main.jsp?catId=784&langId=en.
72
Table 5:2 Analysis of forecast approaches and their application

Skills transferability Cedefop forecast 18 EU sector studies
project’ forecasts (2010 version)
requirements
34 economy sectors (NACE
20 sectors covering whole 41 sectors covering whole
Coverage 1.1) and approximately ⅔ of
economy (NACE rev 2) economy (NACE rev 1.1)
total employment
Forecast
2020 2020 2020
period
One scenario, otherwise Base, Optimistic,
matrixes that include 292 Pessimistic (based on Up to four scenarios per
Scenarios
skills and 219 occupations scenarios of the recovery study
will be too complicated from economic crisis)
Soft, generic hard and
None (skills represented Studies concentrate on soft
Skills analysis specific hard skills analysis
only by level of education) skills analysis
required
Selected ISCO one to 3-digit
Occupations ISCO 4-digit level ISCO 2-digit level level
analysed (219 occupations) (27 occupational groups) (usually 8-12 occupational
groups)
The impact of
economic Considered Considered Not embodied
crisis
Table 5:3 An example of analysis of highly transferable skills across sectors

Occupational
Highly transferable skills Across how many sectors?
group
IV Business Legislative, regulatory (94.4), 9 sectors (Printing and publishing,
professionals Flexibility (94.4), Language (88.9), E- Chemical, pharmaceutical, rubber
(ISCO 241) skills (83.3), Analytical skills (79.6), and plastic manufacturing, Non-
Understanding suppliers and metallic manufacturing, Computer,
customers (79.6), Communication electronics and optical manufacturing,
(77.8), Intercultural (77.8), Stress and Furniture manufacturing, Distribution
time management (75.9), Networking and trade, Transport and logistics,
(70.4), Business development (70.4) Posts and telecommunication)
5.5. Future transferability of skills ―Cedefop forecast
5.5.1. Methodological approach and its limitations

This analysis of future skill needs develops findings of Cedefop projections and
links them with structure and profiles of sectors and occupations created in the
skills transferability project. The task of this analysis was to describe the shift in
demand for skills in the period 2008-20.
73
The process of transition of Cedefop projections data to skills transferability

project methodology includes the following steps:
(a) transition of sectors in Cedefop projections (41 sectors by NACE 1.1) to
sectors in Skill transferability project (20 groups of sectors by NACE rev. 2);
(b) recalculation of growth/decline rates in employment for occupational groups
(2 digits ISCO) in accordance with new sectoral division of EU economy);
(c) recalculation of growth/decline rates in employment for key occupations
(4 digits ISCO) defined for groups of sectors in skills transferability project;
(d) scaling of significance of the growth/decline rates by current share of key
occupations (4 digits ISCO) on total employment in respective group of
sectors;
(e) the calculation of impact of scaled growth/decline rates on demand for skills
associated with key occupations;
(f) the calculation of total change in demand for skills in the period 2008-20.
The process of use of Cedefop projections in skills transferability project had

also some limitations caused by the quality and structure of input data available.
These limitations include:
Problem 1: There is a default presumption that the level and range of skills
required for key occupations will not change during analysed
period. This is a serious simplification of the structural changes
that in reality affect the labour market. But it is not possible to
analyse future shifts in skills needs for each and every key
occupation defined in this project.
Solution: Research team uses current occupational profiles for forecasting
future skills needs in this exercise for two reasons. The first reason
is the lack of relevant information on skills needs development
related to the analysed occupations, the second is the fact that the
results of EU survey carried out among the employers implied that
the occupational profiles of most occupations will not show
substantial changes, it is rather expected that the importance of
some of the key skills will grow.
The problem might be solved by links to the outputs of sectoral
studies or Cedefop project ‘Pilot employer survey on skills need in
Europe’, however, it is vital to coordinate the focus of relevant
projects in terms of examined occupations’ structure.
Problem 2: There are no scales for skills importance in occupational profiles
which means every skill associated with an occupation has the
same level of significance ― this also simplifies the reality.
74
Solution: Unfortunately, there is no tool at European level that can provide

such information. Waging of skills importance can be set by
groups of experts (like sector skill councils) or (in a survey) by
representative sample of employers who employ the respective
occupations. The latter option is used in O*Net and the core of
such European tool is under development in the above mentioned
project, pilot employer survey on skill needs in Europe. This
project deals (among others) with importance of skills in
occupational profiles for 3-digit ISCO, which may add significant
value to the knowledge base regarding skills, their transferability
and importance.
Problem 3: The replacement demand for occupational groups within economic

sectors (e.g. job opportunities created by those who leave existing
jobs either permanently or temporarily) is not considered because
of insufficient data.
Solution: Cedefop should be able to provide data on replacement demand
for occupational groups within economic sectors in the foreseeable
future.
Problem 4: The process of transition of Cedefop occupational groups to those

in key occupations was difficult because the information on future
employment was available for less-detailed ISCO only.
Solution: Cedefop projection disposes of 21 occupational groups, while the
skills transferability project required gathering information on
outlook for 204 occupations. While this task seemed impossible,
initial findings proved that it may be achieved in many cases by
using projections for separate economy sectors. ISCO 2-digit
occupation often represents just one or two ISCO 4-digit
occupations within one sector (see Table 5:4 for an example of
occupations within retail trade sector) and therefore (using the
sectoral approach) the outlook for a significantly larger number of
more specific occupations is feasible.
An example of the retail trade sector shows that 15 largest ISCO 4-digit
groups in retail trade are represented by no less than 11 ISCO 2-digit
occupations. It was possible to use the Cedefop projection for 50-75% of
occupations in each of 20 defined sectors.
75
Table 5:4 The largest occupational groups in retail trade sector
ISCO 2 digit ISCO 4 digit

52 5220
13 1314
12 1224
42 4211
93 9330
34 3415
34 3419
41 4190
41 4131
32 3228
42 4220
22 2224
91 9132
83 8322
91 9151
5.5.2. Outputs and interpretation

These methodological issues will significantly affect reliability of the outputs.
Given the current availability and particularity of data, the analysis can be used
only for a rough estimate and determination of major trends in future skills needs
development.
The outputs can be divided in two parts: occupational outlook and skills
outlook.
Occupational outlook shows that the number of declining occupations (93)
is higher than the number of growing occupations (69) while employment in 51
occupations will change only slightly (25). Total employment in analysed
occupations will decrease by 2.3% in the respective period.
Because of above-mentioned limitations, the research team decided not to
use exact growth/decline rates in employment (the process of transition highly
influenced their reliability). They were aggregated into seven-stage scale, which
enables to identify general trends in expansion demand rather than exact but less
reliable percentage changes. This scale uses these stages:
(25) The fact that the number of occupations expected to show an overall decline in
employment is higher than the number of occupations expected to grow is
significantly influenced by the methodology of their selection. Medium and low
qualified occupations, representing most analysed group of occupations, are
forecasted lower growth than highly qualified occupations.
76
Table 5:5 Growth/decline rates

Stage Description
High growth 25% and more
Medium growth 15% till 24%
Small growth 5% till 14%
No change -4% till 4%
Small decline -5% till -14%
Medium decline -15% till -24%
High decline -25% and more
Major trends regarding occupations and sectors are described in Table 5:6.
Table 5:6 Growth/decline rates in occupations

Number of
Stage Sectoral/occupational trends
occupations
Selected occupations (both lower and highly qualified in sectors
High growth 16 healthcare, media and telco, management and administration
service
Selected occupations distributed among low, medium and high
qualified in number of sectors, namely HoReCa, retail,
Medium growth 20
transportation and telco, management and administration
service
Number of highly qualified technical-oriented occupations in
Small growth 33 industry sectors and management and also some occupations
in HoReCa and retail
Building construction, ICT, household activities, healthcare,
No change 51 telco, management and administration service and specialised
services
Lower-qualified occupations in food, electro mechanical
Small decline 50 engineering and printing and publishing, most occupations in
education, some IT, healthcare and specialised services
Selected occupations (mostly low qualified) in electro
Medium decline 15 mechanical engineering, wood processing, mining and energy
supply
High decline 28 Many occupations in textile, mining and agriculture.
Skills outlook is derived from occupational analysis. Major trends regarding

changes in skill demand are as follows:
(a) generic hard skills will experience slight decline in importance (0-5% in
total appearance), mostly for decline of total employment (in key
occupations). Probably because of projected structural changes in
employment (many jobs will be cancelled in industry ― low and medium
qualified, technical oriented staff) the basic skills in science and technology
will lose most importance (-4.25%);
77
(b) specific hard skills will lose some importance as well (-5.36%), again partly
because of the development of total employment in key occupations and
partly because of structural changes. The analysis of development in
demand for specific hard skills must take two things into account:
(i) relative change in demand (changes in less transferable skills appear
to be more significant);
(ii) total change in demand (highly transferable skills will usually show
smaller changes).
Major specific hard skills that will lose part of their significance for key
occupations are those connected with low or medium qualified occupations
in the industry (e.g. technical drawing; orientation in technical
documentation; appraisal and control of quality of raw materials, semi-
products and products; or handling of machines for metal processing). Many
skills related to agricultural production (e.g. growing of plants and crop-
plants or operation of agricultural and forestry mechanisms) are also among
those which lose importance both relatively and absolutely.
Specific hard skills that will gain importance both from relative and absolute
points of view are usually those connected with rising share of services in
sectoral structure of EU economies and with growing importance of
effectiveness of businesses, for example representing clients in
communication with institutions; bookkeeping, accounting and taxes;
applying knowledge of risk management; elaborating, production and
financial plans; or applying knowledge of economics and entrepreneurial
subjects;
(c) more than half of soft skills show decline in importance. However,
particularly here it is necessary to emphasise one of the limitations of the
approach ― it was not possible to involve also qualitative development in
skills demand regarding respective occupations. It is very likely that many of
the soft skills will be positively affected by the process of growing demand
for skills in most occupations. Some soft skills that seem to be less important
(both relatively and absolutely) in time are also those that appear in profiles
of many occupations that are subject to job cuts.
Nevertheless, the analysis confirms some results of the EU survey among
employers, education and public service institutions (survey was conducted
in the skills transferability project) concerning future demand for soft skills.
Many respondents mention requirements of employers regarding employees’
ability to work independently, actively and with growing care for customer
needs. The major growing soft skills from the analysis are initiative, proactive
78
approach and customer orientation. The fastest growing soft skill is

leadership;
(d) reliability
Because of adopted simplifications, the results of future transferable skills
analysis must be regarded with caution and as an exercise which can be
developed and improved in successive projects;
(e) use of the analysis
These outputs allow performing additional research and some actions aimed
at skills transferability and its support. For example it will be possible to sort
out skills according to their transferability and future prospects which may be
illustrated by the matrix in Table 5:7.
Finally, the interconnection between Cedefop projections and analysis of
skills requirements for key occupations may be used for identification of
possible solutions to future imbalances in demand and supply of skills. For
example, it is possible to identify the level of proximity of occupations, in
terms of skills matching (the higher skills match, the higher their
transferability among occupations and thus occupational mobility). The
future skills analysis shows in which occupations people may have problems
to find employment and the skills matching analysis then may show in which
(prospective) occupations they may be easily and successfully retrained;
(f) summary
The analysis of future transferability of skills is achievable, but it relies on
multiple tools that are either already used or currently being developed.
These tools have to be combined in order to compensate weaknesses of
each approach and to provide detailed and reliable information (Figure 5:2).
Table 5:7 Example of additional outputs
TRANSFERABILITY
Small Medium High
Insufficiency of skills Priority concern Some action may be needed

FORECAST
Balance No need for action
‘Surplus’ of skills Priority concern Some action may be needed
79
Figure 5:2 Possible tools for identification of future skills and their transferability
Labour force survey (LFS) LFS analysis can be used Cedefop projections
Statistics on 3-4 digits for identification of future 27 occupational
ISCO in economy sectors development for 3-4 digits ISCO groups in 41 sectors
Key occupations in sectors and their

future growth/decline in importance
New occupations,
qualitative trends, INTERACTIVE PLATFORM
details about specific FOR IDENTIFICATION OF
occupations that Importance of skills for
TRANSFERABILITY OF
cannot be captured occupational profiles
CURRENT AND FUTURE
by projections SKILLS IN EUROPE
Sector studies ESCO

Employer Survey
Detailed analysis of EurOccupations
Importance of skills Sector Skill Councils
occupational structure
for occupational
and demand and Occupational profiles
profiles
supply of skills in
sectors
References
Cedefop (2009a). Future skill needs in Europe: medium-term forecast. Background
technical report. Luxembourg: Publications Office of the European Union.
Cedefop (2009b). Future skill supply in Europe: synthesis report. Medium-term
forecast up to 2020. Luxembourg: Publications Office of the European Union.
RPIC-ViP et al. (2010). Transferability of skills across economic sectors: role and
importance for employment at European level. Interim report.
Rodrigues, M.J. (2007). Innovation, skills and jobs. Pilot project to develop a
European foresight methodology to identify emergent jobs and their skills
needs. Working document dated 20 March 2007 produced after the
methodological workshop held by the European Commission, Brussels, 1-2
February 2007.
80
Part 2:
Sharing other European
forecasting experiences
The AMS-skills barometer:

an Austrian web-based labour market information tool
Stefan Humpl and Daniel Bacher
Forecasting skills demand and labour market dynamics in the Baltic States
Žilvinas Martinaitis
Demand and supply of labour by education in Norway towards 2030: linking
demographic and macroeconomic models
Bjornstad Roger and Nils Martin Stølen
Methods and results of skills demand and supply forecasting:
the case of Germany
Tobias Maier
FGB model: an innovative tool for the analysis of jobs and skills needs in Italy
Massimiliano Tancioni
The forecast of occupational structure of employment in Poland
Artur Gajdos
Future development of the educational level of the population of
Switzerland
Jacques Babel
The supply of vocational adult education in Finland
Errki Laukkanen
How to anticipate skills and training needs by 2020:
a regional point of view from France
Marie-Béatrice Rochard
Matching formal skills in Slovakia: LFS based evidence
Marek Radvanský
Internet job search data as a possible source of information on skills
demand: example of Slovak university graduates
Miroslav Štefánik
The protective effect of field specific and general skills against over-
education under different conditions of labour supply and demand
Martin Humburg, Andries de Grip and Rolf van der Velden
CHAPTER 6.
The AMS-skills barometer: a web-based
labour-market tool. Making use of forecast
results by stakeholders
Stefan Humpl and Daniel Bacher
The paper illustrates the AMS (26) skills barometer as a web-based labour-
market information tool in the context of the EU agenda for new skills and
jobs and Cedefop initiatives of identifying and anticipating future skill needs.
Cedefop on the one hand produces regular, systematic and quantitative
skills forecasts at pan-European level. European forecasts, however, are
not a substitute for national level projections. Cedefop emphasises that its
value added lies in the provision of a common framework for discussion.
Results are relevant not only to experts and policy-makers, but also to the
broader public.
The AMS-skills barometer on the other hand is a good practice example in
Austria developed to provide a broad group of different users with valid,
comprehensible and well-structured information on current and medium-
term occupation and qualification needs. Available information from
different forecast sources is exploited and merged in a synoptical way. The
chapter discusses the value added of possibly integrating Cedefop results
to link European and national forecasts.
6.1. Introduction
In the rapidly changing economic world, Europe’s future position will be defined
by growth, which should be smart, sustainable, and inclusive, and will also
depend on jobs for the wellbeing of its citizens. Therefore, the forecast of
occupations and qualifications to have a flexible, but well qualified, labour force
for the challenges of this development is one of the key issues in the design of
the future European qualification systems. Several initiatives like new skills for
new jobs are focusing specifically on skills upgrading, matching and anticipation.
It aims at developing more effective ways to analyse and predict which skills will
be required in tomorrow's labour markets and, with this knowledge, developing
and adapting education and training so that the workers gain the skills required.
(26) Arbeitsmarktservice, Austrian Public Employment Service.
82
Cedefop clearly contributes to those initiatives by identifying and anticipating

future skill needs, potential skill mismatches and skills needs in specific sectors.
Its pan-European forecast on skills demand and supply developed in cooperation
with the Skillsnet network provides ‘the most comprehensive and consistent set
of skill projections ever produced for Europe’ (Cedefop, 2008, p. 16). European
forecasts, however, are not a substitute for national level projections. Cedefop
emphasises that its value added lies in the provision of a common framework for
discussion. Such discussion, however, is not only relevant to researchers,
experts and policy-makers. As Cedefop notes: ‘It is important to bear in mind that
the audience for these results is not just policy-makers. It is crucial to escape
from the thrall of planning — and recognise that such information is crucial to
making markets work better by better informing individual choices. This raises
significant issues of dissemination. There is a need for better web-based
tools’ (Cedefop, 2010, p. 81-82).
The question of the use of forecast results will be at the centre of this
chapter. The AMS-skills barometer is the first comprehensive nation-wide online
labour-market information system in Austria. It was designed for the Austrian
Public Employment Service in 2002 and has developed further since then.
Basically the AMS-skills barometer provides a broad group of different users with
valid, comprehensible and well-structured information on current and medium-
term occupation and qualification needs. It is a tool to inform not only ‘expert
groups’ about labour-market needs, but much more so to guide individual people
on jobs and qualifications. Along with recent trends in vocational research the
tool also attempts to pay more attention to skills and competences. In the
following, some background information on the basic structure and methodology
of the AMS-skills barometer as a web-based labour-market information tool will
be given. Especially, the value added of having such a tool for different user
groups will be dealt with. Eventually the mutual benefits of linking the AMS-skills
barometer and the Cedefop pan-European forecasts will be discussed. A final
conclusion outlines possible ways forward.
6.2. Short history of the AMS-skills barometer

The Austrian Public Employment Service (Arbeitsmarktservice, AMS) was
outsourced from the Federal Ministry of Employment, Health und Social Affairs to
become a service enterprise under public law in 1994. Its primary task is to do
labour-market matching by assisting job-seekers and companies in their search
and providing various services such as information, advice or finance. The AMS
83
is Austria’s leading provider of labour-market services (27). It works in close

cooperation with labour and employers organisations (28).
In the late 1990s the AMS reworked its labour-market information system. At
that time, labour-market information was mainly targeted at young people
entering the labour market and encompassed only a few specific occupations.
Job counselling was also complicated by no uniform source of information. The
AMS responded by developing an encompassing labour-market information
system for job and career guidance and counselling. For this purpose the AMS-
skills barometer was implemented with the main objective of providing a broad
group of different users with valid, comprehensible and well-structured
information on current and medium-term occupation and qualification needs. As
job flexibility gained in importance at the expense of stable job requirements
defined by occupations, the new system integrated as well comprehensive
information on skills and competences over time (Humpl and Kargl, 2008).
The AMS-skills barometer as an online tool for labour-market information
was designed and developed by 3s Unternehmensberatung and the Institut für
Bildungsforschung der Wirtschaft (ibw, Institute for research on qualifications and
training of the Austrian economy) being two important players in labour-market
consultancy and research in Austria (Humpl and Kargl, 2008).
6.3. Basic structure of the AMS-skills barometer

The AMS-skills barometer is based on two different systems for labour market
representation: An occupational classification and one of qualifications in the
sense of skills and competences. Both are integrated within the tool, so that
information on skills and competences for specific occupations can be portrayed.
The user can navigate through the tool by using given frames, links and buttons.
Figure 6:1 shows the basic structure of the barometer which will be explained in
more detail in the following.
6.3.1. Occupational information

The occupational classification distinguishes 24 occupational areas
(Berufsbereiche), 95 occupational fields (Berufsfelder) and several occupations in
each field. These occupational areas do not follow any other classification (like
(27) About AMS: http://www.ams.at/english/14595.html.

(28) AMS history: http://www.ams.at/english/14603.html.
84
ISCO or the former Austrian classification of occupations), but was designed for
the purpose of describing the occupational structure of a modern labour market.
Figure 6:1 Basic structure of occupations and skills in the AMS-skills barometer
Hierarchical Occupational Skill

level structure structure
Level 1 24 occupational areas 23 skills areas
95 occupational fields Appr. 230 skills

Level 2
Level 3 Appr. 600 occupations Appr. 8 000

detailed skills
Source: 3s.
Single occupations can also feature in more than one occupational field.
Hence, occupational information follows a three-part hierarchy whereby ‘bigger’
occupational areas are broken down to ‘smaller’ occupational fields and
eventually occupations (Figure 6:1). More detailed information on single
occupations can be found in another online information system of the AMS – the
‘Vocational Information System’ (AMS-Berufsinformationssystem) – which is
linked to the website (around 600 occupations) (29). The AMS-skills barometer
provides also gender specific information for each occupational field (Humpl and
Kargl, 2008).
For each of the 24 occupational areas and 95 occupational fields labour
market needs and trends are illustrated. Figure 6:2 shows the occupational area
‘information technology’ whereby overall trends are visualised with a diagram.
Further information on trends is received by using the button ‘more information’.
Within the occupational area information technology also data on previous and
current vacancies and future trends of the respective occupational fields within
information technology are portrayed. Trends are visualised by arrows and the
current labour-market significance of the occupational fields by dots. Exact
(29) AMS-Berufsinformationssystem: http://www.ams.at/bis/.
85
numbers are only stated for jobs available (at print media and the AMS) in the
last two years and current vacancies at AMS.
Figure 6:2 Occupational area ‘information technology’
Source: AMS-skills barometer.
For each single occupational field, such as ‘software technology and

programming’ categorised under ‘information technology’, estimates of jobs
available in the previous two years, current vacancies and forecasts are provided
for several occupations in the same manner as before (Figure 6:3).
At this level the AMS-skills barometer is also directly linked to jobs available
at the AMS — the ‘eJob-Room’ — being the largest job exchange in Austria (see
right hand side of Figure 6:3 AMS button).
86
Figure 6:3 Occupational field software technology and programming
6.3.2. Qualification information

The tool integrates a second ontology of skills and competences. Similar to
occupational areas the AMS-skills barometer defines 23 skill areas, also
designed especially for the purpose of the description of a modern labour market
and not following any other classification of skills and competences. Those 23
skill areas are again subdivided into approximately 230 skills at the level of
occupational fields. The two taxonomies are also linked at this level (Figure 6:1).
Every occupational field integrates information on relevant knowledge, skills and
abilities.
The skills ontology is highly differentiated as it does not only focus on skills
acquired on the basis of formal education. It integrates skills acquired through
non-formal education and informal learning at the workplace. The skills ontology
87
is visualised by differentiating occupation specific skills (fachliche Qualifikationen)

and non-occupation specific skills (überfachliche Qualifikationen). The latter are
for instance personal and social skills, language skills, generic skills or
transversal skills. Figure 6:4 shows the representation of skills at the AMS-skills
barometer at the level of occupational fields whereby dots illustrate current
labour-market significance of respective skills and arrows the forecasts. More
information on skills (around 8 000 detailed skills) is given at the third level of
occupations linked to the AMS-vocational information system (Humpl and Kargl,
2008).
6.3.3. Regional information

A last important feature regarding the basic structure of the tool is the regional
dimension. The AMS-skills barometer provides regional labour-market
information as well for the nine Austrian federal provinces. Such information was
integrated right from the beginning, even though the main focus of the tool is at
national level. Regional information is still guided primarily by occupational
information, even though skills trends are referred to in texts. Administrative
boundaries are not necessarily the same as regional labour markets. The
barometer partly refers to regional specifics. Recently, the limited database at
this level could be improved by using a regular employer survey carried out by
the AMS (Humpl and Kargl, 2008).
6.4. Methodology
The methodological approach applied during the process of establishing and
updating the AMS-skills barometer is one of editing existing labour-market
relevant information. This is done by a process of searching sources, collecting
and evaluating information critically, joining these together and finally editing
them along the given structures of the barometer. The result, both textual and
graphic, can be looked upon as a synopsis of compiled information (Humpl and
Kargl, 2008).
A general update of the online tool AMS-skills barometer takes place every
March and October. Editing is done by a team of experienced authors. They
follow a common framework of ‘working rules’ that guarantee that all texts within
the barometer are homogenous with regard to content and style. Therefore it is
first and foremost an instrument for representation and only secondary of data
analysis (Humpl and Kargl, 2008).
88
Figure 6:4 Occupational field software technology and programming
The sources of information are both quantitative and qualitative:

(a) yearly analysis of job advertisements commissioned by the AMS;
(b) statistical AMS data on job vacancies;
(c) regular enterprise survey in more than 20 000 firms with at least 20
employees. Survey is carried out by regional offices of AMS. Allows for
quantitative occupational information and information on skills and
competences;
89
(d) AMS standing committee: three yearly sectoral discussions between

experts, company representatives and educational providers for 10 different
sectors;
(e) quantitative data (labour-market statistics from data providers like Statistics
Austria, public insurance association, etc.);
(f) research, analyses and forecasts on skills demand for the Austrian labour
market (at regional and national level);
(g) relevant texts from a broad range of sources ranging from general
publications (even articles from mass media) to specific branch journals;
(h) relevant PhD- and Master-theses.
To use those sources for representing current labour-market needs and

provide for short to medium-term forecasts (up to three years) the editing process
is crucial. To allow for solid forecasts and a common framework, several planes
of analysis have to be considered. First, since many of those sources use labour-
market classifications different from the AMS’s own system, the sources must be
critically adapted to the AMS structure if possible. Second, it is necessary to
consider levels of differentiation. The AMS-skills barometer delivers information
for occupations and skills at a highly detailed level. Other sources often focus on
trends for specific sectors (not necessarily labour-market specific) or focus on
skills on a more general level. Third, sources have different research scales. The
barometer provides information at national and regional level, whereas studies
dealing with certain regions or branches have limited value for forecasts within
the tool. Fourth, differences in time frames have to be dealt with. The barometer
has a time span of up to four years and other sources must be adjusted to this
time frame (Humpl and Kargl, 2008).
The quality of forecasts can be improved by making use of several sources
for occupations and skills. Qualitative expert interviews are often useful for
validating and supplementing available sources. Recently, quantitative data such
as the mentioned large enterprise survey in more than 7 000 companies has
been integrated. This study allows for the quantification of skills at regional level.
Generally, the classification of skills is getting increasingly important at the
expense of the concept of occupations (Humpl and Kargl, 2008).
6.5. User groups of the AMS-skills barometer

The objective of the AMS-skills barometer is to provide a broad group of users
with relevant labour-market information. Among them are individual job-seekers,
labour-market counsellors, enterprises and policy-makers in public agencies.
90
More detailed knowledge about the preferences of user groups has been
gained by regular surveys. The latest user survey was carried out in 2010. It
delivers information on the purpose of visiting the website, time spent using the
tool, how often it is used, usability of the tool, satisfaction, recommendations for
improvement and information on the user itself (Ziegler, 2010).
Results of the previous survey show two main groups of users. Most people
use the tool for the first time for seeking jobs and information about jobs. A
smaller group of regular users not only search for jobs and vocational
information, they also research for labour-market counselling and information.
This means that the tool is also used by a group of ‘expert users’. Such a
conclusion is confirmed by taking a look at personal information of users. About
half of the first time users completed education up to the level of primary or
apprenticeship. Another 30% completed secondary education and 20% tertiary
education. Among the regular users about 40% have tertiary education and only
30% completed primary education or apprenticeship. This shows that among the
regular users the qualification level is much higher (Ziegler, 2010).
The use of the tool by a group of experts is also visible by looking at the
professional background of users irrespective of the pattern of use. Of first time
users, 39% were job-seekers, 17% were still in education, 21% were working for
private companies and 23% for the AMS, educational providers, public
administration, NGOs or social partners. This does not change substantially by
comparing it with the group of regular users (Ziegler, 2010).
The most popular sites at the AMS-skills barometer are vacancies, job
information, labour-market developments, occupational and qualification
structures and information on occupational fields. In this regard, there are no
major differences between new and regular users (Ziegler, 2010).
A challenge is certainly to provide the same kind of information to quite
different user groups. In the future, information for the broader public and for
experts might be separated to allow for more customised representation of
results.
6.6. Comparing Cedefop forecasts and the AMS-skills

barometer
Cedefop forecasts follow a regular, systematic and quantitative approach to skills
supply and demand forecasting based mainly on a regional econometric input-
output model (E3ME). At the same time, Cedefop acknowledges the need to
combine quantitative and qualitative approaches. It calls for a ‘multifaceted
approach’, whereby quantitative and qualitative approaches to anticipation are
91
merged to provide labour-market information to various user groups (Cedefop,

2008, p. 20-21). In comparison, the AMS-skills barometer follows primarily a
qualitative approach on labour-market needs (demand), even though quantitative
sources are integrated. The tool depends to a great extent for its representation
on occupational and qualification information on other sources. A logic next step
is to ask what Cedefop and the AMS-skills barometer have in common and which
differences exist between the two approaches.
Therefore, it is necessary to compare classifications, differentiation level,
scale information and time frames. Those criteria were already outlined in the
section on methodology given that an integration of sources into the barometer
needs to be edited along those lines. In addition, the question of user groups
needs to be addressed.
6.6.1. Classifications
Cedefop forecasts translate employment projections from a multisectoral
macroeconomic model into demand for skills. Cedefop provides comprehensive
information by sector (NACE classification), occupation (ISCO) and qualifications
(ISCED). The AMS-skills barometer designed its own classification on
occupations and qualifications adequately describing the Austrian labour market.
It is not following a sectoral approach, because for the important user group of
job-seekers occupational information is the priority. The barometer is also not
following the ISCO classification, even though it is possible to translate some
ISCO occupations into AMS occupational taxonomy. Pilot projects for that
purpose have been carried out. Besides, forecasts using simple forecast
methods (exponential smoothing, Holt-Winters) were piloted trying to combine
AMS data (AMS-datawarehouse) and macro data (Mikrocensus, based on ISCO
and NACE) (Bacher et al., 2007). As the educational census data are under
development still, and this information would be crucial for further econometric
forecasts, we have to wait for this information for further developments.
6.6.2. Level of differentiation

Cedefop offers a comprehensive macro view of sectors, occupations and
qualifications. But qualifications are only broadly represented by differentiating
low, medium and high qualification. Cedefop notes that at present it is possible
‘to undertake only a broad brush analysis’ (Cedefop, 2008, p. 59). The AMS-skills
barometer, in comparison, provides highly differentiated information on
occupations (around 600 occupations) and skills (around 8 000 detailed skills).
Skills representation is not only based on formal education, but also on non-
92
formal education and informal learning. Hence, the barometer is a tool at micro
and meso level.
6.6.3. Scale level:

Cedefop forecasts focus on European and national level. Thus, forecasts cover
the EU-27 (forecasts in 2008 not yet included Bulgaria and Romania), Norway
and Switzerland. Cedefop emphasises that their forecasts do not substitute
national work on forecasting. ‘Pan-European projections should not compete with
forecasting work of individual countries, which draws on many years investment
in data, systems and knowledge, but offer a common framework and a consistent
set of underlying assumptions. They therefore can provide a context for more
detailed national analyses and forecasts’ (Cedefop, 2010, p. 16).
The AMS-skills barometer focuses mainly on the national level, but
integrates regional information and forecasts for the Austrian federal
provinces (Cedefop, 2010).
6.6.4. Time period

Cedefop provides short- and medium-term forecasts based on the econometric
E3ME sectoral model. In addition, Cedefop is able to provide long-term forecasts
up to 2020 being based on methods of computable general equilibrium (CGE)
models. The AMS-skills barometer, in comparison, provides only short- to
medium-term forecasts up to three or a maximum of four years.
6.6.5. User groups

Cedefop notes that its analyses and forecasts are relevant to a broad audience:
‘the latter should include: policy-makers, stakeholders, social partners, sectoral
organisations, practitioners and individuals. This project is concerned with
developing quantitative models and projections. The results are aimed at all
these audiences, across all the countries of the EU (plus some associated
members)’ (Cedefop, 2008, p. 12).
Despite the relevance of results for a broad audience, one can assume that
forecasts are used primarily by an expert group. Cedefop itself states that results
are not just relevant to policy-makers and underlines the need for ‘better web-
based tools’ to disseminate results. But the interest for results is not just a
question of dissemination. The methodology behind Cedefop forecasts is
complex, results do not allow for manpower planning and macro forecasts are
less tangible for individuals. Such conclusion one can draw on the basis of the
experience with user groups of the AMS-skills barometer.
The use of the tool by experts and individual job-seekers differs
substantially. Individuals using the tool a first time are mainly people searching
93
for a job or re-orientating themselves. They favour information on job vacancies

and information on jobs more than developments on the labour market or
systematic information on occupation and skills. Regular users, among them a
significant group of expert users, are not only interested in job vacancies and
information. They see labour-market developments and information on
occupations and skills as equally important. Experience shows as well that
seemingly contradictory information due to different sources pose problems to
the broader public. This means that the interest for European forecasts is not just
a matter of dissemination, but equally important, one of representation.
6.7. A way forward: integrating European and national

forecasts
The paper illustrated the AMS-skills barometer as a web-based labour-market
information tool in the context of the EU Agenda for new skills and jobs as well as
Cedefop initiatives of identifying and anticipating future skill needs. Cedefop
provides the most comprehensive pan-European skills forecasts including
country workbooks ever produced for Europe. Despite this achievement, Cedefop
views the current results as a starting point. Further analyses of forecast results
as well as methodological and data improvements remain an important task. And
even more imperative is to intensify dialogue with various stakeholders to
establish a common European framework and better disseminate
results (Cedefop, 2010, p. 81).
In this context, the AMS-skills barometer as a tool for labour market
information and skills forecasts was presented. The question on how to possibly
link both approaches remained open. A further integration of both approaches
could be of use concerning the information flow to different user groups, but this
would need more intense and detailed discussions by stakeholders. Possible
synergies should be reflected as points of common interest, but still the
differences in methodologies, scale, level of differentiation, and classifications are
challenges to be overcome in the respective common development.
The barometer’s own taxonomy of occupations and skills is only partly
compatible with most other forecasts including Cedefop classifications. Similarly
are the differentiation levels substantially different. Cedefop provides macro
projections and the barometer primarily micro and meso projections. Its highly
detailed description of occupations and skills is currently not compatible with
Cedefop’s focus on sectors, occupations and broad skills levels. The scale level
and time periods are also quite different. Cedefop offers European and national
forecasts over the short, medium and long term. The barometer focuses on the
national and regional level over the short and medium term.
94
Despite those differences there is a great value added of combining the two
approaches. While classification and differentiation are only partly compatible,
Cedefop results can be used as an additional source for validating national
forecasts at macro level, especially for occupations with ISCO compatibility. For
Cedefop forecasts the AMS-skills barometer and the stakeholders involved can
be viewed as dialogue partners for checking plausibility of results. More
importantly, the integration of Cedefop results would allow for a link between
current illustration of regional as well as national labour market developments
and European trends. The differing time periods are no problem in this respect.
Cedefop offers also short- to medium-term forecasts and the long-term
perspective would be an additional value for the barometer.
While the results are relevant to the broader public in Austria, they are
primarily of interest to an expert group among them educational providers,
counsellors, researchers, the Austrian Public Employment Service and policy-
makers. The barometer could function as a tool for discussing labour-market
trends and developments also at European level. For the broader public the
results need probably to be adequately illustrated. It provides various texts on
labour-market developments and thereby European labour-market information
could be integrated. The consequences of European developments are certainly
of importance to forward looking individuals orientating themselves about future
careers fields.
References
Bacher, D. et al. (2007). Machbarkeitsstudie berufsbezogener Qualifikationsprognosen
auf Basis neuer Datenquellen [Feasibility study on job-related skills forecasts
besed on new data sources]. Pilotprojekt im Auftrag des AMS Österreich.
Vienna: 3s Unternehmensberatung.
Cedefop (2008). Future skill needs in Europe: medium-term forecast: synthesis
Humpl, S.; Kargl, M. (2008). AMS-skills barometer: Austria’s target group oriented
labour market information system. In: Knobel, C. et al. (eds) Regional
forecasting on labour markets. München: Rainer Hampp, p. 148-160.
Ziegler, P. (2010). UserInnbefragung AMS-Qualifikations-Barometer. Ergebnisse
der Online Befragung Mai-Juni 2010 [Survey AMS-qualification barometer:
results of the online survey May-June 2010]. Vienna: 3s
Unternehmensberatung.
95
Web links
AMS — Public employment service Austria
http://www.ams.at/english/14595.html [28.2.2011].
AMS — Qualifikationsbarometer
http://bis.ams.or.at/qualibarometer/berufsbereiche.php [28.2.2011].
AMS — Berufsinformationssystem
http://www.ams.at/bis/ [28.2.2011].
European Commission — From the Lisbon Strategy to ‘Europe 2020’

http://ec.europa.eu/education/focus/focus479_en.htm [28.2.2011].
European Commission — Agenda for new skills and jobs

http://ec.europa.eu/social/main.jsp?langId=en&catId=958 [28.2.2011].
96
CHAPTER 7.
Forecasting skills demand and labour
market dynamics in the Baltic States
What are the specific challenges to forecasting labour-market demand in

the Baltic countries (Estonia, Latvia and Lithuania)? The paper argues that
small, open and flexible Baltic economies are highly sensitive to exogenous
shocks. This creates considerable uncertainty regarding future level and
structure of the demand for skills. As a result there is a gap between policy-
makers’ expectations for detailed and largely accurate estimates and the
capacities of the forecasting exercises to meet these demands. The paper
provides an overview of the national forecasting systems, discusses why
forecasting labour-market demand is particularly challenging the Baltic
countries and assesses how accurate in the short term (2008-09) were the
medium-term forecasts produced by Cedefop for Estonia, Latvia and
Lithuania.
7.1. Introduction
Forecasting exercises tend to inspire high hopes among the European Union
(EU) and national policy-makers. As the Commission argues, forecasting and
anticipation of future skills and labour market needs ‘[…] is a precondition for the
design of efficient employment, education and training policies and individual
career choices’ (European Commission, 2008, p. 5). From the policy-making
perspective, forecasts should be detailed, comprehensive and, above all, they
should be accurate. Yet, as the forecasters do not practise ‘the art of future
telling’, the forecasting exercises have focused on the identification of trends and
possible scenarios, rather than provision of detailed numbers that are yet to be
confirmed by statistical offices in a decade or so. Attempts to close the gap
between the expectations and practice of labour market forecasting have inspired
long-standing discussions on the assumptions, methods and quality of data (e.g.
Neugart and Schömann, 2002).
This paper seeks to contribute to the discussions by focusing on the specific
challenges to labour-market forecasting in the Baltic States (Estonia, Latvia and
Lithuania). The forecasting efforts so far have been concentrated in large OECD
countries, which exhibit moderate economic cycles, relatively stable structure of
the economy and institutionalised labour markets. The small, open and highly
flexible Baltic economies, however, demonstrate steep economic upswings and
97
downturns, unclear competitive advantages and fuzzy labour-market institutions.

How do these contextual factors affect the forecasting of the demand for skills in
the Baltic countries?
The paper argues that forecasting of labour-market demand in the Baltic
countries is indeed highly challenging. The ‘usual’ forecasting problems related to
the assumptions, methodology, and data are topped by dramatic swings in the
level of output and employment. Between 2004 and 2007 the Baltic ‘tigers’
exhibited the fastest GDP growth in the EU. This period was followed by a deep
recession and raise in unemployment in 2008-09. As a result, the forecasts
produced in the Baltic States prior to the global financial crisis have been
subjected to widespread scepticism.
The structure of the paper is as follows: the first part explores the national
forecasting systems. The second part discusses why labour-market forecasting
has proved so difficult in the Baltic States. The third part assesses, how accurate
in the short term are the medium-term forecasts produced by Cedefop (30). The
forecasts were prepared in 2009 and published in 2010, but due to data
restrictions, the latest data used in the forecasts refer to 2007. This provides a
unique opportunity to compare the forecasts for 2008 and 2009 with the actual
data. This allows assessment of the accuracy of the Cedefop forecasts under the
most unfavourable conditions, since 2008 and 2009 were marked by an
unexpected and dramatic economic downturn.
7.2. State of play: experience with forecasting labour

market demand in the Baltic States
Efforts aimed at forecasting labour-market demand in the Baltic States share
several important commonalities. First, it is a relatively recent activity, which
emerged around 2004. Since then anticipation exercises heavily relied on funding
from the European Social Fund and methodological advice from experts of EU-
15. Second, the overall ambition of forecasting was to provide data for the
debates on funding of education and training. This ambition largely remains
unattained in all countries in question due to the low level of detail of forecasts.
‘Small country’ effect has created considerable problems in this respect: most
economic sectors employ less than 100 000 employees and, therefore, standard
surveys fail to capture the composition of the workforce of specific sectors by
(30) The aggregate forecasts for EU-27, Norway and Switzerland are discussed in
Cedefop, 2010.
98
detailed occupation group and level of education. The third commonality: due to
the dynamic nature of the Baltic economies (fast economic growth in 2004-07
and dramatic economic recession in 2008-10) medium- to long-term forecasts
tend to be very inaccurate (even in the short term). This, for example in Lithuania,
caused widespread scepticism regarding the overall utility of forecasting.
7.2.1. Estonia
Anticipation of skills demand in Estonia rests on two types of studies.
First, sectoral studies focusing on the demand for skills in specific industries
(wood processing, metalworking, engineering, machinery and equipment; ICT
and food processing) (Cedefop, 2008). Since 1999 these studies have relied on
the Irish methodology, which seeks to use quantitative and qualitative insights
into strategic trends of specific sectors.
Second, the Ministry of Economic Affairs and Communications prepares
aggregate medium- to long-term employment forecasts. Work in this area started
in 2003 and since then has developed into regular six-year forecasts of labour-
market needs in 34 sectors and five occupational groups. The exercise uses a
macro model Hermin, which is quite similar to the model used for this purpose in
Ireland. The results of the forecast are corrected with sectoral models and expert
assessments (Järve, 2010). The forecasts are updated annually and are primarily
used as a source of information for the discussions regarding the number of
State-funded students.
The main issue encountered during the forecasting exercise is lack of
detailed data on the structure of employment in specific sectors by detailed
occupational group and education level. Small size of the economy and small
samples of the surveys preclude generation of more detailed data. As a result,
the forecasts provide an important framework for deliberations, but are not
detailed enough to support specific decisions (Järve, 2010).
7.2.2. Lithuania
Since 1995 the Lithuanian Labour Exchange provides short-term estimates of the
labour-market demand. They indicate employment prospects (high, medium or
low) for detailed occupational groups and economic sectors in 10 geographical
regions. The estimates are based on the number of registered job openings and
a survey of 4 000 enterprises.
Since 2004 many medium-term forecasts have been developed. Several
studies (31) sought to provide estimates for the whole economy. However, most
(31) BGI Consulting and Sweco EuroFutures, 2008; ETKC, 2007; NPI, 2008.
99
studies were sectoral, i.e. they focused on the future demand in specific
economic sectors (32) or sought to assess the demand for VET or university
graduates (33). The most frequently used forecasting methods include surveys of
employers and qualitative assessments of future demand, while several studies
also engaged in econometric modelling and time series analysis.
A recently published meta-study (BGI Consulting, 2010) on Lithuanian
experience in labour-market forecasting concluded that despite considerable
efforts, the utility of forecasting exercises has been limited. Fragmentation of
efforts prevented systemic accumulation of a knowledge base in forecasting and
understanding of the dynamics of the Lithuanian labour market. Furthermore, the
results of the forecasts have been called into question due to the effects of the
severe economic crisis of 2008-09.
7.2.3. Latvia
In Latvia three types of forecasts have been developed. First, the State
Employment Agency provides short-term estimates by occupational group and
geographical regions. Similar to the Lithuanian system, the short-term forecasts
are based on the surveys of employers (Rastrigina, 2009).
Second, considerable efforts were made to develop economy-wide long-
term forecasts. Between 2005 and 2007 the University of Latvia produced
forecasts of labour-market demand and supply until 2020 (University of Latvia,
2007). The estimates were broken down by 15 economic sectors, 120
professions, 37 aggregated groups of professions, level of education and age.
The exercise was based on the models developed in Sweden. In 2007 the
forecasts concluded that by the year 2013 there will be shortages of labour in 99
professions out of 120 (Dubra and Gulbe, 2008). This widely discussed finding,
however, was dramatically revised in 2009.
Third, the Ministry of Economic Affairs produces annual medium-term labour
force supply and demand forecasts. The estimates are based on the dynamic
optimisation model, which uses three medium-term economic development
scenarios and Eurostat demographic projections. The projections are presented
for 35 aggregated occupational groups and 15 economic sectors (Ministry of
Economic Affairs, 2010).
Last, in addition to the economy-wide forecasts, efforts have also been
made to provide forecasts for labour-force demand in specific economic sectors.
(32) PMMC, 2008a; 2008b; 2008c; 2008d; 2008e; 2008f; 2008g; Lithuanian University of
Agriculture, 2007; ETC, 2007.
(33) VPVI, 2006; 2008; DSTI, 2008.
100
The Higher Education Council also produces forecasts on the future labour-
market demand for university graduates.
Within the policy-making circles the sentiment regarding the existing
forecasting infrastructure remains cautious. On the one hand, some doubts have
been raised regarding the adequacy of forecasting methodology (Rastrigina,
2009; Ecorys, 2010). On the other hand, the dynamic nature of the Latvian
economy imposes considerable uncertainty regarding the reliability of any
medium- to long-term growth scenario.
7.3. Why forecasting is so difficult in the Baltic States

So far much of the academic and policy debate has focused on the overall
strengths and weaknesses of different forecasting approaches, application of
econometric models and technical issues related to availability of detailed historic
data. This section, however, seeks to ask a different question: to what extent
does the reliability of labour-market forecasts depend on the characteristics of the
economy?
To the extent that forecasts of future demand for labour force depend on the
past trends, the ‘ideal’ country for forecasting should exhibit the following
characteristics:
(a) stability — moderate economic cycles and absence of strong external
shocks reduce uncertainty regarding the overall level of future demand for
labour;
(b) clear comparative advantages — due to natural endowments and
institutional structures, countries historically tend to deepen their
comparative advantages, which is important in forecasting the changes in
demand for skills by sector, educational level and profession;
(c) stable labour-market regulation, which affects how the labour market reacts
to external shocks. For example, the costs of firing affect how the firms will
react to decline in aggregate demand.
The Baltic States do not exhibit the above characteristics. First, the
economic cycles in the Baltic countries are far from moderate as they have
exhibited the periods of extremely fast economic growth and steep declines
(Figure 7:1). This is partly explained by high openness of the economies, which
exposes them to the strong impact of external economic shocks.
101
Figure 7:1 Real GDP growth in the Baltics
EE LT LV
15
10
5
Real GDP growth %
0
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
-5
-10
-15
-20
Source: Eurostat.
Second, identification of key sectors, which are likely to drive the future
economic growth of the Baltic economies, remains problematic. In the 1990s
transition from planned to market economies has dramatically reshaped the
structure of the economies. Subsequent economic development was heavily
distorted by large shocks: the Russian crisis of 1998, inflow of cheap credit and
EU structural funds between 2004 and 2007, global financial crisis and the burst
of internal real estate bubbles in 2008-10. This caused large sectoral shifts in
terms of output and employment. As a result it is nearly impossible to pinpoint
any sectors with a historically stable growth curve. Hence, 20 years after
transition considerable uncertainty regarding the competitive advantages of the
Baltic economies remains.
Third, labour-market regulation has remained relatively stable over the past
decade. Nevertheless, attempts to explain how the regulation affects labour-
market outcomes have encountered a paradox. Numerous previous studies (34)
have shown that despite rigid regulation, the Baltic labour markets are very
flexible. This paradox is explained by poor enforcement of the existing regulation.
To sum up, the above discussed characteristics of the Baltic economies
imply considerable difficulties in forecasting labour-market demand in the
medium-long term. As a result one could expect large discrepancies between the
results of the forecasts and actual labour-market demand.
(34) Paas and Emets, 2006; Boguslavas and Blažienė, 2007; Martinaitis, 2008.
102
7.4. Assessment of Cedefop forecasts

This section seeks to assess how accurate are the labour-market demand
forecasts for the Baltic States, which have been developed during the project
‘Forecasting skills supply and demand in Europe’ (hereafter Cedefop
forecasts) (35). The forecasts were published in 2010 and explicitly considered
the effects of the global financial crisis. The forecasts of the labour-market
demand used the data from the labour force survey covering the period between
1998 and 2007. Hence, this section compares the forecasts for 2008 and 2009
with the actual data. The results of comparison are reported as average
percentage difference between forecasted number of employed persons (by
sector, occupational group and education) and actual data of the labour force
survey.
Interpretation of the findings should consider several caveats. First, 2008
and 2009 were marked by an unexpectedly large economic contraction. Hence,
the analysis tests the Cedefop forecasts under the least favourable conditions.
Second, the forecasts produced in the Baltic States prior to mid-2008 did not see
the unfolding economic crisis (and therefore forecasted steep increase in the
demand for labour) or severely underestimated the scale and effects of the
downturn. Therefore, the analysis below does not seek to compare Cedefop
estimates with results of efforts taken in the Baltic States.
The number of employed persons is an aggregate measure of labour force
demand. Comparison of forecasts with historical data (Table 7:1) shows that the
forecasts for 2008 did not deviate from actual number of employed persons by
more than 0.5%. However, the differences for 2009 are considerably larger.
Further, the forecasts underestimated the impact of crisis on employment: actual
drop in employment in 2009 was steeper than forecasted. Poor enforcement of
labour-market regulations and the resulting high flexibility of the Baltic labour
markets could explain this result: a drop in aggregate demand has a larger
negative effect on employment in the Baltic States compared to the EU-15, since
the costs of firing in the former are considerably lower than in the latter.
(35) The aggregate forecasts for EU-27, Norway and Switzerland are discussed in
Cedefop, 2010.
103
Table 7:1 Forecasted and actual number of employed persons

Number of employed persons % difference between forecasts
(thousands) and actual data
2008 2009 2008 2009
Forecast 659.1 619.6
EE 0.39 % 3.84 %
Actual 656.5 595.8
Forecast 1.119.6 1.049.2
LV 0.44 % 6.3 %
Actual 1.124.5 983.1
Forecast 1.521.6 1.459.8
LT 0.11 % 3.01%
Actual 1.520.0 1.415.9
Source: Own calculations based on Eurostat data and Cedefop forecasts.
What are the forecasted changes in terms of the structure of the demand for
skills? The Cedefop forecasts for EU-27 emphasise three general trends:
(a) demand in services is expected to grow, while employment in primary
sectors should substantially decline;
(b) employment in knowledge- and skills-intensive jobs should increase, while
the number of jobs for traditional agricultural, clerical and craft and related
trades workers will decline;
(c) demand for highly-qualified workers should increase, while the demand for
persons with low qualifications should decline.
In sum, during the next decade the structure of the European labour markets
should shift towards high value-added knowledge-intensive services. To what
extent have the developments in the Baltic States in 2008-09 matched these
expectations?
Table 7:2 reports forecasted and actual number of employed persons by
occupational group. In line with above discussed general forecasted trends, the
share of managers, professionals and technicians has increased (although in
absolute numbers it has declined), while the share of crafts and related trades
workers has rapidly decreased in all three Baltic States. However, the share of
skilled agricultural workers has not declined as fast as forecasted. A more
detailed comparison of the forecasts and actual developments suggest that the
largest discrepancies have occurred in three occupational groups. First, the
forecasts suggested dramatic decline in the number of employed technicians and
associated professionals in Lithuania. Employment in this occupational group,
however, has increased. Second, while the number of employed skilled
agricultural and fishery workers has declined in Estonia and Latvia (although not
as fast as forecasted), unlike expectations, it has increased in Lithuania. The
latter observation could be explained by historical trends: dramatic rise of
unemployment tends to be related with increase in employment in the agricultural
104
sector. On the one hand, this is likely to reflect hidden unemployment, when
unemployed persons move to rural areas in search of part-time jobs. Rise of food
prices in 2009 could also provide a partial explanation of the increase in the
number of skilled agricultural and fishery workers in Lithuania. Last, the forecasts
indicated a decline in the number of persons employed in elementary
occupations. The data show that the forecasts were too optimistic and
employment in elementary occupations dropped faster than expected. Flexibility
of the labour markets and high substitutability of the skills of persons in
elementary occupations are likely to be the main factors behind these trends.
Table 7:2 Forecasted and actual number of employed persons by occupational

groups
Estonia Latvia Lithuania
2008 2009 2008 2009 2008 2009
Legislators, Forecast 82 77 98 93 156 152
senior officials Actual 80 72 101.1 92 159 143
and managers % difference 3.38 5.93 -3.16 0.97 -1.86 5.66
Forecast 98 97 155 146 288 284
Professionals Actual 90 96 164 167 274 256
% difference 7.87 1.45 -5.85 -14.68 4.75 9.96
Technicians Forecast 85 84 186 176 139 137
and associate Actual 85 83 186 150 170 174
professionals % difference 0.41 0.82 0.09 14.86 -22.17 -27.29
Forecast 32 30 52 50 65 63
Clerks Actual 35 33 53 51 69 63
% difference -8.58 -9.01 -0.51 -2.81 -5.71 0.09
Service workers Forecast 83 81 137 131 208 205
and shop and Actual 83 75 139 131 183 187
market sales
workers % difference -0.54 6.87 -1.50 0.06 11.67 8.92
Skilled Forecast 10 8 46 38 84 81
agricultural and Actual 12 9 46 43 79 93
fishery workers % difference -21.27 -8.11 0.06 -12.81 6.38 -15.61
Craft and Forecast 105 87 195 180 277 247
related trades Actual 111 84 170 124 288 233
workers % difference -5.57 3.34 12.55 31.26 -4.11 5.69
Plant and Forecast 90 85 111 104 143 136
machine Actual 93 85 112 93 156 141
operators and
assemblers % difference -3.41 -0.49 -1.11 11.31 -8.90 -4.02
Forecast 70 68 139 129 157 150
Elementary
Actual 63 56 150 127 139 122
occupations
% difference 9.53 18.25 -8.08 0.89 11.55 18.90
NB: Data presented refer to thousands of persons, while the % difference was calculated using the most
detailed data available.
105
Table 7:3 reports forecasted and actual trends in employment by selected

sectors. As forecasted, employment in manufacturing, construction and
distribution and transport considerably declined in 2008-09. However, in contrast
to the forecasts, employment in agriculture did not dramatically decline in Estonia
and Latvia and increased in Lithuania. This could be explained by the above
discussed factors: hidden unemployment and higher global demand for
agricultural products. Furthermore, it seems that employment in business and
other services has decreased faster than forecasted (data are not provided due
to high uncertainty regarding its reliability).
Table 7:3 Forecasted and actual number of employed persons in selected

sectors
2008 2009 2008 2009 2008 2009
Agriculture, Forecast 25 21 106 91 120 115
forestry and Actual 25.3 24.0 88.6 85.4 119.8 130.5
fishing % difference -1.30 -14.40 16.54 6.18 0.17 -13.28
Forecast 136 120 170 159 266 262
Manufacturing Actual 135.0 113.9 165.7 135.9 260.4 226.0
% difference 1.04 4.89 2.30 14.35 2.04 13.73
Forecast 72 58 126 117 165 120
Construction Actual 81.0 58.3 128.7 80.1 166.5 122.6
% difference -12.50 0.13 -2.16 31.29 -0.91 -2.23
Forecast 134 136 215 211 331 330
Non-marketed
Actual 129.4 132.2 231.4 208.1 329.9 326.5
services
% difference 3.49 2.79 -7.54 1.39 0.34 0.91
Forecast 178 174 327 313 417 415
Distribution &
Actual 181.325 167.3 336.125 299.6 429.875 400
transport
% difference -1.81 3.63 -2.79 4.37 -3.06 3.57
NB: Data presented refer to thousands of persons, while the % difference was calculated using the most
Table 7:4 reports forecasted and actual changes in the number of employed
persons by qualification. In absolute numbers all groups of workers have
witnessed a decline in employment prospects. However, in line with the general
expected trends, the share of highly-skilled workers has increased at the
expense of persons with low qualifications, i.e. demand for highly-qualified
persons has declined less than the demand for persons with low qualification.
The number of persons with low qualification has decreased faster than
forecasted, while the reverse is true in the case of demand for highly-qualified
persons.
106
Table 7:4 Forecasted and actual number of employed persons by qualification

2008 2009 2008 2009 2008 2009
Forecast 68 63 155 147 119 112
Low
Actual 68 52 140 104 95 79
qualification
% difference -0.09 17.84 10.09 29.32 20.09 29.91
Forecast 361 335 686 638 879 838
Medium
Actual 361 313 681 589 896 818
qualification
% difference 0.02 6.54 0.70 7.65 -2.00 2.44
Forecast 230 222 278 264 523 509
High
Actual 231 228 304 290 526 517
qualification
% difference -0.14 -2.76 -9.09 -9.73 -0.56 -1.59
NB : Data presented refer to thousands of persons, while the % difference was calculated using the most
To sum up: how accurate in the short term are the Cedefop projections for
the Baltic countries? When assessing the results, one should consider that the
forecasting exercise took place during one of the largest global economic
recessions, which created high uncertainty regarding the demand for labour in
small dynamic and flexible economies. However, the overall conclusions are two-
fold. On the one hand, the broad forecasted trends seem to be correct: persons
with high qualification in knowledge- and skills-intensive jobs seem to have
weathered the ‘economic storm’ better than low-qualified workers in elementary
occupations. However, in contrast to the forecasts, employment in agriculture did
not dramatically decline (it has increased in Lithuania). Further, considerable
growth of demand in high-value added service sectors seems to be overstated at
least in the short term.
On the other hand, the forecasts considerably differ from historic data when
one compares the absolute numbers. The more detailed the breakdown of
demand, the larger are the discrepancies. Also, the forecasts for 2008 were more
accurate than the ones for 2009. Hence, if the detailed forecasts are used for
manpower planning decisions, they could lead to wrong choices.
7.5. Implications and further discussion

Several broad lessons stand out from the above discussion. First, consolidation
of efforts within national forecasting systems helps capacity building and the
learning process. In Lithuania many medium-term forecasts were developed, but
fragmentation of efforts prevented systemic accumulation of know-how and
establishment of forecasting infrastructure. Concentration of efforts within a
107
single ministry in Estonia and subsequently in Latvia helped development of a

stable forecasting system.
Second, the characteristics of the Baltic economies imply considerable
difficulties in forecasting labour-market demand in the medium- to long-term.
Small, open and flexible economies are highly sensitive to exogenous shocks,
which creates considerable uncertainty regarding future level and structure of the
demand for skills. Individuals seem to be adapting to the uncertainty by investing
in highly transferable general skills and shunning away from industry-specific
skills, which are developed in vocational education and training
systems (Martinaitis, 2010). The policy-makers, on the other hand, tend to
muddle through the uncertainty with the hope of better forecasting systems. The
assessment of Cedefop forecasts revealed, that the estimates capture the overall
trends in labour-market demand. However, the detailed estimates in absolute
numbers considerably deviate from actual developments.
Last, qualitative analysis of future demand for skills is still underdeveloped
compared to the quantitative estimates in the Baltic countries. Considering the
dynamic nature of the economies and chronic lack of high quality and detailed
data, qualitative insights into the changing nature of skills requirements could
considerably contribute to reduction of uncertainty for policy-makers and
learners.
References
BGI Consulting and Sweco EuroFutures (2008). Kvalifikacijų poreikio Lietuvos
darbo rinkoje vidutinės trukmės prognozavimo metodika ir prognozės pagal
ūkio sektorius [The methodology for medium-term forecasts of demand for
qualifications in Lithuania and forecasts by economic sectors]. Vilnius.
BGI Consulting (2010). Specialistų ir kompetencijų poreikio žemėlapio studija
[Mapping demand for specialists and competences]. Vilnius.
Cedefop (2008). Systems for anticipation of skill needs in the EU Member States.
Thessaloniki: Cedefop (Cedefop working paper No 1).
to 2020. Luxembourg: Publications Office for the European Union.
European Commission (2008). New skills for new jobs: anticipating and matching
labour market and skills needs. (Communication from the Commission to the
European Parliament, the Council, the European Economic and Social
Committee and the Committee of the Regions). Luxembourg: Publications
Office of the European Community (COM(2008) 868 final).
DSTI (2008). Aukščiausios kvalifikacijos specialistų (magistrantų) pasiūlos ir
paklausos atitikimo sisteminis įvertinimas [Assessment of the match between
108
demand and supply of highest qualification specialists (master-level

graduates)]. Vilnius: DSTI, Institute of labour and social research.
Dubra E.; Gulbe, M. (2008). Forecasting the labour force demand and supply in
Latvia. Technological and economic development of economy, Vol. 14, No 3,
p. 279-299.
Ecorys (2010). Sector councils on employment and skills at EU level. Country
reports. Reports prepared for the European Commission, DG Employment,
Social Affairs and Equal Opportunities. Rotterdam: Ecorys.
ETK (2007). Energijos išteklių naudojimo, energiją vartojančių įrenginių ir sistemų
projektavimo, naudojimo ir priežiūros specialistų rengimo regioninio ir
struktūrinio poreikio studijos parengimas [Assessment of structural demand
for specialists in the area of energy resources and energy infrastructure].
Vilnius: ETK, Ekonominės konsultacijos ir tyrimai [Economic consulting and
research].
ETKC (2007). Darbo rinkos analizė ir pasiūlymai darbo jėgos trūkumo problemai
spręsti, atsižvelgiant į šalies ūkio plėtros prognozę 2008-15 metams [Labour
market analysis and recommendations concerning the shortages of labour
force taking into account the forecasts of economic development for 2008-
15]. Vilnius: ETKC, Centre for business and legal consultations.
Boguslavas, G.; Blažienė, I. (2007). Lithuania. In: Cazes, S.; Nesporova, A. (eds)
Flexicurity: a relevant approach in central and eastern Europe. Geneva:
International Labour Office, p. 149-182.
Järve, J. (2010). Estonian experience in anticipating skills needs. New skills
network.
Lithuanian University of Agriculture (2007) Forecasts for the demand of qualified
workers in agricultural sector for 2013-15 [Specialistų ir kvalifikuotų
darbuotojų poreikio ir rengimo prognozės žemės ūkio srityje 2013-15
metams]. Vilnius.
Martinaitis, Ž. (2008). Flexicurity: will the European medicine cure Lithuanian
labour market? Lithuanian political science yearbook, 2008. Vilnius: Vilnius
University Publishing House.
Martinaitis, Ž. (2010). The political economy of skill formation: explaining
differences in central and eastern Europe. Vilnius University Social
Sciences, Political science 02S (Doctoral dissertation).
Ministry of Economic Affairs (2010). Informational report on medium-term
forecasts of workforce supply and demand [Informatīvais ziņojums par
prognozēm darbaspēka pieprasījuma un piedāvājuma atbilstībai vidējā
termiņā]. Riga.
NPI (2008). Methodology and analysis of labour force demand in Lithuania
[Specialistų poreikio tyrimų metodologija ir prognozės Lietuvoje]. Vilnius:
NPI, National Development Institute.
109
Neugart, M.; Schömann, K. (2002). Forcasting labour markets in OECD

countries. Measuring and tackling mismatches. Cheltenham, Northampton:
Edward Elgar.
PMMC (2008a). Sectoral study for manufacture of petroleum, chemicals, rubber,
plastics and mineral products [Kuro, chemikalų, guminių, plastikinių ir
mineralinių produktų gamybos sektoriaus studija]. Vilnius: PMMC,
Methodological Center for Vocational Training.
PMMC (2008b). Sectoral study for manufacture of food and beverages [Maisto
produktų ir gėrimų gamybos sektoriaus studijama]. Vilnius: PMMC,
Methodological Center for Vocational Training.
PMMC (2008c). Sectoral study for manufacture of wood products [Medienos
sektoriaus studija]. Vilnius: PMMC, Methodological Center for Vocational
Training.
PMMC (2008d). Sectoral study for sports [Sporto sektoriaus studija]. Vilnius:
PMMC, Methodological Center for Vocational Training.
PMMC (2008e). Construction sectoral study [Statybos sektoriaus studija]. Vilnius:
PMMC (2008f). Tourism sectoral study [Turizmo sektoriaus studija]. Vilnius:
PMMC (2008). Agriculture sectoral study [Žemės ūkio sektoriaus studija]. Vilnius:
Rastrigina, O. (2009). Improving the match between skills supply and labour
market demand: the Welsh approach. Statements and comments. Riga:
Stockholm School of Economics in Riga.
Paas, T.; Emets, R. (eds) (2006). Labour market flexibility, flexicurity and
employment: lessons of the Baltic States. New York: Nova Science.
University of Latvia (2007). Research of long-term forecasting system of the
labour market demand and analysis of improvement options. Riga.
VPVI (2006). The supply and demand of master level graduates [Magistrantūros
ir Lietuvos ūkio poreikių atitikimas]. Vilnius: Public Policy and Management
Institute.
VPVI (2008). The demand for vocational education in Lithuania [Profesinio
mokymo paslaugų paklausa Lietuvoje]. Vilnius: Public Policy and
Management Institute.
110
CHAPTER 8.
Demand and supply of labour by education
in Norway towards 2030 ― Linking
demographic and macroeconomic models
Roger Bjørnstad, Marit L. Gjelsvik, Anna Godøy, Inger Holm
and Nils Martin Stølen
Updated projections on labour demand and supply in Norway towards 2030

are presented. The macroeconomic model Modag gives projections for
labour demand by education within five aggregated educational
levels/groups. To compare the projections for demand with projections for
supply we have used the dynamic microsimulation model Mosart based on
constant educational propensities and labour force participation rates. The
projections show that the previous trends of increasing demand for workers
with a tertiary education and higher secondary vocational education will
continue towards 2030. Demand for labour with a lower tertiary education
may increase more than the corresponding increase in supply. The need
for labour with primary, lower secondary and higher secondary general
education is decreasing. A module translating labour demand by industry
into labour demand by education at detailed level is linked to Modag. This
shows a particularly strong growth in demand for candidates in economics
and administration and nursing and caregiving at a lower tertiary level.
8.1. Main structure of the model system

Because of globalisation and technological progress, Norway has seen a
considerable growth in demand for labour with higher skills and educational
levels the past decades. Labour supply has followed demand rather closely, and
unemployment and wage differences have stayed relatively equal. In many other
OECD countries, supply has not grown correspondingly. This has resulted in
increasing differences either in unemployment or in wages between high and
low-skilled workers. The past trends in educational upgrading will probably
continue, and further stability in the labour market requires that labour demand
and supply matches also in the future. Both for future students, who must decide
on education, and the authorities, who must plan the educational capacity,
industrial development and welfare reforms, etc., projections on demand and
supply of labour by education are useful.
Statistics Norway has produced such projections since 1993, and the paper
presents updated projections towards 2030. The main structure of the model
111
system used for projecting supply and demand for different kinds of labour by
Statistics Norway is presented in Figure 8:1. According to what Cedefop (2008)
calls ‘best practice’ in projecting demand for labour, the Norwegian projections
are based on a macroeconomic model. The Modag model has a core of input-
output relations to capture the interaction between the different industries. In the
version of the model used for this analysis labour is divided by five educational
groups. Production in the different industries is dependent on the macroeconomic
functioning of the Norwegian economy and on exogenous assumptions about
economic development abroad, further growth in industries based on natural
resources and the further growth in public services. In addition to the level of
production, demand for labour in the different industries depends on
technological progress and the possibility of substitution with other inputs.
In a separate submodel, the aggregate demand for labour in each industry is
further spread by educational fields using exogenous assumptions about how the
employment shares within each main educational group may develop. The
continuation of trends observed from the past decade regarding the composition
of labour is used as a standard assumption. On the demand side our projections
correspond rather closely to the European joint project managed by Cedefop that
aims to project demand and supply for different kinds of labour for EU-27,
Norway and Switzerland (Wilson et al., 2008).
On the supply side of Modag labour force by the five educational categories
is determined by linking equations for labour-market participation by age and
gender to the corresponding groups of individuals in working age. A discrepancy
between total supply and demand, expressed by the corresponding rates of
unemployment, affects wage formation. Because wages affect labour-market
participation rates and demand for labour, this link helps in the direction of
achieving consistency between the labour demand and supply by education.
However, in line with observations of real life, the equilibrium correction
mechanisms are not strong enough to secure a steady path for the economy and
stable unemployment. Therefore, we have chosen to let the labour force for the
five educational groups in the projections be determined from the demand side
instead. Hence, future mismatches in the labour marked cannot be analysed
using the Modag-based projections alone.
However, labour supply by education is also projected with the
demographic-based dynamic microsimulation model Mosart. From a
representative sample of the population in a base year, Mosart simulates the
further life course for each person in this initial population by using estimated
transition probabilities. The transition probabilities are kept constant, and the
projections then show what will happen if everything continues as recently
112
observed. Some degrees of consistency between the two labour supply

projections are secured because Statistics Norway’s official demographic
projections by age and gender form a common basis for them. Nonetheless,
while labour supply is in line with labour demand in Modag, educational
propensities and detailed participation rates that differ by age, gender and
education are assumed to be fixed in the Mosart projections. Thus, there may be
discrepancies between the projections from the models regarding labour supply,
and these discrepancies may be viewed as indications of potential mismatches in
the future labour market.
Figure 8:1 Main structure of the model system
Demographic projections by age and gender
Modag Modag Mosart Mosart

Employment by Labour force by four Labour force by Population by four
industry and four educational levels (five four educational educational levels,
educational levels groups), consistent levels, constant constant educational
(five groups) with steady participation rates propensities
unemployment
unemployment
Demand for labour by

industry and 30 Mismatch by four educational levels
educational fields/levels,
continuation of shares of
employment by industry
and education
Source: Statistics Norway.
The five main groups of education are presented in Table 8:1. The levels are
in accordance with the typical design of the Norwegian educational system and
correspond closely with international standards for education (ISCED97). At
detailed level, each of the three upper levels of education is divided into about 10
fields reaching a total of 30 educational groups. The main criterion for the division
was to subdivide educational groups to present the greatest possible
homogeneity within subgroups regarding supply and demand for labour.
113
8.2. Modag — Model of the Norwegian economy

Modag is a macroeconometric model for the Norwegian economy developed at
Statistics Norway. In this section, we provide a short description of the model.
Cappelen (1992) provides a more detailed presentation of an earlier version of
Modag, and Boug and Dyvi (2008) present the newest version written in
Norwegian. The standard version of Modag has homogeneous labour. In the
version we use here, labour is disaggregated into five educational groups.
Bjørnstad and Skjerpen (2006) describe the education-specific labour demand
and wage setting.
Table 8:1 Classification by main group of education and corresponding

numbers of employed in 2006 (1 000 persons)
Code Field of education Employed
0 Total, including unknown 2362.1
1 Primary and lower secondary education (ISCED 0-2) 499.3
2a Upper secondary education general programmes (ISCED 3 and 4) 462.7
2b Upper secondary education vocational programmes (ISCED 3 and 4) 607.2
3 Tertiary education, lower degree (ISCED 5, lower degree) 591.7
4 Tertiary education, higher degree (ISCED 5, higher degree and ISCED 6) 177.9
9 Unknown 23.3
Modag is relatively disaggregated and distinguishes between 45 different

products and 21 different sectors. The model specifies many final applications of
the products, and these products have different prices depending on supply
(home- or foreign-produced) and utilisation (export or home market). The
Norwegian national accounts form the conceptual framework and the empirical
basis of the model. Specifically, Modag balances all products in terms of input
and output. These relationships connect — at the most detailed level in the
model — the supply and usage of the products to different activities in the
economy. Modag also contains a consistent account of income and capital
balances within each institutional sector, and in the economy as a whole. At the
same time, the input-output structure and the account-based relationships are
supplemented with econometric equations describing how the agents in the
economy tend to respond to different options. While the long-term relations are
based on economic theory, the dynamic adjustments towards the long run are
largely quantified to fit data.
The main structure of the model implies that prices — along with interest
rates, exchange rates and wages — determine the demand from private
consumption and investment and foreign demand. Public sector’s income and
expenditures are richly represented in the model. However, the public use of
114
resources and various tax rates are exogenous. In 11 of the industries, the
demand for variable input factors is based on a Cobb-Douglas production
function in materials, a CES-aggregate in energy and a CES-aggregate in high-
and low-educated labour. The model assumes that there is no substitution
between the different types of labour within the two aggregates (36). Total real
capital is treated as a quasi-fixed input, and Hicks-neutral technological change is
represented by a linear trend. For seven of the industries, individuals with
vocational training at higher secondary level are regarded as low-educated
labour, while the other four — which in sum is much larger —are regarded as
high-educated labour. This choice is based on the empirical results obtained by
Bjørnstad and Skjerpen (2006). The assessment of high- versus low-educated
workers depends on relative wages, the stock of machine capital and production
volume following a linear trend, which is assumed to represent skill-biased
technological change (but possibly also other factors).
The model involves a traditional Keynesian multiplier, where higher income
leads to more spending, which increases production in the next round and
activates further growth in employment and income. In this way, the initial change
in aggregate demand causes a change in aggregate output for the economy that
is a multiple of the initial change. The operational version of the model does not
contain mechanisms that ensure full resource utilisation or balance in the
external account or in public sector budgets. We believe this property is in line
with the functioning of the Norwegian economy, where centralised wage
negotiations and an active fiscal and monetary policy play a central role. Thus, in
applying the model, the user must decide on a fiscal policy that is sustainable.
8.3. Projections of the Norwegian economy towards

2030
To study the effects of economic development on labour demand, we have
projected a path for the Norwegian economy that extends to 2030 using Modag
simulations. Our projections are based on final national account figures up until
2006 and preliminary figures for 2007. The Modag simulations start in 2008, but
we have generated residuals so that the simulated figures should match the
preliminary account figures for 2008 and 2009 fairly well. The path is based on a
rather quick adjustment from the low growth we have observed through 2009
(36) In the other industries in which the demands for variable inputs are modelled, there
is no possibility for substitution between any different types of labour.
115
following the international financial crisis, to a more balanced economic

development where employment, wages, consumption, prices and activity grow
at rates close to the prevailing trend. This ensures that the demand for labour by
education in our analysis results from structural rather than cyclical conditions.
During the upswing period 2004-08 there has been an extensive labour
supply growth in Norway. There are multiple explanations, but most importantly
the expansion of the EU on 1 May 2004 has increased the flow of foreign workers
and immigrants seeking work in Norway, particularly from central east European
and Baltic countries. In addition participation rates have increased. A substantial
share of the population in Norway of working age is participating in working life. It
has among the highest participation rates in the world. However, the average age
within the cohort is continually increasing, and the demographic situation will be
less favourable as we move towards 2030. But immigration is assumed to stay at
a high level. Even though labour supply is determined from the demand side in
our projections, total labour supply is reasonably consistent with these trends. In
our projections, the participation rate falls gradually from a level of about 73% in
2009 and 2010 to about 69% towards 2020. This development is fairly in line with
what we would expect in light of the cyclical movements and the demographic
changes during this period. Then, the fall in the participation rate stops and it
even turns to increase after 2025. At the end of the period, the participation rate
reaches 70%. The increase mainly comes from an increase in labour demand in
the public health and social sector as the elderly population grows (keeping in
mind that labour supply is determined by employment plus unemployment).
Figure 8:2 shows employment by sectors of the economy as a share of total
employment. The sectors include the primary industries (37), manufacturing (38),
private services (39), construction industry, public sector (40) and the remaining
industries (41). The extent of public activity is expected to grow both in quantity
and as a share of total employment. The public sector’s share of total
(37) Primary industries consist of agriculture, forestry, hunting and fishing and fish
farming.
(38) Manufacturing includes production of consumer goods, products of input and
investment, industrial raw materials, workshop products, ship and oil platforms and
production and refining of petroleum products, etc.
(39) Private services include services in banking and insurance, retail, domestic transport
and communication including air transport, housing services and other private
service activities.
(40) Central and local government including the military.
(41) Among the remaining industries are the extraction of crude oil and natural gas,
foreign shipping and production of electric energy.
116
employment rises from 30% in 2009 to nearly 37% in 2030. This growth also
illustrates the demand for public health and caregiving services as the population
of elderly people grows.
Figure 8:2 Employment by sector, share of total employment
50%
Private services
Public sector
Manufacturing
40%
Primary sector
Construction
30% Remaining sectors
20%
10%
0%
1970
1974
1978
1982
1986
1990
1994
1998
2002
2006
2010
2014
2018
2022
2026
We have assumed that the oil price stays around USD 80 per barrel in the
period towards 2015, and then increases gradually to USD 100 per barrel in
2030. The petroleum production in Norway has already reached its peak, and is
expected to decline further the years ahead. Although the revenues from the
petroleum activity will remain high for many years still, the possibility of financing
the rapid increase in public pension expenditures after 2025 abates. The
Norwegian government has adopted a fiscal rule that the deficit in the
government sector, corrected for oil and gas revenues and cyclical conditions,
should be approximately 4% of the value of the government pension fund –
global in the beginning of the budget year. This corresponds to the expected real
surplus of the fund. The rest of the petroleum revenues are invested in the fund.
In the projections, the government is assumed to stick to this rule. Even though
the growth in public employment continues during the entire period, the growth in
standards in public services is relatively moderate, also before 2020. Instead we
have assumed a reduction in income tax rates up until 2025. After 2025 we have
assumed gradual increases in the income tax rates again
117
The growth in households’ disposable real income remains relatively high

through the entire projection period. The growth in demand for private services
also stays at a relatively high level. However, because of the growth in the public
sector, the private service sector measured as a share of total employment will
decline. The primary industries and manufacturing are projected to continue their
downward trend. The remaining industries also reduce their activity, especially in
the extraction of crude oil and natural gas. The construction industry, on the other
hand, is expected to expand somewhat.
8.4. Demand for labour by education

Figure 8:3 shows historical and projected employment, aggregated across all
industries in the economy for each of the five educational groups in percent of
total employment. Statistics Norway has supplied us with data consistent with the
present national account figures for the period 2000-07. In the figure we have
linked these numbers with the data used in Bjørnstad and Skjerpen (2006).
Bjørnstad and Skjerpen used an older version of Modag, but had the same
disaggregated labour market as in the version used here. In their analyses
Bjørnstad and Skjerpen used provisional figures provided by the unit for national
accounts at Statistics Norway, and they relied on their own imputations in the
cases where data were not available to obtain figures for the period 1972-97.
They then used Modag together with ordinary national account figures to
generate data on employment and wages by education for 1998-2000. Data in
the larger groups are linked by adding the absolute differences between these
two data sets in 2000 also in the period 1972-99. Data in the smaller groups are
linked by adding the relative differences.
The number of employed persons with unknown education grew from about
40 000 in 2000 to 180 000 in 2007. The growth was particularly high in 2006 and
2007 and reflects the considerable labour immigration into Norway from central
east European and Baltic countries. These immigrants have largely not been
registered with an education at the register of the population’s highest level of
education. In the estimated labour demand equations in Modag, those with
unknown education were included with primary and lower secondary education.
Therefore, we have chosen to do the same in our projections. In the figure, we
show employment numbers for this group both with and without workers with
unknown education, but only during the period 2000-07.
According to our projections past trends of increasing employment for skilled
labour will continue. Analogously, demand for less-educated workers will decline.
Hence, we assume that the relatively flat development in the employment of
118
primary- and secondary-educated workers will shift in the years ahead. We

believe that the stable shares are due to specific problems regarding the
registration of immigrants’ education. Many new citizens have likely been
registered with only primary education when they, in fact, might have had
vocational training.
The educational upgrading continues in all sectors of the economy. While
about 27% of the employed had a primary or lower secondary education in 2007,
this share declines to 18% in 2030. Employees with an upper secondary
vocational education increase in percentage points from 25 to 30 in the same
period. This rise is particularly connected to the strong employment growth in the
private services sector. As explained above, employees with an upper secondary
vocational education are mostly regarded as skilled labour, and employees with
an upper secondary general education are unskilled labour. Therefore, those with
general programmes are expected to decline in numbers in the future, even
though they have been increasing in number until 2000. The projections show
that this group will decline from 19% of the total employment in 2007 to 13% in
2030. The decline is most evident in the private services sector. Hence,
employment in this sector is reorienting towards those with a vocational
education, in addition to those with generally higher education.
Modag does not contain any detailed information on employees’ fields of
education. Thus, we have created a submodel to disaggregate further. In this
submodel, we match the NAV National Register of Employers and Employees
with the Norwegian register of the population’s highest level of education. The
match gives us data on industry-specific employment by educational field back to
1986. By calculating education-specific employment shares and assuming a
continuing trend in these shares in the private sectors, we are able to project
industry-specific employment by field of education by multiplying the shares with
the national account employment figures in each industry. In the public sector, we
distributed employment into educational fields according to information from three
other projection models at Statistics Norway which are designed for this. They
are called Makko, Lærermod and Helsemod. Figures 8:4, 8:5 and 8:6 show the
projected number of employees at national level by educational field at higher
secondary level, lower tertiary level and higher tertiary level, respectively.
Figure 8:4 indicates growth in the employment of persons with vocational
education at higher secondary level. The demand for persons with specialisation
in electronics, mechanics work and machinery and in other fields is predicted to
show the strongest growth. For these fields, employment is projected to increase
by nearly 65 000 persons from 2006 to 2030. In 2030, workers with an education
in electronics, mechanical work and machinery may constitute the largest group
119
of vocational education at higher secondary level, with a share of nearly 30%.

The demand for candidates with specialisation in electronics, mechanical work
and machinery primarily originates from the service sector and the building and
construction industry, while the demand for candidates specialising in other fields
originates from public sector and private service sector. Further, the employment
of persons with specialisation in both nursing and caregiving and building and
construction clearly increases throughout the prediction period.
Figure 8:3 Employment by level of education as a percentage of total

employment, 1986-2030
Primary and lower secondary education without unknown

Primary and lower secondary education and unknown
Upper secondary education, general programmes and economics and administration
Upper secondary education, vocational programmes
Tertiary education, lower degree
Tertiary education, higher degree
70%
60%
50%
40%
30%
20%
10%
0%
1972
1976
1980
1984
1988
1992
1996
2000
2004
2008
2012
2016
2020
2024
2028
The employment of individuals with general education programmes and

economic and administrative education at higher secondary level is, on the other
hand, predicted to fall of the order of 45 000 and 30 000, respectively, during the
projection period. Accordingly, the employment shares of individuals with general
education programmes of a higher secondary level are projected to fall from 12%
of total employment in 2006 to about 8% in 2030. For candidates in economics
and administration at higher secondary level, the employment share falls from
120
nearly 8% in 2006 to a little more than 5% in 2030. This development is

connected with vocational training being regarded as specialised labour in some
of the large private services industries, while general education programmes and
economic and administrative education at this level are mainly regarded as
unskilled in the labour market.
Figure 8:4 Employment by education at upper secondary school level, 1 000

persons
General programmes General programmes, folk high school

Economics and administration Other fields of science, technique and crafts
Electronics, mechanics work and machinery Nursing and caregiving
Other fields Building and construction
300
250
200
150
100
50
0
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
2018
2020
2022
2024
2026
2028
2030
Figure 8:5 shows the development in demand for labour among the fields at
lower tertiary level.
The figure indicates employment growth within all fields, particularly for
candidates specialising in economics and administration. By 2030, the
employment of individuals with this qualification is expected to expand by almost
100 000 persons, making this occupational group the largest at the lower tertiary
level. More than one in five individuals holding a lower degree at tertiary level is
specialised in economics and administration in this year. The demand for this
type of skilled labour may especially grow in the service industry, but the banking
121
and insurance industry, public sector and the manufacturing industry all demand
a large share of the candidates in economics and administration.
Figure 8:5 Employment by education at tertiary level with a lower degree, 1 000
persons
Education Other fields of science
Economics and administration Social sciences
Nursing and caregiving Other tertiary education
Engineering Preliminary examination
Other fields of health and social services Law
Humanities and arts
250
200
150
100
50
0
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
2018
2020
2022
2024
2026
2028
2030
There is also a strong increase in demand for teachers and health personnel
from the public sector. Moreover, the service sector employs many teachers, and
this development is assumed to continue. The employment of persons
specialising in education is projected to increase by about 55 000, while the
employment of persons specialising in other fields of health and social services
and in nursing and caregiving is projected to increase by about 85 000 in total. In
addition, the demand for skilled labour specialising in science (excluding
engineering) is projected to expand by nearly 30 000 people, increasing its share
of total employment of candidates with a lower tertiary education from 5% in 2006
to 6% in 2030. This demand increase mostly stems from the service industry.
Employment of candidates with specialisation in humanities and the arts is also
122
predicted to increase by about 25 000 from 2006 to 2030. The demand increase
for this occupational group also primarily stems from the service sector.
Moving on to the employment of persons at the highest educational level,
Figure 8:6 shows the development in each educational field. Demand increases
markedly within most fields at this level. The growth is most pronounced among
individuals specialising in other fields of science, with an employment increase of
nearly 20 000 persons. There is also a strong demand for graduate engineers.
Counting these two fields together, it may be necessary that 35% of the
employees at tertiary level in 2030 hold a degree in science and engineering; in
addition, the service sector clearly has the most demand for these skills. In the oil
and gas industry, which employs a considerable share of the graduate engineers,
the demand for engineers is assumed to fall. This decrease is connected to the
production decrease we anticipate in this industry. The growth in the employment
of candidates with specialisation in economics and administration has been
pronounced in the observation period 1986-2006. We also expect a marked
increase for this type of skilled labour in the projection period, mostly stemming
from the service sector. From a share of 6% of employees with a higher degree
at tertiary level, employment of persons specialising in economics and
administration increases to about 10% in 2030. The employment growth of
teachers, on the other hand, is predicted to be relatively modest in the following
years, while the employment of dentists is expected to stay constant.
8.5. Comparing labour supply with labour demand

To compare labour supply with labour demand, supply is projected by using the
dynamic microsimulation model Mosart based on constant educational
propensities and labour-market participation rates. Mosart is a dynamic
microsimulation model used for projections of demographic development, supply
of labour and pension expenditures in Norway, and it is described by Fredriksen
(1998). In the past decade, the model has been extensively used in the analyses
of effects of the reform of the Norwegian pension system.
From a representative sample of the population in a base year, the model
simulates the further life course for each person in this initial population. The life
course is simulated by possible transitions from one state to another given by
transition probabilities depending on each individual’s characteristics. The
transition probabilities are estimated from observed transitions in a recent period
and are calibrated to match aggregated statistics. Normally, the transition
probabilities are kept constant, and the projections then show what will happen if
everything continues as recently observed. Alternative scenarios based on
123
different assumptions may be produced. These alternative scenarios then show

the robustness of the reference scenario (sensitivity analyses) and/or
consequences of alternative policies.
Figure 8:6 Employment by education at tertiary level with higher degree, 1 000
persons
Graduate engineering Medicine

Other fields of science Social sciences
Humanities and arts Dental studies
Other tertiary education Education
Law Economics and administration
60
50
40
30
20
10
0
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
2018
2020
2022
2024
2026
2028
2030
Events included in the simulation are migration, deaths, births, marriages,

divorces, educational activities, labour force participation and retirement.
Education is included in the model for two reasons. First, at Statistics Norway,
there is a long tradition of projecting the number of persons and supply of labour
by education, and Mosart has been used for these projections since the
beginning of the 1990s. Second, education serves as an important explanatory
factor for different kinds of behaviour in the simulation, especially because
education may capture the effects from different events during the life course, as
well as from fixed characteristics.
124
Because few persons start and complete a new education after the age of
30 (or even younger), a different level of education among the younger and older
workers is one of the main driving forces in the model. With a dynamic
microsimulation model, it is also possible to consider that the age for entering the
labour market may range between the age of 18 and 30, and some educational
activities may take place between the age of 30 and 40. Education seems to be
very important for labour-market participation and the age of retirement.
Projections of the population by education as a basis for corresponding
projections of labour supply in this paper are based on a representative sample
for the Norwegian population that is mainly calibrated to the situation in 2008.
Because of the fluctuating propensities to enter education over the business
cycle, an average for the years 2004-08 is used. The demographic assumptions
are based on Statistics Norway’s demographic projections from June 2009. A
total fertility rate of 1.9 and net immigration decreasing from about 40 000
persons per year at present to about 20 000 persons per year after 2040 suggest
that the size of the younger and middle-aged cohorts stabilises towards 2050.
The results of the projections for the number of persons aged 16 to 74 by
level of education are presented in Figure 8:7. A declining trend in the share of
persons with primary and lower secondary education from the past decades
continues, but to a weaker degree. While 54% of the population in the age range
16 to 74 had this level of education in 1993, the share had fallen to 35% in 2008
and is projected to decrease further to 20% in 2030.
However, the share of persons with higher secondary education increased
from 24% in 1993 to 33% in 2008. Due to a rising share of persons completing
tertiary education, further growth slows, and the share is projected to reach 37%
in 2030. Due to the large expansion of capacity in tertiary education during the
1990s, the number of individuals with this level of education has shown
increasing growth since 2000. In 2030, the share of individuals with a lower
degree tertiary education aged 16-74 may reach 24%, while the corresponding
share of persons with higher tertiary education then may reach 7%.
The level of education for immigrants to Norway is only weakly recorded in
Norwegian administrative registers. In a situation where immigration has
increased significantly in the past years and is expected to stay at a higher level
in the years to come relative to the situation before 2005, the projections with
Mosart show a significant increase in the number of individuals with unknown
education levels.
Because the labour supply projections from Modag in Section 8:4 are
determined from the demand side (with stable unemployment), a comparison with
the corresponding labour supply projections from Mosart is an implicit analysis of
125
potential mismatches in the future labour market. To make this comparison we

must aggregate the groups identically. While the large and growing group of
mainly immigrants is classified with unknown education according to Mosart,
those with an unknown education are included among those with primary and
lower secondary education in Modag. The two groups with upper secondary
education in Modag are further aggregated to correspond with the figures from
Mosart. In addition, we adjust the level of labour supply by education in Mosart to
the observed level in the base year of Modag (2008) to ensure an identical
starting point. Deviations between the two labour supply projections indicate that
educational propensities (and capacity if necessary) and/or participation rates
need to change to secure a balanced labour market in the future.
Figure 8:7 Population by level of education by 1 000 persons
1800 Primary and lower

secondary education
1600
Upper secondary
1400 education
Tertiary education,
1200 lower degree
1000 Tertiary education,

higher degree
800 Unknown
600
400
200
0
1993 1996 1999 2002 2005 2008 2011 2014 2017 2020 2023 2026 2029
The results are presented in Figure 8:8. The projections show a more rapid
increase in demand for labour with a lower degree of tertiary education than the
corresponding increase in supply. This means that it could be beneficial to
increase capacity in higher education somewhat to meet the excess demand. For
higher degree of tertiary education, supply and demand seem to grow more
balanced. Also, the projections show a smaller growth in demand for workers
with upper secondary education compared to supply, and the decrease in
demand for primary educated workers is more pronounced compared to supply.
Hence, to avoid future mismatches in the Norwegian labour market, the
projections show a need for a more rapid educational upgrade compared to what
follows from constant education propensities and participation rates.
126
Figure 8:8 Labour supply by education from Modag compared to adjusted labour
supply from Mosart
8.6. Concluding remarks

Statistics Norway’s projections on future demand and supply of labour by
education is an ongoing project. The models are continuously modified and
further developed, and new data are considered. The projections are uncertain,
because projection period is long and they are based on assumptions that might
be discussed. Therefore, the results must be used with caution.
Nevertheless, judged by the response, it seems to be important that such
quantitative projections are made. Compared to similar projections in other
countries, the models used in the projections in this chapter are well developed
for the purpose. To our knowledge, no other projections are made using a
macroeconomic model with heterogeneous labour, where relative labour demand
depends on relative wages and where relative wages depends on mismatches in
the labour market. Possibilities for mismatch are also studied by comparing
implied labour supply projections at steady unemployment, with labour supply
127
projections according to a dynamic microsimulation model assuming constant

education propensities and participation rates.
References
Bjørnstad, R.; Skjerpen, T. (2006). Trade and inequality in wages and
unemployment. Economic modelling, Vol. 23, Issue 1, p. 20-44.
Boug, P.; Dyvi, Y. (2008). Modag – En makroøkonomisk modell for norsk
økonomi. [Modag, a macroeconomic model for the Norwegian economy].
Social and economic studies No 111, Statistics Norway.
Cappelen, Å. (1992). Modag, a medium term macroeconometric model of the
Norwegian economy. In: Bergman, L.; Olsen, Ø. (eds) Economic modelling
in the Nordic countries. Amsterdam: North-Holland.
Cedefop (2008). Skill needs in Europe: focus on 2020. Luxembourg: Publications
Office of the European Union (Cedefop Panorama series; 160).
Fredriksen, D. (1998). Projections of population, education, labour supply and
public pension benefits: analyses with the dynamic microsimulation model
Mosart. Social and economic studies, No 101, Statistics Norway.
Wilson, R. et al. (2004). Working futures. Wath on Dearne: Sector Skills
Development Agency.
128
CHAPTER 9.
Methods and results of skills demand and
supply forecasting — The case of Germany
Tobias Maier
In June 2010 detailed results of two new forecasts of the future

occupational and qualification structure in Germany were published. One is
the (European) Cedefop forecast of Skills supply and demand in Europe,
and the other the (German) QUBE project on Occupations and
qualifications in the future. This article gives an introduction to the methods
of the QUBE project (42) carried out by the Federal Institute for Vocational
Education and Training (BIBB) and the Institute for Employment Research
(IAB) with the Fraunhofer Institute for Applied Information Technology (FIT)
and the Institute of Economic Structures Research (GWS), and compares
the QUBE results with those of the Cedefop forecast.
The QUBE project contrasts labour force supply and demand projections
for occupational fields by considering occupational mobility. Even though
the taxonomy and methods of the QUBE project differ from those of the
Cedefop forecasts, both projections rely on similar databases and forecast
similar trends for skills demand and supply development in Germany (43).
9.1. Introduction to the methods of BIBB-IAB

qualification and occupational field projections (44)
To identify potential future problem areas in qualification levels and occupations,
demand and supply forecasts in the BIBB-IAB projections are based on the same
(42) Information on the project is available at: www.qube-projekt.de (in German only).
(43) The BIBB-IAB Qualification and occupational field projections (Helmrich and Zika,
2010) from the QUBE project are coordinated projections of supply and demand in
the German labour market up to the year 2025 with the objective of identifying
potential future problem areas in qualification levels and occupations. Therefore, the
projections on the demand side as well as on the supply side are based on the same
data and taxonomy. By considering the occupational flexibility of the labour force, the
QUBE consortium is able to contrast demand with supply. However, this kind of
balancing has to be assessed cautiously as demand and supply projections do not
interact with each other at the current state of the methodology.
(44) The section is mainly an extract from Helmirch et al. (2010). More detailed
information about the methodology of the model set-up can be found at
http://www.bibb.de/dokumente/pdf/BIBB-
IAB_Qualification_and_Major_Occupational_ Field_Projections.pdf.
129
data and taxonomy. The data basis as well as the taxonomy will be described in
the following. Thereafter, the methods of demand and supply forecasts of the
model set-up will be explained.
9.1.1. Data basis and taxonomy

Structural information about age, gender and education of the economically
active population and persons in employment by occupation and industrial sector
has been obtained from several years’ microcensus surveys (45). For the survey
year 2005, it was even possible to reconstruct the ‘initial vocational qualification’
of the workforce by coding the specialisation of the labour force during their
education into an occupational classification. As the microcensus results tend to
underestimate the size of the workforce compared to the German system of
national accounts (Köhne-Finster and Lingnau, 2008), a structurally neutral
upward adjustment of the microcensus figures to the national accounts level was
undertaken for the projections.
On the level of occupations, the classifications of BIBB occupational fields
(Tiemann et al., 2008) were applied to both the demand and the supply side.
These consist of 54 occupational fields which are grouped at the level of the
occupational categories (three-digit codes) from the official German classification
of occupations 1992 (KldB 92) on the basis of comparable job characteristics and
branch dominance. Thus, in contrast to the KldB 92 occupational classification
scheme, they show greater intrahomogeneity and, at the same time, greater
interheterogeneity in their main focus of activity. Mainly for presentational but
also for methodological (46) reasons, initially results are only reported on the level
of the 12 major occupational fields (MOF), which are similarly based on the BIBB
occupational fields (Bott et al., 2010).
On the level of qualifications, the population was differentiated into four skill
levels (as measured by highest formal qualification) following the ISCED
classification framework, namely:
(a) people with no vocational qualification (ISCED 1, 2, 3a);
(b) people with a skilled initial vocational qualification (ISCED 3b, 4);
(45) The microcensus provides official representative statistics of the population and the
labour market in Germany. A description of the microcensus in English is available at
the URL of the German Federal Statistical Office:
http://www.destatis.de/jetspeed/portal/cms/Sites/destatis/Internet/EN/press/abisz/Mik
rozensus__e,templateId=renderPrint.psml.
(46) The sample size for some occupational fields is small, especially if persons in
employment are disaggregated by education and/or age and gender.
130
(c) people with a master craftsman, technical engineer or a trade and technical
school qualification (ISCED 5b);
(d) people with an academic degree (ISCED 5a, 6).
The major difference compared with international forecasts or forecasts in

other countries is the distinction between ISCED 3a and 3b (Cedefop, 2010,
p. 109-110), where people with ISCED 3a are assigned to the category ‘with no
completed vocational education and training (VET)’. But there is one main reason
for this: even though it might be obvious that people in ISCED 3a have, for
example, higher cognitive competences than people at ISCED 0-2, they do still
not have a vocational specialisation; yet the main focus of the QUBE project is to
identify possible future mismatches on the occupational level. It is therefore
necessary to assign people with ISCED 3a together with ISCED 0-2 into the
category ‘no completed VET (47).
To make the available structural information for demand and supply
forecasting in Germany clearer, Table 9:1 shows a breakdown of the German
population according to relevance for the labour market. All fields marked light
blue reflect the essential information for the supply forecast: economically active
and inactive population by age, gender, skill level and VET within 12 MOF. The
necessary information for the demand forecast is marked dark blue: persons in
employment within 59 economic sectors and 12 MOF. The distribution of the
labour force from education into employment within 12 MOF is represented by an
occupational flexibility matrix (marked grey). For the forecasting period, the
amount of potential unemployed persons can be derived by simple definition:
labour force minus persons in employment.
9.1.2. Forecasting demand

The demand projection is based upon actual figures for the years from 1996 to
2006 and projects the realised demand (persons in employment) up to 2025 for
59 industrial sectors and 54 occupational fields. Within each occupational field all
persons in employment are disaggregated into four skill levels (highest formal
qualification). The starting point for the demand projection is the workforce
demand projection up to 2025, which was compiled using the IAB/Inforge model
(47) ISCED 3a ones are designed to provide direct access to ISCED 5A. They do not
provide a vocational specialisation in comparison to ISCED 3b programmes (ISCED
3c programmes are rare in Germany). The skill classification framework applied in
the projections is therefore not fully comparable with the usual low-, medium- and
highly-skilled classification. Instead, it differentiates between people with ‘no
completed VET’ (1, 2, 3a), people with a ‘completed VET’ (ISCED 3b, 4), with a
‘higher VET’ (ISCED 5b) and people with an ‘academic degree’ (ISCED 5a, 6).
131
(Hummel et al., 2010; Meyer et al., 2007) and disaggregated by 59 economic

sectors, on the basis of model calculations with the IAB/Inforge model. This is a
macro econometric input/output model, constructed on the bottom-up principle
and differentiated according to 59 industrial sectors, 59 product groups and 43
types of use. Foreign trade is incorporated endogenously since the model is an
integral element of the Ginfors global model (50 countries, 26 product groups)
developed by GWS mbH (Meyer and Lutz, 2007). In the labour market, the labour
supply and the potential labour force is exogenous. Labour demand is modelled
endogenously on an hourly basis, i.e. the volume of labour necessary for
production is determined first, and only then converted to a per capita figure.
Table 9:1 Structural information for demand and supply forecasting
Demand within 59 economic sectors
Unemployed
Persons aged over 15 in employment within
12 MOF, disaggregated by 4 skill levels
1 2 3 4 5 6 7 8 9 10 11 12
1
By definition: (labour force) - (persons in employment)

VET within 12 MOF (ISCED 3b-6)
2
3
Qualification (four skill levels) and VET
Following retirement, aged over 65

Supply by age cohorts and gender
(pre-school, general schooling)
(some still part of the workforce)

5
Persons aged under 16
Economically inactive
7
8
9
Occupational flexibility matrix
10
11
12
No VET
(ISCED 0-3a)
(Still) in
training
~ 11.5 ~ 4.6 ~ 12.4 ~ 15.9

~36.1 million
million million million million
> > > > > > > > > > > > > > > > > > age cohorts and status > > > > > > > > > > > > > > > > >
Source: Microcensus 2005 of the German Federal Statistical Office; own illustration.
9.1.3. Forecasting supply

The labour supply projection was calculated with two independent models with
the aim of model plurality, both of which, like the demand projection, were based
on the same classification schemes and data generation runs. These consist of
the BIBB-Demos model from GWS, which has certain points of overlap with the
132
IAB-Inforge model and the BIBB-FIT model, an established model from FIT,
already in use for the well-known forecasts by the Bund-Länder Commission for
educational planning and research promotion (BLK, 2001). In addition to the
information on educational participation and participation in employment from
microcensus data, the absolute change and relative fluctuation of age cohorts, in
which women and men are shown separately, are derived from the 12th
coordinated population projection (48) by the German Federal Statistical Office.
Both models allow for the adjustment of the new standard pension age from 65 to
67 but still consider active members of the workforce over this age.
The BIBB-Demos model (Drosdowski et al., 2010) determines the labour
supply, considering several interconnected processes. The demographic trend
determines the distribution across age-cohorts and gender as well as the size of
the potential labour force, namely the number of people aged over 14. Connected
to this is the qualification process (four skill levels), which is notable for a rising
level of qualification, particularly among women, and which also determines the
length of time taken for education. The decision to participate in the workforce is
not only age- and gender-specific but also dependent on the formal qualifications
achieved. Workforce participation is also subject to changes over time. Based on
this information, the choice of an educational specification, an initial vocational
training occupation within the determined 54 occupational fields, is made. It is
ultimately the outcome of a gender-, qualification- and age-specific distribution.
The BIBB-FIT projection is based on several transition models in which,
taking a baseline population as a starting point, future stocks of labour force are
modelled by means of entries and exits. The chosen approach thus relies on
three basic elements: a population projection, a transition model of the education
and training system to quantify the new labour force supply, and an analysis of
workforce participation to determine the remainder.
The projection of the new labour force supply from the VET system in terms
of four skill levels and 54 occupational fields is based on modelling the VET
system including higher education establishments in a way that reflects the
numbers enrolled at the individual training establishments as well as the
transitions to and from the individual educational establishments and the labour
market. This model builds largely on concepts and results from the IAB system of
educational accounting (Bildungsgesamtrechnung; Reinberg and Hummel, 2002)
as well as on the benchmarks of the projection of school pupils and leavers by
the Standing Conference of the Ministers for Education and Cultural Affairs of the
(48) Both models rely on variant 1-W1 with a net migration flow of +100 000 persons
every year.
133
Länder (Kultusministerkonferenz, KMK) and of the KMK projection of university

entrants from the year 2009 (KMK, 2009).
The future new labour force supply from the education system and migration
is, however, only one of the factors that influence the future qualification
structure. It will also increasingly be determined by those who exit from
employment. Here, in particular, the assumptions made about future qualification-
, age- and gender-specific workforce participation also come to bear. The rapid
expansion in workforce participation in past years is extrapolated only moderately
in the projection. To differentiate between occupations, classification of both the
existing labour force and the new labour force supply according to occupational
field was undertaken and extrapolated for the projection period.
9.1.4. Bringing demand and supply together

The microcensus sample 2005 was the first year of the microcensus surveys that
allowed the reconstruction of an initial vocational qualification of the labour force
by coding the specialisation of persons during their education into the
occupational classification scheme KldB 92. Using this information, it was
possible to compute shares of people who worked within the MOF they originally
trained in — the stayers — and shares of people who moved away from their
initial occupation into another MOF — the movers. This occupational mobility is
reflected by an occupational flexibility matrix (see Table 9:1 grey area and Table
9:4) which helps to contrast demand and supply projections. The occupational
flexibilities are assumed to be stable for the whole forecasting period (49) for four
skill levels and three age cohorts (15 to 34, 35 to 49 and 50+; see also Maier
et al., 2010). Hence, future participants of the labour force are distributed into the
12 MOF according to their education-specific (initial vocational qualification), skill-
specific (highest formal qualification) and age-specific occupational flexibility. The
results of the projections are described in the following section.
9.2. Results of the QUBE project

The overall results of the demand and supply forecast can be seen in Table 9:2.
Whereas the demand for labour is slightly rising from 2010 until 2020 before it
goes down slowly, both supply projections forecast a strong decrease of labour
(49) This may seem rather unrealistic in view of changing demand for and supply of skills
in the future, but as no long-range information on occupational mobility (from initial
vocational qualification) is available for Germany, this is the only possible and most
pragmatic approach at the moment.
134
supply from 2009. A theoretical labour shortage will either occur in 2023 (BIBB-
FIT model) or right after the end of the projection period (BIBB-Demos model).
Table 9:2 Demand and supply of labour (in millions)

2005 2010 2015 2020 2025
Persons in employment (IAB-Inforge) 38.9 39.8 40.2 40.2 39.7
Labour force (15+) BIBB-FIT 43.3 43.2 42.5 41.2 39.3
Labour force (15+) BIBB-Demos 43.3 43.7 43.3 42.3 40.4
Source: Helmrich and Zika (2010, p. 19); own illustration.
The demand and supply forecast do not interact with each other. Thus, for
example, a shortage of economically active persons has no effect on the persons
in employment. This is rather unrealistic as either the demand or supply side is
bound to cause reactions on the other side. However, as the necessary adaptive
reactions could not yet be quantified, the results may seem ‘unrealistic’, in total
and in some labour market segments toward the end of the forecasting period.
Nevertheless, such results can be useful as they help to identify future problem
areas if the development of labour supply and demand continues as in the past.
9.2.1. Results on the skills level

Results on the level of skills are presented in Table 9:3. To make the results
more comparable to the results of the Cedefop forecast, ISCED 5b and ISCD 5a
and 6 are grouped together into one category ‘tertiary sector’. Here, the major
differences of the supply projections come to bear. Whereas BIBB-Demos
extrapolates a long-term qualification trend and, therefore, only a slight increase
in persons at ISCED 5 and 6, BIBB-FIT puts a stronger focus on recent
developments in the German educational system. The shortened duration of
schooling at university-track secondary schools and the resulting double cohorts
of school leavers entitled to higher education are reflected in a rising supply of
academically qualified new entrants to the workforce, if a constant ratio of
university entrants is maintained. Compared to new entrants into the labour force,
persons leaving the workforce have a higher share of completed VET (ISCED 3b
and 4). In consequence, the supply of persons with ISCED 3b and 4 will diminish
in contrast to persons with ISCED 5 and 6.
If the status quo persists, people without completed VET will continue to be
affected by high underemployment. For the medium-skilled, the results indicate
that there will be a future shortage of labour occurring either between 2015 and
2020 (BIBB-FIT) or towards the end of the projection period (BIBB-Demos). The
results do not indicate whether there will be a skill upgrade in the demand of
certain jobs or whether the highly-skilled will be employed inadequately if they try
135
to fill the medium-skilled gap. The status quo forecasts only tell us that there will
be either an extremely tight labour market for the highly-skilled (BIBB-Demos) or
an oversupply (BIBB-FIT) if the demand for, and supply of, skills develop as in
the past but independently from each other.
Table 9:3 Results of qualification projections (in millions)

2005 2010 2015 2020 2025
Persons in employment/job opportunities (IAB-Inforge)

ISCED 1, 2 and 3a 5.9 5.7 5.6 5.5 5.3
ISCED 3b and 4 20.6 21.3 21.5 21.6 21.3
ISCED 5a, 5b and 6 10.0 10.2 10.4 10.4 10.3
(Still) in training but working 2.3 2.5 2.7 2.7 2.7
In total 38.9 39.8 40.2 40.2 39.7
Labour force (15+) BIBB-FIT

ISCED 1, 2 and 3a 7.3 7.1 6.7 6.3 5.8
ISCED 3b and 4 23.1 22.9 22.2 21.0 19.5
ISCED 5a, 5b and 6 10.5 11.0 11.5 11.9 12.1
(Still) in training but willing to work 2.5 2.2 2.1 2.0 1.9
In total 43.3 43.2 42.5 41.2 39.3
Labour force (15+) BIBB-Demos

ISCED 1,2 and 3a 7.3 7.3 7.1 6.9 6.6
ISCED 3b and 4 23.1 23.4 23.2 22.5 21.5
ISCED 5a, 5b and 6 10.5 10.7 10.9 10.8 10.5
(Still) in training but willing to work 2.5 2.3 2.1 2.0 1.9
In total 43.3 43.7 43.3 42.3 40.4
Source: Helmrich and Zika (2010, p. 21-26); own illustration.
9.2.2. Results of occupational field projections

If we look at the demand for labour in the 12 defined MOF and we consider only
persons with a completed educational specialisation for those 12 MOF, we can
see that there was already a massive theoretical labour shortage at the beginning
of the projection period in ‘occupations involving the trading and sales of goods’,
‘occupations involving traffic, warehousing, transport, security, guarding’, ‘hotel,
restaurant and cleaning occupations’ and ‘office, commercial service
occupations’ (see dark blue bars in Figure 9:1). Otherwise, there was also an
excess of labour supply in ‘processing, manufacturing and repair occupations’
and ‘technical scientific occupations’. This is possible because occupational
flexibilities and persons without a completed VET qualification are not considered
in Figure 9:1.
136
Figure 9:1 Total demand in MOF and supply of trained persons (ISCED 3b-6)
2005
BIBB-Demos 2025
BIBB-FIT 2025
Raw material extraction occupations
Processing, manufacturing and repair occupations labour shortage
Occupations involving the control and maintenance of machines and plants
Occupations involving the trading and sale of goods
Occupations involving traffic, warehousing, transport, security, guarding
Hotel and restaurant and cleaning occupations
Office, commercial service occupations
Technical and scientific occupations
Legal, management and economic occupations excess labour supply
Artistic, media, humanities and social science occupations
Health and social occupations, body care providers
Teaching occupations
-4000 -3000 -2000 -1000 0 1000 2000 3000
Source: BIBB-IAB qualification and occupational field projections, www.qube-projekt.de; own illustration.
The occupational flexibility matrix is an essential tool to compare the results

on the demand and supply sides on the occupational level. The mobility matrix
(Table 9:4) shows that persons with no completed VET (50) (ISCED 0-3a) mainly
move into ‘processing, manufacturing and repair occupations’ (16.1%),
‘occupations involving the trading and sales of goods’ (11.3%), ‘occupations
involving traffic, warehousing, transport, security, guarding’ (15.1%), ‘hotel,
restaurant and cleaning occupations’ (25.5%) and ‘office, commercial service
occupations’ (10.8%). Except for ‘processing, manufacturing and repair
occupations’, persons with no completed VET mainly move into occupations
where the demand could not be satisfied by medium- or highly-skilled persons
(Figure 9:1). The great inflow of low-skilled workers into ‘processing,
manufacturing and repair occupations’ goes hand in hand with a great outflow of
15.6% persons originally trained in this MOF into ‘occupations involving traffic,
warehousing, transport, security, guarding’. The gap between demand and
supply in 2005 in ‘occupations involving traffic, warehousing, transport, security,
guarding’, stated in Figure 9:1, was therefore filled by low-skilled persons and
persons with vocational education in ‘processing, manufacturing and repair
occupations’. This also explains the excess labour supply in ‘processing,
manufacturing and repair occupations’: more than half of persons originally
trained within this MOF did not work in their MOF, but moved into other
(50) People in this category cannot be assigned to a skilled initial vocational qualification.
137
occupational fields instead. Likewise, in ‘technical and scientific occupations’,

only 52% of the employed persons in 2005 with a technical and scientific
background worked within the MOF of ‘technical and scientific occupations’.
However, the flexibility matrix does not contain information why those people left
their trained occupation. Occupational mobility can occur for several reasons, for
example unemployment in the occupation people have trained for, or better
wages in other occupations (see Maier et al., 2010, p. 156 et seq.). The
occupational flexibility matrix thus gives no explanation for the mobilities but
makes it possible to show which kinds of opportunities are connected with
vocational education in a certain occupation and where competition between
occupations can occur.
Considering the occupational mobility processes as computed for the year
2005 the results of occupational projections in 2025 can be interpreted as follows
(see Figure 9:2): there will be a possible shortage of labour occurring in
‘occupations involving traffic, warehousing, transport, security, guarding’, ‘hotel,
restaurant and cleaning occupations’, ‘legal, management and economic
occupations’, ‘artistic, media, humanities and social science occupations’ and
‘health and social occupations, body care providers’. Tight labour market can
occur in ‘raw material extraction occupations’, ‘technical and scientific
occupations’ and ‘teaching occupations’. The results have to be assessed with
caution as they are not the outcome of an endogenous forecast of demand and
supply. Nevertheless, additional information on mobility processes is helpful for a
better understanding of the outcome.
138
Table 9:4 Occupational mobility (in %)

Proportional values for change from VET into MOF
Number
VET (ISCED 3b-6) in one of the following MOF MOF MOF MOF MOF MOF MOF MOF MOF MOF MOF MOF MOF ∑
of MOF
1 2 3 4 5 6 7 8 9 10 11 12 MOF
1 Raw material extraction occupations 49.5 8.6 3.0 6.3 12.8 6.1 5.1 2.6 2.1 0.7 2.3 0.9 100
2 Processing, manufacturing and repair occupations 1.9 46.3 7.8 6.2 15.6 5.7 4.2 7.4 2.0 0.9 1.5 0.5 100
Occupations involving the control and maintenance of
3 1.3 14.0 44.3 5.2 11.9 4.7 4.4 7.8 2.0 2.4 1.7 0.4 100
machines and plants
4 Occupations involving the trading and sale of goods 0.8 2.9 1.2 50.4 6.2 9.5 19.8 1.3 3.4 1.1 3.1 0.4 100
Occupations involving traffic, warehousing, transport,
5 1.1 6.1 2.0 4.3 65.3 3.8 11.5 2.2 1.4 0.5 1.3 0.5 100
security, guarding
6 Hotel and restaurant and cleaning occupations 3.1 5.0 2.6 9.7 8.2 56.4 7.3 1.1 1.8 0.5 3.7 0.7 100
7 Office, commercial service occupations 0.5 1.2 0.5 8.6 3.7 3.5 71.3 2.0 4.8 1.1 2.3 0.5 100
8 Technical and scientific occupations 0.7 8.9 3.3 5.0 4.1 2.3 8.0 52.0 7.3 3.2 1.7 3.5 100
9 Legal, management and economic occupations 0.2 0.8 0.2 7.3 2.3 1.5 26.0 4.2 49.3 4.1 1.4 2.7 100
Artistic, media, humanities and social science
10 0.3 2.2 0.8 6.1 2.3 2.8 10.2 5.1 5.9 46.9 4.3 13.0 100
occupations
11 Health and social occupations, body care providers 0.4 2.1 0.4 3.9 1.7 4.0 6.3 0.8 1.1 0.9 74.6 3.7 100
12 Teaching occupations 0.3 1.2 0.3 1.9 1.5 2.1 4.3 1.2 1.3 2.2 4.5 79.3 100
0a No completed VET (ISCED 0-3a) 3.2 16.1 6.0 11.3 15.1 25.5 10.8 2.5 1.7 1.9 5.2 0.8 100
0b (Still) at school/in training 2.1 19.8 5.4 14.5 5.8 13.0 17.2 4.8 0.7 3.1 12.2 1.3 100
In total (share on total employment) 2.4 13.8 5.0 11.0 9.2 10.3 17.4 8.5 4.6 3.0 11.1 3.7 100
NB: 7.3% of persons with vocational education in ‘technical and scientific occupations’ move into ‘legal, management and economic occupations’.
Source: Microcensus 2005 of the German Federal Statistical Office; illustration: Maier et al. (2010, p. 170).
139
Projections are based on per capita data and assuming no adaption of work-
time volume for the future, which could help to overcome the theoretical labour
shortage in ‘occupations involving traffic, warehousing, transport, security,
guarding’ and ‘hotel, restaurant and cleaning occupations’. ‘Legal, management
and economic occupations’ already profit from a great inflow from persons with a
different educational background. Therefore, a labour shortage may be countered
by a rise in occupational mobility in this occupation. Similar adaption processes
may be plausible for ‘artistic, media, humanities and social science occupations’.
Labour shortage in ‘health and social occupations, body care providers’ has to be
seen in a different light: due to the need for specific education within this MOF
and the resulting restricted mobility into and out of these occupations, labour
shortage may be avoided only by means of increased numbers of persons
vocationally educated in ‘health and social occupations, body care providers’.
Figure 9:2 Demand and supply including occupational mobility (ISCED 0-6)
2005
BIBB-Demos 2025
BIBB-FIT 2025
Raw material extraction occupations
Processing, manufacturing and repair occupations labour shortage
Occupations involving the control and maintenance of machines and plants
Occupations involving the trading and sale of goods
Occupations involving traffic, warehousing, transport, security, guarding
Hotel and restaurant and cleaning occupations
Office, commercial service occupations
Technical and scientific occupations
Legal, management and economic occupations excess labour supply
Artistic, media, humanities and social science occupations
Health and social occupations, body care providers
Teaching occupations
-1000 -500 0 500 1000 1500
Source: BIBB-IAB qualification and occupational field projections, www.qube-projekt.de; own illustration.
9.3. Differences and common trends between

Cedefop and BIBB-IAB forecasts
BIBB-IAB qualification and occupational field projections rely on similar data
sources as the Cedefop forecasts, namely demographic data, national accounts
and the European labour force survey (cf. Cedefop, 2010, p. 24). Information on
industries, education, occupation as well as age and gender in the Cedefop
forecast are gained from the labour force survey from Eurostat (Stehrer and
140
Ward, 2009). The German the labour force survey was carried out as part of the
German microcensus (Eurostat, 2007, p. 8). Hence, structural information on
occupational choice should be similar. Nevertheless, on the occupational level
results cannot be compared as Cedefop uses the ISCO classification and the
QUBE project relies on BIBB’s occupational fields (Tiemann et al., 2008).
Comparisons of the results for Germany therefore have to be drawn based on
skill levels.
Table 9:5 Comparison of results between BIBB-IAB and Cedefop forecasts:

demand side
Persons in Levels (000s) Change (000s) Change (%)
employment/total job
openings 2000 2010 2020 2000-10 2010-20 2000-10 2010-20
(2010-20)
All qualification
Cedefop 39 144 39 619 39 992 475 373 1.2 0.9
IAB-Inforge 39 144 39 788 40 230 644 442 1.6 1.1
High qualification
Cedefop (ISCED 5-6) 9 928 10 330 11 116 402 786 4.0 7.6
IAB-Inforge (ISCED 5-6) 9 452 10 222 10 429 770 207 8.1 2.0
Cedefop (ISCED 3-4) 22 352 23 734 24 190 1 382 456 6.2 1.9
IAB-Inforge (ISCED 3b, 4) 19 815 21 290 21 596 1 475 306 7.4 1.4
Low qualification / no completed VET

Cedefop (ISCED 0-2) 6 864 5 555 4 687 -1 309 -868 -19.1 -15.6
IAB-Inforge (ISCED 0-3a) 6 121 5 738 5 479 -383 -258 -6.3 -4.5
Source: Cedefop (2010, p. 100-101); Helmrich and Zika (2010, p. 21-26); own illustration.
Table 9:5 presents projected developments of persons in employment in

total and by skill level (highest formal qualification). Considering the results for all
qualifications it is striking that both forecasts point in the same direction and also
project an almost similar trend. In the Cedefop forecast persons with medium
qualification have ISCED 3-4, whereas the IAB-Inforge model only includes
persons with ISCED 3b and 4 in this category. Hence, ISCED 3a is assigned to
the ‘no completed VET’ category in the IAB-Inforge model and to the medium-
skilled in the Cedefop forecasts. Thus, results on the medium and low
qualification level are not fully comparable between the two forecasts. This might
also be the reason why Cedefop forecasts predict a much more pronounced
decrease in persons with a low qualification. The demand for highly-skilled
labour, which should be comparable between the two approaches, is forecast to
increase more sharply in the Cedefop projection.
141
With regard to the supply forecasts, the Cedefop forecast assumes in total
around 2.3 million fewer persons in the labour force than there were in Germany
in 2000. The reason for this can be that either the labour force survey data have
not been adjusted to national accounts level like the microcensus in the BIBB-IAB
forecasts, which is necessary because for some reason the microcensus tends to
put the size of the workforce lower than the German system of national accounts;
or persons still at school or in training but members of the workforce have not
been included in the total. Looking at the figures in Table 9:3, it seems more
plausible that persons still in training have been omitted from the chart in
Cedefop (2010, p. 88-89), because the difference is more or less the number that
is displayed for people still at school or training, but also members of the
workforce in the QUBE project. To compare the projected development of both
forecast approaches, it is therefore recommended to look at the percentages of
change in Table 9:6 (in bold). On the results in total, the Cedefop forecast seems
to be more optimistic about the decrease in the labour force, but in general, all
three models projected similar trends. On the supply of highly-skilled persons, the
Cedefop forecast is much closer to the BIBB-FIT model than to the BIBB-demos
model. The developments of people with medium and low qualifications/‘no
completed VET’ are, again, not fully comparable due to the difference in the
category-assignment of ISCED 3a in each approach. This, again, could explain
why the supply of low-skilled decreases much more sharply in the Cedefop
forecast than in the BIBB-Demos or BIBB-FIT model.
9.4. Conclusion and further development

This chapter introduced the methods of the BIBB-IAB qualification and
occupational fields projection. Results of the QUBE project have been reported
and a comparison with the results of the Cedefop forecasts has been given. The
comparison showed that both forecasting approaches draw on similar data
sources and that the results in total and on qualification level point in the same
direction. International comparisons of Germany regarding skill specific
developments are considered more reliable if they are consistent at national
level. While similar developments on the qualification level in the QUBE project
and in the Cedefop approach suggest that occupational developments in the
BIBB-IAB forecasts could also be used for the national German results within the
pan-European approach. This is possible because in the QUBE project, the
qualification level is strongly connected with the choice of an occupation (Section
9:1). To promote an international comparison on the occupational level in the
future, the QUBE project has set itself the goal of using the new German
142
classification of occupations 2010 (Wiebke et al., 2010) which is more consistent

with the ISCO classification.
Table 9:6 Comparison of results between BIBB-IAB and Cedefop forecast:

supply side
Levels (000s) Change (000s) Change (%)
Labour force (15+) 2000-20
2000 2010 2020 2000-10 2010-20 2000-10 2010-20
All qualification
Cedefop 39 447 41 458 40 815 2 011 -643 5.1 -1.6
BIBB-Demos 42 175 43 658 42 254 1 483 -1 404 3.5 -3.2
BIBB-FIT 42 175 43 152 41 180 977 -1 972 2.3 -4.6
High qualification
Cedefop (ISCED 5-6) 9 663 10 287 10 960 624 673 6.5 6.5
BIBB-Demos (ISCED 5-6) 9 952 10 716 10 845 764 129 7.7 1.2
BIBB-FIT (ISCED 5-6) 9 952 10 981 11 879 1 029 898 10.3 8.2
Cedefop (ISCED 3-4) 22 531 24 786 24 716 2 255 -70 10.0 -0.3
BIBB-Demos (ISCED 3b, 4) 21 784 23 390 22 549 1 606 -841 7.4 -3.6
BIBB-FIT (ISCED 3b, 4) 21 784 22 926 21 032 1 141 -1 894 5.2 -8.3
Low qualification / no completed VET

Cedefop (ISCED 0-2) 7 253 6 384 5 139 -869 -1 245 -12.0 -19.5
BIBB-Demos (ISCED 0-3a) 6 681 7 258 6 873 577 -384 8.6 -5.3
BIBB-FIT (ISCED 0-3a) 6 681 7 082 6 274 401 -808 6.0 -11.4
Source: Cedefop (2010, p. 88-89), Helmrich and Zika (2010, p. 21-26); own illustration.
The greatest challenge for further development of the QUBE project will be
to implement endogenous balancing processes between demand and supply.
This comprises to consider wages, volume of work and a dynamisation of the
flexibility matrix, which plays a central role within the forecasting framework. Pan-
European steps to manage the interaction and balancing process are therefore
expected to improve exchange of ideas between both forecasting approaches.
Abbreviations
Bundesinstituts für Berufsbildung
BIBB
[Federal Institute for Vocational Education and Training]
Institut für Arbeitsmarkt- und Berufsforschung
IAB
[Institute for Employment Research]
Fraunhofer-Institut für Angewandte Informationstechnik
FIT
[Fraunhofer Institute for Applied Information Technology]
Gesellschaft für wirtschaftliche Strukturforschung
GWS
[Institute of Economic Structures Research]
MOF major occupational fields
143
ISCED International standard classification of education (1997)

ISCO International standard classification of occupations
References
BLK (2001). Zukunft von Bildung und Arbeit: Perspektiven von Arbeitskräfte-
bedarf und -angebot bis 2015 [Future of education and work: perspectives of
labour demand and supply by 2015]. Bonn: BLK, Bund-Länder Commission
for educational planning and research promotion (Materialien zur
Bildungsplanung und zur Forschungsförderung [Materials for educational
planning and research promotion], Vol. 104).
Bott, P. et al. (2010). Datengrundlagen und Systematiken für die BIBB-IAB-
Qualifikations- und Berufsfeldprojektionen [Databases and classifications of
the BIBB-IAB qualifications and professional field projections]. In:
Helmrich, R.; Zika, G. (eds) Beruf und Qualifikation in der Zukunft: BIBB-
IAB-Modellrechnungen zu den Entwicklungen in Berufsfeldern und
Qualifikationen bis 2025. [Qualifications and occupation field projections:
BIBB-IAB-model calculations on the developments in vocational areas and
qualifications by 2025]. Bielefeld: Bertelsmann, p. 63-80.
Drosdowski, T. et al. (2010) Entwicklung der Erwerbspersonen nach Berufen und
Qualifikationen bis 2025: Modellrechnungen mit dem BIBB-Demos-Modell.
[Development of the labour force by occupation and qualification until 2025:
model calculations with the model-BIBB-DEMOS] In: Helmrich, R.; Zika, G.
(eds) Beruf und Qualifikation in der Zukunft [Occupation and qualification in
the future]. BIBB-IAB-Modellrechnungen zu den Entwicklungen in Berufs-
feldern und Qualifikationen bis 2025. [Qualifications and occupation field
projections: BIBB-IAB-model calculations on the developments in vocational
areas and qualifications by 2025] Bielefeld: Bertelsmann, p. 125-152.
Eurostat (2007). Labour force survey in the EU, candidate and EFTA countries:
main characteristics of the national surveys 2005. Luxembourg: Publications
Office of the European Union (Eurostat Methodologies and Working papers).
Helmrich, R.; Zika, G. (eds) (2010). Beruf und Qualifikation in der Zukunft
[Occupation and qualification in the future]. BIBB-IAB-Modellrechnungen zu
den Entwicklungen in Berufsfeldern und Qualifikationen bis 2025
[Qualifications and occupation field projections: BIBB-IAB-model calculations
on the developments in vocational areas and qualifications by 2025].
Bielefeld: Bertelsmann.
Helmrich, R. et al. (2010). BIBB-IAB qualification and major occupational field
projections: notes on the methodology of a cooperation project. Bonn: BiBB,
144
Bundesinstituts für Berufsbildung [Federal Institute for Vocational Education

and Training].
Hummel, M. et al. (2010). Der Arbeitskräftebedarf nach Wirtschaftszweigen,
Berufen und Qualifikationen bis 2025 [Labour demand by industry sector,
occupation and qualifications by 2025]. In: Helmrich, R.; Zika, G. (eds) Beruf
und Qualifikation in der Zukunft [Occupation and qualification in the future].
BIBB-IAB-Modellrechnungen zu den Entwicklungen in Berufsfeldern und
Qualifikationen bis 2025 [Qualifications and occupation field projections:
BIBB-IAB-model calculations on the developments in vocational areas and
qualifications by 2025]. Bielefeld: Bertelsmann, p. 81-102.
KMK (2009). Vorausberechnung der Studienanfängerzahlen 2009-20:
Zwischenstand [Prediction of student number 2009-20: preliminary result].
Bonn: Sekretariat der Ständigen Konferenz der Kultusminister (KMK) der
Länder in der Bundesrepublik Deutschland Referat IV D [Secretariat of the
Standing Conference of the Ministers of Education and Cultural Affairs of the
Länder in the Federal Republic of Germany Section IV D].
Köhne-Finster, S.; Lingnau, A. (2008). Untersuchung der Datenqualität
erwerbsstatistischer Angaben im Mikrozensus. Ergebnisse des Projektes
‘Nachbefragung im Mikrozensus/LFS’ [Analysis of quality of data acquired in
the microcensus: results of the project ‘Follow-up of the microcensus/LFS’].
Wirtschaft und statistik, Vol. 12, p. 1067-1088.
Maier, T. et al. (2010). Flexibilität zwischen erlerntem und ausgeübten Beruf
[Flexibility between learned and practised occupation]. In: Helmrich, R.;
Zika, G. (eds) Beruf und Qualifikation in der Zukunft [Occupation and
qualification in the future]. BIBB-IAB-Modellrechnungen zu den
Entwicklungen in Berufsfeldern und Qualifikationen bis 2025 [Qualifications
and occupation field projections: BIBB-IAB-model calculations on the
developments in vocational areas and qualifications by 2025]. Bielefeld:
Bertelsmann, p. 153-180.
Meyer, B.; Lutz, C. (2007). The Ginfors model: model overview and evaluation.
Osnabrück: GWS mbH [Institute of Economic Structures Research].
Meyer, B. et al. (2007). National economic policy simulations with global
interdependencies: a sensitivity analysis for Germany. Economic Systems
Research, Vol. 19, No 1, p. 37-55.
Reinberg, A.; Hummel, M. (2002). Die Bildungsgesamtrechnung des IAB [IAB
education accounts]. In: Kleinhenz, G. (ed.). IAB-Kompendium Arbeitsmarkt-
und Berufsforschung. Beiträge zur Arbeitsmarkt- und Berufsforschung [IAB’s
Compendium on labour market and employment research. Contributions to
the labour market and occupational research], 250, p. 491-506.
Stehrer, R.; Ward, T. (2009). Using LFS data for estimating changes in skill
demand and supply at a pan-European level. Paper presented at the
workshop on forecasting skill supply and demand in Europe: interim results.
145
Technical workshop organised by Cedefop in Valetta, 15 and 16 October

2009.
Tiemann, M. et al. (2008). Berufsfeld-Definitionen des BIBB [BiBB definition of
professional field]. Bonn: BiBB, Bundesinstituts für Berufsbildung [Federal
Institute for Vocational Education and Training] (Wissenschaftliches
Diskussionspapier [Scientific discussion paper]).
Wiebke, P. et al. (2010). Klassifikation der Berufe 2010: Entwicklung und
Ergebnis [Classification of occupations, 2010: development and results].
Nuremberg: Federal Employment Agency, Statistics (Methodical report).
146
CHAPTER 10.
The FGB-LM model: structure and recent
forecasts of the Italian labour-market stocks
and flows
Giuseppe Ciccarone and Massimiliano Tancioni
This paper describes a microfunded econometric model recently developed

by the Fondazione Giacomo Brodolini to produce good quality forecasts of
labour-market stock and flows. Some results from a recent application of
the model to Italian data are also presented and briefly discussed, focusing
on the predicted evolution of the employment stock and of the hiring
process in major disaggregation, such as cohorts, educational attainments
and professional profiles. The presentation has the main scope to exemplify
the potential uses of the model in the evaluation of future labour-market
trends and in the simulation of policy options.
10.1. Introduction
In this paper we present the main features of the econometric model developed
by the Fondazione Giacomo Brodolini for the Italian labour market (FGB-LM),
providing some results of a recent application. This model, which is tailor-made
for the simulation and forecasting of the labour-market dynamics at high level of
disaggregation, is composed of two main building blocks (51). A theory-based
core, or pilot, new Keynesian dynamic stochastic general equilibrium (NK-DSGE)
model (52) describes the evolution of the fundamental macro variables of the
system. ‘Satellite’ blocks of equations, derived from empirical relations, define the
breakdown of the employment stocks and flows by sector, region, age,
occupation and qualification. The aim of this ‘hybrid’ model is to devise a formal
structure with sound theoretical foundations and, at the same time, able to
minimise estimation and forecast errors.
The ‘pilot’ model structure is thus derived from the solution of constrained
maximisation problems by firms and workers under fully specified theoretical
hypotheses regarding the market structure, the production technology and
(51) A more technical exposition of the FGB-LM model is in Giuli and Tancioni (2009).
(52) For a technical description of the of NK-DSGE models see, e.g. Smets and Wouters
(2003; 2007); Pagan (2003); Erceg et al. (2005); Del Negro et al. (2007); Coenen
et al. (2007); Forni et al. (2007) and Christoffel et al. (2009).
147
individual preferences. The main distinctive feature of the model lies in the
hypotheses describing the functioning of the labour market, which is modelled
according to the Mortensen and Pissarides (1994) approach, which has recently
been introduced into the NK-DSGE modelling apparatus (Blanchard and Galì,
2010; Riggi and Tancioni, 2010).
To improve the forecasting performances, the standard hypothesis of model-
consistent expectations is replaced by that of data-consistent expectations,
described by a VECM (53) structure in which only the cointegration space is
identified according to the standard long-term theoretical prediction of balanced
growth (54). The VECM is estimated and simulated before core model estimation,
so as to take expectations as exogenous at the beginning of the forecast.
A further distinctive model feature is the use of estimated empirical relations
for the different events of exit from the employment status, defining the aggregate
separation rate. By specifying the exit rates as functions of both economic and
institutional variables, the model is able to capture the main effects on
employment produced by interventions on labour-market regulation and on the
social security system.
The other main motivations underlying our modelling approach can be
summarised as follows:
(a) lack of long-time series for particular variables and disaggregations (flow
data) (55) does not make data-intensive methods, such as VARs, a viable
option, whereas the possibility to calibrate some parameters in DSGE
models allows us not to rely on particularly large samples;
(b) theoretical identification allows us to perform policy simulations, provides
optimal policy-making design and improves policy communication
transparency.
The model employs macroeconomic data from national and international

official sources. In particular, separation and hiring rates (by age, profession,
qualification) are obtained from the elementary data of the Italian National
Statistical Institute. The model outputs are forecasts for all the model variables:
(53) Vector error correction model.

(54) The joint data density, under the standard hypotheses of stationarity, linearity and
normality, has a VAR (vector autoregression) representation and this ensures data
consistency. Considering DSGE model stability properties (long-term balanced
growth), a VECM ensures consistency among expectational variables and model
properties.
(55) For example, regional historical data are available only on a yearly basis and for a
limited period.
148
macroeconomic aggregates, labour force and employment stocks and flows

(separations and hiring) by age, sector, profession, and qualification.
The generalised method of moments is applied for the estimation of the
‘pilot’ model, whereas the blocks of equations defining the empirical relations are
estimated using the FIML (56) or the SURE (57) estimators.
In Section 10.2 we describe the ‘pilot’ structure and the relations linking the
national and regional levels, together with those defining the state transitions, i.e.,
labour-market entry and exit flows, and the relations defining the disaggregation
represented in the model. Section 10.3 goes briefly into the procedures for
estimation and econometric simulation of the model, providing reference to the
data used and a concise overview of the information produced by the model.
Section 10.4 presents the main model forecasts of the Italian labour-market
variables. Section 10.5 concludes.
10.2. The model
10.2.1. The ‘pilot’ structure

NK-DSGE models with a Walrasian labour market display a lack of involuntary
unemployment. Labour input fluctuations along the cycle are hence fluctuations
in hours worked guided by variations in the marginal rates of substitution
between work and leisure rather than fluctuations in the extensive margin (i.e.
employment fluctuations tout-court). These models also face difficulties in
replicating the observed comovements between the macroeconomic labour-
market variables. The real wage exhibits low volatility, while it is the extensive
margin that undergoes the major cyclic fluctuations. Finally, being the real
marginal costs the variable driving inflation (which lies at the heart of the New
Keynesian Phillips curve), these models lack of inflation persistence because real
wages are instantly adjusted to clear the market, determining adjustments of
inflation to its steady state value that are far more rapid than suggested by the
empirical evidence.
The recent developments of NK-DSGE models have tackled these
difficulties through two main theoretical extensions: the introduction of
nominal/real rigidities in wage determination (Erceg et al., 2000); the introduction
of real frictions in the match between labour supply and demand à la Diamond
(56) Full information maximum likelihood.

(57) Seemingly unrelated regression equation.
149
(1982a, b) and Mortensen and Pissarides (1994) (e.g. Blanchard and Galì, 2010).
The first amendment is able to reconcile the dynamic properties of the NK-DSGE
models with the principal comovements of the macroeconomic variables over the
cycle and to improve the persistence of the convergence processes to the steady
state. The second extension focuses on the matching between labour demand
and supply, and on the empirical finding that job destruction and job creation
coexist in the various phases of the business cycle. According to this approach,
the matching of labour demand and supply resulting from the maximising
behaviour of firms and households is described by a matching function, whose
arguments are the vacancies posted by firms and the number of job-searchers
(given by the stock of unemployed in the previous period plus the workers who
have seen their working relationship terminated in the present period). Wages, on
the other hand, are determined through a Nash bargaining between unions and
firms on the surplus generated by the realisation of a match (58). The main virtue
of this approach (in addition to the presence of involuntary unemployment
throughout the cycle) lies in the possibility to replicate the main comovements
between unemployment and vacancies, and between the flows of job creation
and job destruction (Trigari, 2006; 2009).
The FGB-LM model incorporates these two theoretical developments into a
unifying DSGE framework. The wage is subject to a real/nominal rigidity which
improves the persistence of the convergence processes to the long-term
equilibrium; the presence of search and matching frictions allow for the
coexistence of processes of destruction and creation of new jobs, the quantitative
evaluation of which is the ultimate aim of the approach.
The economic system is populated by three types of agents: households,
firms and policy-makers. Households can be rationed in the credit market, giving
rise to liquidity constraints that introduce a one-period perspective in
consumption choices (59). We hence assume that only a fraction of the
households can borrow and/or lend in the financial markets. The households that
(58) The presence of hiring costs generates a surplus for every working relationship
created.
(59) This hypothesis has been introduced to account for the empirical evidence of
consumption dynamics closely correlated with current income, i.e.rule-of-thumb
consumption (see, e.g. Campbell and Mankiw, 1989; 1990; 1991; Di Bartolomeo
et al., 2011).
150
cannot do so can only pay for their present consumption with the current
disposable income (60).
Given our modelling goals, heterogeneity needs to be defined also in terms
of gender and age but, for the sake of simplicity, the demographic heterogeneity
is formally represented only in the ‘satellite’ model, where the proportion of each
cohort is defined by demographic hypotheses regarding fertility, life expectation
and migratory flows.
For improving the consumption persistence observed in the data, the model
also assumes the presence of external habits in households’ utility from
consumption (61) (defined in log-linear form, whereas labour disutility is modelled
as a constant relative risk aversion, CRRA, function). Utility maximisation under
the budget constraint leads to a classical Euler condition in which, for the
households able to access the financial markets, the intertemporal allocation of
consumption depends on the expected real interest rate and, for the households
subject to financial constraints, on a one-period condition depending solely on the
current real wage. The aggregation of these first-order conditions determines,
together with a standard capital (thus investment) law of motion, the aggregate
demand dynamics. As for the labour supply, households are ready to supply their
labour services to firms up to the point at which the current real wage equals the
marginal rate of substitution between work and consumption.
Intermediate goods are produced by a continuum of firms, each of which
produces a differentiated good through a production function with constant
returns to scale. Employment in a firm evolves on the basis of the following law of
motion:
Nt  i   1    Nt 1  i   H t  i  (1)
where   0,1 is a separation rate indicating the probability of transition for

a worker from the pool of the employed to that of the unemployed and H t  i 
represents the fraction of unemployed workers that are employed by the i-th firm
in period t (hiring of new labour). As the use of elementary data allows for the
identification of different events of separation, completely specified in the
simulations (firing, retirement, end of temporary labour contracts, work-related
(60) The main consequence of introducing liquidity constraints is a violation of the Barro-
Ricardo equivalence which entails the possibility of crowding-in effects in public
spending.
(61) Formally, the hypothesis of habit persistence introduces an autoregressive
component in consumption.
151
injury, mortality and other causes) the separation rate is actually split into six
different rates (62).
Following the approach of Blanchard and Galì (2010) and Riggi and Tancioni
(2010), labour-market frictions are introduced by assuming that hiring costs can
be expressed as an increasing function of the job finding rate.
The aggregate resource constraint equates consumption and investment to
gross national income minus public spending for goods and services (its trend
emerging from an exogenous process) and the quantity of real resources which
are employed to hire new labour (hiring costs), given the quantity of employment
desired by the firms. Labour hiring thus equals the replacement demand, i.e., the
quota which has separated in the previous period, plus the net employment
variation (63).
The firms’ minimisation of costs gives rise to a marginal cost function linking
marginal costs to the total costs of labour and to the hiring costs, both normalised
by productivity, considering the fact that these costs are not to be borne in the
successive period by the firms that have not separated from workers.
Monopolistic competitive firms in the final goods sector aggregate the
intermediate goods acquired at their marginal cost and fix their price adding a
mark-up to the marginal costs. In each period only a proportion of firms in the
final sector can change the price of their differentiated good, setting it at the
optimal level (Calvo rule). From the linearisation around the steady state of the
first-order condition for profit maximisation the New Keynesian Phillips curve is
obtained:
1   1   
 t  Et  t 1  mct   t
p p
(2)
p
where  t and mct are the log deviations of inflation and of the marginal

costs from their steady state values, and  t is a cost-push shock.
The monetary policy authority adopts a Taylor rule on the basis of which the
rate of interest gradually adjusts to deviations of inflation and income from their
target values.
(62) The Cedefop model also considers replacement demand for exits in hiring. However,
only permanent separations are considered (Cedefop, 2010).
(63) In the steady state, and assuming a stationary population, the system would thus see
a constant level of employment, with equivalent exit and entry flows (Bagnai et al.,
2006).
152
Under the hypothesis of a balanced budget, taxes finance the purchase of

differentiated goods by the public sector, with the trend in public spending
assumed to be first-order autoregressive and exogenous.
The hypothesis of sticky wages is introduced by modelling the labour-market
similarly to the commodity market. There exists a continuum of differentiated
work services indexed in the unit interval, each of which is used by each firm.
Every household is specialised in one type of work and sets the nominal wage at
which the members of the household are willing to work (alternatively, we may
imagine a continuum of monopolistic unions, each of which is representative of a
household). Just as happens in the commodity market, in each period only a
certain proportion of households/union can renegotiate their wages. This is done
in such a way as to maximise the present value of the difference between the
real wage and the marginal rate of substitution between labour and leisure time,
weighted to consider the probability of not being able to change the wage in
some future periods.
The linearisation of the first-order condition for the household/union problem
provides the equation describing the wage dynamics:
1   w 1  w 
 tw  Et  tw1 
 w 1   w 
smstA  wtreal    t
w
(3)
where  twreal   tw   t . A variation in nominal wages positively depends on

the (logarithmic) difference between the aggregate marginal rate of substitution
smstA and the real wage.
To study the dynamic properties of the model, the system equations are
linearised around the steady state that would prevail in the absence of shocks
and nominal frictions. To represent the growth of the system, in technical
implementation it is assumed that the steady state evolves over time following a
linear deterministic trend. A stationary stochastic technology is then assumed
around a first-order deterministic trend to capture the long period evolution in
technical progress and demography (Smets and Wouters, 2007).
10.2.2. The ‘satellite’ model

The equations in the ‘satellite’ model occupy a recursive position vis-à-vis the
block of structural equations of the ‘pilot’ model. In the representation of the
relations between stocks and flows, consistency between the blocks is
guaranteed by applying an equivalent methodological approach in the two blocks
(as the second block decomposes the magnitudes of the first one), and by
imposing a series of numerical consistency constraints between aggregated and
153
disaggregated variables. To maximise the forecasting precision of the

disaggregation, the relations represented in the ‘satellite’ block are constructed
on eminently statistical and econometric bases, albeit with the constraint of
respecting the results of simulation and forecasting at the aggregate level
produced by the ‘pilot’ structure.
For each macroeconomic variable represented in the ‘pilot’ model, regional
disaggregation is obtained through the estimation of a system of R regional auto-
regressive distributed lags simultaneous equations (64). For the cointegration
between national and regional variables (i.e., balanced growth among regions),
the auto-regressive distributed lags pattern has a representation in terms of both
long-term static relationships and error-correcting dynamic relationships (Pesaran
and Shin, 1999). By introducing a linear trend it is possible to deterministically
capture the shifting dynamics that may occur in regional evolution. In this case,
the presence of cointegration determines a dynamic statistical equilibrium around
the linear trend.
Sectoral disaggregation of the macroeconomic variables is achieved
estimating a system of N simultaneous equations, which are again defined in
auto-regressive distributed lags formulation. To capture the ongoing trends in
sector evolution, a second-order trend polynomial, able to represent any
quadratic trend, is estimated (65).
Simulation and forecasting of disaggregated labour-market exits
(outflows) (66) are obtained using a consistent extension of the fundamental
stock-flow relation (1), which associates the stock of employed in the current
period with the stock of employed in the previous period, considering the
separation rate and the hiring process (labour matches).
The definition of the six event-specific separation rates — which are
assumed to evolve over time and specified in terms of regional, sectoral and
cohort disaggregation — requires processing on the elementary data of the
Italian National Statistical Institute. As for the separation rate due to biological
causes, use is made of the projected survival rates provided by official sources,
while for the evolution of the separation rates relative to the end of temporary
(64) The autoregressive distributed lag formulation determines a relationship between the
dependent variable, its past values (autoregressive component) and the
contemporary and past values of the explicative variables (distributed lag
component).
(65) This choice is supported by the historical structural modifications observed in the
data.
(66) Outflows at the aggregate level are obtained by multiplying the stock of employed in
the previous period by the separation rate.
154
contracts, work-related injury and other causes, a second-order time polynomial

is estimated and projected to the forecast horizon. A more complex formulation is
used to estimate the exit rate due to retirement and firing. In the former case, the
rate is assumed to be a function of regulation dummies, which capture the effects
of the ongoing and future regulatory modifications on the eligibility conditions for
retirement and on the expected pension income, which in turn depends on the
cohort-specific life expectation. The separation rate due to firing is assumed to
depend on the cyclic phase, defined in terms of the deviation of the growth rate
from its long-term value.
The employment stock and flows determined at regional and sectoral level
are divided over the various ages using a system of distribution equations (67).
The disaggregation of the stock of employment and of employment
requirement by occupation and by qualification is obtained by breaking down
these two magnitudes, defined at regional and sectoral levels, with a system of
weights assumed to evolve over time, according to a second-order time
polynomial.
10.3. Estimation, data sources and model output

Different estimators are used to estimate the coefficients of the model. The
structural equations defining of the ‘pilot’ model are estimated by employing the
generalised method of moments. The use of this method estimator is justified by
the theoretical characterisation of these equations, as it naturally emerges from
the optimising approach characterising the first block of the model (Favero,
1996) (68).
With respect to the equations defining the satellite structure, the FIML (69)
estimator is generally employed, while using Zellner’s (1963) SURE (70) estimator
for systems in which the equations have the same construction, such as in the
regional disaggregation blocks, to capture the latent correlation in the error
structure and improve efficiency. This choice is appropriate to consider the latent
cross-correlation otherwise unrepresented in the systematic part of the
simultaneous systems.
(67) For a definition of the distribution function, see Bagnai et al. (2006).
(68) The orthogonality conditions between instruments and the regressions residuals
proved to be perfectly consistent with the constraints imposed by the theory.
(69) Full information maximum likelihood.
(70) Seemingly unrelated regression equation.
155
The ‘pilot’ model is estimated with quarterly time series for the main
macroeconomic aggregates (GDP, private consumption and investment, public
consumption, real wages, GDP deflator, short-term interest rate, population,
labour force, employment and unemployment rate) which are readily available at
both the national and international level from official statistical sources. The
marked degree of disaggregation in the output produced by the second block
equations calls instead for the use and processing of regional account data and
of the elementary labour force survey data of the Italian National Statistical
Institute. As for the latter source, the necessary disaggregations are in terms of
decomposition by age of the population, of the labour force, of employment and
unemployment. Processing of elementary data is also necessary to identify the
various modes of exit from work, through which the specific transition rates are
obtained. Processing is finally needed to determine the weights to be applied in
defining the equations linking the aggregate employment stocks and flows with
the distribution by cohort, occupation and qualification.
10.4. Labour-market forecasts for the Italian economy

In this section we present some results provided by a recent application of the
FGB-LM model to Italian data. The ‘pilot’ model is estimated over the sample
period 1981:1-2010:2, and the satellite blocks over the sample period 1992:1-
2009:4. The forecast considers the period 2010:3-2015:3.
The main exogenous components in the model, i.e., public expenditure,
population and total factor productivity, are projected to the forecast horizon
using estimated ARIMA (71) processes. The variables controlling for the evolution
of the regulatory set-up are held fixed at the last period’s observation. Labour
force dynamics is described by an auto-regressive distributed lags equation in
which population enters as log-term forcing variable and the lagged
unemployment rate is an exogenous component entering the VEC (72)
representation only. Expectations are obtained by simulating the unrestricted
VECM mentioned in the introduction.
10.4.1. The macroeconomic environment

Figure 10:1 reports the historical and forecasted evolution (grey area) of six
mayor macroeconomic aggregates expressed in real terms: GDP, private
(71) Autoregressive integrated moving average.

(72) Vector error correction model by Johansen.
156
consumption, private investment, government expenditure, net export and the

real wage.
The picture for GDP indicates a gradual and basically moderate recovery for
the forecasted period, mostly due to the post-crisis jump in private investment.
Consumption dynamics is predicted to remain slightly negative in the first
half of 2011 and almost flat for the subsequent four periods. Convergence to the
(balanced) long-term rate of growth is only achieved in the second half of 2012.
The hypothesis of liquidity constraints and thus of non-Ricardian (or rule of
thumb) consumers is decisive for explaining the predicted pattern of private
consumption. Under this hypothesis, which is decisively supported by the data,
the peculiar evolution of consumption can be attributed to the weak dynamics of
the real wage. Note that the estimates indicate that nearly 35% of Italian
households do not smooth consumption, i.e., a relevant fraction of them makes
consumption decisions on the basis of current income.
The presence of liquidity constraints implies a second indirect effect on
consumption, related to the fact that, under rule of thumb behaviour in
consumption, public expenditure can have crowding-in effects. Since the
predicted dynamics of government expenditure is below the balanced growth
trend, private consumption is likely to be negatively affected.
The dynamics of net exports indicate a persisting negative balance for a
great part of the forecast period, with initial negative effects on aggregate
demand.
Table 10:1 summarises the predicted dynamics for these variables (splitting
the net export series into its two components, export and import), reporting the
annual growth rates for the period 2010-14.
At the time the forecast was made, only the first two quarters of 2010 were
known, thus the 2010 annual growth rates are obtained considering the model
predictions for the third and fourth quarter. This allows us to evaluate, albeit over
a reduced time span, the forecasting performances of the FGB-LM model.
Despite the particularly critical time being considered, actually a post-crisis/pre-
recovery period, the model showed a good performance, pointing to an annual
growth rate of GDP of nearly 1.2%, which is the actual GDP growth rate for 2010
recently certified for Italy by the national statistical office and by international
institutions.
The good predictive capabilities of the ‘pilot’ model are basically confirmed
also considering the dynamics of the other variables.
157
Figure 10:1 Macroeconomic environment. Forecast period 2010:3-2015:3
Billions Billions
360 220
200
320
180
280 160
140
240
120
200 100
1985 1990 1995 2000 2005 2010 2015 1985 1990 1995 2000 2005 2010 2015
Real GDP Real consumption
Billions Billions
70 70
60 60
50 50
40 40
30 30
1985 1990 1995 2000 2005 2010 2015 1985 1990 1995 2000 2005 2010 2015
Real investment
Real government expenditure
Billions
16 6500
12
6000
8
5500
4
5000
0
-4 4500
-8
4000
1985 1990 1995 2000 2005 2010 2015
1985 1990 1995 2000 2005 2010 2015
Real net export
Real wage
Source: FGB-LM model forecasts.
10.4.2. Labour market stocks and flows

Figure 10:2 reports the historical and predicted evolution of the labour force, of
employment and of the unemployment rate. Considering the simulation period,
after an initial stagnation in the last two quarters of 2010, the labour force is
158
expected to linearly evolve over its long-term path. Differently, a further moderate
reduction is expected for the employment stock until 2011:1, followed by a period
of substantial stability or moderate increase until 2012:4. A clear recovery in the
employment stock will take place only in the second quarter of 2013.
Table 10:1 Predicted annual growth rates for the major macroeconomic
aggregates
GDP CONS INV GOV EXP IMP WAGE
2010 1.26 -0.32 2.45 -0.19 9.18 7.35 0.34
2011 1.43 -0.52 2.12 0.71 9.32 5.59 0.92
2012 1.49 1.23 3.12 0.61 2.23 2.20 0.42
2013 1.49 1.13 2.69 0.56 4.18 3.49 0.60
2014 1.32 1.13 2.14 0.52 3.01 2.53 0.75
Given these evolutions, the unemployment rate is expected to increase to

nearly 9.5% at the end of 2012 (from 8.7% in 2010:2) and then weakly decrease
to nearly 9.2% at the end of the forecast window.
The expected weak performances of the Italian labour market are mostly
related to the weakness of the predicted macroeconomic dynamics. The
expected average growth rate for GDP in the simulation window is below 1.5%,
basically in line with the long-term total factor and labour productivity growth
rates. This rules out the possibility of observing a clear process of recovery in the
labour market stocks. The expected increases in employment beginning in 2013
must in fact be attributed almost entirely to the increase in the labour force (thus
supply) stock.
Concerning the labour market flows, Table 10:2 reports the expected annual
outflows and inflows (hiring) in employment for the period 2010-14, distinguishing
among the different events of exit and reporting the difference between labour
outflows (replacement demand) and hiring (replacement demand plus net
employment variation, or expansion demand).
Consistent with the predicted dynamics of the labour-market stocks, the
expected evolution of labour hiring is basically explained by that of labour
outflows. In other terms, in the next four years hiring is predicted to be mostly
driven by replacement demand. Outflows for firing and end of temporary contract
are expected to increase in the next years, due to the increase in flexibility and
the weak macroeconomic performances (firing) and to the increasing fraction of
temporary workers in employment (end of temporary contract). Outflows for
retirement are expected to reduce until 2013 and then slightly increase. The initial
reduction is due to the ongoing pension reform for which, during the simulation
159
period, the eligibility conditions for retirement become more restrictive in terms of
age and number of years of contribution to the public pension funds. Outflows for
un-modelled events are expected to reduce due to the estimated negative trend
in the specific exit rate.
Figure 10:2 Labour market stocks and unemployment rate (2010:3-2015:3)

Millions
26 0.12
25 0.11
24 0.10
23 0.09
22 0.08
21 0.07
20 0.06
19 0.05
1985 1990 1995 2000 2005 2010 2015 1985 1990 1995 2000 2005 2010 2015
Labour force Employment Unemployement rate
Table 10:2 Predicted outflows and hiring (thousands), 2010-14

Firing ETC* Retirement Injury Other Total out Hiring Hiring out
2010 599 1 271 638 41 219 2 770 2 695 -74.6
2011 646 1 286 620 41 212 2 807 2 866 58.9
2012 709 1 328 606 41 197 2 883 2 964 81.3
2013 780 1 359 605 41 182 2 968 3 129 161.4
2014 862 1 378 613 41 168 3 064 3 225 161.7
* ETC — end of temporary contract.
Total outflows (replacement demand) is predicted to be between nearly

2.8 millions in 2010 and 3.1 millions in 2014, a value decisively above that implicit
to the Cedefop’s forecast for Italy for the period 2010-20 (Cedefop, 2010),
suggesting that the average annual replacement demand is of near 0.8 millions.
The reason for this difference is that Cedefop-RDMOD considers only permanent
outflows, such as retirement, deaths and migrations (Cedefop, 2010), while the
FGB-LM model considers also potentially transitory outflows, such as firings and
interruptions due to end of temporary contracts (73). Considering this major
(73) Outflows due to end of temporary contracts are not actual dismissals, since part of
these contracts are renewed within the period of observation (quarter). Under a more
160
difference, our forecasts for replacement demand turn basically in line with
Cedefop’s forecasts for replacement demand.
Hiring of new labour is given by replacement demand (equal to the number
of total outflows) plus the variation in employment required by the system
(expansion demand). Hiring is slightly below total outflows in 2010 and slightly
above in the remaining periods. The last column in Table 10:2 reports the
difference between hiring and outflows.
Interestingly, the employment stock variation is on average 3.5% of hiring,
thus nearly 96.5% of the job openings is due to replacement demand. The
relevance of replacement demand in determining the job openings highlights the
informative limitations of analyses targeted only to the evolution of the labour
market stocks, and thus the importance of extending the focus of modelling and
forecasting to labour market flows.
Figure 10:3 shows the cohort composition of the predicted employment
stock and of hiring in 2010:4 and five years later in 2014:4.
Figure 10:3 Employment and hiring by age class, 2010 and 2014
Millions
2010Q4 2014Q4 Thousands 2010Q4 2014Q4
4.5
180
4
160
3.5
140
3
120
2.5
100
2
1.5 80
1 60
0.5 40
0 20
-0.5 0
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
The histograms for the employment stock clearly show the presence of a
shift towards the right in the age composition of employment. This shift has two
major determinants: the ongoing modifications in eligibility condition for
retirement; and the effects of the 2008-09 crisis, that mostly affected younger
workers and workers on temporary labour contracts.
structural perspective, the inclusion of such an important component of outflows can

lead to an overvaluation of the replacement demand.
161
The graph for the cohort composition of hiring highlights an increase in job
openings for all the age classes, due in part to the increase in replacement
demand and in part to the moderate recovery in employment, thus to the increase
in expansion demand.
10.4.3. Employment stocks and flows by qualification

Tables 10:3 and 10:4 summarise the evolution of employment and hiring by
qualification, respectively, for the period 2010-14. The last row of the employment
stock table reports the percentage total change in the forecast interval, while that
of the table for hiring reports the average annual variation.
The 12 educational attainments of the Italian labour force survey
represented in the model are grouped into six classes for expositional
convenience, basically following the ISCED classification:
1) pre-primary and primary education;
2) lower secondary education;
3) upper secondary education;
4) tertiary education;
5) lower post-tertiary education (master, specialisation school);
6) upper post-tertiary education (PhD).
Table 10:3 Employment stock by qualification (thousands), 2010-14
Time Qualific. 1 Qualific. 2 Qualific. 3 Qualific. 4 Qualific. 5 Qualific. 6

2010 1437 7086 10146 3645 239 41
2011 1431 7058 10180 3698 242 41
2012 1425 7031 10225 3763 247 42
2013 1424 7021 10306 3847 253 43
2014 1422 7004 10387 3939 259 44
Δ% -1.0 -1.1 2.4 8.1 8.3 8.8
NB: Δ% indicates the percentage total change from 2010 to 2014.
For a more immediate comprehension of the employment performances of

the specific qualification groups, consider that the total employment variation
from 2010 to 2014 is predicted to be nearly 2%. This implies that the employment
change for the first two qualifications (pre-primary, primary and lower secondary
education) — a relevant fraction of the Italian employment stock (nearly 38%) —
is predicted to be negative in absolute terms and three percentage points lower
than the aggregate change. An increase only slightly above the total change in
aggregate employment is predicted for workers with upper secondary education,
while for those with tertiary education and over a more than 8% increase is
162
expected. Following these changes, the fraction of workers with a tertiary

education and over is expected to grow from 14.7% in 2010 to 15.6% in 2014.
The evolution of the employment stock for the different education levels is
reflected by hiring of new workers by qualification. The increase in replacement
and expansion demand determines an increase in hiring for all the qualifications,
even if it is heterogeneous among levels.
Table 10:4 Hiring by qualification (thousands), 2010-14

Time Qualific. 1 Qualific. 2 Qualific. 3 Qualific. 4 Qualific. 5 Qualific. 6
2010 105 730 1243 530 72 7.7
2011 112 777 1325 568 77 8.3
2012 115 801 1369 591 80 8.6
2013 121 842 1443 629 85 9.3
2014 125 864 1485 653 89 9.7
Δ% 4.4 5.3 4.6 5.5 5.6 6.3
NB: Δ% indicates the percent average annual variation in the period 2010-14.
As compared to an average annual growth rate for hiring close to 4.6%

(2011-14), the first three qualifications denote a lower dynamics, while the
dynamics of the fourth, fifth and sixth qualifications are above the aggregate
dynamics. Hiring of workers with a PhD degree (sixth qualification) requires an
average annual growth rate of 6.3%. In 2014, 9 700 top-qualified workers will be
required in the place of 7 700 thousand being required in 2010. This result has
serious policy implications as it requires important changes in the organisation
and financing of the higher education sector (74).
To highlight the differences emerging when considering hiring of new labour
in the place of the employment stock, Figure 10:4 displays the composition by
qualification of the employment stock and of hiring in 2014. Hiring shows lower
values than employment for lower qualifications and higher values for higher
qualifications, as a result of the fact that hiring is the nearest proxy for the
demand for new labour.
(74) In our simulation we are not considering the qualification upgrade that would hold
even in the presence of a stationary evolution of the composition among sectors of
the labour market.
163
Figure 10:4 Employment and hiring by qualification, 2014
Q3 Q4 Q4
46% 17% 20% Q1
Q5
Q3 4%
Other 1% Other
46%
Q1 1% 7%
Q5
6% 3%
Q2 Q6
Q2
0% Q6
30% 27%
0%
10.4.4. Employment stocks and flows by occupation

Tables 10:5 and 10:6 contain the predicted annual values of the employment
stock and of hiring by occupation (or professional profile). Even in this case, the
last row of the employment stock table reports the percentage total change in the
forecast interval, while that of the table for hiring reports the average annual
variation.
To simplify the exposition, the 37 professional profiles observed in the Italian
labour force survey and represented in the model are grouped into seven major
occupations, broadly following the ISCO classification:
1) legislators, senior officials, managers and entrepreneurs;
2) mathematical, health social and life sciences specialists and
professionals;
3) technicians and associate professionals;
4) clerks;
5) specialised workers in different fields;
6) machine operators and assemblers;
7) unqualified and elementary occupations.
Compared to the nearly 2% growth in total employment in the five years
period considered in the simulation, occupations 2, 3, 4 and 7 are expected to
grow above average, while occupations 1, 5 and 6 below average. Substantial
reductions are expected for machine operators (sixth occupation) and, to a lesser
extent, for specialised workers (fifth occupation).
This prediction is justified by the predicted sectoral dynamics of employment,
in which a further contraction of the industrial sectors (where the fifth and sixth
professional profiles are more important) and a clear expansion of the service
sectors (where the second, third and fourth professions are more represented)
are expected.
164
Table 10:5 Employment stock by occupation (thousands), 2010-14

Time Occup. 1 Occup. 2 Occup. 3 Occup. 4 Occup. 5 Occup. 6 Occup. 7
2010 991 2504 4935 2398 7442 1977 2348
2011 991 2547 4982 2416 7410 1945 2363
2012 992 2597 5039 2437 7377 1909 2383
2013 996 2662 5119 2468 7362 1873 2416
2014 999 2734 5204 2500 7338 1832 2451
Δ% 0.8 9.2 5.4 4.3 -1.4 -7.3 4.4
The predicted evolution of hiring by occupation confirms the heterogeneity

expected for the different professional profiles in the employment stock. In
particular, even if the average annual growth rate for the period 2011-14 is in this
case expected to be positive for all the job typologies (mostly because of the
expected growth in replacement demand), the annual growth rate is above the
4.6% average for hiring in job typologies 2, 3, 4 and 7, and below average for the
remaining typologies.
An above average increase is expected for hiring of unqualified/elementary
workers, mostly due to the expected recovery in the primary and, to a reduced
extent, in sales and household services.
Table 10:6 Hiring by occupation (thousands), 2010-14

Time Occup. 1 Occup. 2 Occup. 3 Occup. 4 Occup. 5 Occup. 6 Occup. 7
2010 118 297 586 285 889 236 279
2011 126 320 629 305 940 248 299
2012 129 337 655 317 965 251 310
2013 136 361 697 337 1010 259 329
2014 140 380 725 349 1031 259 342
Δ% 4.3 6.3 5.4 5.2 3.8 2.4 5.2
The relevant changes predicted for the composition of employment and, in

particular, of the hiring process highlight the utility of forecasting tools able to
represent these stocks and flows considering their major articulations. The ability
to foresee future changes in the composition by qualification and professional
profile can in fact provide valuable information for policy-makers interested in the
formulation and calibration of the necessary policy actions.
165
10.5. Conclusions
The paper has provided a general description of the forecasting model developed
by the Fondazione Giacomo Brodolini. We have described and motivated the
main distinctive characteristics of the model, that is composed of a microfounded
‘pilot’ NK-DSGE structural model and a ‘satellite’ second block representing the
detailed dynamics of the economy and of the labour market. Sticky wages, hiring
costs and involuntary unemployment characterise the labour market. The
discussion has addressed the technical choices underlying the estimation of
model parameters and has described the data sources used and model output.
We have subsequently presented the results of a recent application of the
model in a mid-term forecast up to 2015:3. To improve the readability of the
simulation results, a brief presentation of the expected macroeconomic
environment has preceded the discussion of the results for the labour market. To
simplify the exposition, the attention has mainly focused on the employment
stock and the hiring process, detailing their articulation into six major educational
attainments and seven major professional profiles.
The evidence provided for the macroeconomic environment has shown the
model ability to adapt to the data, and to provide internally consistent results. The
‘pilot’ model exhibits good estimation capacities which, together with the
theoretical transparency of results due to the full specification of the hypotheses
on behaviours and technology (the microfundation), hold out considerable
promise in forecasting and simulating the effects of both national and regional
labour-market policies.
The evidence provided for the stock and flow labour market variables has
exemplified some potential directions in the use of the FGB-LM model. These
range from the simulation of the labour-market effects of potential policy
interventions in the regulatory set-up, such as the modification of the
requirements for accessing the public pension system, to the forecasting of
qualification and professional needs associated to different scenarios, so as to
implement the most appropriate policy actions which are needed to meet the
future requirements of the labour market.
166
Abbreviation list
DSGE dynamic stochastic general equilibrium
FGB-LM model Fondazione Giacomo Brodolini labour market model
GDP gross domestic product
NK-DSGE new Keynesian dynamic stochastic general equilibrium
VECM vector error correction model
References
Bagnai, A. et al. (2006). FGB-STEP: un modello di simulazione per l'analisi del
mercato del lavoro [FGB-STEP: a simulation model for the analysis of labor
market]. Economia e lavoro, Vol. 40, No 3, p. 123-149.
Blanchard, O.; Galì, J. (2010). Labour markets and monetary policy: a new
Keynesian model with unemployment. American economic journal:
macroeconomics, Vol. 2, No 2, p. 1-30.
Cedefop (2010). Qualifications supply and demand in Europe. Medium-term
forecast up to 2020. Luxembourg: Publications Office of the European
Union.
Campbell, J.Y.; Mankiw, N.G. (1989). Consumption, income, and interest rates:
reinterpreting the time series evidence. In: Blanchard, O.J.; Fischer, S. (eds)
NBER Macroeconomics annual, Vol. 4, p. 185-216. Cambridge: MIT Press.
Campbell, J.Y.; Mankiw, N.G. (1990). Permanent income, current income, and
consumption. Journal of business and economic statistics, Vol. 8, No 3,
p. 265-279.
Campbell, J.Y.; Mankiw, N.G. (1991). The response of consumption to income: a
cross-country investigation. European economic review, Vol. 35, Issue 4,
p. 723-767.
Christoffel, K.; et al. (2009). The role of labour markets for Euro area monetary
policy. Frankfurt am Main: European Central Bank (ECB Working paper,
No 1035).
Coenen, G. et al. (2007). Tax reform and labour market performance in the Euro
area: a simulation-based analysis using the new area-wide model. Frankfurt
am Main: European Central Bank (ECB Working Paper, No 747).
Del Negro, M. et al. (2007). On the fit of new Keynesian models. Journal of
business and economic statistics, Vol. 25, No 2, p. 123-143.
Di Bartolomeo, G. et al. (2009). Monetary policy, rule-of-thumb consumers and
external habits: a G7 comparison. Applied economics, Vol. 43, Issue 21,
p. 2721-2738.
167
Erceg, C.J. et al. (2000). Optimal monetary policy with staggered wage and price
contracts. Journal of monetary economics, Vol. 46, No 2, p. 281-313.
Erceg, C.J. et al. (2005). SIGMA: a new open economy model for policy analysis.
Washington, D.C.: Board of Governors of the Federal Reserve System
(International finance discussion papers, No 835).
Favero, C.A. (1996). Econometria: modelli ed applicazioni in macroeconomia
[Econometrics: models and applications in macroeconomics]. Rome:
Carocci.
Forni, L. et al. (2007). The general equilibrium effects of fiscal policy. Rome:
Banca d’Italia [Italian Central Bank]. (Temi di discussione, Working Papers,
No 652) Banca d'Italia.
Giuli, F.; Tancioni, M. (2009). Il nuovo modello di previsione dei flussi del mercato
del lavoro FGB-MDL: aspetti di metodo e di struttura [the new FGB-MDL
forecasting model for labour market flows: aspects of method and structure].
Economia e lavoro, Vol. 43, No 3, p. 29-50.
Mortensen, D.T.; Pissarides, C. (1994). Job creation and job destruction in the
theory of unemployment. Review of economic studies, Vol. 61, No 3,
p. 397-415.
Pagan, A. (2003). Report on modelling and forecasting at the Bank of England.
Bank of England quarterly bullettin, Spring, Vol. 43, Issue 1, p. 60-88.
Pesaran, M.H.; Shin, Y. (1999). An autoregressive distributed lag modelling
approach to cointegration analysis. In: Strom, S. (ed.) Econometrics and
economic theory in the 20th century: the Ragnar Frisch centennial
symposium, chapter 11. Cambridge: Cambridge University Press.
Riggi, M.; Tancioni, M. (2010). Nominal v. real wage rigidities in new Keynesian
models with hiring costs: a Bayesian evaluation. Journal of economic
dynamics and control, Vol. 34, Issue 7, p. 1305-1324.
Smets, F.; Wouters, R. (2003). An estimated dynamic stochastic general
equilibrium model of the Euro area. Journal of the European economic
association, Vol. 1, No 5, p. 1123-1175.
Smets, F.; Wouters, R. (2007). Shocks and frictions in US business cycles: a
Bayesian DSGE approach. American economic review, Vol. 97, No 3,
p. 586-606.
Trigari, A. (2006). The role of search frictions and bargaining for inflation
dynamics. Milan: IGIER, Innocenzo Gasparini Institute for Economic
Research, Bocconi University (IGIER Working paper, No 304).
Trigari, A. (2009). Equilibrium unemployment, job flows and inflation dynamics.
Journal of money, credit and banking, Vol. 41, No 1, p. 1-33.
168
CHAPTER 11.
The forecast of occupational structure of
employment in Poland
Artur Gajdos
The main purpose of the following paper is comparative presentation of the

forecasts of occupational structure of employment in Poland conducted by
Cedefop and obtained within the labour demand forecasting system (LDFS)
in Poland. LDFS is an expert interactive system that allows analysis and
forecasting of time-series-cross-section economic variables, which is used
in the research and teaching process at the Chair of Spatial Econometrics
at the University of Lodz.
The study also presents suggestions for applying a time-series-cross-
section data analysis method in forecasting employment structure in Poland
in a regional cross-section (NUTS2). The results obtained indicate that
these methods can be used in the forecasting process. The use of location
quotient and shift-share analysis allows additional information, which may
be used in time-series-cross-section data forecasting in a territorial cross-
section. The analyses and forecasts on regional level are at present an
inherent element of European Union policy, including measuring cohesion
policy effectiveness.
11.1. Introduction
The study concentrates on two main problems. The first is forecasting
occupational employment structure at national level and comparing the forecast
results obtained within the project Skills supply and demand in Europe, and
forecasts made in the project Labour demand forecasting system in Poland. The
other is an analysis of possibility of using the methods of cross-section (as well
as spatial) data analysis in the process of occupational employment structure
forecasting. Structural analyses require the application of suitable data analysis
methods so that obtained results reflect global processes, but also include the
specificity of studied objects in cross-section and space. In this context it is
reasonable to use methods of time-series-cross-section analysis of statistical
data. The use of these methods makes it possible to include in the forecasting
process additional information specific for the studied objects, which in the case
of uniform treatment of the objects might be omitted.
169
11.2. Forecast comparison

The comparative analysis of the forecast results is to indicate the conclusions
which are essential from the point of view of forecasting process improvement
and which enable the modification of the applied methods to achieve better
forecast results in the future.
The first step was the comparison of forecast results made in Poland in 2006
with the Cedefop forecasts from 2010 (Cedefop, 2010), and the actual data
(labour force survey 2010). Then the forecast results made by Cedefop for 2020
were discussed and compared with employment structure forecasts for 2011-15.
The forecasts generated in 2006 within the labour demand forecasting
system (LDFS) in Poland were conducted in the cross-section of major
occupational groups with the use of data from labour force survey from 1995-
2005. The forecasts included employment cross-section (submajor and minor
groups), education level and economic sector. The number of the employed in
particular cross-sections was a forecast variable. The project included building an
interactive analytical system — LDFS. This is a combination of database,
analytical software and internet interface that enables online data analysis.
The system uses MS SQL Server database and Matlab software. Work in
the system is based on making analyses (forecasts) using web browser interface.
All calculations are server-side, where the database and analytical software are
placed. The analyses results are recorded in the database and may be published
as a website.
At present the system works with macroeconomic data and data from the
labour force survey. The system enables data analysis in the form of time-series-
cross-section using methods of time series analysis (linear and non-linear trend
models, seasonality, exponential smoothing, single- and multi-equation models,
autoregression models and vector-autoregression models). The analyses results
are presented as tables and graphs. It is also possible to generate combined
forecasts (weighted average from individual forecasts).
The expansion of the system is planned, which will enable analysing data
from other sources (e.g. Eurostat database, OECD), the use of spatial data
analysis methods, and presentation of data in the form of maps and dynamic
data visualisation.
To compare the forecast results for 2010 and evaluate the forecast errors,
data published by Eurostat from the labour force survey in major occupational
groups cross-section, forecasts published by Cedefop (2010), and forecasts
made in 2006 in the LDFS project were used. Nominal differences between
observed and predicted values of structure indicators for the major occupational
groups were analysed.
170
Figure 11:1 LDFS — Structure
DATA BASE
DATA BASE ANALYTICAL TOOL
ANALYTICAL TOOL
DATA BASE DATA BASE ANALYTICAL TOOL
ANALYTICAL TOOL
Database DATA BASE Analytical tool
ANALYTICAL TOOL
DATA BASE ANALYTICAL TOOL
DATA DATA DATA

DATA
DATA
DATAData
QUESTION - RESULT
QUESTION - RESULT
SERVER WWW INFORMATION

QUESTION - RESULT
QUESTION
QUESTION
SERVER - RESULT
WWW - RESULT INFORMATION
Question – Result
QUESTION - RESULT
Server
SERVERWWW
WWW
SERVER WWW
INFORMATION Information
INFORMATION
Source: Gajdos (2004, p.189).
Figure 11:2 LDFS — Database
Source: http://www.sap.uni.lodz.pl (10.2.2011).
171
Figure 11:3 LDFS — Forecasting
Source: http://www.sap.uni.lodz.pl (10.2.2011).
Figure 11:4 Forecast errors (a) in 2010 (major ISCO groups)
Cedefop (5.5 percentage points) LDFS (5.4 percentage points)

1.5%
1.0%
0.5%
0.0%
-0.5%
-1.0%
-1.5%
-2.0%
1 2 3 4 5 6 7 8 9
NB: 1-9 = major ISCO groups.

a
( ) Percentage error: PE= [(actual value – forecast)/actual value]*100%.
Source: Own calculations based on Eurostat, Cedefop and LDFS data.
The highest nominal error values were obtained for groups characterised by
the greatest change dynamics between 2000 and 2010. Especially high
deviations of predicted values from the observed ones were obtained for the
172
groups of: professionals (2), service workers and shop and market sales workers
(5), and skilled agricultural and fishery workers (6).
The sum of absolute errors in all analysed major occupational groups for
Cedefop and LDFS forecasts is similar and is about 5.5 percentage points. This
indicates a similar forecasting precision in the analysed cases.
In the following year an attempt was made to discuss the forecast results
generated by Cedefop for 2020 compared to the forecast of occupational
structure of employment in Poland made for the years 2011-15 in LDFS.
The analysis of historical data for 2000-10 made it possible to generate,
using a segmented linear trend model, the value of the share of particular major
occupational groups between 2011 and 2015. The confrontation of the obtained
results with the Cedefop forecasts indicates major occupational groups, for which
the forecasts largely differ from the trends observed between 2000 and 2010 and
the values predicted in 2011-15.
The most essential differences were observed in the major occupational
group of professionals (2) and clerks (4). In the case of professionals the
predicted share in the number of the employed in 2020 is close to the value
observed in 2010. Due to the importance of this group in the development of
modern economy, it should be determined whether a setback of the observed
upward tendency for 2000-10 may be expected. In the case of clerks, the
forecast for 2020 indicates a high share of this group, much higher than the level
observed for 2000-10.
When organising a forecasting process, particular attention should be paid to
the occupational groups which undergo the most dynamic changes and the share
of which significantly influences the economical development dynamics, and
which were characterised by the greatest forecasting errors.
11.3. Regional employment structure analysis (NUTS2)

Between 2000 and 2010 significant differences in the dynamics of changes in the
total number of the employed according to provinces (regions) were observed in
Poland. Moreover, important differences in the employment structure by region
and various directions of this structure changes were identified. Therefore, an
attempt to describe regional employment structure was made. On the basis of the
share rate analysis it is possible to indicate the regions where the occupational
employment structure is similar to the national one: Malopolska province,
Pomeranian province, Wielkopolska province, and those where the structure is
significantly different: the Lublin region and Silesia.
173
Figure 11:5 Cedefop and LDFS — Forecast comparison
Legislator, senior officials and managers Professionals Technicians and associate professionals
8% 18% 13%
16.8%
7% 12%
7% 16% 12%
6% 14% 11%
5% 12% 10%
2000 2005 2010 2015 2020 2000 2005 2010 2015 2020 2000 2005 2010 2015 2020
Clerks Service workers and Service workers and

shop and market sales workers shop and market sales workers
8.5% 13% 20%
8.5% 12.9%
12% 15%
7.5%
11% 10%
6.5% 8.2%
10% 5%
2000 2005 2010 2015 2020 2000 2005 2010 2015 2020 2000 2005 2010 2015 2020
Craft and related trades workers Plant and machine operators and assemblers Elementary occupations
18% 12% 9%
8.2%
16.3% 10.2%
16% 10% 8%
14% 8% 7%
2000 2005 2010 2015 2020 2000 2005 2010 2015 2020 2000 2005 2010 2015 2020
2000-10 2011-15 2020
Source: Own calculations based on Eurostat, Cedefop and LDFS data.
One can also observe the structure change processes in various directions
leading to a greater and greater structure difference. This study confirmed the
thesis about the lack of convergence process in the occupational employment
structure. Only a major occupational ISCO group of legislators, senior officials
and managers (ISCO 1), craft and related trades workers (ISCO 7) and plant and
machine operators and assemblers (ISCO 8) show between 2001 and 2008 a
smaller regional differentiation. On the other hand, a major occupational group of
skilled agricultural and fishery workers (ISCO 6) shows between 2000 and 2008
a growth of structure differentiation by region. In most other major occupational
groups a relatively stable structure differentiation level is observed between
regions or differentiation changes have different directions in selected
subperiods.
174
Figure 11:6 Sigma-convergence, 2000-08
25% 20% 25%
20%
20% 15%
15%
15% 10% 10%

2000 2002 2004 2006 2008 2000 2002 2004 2006 2008 2000 2002 2004 2006 2008
Service workers and Skilled agricultural and fishery workers

Clerks
shop and market sales workers
25% 20% 65%
20% 60%
15%
15% 55%
10% 10% 50%

2000 2002 2004 2006 2008 2000 2002 2004 2006 2008 2000 2002 2004 2006 2008
25% 30% 20%
25%
20% 15%
20%
15% 15% 10%

2000 2002 2004 2006 2008 2000 2002 2004 2006 2008 2000 2002 2004 2006 2008
Source: Own calculations based on labour force survey data.
On the basis of regional analysis of occupational structure one may observe

a spatial concentration of selected occupational groups. A major occupational
group of skilled agricultural and fishery workers (ISCO 6) is concentrated in the
eastern part of the country, elementary occupations (ISCO 9) in the north, while
plant and machine operators and assemblers (ISCO 8) in the south-west.
Another example of applying a time-series-cross-section data analysis
method in the process of describing space and structure is the use of location
quotients which enables defining a cross-section variable localisation and spatial
object specialisation.
The location quotients analysis in 2008 points out major occupational groups
which are not localised (have a similar share in the labour market in most
regions), for example clerks, and highly localised groups, for example skilled
agricultural and fishery workers.
One may also point out regions which have occupational structure of
employment similar to the national one, for example Wielkopolska region, and
those where the structure is significantly different, for example Silesia and the
Lublin region.
175
Figure 11:7 Major occupational groups — Spatial concentration in 2008
Legislators,
Legislators, senior
senior
Legislators, senior
Legislators,
officials
officials
officials
senior and
and
officials and Professionals
Professionals
Professionals
Technicians
Technicians
Technicians
and
and
Technicians andassociate
associate
associate
and associate
Professionals
and managers
managers
managers
managers professionals
professionals
professionals
professionals
Legislators,
Legislators, senior
senior officials
officials and
and Professionals
Professionals Technicians
Technicians and
and associate
associate
Legislators, senior officials and
managers
managers Professionals Technicians and associate
professionals
professionals
managers
Legislators, senior officials and Professionals professionals
Technicians and associate
managers
Legislators,
Legislators, senior
senior officials
officials andand Professionals
Professionals professionals
Technicians
Technicians andand associate
associate
managers
managers professionals
professionals
Clerks
Clerks
Clerks Service
Service
Serviceworkers
workers
workers and
and shop
shop
and and
and
shop Skilled
and Skilled agricultural
agricultural
Skilled and
agricultural and and
fisheryfishery
fishery
market
market
market sales
sales workers
workers
sales workers workers
workers
workers
Clerks
Service
Service
workers
workers and
and
shop
shop
and Skilled
agricultural and
and
fishery
Clerks
ClerksClerks Service workers
Service and
workers shop
and and
shop andSkilled agricultural
Skilled and
agricultural fishery
and fishery
and market
market
market sales
sales
sales workers
workers
workers fishery workers
workers
workers
market sales workers workers
Clerks Service workers and shop and Skilled agricultural and fishery
Clerks Service workers and shop
andand Skilled
Clerks Service workers
market sales and shop
workers workersandand
agricultural fishery
fishery
market
market sales
sales workers
workers workers
workers
Craft and related trades Plant and machine operators

Craft and
Craft andrelated trades
related trades Plant and
Plant machine
and operators
machine operators
workers and assemblers Elementary occupations
workers
workers and assemblers
and assemblers Elementary occupations
Craft Craft
and
Craft and and
relatedrelated
relatedtrades trades
trades PlantPlant
and
Plant andand machine
machine operators
operators
machine operators
workersworkers
workers and
andand assemblers
assemblers
assemblers Elementary
occupations
Craft and related Plant and machine
trades
Craft and
Craft
Craft and workers
related
and trades
related
related trades
trades operators
Plant and
Plant
Plant andandand assemblers
machine operators
machine
machine operators
operators
workers
workers
workers and assemblers
andand assemblers
assemblers Elementary occupations
Elementary
occupations
Other methods to be used for comparing structures may be coefficients of

structure similarity in statical analysis or shift-share analysis, which enables the
separation of structural and geographical factor in dynamic analysis of structural
changes (Gajdos, 2011).
The analysis of data from shift-share analysis table enables the indication of
regions characterised by change dynamics of the total number of the employed
much different from the national average, for example North Pomeranian region,
the Mazowsze region, and those where the dynamics is close to the national one,
for example Warmia and Mazury province and the Lodz region. It is also possible
to distinguish a structural and geographical effect.
176
Table 11:1 Location quotients in 2008

Major occupational groups
NUTS2
1 2 3 4 5 6 7 8 9
Lower Silesia 0.86 0.89 1.06 1.13 1.06 0.40 1.27 1.31 0.97
Kujawy-Pomerania 0.99 0.72 0.86 1.00 0.94 1.28 1.01 1.03 1.35
Lublin 0.79 0.96 0.93 0.75 0.78 2.25 0.73 0.66 1.02
Lubuskie 0.83 1.00 1.13 0.75 1.00 0.43 1.20 1.39 1.08
Lodz 1.11 0.97 0.83 1.07 0.88 1.21 0.86 1.19 1.05
Malopolska 1.22 1.04 0.79 0.97 0.96 1.20 0.96 0.93 1.01
Mazovia 1.10 1.29 1.15 1.19 0.98 0.96 0.73 0.82 0.87
Opole 0.91 0.94 1.07 0.89 1.04 0.66 1.19 1.18 1.02
Podkarpacie 0.82 0.85 0.96 0.80 0.95 1.83 0.84 1.04 0.75
Podlasie 0.67 0.90 1.05 0.93 1.09 1.71 0.70 0.90 1.00
Pomerania n 1.03 0.98 1.17 1.10 1.09 0.41 1.19 0.94 1.14
Silesia 1.12 1.09 1.09 1.14 1.20 0.16 1.23 1.11 0.86
Swietokrzyskie 0.85 0.87 0.79 0.64 0.85 1.99 1.01 0.81 0.94
Warmia-Masuria 0.60 0.93 1.04 0.71 1.11 0.82 1.20 1.10 1.23
Wielkopolska 1.19 0.79 0.85 1.01 1.02 1.00 1.17 1.04 1.03
West Pomerania 0.94 1.16 1.29 0.94 1.04 0.31 1.15 0.85 1.29
Source: Own calculations based on LFS data.
Table 11:2 Shift-share analysis in 2004-08 (major ISCO groups)
Change Total result Structural Geographical

NUTS2
2004-08 (region – Poland) result result
Lower Silesia 18.3 4.2 3.9 0.3

Kujawy-Pomerania 5.2 -8.9 1.2 -10.0
Lublin 9.4 -4.7 -6.6 1.9
Lubuskie 5.0 -9.1 3.4 -12.5
Lodz 16.4 2.4 -0.3 2.7
Malopolska 6.8 -7.3 -2.3 -4.9
Mazovia 23.7 9.7 -1.2 10.9
Opole 19.5 5.4 0.4 5.0
Podkarpacie 17.8 3.8 -4.5 8.3
Podlasie 21.1 7.0 -6.6 13.6
Pomerania 22.5 8.5 2.6 5.8
Silesia 10.9 -3.1 4.8 -7.9
Swietokrzyskie 21.5 7.5 -5.6 13.1
Warmia-Masuria 14.4 0.4 1.5 -1.2
Wielkopolska 6.8 -7.2 2.0 -9.2
West Pomerania 1.7 -12.3 2.7 -15.0
Poland 14.0
177
The information obtained in the process of time-series-cross-section (spatial)

data analysis may be a complementary source of specific, additional information,
the use of which in the forecasting process may lead to getting better forecast
results. It is possible to apply these methods in the regional structures forecasting
process (national), as well as the European level (international).
11.4. The forecast of regional occupational structure of

employment in Poland
Test results of occupational structure forecasts in the cross-section of major
occupational groups in selected provinces are presented below. The Lublin
province is a region with the employment structure is significantly different from
the rest of the country. Wielkopolska region is characterised by employment
structure similar to the national one.
Figure 11:8 Forecast for major occupational groups — Lublin region
7% 18% 12%
5% 14% 9%
3% 10% 6%
2000 2004 2008 2012 2000 2004 2008 2012 2000 2004 2008 2012
Service workers and Skilled agricultural and fishery workers

Clerks
8% 14% 40%
6% 11% 25%
4% 8% 10%
2000 2004 2008 2012 2000 2004 2008 2012 2000 2004 2008 2012
9% Elementary occupations
Craft and related trades workers Plant and machine operators and assemblers
20% 12% 9%
7%
15% 8% 7%
5%
10% 4% 5%
2000 2004 2008 2012 2000 2004 2008 2012 2000 2004 2008 2012
Poland Lublin region
Source: Own calculations (LDSF) based on labour force survey data.
178
Figure 11:9 Forecast for major occupational groups — Wielkopolska region
9% 18% 14%
7% 14% 11%
5% 10% 8%
2000 2004 2008 2012 2000 2004 2008 2012 2000 2004 2008 2012
Service workers and

Clerks Skilled agricultural and fishery workers
8% 14% 20%
7% 11%
14%
6% 8%
8%
2000 2004 2008 2012 2000 2004 2008 2012
2000 2004 2008 2012
21% 12% 10%
18% 10% 8%
15% 8% 6%
2000 2004 2008 2012 2000 2004 2008 2012 2000 2004 2008 2012
Poland Wielkopolska region
Source: Own calculations (LDFS) based on labour force survey data.
11.5. Conclusion
The paper identifies important areas concerning improvement of forecasting
process quality. It indicates the occupational groups, for which the forecasting
must be done in a particularly careful way due to their importance in the
economical transformation process (professionals, service workers and
agricultural workers).
The paper presents suggestions for using time-series-cross-section data
analysis methods for forecasting employment structure in Poland by regions
(NUTS2).
The obtained results indicate that these methods can be used in the
forecasting process. The application of location quotients and shift-share analysis
allows additional information which may be used for time-series-cross-section
data forecasting in the territorial cross-section. The analyses and forecasts on
179
regional level are at present an inherent element of the European Union policy,
including measuring the effectiveness of cohesion policy.
References
Gajdos, A. (2004). System Prognozowania Popytu na Pracę w Polsce – wersja
testowa, w: SPPP, Prognozy popytu na pracę w Polsce [Labour demand
forecasting system in Poland – the test version, in LDFS: Labour demand
forecast in Poland], part III. Suchecki, B. (ed.). Warsaw: RCSS, p. 188-204.
Gajdos, A. (2011). Time-series-cross-section analysis of employment structure
according to major occupational groups in Poland. Acta Universitatis
Lodziensis Folia Oeconomica (forthcoming).
180
CHAPTER 12.
Forecasting the development of the
educational level in Switzerland
Jacques Babel
The Swiss Federal Statistical Office (FSO) produces annual forecasts for
the entire education system. These would in principle constitute an ideal
foundation for the scenarios for the educational level, scenarios for which a
new model has just been developed. This stock-flow model also
incorporates the international flows of human capital, a phenomenon that
should not be neglected in Switzerland.
More than the reliability of the quantitative results of the scenarios for the
educational level, detailed analyses show that it is their interpretation that
may present some difficulties owing to inconsistencies between the
development of the educational level and the flows of new graduates. The
political importance of accurately predicting the future development of the
skills supply and thus the educational level means that it is essential to
ensure that the scenarios for the educational level are consistent with the
scenarios for the numbers of pupils and students.
12.1. Overview of the FSO’s education forecasts

In the education sector, forecasts play an important role in planning for the
education system (budget, etc.) and the first works on this subject in Switzerland
date from the 1970s. For 10 years, these have clearly assumed greater
importance, with 10-year forecasts being made annually for the entire education
system. This forecasting is carried out to very detailed degrees of (regional or
institutional) granularity, such as for each university in Switzerland and
approximately 30 educational sectors (FSO, 2010a; 2010b; 2011a).
The flows of the education system are represented in greater detail in the
model. These flows account as fully as possible for mobility, such as between or
within universities, between or within qualifications or various transition periods.
The modelling output is produced in the form of rates of transition between many
states.
Many outputs are provided by the project, whether in relation to numbers of
pupils or students, numbers of qualifications, information about human resources
or forward-looking indicators.
Some hypotheses are developed to account for structural phenomena,
reforms in progress, upgrading of courses or observed trends and flows of
181
students from abroad. The quality of the forecasts is subjected to permanent

monitoring.
One important consequence of these works is a detailed knowledge of the
dynamics of the Swiss education system. The most important result for works on
society’s future skills requirement is a reliable forecast of the future number of
graduates from the Swiss education system.
12.2. Educational level forecasts — Model

The scenarios for the educational level have so far formed part of the scenarios
for the development of the Swiss population (FSO, 2010d). In these works,
demographic scenarios provide input to the works on the scenarios for the
educational level. These are then used to determine the future active population
of Switzerland for the next 50 years.
In 2009, the forecasting model was completely revised (FSO, 2009). The
current form is a multistate hierarchical model (see, e.g. Lutz et al., 1999). This
means that in each state a probability of transition to another state is determined.
The model is hierarchical in the sense that the educational level can only
progress. The states or strata are age, sex, nationality (Swiss or foreign,
considering naturalisations via probabilities of transitions) and the highest
educational level. For the latter, four levels are being considered from this year,
namely levels ISCED 1 (low qualification), 3-4 (medium qualification), 5A (high
qualification: universities) and 5B (high qualification: professional education and
training). The modelling is fully consistent with the demographic scenarios. Some
simplifying assumptions are nevertheless made concerning two minor
phenomena in Switzerland as concerns 25-64 year-olds: it is implicitly assumed
that mortality for these age groups does not depend on the educational level, as
with the immigration or emigration of Swiss nationals.
These works have two further characteristics:
(a) they aim to account as precisely as possible for migratory flows (Section
12.2.1);
(b) although it is a stock-flow model, both the stock and the flow are determined
using the Swiss labour force survey (SLFS) (FSO, 2004). This ensures that
the model is internally consistent (Section 12.3).
12.2.1. Accounting for migration flows by educational level

Particular emphasis is placed on the appropriate treatment of migration, as this
has played an important role in Switzerland for many years. Since 2003, a
substantial, highly qualified immigration has been observed in this country.
182
Between 2003 and 2009, the migratory balance averaged 65 000 with, according
to the SLFS, 60% of immigrants holding a tertiary-level diploma; that is, distinctly
higher proportions than for the 25-64-year-old Swiss population. In contrast, the
proportion of immigrants without post-compulsory education was 12%, a distinctly
higher level than for the Swiss population of the same age group (75).
Another reason for taking full account of migration is the fact that these
forecasts are also designed to describe the future population of Switzerland and
therefore also the composition of its foreign population.
For immigration, the available data sources are the demographic records, as
well as the SLFS, for the educational composition of the population.
There is no quantitative information available concerning emigration. The
method used to work out the age composition for the educational level of
emigrants therefore consists in comparing the observed development of the age
composition of the educational level of Switzerland’s foreign population between
2003 (baseline for the model) and 2009 with that predicted by the model for
various hypotheses. A very good adjustment is obtained by simply assuming that
the educational level by emigrants’ age is the same as that of immigrants for
those aged 50 years and under (FSO, 2009). This adjustment is much better than
if the educational level of emigrants is assumed to be the same as that of the
foreign population of the same age. These results correspond with analyses
made, for example, by taking emigrants’ origin as a basis. They also demonstrate
that the highly qualified population currently entering Switzerland is mobile and it
is also the population that most commonly emigrates.
These analyses also enable us to assess the ‘brain-gain’ in Switzerland in
recent years and also in some sense to determine how the demand in the labour
market for highly qualified workers has been partly met by immigration. It is found
that the ‘migratory balance’ of tertiary-level graduates between 2004 and 2009
stood at 28 000 per year (48 000 arrivals versus 20 000 departures). This figure
is highly significant when compared to the approximately 50 000 tertiary-level
diplomas awarded annually in Switzerland by the universities or advanced
professional training programmes.
This importation of human capital also has had a significant effect on the
educational level of Switzerland’s foreign population (in six years, a nine-point
increase in the proportion of tertiary-level graduates). The proportion of tertiary-
(75) The SLFS gives also information on the nationality of the immigrants. It indicates
larges differences, with percentages of immigrants with tertiary education ranging
from around 70% for instance for German or French citizens, to values around 15%
for citizens of several countries (the sample size does not permit to go much in
detail).
183
level graduates in this population is currently reaching the same level as for
Swiss nationals, although the probability of foreigners having attended school in
Switzerland and having reached tertiary level is distinctly lower than that of the
Swiss (FSO, 2010d). The effect of the ‘brain-gain’ on the proportion of tertiary-
level graduates in the entire Swiss population is not insignificant either, ranging
from 1-1.5 points over the period under consideration. According to Sheldon and
Straubhaar (2008), this immigration has substantially contributed to the increase
in productivity and growth in recent years.
12.2.2. Main results of the scenarios for the educational level

The results of the three scenarios considered by FSO for the educational level
are described in detail in FSO (2010d). The ‘high’ scenario assumes an ever-
increasing transition to tertiary level (universities or higher vocational training), a
continuation in the increase in the educational level of the immigrant population
and a high hypothesis on the current rates of transition to tertiary level for
foreigners with a gradual reduction in the gap between Swiss nationals and
foreigners with respect to the probability of obtaining a diploma at upper
secondary or tertiary level. The reference scenario (medium) is based on much
more modest hypotheses that are relatively close to a status quo in terms of
education. It relies chiefly on the assumption that the upgrading process from
upper secondary to tertiary level, which several training and education
programmes in teaching, health, social work and the arts have undergone, is
coming to an end in Switzerland. To set the baseline rates of transition to tertiary
level, the ‘low’ scenario considers first the uncertainty about the actual relevance
of the number of new graduates measured in the labour force survey. It supposes
a decrease by 25% of the probability of entering the tertiary level at any age in
comparison with the baseline levels adopted in the other scenarios, a very slight
decrease in the probability of obtaining an upper secondary diploma and a
gradual decline over time of the probability of obtaining a tertiary level diploma, a
‘low’ hypothesis on the probability for the foreign resident population to obtain a
tertiary level degree and a decline in the educational level of the immigrant
population.
For the Swiss population as a whole (Figure 12:1), a very strong increase in
the educational level is anticipated for the coming years. According to the
reference scenario, the proportion of those holding a tertiary-level qualification in
the 25-64-year-old population is expected to rise from 35% to 46% in 2020 and
from this date to exceed the proportion with an upper secondary-level
qualification. The foreign population should witness a similar increase in its
proportion of tertiary-level graduates to that of Swiss nationals. The proportion of
184
those without post-compulsory education among foreigners is also set to decline.

However, as a result of the lower probabilities for foreigners than the Swiss of
obtaining an upper secondary level or tertiary-level diploma, a base level of 15%
of people with no post-compulsory education in this population may subsist over
the medium to long term.
Figure 12:1 2010-60 scenarios for the population development: level of education
of population aged 25 to 64 in Switzerland
80%
◄ Forecasts ► Observations
Without post-compulsory education
70%
Upper secondary level
Tertiary level
60%
Scenario A-00-2010 'm edium '
50% Without post-compulsory education
Tertiary level
40%
Scenario B-00-2010 'high'
Tertiary level
20%
Scenario C-00-2010 'low '

0% Tertiary level
1990 2000 2010 2020 2030 2040 2050 2060
NB: Forecasts: as in July 2010.

Source: FSO, scenarios for the population development and education prospects.
12.2.3. Comparing the FSO and the Cedefop (2010) forecasts

A comparison between the population’s educational level according to the FSO
(2010d) and Cedefop (2010) forecasts is made in Figure 12:2. The overall results
for 2020 are very similar, differing by only 1 percentage point (for an 11-point
increase during the same period) (76). In the age composition some major
differences are observed. While good agreements are obtained for some age
groups, Cedefop (2010) clearly anticipates more graduates among 25-29-year-
olds (10 points more than FSO, 2010d) and distinctly fewer among 40-44-year-
olds (10 points less). This difference may originate from the fact that the ‘flow’ of
new graduates used by Cedefop (2010) is not completely consistent with the
stock, probably with an overestimate by Cedefop (2010) of the proportion of
(76) This difference is not really higher than that recorded for 2009 and is probably
attributable to slightly different definitions.
185
tertiary-level diplomas issued in Switzerland to those of 30 years of age and

under (Figure 12:1). A similar result is obtained for those without post-compulsory
education, with a very good agreement on the overall proportions but major
differences in the age distributions.
Figure 12:2 Percentage of high qualification in the Swiss population, by age
%
60
Cedefop (2010)
2009
50
2020
40 FSO (2010)
2009
2020 (reference scenario)
30
20
10
0
25-64 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 Age
Source: FSO, education prospects.
12.3. What are the future flows of new graduates?

The correspondence between the FSO (2010a-d) and the Cedefop (2010)
forecasts, as well as the extensive preliminary testing, show that neither the
forecasting model nor the quality of the projections for the educational level
should pose real problems in themselves. Nevertheless, the results of these
forecasts raise some questions, in the sense that the high proportion of tertiary-
level graduates anticipated for 2020 for the 25-64-year-old population seems at
first sight to contradict the picture obtained by examining certain indicators, such
as the rate of new tertiary-level graduates in the population, an indicator that
currently stands at approximately 50% (FSO, 2011b).
This possible discrepancy, thus, raises several questions concerning the
true relevance of the observed educational level, as obtained from responses to
the SLFS interviews. Is the development of the educational level (stock) truly
compatible with the numbers of new graduates (flow)? How does the educational
level that people state that they have compare with the level that we think they
have (according to standard definitions)? What exactly do we mean when we
186
refer to educational level? These questions are not purely methodological

because such indicators may form the basis for policy decision-making
processes.
The only way to answer these questions is to adopt a common frame of
reference and to make a detailed comparison of the flows of new graduates
between the various sources. These flows can be obtained specifically in the
three following ways:
(a) SLFS-T. Using macroscopic data from the SLFS (FSO, 2009): this method
consists of ‘differentiating’ the ‘stock’ for each category of age, sex or
nationality (Swiss or foreign). It, therefore, entails postulating that for a
defined population, the development in the number of graduates between
two consecutive years provides a good estimate of the number of new
graduates in the group analysed. It is of interest because where this
measure is possible, it provides by definition distributions of new graduates
that are consistent with the development of the educational level. However, it
presents many difficulties. To prevent possible bias, the populations
analysed must be so defined as to eliminate the migratory or naturalisation
effects. Also, this method is only applicable for few study levels or transitions
(three educational attainments and two transitions in FSO, 2009);
(b) SLFS-L. Using the SLFS panel data: this method consists of drawing on the
SLFS panel, i.e. on the longitudinal data, and in measuring the numbers of
people who report having made a change or not in higher education
completed between consecutive years. It has many advantages over the
previous method, such as its capacity to consider many study levels or
transitions. It may, however, be susceptible to possible effects of the ‘design’
of the panel, for example where the method of re-interviewing people in the
sample group would cause possible bias;
(c) EduSys. Using data from the education system: This method, which
provides an objective measure of the population’s education, is at first sight
the simplest. It consists of considering all the new graduates of the
education system using the detailed administrative statistics gathered in the
country. In Switzerland, this flow of new graduates cannot truly be
determined with great accuracy owing to the possibility of double-counting
and sometimes deficient information about age distributions for certain types
of diploma. The possibility of small problems of coverage also cannot be
excluded.
In the following sections, comparisons are made for different transitions.

They are restricted to the population of Swiss nationality, a category for which
good-quality comparisons can be made. Several processing stages are
187
necessary so that the various sources can be compared on a common basis:

transformation, calculated or estimated, of the sources into a single period of
reference (using the dates of the SLFS interviews as reference), estimates of
double-counting (for the upper secondary level and advanced professional
training), consistent age definitions and sometimes estimates of the proportions
of Swiss nationals. To limit statistical variations, comparisons are made by
pooling data on the transitions made between pairs of successive years from the
years 2006 to 2008.
12.3.1.1. From low to medium qualification

Figure 12:3 presents the numbers of first upper secondary level diplomas
obtained, i.e. the transitions made between the low and medium qualification
levels according to the Cedefop (2010) categories. The three methods used
provide very close results. However, a slightly lower value seems to be obtained
for the measurements made using the SLFS panel group for those aged 20 years
and over.
Despite these excellent agreements, an uncertainty of around three
percentage points remains as regards Swiss nationals who do not obtain any
post-compulsory diploma.
Figure 12:3 Number of degrees awarded in the upper secondary programmes

(ISCED 3-4, medium qualification)
Num ber of degrees

Total 2006-08 sw iss national population
80000
70000
60000
50000
40000
30000
20000
10000
-10000
15 17 19 21 23 25 27 29 Age
SLFS(L): Longitudinal micro (N=485) SLFS(T) : Longitudinal macro EduSys: education system data
188
12.3.1.2. To high qualifications (universities, ISCED 5A/6)

Given that the SLFS-T method can only be used for the tertiary level as a whole,
the comparisons made in this and the following paragraph only concern the
results according to SLFS-L and EduSys.
As concerns the Swiss institutions of higher education, the comparison is
clearly sounder than for the transitions to the upper secondary level. In fact, the
data for the university system allow all the double-counting to be completely
removed and require no estimate for the determination according to EduSys. An
excellent agreement between methods is obtained concerning the number of
transitions between the upper secondary level and the institutions of higher
education (Figure 12:4). The total numbers only differ by 9% and the age
distributions are very similar.
More surprising is the result obtained for transitions between advanced
professional education training (5B) and institutions of higher education (5A).
While the creation of the institutions of higher education of applied sciences in
the late 1990s is a phenomenon that has resulted in approximately 1 000
diplomas per year being recognised retrospectively and approximately 1 300 new
diplomas per year being awarded in institutions of higher education to those who
have previously completed professional education training studies, SLFS-L
indicates that a five times greater number, namely approximately 10 000 people
per year, could have made this transition. Given that the statistics only include
continuing education programmes lasting one year full-time (60 ECTS or more),
one possible interpretation is that the divergence arises from short continuing
education programmes. (77)
12.3.1.3. To high qualifications (ISCED 5B, professional education and training)

The advanced professional training sector (ISCED 5B, professional education
and training) is very important in Switzerland because it represents about half of
the tertiary-level diplomas issued every year. It is also a highly complex sector
because it offers a very broad range of courses, and it is difficult to work out
using the education statistics. The situation is also complicated by the fact that
certain variables are not always available and recourse has to be made to
estimates, especially as concerns age distributions (see also FSO, 2011b). The
comparison below, Figure 12:5, between the various sources shows a fairly good
agreement for people aged less than 30 years. Above these ages, more people
are clearly recorded reporting in the SLFS having obtained a diploma of this type
than are measured in the official education statistics.
(77) Recent computations based on the new SLFS and the 2010-11 data seem to
indicate a better agreement for institutions of higher education than depicted here.
189
Figure 12:4 Number of first degrees awarded in the Swiss universities (ISCED 5A-6,
high qualifications)
Num ber of first degrees

Total 2006-08, sw iss nationality, moving average (5)
10000
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
15 20 25 30 35 40 45 50 55 60 65 Age
SLFS(L) : from sec. II to 5A (Total: 79 192, N=196)

Edusys (eucation system data): from sec. II to 5a (Total/N: 72 492)
SLFS(L) : from 5B to 5A (Total: 29 772, N=96)
In total, the number of diplomas differs by 40% and the age distributions are
very clearly different. The origin of the differences is currently unknown. There
are several possible hypotheses. Given that it seems improbable that there
should be a large lacuna in the education statistics concerning courses of a
longer duration than one year, one hypothesis may be that the people
interviewed in the SLFS are reporting continuing education programmes of short
duration in the SLFS.
12.4. These results obtained using the SLFS-L for the

2006–09 period confirm the results obtained by
the FSO (2009) for three educational levels (and
therefore with no distinction between ISCED 5A
and 5B), but for the period from 1996 to 2008.
Next step
The Cedefop (2010) works are very important, specifically because they combine
aspects connected with the skills supply with those connected with demand.
Given that works on the educational level, and therefore on the supply, may also
serve as a basis for strategic considerations, it is important to make combined
and consistent forecasts both for the educational level and for the education
190
system. Depending on the available data it may constitute a challenge, but this
work is necessary for several reasons.
(a) these works enable us to develop better hypotheses and therefore better
forecasts for the educational level, given that the flows of new graduates
stem directly from the scenarios for the education system. Because of this
consistency, they can also provide more comprehensible results;
(b) the forecasts are easier for decision-makers to use because it is then
possible to connect future developments in the population’s educational level
with future developments in the numbers of pupils or students.
Figure 12:5 Number of first degrees from professional education and training
programmes (ISCED 5B, high qualification)
Num ber of first degrees

Total 2006-07, sw iss nationality, moving average (5)
3500
3000
2500
2000
1500
1000
500
0
15 20 25 30 35 40 45 50 55 60 65
Age
Edusys: 2006-07 (estimation), (Total / N: 40 200) SLFS (L) : (2006-08)*2/3, (Total: 57 400, N = 250)
The discrepancies currently recorded in the numbers of diplomas show the

difficulties of achieving this objective, for either the future development is
inconsistent with and higher than the future outputs in the education system, or
the current educational level, which serves as a starting-point, is not compatible
with the education system. Figure 12:6 shows the projected future development
in the educational level of the Swiss population according to the data source
used. The forecast of the number of people without post-compulsory education is
very robust in relation to the source used. This is due to the fact that the various
sources have a very high consistency and the reduction in the proportion of
people without post-compulsory education is principally due to the replacement of
the old generations, who have distinctly less education than the young
191
generations. At the tertiary level, the difference is significant and is reaching

around four points towards 2020. This difference concerns both the entire tertiary
sector and the university sector.
Figure 12:6 Level of education of the population of Switzerland as a function of

the data source used for the flow
Group: population w ith sw iss nationality, aged 25-64, model w ithout migrations/naturalisations
70%
Continuing education
60% in 5A/5B ?
50%
Observed
Low qualification
40% Medium qualification
Transitions from 5B to 5A/6, High qualification
continuing education?
30% SLFS-L
Low qualification
High qualification
20%
Universities
EduSys
10% Low qualification
High qualification
0% Universities
1995 2000 2005 2010 2015 2020
Owing to the limits indicated above, the only possible interpretation currently
that is consistent with the education system data is to communicate the higher
limits. Accordingly, in 2020 the SLFS should show proportions of tertiary-level
graduates of 47% and of university graduates of 28%. The same proportions,
calculated consistently with the outputs of the education system, should however
remain lower than 44% and 24% respectively.
12.5. Closing remarks

As concerns the educational level, a very good agreement is found with Cedefop
(2010) results, if the age composition is abstracted. This discrepancy in the age
composition shows that the Cedefop model might perhaps be improved by
considering age-related flows that would be consistent with the stock. That
192
means considering the large proportion of people who, in Switzerland for

example, obtain a first tertiary-level diploma after the age of 30 years.
To make good use of results of the scenarios for the educational level, it is
important to run controls on the compatibility between the flows causing the
developments in the proportions of graduates in the population and those
delivered by the education system. In Switzerland detailed comparisons were
made by ‘differentiating’ the ‘stock’ or taking the SLFS panel data as a basis.
These comparisons show appreciable gaps, especially for those over 30 years of
age, and indicate that the development of the educational level is faster than it
would be based on the numbers of diplomas recorded in the education statistics.
It is probable that such inconsistencies between the ‘stock’ and the flow of new
graduates exist in other countries. Comparative works of this kind must also have
an impact on the quality of the data sources. It is in fact for this purpose that an
additional category was recently introduced in the microcensus of ‘basic
education and continuing education’ that is being carried out in 2011.
As the educational level is a key indicator for the measurement of the stock
of human capital, and its future development is at the heart of political
considerations, scenarios for the educational level should be consistent with the
scenarios for the numbers of pupils and students. In the short term, this can be
achieved relatively easily as long as the flows of graduates are consistent
according to the various sources. In the long term, this also requires formulating
consistent hypotheses between the various models.
References
Cedefop (2010). Skills supply and demand in Europe: medium-term forecasts up
FSO (2004). L'enquête suisse sur la population active (ESPA): concepts, bases
méthodologiques, considérations pratiques [The Swiss survey of the active
population (ESPA, SLFS): concepts, methodological foundations, practical
considerations]. Neuchâtel: Office fédéral de la statistique.
FSO (2009). Evolution future du niveau de formation de la population de la
Suisse [Future development of the educational level of the Swiss
population]. Neuchâtel: Swiss Federal Statistical Office.
FSO (2010a). Scénarios 2010-19 pour l’école obligatoire [Scenarios 2010-19 for
compulsory schooling]. Neuchâtel: Office fédéral de la statistique.
FSO (2010b). Scénarios 2010-19 pour le degré secondaire II [2010-19 scenarios
for the higher secondary level]. Neuchâtel: Swiss Federal Statistical Office.
193
FSO (2010c). Les scénarios de l’évolution de la population de la Suisse, 2010-60

[Scenarios for the development of population in Switzerland, 2010-60].
Neuchâtel: Swiss Federal Statistical Office.
FSO (2011a). Scénarios 2011-20 pour les hautes l’écoles et les diplômés dans la
population [University scenarios 2011-20 and university graduates in the
population]. Neuchâtel: Swiss Federal Statistical Office.
FSO (2011b). Titres de la formation professionnelle supérieure: un état des lieux
statistique [Qualifications in advanced professional training: a statistics
review], Neuchâtel: Swiss Federal Statistical Office.
Lutz, W. et al. (1999). Demographic dimensions in forecasting: adding education
to age and sex. In: Lutz, W. et al. Frontiers of population forecasting. New
York: Population Council.
Sheldon, G.; Straubhaar, T. (2008). Was bringt uns die neue Zuwanderung? Eine
Kosten-Nutzen Betrachtung. [What does the new immigration bring us: a
cost-benefit analysis]. In: Müller-Jentsch, D. (ed.) Die neue Zuwanderung
[The new immigration]. Genève: Avenir Suisse.
194
CHAPTER 13.
Education — Training complementarity in
Finland
Erkki Laukkanen
Following the division to three qualification levels based on the international

standard classification of education (ISCED), the supply of adult education
is estimated for the period 1990 to 2006. In the 1990s, adult education and
training of better educated was superior to less educated individuals. This
was because of education-training complementarity. But in the 2000s, the
difference between skill-levels has decreased, since vocational adult
education and training of those with low skill-level has relatively increased.
Training-training complementarity, instead, has prevailed, i.e. work-oriented
training generates more work-oriented training. An explanation to
decreased education-training complementarity is the forecasted shortage of
qualified employees in the long term. The shortage has motivated
redirecting training to those with no vocational education. The lasting
training-training complementarity is to be seen as a consequence of this
redirecting: upgrading skills of those with no vocational education favours
short work-oriented courses.
13.1. Study set-up

The situation in Finland is challenging. The education level of young cohorts is
higher than older ones. One key reason is that education attainment of younger
age groups is among the highest in the world. Today the demarcation line is
rather close to 50 years of age. This dichotomy is a consequence of the
decisions made in the 1960s and 1970s, when the supply of the secondary and
tertiary level education was substantially increased. Educational attainment of
people who were educated in the 1970s is considerably better than the
educational attainment of people who were educated earlier. Since the 1980s,
the basic idea of Finnish adult education has been upgrading, i.e. directing
people towards the next higher level of education (Lifelong learning Committee,
1997).
From the policy point of view, it is favourable that this challenge is also
recognised among the Finnish population, with rather small differences in
assessment between the young and the old, as depicted by Aro et al. (2005). In
the data, more than 90% responded affirmatively to several statements
concerning the importance of education. On such a consensus it is possible to
195
constitute policy interventions that otherwise would not be possible. These

interventions are targeted at people with no vocational degree.
The results from the last 10 to 15 years are ambiguous. However, adult
education has expanded quite well compared to other OECD countries. Every
second person takes at least some adult education during 12 months reference
time. Especially, the volume of employer-provided training has increased in
general, and in the 2000s, the focus has been on those who enter the labour
market without a vocational degree.
Since 1994, it has been possible to aim at competence-based qualifications
without formal education procedures. In the 2000s, the volume of this system has
expanded considerably favouring those with no formal educational qualification.
Besides, starting from 2003, and ending in 2009, the lifting campaign focused on
lifting the qualifications of the educationally weakest.
The statistics show that most of the vocational adult education and training
takes place outside the formal, degree-based education system. In the 1990s,
the share of employer-provided training of all vocational training has increased.
This is not problematic as such, but since only some employer-provided training
is degree-oriented, vocational adult education has not directed people to next
higher level of education as planned in the 1980s (Laukkanen, 2010). Formal
skills upgrading has not taken place.
In the 2000s, the general picture has changed, as depicted below. Training
days for well-educated (skill-level 3) have decreased, days for typically educated
(skill-level 2) remained constant, and days for poorly educated (skill-level 1)
increased. This may be because of new policy interventions discussed above,
but also other reasons, arising from other changes in the labour market, may
exist.
Altogether, deviations from the old pattern suggest that education-training
complementarity has weakened. In the following this possibility is modelled with
data from the Finnish adult education surveys (AES), from 1990, 1995, 2000 and
2006. The methodology is microeconometric. The modelling is based on
Heckman’s (1972) two step estimator. In step one, selection for training is
estimated. In step two, the number of training days is estimated with a selection
correction term constructed from the first step estimation. By comparing results
for four successive years, it is possible to examine in some detail, what has
changed. Finally, as policy conclusions, possible reasons for these changes are
discussed.
196
Figure 13:1 Days in vocational adult education and training by skill level during
the last 12 months
14
12
skill3
10
skill2
8
6
skill1
0
1990 1995 2000 2006
Source: Adult education surveys 1990, 1995, 2000, 2006.
13.2. The literature

How to identify wage effect of training when wages are observed for the trained?
This problem of selection bias is well known from other research topics of labour
economics and Heckman’s (1979) two-step estimation has become the standard
method for correcting this problem. The same method is on offer for the
correction of selection for training bias in estimation of training days. In the first
step, a probit model for positive annual training days is estimated. In the second
step, the determinants of training days acquired during the last 12 months are
estimated, including a selection correction term formed from the predicted values
of the first step model.
Besides this, selection for training and determination of training days are
contextual issues. In Finland, around 80% of all vocational training instances and
around 50% of all days are employer-provided training, the need of which is
assessed in cooperation between employers’ and workers’ representatives.
Assessment includes informal relations arising from the people and tradition, and
formal relations, such as meetings of cooperation bodies and documentation of
the common understanding regarding the need for training. But the cooperation
does not include such specialisation as the use of union learning representatives
like in the UK (Davies, 2008).
Thus, direct guidance from outside the undertakings should not much affect
the selection for training within undertakings. But then there is the question of
197
indirect guidance, through collective bargaining. In Finland, collective bargaining

takes place between unions, and workplace bargaining is conditional to unions’
approval. Moreover, collective agreements are generally binding, i.e. they apply
to employees of an unorganised employer too. It follows that with around 70%
union density bargaining coverage is 90-95%. This enables an institutional
compression of wages, which decreases individuals’ incentives to get training,
but increases firms’ incentives to invest in training (Malchow-Moeller and
Skaksen, 2003).
Harcourt and Wood (2003, 2007) go further, and name Finland as a
coordinated market economy, as contrasted to a liberal market economy, such as
the UK. It is typical that in coordinated market economies, firms resource their
internal labour markets by employment protection, which then provides workers
with incentives to stay and develop their skills in the firm. In internal labour
markets, firm-specific skills are created in the long term, through a combination of
training, classroom instruction, work experience, mentoring, induction and other
mainly informal processes. For employers this can be costly, but if firms are able
to share the costs with workers, as compensation for employment protection, and
the government, as compensation for positive externalities, cooperation can
induce a superior development in productivity. Thus, in such labour market there
should be more training.
Actually, we do not know to which extent the above principles are present in
the Finnish labour market. In the EU, the framework concept for discussions like
this is flexicurity, a concept seeking some kind of balance between flexibility,
underlined by employers, and security, underlined by trade-unions. Within this
framework training includes both points of view, flexibility and security, making it
difficult to name the actual driver of change. Despite its ambiguities, discussed by
Keune (2008) and Rubery et al. (2008), it challenges the values of employment
protection, and — at least indirectly —the status of seniority, years with the
present employer.
Unfortunately, only few of these aspects are included in the data. This is a
general data availability problem. Such a data source that would include all the
contextual variables hardly exists.
13.3. The data

In the Cedefop forecasts for skills demand and supply, the focus is on education
in general, i.e. there is no division between education at a young age and adult
education. In most models the study set-up is based on macroeconomics and
time series data. The basic units of data are aggregative in nature. In the case of
198
adult education and training this is not possible, since the basic units of data are
persons and their characteristics.
By means of labour force survey it would be possible to construct time series
data for persons. But, unfortunately, training data of the labour force survey are
not reliable. As compared to data from Finnish adult education surveys of 1990,
1995, 2000 and 2006, the labour force survey leads to around 10 percentage
points lower annual training participation rate. This is because of differences in
data collection (78).
The adult education survey includes information on all human capital
variables, such as formal, degree-based education, years and nature of work
experience as well as adult education. It also includes in-service training (79) ,
measured by days during the last 12 months. Besides in-service training, paid for
by the employer, vocational education and training includes self-motivated
training, paid for by the student, and training for labour market reasons, paid for
by Ministry of Labour. In the following, training includes all these forms of
vocational education and training.
In Table 13:1, there are two sets of data: one for the first step estimation and
another for the second step estimation. There are some common variables in
both data sets, but the figures differ, since they depict to different samples. The
first data set includes those who have had at least some work experience. The
second data set includes those who have taken at least some vocational training
during the reference year, 12 months before the interview of the adult education
survey at given years (80).
From 1990 to 2006 training has expanded from 36 to 46% of the population.
Meanwhile, average days of training has increased only from 15.6 to 18.5 days.
The standard deviations, however, are large referring to the fact that training is
unevenly distributed. The typical range is from 2 to 5 days, especially in the case
of in-service training, but then some receive training even 270 days in one year’s
time. For those employees training is full-time education. The necessary thing to
(78) In the adult education survey, sample persons’ experiences are examined according
to an extensive questionnaire and a face-to-face interview. In such case, sample
persons may feel some social pressure to report at least some training, however, all
relevant activity is reported accurately. In the labour force survey, in-service training
is just one issue among tens of others. The survey is done by telephone, without
separate expressions of education types and contents. There are no questions about
other adult education forms, indicating the relative position of employer provided
training. For these differences in data collection, the adult education survey reports
higher participation rates for in-service training than the labour force survey.
(79) The Finnish parallel to Anglo-American on-the-job training.
(80) In the following, for reasons of expression, I call all this activity training.
199
do, when modelling the causality between training days and dependent variables,
is to transform days into natural logarithm of days, and thus to make the
distribution more normal.
Most of the dependent variables are indicators taking value 1 if the condition
holds, and otherwise 0. Besides skill-levels, there are indicators for labour-market
position (entrepreneur, wage earner or precarious), type of upgrading and
gender. Precarious, as a labour-market position, refers to those without an
established position of either an entrepreneur or a wage earner, such as students
and unemployed. Forms of training are degree-oriented education and in-service-
training, which are not mutually exclusive. In 2000, only around 3% of in-service
training was degree-oriented. By 2006, percentage had risen to 13. Considering
this, it is possible to estimate that the other education, the comparison group, has
decreased from 35% in 1990 to 14% in 2006.
From 1990 to 2006, skill-level 1 has decreased from 19%, while skill-level 3
has increased from 24 to 36%. The rest, from 35 to 44%, are those with skill-level
2, with upper secondary education. This is to be considered as an extensive skills
upgrading. However, there is only a limited increase in the number of training
courses acquired during working life. From 1990 to 2006 they have increased
from 10.3 to 11.4 courses.
13.4. Selection for training

In the first step, I estimated the probability for positive number of training days. In
the probit model, the independent variable is marked by one if training days >0,
and by zero otherwise. This model is needed for two reasons. First, to show how
well the available variables explain selection for training. Second, the predicted
values of this estimation are used to form a selection correction term, included in
the second step estimation for the number of training days (81).
The results of Table 13:2 show that the history of vocational adult education
is a key determinant in selection for training. The more courses before the
reference year, the more likely there has been training during the reference year
as well. But over time, there seems to be a tendency to smaller effect, especially
if there have been only few courses prior the reference year. In 2006, one
previous course increased the probability of training by 18%, and more than 10
(81) If the coefficient estimate then proves to be statistically significant, the term corrects
the bias rising from the fact that some people are more probably selected than
others, independent of self-selection.
200
courses by 54%. Thus, training-training complementarity has prevailed, at least in

relative terms.
Table 13:1 The data

1990 1995 2000 2006
Variables
Mean Std.d. Mean Std.d. Mean Std.d. Mean Std.d.
First step
Training days > 0 0.36 0.37 0.44 0.46
Male 0.50 0.50 0.51 0.50
Position
Entrepreneur 0.13 0.13 0.11 0.10
Wage earner 0.64 0.54 0.59 0.63
Other 0.23 0.33 0.30 0.27
Skill-level
Skill-level 1 0.41 0.31 0.27 0.19
Skill-level 2 0.35 0.38 0.40 0.44
Skill-level 3 0.24 0.30 0.33 0.36
Previous training
1 course 0.10 0.08 0.07 0.06
2-3 courses 0.15 0.13 0.13 0.13
4-10 courses 0.20 0.22 0.23 0.23
>10 courses 0.27 0.35 0.42 0.44
Number of
3070 3510 2937 4119
observations
Second step
No of training days 15.60 26.80 18.70 38.30 18.50 36.00 18.50 38.60
Male 0.49 0.47 0.48 0.45
Years of age 39.70 9.00 40.90 9.20 41.40 9.20 40.90 10.50
Previous courses 10.30 5.30 10.90 5.20 11.20 5.10 11.40 4.90
Position
Entrepreneur 0.07 0.09 0.08 0.07
Wage earner 0.92 0.81 0.82 0.82
Precarious 0.01 0.10 0.10 0.11
Skill-level
Skill-level 1 0.21 0.18 0.14 0.12
Skill-level 2 0.36 0.36 0.37 0.38
Skill-level 3 0.44 0.47 0.48 0.50
Training form
Degree oriented 0.10 0.05 0.05 0.17
In-service-training 0.61 0.76 0.78 0.79
Other training 0.35 0.21 0.20 0.14
Number of
1102 1302 1287 1889
observations
201
Skill-levels matters too, but there changes have taken place only in the
2000s. In the 1990s, those with skill-level 2 had 5 to 6% bigger probability to get
training than those at skill-level 1. But in the 2000s, the premium decreased to
zero. Also those with skill-level 3 have lost some of their premium, but still in
2006 it was 11% compared with those with skill-level 1.
Also age is relevant. Ageing decreases the probability of training, especially
after 54 years, as compared to younger than 35 years of age. And over time, this
feature has got stronger suggesting that age discrimination is getting stronger.
However, in this respect, the results for year 2006 are not worse than for 2000.
Wage earners have retained their premium relative to entrepreneurs and
precarious labour, i.e. those with no established labour-market position. In 2006,
wage earners’ premium was estimated to be 20% (32-12) compared with
entrepreneurs and 32% compared with precarious labour. All the time, excluding
year 2000, males’ marginal probability for training has been smaller than that for
females.
Table 13:2 Marginal probabilities to positive number of training days (%)

1990 1995 2000 2006
Dependents Std. Std. Std. Std.
dF/dx dF/dx dF/dx dF/dx
Err. Err. Err. Err.
Male -0.07** 0.02 -0.04* 0.02 -0.03 0.02 -0.09** 0.02
CF age < 34
- -
Age 35-44 -0.08** 0.02 0.02 0.03 -0.09** 0.02
0.08** 0.08**
- -
Age 45-54 -0.11** 0.02 0.02 0.03 -0.16** 0.02
0.11** 0.15**
- -
Age > 54 -0.20** 0.02 0.02 0.02 -0.34** 0.02
0.25** 0.37**
CF precarious
Entrepreneur 0.30** 0.06 0.10 0.03 0.05 0.04 0.12** 0.03
Wage earner 0.35** 0.02 0.30** 0.02 0.29** 0.02 0.32** 0.02
CF skill-level 1
Skill-level 2 0.05** 0.02 0.06* 0.02 0.06* 0.03 -0.01 0.03
Skill-level 3 0.14** 0.03 0.15** 0.03 0.17** 0.03 0.11** 0.03
CF none
One previous course 0.41** 0.05 0.29** 0.05 0.18** 0.05 0.18** 0.05
2-3 previous courses 0.46** 0.04 0.32** 0.04 0.25** 0.04 0.24** 0.03
4-10 previous courses 0.61** 0.04 0.43** 0.03 0.39** 0.04 0.34** 0.03
>10 previous courses 0.77** 0.03 0.62** 0.03 0.59** 0.03 0.54** 0.03
Number of observations 3070 3510 2937 4119
Pseudo R2 0.40 0.30 0.29 0.24
Robust standard errors, * and ** are statistically significant coefficients at the level of 0.05 and 0.01
CF = Compared to.
202
There is one more result to report, about the model itself. The coefficient of
determination, Pseudo R2, has decreased over time to almost half. This refers to
other latent factors, not available in the data. Thus, the above results are robust
only for given data. More detailed data, if they were available, could change the
picture.
13.5. Training days

In the second step, I estimated the amount of training acquired during one year
reference time before the interview round. The estimator is ordinary least squares
with robust standard errors. In Table 13:3, the coefficient estimates depict to %,
as the independent variable is logarithm of training days. The available
dependents explain from 10% in 1990 to 30% in 1995 of the total variation of
training days (82). Since 1995, there has been a systematic decrease over time,
much stronger than in Table 13:2.
The results suggest that males have acquired less training days than
females, especially in 1990 and 2006. In the 1990s, the premium for young, less
than 34, was significant and increasing. In the 2000s, something has changed,
since only those older than 54, have received less training days than the
youngest. For entrepreneurs there seems to be a persistent training deficit. They
have acquired less training days than wage earners and precarious labour, i.e.
those with no established labour-market position. Since 1995 the difference
between wage earners and precarious labour, 60 to 90%, has been statistically
insignificant or very near to that.
When it comes to skill-levels in the 1990s, those with skill-level 3 acquired
50 to 70% more training days than those with skill-level 1. But even then there
were no statistically significant differences between skill-levels 1 and 2. In the
2000s, differences between skill-level 1 and 3 disappeared too. By remodelling,
however, I found out that skill-level 3 has become heterogeneous in a way that
only the highest third-level education favours training. Bachelors’ degrees and
such do not. Thus, in the 2000s, education-training complementarity has
weakened.
Contrary to this, the number of lifetime courses in vocational adult education
and training has retained its significance. From 1990 to 2006, one previous
training course increased training days during the reference year by 9 to 11%.
(82) The per centage would have been higher if more variables and instruments were
used. But I wanted to keep it simple and avoid problems like colinearity between
dependent variables.
203
Thus, 10 courses, which is the average for sample persons in Table 13:1, would
double the number of training days. These results, as results in Table 13:2,
suggest that training-training complementarity has prevailed.
Perhaps the most interesting impact comes from training forms. In 1995, the
first year after a long and deep recession, degree-oriented training tripled the
number of training days, as compared to those supplementary, non degree-
oriented, training. Since that the difference has decreased to around half.
Training days in in-service training have always been fewer than days in
supplementary, non degree-oriented, training. Since 1995, this difference has
decreased too. But still in 2006, training spells were around half of those in
supplementary, non degree-oriented, training (83).
Finally, there is the selection correction term. It seems to be necessary for all
other years but 2006, since it is statistically significant and positive, i.e. those
selected to training are people who typically receive more training days. In other
words: in the 1990s, selection for training was not random, but based on the
characteristics of the people. In 2006, those characteristics did not anymore
affect the number of training days. The delivery of training had become
independent of these characteristics.
Now, focusing on the title of this study, how to summarise changes in
education — training complementarity and training — training complementarity?
In Figure 13:2, I have done it in the following way. I first re-estimated the model of
Table 13:3 by substituting years of education for skill-levels, i.e. sill-level 1 is 9
years, skill-level 2 is 12 years, and skill-level 3 is 16 years. Then I dropped off all
other variables but the two key variables, years of (formal) education and the
number of vocational training courses, and depicted their coefficient estimates as
per cent. Thus, the bars express the effect of one more year of (formal) education
and one more course of vocational adult education.
Results of Figure 13:2 are in line with results of Table 13:3. Training-training
complementarity, as measured by the effect of one more training course before
the 12 months reference period on training days acquired during the reference
period, has prevailed. From 1990 to 2006 in has been steady in a range from 5 to
6%. On the other hand, education–training complementarity, as measured by the
effect of one more year of (formal) education on training days during one year’s
reference period, is twofold. First, from 1990 to 1995, there was an increase from
6 to around 10%. Then, from 1995 to 2006, there was a decrease to around 3%.
(83) I also tested the interaction term of degree-oriented training * in-service training. It
proved to be statistically insignificant also for 2006, when degree-oriented in-service
training had taken a relevant role as a form of vocational adult education.
204
Table 13:3 Determinants of training days (%)

Independent: 1990 1995 2000 2006
Std. Std. Std. Std.
days in training Coef. Coef. Coef. Coef.
Err. Err. Err. Err.
Dependents
Male -0.36** 0.10 -0.12 0.08 -0.17 0.09 -0.21* 0.09
CF age < 34
Age 35-44 -0.28* 0.12 -0.22* 0.10 -0.32** 0.11 -0.12 0.11
Age 45-54 -0.35* 0.18 -0.39** 0.13 -0.63** 0.15 -0.21 0.15
Age > 54 -1.08** 0.39 -1.20** 0.28 -1.37** 0.29 -0.71* 0.32
CF precarious
Entrepreneur -0.37 0.37 -0.91** 0.19 -0.84** 0.22 -0.88** 0.20
Wage earner 0.25 0.51 0.90** 0.33 0.62 0.35 0.88* 0.37
CF skill-level 1
Skill-level 2 0.11 0.09 0.13 0.09 0.13 0.11 -0.09 0.10
Skill-level 3 0.50** 0.13 0.71** 0.12 0.61** 0.13 0.19 0.13
Previous training
No of courses 0.10** 0.03 0.11** 0.02 0.11** 0.02 0.09** 0.03
CF other training
Degree oriented 0.81** 0.11 2.12** 0.16 1.65** 0.16 0.93** 0.08
In-service-training -0.17* 0.07 -1.37** 0.16 -0.93** 0.17 -1.04** 0.18
Other
Selection correction 1.57* 0.65 1.46** 0.44 1.56** 0.48 0.99 0.58
Constant 0.07 1.04 0.02 0.69 0.01 0.77 0.51 0.84
Number of
1101 1302 1287 1825
observations
Adj R-squared 0.10 0.30 0.20 0.15
Robust standard errors, * and ** are statistically significant coefficients at the level of 0.05 and 0.01.
CF = Compared to.
Conclusions
Traditionally, the volume of vocational adult education at higher level of education
has been superior compared with people with no vocational education and
people with upper secondary level education. This is because of education-
training complementarity: well educated tend to get more training than others.
In Finland, this has been very true until the end of the 1990s. But in the
2000s, the picture has become more ambiguous. The results suggest that
previously acquired positions, such as educational attainment and the number of
previous courses in vocational adult education, do not matter as much as before.
Education-training complementarity has weakened in both models, in selection
for training and in determination of training days. Training-training
complementarity has weakened only in selection of training, and only in absolute
205
terms. In relative terms, between training intensity categories, complementarity

has prevailed there too.
Figure 13:2 Education — training and training — training complementarity over

time, as measured by the effect of education years and training
courses on training days (%)
One more year formal education One more course of vocational training
12
10
0
1990 1995 2000 2006
Why is this?
Education-training complementarity has probably decreased because of new
types of policy interventions, such as the competence-based exams without
formal schooling procedures and the so-called lifting campaign, targeted at those
with no formal educational qualification. Otherwise it is difficult to explain why
coefficients of determination (R2 and Pseudo R2) decrease over time, and
especially in the 2000s.
The primary reason for these policy interventions has been the forecasted
skill deficit in the long term. To tackle this problem it has been important to equip
the low-qualified with competences that enable them to continue until the
stipulated pension age. Thus, these new interventions may also be the reason
why training-training complementarity has prevailed, i.e. upgrading the low-
qualified usually takes form of guiding people to courses, short and long.
Besides, there is some evidence that employers have increasingly targeted in-
service-training for their key personnel.
Since low-qualified labour is also a productivity problem, fighting labour
shortage has been a joint operation for the government and labour-market
organisations. Unfortunately, also this consensus seems to be selective in
nature. It does not include precarious labour, i.e. those with no established
206
labour-market position, at least in selection for training. Their training days are
around the same as those for wage earners, but the days arise from different
training form, such as self-motivated training paid for by themselves. The training
days of wage earners are paid for by the employers, at least in the case of in-
service-training.
Selection for training and determination of training days are processes that
are to be researched more carefully. Data from the adult education survey need
to be expanded with variables reflecting the workings of the labour market
institutions, especially within workplaces. In this study those few available
structural variables, such as seniority and union membership, did not play any
significant role. Therefore, it is possible that the supply of vocational adult
education and training is not very sensitive to changes in social relations within
undertakings. For the supply of in-service-training they probably matter more.
Given the persistent kind of labour shortage, it is possible to forecast that
low-qualified remain the focus of adult education and training also in the future.
But from these grounds it is not possible to say what will happen to average days
in training. It depends on policy evaluations regarding the impacts of training.
References
Aro, M. et al. (2005). Education or learning on the job? Generational differences
of opinions in Finland. International journal of lifelong education, Vol. 24,
No 6, p. 459-474.
Davies, S. (2008). Union learning reps in a high-tech industry. Journal of
education and work, Vol. 21, No 4, p. 297-315
Harcourt, M.; Wood, G. (2003). Under what circumstances do social accords
work? Journal of economic issues, Vol. 37, No 3, p. 747-767.
Harcourt, M.; Wood, G. (2007). The importance of employment protection for skill
development in coordinated market economies. European journal of
industrial relations, Vol. 13, No 2, p. 141-159.
Heckman, J. (1979). Sample selection bias as a specification error.
Econometrica, Vol. 47, Issue 1, p. 153-161.
Keune, M. (2008). Between innovation and ambiguity: the role of flexicurity in
labour market analysis and policy making. Brussels: European Trade Union
Institute (Working paper, 2008;04).
Laukkanen, E. (2010). Wage returns to training: evidence from Finland. Turku:
Åbo Akademi (Research for doctoral thesis in economics).
Lifelong learning Committee [Elinikäisen oppimisen komitea] (1997). The joy of
learning: a national strategy for lifelong learning. Finnish Ministry of
Education (Committee Report, 14).
207
Malchow-Moeller, N.; Skaksen, J. (2003). Labour market implications of a

compressed wage structure when education and training are endogenous.
Copenhagen: Centre for Economic and Business Research (Working paper
No 2003;09).
Rubery, J. et al. (2008). Surviving the EU? The future for national employment
models in Europe. Industrial relations journal, Vol. 39, Issue 6, p. 488-509.
208
CHAPTER 14.
How to anticipate the evolution of qualifications
and training needs: a regional point of view
The ORFE ’s (84) model to anticipate recruitment needs by 2020 is a tool to

support decision-making. The calculation method is simple in order for
everyone to be able to use the tool. The central hypothesis of the model is
that trends continue, unless events modify, accentuate or reverse them. It
is a part of all diagnosis and decision supporting tools, and offers the
possibility to follow the impact of decisions, allowing to rectify them. If we
compare the actual statistics observed with our projections in the previous
round (until 2015), the results confirm our predictions. Differences are
either linked to cyclical variations, or to some measures taken by regional
stakeholders. The results are very close to those predicted by the Cedefop
model: increase of qualifications, gaps between recruitment needs and
graduates, persistent tensions in certain professions.
14.1. Introduction
Anticipating economic changes has become a central issue for regional
stakeholders. The increasing globalisation of our economy, the complexity of
processes at work, the acceleration of changes and especially the accentuation
of constraints on environment and economic austerity raise questions on how
efficient decisions are. Leading a regional forecasting exercise is particularly
demanding. It means leading stakeholders who have different interests to share
and consider the consequences of their own decisions, to possibly modify them
to serve the collective interest. It also implies making a shared diagnosis,
determining directions and making choices on possible options. From an
institutional point of view, this means the implementation of concerted actions to
achieve better results.
This work therefore needs to be rigorous. Stakeholders are not statisticians.
Elaborate tools need to be created, that answer well-formulated questions, as
well as simple tools, which do not hide the complexity of the observed
phenomena. These synthesising tools need to allow everyone to express
themselves and to base their analysis on the observed processes. In terms of
(84) Observatoire regional formation emploi [Regional observatory of training and

employment].
209
economic observation, there is no ‘truth’. Knowledge is acquired by analysing the

situation through each stakeholder’s point of view, representing different sides of
the ‘lens of truth’: therefore a systemic analysis is favoured.
We excluded the idea of building a macroeconomic model that would use
multiple hypotheses that would simultaneously cover for demographical
phenomena, demand and supply for qualifications and training. Regional
stakeholders are representatives of employees, companies, regional politicians,
representatives of State services in the centre region and persons in charge of
administrative services that implement regional and national policies among
others. They need to capture not only the modelling results but also understand
how they were obtained and to analyse the mechanisms at work on which their
decisions can have an influence.
Moreover, beyond the forecasting exercise, the region needs detailed facts
and figures to follow the results of actions that are undertaken. There is therefore
a need for a direct link between the model of employment projection and the
diagnosis and monitoring tools. That is how the tools of the Regional Observatory
of Employment and Training have been built over time.
In the first part of this chapter, we will present the differences between our
approach and macroeconomic models such as the one by Cedefop, and then
present our model and the results. Finally we will discuss the relationship and the
complementarities of this model with all of the diagnosis and monitoring tools that
we offer.
14.2. A specific method

Our method differs from the work of Cedefop and other forecasting models at
national level on several points:
(a) the geographical unit
At regional level, means of action on economic strategy are increasingly
limited because of globalisation. Many companies have their headquarters
outside the region, often abroad. The challenge for the region is to provide
the means to respond, in terms of qualifications, to companies’ needs, and
thus to implement training courses for young people but also for employees
and job-seekers; this can validate their experience or enable them to change
jobs so that they can access job opportunities closer to their territory. That is
why we chose to analyse the impact of economic changes in the region in
terms of employment, recruitment and training needs. With our projections,
we are trying to anticipate the impact of job creations during periods of
growth, and massive job cuts in times of crisis in the labour market.
210
We observed that more frequent and stronger cyclical variations have an

impact on how trends have evolved in recent years. The steady decline of
industrial jobs over the past 20 years was followed by a decrease in the
training offered to individuals within industrial jobs, especially since the
2000s. Also, the economic crisis like the one we experienced in 2009 can
durably impact regional production capacities, not only because some
companies are disappearing definitively from the regional territory, but also
because the jobs that are created during periods of growth require more
skills than the available supply of people with these qualifications;
(b) the aim
ORFE’s model is a tool to support decision-making. We do not aim at
estimating an employment level by 2020 as, for instance, the Cedefop model
does. Our tool is also designed to give us an idea of our region on a long-
term perspective, and also as a basis for collaborative decision-making and
monitoring of decisions;
(c) the model
We use a very simple model; our calculation methods are easily explained.
Everyone can understand what a figure represents. Results serve as a basis
for reflection in the framework of the regional plan for vocational training.
The simplicity of our assumptions allows everyone to understand them and
to express their views. These results are then discussed with regards to the
economic decisions that are or may be taken outside of the region, for
example national decisions and European recommendations that will impact
the organisation and level of regional employment. For example, the
reduction of civil servants will strongly impact the region because jobs in the
administration, education and health sector represent nearly 30% of regional
jobs. We then compare the results with those of other forecasting models at
the global (BIPE, 2007), national (CAS PQM, Futurible, 2009) and European
(Cedefop) levels, to work with local stakeholders on the impact of different
scenarios regarding the state of the region by 2020;
(d) the scenarios
We build scenarios around a baseline scenario with two variants, one
optimistic and one pessimistic.
We found out that it is always difficult for stakeholders who did not directly
participate in the development of our model to appropriate the results, especially
if they are used to relying on mathematical forecasting models. To minimise the
risk of making wrong decisions, they usually prefer to choose the trend
hypothesis as the most likely one to which we must prepare. However,
211
implementing actions to prepare for a trend hypothesis implies that we can be

sure that this is the best one or the most desirable.
This was the case in 1960-70, when many forecasting prospective studies
were conducted. Employment was strongly growing and the main problem was to
quickly adapt to technological developments and production processes to avoid a
slowing of growth. Analysing trend changes has allowed for a better adaptation of
occupations.
Since 1980-90, globalisation and just-in-time production have altered the
situation which has been marked by economic crises. We know today that some
manufacturing sectors are in crisis for many years. The question we must ask is
whether this evolution is inevitable and if we must adapt. But making decisions to
adapt means implementing actions that will exacerbate the trends. The question
is more along the lines of Is this trend desirable? rather than What is the scenario
that has the greatest chance of being realised? In more concrete terms, should
we remove training courses for manufacturing because it is in crisis, even if it
could aggravate the crisis? Or should we think about measures that could curb
this trend and encourage growth in certain subsectors? It has therefore become
necessary to analyse the conditions for carrying out various scenarios to avoid
accelerating negative unwanted trends.
14.3. Presentation of the model used in the centre

region
ORFE proposes a medium-term employment projection model, which allows
understanding how the creation of jobs influences the job market. It offers the
possibility of following and reorienting the strategic decisions made as a
result (85).
This method of prospective reflection relies on expertise. The work done by
ORFE since 2001 allows it to have detailed analysis on the territories, job types
and sectors of activity. This model relies on two ORFE tools: the Cahiers, that
seeks to reveal the trends (86) at work over the last 20 years on employment and
training, and the regional indicators of employment and training (IREF) which
follows the evolutions on an annual basis (87). This model was created in 2003,
(85) This article presents the results of the employment projection model looking forward
to 2010, conceived in 2005 and used until 2008 (ORFE, 2005; 2008).
(86) The Cahiers: http://www.etoile.regioncentre.fr/ORFE.
(87) The IREF: http://www.etoile.regioncentre.fr/ORFE.
212
through the work undertaken for the Plan régional de développement des
formations (PRDF 2003-10) [Regional plan for training development), conducted
by the central region of France.
Box 14:1 The centre region, near the Paris region, between industry and rural life
The centre region is located south-west of the Paris basin. With 2.3 million
inhabitants, it stretches south to Berry (Bourges). There are two major towns along
the Loire: Orléans (capital of the region, 116 000 inhabitants) and Tours (142 000
inhabitants). It is the sixth industrial region in France. Overall employment in the
regional industry is increasing less rapidly that the national average. While continuing
to lose jobs between 2001 and 2006, the industrial sector is well-represented in the
region (about 20%). Pharmacy, perfumery and cleaning sectors have 2½ times more
employees than average in France. Household equipment, industry, mechanical
equipment, chemicals, rubber, plastic are one and a half times larger (about 1.54%).
Yet the industrial sector today is composed of small and medium-sized
subcontracting companies, and some larger companies in terms of employees but
which are dependent on external headquarters. Furthermore, among the over-
represented economic sectors in the region, at least 30% of employees work in the
five largest companies in the region, which represents a significant risk in case of
recession. Decisions are indeed often taken at an international level according to a
global perspective.
The agricultural sector, with a performance of more than 1.5 times the national
average, is decreasing in favour of big farms. Regional agricultural products remain
little promoted by regional food-processing industries. Employment in the agricultural
sector is falling faster that the national average.
The tertiary sector has become increasingly important in the regional economy, but
employment is growing slower than the national average.
Overall, employment in the centre region has increased more slowly than the national
average since 2000.
14.3.1. The assumptions

As most macroeconomic models are based on a trend hypothesis, we are also
referring to a trend hypothesis that allows us to base the discussion in reference
to forecasting studies at national and European level. However our regional
perspective is different: we propose to discuss the regional trend hypothesis
considering that trends have great chances of continuing, unless there are events
which could lessen, accentuate or reverse them. The focus of our work is
therefore to consider what might change and to focus on leverages on which it is
possible to act.
This comes back to constructing:
(a) a central trend-based hypothesis based on the idea that ‘if all things equal,
the trends observed will continue’. Trend projection thus makes it possible to
213
visualise, in a simple manner, the expected creations or eliminations of

employment if no event occurs either unexpectedly or naturally.
(b) two situational hypotheses. With respect to the hypothetical trends, two
situational projections are performed, one for a favourable context, the other
for an unfavourable context.
As it is always difficult to remove oneself from the present and the recent
past to think about the future, these projections rely on the evolutions observed in
recent periods. The projections made refer to factual and concrete data, by
calling upon recent memory.
The first model proposed projections by 2010. At each reversal (crises and
recoveries), we compared employment level observed to the employment level
projected according to our assumptions. The projections are similar, even
consistent with what was observed in 2005, end of a period of stability. Results
are consistent with the trend hypothesis, except in the construction sector that
benefited from a reduction in VAT and in operational services, because
companies preferred to use temporary work rather than recruit directly after the
crisis. In 2007, they are below the trend hypothesis because after the 2001-03
crisis, the region did not regain its previous level of employment. In 2009, the
differences are very small compared to the most unfavourable projection. This
tool proves to be a real means of monitoring trends and the impact of decision-
making.
The model by 2020 identifies three hypotheses:
(a) a stability hypothesis, projection obtained if the region has after the 2009
crisis the growth rate observed between 2001 and 2005 until 2020;
(b) a favourable hypothesis, projection obtained if the growth observed between
2005 and 2007 would continue until 2020. The centre region would then
come back to the national average growth. We therefore chose to retain this
period, the closest to the French trend, as the favourable period for our
projection in the centre region. It also allows to make a link with national
projections;
(c) a crisis projection, covering the 2007-09 period.
14.3.2.Modelling
Our model is simple. An annual average growth rate must be applied to
employment in the base year until year +7 or +10.
Recruitment needs are calculated as the sum of net job creations
(employment estimated by the model in 2020 minus employment observed in
2007) and retirement estimations (calculated on two assumptions about
214
retirement age here: one starting at the age 62 and one at age 65 in 2020).
Currently, the French are retiring at the age of 60.
14.3.3. The results
Employment projections
They show high variations according to the different hypotheses in almost all
sectors (Figure 14:1).
Figure 14:1 Employment projections by 2020
140000
S01_05
120000 S05_07
S07_09
100000
80000
60000
40000
20000
0
A0 B0 C1 C2 C3 C4 D0 E1 E2 E3 F1 F2 F3 F4 F5 F6 G1 G2 H0 J1 J2 J3 K0 L0 M0 N1 N2 N3 N4 P1 P2 P3 Q1 Q2 R1 R2
A0 Agri cul ture, fores try, fi s hi ng

B0 Ma nufa cture of food products , bevera ges a nd toba cco G2 El ectri ci ty a nd ga s s uppl y; col l ection a nd di s tri bution of
C1 Ma nufa cture of cl othi ng a rticl es a nd l ea ther products wa ter
C2 Publ i s hi ng, pri nting a nd reproduction of recorded medi a H0 Cons truction
C3 Ma nufa cture of pha rma ceutica l s products , perfumes , s oa p J1 Sa l e, ma i ntena nce a nd repa i r of motor vehi cl es a nd
a nd cl ea ni ng prepa ra tion motorcycl es
C4 Ma nufa cture of domes tic equi pment J2 Whol es a l e tra de a nd commi s s i on tra de
D0 Ma nufa cture of motor vehi cl es J3 Retai l tra de, repa i r of pers ona l a nd hous ehol d goods
E1 Bui l di ng of s hi ps a nd boa ts , ma nufa cture of ra i l wa y K0 Tra ns ports
l ocomotives , rol l i ng s tock L0 Fi na nci a l a ctivi ties
E2 Ma nufa cture of metal products , ma chi nery a nd equi pment M0 Rea l es tate a ctivi ties
E3 Ma nufa cture of el ectri c a nd el ectroni c equi pment N1 Pos t a nd tel ecommuni ca tions
F1 Mi ni ng a nd qua rryi ng except energy produci ng ma teri a l s , N2 Cons ul tancy a nd a s s i s tance a ctivi ties
ma nufa cturi ng of other non metal l i c mi nera l products N3 Renting a nd other bus i nes s a ctivi ties
F2 Ma nufa cture of textil es N4 Res ea rch a nd devel opment
F3 Ma nufa cture of wood, wood products , pul p, pa per a nd P1 Hotel s a nd res taura nts
pa per products P2 Recrea tiona l , cul tura l a nd s porting a ctivi ties
F4 Ma nufa cture of chemi ca l s , rubber, pl a s tic a nd chemi ca l P3 Other pers ona l a nd domes tic s ervi ces
products Q1 Educa tion
F5 Ma nufa cture of ba s i c metal s a nd fa bri ca ted metal Q2 Hea l th, s oci a l work
products R1 Publ i c a dmi ni s tra tion a nd defence
F6 Ma nufa cture of el ectri c a nd el ectroni c components R2 Activi ties of members hi p orga ni za tions a nd extra
G1 Extra ction of coa l , crude petrol eum, ga s a nd ura ni um;
ma nufa cture of coke, refi ned petrol eum products , a nd
nucl ea r fuel
Net job creations

For the stability scenario, job creations would be negative for industry. They
would be quite moderate in the tertiary sector and construction (Figure 14:2).
215
The relative growth scenario would generate significant job creations in the
sectors of construction, transportation, and services for companies (counselling,
assistance and operational services). In education, health-social work and
administration, planned job creations are more moderate.
The crisis scenario, a continuation of the trend between 2007 and 2009,
shows how important job losses could be in all sectors if such a crisis would
happen again.
Figure 14:2 Net job creations and retirements by 2020
4000
CN - S01_05
3000 CN -S05_07
CN -S07_09
2000 Retirements i n 2020 (62)
Retirements i n 2020 (65)
1000
-1000
-2000
-3000

G2 El ectri ci ty a nd ga s s uppl y; col l ection a nd di s tri bution of
B0 Ma nufa cture of food products , bevera ges a nd toba cco
wa ter
C1 Ma nufa cture of cl othi ng a rticl es a nd l ea ther products
H0 Cons truction
C2 Publ i s hi ng, pri nting a nd reproduction of recorded medi a
J1 Sa l e, ma i ntena nce a nd repa i r of motor vehi cl es a nd
C3 Ma nufa cture of pha rma ceutica l s products , perfumes , s oa p
motorcycl es
a nd cl ea ni ng prepa ra tion
J2 Whol es a l e tra de a nd commi s s i on tra de
C4 Ma nufa cture of domes tic equi pment
J3 Retai l tra de, repa i r of pers ona l a nd hous ehol d goods
D0 Ma nufa cture of motor vehi cl es
K0 Tra ns ports
E1 Bui l di ng of s hi ps a nd boa ts , ma nufa cture of ra i l wa y
L0 Fi na nci a l a ctivi ties
l ocomotives , rol l i ng s tock
M0 Rea l es tate a ctivi ties
E2 Ma nufa cture of metal products , ma chi nery a nd equi pment
N1 Pos t a nd tel ecommuni ca tions
E3 Ma nufa cture of el ectri c a nd el ectroni c equi pment
N2 Cons ul tancy a nd a s s i s tance a ctivi ties
F1 Mi ni ng a nd qua rryi ng except energy produci ng ma teri a l s ,
N3 Renting a nd other bus i nes s a ctivi ties
ma nufa cturi ng of other non metal l i c mi nera l products
N4 Res ea rch a nd devel opment
F2 Ma nufa cture of textil es
P1 Hotel s a nd res taura nts
F3 Ma nufa cture of wood, wood products , pul p, pa per a nd
P2 Recrea tiona l , cul tura l a nd s porting a ctivi ties
pa per products
P3 Other pers ona l a nd domes tic s ervi ces
F4 Ma nufa cture of chemi ca l s , rubber, pl a s tic a nd chemi ca l
Q1 Educa tion
products
Q2 Hea l th, s oci a l work
F5 Ma nufa cture of ba s i c metal s a nd fa bri ca ted metal
R1 Publ i c a dmi ni s tra tion a nd defence
products
R2 Activi ties of members hi p orga ni za tions a nd extra
F6 Ma nufa cture of el ectri c a nd el ectroni c components
terri tori a l a ctivi ties
nucl ea r fuel
Recruitment needs
Recruitment needs are calculated by adding net job creations and retirements
(Figure 14:3).
In the stability scenario, recruitment needs will be very limited for industry.
However, although job creations would remain limited, recruitment needs could
be a problem because of retirements in the transportation sector, counselling and
216
assistance and operational services as well as in construction, trade-retail, health

and public administration.
For the relative growth scenario, recruitment needs increase slightly for
industry. However, they increase significantly in construction, transportation,
counselling and assistance and operational services, under the influence of job
creations and retirements.
Finally, in the crisis scenario, there would mainly be job losses in industry,
transportation and operational services. The other tertiary sectors would also be
impacted but to a lesser extent because retirements would mainly create
recruitment needs in trade, health-social work and administration, even if
replacing one civil servant in two is considered.
The variation between both ages of retirement shows that recruitment needs
change between retirement at 62 and retirement at 65 mainly in the tertiary
sector. This is due to early retirement measures that have been widely used by
struggling companies of the industrial sector.
14.3.4. The main contribution of this model is the identification of a

structural lack of qualifications in some sectors
The results of this model and the views of professional branches, such as Pôle
Emploi (National Employment Office) and training organisations, have led us to
wonder about the nature of jobs that suffer the most from economic shocks. We
therefore measured the tension on these jobs with a criterion generally used: the
ratio between job vacancies registered at Pôle Emploi and the number of job-
seekers (tension rate = supply of jobs/demand for jobs). Knowing that all job
vacancies are not submitted to Pôle Emploi, the calculated ratio is
underestimated, which further accentuated the observed phenomenon.
We first looked at jobs that were both in tension in 2007, the highest period
of growth, and that maintained a high growth rate even in 2009, in the heart of
the latest crisis. The persistence of a high growth rate even in times of crisis
leads us to describe these shortages as ‘structural’. In industry, this concerns
maintenance jobs, adjusters and technician jobs (Table 14:1). These are jobs
that require a high level of technicality that can only be acquired through initial
training, but that were also accessible in the 1970s and 1980s through
experience and internal training in large and medium-sized companies. Today,
given that the training offers as well as employment stability are decreasing,
these tensions pose a real problem.
217
Figure 14:3 Recruitment needs (retirement 62/65)
6000
5000 B.R. S 01_05 (Retirement 62)

B.R. S 01_05 (Retirement 65)
4000
3000
2000
1000
-1000

G2 El ectri ci ty a nd ga s s uppl y; col l ection a nd di s tri bution of
B0 Ma nufa cture of food products , bevera ges a nd toba cco
wa ter
C1 Ma nufa cture of cl othi ng a rticl es a nd l ea ther products
H0 Cons truction
C2 Publ i s hi ng, pri nting a nd reproduction of recorded medi a
J1 Sa l e, ma i ntena nce a nd repa i r of motor vehi cl es a nd
C3 Ma nufa cture of pha rma ceutica l s products , perfumes , s oa p
motorcycl es
a nd cl ea ni ng prepa ra tion
J2 Whol es a l e tra de a nd commi s s i on tra de
C4 Ma nufa cture of domes tic equi pment
J3 Retai l tra de, repa i r of pers ona l a nd hous ehol d goods
D0 Ma nufa cture of motor vehi cl es
K0 Tra ns ports
E1 Bui l di ng of s hi ps a nd boa ts , ma nufa cture of ra i l wa y
L0 Fi na nci a l a ctivi ties
l ocomotives , rol l i ng s tock
M0 Rea l es tate a ctivi ties
E2 Ma nufa cture of metal products , ma chi nery a nd equi pment
N1 Pos t a nd tel ecommuni ca tions
E3 Ma nufa cture of el ectri c a nd el ectroni c equi pment
N2 Cons ul tancy a nd a s s i s tance a ctivi ties
F1 Mi ni ng a nd qua rryi ng except energy produci ng ma teri a l s ,
N3 Renting a nd other bus i nes s a ctivi ties
ma nufa cturi ng of other non metal l i c mi nera l products
N4 Res ea rch a nd devel opment
F2 Ma nufa cture of textil es
P1 Hotel s a nd res taura nts
F3 Ma nufa cture of wood, wood products , pul p, pa per a nd
P2 Recrea tiona l , cul tura l a nd s porting a ctivi ties
pa per products
P3 Other pers ona l a nd domes tic s ervi ces
F4 Ma nufa cture of chemi ca l s , rubber, pl a s tic a nd chemi ca l
Q1 Educa tion
products
Q2 Hea l th, s oci a l work
F5 Ma nufa cture of ba s i c metal s a nd fa bri ca ted metal
R1 Publ i c a dmi ni s tra tion a nd defence
products
R2 Activi ties of members hi p orga ni za tions a nd extra
F6 Ma nufa cture of el ectri c a nd el ectroni c components
terri tori a l a ctivi ties
nucl ea r fuel
Some other jobs are in ‘tension’ in times of growth (Table 14:2). However,
they strongly suffer from the consequences of crises. These jobs are generally
accessible to beginners, usually operator jobs. Strong variations discourage
young people to choose these jobs and encourage job-seekers to change sectors
when they can.
Jobs in the tertiary sector such as nurses and caregivers, for example, are
also in ‘tension’. This is causing major problems and is likely to rise even more in
the forthcoming years because of retirements and an ageing population.
218
Table 14:1 Example of ‘structural’ professions in need

2007 2009
Tension Tension
Supply Demand Supply Demand
rate rate
52332-Maintenance technician for
thermic, climatic and refrigerating 160.8 328 204 80.9 266 329
systems
44313-Adjusters 145.2 228 157 73.4 105 143
44341-Poly-maintenance technician 128.6 562 437 70.2 394 561
52333-Electronic maintenance
121.2 292 241 70.7 200 283
technician
52133-Electric and electronic
105.6 38 36 155.2 45 29
designer
52231- Production technician for
102.4 126 123 73.6 106 144
process industries
51211-Maintenance managerial
95.7 89 93 87.3 96 110
technician
Source: Pôle Emploi.
Table 14:2 Example of ‘cyclical’ professions in need

2007 2009
Tension Tension
Supply Demand Supply Demand
rate rate
44112-Metal cutting out technician 174.1 47 27 52.4 22 42
44211-Operator on automatic
machines for electric and electronic 157.5 230 146 36.6 64 175
production
45113-Operator on forming
144.2 274 190 42.3 82 194
machines for plastic and rubber
44121-Operator-adjuster on machine
126.5 1056 835 22.5 261 1 162
tools
45212-Operator on metal production 125.8 83 66 10.3 8 78
52211-Production technician in
mechanical construction and 107.5 57 53 35.9 23 64
metalwork
44311-Maintenance mechanic 105.2 365 347 39.7 149 375
44132-Solderer 105.0 953 908 34.6 406 1 172
Source : Pôle Emploi.
14.3.5. Extension of this model: coordination with other diagnosis and

decision-support tools
The results of this model led us to look at how the consequences of cyclical
employment variations could be reflected in the labour market, by employment
area (the region is divided in 23 employment areas) and by economic sector (36).
219
Beyond similarities, we identify specific characteristics that can result in very

difficult situations.
So we created the tool to support decision-making to help anticipating
economic changes. It is a simple instrument to measure the impact of
employment changes in the labour market. These are broken down by
employment area and economic sector.
In some areas and some sectors, job offers exceeded job applications in
2007, meaning that the territory or the sector could not cope with too much
activity.
In other areas, although net job creations are negative, there is still a
significant number of recruitments. Beyond job losses, there may be recruitments
either for replacements, retirements or for new investment that can generate a
qualification need not found in the company.
These tools show that generally, when a country or an industry is subjected
to a very high tension rate, the crisis that follows is marked by job losses. During
the subsequent recovery, employment never reaches the previous growth level.
This leads us to confirm that the lack of skills can create a structural
maladjustment.
These tools, updated annually, are used to monitor employment changes,
recruitments, labour-market changes, tensions and initial training offers.
14.4. Global issues at stake: preserving the industrial

specificity of the region
14.4.1. Coping with higher skills

The ability of a territory to evolve is related to the socioeconomic history of its
inhabitants. At the same time that the number of employees has declined in the
centre region following the closure of companies and the relocation of
workshops, industry has been ‘reskilled’. The managerial rate has increased, as
well as qualifications for workers, making adjustment more difficult in times of
growth.
Securing career paths to make flexibility and upskilling compatible

In the centre region, outsourcing was considered an asset given that the low
qualification level allowed for a faster professional integration and a greater
flexibility in the labour market. The flexibility of employees on which industry was
relying came in large part from the low qualifications used in workshops. Today,
the issue of skill transferability is raised.
220
The qualification level for jobs has risen, followed by industrial restructuring.
The latter, being accompanied by job losses, has led to a lower attractiveness of
industrial training. If we add to this phenomenon the population decline, it is
understandable that tensions in the labour market are increasing during
economic recoveries; this may prevent companies from taking the opportunity to
meet more orders.
For the local authority, the challenge is to promote among companies the
awareness of the impact of job losses on youth career paths, and eventually on
the loss of regional expertise. Similarly it may play a role in the awareness
among the working population on the necessity to be trained to maintain their
employability. The training offer would have to diversify to facilitate validations in
their various forms.
Considering outsourcing as a support for economic development

The strong subcontracting tradition in the region is not necessarily a source of
weakness. Some countries that have indeed benefited from subcontracting have
acquired a know-how that has gradually made them more competitive than the
countries that had provided that subcontracting activity (particularly China).
Progressively, contracting companies are undermining their research and
development capacity while losing the detailed knowledge of the production
process, especially if they are far away.
Thus the challenge for the region is not only to preserve industrial activities
but also to develop services including research and development.
14.4.2. Serving needs in the tertiary sector

Employment in the tertiary sector is growing more slowly than in the previous
period (1990-99). The qualification rate is increasing at the same time as the
work organisation normalises and standardises. The challenge then is to qualify
employees that would try to integrate the labour market after a period of
inactivity, validating skills in particular for individual services. It is also about
developing a tertiary sector of higher qualification in order to ensure new
activities or complementary activities in vulnerable areas because of their strong
industrial representation.
Conclusion
Our results raise the question of skill transferability. They ask about lifelong
learning: initial training but also continuing training. How to detect and develop
training that offers the greatest opportunity for skill transferability? It is a research
221
topic that perhaps needs to be considered. In the 1970s many studies showed
that training programmes for clothing industry were very useful for women’s
access to careers in electronics. Similarly, training courses for car repairing led
some young graduates towards jobs in the industrial sector.
The model used in the centre region is deliberately simple, but it does not
simplify the discussion: it is based on trend observation tools since 1982 and
refers to annual tools that monitor and analyse these trends since 2003 at
regional level, as in professions (20 professional fields), economic sectors (36)
and employment areas (23).
The method is to create tools to address new issues: diagnostic or decision-
support tools (forecasting and monitoring). Each new tool is based on the
previous one. It is articulated to provide new and additional light to important
issues.
References
BIPE (2007). Quelle France en 2035? Choisir aujourd’hui [What France in 2035?
Choosing today]. Lettre du BIPE, Janvier 2007.
Futuribles (2009). Rapport Vigie 2010. Horizon 2020-30: tendances lourdes et
incertitudes majeures [Earthwatch report, 2010. Horizon 2020-30: trends and
key uncertainties]. Paris: Futuribles (monographies).
ORFE (2008). L’impact des départs en retraite sur les besoins en recrutement en
Region Centre. [The impact of retirement on recruitment needs in the centre
region]. Drafted within the framework of the Plan régional de développement
des formations (PRDF) [Regional plan for training development], January
2005.
ORFE (2005). From employment to training needs, looking forward to 2010.
ORFE (2010). Prospective: modelling to anticipate training needs.
Links
ORFE: OAD, outils d’aide à la décision [ADO, decision-support tools]
Economic sectors
http://www.etoile.regioncentre.fr/GIP/site/etoilepro/Secteurs-activites
Employment areas:
http://www.etoile.regioncentre.fr/GIP/site/etoilepro/zone-emploi#Para1
222
ORFE: IREF, Indicateurs de la relation emploi-formation en région Centre [IREF,

indicators of employment-related training in the Centre region]
Les domaines professionnels [Professional fields]
http://www.etoile.regioncentre.fr/GIP/site/etoilepro/accueiletoilepro/ORFE/Em
ploi/domaine-professionnel,
Dernières versions des IREF [Latest versions of IREF]
http://www.etoile.regioncentre.fr/GIP/site/etoilepro/IREF-dernie
223
CHAPTER 15.
Matching formal skills in Slovakia —
LFS-based evidence (88)
Marek Radvanský
This paper analyses the development of job matching between skill supply
(education) and demand (type of work performed) in Slovakia during 1998-
2009. This period was chosen according to the quality of data provided by
the labour force survey (LFS) database and it also relates to the reform of
regional administration in 1997 (rescaling NUTS III regions). The main
focus of this paper is the analysis of employees with more higher education
than is generally needed or expected by the performed activity
(overeducated) and of the share of employees with lower education than is
generally expected for the job (undereducated). To our broader analysis,
we include only employees older than 25 years so we can avoid major
problems with employees with unfinished tertiary education but who are
already working at the time. The analysis also does not include members of
the armed forces.
15.1. Introduction
The importance of matching skills and jobs across European countries rose
during the crisis period. The structure of the Slovak economy is changing
dynamically, labour market needs are changing according to this development
and the total educational level during the last decade has significantly risen. The
dynamics of this progress has to be observed from different angles to help match
future skills supply with demand. This concerns for example the adaptation of
curricula of tertiary education so that the highest number of graduates use their
knowledge and specific skills obtained at university in their jobs. Overeducation
and undereducation of the labour force can result in ineffective use of economic
potential. Future analysis will cover this inefficiency.
The complexity of this problem from a methodological and statistical point of
view is being periodically discussed (Cedefop, 2010a; 2010b). The most suitable
data set to analyse skills matching at European level are from the European
working conditions survey. The main advantage of this data is highly suitable
structure for analysing skills matching at both formal education and informal skill
(88) This work was supported by the agency to support research and development under
contract number APVV-0649-07.
224
level compared to job requirements. Unfortunately, Slovakia was only included in

this survey in 2001 and 2005. The remaining problem is that the latest available
data are from 2005 (the 2010 survey has not been evaluated yet). Due to
periodicity of the survey in the interval of five years, the data do not show
dynamics of structural changes in time.
Another suitable approach to analyse this problem could be the use of
widely available data from the national labour force survey (LFS). LFS provides
data restricted by their limited informational value, but the advantage of this
approach is the availability of LFS on a quarterly basis and their comparability
with other surveys of EU countries. With the information provided by LFS we
cannot compare the suitability of skills to the job requirements/occupation
directly, but we can compare the educational level (growth of formal skills) and
the employment in particular working groups. To create a detailed overview of
suitability of formal skills in terms of matching education with work, we use
specific information obtained by creating cross-tables between the educational
level according to the international standard classification of education (ISCED)
and job classification according to the international standard classification of
occupations (ISCO). This approach was also used in Measuring quality of
employment (UNECE, 2010).
In terms of methodology the data divided by occupations can be assigned to
a particular class of demand for skills. Due to skills upgrading the association
between ISCO and ISCED might not be stable over time, especially at more
detailed digit classification level. On main category level this issue can be only
hardly considered. For simplification, we will aggregate nine main ISCO classes
of occupation (the analysis could be widened to include a more detailed level in
next stages) to four specific classes of skills demanded (if we do not include
members of armed forces, Table15:1). Focusing on education, skill level 1
generally requires only lower than secondary education, skill level 2 requires
secondary education (specialised or general), level 3 requires at least the first
stage of tertiary education (bachelor) and level 4 requires a university degree.
From an analytical point of view, given the limited use of the first step of tertiary
education (bachelor) in Slovakia in the observed period, we can divide the
demanded skill level by transforming the ISCO classification into three
categories:
(a) high (ISCO 1-3);
(b) medium (ISCO 4-8);
(c) low (ISCO 9).
225
Table 15:1 Conversion of occupation classification (ISCO) according to required

formal skills level (education)
Required level of skills
Main class of classification ISCO-08
(education)
High level of skills (education)
1 Legislators, senior officials and managers 3+4
2 Professionals 4
3 Technicians and associate professionals 3
Medium level of skills (education)
4 Clerks
5 Service workers and shop and market sales workers
6 Skilled agricultural and fishery workers 2
7 Craft and related trades workers
8 Plant and machine operators and assemblers
Low level of skills (education)
9 Elementary occupations 1
Not applied
0 Armed forces 1, 2, 3, 4
Source: Statistical Office of the Slovak Republic and author.
The share of employees in a job that requires low level of skills in Slovakia
(ISCO 9) is currently less than 10% of overall employment and a long-term
declining trend is also observed (Figure 15:1). In the past, the gender ratio
disadvantaged women (in 1994 there was share of elementary educated women
higher by eight percentage points than men working population), while currently
gender difference in low education level as well as at occupation is less than 2%.
Most employees are working in occupations requiring medium level of skills
(education). In 2009 it was 55% of all employed persons, while the proportion
was significantly higher for men (60.5%) than for women (47.5%). Men usually
work in category 7 (qualified workers and craftsmen, 27.6%) and 8 (operators
and assemblers of machinery and equipment, 20%). Almost half of women
working in the category requiring medium skills, work in category 5 (workers in
services and trade) with a total share of 22.7%. The share of the occupations that
require higher level of skills, is currently at 35.8% of the total population
(employed), which is slightly increasing if one take a long-term view. There is a
significantly higher percentage of women in this category; they are mostly
employed in category 3 (technicians and professionals) which is mainly
determined by the high share of health and educational staff.
226
Figure 15:1 Development of employees by the classification of occupation (ISCO)
100%
90% 9 Elementary occupations

8 Plant and machine operators and assemblers
80%
7 Craft and related trade workers
70% 6 Skilled agricultural and fishery workers
60% 5 Service workers, shop and market sales workers
4 Clerks
50% 3 Technicians and associate professionals
40% 2 Professionals
1 Legislators, senior officials and managers
30%
20%
10%
0%
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
In terms of classification of education the situation is more complicated. Until

2008, the national statistics used a different classification of education; however
individual levels can be assigned reversely to six selected levels of classification,
while some levels are further divided in sublevels. For the provided analysis we
use three levels of attained education (Table 15:2):
(a) low (ISCED 0-2);
(b) medium (ISCED 3-4);
(c) high (ISCED 5-6).
Table 15:2 Conversion of classification of education classes (ISCED)

Class Original Description
Low level of education
ISCED 0 0 Pre-primary education, without education
ISCED 1 0 First stage of basic education
ISCED 2-2B 1,2,3 Second stage of basic education, lower secondary
Medium level of education
ISCED 3-3B 4,5,6 Vocational with A levels
ISCED 3C Secondary without A levels
ISCED 4 Post-secondary non-tertiary education
High level of education
ISCED 5-5A 7 Bachelor degree
ISCED 5B 8 Master degree
ISCED 6 9 PhD degree
Source: UIPSŠ, Unesco.
227
In terms of attained education level, Slovakia has a very low share of

economically active population with basic education — 5.9%, with slightly higher
share of women (6.5%). It has also one of the highest shares (80%) of people
with secondary education in Europe and a marginal share of lower secondary
education (ISCED 2C), which is not separately observed by national statistics.
Nevertheless, possible bias caused by this statistical error is by our estimation
not higher than 0.1%. For this reason, the medium level of skills is predominant in
all groups of population. There is also more than 14% of economically active
population with higher level of education and a higher share of women with high
level of education in the total economically active population compared to men.
From this perspective it is clear that there are more economically active women
with higher education than those with lower level of education in comparison to
men.
Figure 15:2 Development of economically active population classes according to

classification of education (ISCED)
100% ISCED 6
ISCED 5
80% ISCED 4
ISCED 3
60% ISCED 1+2
ISCED 0
40%
20%
0%
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
For a basic overview it is necessary to include only four combinations based

on two different levels of education (ISCED 1-4, employees with lower/medium
education; and ISCED 5-6, employees with higher education) and two groups
according to occupation category ISCO (ISCO 1-3, occupation with need of
higher level of skills; and ISCO 4-9, occupations with need of lower level of
skills). In view of the previous discussion, we will therefore consider separately
high level and low together with medium level. Considering the simplification of
using information on needed skills only according to occupation classification by
ISCO we are aware of a possibility of a problem arising from incompleteness and
reduced quality of information, but from analytical point of view, results will be
228
able to capture sufficiently the current trends, because we are expecting that the
error is likely to be constant over time.
Figure 15:3 Scheme of basic matching skills analyses by level of education

Level of education (ISCED)
Skills match
Overeducated ISCO 1-3/ISCED 5-6
Skills match Undereducated

ISCO 4-9/ISCED 1-4
Required level of skills (ISCO)
Source: Author.
In this chapter we will therefore consider and include several types of

information about obtained education level (skills) and type of occupation
according to the Slovak labour force survey. Individual data for Slovakia will be
used, therefore the analysis includes not only aggregate information about the
Slovak labour force and gender differences, but different age groups and regional
developments are also considered.
From a methodological point of view we will therefore continue with analysis
containing three educational levels and three groups of occupation category
based on expected level of skills — low, medium and high. For a better idea of
the expected results see Figure 15:4.
For a basic overview of matching formal skills we chose the latest available
year — 2009. Despite the fact that this year is, from an employment point of view,
strongly influenced by the ongoing crisis, the structural changes in light of
previous trends were not significant. We can see in Figure 15:5, that the share of
medium level of education (ISCED) is highest which is in line with our previous
expectations. From aggregate point of view, the matching of education and skills
at national high/high level is slightly over 38%, at medium/medium level almost
92% and at low/low 74%. The rate of underqualification (lower level of education
than expected level of skills) is highest at group high/medium. Most of this
disproportion is caused by the fact that formal education is not the only way to
obtain qualification, on-job training and informal education play a significant role,
229
especially when considering older employees. A striking fact is the quite high rate
of overeducation at low/medium level with a greater proportion of people
employed with low education working in occupations requiring medium level of
education. It will be important to see the distribution of age groups; we should
expect that people with low level of education and longer practice will be
employed in occupations with medium skills. From a gender point of view, the
biggest differences can be found in the category of undereducated
(high/medium). We can assume that this is because of the higher share of
medical and teaching staff with a higher ratio of women. Men mainly remain in
the medium/medium category.
Figure 15:4 Scheme of matching analysis of required skills and obtained

education level
Highly overeducated Overeducated Skills match

ISCO 9/ISCED 5-6 ISCO 4-8/ISCED 5-6 ISCO 1-3/ISCED 5-6
Overeducated Skills match Undereducated

Skills match Undereducated Highly undereducated

Source: Author.
15.2. Analysis of skills and education match by age

groups
This section is devoted to analysis of matching skills and attained education by
age groups. The age group of 15 to 24 years was not included. In this category
people who have generally already finished their formal education are included
(either secondary or tertiary).
Looking at the group of low education skills match, the rate of people
employed in this category declined continuously. The level of education has been
rising and the end of the 1990s witnessed two trends: the withdrawal from the
labour market of older people with lower education and at the same time a
decline in the rate of people with lower than secondary education. These trends
230
can be best illustrated in the category 55+. Also the total share of the category
with lower education and medium level of required skills is declining. In the same
category we expect a higher share of older employees who obtained experience
(in time) to perform more qualified occupations (with share below 4%).
Figure 15:5 General overview of formal skills matching (2009) as % of total

employment
Highly overeducated 0.1% Overeducated 1.1% Skills match 1.1%

High
Men 0.1%, women 0% Men 1.2%, women 1% Men 1.2%, women 1%

Medium
Overeducated 1.1% Skills match 1.1% Undereducated 1.1%

Skills match 1.1% Undereducated 1.1% Highly undereducated 1.1%

Low
Low Medium High
Source: Author.
A relatively stable or slightly different behaviour appears among people with

a medium level of skills, who carry out simple work, in view of labour market they
are overeducated. By 2004 there was no predominant age group in this category,
however recently we perceived a gradual breakdown by age groups. Expected
further decline in number of people with lower education will cause a growing
share of overeducated. The alternative of increasing overeducation of this
category represents potential immigration inflow of people with lower education.
The biggest group, with almost 50% of all employed, are people with
medium level of education performing appropriate work thus with suitable
occupation. Given the share of this group and quite broad scope with regard to
attained education, it would be advisable to subdivide this category and analyse it
separately (e.g. vocational and secondary general education). In terms of age
groups, we can see gradual decline in differences, with lower share of older
employees — who were included in previous categories.
The share of undereducated employees in category medium/high is
relatively high (between 20-25 %). A significant proportion of this imbalance could
be covered by the extension of the first level of tertiary education (bachelor
231
degree). Taking a long-term point of view, we can identify a moderately declining

trend that can be connected to an increase in the share of the population with
tertiary education.
Overeducation in the category higher/medium is on average 1.5%. It is
slightly higher among the age group of 25 to 34 years, and we expect that this is
related to entry into the labour market and the carrying out of temporary work with
lower needs to obtain experience.
The share of category higher/higher is closely related to the number of
graduates and the proportion of workers with tertiary education, which is the
limiting factor from this perspective (Figure 15:2). In age group 25-34 the number
of universities and graduates increased, it almost doubled over the past 10 years.
There is a relatively high share in the category 55+ for the longer persistence of
highly educated population in the labour market as they have better health and
lower level of compensation when retiring.
Considering all groups of matching formal qualification and attained
education, the share accounts for 65% of population. Two thirds of employed
people have the suitable skills according to their occupation and level of
education. The highest rate of matching is among the age group of 25-34.
Figure 15:6 Share of employees with skills matched according to obtained and
required level of education and age group
75% 25-34
35-44
70% 45-54
55+
65%
60%
55%
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Source: Author.
The largest share of undereducated is in group medium/higher. Given the

increasing rate of population with tertiary education, we expect further decline in
this category (Figure 15:7). The important aspect will be the analysis of individual
regions.
The rate of overeducation is relatively stable below 10%. Significant part of
this share is people in category medium/lower, because of small number of
232
people with less than secondary education. In connection with the increasing
share of people with tertiary education and lower demand in some sectors, we
expect future increase in category higher/medium. Currently it affects roughly
10% of graduates.
Figure 15:7 Share of employees with lower level of education than required by the
actual position by age group — undereducated
40% 25-34
35-44
45-54
30%
55+
20%
10%
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Source: Author.
Figure 15:8 Share of employees with higher level of education than required by
the actual position by age group — overeducated
14% 25-34
35-44
12%
45-54
10% 55+
8%
6%
4%
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Source: Author.
15.3. Analysis of expected skills and education match

by regions
Analysis of matching of expected skills and attained education by regions has
several specificities. The most striking factor is the distribution of economically
active population by education. It differs significantly due to differences in
population with tertiary education. Largest universities are concentrated in
Bratislava region. Significant part of students coming to study also remains in the
capital after the end of study. At the same time demand for qualified labour force
in the capital city increases. In Bratislava region the share of people with tertiary
233
education almost double the average of Slovakia and reaches more than 30%
(Figure 15:9). In other regions, the share of people with tertiary education is
between 12 and 15% and the 15% share is only exceeded in Žilina region. The
lowest share of people with tertiary education (12.1%) is surprisingly in Nitra
region.
Figure 15:9 Share of highly educated in the economically active population

disaggregated by region
35%
Slovak Republic
30%
Bratislava
25% Trnava
Trenčín
20%
Nitra
15% Žilina
Banská Bystrica
10% Prešov
Košice
5%
0%
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Source: Author.
The structure of categories of occupation and the structure of education in

individual regions are shown in Figure 15:10. The share of occupations requiring
high level of skills is highest in Bratislava region (almost 60%), followed by
Banská Bystrica region with 39% and Trnava region with 37%. Other regions
have share of occupations requiring high level of skill at 30-35% and the lowest
rate is in Prešov region. The rate of occupations requiring medium level of skills
ranges between 53 to 60%. The exception is Bratislava region with only 35%,
lower than the share of jobs requiring high level of skills. Number of jobs requiring
only low level of skills differs from 6% in Bratislava region to 11% in Žilina and
Košice regions.
In terms of structure of labour market in particular regions, the share of
persons with tertiary education ranges from 14 to 18% with the exception of
Bratislava region with 34%. Between 77 to 82% of population in regions attained
the medium level of education. Low level of education performs from 2 to 6% of
all employed in particular regions. Košice has the lowest share of low education,
2.5%.
234
Figure 15:10 Share of jobs according to standard classification of occupations and

share of employees according to classification of education at Slovak
NUTS III regions (2009)
100%
Occupation
90% high
80% Occupation
70% medium
60% Occupation
low
50%
Education
40%
high
30%
Education
20% medium
10% Education
low
0%
Bra tis l a va Trna va Trenčín Ni tra Ži l i na Ba ns ká Preš ov Koš i ce Sl ova k

Bys tri ca Republ i c
Source: Author.
The share of people employed with low level of education and occupied in a
job requiring low level of skills (low/low match) declined in all regions during the
analysed period. The highest rate of match was in Nitra region, which is
determined by higher share of population with lower level of education. Overall,
the rate of match was in 2009 in the interval from 1 to 3%. Similar trend could be
found in the rate of undereducated at low/medium level. The lowest rate of
undereducated is in Prešov and Košice regions, where only few employed with
low level of education work in occupations that require medium level of skills. In
Prešov region fewer than 25% employees with lower level of education work in
occupations requiring medium level of skills, in Košice region this rate is 37.7%.
In other regions this rate moves between 45-55%.
Overeducation medium/low skilled employed was stable during the observed
period in individual regions. In 2009 in individual regions, the rate of
overeducated was from 4 to 9%. The rate of match of employees with medium
level of education having a job that requires medium level of skills is moving
steadily in range 49-56% of all employed. The exception is Bratislava region with
the share of 31%, which is caused by significantly lower rate of people with
secondary education on total employment. In Bratislava region 50% of people
with secondary education are employed in occupations that require medium skills
and 43% in occupations that require high skills. In other regions on average 67%
of employed work in jobs with a need of medium skills and only 24% in jobs
needing high level of skills.
235
The highest rate of undereducation is thus in category medium/high. The

highest rate of secondary-educated employees in jobs requiring high level of
skills is in Bratislava region (43%), followed by Banská Bystrica region with the
share of 30% and Trnava region with 27.4%. The lowest share is found in Prešov
region with less than 20% share. Proportion of this mismatch could be reduced
by greater use of first stage of tertiary education which is conditional on the part
of people with secondary education gaining more skills (formally). The number of
people with bachelor degrees not continuing in further study would be almost the
same as the number of graduates.
The situation is markedly different in the rate of overeducation, persons
employed with tertiary education (high formal skills) working in jobs that require
only medium level of skills, the tertiary education is in this case unimportant. In
observed period, we can see slightly increasing trend in overeducated graduates
in particular regions. Some of this can be the result of non-availability of
adequate job opportunities after coming back from studies. In Slovakia on
average every 10th employee with tertiary education works in a position that
formally does not require this kind of education. This ratio is lowest in Bratislava
region (slightly less than 8%) and highest in Košice region (11.6%).
In individual regions, the ratio of people with tertiary education with jobs
requiring high level of skill (match high/high) is slightly increasing. This is the
result of increasing number of graduates. This ratio is on average 90% (of tertiary
educated), thus the effectiveness of tertiary education is at a relatively high level.
In different regions the share of these employees of total number of people
employed is around 15%, with exception of Bratislava region with the share of
these employees at 31%.
The share of highly overeducated or undereducated employees with the
level of education and required skills at high/low and low/high levels does not
reach 1% of people employed in the regions of Slovakia. In this respect we will
not pay close attention to this.
The share of employed with the match of qualification and work performed
has in observed period increased slightly in all regions of Slovakia and is
between 64-74 % of all employees. The rate of total undereducation is gradually
declining with rising total level of education, the highest remains in Bratislava
region at 30% which is caused by the demand for highly educated (tertiary)
employees. There are slightly less than 10% of overeducated employees in
Slovakia.
236
15.4. Analysis of expected skills and education match

by classification of economic activities by SK
NACE
An interesting part of present analysis is the analysis of expected skills and
education match by classification of economic activities by SK NACE. We
analysed seven aggregated sectors at first level of SK NACE as follows:
A, B Agriculture and fishing
C, D, E Mining and quarrying, manufacturing, electricity, gas and water
supply
F Construction
G Wholesale and retail trade, repair of motor vehicles, motorcycles
and household goods
H, I Hotels and restaurants, transport, storage and communication
J, K Financial intermediation, real estate, renting and business
activities
L-O Public administration, compulsory social security, education,
health and social work, other community, social and personal
service activities.
The structure of employed by classes of economic activities and the share of

employed by classification of education in individual sectors in 2009 is compared
in Figure 15:11. In terms of education, the highest share of employed with low
qualification is in agriculture and fishing (A, B) (13%), in other sectors the share
moves from 1.9% in sales (G) to 4.4% in industry (C, D, E). The lowest share of
employees with secondary education is financial services (J, K) at 56.6% and
highest in construction (F) with 89%. In terms of tertiary education, the highest
share of employees is in financial services (40.9% of all employees in sector J, K)
and in public administration with 34.5%.
Looking at required skills by classification of occupations, we can see
relatively high demand for employees with medium skills in industry, construction
and sales (A to I). The highest demand for employees with higher level of skills
(almost two thirds) is in financial services and public services (J to O). Higher
demand for low qualified employees (25.6%) can be only found in agriculture and
fishing (A, B).
237
Figure 15:11 Share of employees according to classification of occupation and

classification of education according to selected sectors by SK NACE
classification (2009)
100%
Occupation
90% high
80% Occupation
medium
70%
Occupation
60% low
50%
Education
40% high
30% Education
medium
20%
Education
10% low
0%
A,B C,D,E F G H,I J,K L-O
Source: Author.
In Slovakia agriculture and fishing differs significantly from other sectors due
to their specific needs. This sector has the second lowest share of employees
with tertiary education and lowest share of workers with this education, who work
in an adequate position requiring high level of skills. Therefore, one third of
employees with tertiary education in this sector are overeducated. This fact can
also be caused by the number of graduates without adequate opportunities on
labour market. Looking at employees with secondary education, we can find
significant gender disparities. The match in skills and education at
medium/medium level can be found in one half of the employed but this ratio is
only one third when it comes to women. The share of women with secondary
education performing jobs requiring higher level of skills (17.9%) is almost double
the share of men in this category (9.1%). More than a half of women with low
level of education work in this sector, while this share is only 39% for men. In
total roughly 75% of employees in this sector are men and 25% women.
Mining and quarrying, manufacturing and electricity, gas and water supply
(C,D,E) in 2009 is of a similar nature when it comes to structure of required skills
and attained education as the whole sector of services and manufacturing (C-I).
In this sector there is a low share of people with tertiary education (8.9%). The
sector has high share of employees with secondary education and the
medium/medium match is 64.1%. From gender perspective the differences in
various categories are not significant enough, in total 64% men and 36% women
work in this sector.
238
Figure 15:12 Formal skills match at mining and quarrying, manufacturing,

electricity, gas and water supply sector (C, D, E), 2009
Highly overeducated 0% Overeducated 0.8% Skills match 8.1%

High
Men 0%, women 0% Men 0.8%, women 0.7% Men 8.9%, women 6.7%
Medium

Men 5.1%, women 9.2% Men 67%, women 58.8% Men 14.8%, women 18.3%

Low
Men 1.3%, women 2.4% Men 2%, women 3.8% Men 1.2%, women 1%
Low Medium High
Source: Author.
The construction sector is very specific mainly due to gender imbalance.

Almost 95% of employees are men and only 5.3% are women. Women work
mainly in more qualified positions requiring high level of skills (76.8%). The sector
has the highest ratio of employees with secondary education (88.7%) and second
highest ratio of medium level of skills required (69%). At the same time it has the
second highest share of people employed in auxiliary positions requiring lower
level of education (13.9%), second after agriculture.
Wholesale and retail trade has the highest share of people with tertiary
education (10.9%) in the whole manufacturing and services sector and these
higher positions are mainly taken by men. At the same time it has relatively high
share of positions that require high level of skills (30.8%). There is the lowest
share of employees with low level of education (1.9%) in wholesale and retail
trade and at the same time low level of skills required (2.8%). Globally, more than
60% of employees in this sector are women and less than 40% are men.
Transport and storage sector together with hotels and restaurants (H,I) has
the highest share of positions with medium level of skills needed. Men usually
work at the medium positions, there is slightly higher share of women in positions
that require higher level of skills (24.7% women, 23.6% men), and similar
situation is in positions with low level skills needed (7.9% women against 3.8%
for men). The ratio of men working in this sector is 60% and a bit less than 40%
for women — inverse ratio as in wholesale and retail trade.
239
Figure 15:13 Formal skills match in construction sector (F), 2009
Highly overeducated 0.1% Overeducated 0.5% Skills match 6.6%

High
Men 0.1%, women 0% Men 0.5%, women 0% Men 5.2%, women 30.4%
Medium

Men 11.7%, women 7.3% Men 69.7%, women 14.3% Men 8.5%, women 46.4%

Low
Men 2.5%, women 0% Men 1.8%, women 1.5% Men 0.1%, women 0%
Low Medium High
Source: Author.
The financial sector (J, K) has the highest share of employees in positions
that require higher skills (70.1%). In perspective of education attained, we can
observe higher ratio of men with tertiary education having adequate position
(41.6%). Share of undereducated women is found in this sector (40.4%), while
almost two thirds of women with secondary education work in positions with
higher skills needed. In this sector there is almost balanced employment of men
(48%) and women (52%).
The public services are second after the financial sector, where the ratio of
employees in positions with higher skills required is dominant. The structure of
employed is similar to those of financial sector (Figure 15:14). In this sector we
can find less than 30% of men and more than 70% of women.
The match of skills and education in various sectors moves from 57 to 78%.
The lowest rate of match can be found in public services (J, K) and financial
services (L-O), mainly due to lack of highly skilled labour force. In these sectors,
there is much higher demand for employees with higher education, of which there
is shortage in the labour market. On the contrary, there is a different problem in
agriculture, where we can observe higher demand for workers with low
qualification.
Undereducation in various sectors is closely related to previous information;
it applies to sectors with need of highly skilled workers.
In Slovakia overeducation in individual sectors is still relatively small, it
reaches on average less than 10%. It is significantly higher only in agriculture
240
with more than 20% and in construction where there is highest demand for
auxiliary and low qualified workers.
Figure 15:14 Formal skills match public administration (L-O), 2009
Highly overeducated 0.1% Overeducated 2% Skills match 32.4%

High
Men 0.2%, women 0% Men 3.1%, women 1.5% Men 36.8%, women 30.6%
Medium

Men 8.4%, women 7.7% Men 29.5%, women 22.2% Men 19.4%, women 33.4%

Low
Men 2.2%, women 2.9% Men 0.5%, women 1.5% Men 0%, women 0.2%
Low Medium High
Source: Author.
Figure 15:15 Share of employees with formal skills match by selected SK NACE
sector
80%
A,B
C,D,E
75%
F
70% G
H,I
65% J,K
L-O
60%
55%
50%
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Source: Author.
241
Figure 15:16 Share of employees with lower education level than required in
occupation at selected SK NACE sector — undereducated
40%
A,B
C,D,E
35%
F
30% G
H,I
25% J,K
L-O
20%
15%
10%
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Source: Author.
Figure 15:17 Share of employees with higher education level than required in
occupation at selected SK NACE sector — overeducated
25%
A,B
C,D,E
20%
F
G
15%
H,I
J,K
10% L-O
5%
0%
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Source: Author.
Figure 15:18 shows the overeducation of people with tertiary education in

particular sectors. More than 31% of employees with tertiary level of education in
agriculture work in positions with lower required qualification, which leads to high
inefficiency in this sector. High level of overeducation can also be found in
transport and hotels and restaurants sector (H,I), 16.4%, and in wholesale and
retail trade, 14.8%.
242
Figure 15:18 Share of overqualified with high education, 2009
35%
30%
25%
20%
15%
10%
5%
0%
A,B C,D,E F G H,I J,K L-O
Source: Author.
Summary
The main aim of this paper was to focus on the matching between education and
demand for a qualified workforce in the Slovak economy. The analysis was
based on cross-tabulation between level of education represented by ISCED and
development of classification of occupations ISCO. Problem was simplified from
analytical point of view to three levels of occupation, which require specific
attained education: high (ISCO 1-3), medium (ISCO 4-8) and low (ISCO 9).
Similarly was treated simplification and aggregation of attained education on
three levels: high (ISCED 0-2), medium (ISCED 3-4) and low (ISCED 5-6).
Gender differentiation was also selected as important factor of employment,
especially at sectoral level, where this differentiation was highly significant.
Section 15:1 engaged in a general overview of the dependence between
qualification and type of work performed. It was found that in Slovakia almost
50% of overall employment at age 25+ has matched their attained formal
qualification and specific demand on medium level (92% of medium educated are
providing appropriate jobs and almost 28% of them work in a position which
requires higher education — undereducated) and almost 14% on high level of
education in general (92% of highly educated have appropriate jobs).
At the same time we should observe high share of underqualification mainly
at positions requiring high level of achieved formal skills (education) taken by
workers with medium level of education (22% of total employment). These skills
could be obtained also informally, but analysis shows that there is necessity for
wider use of first grade at tertiary education, bachelor title in Slovakia.
243
Section 15:2 compared four different age groups, whereas age group 15 to
24 was not considered. An interesting fact observed was the higher share of
working groups with higher formal education at category 55+. Despite a lower
share of highly educated people at older cohorts, older people stay in the labour
market for a longer period and those with lower education have much higher
difficulties to be employed after losing their job.
Section 15:3 aimed at the matching of formal skills at regional level (NUTS
III). It did not show highly significant differences between self-governed regions,
as was shown in the sectoral comparison. Significant differences were only found
for Bratislava region where almost a double share of highly educated employees
was found compared to the average of other regions. This differentiation is
mainly based on the fact that Bratislava is the capital city and a high portion of
government institutions reside there. Also, Bratislava represents the financial
centre of the country. Main universities are located in this area. From a structural
point of view regarding education matching the Banská Bystrica region is
performing surprisingly well. This region belongs to the most problematic regions
from a socioeconomic point of view with very high level of unemployment.
Section 15:4 analysed differentiation of formal skills match between
education and employment from a sectoral point of view using aggregated
classification of employment SK NACE. Significant differences not only on
aggregate view but also from a gender point of view were found. As an example,
the construction sector was mentioned with 95% of male employees, women
employed in this sector are mostly in higher positions considering skills needs
and required education. Lowest match between needs of education and formal
skills of employees is financial sector and public administration sector, where an
important disproportion between needs and supply of highly educated workforce
was presented.
This paper does not consider unemployment. This slightly affects the results.
However, the issue of unemployed could be solved by setting their working skills
by last performed job, which is also inaccurate. Major share of unemployed
attained only lower qualification level, education and skills. For further analyses
more detailed level of medium education should be considered, because most
employed are in this category. Also the analysis of different specialisation of
higher educated employees should be provided, which could determine more
efficient allocation of funding for universities. The use of such analyses can be
presented in very simple way in the sector of agriculture. More than 31% of highly
educated employees in sector of agriculture actually work in a position, which
does not require high education level — they are overeducated. These results
244
are still not strong from analytical point of view, but show a potential way of
thinking.
Provided analyses represent greater simplification of described problem, but
they present possibilities and usefulness of such analyses for each Member
State. Detailed analyses of education and skills supply and demand at national
level using existing methodology could provide more detailed results to national
decision sphere than international comparisons. Simplification to three major
groups on the side of demand and supply could be partially inaccurate, but from
the point of comparison we are expecting their value to be significant.
References
Cedefop (2010a). The right skills for silver workers: an empirical analysis.
Luxembourg: Publications Office of the European Union.
Cedefop (2010b). The skill matching challenge: analysing skill mismatch and
policy implications. Luxembourg: Publications Office of the European Union.
UNECE, Task force on the measurement of quality of employment (2010).
Measuring quality of employment. Geneva: United Nations.
Statistical Office of the Slovak Republic (1998-2009). Labour force surveys.
Bratislava: Statistical Office of the Slovak Republic.
245
CHAPTER 16.
Internet job search data as a possible
source of information on skills demand (with
results for Slovak university graduates) (89)
Miroslav Štefánik
This paper examines Internet job search data as a possible source of

information on skill demand. It proposes two methods of processing such
data. The classical method is based on a content analysis of
advertisements, the CV method on a content analysis of CVs uploaded by
individuals looking for a job. The analysis is restricted to tertiary education
graduates. Based on a representativeness analysis two specific groups in
terms of occupation and economic sector are selected. It is technicians in
public services and professionals in construction.
Results produced by both methods identified some key skills in relation to
selected groups. Despite this, many methodological difficulties arise. These
are shortly addressed in the discussion.
16.1. Introduction
Presented paper describes an alternative source of information on demand for
skills, which is ‘Internet job search data’ administrated by website companies. It
describes and compares two methods of gathering such data, as well as two
possible approaches in exploiting such data. This paper tries to present the
restrictions and advantages of these two kinds of analysis. For this purpose we
use data which were made available by a company administrating a job search
website in Slovakia. The additional cost of getting such data is very low, which
makes this source of information significantly cheaper compared to surveys
(skills survey, employers’ survey, or graduates’ survey).
Web pages designed for intermediate job search process usually perform
two procedures. The first one is a traditional collection of job advertisements and
their display on the web. This is in principle the same as newspapers do for
centuries. The second procedure was enabled later; nearly one decade ago, by
the advancement of web technologies. It consists of picking the information about
the applicants, using a standardised CV form. These CVs are subsequently
(89) This work was supported by the Slovak Research and Development Agency under
the contract No APVV-0541-10.
246
offered to employers looking for an employee. Employers can browse the

database with all the attributes (skills) stated in the CV for free. If they find an
applicant, who is appropriate for the position they are filling, they pay for the
display of contact details, so they can contact the applicant. Companies running
the job search websites have information on the attributes in the CVs and also on
the number of displays of each CV.
The first way of gathering data is the classical method. The second way is
the CV method. Both patterns of Internet mediated job search can be found
internationally; there are, therefore, reasons to assume that they could become
an international source of information on skills demand. The following text
focuses on the comparison of these two procedures as a possible source of
information on skills demand. When comparing, we describe shortly the method
used to process data from each source. Afterwards we look at the
representativeness, so we can asses to what extent the acquired information
refers to the whole population. Finally we compare the results.
16.1.1. Classical method

Data picked via the classical method which were made available for the analysis
contain:
(a) basic information on the job vacancy (position name -string and categorised,
location, type of contract, etc.);
(b) requirements of the applicant (level of required education, years of previous
experience, required skills -string and categorised, etc.);
(c) information on the company (location, number of employees, business area).
These data will be handled to provide information on skill requirements

employers put on employees, when filling open vacancies. For this we basically
make two steps. First, we select a specific group (in terms of occupation and
sector) for which we would like to identify the skill requirements. Second, we rank
required skills according to how often they appear in the advertisements.
For the analysis, all advertisements uploaded during 2009 and 2010 were
used.
16.1.2. CV method
Data picked up by the CV method which were made available contain information
published in standardised CVs on:
(a) gender;
(b) age;
(c) achieved formal level of education (e.g. university and faculty);
(d) work experience;
247
(e) skills (language skills, economic knowledge, office software, advanced

computer skills, etc.);
(f) industry, where the individual is looking for a job;
(g) position, which the individual is desiring;
(h) region, in which the individual is willing to enter a job.
This information is complemented by information on how many times were

the contact details of each CV displayed by the employers in 2009.
Disposable data allow constructing a generalised linear model, in which the
number of displays becomes the dependent variable and the attributes from the
CVs are potential independent variables. Thanks to this, we can rank the skills
according to the outcome of such equation.
This technique can also be used separately for particular groups. In
comparison to a simple ranking based on frequencies of appearance, which was
used when processing the classical method data, construction of a generalised
linear model allows us to identify some side effects as for example gender or
age. To make the results of both methods comparable we will focus purely on
skills.
When using the CV method, we operate with CVs, which were uploaded
during 2009 and the number of contact details displays counted within the year
2009. The most common length of CV publication is three months, but it can be
prolonged. The length of the publication does not have any significant effect on
the number of contact details displays of a CV.
16.2. Methodological issues

Lists of working positions and skills from which employers and employees pick
when constructing their advertisements, respectively CVs are the same. They
comprise over 500 positions and over 160 skills. Such detailed information
cannot be used for the analysis. Positions were, therefore, recoded to correspond
to nine (1-digit) ISCO groups. This will be complemented with seven identified
sectors of economic activity, which can be linked to international statistical
classification of economic activities NACE. Nine occupational groups and seven
identified sectors allow us to identify 63 specific groups, in which we can focus on
the skill requirements. This can be done on classical method data, as well as CV
method data (90).
(90) More than one position may be chosen when completing the CVs and
advertisements. In the further analysis, the CVs and advertisements are linked with a
248
Over 160 skills are grouped into language skills, administrative and
economical skills, general computer skills and skills related to particular software
or a programming language. All 160 skills from the list cannot be incorporated
into a generalised linear model, which has to be constructed when processing the
CV method data. Therefore, the model will use only those skills which were the
most frequently mentioned in the advertisements targeting the same group.
Due to differences in Internet use across age and educational groups, we
can expect some biases in the representation of some groups among the CVs.
To avoid possible complication, the analysis therefore focuses only on tertiary
education graduates, who present a relatively homogenous group in terms of age
and achieved level of education. On the other side, tertiary education graduates
are considered to be a disadvantaged group on the Slovak labour market, with
higher unemployment rates and lower wages compared to Slovak average.
Based on this, we can assume that the requirements applied to the graduates are
even stricter and that employers are even pickier, when hiring a graduate, than
an experienced employee. On the other hand some positions will drop out from
the analysis.
To filter the tertiary education graduates sample among the advertisements
we used only those advertisements which stated that the required educational
level is tertiary education and that the advertised vacancy can be filled also by a
person with zero years of experience. To filter the tertiary education graduates
among the CVs, we selected only CVs which declared to have finished tertiary
education and did so after the year 2004.
16.3. Representativeness
For further analysis we use data picked up by a company administrating the most
popular job search website in Slovakia. The market share of the company is
around 80%. This gives us a good prerequisite to acquire relatively
representative data.
Assuming (91) that the occupational and sector shares of vacancies are
similar to the structure of total employment on the Slovak labour market, we will
compare the shares of the CVs and advertisements to the structure of overall
employment. To achieve better match we will look only at the tertiary education
particular position according to the most frequent ISCO 1-digit level category. An
analogous method was applied also in linking with economic sector.
(91) This was confirmed also by a previous empirical analysis.
249
graduates CVs and advertisements, comparing their occupational and economic

sector shares to the structure of employment of people with tertiary education.
The information on employment structure was obtained from the labour force
survey. Unfortunately, we have no information on employment structure of tertiary
education graduates, for this reason we use the whole tertiary educated
population in employment.
Table 16:1 Shares of occupational groups of tertiary educated in employment,

advertisements and CVs
Labour force
Advertisements CVs
survey
Frequency % Frequency % Frequency %
Managers and legislators 16 144 18.19 16 758 16.65 3 125 12.65
Professionals 48 714 54.90 42 550 42.26 11 155 45.17
Technicians and associate
16 715 18.84 21 507 21.36 4 282 17.34
professionals
Clerks 2 723 3.07 11 622 11.54 3 886 15.73
Service workers 3 035 3.42 6 077 6.04 1 817 7.36
Skilled agricultural workers 40 0.05 24 0.02 1 0.00
Craft and related trades
393 0.44 835 0.83 169 0.68
workers
Operators and assemblers 371 0.42 286 0.28 18 0.07
Elementary occupations 593 0.67 1 018 1.01 245 0.99
Total 88 728 100.00 100 677 100.00 24 698 100.00
Source: Labour force survey 1998-2009 and web job search data.
Table 16:1 shows the occupational groups which are strongly

overrepresented, or underrepresented in the examined population. For example,
service workers present only 3.42% of total employment of tertiary educated, but
6.04% of all advertisements referring to tertiary educated and 7.36% of all tertiary
educated persons’ CVs inquire a job as a service worker.
To assess the ability of classical and CV method to gather data on
occupation we will run a chi-square goodness of fit test. The test confirmed that
the gathered data fit the overall population of tertiary educated in employment.
Both methods proved to provide roughly representative data when looking at nine
basic occupational categories.
Besides the occupational perspective, we assess the representativeness of
data also in relation to the economic sector. Table 16:2 shows the shares of
economic sectors on employment of tertiary educated, advertisements looking for
employees with tertiary education and CVs of tertiary educated looking for a job.
250
Table 16:2 Shares of economic sectors on employment of tertiary educated,

advertisements and CVs
Labour force survey Advertisements CVs
Frequency % Frequency % Frequency %
Agriculture and mining
1 969 2.28 2 780 2.67 356 1.39
(A-C)
Industry and utilities
4 490 5.19 3 512 3.37 548 2.14
(DA-DJ+DN)
Electro-machinery and
6 590 7.62 6 345 6.10 2 138 8.34
utilities (DK-DM +E)
Construction (F) 2 983 3.45 3 431 3.30 748 2.92
Sales (G-H) 7 120 8.23 18 258 17.54 1 798 7.01
Services (I-K) 16 753 19.37 61 846 59.42 17 383 67.81
Public Services (L-O) 46 567 53.85 7 911 7.60 2 663 10.39
Total 86 472 100.00 104 083 100.00 25 634 100.00
The shares of economic sector uncovered huge differences in

representation of services and public services. Based on these findings, the
conclusion can be made that employers in services and public services do not
use Internet job search to fill in vacancies to the same extent as in other sectors.
However, applicants looking for jobs adjusted their search, and they are not using
Internet job search to look for jobs in these sectors to the same extent as in other
sectors. Applicants and employers offering positions use other ways of search
than Internet job search. On the other hand both future employers and future
employees, know that jobs in (private) services are advertised on the Internet.
They, therefore, use Internet job search to look for this kind of jobs/employees to
a great extent.
By the combination of nine occupational and seven economic sector
categories we can identify 63 various groups. Due to the restriction of the
analysis focusing on tertiary education graduates, some combinations of
occupational and sector groups become marginal. For example, combinations
involving occupations ISCO 6-9, which are dominantly filled by people with lower
levels of education. Table 16:3 shows the best represented combinations of
occupation and economic sector. These are ranked according to the difference
index (92).
(92) The difference index is counted as: DI=absolute value (labour force survey share-
advertisements share)+absolute value (labour force survey share-CV share)/labour
force survey share.
251
Table 16:3 Selected combinations of occupation and economic sector and their
share in the total population
Labour force Adver-
CV Differ-
survey tisements
Occupation Economic sector ence
% of % of % of index
Count Count Count
Total Total Total
Technicians and
associate Public services (L-O) 4 854 5.7 2 945 6.7 1 180 4.8 0.18
professionals
Professionals Construction (F) 1 469 1.7 956 2.2 603 2.4 0.30
Industry and utilities
Professionals 1 364 1.6 463 1.1 334 1.4 0.31
(DA-DJ +DN)
Managers and Electro-machinery and
1 284 1.5 419 1.0 135 0.5 0.34
legislators utilities (DK-DM +E)
Managers and
Services (I-K) 3 111 3.6 2 059 4.7 2 137 8.7 0.36
legislators
Technicians and
associate Sales (G-H) 1 994 2.3 1 349 3.1 208 0.8 0.36
professionals
Professionals Sales (G-H) 1 129 1.3 342 0.8 180 0.7 0.39
Professionals 2 036 2.4 1 707 3.9 1 557 6.3 0.66
Technicians and
associate Services (I-K) 4 444 5.2 3 941 9.0 2 426 9.8 0.78
professionals
Technicians and
associate 2 790 3.3 359 0.8 275 1.1 0.78
professionals
Managers and
Sales (G-H) 2 883 3.4 3 241 7.4 687 2.8 1.18
legislators
Professionals Public services (L-O) 33 113 38.8 1 728 3.9 1 281 5.2 1.24
Agriculture and mining
Professionals 558 0.7 690 1.6 169 0.7 1.29
(A-C)
Service workers Sales (G-H) 669 0.8 899 2.1 465 1.9 1.64
Clerks Sales (G-H) 437 0.5 694 1.6 121 0.5 2.20
Professionals Services (I-K) 7 800 9.1 14 306 32.6 7 031 28.5 2.78
Clerks Services (I-K) 732 0.9 3 303 7.5 3 702 15.0 7.47
Service workers Services (I-K) 329 0.4 1 478 3.4 1 227 5.0 7.55
Table 16:3 shows that there are significant differences between the
representation of various groups. For groups in the top of the table, both types of
Internet job search data can bring some interesting information on the demand
for skills. On the other side groups situated in the bottom of the table, or totally
excluded from the table, are not represented sufficiently in the available job
search data. The information acquired in relation to these groups is, therefore,
less valid. For this reason we will choose only first two groups for further analysis,
252
in attempt to dig out some information on skills demand within these groups. The
underrepresentation of public services disappeared as we focused on technicians
in public services. The second, best represented group are professionals in
construction.
A more specific analysis of representativeness can be done on the CV
method data. These data can be compared with the information on the structure
of Slovak university graduates after 2004. This enables us to look at gender, year
and university/faculty of graduation. The results of this representativeness
analysis were surprisingly good with only one significant bias based on the type
of faculty of graduation. Graduates of some faculty types are underrepresented in
the sample. The underrepresentation of these faculty types can be explained by
the structure of study programmes. There are reasons to assume that, graduates
from most of artistic and theological study programmes are using other ways of
looking for a job, than web job search. This is valid, for example, for medicine
doctors, but not for all medical faculties graduates. Educational faculties
graduates are using web job search when looking for a job in other than
educational field. On the other side, CVs of economical and technological
faculties’ graduates are due to opposite reasons overrepresented (93).
16.4. Results
Further analysis of skill demand using Internet job search data will always try to
focus on a particular group. Its potential in producing information on a more
general level is limited. The reasons are partially connected to the
representativeness of gathered data. The representativeness analysis was
nevertheless useful, when information was provided on which groups are better
represented in the Internet job search data and which are more problematic in
this context. Internet job search is still in the process of penetrating the labour
market. Some particular areas of the labour market are more, some less
penetrated. Differences in penetration are an important limitation to the analysis
of Internet job search. Despite that, many areas of the labour market can be
satisfactorily analysed using the Internet job search data. We will focus our
further analysis on two groups selected according to their representation in
available job search data. These are:
(a) technicians and associate professionals in public services;
(b) professionals in construction.
(93) For more specific information on this analysis find Štefánik (2010).
253
16.4.1. Technicians and associate professionals in public services

Using the classical method to analyse technicians and associate professionals in
public services we can see that the most commonly required skills appearing in
the advertisements are not surprising. Employers most frequently require skills
related to office software such as Microsoft Excel and Word, together with
language skills, such as English language. The most surprising are the
requirements for skills necessary to operate the cash register and Slovak
language skills.
Table 16:4 Skills frequently mentioned in advertisements looking for technicians

in public services
Responses % of cases
Microsoft Word 1 674 62.60
English language 1 603 59.95
Microsoft Excel 1 568 58.64
Internet (e-mail, www) 1 146 42.86
Microsoft Outlook 854 31.94
Microsoft PowerPoint 538 20.12
Microsoft Windows 423 15.82
Slovak language 207 7.74
Lotus Notes 205 7.67
German language 144 5.39
Cash register 31 1.16
Open Office 29 1.08
Business correspondence 23 0.86
French language 22 0.82
Russian language 22 0.82
Invoicing 20 0.75
Warehouse management 12 0.45
Hungarian language 11 0.41
Double-entry bookkeeping 10 0.37
Adobe Photoshop 10 0.37
Microsoft Access 8 0.30
Human resources 8 0.30
UNIX/Linux 7 0.26
HTML 6 0.22
Typing 6 0.22
NB: Skills mentioned in results of both kinds of analysis are marked in bold.
Source: Analysed web job search data.
Skills mentioned in results of both kinds of analysis are marked in bold. The
CV method brought fairly different results. Nevertheless, the most frequently
mentioned skills from Table 16:4 can also be found in Table 16:5. The
contribution of some skills to the attractiveness of a CV, measured by the number
of contact details displays, is often negative. This is when the B regression
254
coefficient is negative. Such situation can be seen for example for Slovak
language. This means that CVs which mention their Slovak language skills are
less attractive in terms of number of contact details displays. This is valid also for
skills in HTML coding; Microsoft Word and Invoicing, when all these skills were
mentioned by the advertisements, but in fact are negatively related to the number
of displays. This discrepancy can be explained in various ways, one of which can
be that these skills are implied for CVs looking for this kind of position and their
explicit mentioning is linked with a lack of other skills, which affect the number of
contact details displays positively.
Table 16:5 Skills ranked according to the Wald Chi-Square in the generalised
linear model applied to technicians in public services
Standard
B Wald Chi-Square Sig.
error
Slovak language -1.723 0.051 1144.62 0.000
SAP 1.126 0.122 85.09 0.000
Polish language -1.032 0.135 58.58 0.000
Microsoft Outlook 0.299 0.045 44.13 0.000
LAN/WAN administration -2.016 0.324 38.79 0.000
Czech language 0.426 0.069 38.45 0.000
French language 0.495 0.080 38.16 0.000
Internet (e-mail, www) 0.271 0.050 29.09 0.000
HTML -1.468 0.273 28.88 0.000
German language 0.223 0.044 25.76 0.000
Adobe Photoshop 0.375 0.078 22.95 0.000
Microsoft Word -0.548 0.115 22.80 0.000
Adobe Illustrator -1.577 0.338 21.76 0.000
English language 0.275 0.063 18.87 0.000
Italian language 0.45 0.106 18.00 0.000
Invoicing -0.369 0.089 17.15 0.000
PHP 3.923 1.002 15.33 0.000
.NET 2.597 0.667 15.18 0.000
Warehouse management 0.427 0.114 14.13 0.000
MS-DOS user 0.496 0.137 13.06 0.000
Microsoft Excel 0.285 0.086 11.06 0.001
Microsoft Windows 0.147 0.046 10.35 0.001
Client/server administration 1.511 0.474 10.19 0.001
Windows server administration -1.765 0.646 7.46 0.006
Adobe InDesign 0.923 0.340 7.35 0.007
Business correspondence 0.199 0.075 7.05 0.008
Pro/Engineer 0.661 0.257 6.64 0.010
255
The CV analysis ranks the skills according to what difference they make in
relation to number of contact details displays. Combined with the results of the
classical analysis we can dig out valid information on what employers are looking
for.
16.4.2. Professionals in construction

Looking at professionals in construction, we can find an increase in importance of
specialised skills related to software used in construction business, such as
AutoCAD, Cenkros plus, or MicroStation. As compensation, the importance of
skills related to general office software, such as Microsoft Word or Excel, is lower
compared to previous group. The importance of language skills, which appeared
when looking at technicians in public services, was confirmed also when
analysing professionals in construction.
Table 16:6 Skills frequently mentioned in advertisements looking for

professionals in construction
Responses % of cases
Microsoft Excel 720 62.66
Microsoft Word 718 62.49
English language 517 45.00
Internet (e-mail, www) 481 41.86
Microsoft Outlook 421 36.64
AutoCAD 329 28.63
German language 182 15.84
Microsoft Windows 150 13.05
Microsoft Powerpoint 138 12.01
Cenkros plus 104 9.05
Slovak language 83 7.22
OpenOffice 62 5.40
Invoicing 42 3.66
Microsoft Access 32 2.79
MicroStation 31 2.70
Business correspondence 22 1.91
Microsoft Project 19 1.65
Slovak language, Microsoft PowerPoint and Invoicing show negative

regression coefficients despite the fact they appeared in the advertisements. This
can be ascribed to the wider context of these skills, as we did not include
demographic features into the equation to put in as much skills into the equation
as possible. Demographic features, such as gender could shed some more light
into the explanation of negative contributions of, for example, Invoicing.
256
Table 16:7 Skills ranked according to the Wald Chi-Square in the generalised
linear model applied to professionals in construction
Standard Wald Chi-
B Sig.
error Square
Slovak language -1.644 0.0545 909.41 0.000
Internet (e-mail, www) 0.664 0.0553 144.03 0.000
Hungarian language 0.565 0.0843 44.93 0.000
Microsoft Windows 0.342 0.0531 41.39 0.000
Lotus Notes 0.767 0.1305 34.54 0.000
German language 0.316 0.0554 32.52 0.000
SAP -1.428 0.2518 32.18 0.000
ArchiCAD -0.376 0.0691 29.69 0.000
Autodesk Revit Architecture -0.471 0.0951 24.56 0.000
English language 0.537 0.1107 23.59 0.000
CorelDRAW -0.341 0.0757 20.26 0.000
Cenkros plus 0.280 0.0648 18.72 0.000
PHP -1.687 0.4033 17.49 0.000
Microsoft Powerpoint -0.230 0.0567 16.50 0.000
JavaScript 1.061 0.2655 15.96 0.000
Single-entry bookkeeping -0.470 0.1226 14.70 0.000
ArCon -0.524 0.1386 14.31 0.000
Allplan -0.422 0.1137 13.79 0.000
Adobe Photoshop 0.206 0.0630 10.69 0.001
Windows server administration -0.748 0.2419 9.56 0.002
Mac OS -0.745 0.2480 9.03 0.003
Double-entry bookkeeping 0.392 0.1326 8.72 0.003
CAD 0.451 0.1543 8.54 0.003
Client/server administration 0.736 0.2713 7.35 0.007
Invoicing -0.215 0.0813 6.97 0.008
Corel Photo-Paint 0.302 0.1274 5.61 0.018
It is possible to include variables, such as gender or region in the analysis,

nevertheless this project is focused on skills demand identification and the
equation was constructed to look at individuals in terms of skills.
16.5. Conclusions
This paper tries to introduce two methods of processing Internet job search data.
The availability of Internet job search data leads to questions about its possible
usage in skill demand analysis. In relation to possible usage of such data, the
question of representativeness arises. The representativeness of Internet job
search data is limited primarily by the penetration of Internet usage. For this
257
reason we restricted our analysis to tertiary education graduates, which presents

a group with a fairly high level of Internet usage.
Two different methods of picking data present two different sampling
methods. The structure of data gathered by both methods was compared to the
population of tertiary educated individuals in employment. The comparison was
realised in terms of occupation and economic sector. Looking at occupational
structure we can find overrepresentation of clerks and service workers and a
slight underrepresentation of managers and professionals. Looking at economic
sector structure we can find mostly the overrepresentation of services and
underrepresentation of public services. The dominant pattern of bias seems to be
similar for both, classical as well as CV method. Looking at the combinations of
nine occupational and seven economic sector categories some groups appeared
to be represented better than others. We picked two of them, which appeared the
best in terms of representation and provided results on skills demanded by
employees when looking for employees filling positions in this ‘occupation-
economic sector’ group. We selected technicians in public services and
professionals in construction and produced results for these two groups using the
classical, as well as CV method.
When comparing the classical and CV method of processing Internet job
search data we get fairly different results. Skills identified as important by both
methods can be considered as crucial. For both cases, these are mainly
language skills, with the dominance of English, German and Slovak. For the
technicians in public services, the skills identified as important by both methods
can be named as general office skills, often related to popular office software
such as Microsoft Word, Excel or Outlook. For professionals in construction,
general office skills are less important. Crucial skills of this group can be
described as specialised skills related to construction. These involve skills
necessary to work with graphical software used in construction such as AutoCAD
of ArchiCAD, but also office software specialised for construction businesses
such as Cenkros plus.
While the classical method brings results directly based on the formulations
of employers’ needs, the CV method analyses the behaviour of employers, when
looking at the number of contact details displays. As a result, the classical
method brings more commonly known findings. The CV method tells which skills
actually make the difference. Both methods are able to bring more valuable
information when looking at a more specific group.
258
16.6. Discussion
Internet job search data and both compared methods have their limits. In respect
to representativeness these limits appear to be fairly similar. Some occupations,
and some economic sectors, are underrepresented in the sample of
advertisements and CVs. The pattern seems to be the same, when services and
clerks and service workers are overrepresented; and when public services and
managers and professionals are underrepresented. Similar bias can be observed
also when looking at the type of university/faculty of graduation, with artistic,
theological and medical faculties being underrepresented in the sample. These
biases can be linked to differences in Internet use, in particular when looking for
a job. In some fields we can expect further penetration of Internet use and
consequential disappearance of particular biases.
In general, classical and CV method showed their ability as sampling
methods. They are able to produce valuable information on skill demand for
some particular groups. Produced information can be used for example in
producing skill profiles (in case of classical method), or identifying skill shortages
(in case of CV method).
Information on skill requirements always becomes clearer when looking at a
particular group. On the general level the skill requirements can be only vague
and roughly defined. One of the advantages of Internet job search data in general
are high counts (over 25 000 CVs and over 160 000 advertisements), which allow
the analysis to go into more specific groups and to identify their skill requirements
more precisely. However, important limits arise precisely due to the very same
fact that the data were picked for a different purpose than a skills demand
analysis. An example can be drawn as we run into difficulties transforming used
occupational categories, created according to the demand of job search company
clients, to international classification of occupations (ISCO). Too broad
categories, such as account manager had to stay out of the analysis.
Methods of processing data, presented in this paper, show only two possible
ways to use Internet job search data. Data contain more information than was
used in the proposed analysis. Used data involve information on the position
which is going to be filled. This information is categorised when employers
uploading the advertisement select from a prepared list of working positions. The
data also contain similar information on required skills, which are also picked
from an existing list. Besides this categorised information the advertisement
contains also a text description of the position and required skills. This
information can be stored as a string variable and could be processed as well,
but for the analysis we used only the categorised data, which can be processed
259
easily using quantitative methods. The production of information from string data
requires sophisticated methods, but it is possible.
We restricted our analysis to the appearance of a skill, within an
advertisement or a CV. The data, after few preparations, allow us to look also at
the required level of skill. This is linked to several methodological problems which
can be solved.
Both methods also allow us to involve wider context into the analysis. In
case of classical method we can further specify the target group for the analysis,
for example: ‘women professionals in construction’. In case of CV method we can
simply involve some more variables (such as gender, age, region, etc.) into the
equation.
Reference
Štefánik, M. (2010). Slovak regional differences in the process of university
graduates recruiting for a job. In: Tiruneh, M.W.; Radvanský, M. (eds)
Regional disparities in central and eastern Europe: theoretical models and
empirical analyses: peer-reviewed international conference proceedings.
Bratislava: Institute of Economic Research, Slovak Academy of Sciences.
260
CHAPTER 17.
The protective effect of field-specific and
general skills against overeducation under
different conditions of labour supply and
demand
Martin Humburg (94), Andries de Grip and Rolf van der Velden
Labour supply and demand imbalances lead to educational mismatch

(overeducation) in the labour market. This paper explores the relationship
between graduates’ skills and the risk of overeducation under different
labour demand and supply conditions. We distinguish two types of skills,
field-specific and general skills, as well as two labour market segments, the
occupational domain of a particular field of study and a general
occupational domain. We find that the level of protection field-specific skills
offer against the risk of overeducation increases with the degree of slack in
the occupational domain of the graduate’s field of study. The level of
protection general skills offer against the risk of overeducation increases
with the degree of slack in the general occupational domain.
17.1. Introduction
Labour supply and demand imbalances lead to educational mismatches in the
labour market. When labour supply exceeds demand, workers are at risk of
getting jobs for which they are (formally) overeducated. Overeducation has been
shown to have negative consequences for societies as a whole, but also for the
individuals concerned (Cedefop, 2010; Groot and Maassen van den Brink, 2000;
McGuinness, 2006).
In this article, we make use of a unique international college graduate survey
in 17 European countries to explore how overeducation is related to labour
market entrants’ field-specific and general skills. We thereby attempt to make
inferences about how the relationship between overeducation and field-specific
and general skills is influenced by conditions of labour supply and demand. Our
(94) Corresponding author: m.humburg@maastrichtuniversity.nl

This article and other versions of it have benefited from discussions at the Second
ELM conference at Maastricht University, as well as at the 2010 IAB PhD workshop
in Nuremberg. The authors are grateful to Jim Allen and Ben Kriechel for their useful
comments. Any errors or inaccuracies are, of course, the sole responsibility of the
authors.
261
hypothesis is that higher job competition and selectivity in times of excess

graduate supply causes the protective effect of field-specific and general skills to
increase.
The remainder of this article is structured as follows: Section 17:2 presents
the conceptual framework. In Section 17:3 we discuss our data and in Section
17:4 we present the results of our analysis. Section 17:5 then concludes.
17.2. Conceptual framework and hypotheses

The risk of overeducation of high and low skilled workers tends to converge in
tight labour markets and diverge in slack labour markets. Models brought forward
to explain why low skilled workers experience greater cyclical variation in
overeducation often refer to increasing job competition or the upgrading of hiring
standards during recessions. According to the job competition model (Thurow,
1975), wages are rigid, jobs are ranked according to the level of skills they
require and workers are ranked according to their skills. The highest skilled
worker is then assigned to the best job, the second highest skilled worker to the
second best job, and this process continues until there are either no jobs or no
workers left to be assigned. Economic or demographic shocks affect the length of
the labour queue and the degree to which low skilled workers have to compete
with high skilled workers for particular jobs. When the labour queue is longer than
the job queue the low skilled at the bottom of the queue are pushed into
unemployment. The occupational upgrading hypothesis (Reder, 1955) proposes
a slightly different adjustment mechanism but comes to the same result as the
job competition model. While the job competition model is based on the idea that
employers passively hire the highest skilled worker available, the occupational
upgrading hypothesis assumes that employers actively raise hiring standards in
times of excess labour supply. Both models predict that jobs previously available
for low skilled workers are filled with higher skilled workers in slack labour
markets, pushing the former into unemployment or into jobs for which they are
(formally) overeducated. Teulings and Koopmanschap (1989) show that crowding
out of workers with lower levels of education by workers with higher levels of
education actually took place in the Netherlands during the 1980s, yet, van Ours
and Ridder (1995) argue that job competition was limited to higher versus higher
vocational educated workers. Devereux (2002) finds that the mean level of
education within occupations is increasing during recessions in the US and
concludes that this finding is in line with the occupational upgrading hypothesis
and the predictions of the job competition model. Keane and Prasad (1993) show
262
that workers with college degrees were protected from cyclical variation in
employment in the US in the 1970s.
Traditionally, studies interested in the differences in the cyclicality of labour
market outcomes such as employment and job quality have defined skills as
years of schooling or the level of education (95). In this article, we are interested
in what happens to the risk of overeducation of graduates from higher education
if an economic or demographic shock hits the labour market. We therefore focus
on skill components that distinguish graduates with the same level of education.
We hold the level of education fixed (higher education) and distinguish two types
of skills, field-specific and general skills. These skill types have been shown to
affect the transition from education to work in terms of wages and the risk of
unemployment (96), as well as the risk of overeducation (97).
Besides distinguishing two skill types, we distinguish two labour market
segments: the occupational domain of a particular field of study and the general
occupational domain. We assume that a graduate’s rank in the occupational
domain of his particular field of study is based on his field-specific skills. This
labour market segment contains occupations, which require a very specific set of
skills, such as medical doctors, pilots, or engineers. In the occupational domain
of a particular field of study, field-specific skills are the dominant factor for labour
market success because they are instantly deployable and are associated with
low costs for further field-specific training. Alternatively, graduates have the
choice to work in the general occupational domain. We assume that a graduate’s
rank in the general occupational domain is based on his general skills. This is
because the occupations contained in this labour market segment are broad in
nature and require graduates to have low general training costs. Examples of
general jobs are trainee programmes in large firms, managing positions or
general administrative positions.
The segmentation into a labour market where more field-specific skills are
required and a labour market where more general skills are required has
important implications for the formulation of our hypotheses. Shocks taking place
in the field-specific labour market will affect the relationship between field-specific
skills and labour market outcomes whereas shocks in the general labour market
will affect the relationship between general skills and labour market outcomes.
(95) Devereux (2002); Keane and Prasad (1993); Okun (1981); Teulings and
Koopmanschap (1989); van Ours and Ridder (1995).
(96) Bishop (1995); Campbell and Laughlin (1991); Goux and Maurin (1994); Heijke et al.
(2003); Kang and Bishop (1989); Mane (1999); Payne (1995); Ryan (2001).
(97) Chevalier and Lindley (2009); Verhaest and van der Velden (2010).
263
When the number of jobs at the tertiary level in the occupational domain of a
particular field of study is lower than the number of graduates in this field, due to
an economic and/or demographic shock, graduates with the least field-specific
skills will not necessarily become unemployed, but they will tend to stream into
jobs previously available for medium educated workers of that field. One should
therefore be able to observe that the comparative advantage of having high as
opposed to low field-specific skills increases with the degree of slack in the
labour market. The same mechanism is at work in the occupational domain
demanding more general skilled workers. When the degree of slack in these
general occupations increases, employers will fill their vacancies with the best
workers available and more and more graduates with low levels of general skills
will accept jobs previously available for medium educated workers. When
aggregate unemployment increases, one should therefore observe an increase in
the comparative advantage of having high as opposed to low general skills.
17.3. Data
Figure 17:1 shows that even among college graduates the risk of overeducation
is substantial. Our analysis is based on original and representative data from the
Reflex and Hegesco surveys among graduates from 17 European countries (98).
The questionnaire was sent to higher education graduates five years after
graduation. Our sample contains 11129 individuals.
In the questionnaire, respondents were asked to rate their skill level on a
scale from 1 (very low) to 7 (very high). ‘Mastery of own field or discipline’ refers
to graduates’ level of theoretical and practical knowledge in their own field and to
the ability to apply this knowledge in practice. We use this skill as an indicator for
field-specific skills. ‘Analytical thinking’ refers to the ability to generalise from a
concrete problem to abstract ideas, and to manipulate these ideas in one’s mind
(98) Reflex was conducted in 2005 among 15 European countries and Japan. Hegesco is
the extension of Reflex to four new EU Member States and Turkey conducted in
2009. In our analysis we only focus on European countries to ensure comparability.
We excluded Sweden and Portugal because their survey design substantially
deviated from the rest of the survey. For the remaining countries, we only include
individuals who were less than 36 years old at the time of the survey to avoid
unobserved pre-university labour market experience to be influencing the results.
Moreover, we exclude all individuals who were not living or working in their home
country at the time of the survey or who enrolled in further education after the initial
education they reported on. The number of observations per country varies between
382 and 995.
264
to arrive at a solution, not only to the original problem, but to a whole class of
similar problems. This skill matches our definition general skills quite well.
Figure 17:1 Incidence of overeducation among college graduates five years after
graduation (%)
20%
18%
16%
14%
12%
10%
8%
6%
4%
2%
0%
AT BE CZ EE FI FR DE HU IT LT NL NO PL ES SI CH UK
Source: Own calculations on the basis of Reflex and Hegesco data set. Data of Hungary, Lithuania, Poland
and Slovenia from 2009, the rest of the countries from 2005.
Our definition of the dependent variable is straight forward. We consider

graduates overeducated if they indicated that the type of education most
appropriate for the job they are doing is below the tertiary level (99).
We use the unemployment rate at the time of the surveys as a proxy for
labour market conditions. We include two different unemployment rates in our
regression: the unemployment rate of the respondent’s country (overall
unemployment rate) as well as the unemployment rate of the respondent’s field
of study within that country (unemployment rate of the occupational domain of the
field of study). Both unemployment rates are calculated on the basis of the
combined Reflex and Hegesco data. Their values per country and field of study
are displayed in Table 1 (100). Using international variation for the identification of
(99) For a discussion on measurement issues in assessing overeducation, see Dolton

and Vignoles (2000) and Hartog (2000).
(100) In the regressions, we include the unemployment rate in the occupational domain of
a particular field of study (the field-specific unemployment rate within a country) as
the deviation from the overall unemployment rate (the country-level unemployment
rate), and we include the overall unemployment rate as the deviation from the
unemployment rate of the whole sample. This ensures that both unemployment rates
are uncorrelated and has advantages concerning the interpretation of the regression
results as pointed out in Section 17:4.
265
effects of skills and their interaction with labour market conditions has
advantages over national studies but also obvious limitations (Cedefop, 2010).
The main advantage is that measures generated from international data offer
variation usually unavailable within a single country and provide insights into
long-term, general equilibrium effects. A clear limitation of cross-country, cross-
field evidence is possible omitted country-level and field-level variables, such as
institutional differences in ability sorting or employers’ beliefs.
As control variables we only use variables which influence the probability of
being overeducated because of signalling or network effects but which are not
necessarily outcomes of skills. We include gender, age, age squared, a dummy
whether the father has higher education, a dummy whether the respondent has a
second level higher education degree (101), a dummy whether the respondent had
study-related work experience during higher education and a dummy whether the
respondent had non study-related work experience during higher education.
17.4. Results
Figure 17:2 shows a scatter plot of the coefficient of general skills and the overall
unemployment rate. The coefficients were obtained by countrywise regressing a
dummy variable that takes the value 1 if individuals are overeducated on general
skills and the set of control variables described above. The fitted line suggests
that the protective effect of general skills against overeducation is higher when
aggregate unemployment is higher.
Figure 17:3 presents a scatter plot of the coefficient of field-specific skills
and the unemployment rate of the occupational domain of the field of study. The
coefficients were obtained by regressing a dummy 1 if overeducated on field-
specific skills and a set of control variables per field of study. The fitted line
suggests that the protective effect of field-specific skills against overeducation is
higher when unemployment in the occupational domain of the field of study is
higher.
(101) A second level higher education degree is a degree at ISCED 5A allowing direct
access to doctoral studies.
266
Table 17:1 Overall unemployment rates and unemployment rates in the

occupational domain of particular fields of study per country
Mean E HA SJI BL SMC EMC AV HW S
AT 4.3 2.4 6.1 4.4 4.5 3.6 3.2 5.8 4.8 3.7
BE 2.2 1.9 3.3 2.2 1.1 3.0 2.0 2.1 1.2 1.9
CZ 2.6 3.1 3.9 1.6 1.9 2.6 2.1 2.5 2.6 2.1
EE 1.9 1.0 2.3 1.7 1.2 1.2 1.1 2.3 2.4 2.5
FI 4.4 4.4 4.9 3.3 4.0 4.1 3.1 3.5 5.8 6.5
FR 7.6 6.6 10.5 8.7 8.2 9.7 5.7 7.2 5.9 7.5
DE 4.8 4.7 5.9 4.7 4.1 7.3 4.6 4.1 3.8 4.0
HU 5.0 6.7 4.2 5.3 4.9 4.6 4.5 4.9 4.1 5.9
IT 7.6 9.0 10.2 9.1 7.4 8.4 3.6 8.1 6.6 7.4
LT 3.2 3.8 4.4 2.7 2.7 2.8 2.7 - 2.9 3.0
NL 4.2 2.6 6.0 4.0 3.6 6.1 3.5 4.7 2.8 4.1
NO 2.7 2.4 3.3 3.0 2.1 2.5 2.2 2.6 2.1 3.1
PL 2.2 3.3 2.0 1.5 2.1 2.3 1.7 2.0 1.8 1.8
ES 8.9 9.8 12.5 9.9 7.8 10.3 5.1 11.3 7.5 8.0
SI 3.8 3.9 4.1 3.3 3.6 3.5 2.4 5.1 1.6 5.4
CH 4.4 3.6 5.2 5.3 4.2 6.4 3.2 4.3 3.1 -
UK 3.6 2.7 5.6 3.5 3.3 2.6 3.8 4.0 2.5 3.8
NB: Values are posterior means derived from Reflex/Hegesco data. E=Education, HA=Humanities and Arts,
SJI=Social Sciences, Journalism and Information, BL=Business, Law, SMC=Science, Mathematics and
Computing, EMC=Engineering, Manufacturing and Construction, AV=Agriculture and Veterinary,
HW=Health and Welfare, S=Services.
Figure 17:2 Coefficient of general skills and overall unemployment rate (%)
0.02
DE
0.01
EE BE NO
CH
0
PL NL FR
-0.01
CZ LT FI
-0.02 SI IT
AT
-0.03 HU
UK
-0.04
-0.05
-0.06
ES
-0.07
1 2 3 4 5 6 7 8 9 10
Data: Reflex and Hegesco. AT=Austria, BE=Belgium, CH=Switzerland, CZ=Czech Republic, DE=Germany,
EE=Estonia, ES=Spain, FI=Finland, FR=France, HU=Hungary, IT=Italy, LT=Lithuania,
NL=Netherlands, NO=Norway, PL=Poland, SI=Slovenia, UK=United Kingdom.
267
Figure 17:3 Coefficient of field-specific skills and unemployment rate in

occupational domain of field of study
0.02
AV
0.01
S
0
BL
EMC
-0.01
HW
SJI
-0.02 HA
E SMC
-0.03
2 3 4 5 6 7 8
Data: Reflex and Hegesco. E=Education, HA=Humanities and Arts, SJI=Social Sciences, Journalism and
Information, BL=Business, Law, SMC=Science, Mathematics and Computing, EMC=Engineering,
Manufacturing and Construction, AV=Agriculture and Veterinary, HW=Health and Welfare,
S=Services.
The regression results presented in Table 17:2 confirm that these

relationships are statistically significant. Model 1 shows that, on average, both
skill types are negatively related to the probability of being overeducated (102).
Holding all other things constant, a one standard deviation higher endowment of
field-specific skills reduces the risk of being overeducated by 1.3%. A one
standard deviation higher endowment of general skills reduces the risk of being
overeducated by 1.7%. Both unemployment rates are positively related to the
probability of being overeducated for one’s job. Moreover, as expected, both the
coefficient of the interaction of general skills (analytical thinking) and the overall
unemployment rate as well as the coefficient of the interaction of field-specific
skills (mastery of own field) and the unemployment rate in the occupational
domain of the field of study are negative and significant at the 1% level in
model 2. This indicates that when the degree of slack in the general occupational
(102) We also tested if graduates working in jobs that match their level of education have
higher skills because they receive more training than those who are overeducated
(van Smoorenburg and van der Velden, 2000). We therefore reran our estimation
once with hours of training in the past four weeks and once with a dummy indicating
the participation in training in the last 12 months to take account of the effect of
training on skills. Including these variables did not substantially change the
coefficients of our variables of interest, indicating that our skill variables are not
picking up training effects.
268
domain increases, the level of protection against the risk of overeducation

general skills offer graduates also increases.
Table 17:2 Linear probability model of being overeducated five years after
graduation
Model 1 Model 2
-0.013*** -0.014***
Mastery of own field (standardised)
(0.003) (0.003)
-0.017*** -0.019***
Analytical thinking (standardised)
(0.003) (0.003)
0.016*** 0.016***
Overall unemployment rate
(0.001) (0.001)
Unemployment rate in occupational domain of 0.013*** 0.013***
field of study (0.003) (0.003)
Overall unemployment rate X -0.007***
Analytical thinking (0.002)
Unemployment rate in occupational domain of -0.008***
field of study X Mastery of own field (0.003)
Controls yes yes
R² 0.040 0.043
N 11 129 11 129
NB: Coefficients reported are estimates from a linear regression of a dummy 1 if overeducated on our
independent variables, robust standard errors in parentheses
(significance levels *** 0.01, ** 0.05, * 0.1).
Controls included are gender, age, age squared, father having higher education, respondent having a
second level degree, study related work experience during higher education, and non study related work
experience during higher education.
Data: Reflex/Hegesco.
Moreover, the level of protection against the risk of overeducation field-

specific skills offer graduates increases when the unemployment rate in the
occupational domain of the field of study rises.
17.5. Conclusion
In this article, we investigated the relationship between graduates’ field-specific
and general skills and the risk of being employed in a job for which they are
overeducated.
One of the main arguments brought forward by this paper is that two labour
market segments can be distinguished: a labour market segment where field-
specific skills determine the allocation of graduates to jobs (the occupational
domain of a particular field of study) and a labour market segment where general
skills determine the allocation of graduates to jobs (the general occupational
domain). From this followed that heterogeneity concerning the effect of field-
269
specific skills on overeducation should stem from variation in excess supply of

graduates in the occupational domain of the corresponding field of study whereas
heterogeneity concerning the effect of general skills on overeducation should
stem from variation in excess supply of graduates in the general occupational
domain.
We found that graduates with low levels of field-specific or general skills
experience greater cyclical variation in job quality (overeducation). More
precisely, the probability of being overeducated of graduates with high and low
levels of field-specific skills converges when unemployment in the occupational
domain of the field of study is low and diverges when the degree of slack in this
specific labour market segment rises. The probability of being overeducated of
graduates with high and low levels of general skills converges when overall
unemployment is low and diverges when general unemployment rises.
Besides looking at differences in the cyclical employment and job quality
variation between graduates with different skill endowments, we explored the
average relationship between field-specific and general skills and graduates’
probability of being overeducated. Our study showed that skills matter.
Graduates with high field-specific skills as well as graduates with high general
skills (or both) were more often allocated to jobs matching their level of education
than graduates with low levels of field-specific and general skills.
The results of our study are encouraging. We realise, however, that the
effects we find could be driven by institutional differences between countries or
fields of study. We therefore think that using cross-sectional data to test our
conceptual framework can only be a first step. Further research exploiting cyclical
variation within fields and within countries over time is needed to better establish
causality.
References
Bishop, J. (1995). Expertise and excellence. Ithaca, NY: Cornell University
(CAHRS Working paper series; 95-13).
Campbell, P.B.; Laughlin, S. (1991). Participation in vocational education: an
overview of patterns and their outcomes. Columbus, OH: National centre for
research in vocational education.
Cedefop (2010). The skill matching challenge: analysing skill mismatch and
policy implications. Luxembourg: Publications Office of the European Union.
Chevalier, A.; Lindley, J. (2009). Overeducation and the skills of UK graduates.
Journal of the Royal statistical society, Series A (Statistics in society), Vol.
172, No 2, p. 307-337.
270
Devereux, P.J. (2002). Occupational upgrading and the business cycle. Labour,
Vol. 16, No 3, p. 423-452.
Dolton, P.; Vignoles, A. (2000). The incidence and effects of overeducation in the
UK graduate labour market. Economics of education review, Vol. 19, No 2,
p. 179-198.
Goux, D.; Maurin, E. (1994). Éducation, expérience et salaire: tendances récentes
et évolution de long terme [Education, experience and earnings: recent trends
and long-term developments]. Économie et Prévision, Vol. 116, No 5, p. 155-178.
Groot, W.; Maassen van den Brink, H. (2000). Overeducation in the labour
market: a meta-analysis. Economics of education review, Vol. 19, No 2,
p. 149-158.
Hartog, J. (2000). Over-education and earnings: where are we, where should we
go? Economics of education review, Vol. 19, No 2, p. 131-147.
Heijke, H. et al. (2003). Fitting to the job: the role of generic and vocational
competences in adjustment and performance. Labour economics, Vol. 10,
No 2, p. 215-229.
Kang, S.; Bishop, J. (1989). Vocational and academic education in high school:
complements or substitutes? Economics of education review, Vol. 8, No 2,
p. 133-148.
Keane, M.; Prasad, E. (1993). Skill levels and the cyclical variability of
employment, hours, and wages. IMF staff papers, Vol. 40, No 4, p. 711-743.
Mane, F. (1999). Trends in the payoff to academic and occupation-specific skills:
the short and medium run returns to academic and vocational high school
courses for non-college-bound students. Economics of education review,
Vol. 18, No 4, p. 417-437.
McGuinness, S. (2006). Overeducation in the labour market. Journal of economic
surveys, Vol. 20, No 3, p. 387-418.
Okun, A.M. (1981). Prices and quantities: a macroeconomic analysis.
Washington DC: The Brookings Institution.
Payne, J. (1995). Options at 16 and outcomes at 24: a comparison of academic
and vocational education and training routes. Sheffield: Dept. for Education
and Employment (Youth Cohort report series; No 35).
Reder, M.W. (1955). The theory of occupational wage differentials. The American
Economic Review, Vol. 45, No 5, p. 833-852.
Ryan, P. (2001). The school-to-work transition: a cross-national perspective.
Journal of economic literature, Vol. 39, No 1, p. 34-92.
Teulings, C.; Koopmanschap, M. (1989). An econometric model of crowding out of
lower education levels. European economic review, Vol. 33, No 8, p. 1653-1664.
Thurow, L. C. (1975). Generating inequality. New York: Basic Books.
van Ours, J.C.; Ridder, G. (1995). Job matching and job competition: are lower
educated workers at the back of job queues? European economic review,
Vol. 39, No 9, p. 1717-1731.
271
van Smoorenburg, M.S.M.; van der Velden, R.K.W. (2000). The training of school-
leavers: complementarity or substitution? Economics of education review,
Vol. 19, No 2, p. 207-217.
Verhaest, D.; van der Velden, R.K.W. (2010). Cross-country differences in graduate
overeducation and its persistence. Research memoranda, No 7. Maastricht:
ROA, Research centre for education and the labour market.
272
List of contributors
Jacques Babel
Swiss Federal Statistical Office
Neuchâtel, Switzerland
jacques.babel@bfs.admin.ch
http://www.bfs.admin.ch/
Daniel Bacher
3s Unternehmensberatung GmbH
Vienna, Austria
bacher@3s.co.at
www.3s.co.at
Roger Bjørnstad
Statistics Norway
Oslo, Norway
Roger.Bjornstad@ssb.no
http://www.ssb.no/
Jiří Braňka
National Training Fund
Prague, Czech Republic
branka@nvf.cz
http://www.nvf.cz/
Giuseppe Ciccarone
Sapienza University of Rome &
Fondazione G. Brodolini
Rome, Italy
giuseppe.ciccarone@uniroma1.it
http://uniroma1.it/
www.fondazionebrodolini.it
Andries de Grip
Research Center for Education and the Labour Market (ROA)
Maastricht, the Netherlands
IZA
Bonn, Germany
A.deGrip@maastrichtuniversity.nl
http://www.roa.unimaas.nl/
http://www.iza.org/en/webcontent/index_html
Artur Gajdos
University of Lodz
Lodz, Poland
gajdos@uni.lodz.pl
http://www.uni.lodz.pl/
273
Marit L. Gjelsvik
Statistics Norway
Oslo, Norway
http://www.ssb.no/
Anna Godøy
Statistics Norway
Oslo, Norway
http://www.ssb.no/
Inger Holm
Statistics Norway
Oslo, Norway
http://www.ssb.no/
Martin Humburg
m.humburg@maastrichtuniversity.nl
Stefan Humpl
3s Unternehmensberatung GmbH
Vienna, Austria
humpl@3s.co.at
www.3s.co.at
Ben Kriechel
Maastricht University / ROA
Maastricht, The Netherlands
ben.kriechel@maastrichtuniversity.nl
Kvetan Vladimir
European Centre for the Development of Vocational Training (Cedefop)
Thessaloniki, Greece
vladimir.kvetan@cedefop.europa.eu
http://www.cedefop.europa.eu/EN/index.aspx
Erkki Laukkanen
SAK
Helsinki, Finland
erkki.laukkanen@sak.fi
http://www.sak.fi/english/whatsnew.jsp?location1=1&lang=en
Tom Leney
Institute of Education
London, United Kingdom
TomLeney@btinternet.com
http://www.ioe.ac.uk/
274
Tobias Maier
Federal Institute for Vocational Education and Training
Bonn, Germany
tobias.maier@bibb.de
http://www.bibb.de/
Public Policy and Management Institute
Vilnius, Lithuania
zilvinas@vpvi.lt
http://www.vpvi.lt/
Hector Pollitt
Cambridge Econometrics
Cambridge, United Kingdom
hp@camecon.com
http://www.camecon.com/Home.aspx
Marek Radvanský
Institute of Economic Research, Slovak Academy of Sciences
Bratislava, Slovakia
marek.radvansky@savba.sk
http://www.sav.sk/
Antonio Ranieri
antonio.ranieri@cedefop.europa.eu
Regional Observatory of Training and Employment - GIP Alfa Centre
Orléans, France
mb.rochard@alfacentre.org
http://www.alfacentre.org/
Miroslav Štefánik
Institute of Economic Research, Slovak Academy of Sciences
Bratislava, Slovakia
miroslav.stefanik@savba.sk
http://www.sav.sk/
Nils Martin Stølen

Statistics Norway
Oslo, Norway
nils.martin.stolen@ssb.no
http://www.ssb.no/
275
Massimiliano Tancioni
Sapienza University of Rome &
Fondazione G. Brodolini
Rome, Italy
massimiliano.tancioni@uniroma1.it
http://uniroma1.it/
www.fondazionebrodolini.it
Rolf van der Velden

R.vanderVelden@Maastrichtuniversity.nl
Rob Wilson
Institute for Employment Research
University of Warwick
Coventry, United Kingdom
r.a.wilson@warwick.ac.uk
http://www2.warwick.ac.uk/
Alena Zukersteinova
alena.zukersteinova@cedefop.europa.eu
276
Building on skills forecasts —
Comparing methods and
applications
Luxembourg:
Publications Office of the European Union
2012 – VI, 276 p. – 21 x 29.7 cm
ISBN 978-92-896-0892-3
ISSN 1831-5860
doi: 10.2801/36961
Cat. No: TI-BC-11-010-EN-N
Free download at:

http://www.cedefop.europa.eu/EN/Files/5518_en.pdf
Free of charge – 5518 EN –

Building on skills forecasts — Comparing
TI-BC-11-010-EN-N
methods and applications
This publication is based on contributions to the international expert

conference, Building on the skills forecast: comparing methods and
applications, organised by Cedefop in February 2011, which brought
together a forum of more than 60 researchers and experts in labour-
market analysis. This inspiring and constructive event aimed at
sharing the latest insights on how Cedefop’s forecasting results are
used at national level, and what other (innovative) activities are taking
place in individual Member States.
The publication helps to identify the challenges and offer solutions to
skills and labour-market forecasting in general and to Cedefop
projections in particular; it also discusses various national approaches,
methods and results and their comparisons. It provides sound
evidence and suggests innovative ways of thinking about future skills
needs and supply. The findings will inspire further development of
Cedefop methodology used to project future skill demand and supply.
Europe 123, 570 01 Thessaloniki (Pylea), GREECE

Postal address: PO Box 22427, 551 02 Thessaloniki, GREECE
Tel. +30 2310490111, Fax +30 2310490020
E-mail: info@cedefop.europa.eu
visit our portal www.cedefop.europa.eu
5518 EN
http://www.cedefop.europa.eu/EN/Files/5518_en.pdf

5518 en

Uploaded by

Copyright:

Available Formats

5518 en

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

5518 en

Uploaded by

Copyright:

Available Formats

European Centre for the Development

Luxembourg: Publications Office of the European Union, 2012

Europe 123, 570 01 Thessaloniki (Pylea), GREECE

Christian F. Lettmayr, Acting Director

This publication results from proceedings of the international expert conference

Note from the editors

PART 1: BUILDING ON CEDEFOP’S FORECASTING RESULTS .................................. 36

PART 2: SHARING OTHER EUROPEAN FORECASTING EXPERIENCES ......................... 81

11. The forecast of occupational structure of employment in Poland

List of tables and figures

15:2 Conversion of classification of education classes (ISCED) ..................... 227

4:8 Concentration index of non-national workers in elementary

12:5 Number of first degrees from professional education and training

15:18 Share of overqualified with high education, 2009 .................................... 243

1.2. Methodological framework and approach

The first module is a multi-sectoral macroeconomic model (E3ME) which is

(2) More detailed information is available from Internet:

Source: Cedefop (2010).

The interaction between labour supply and demand by qualifications is a

1.3. Results: 2011 update on future skills supply and

1.3.1. Baseline results

Figure 1:2 Impact of the recession on employment, EU-27+

Source: Cedefop (2011a).

Primary sector & utilities

Distribution & transport

Business & other services

Source: Cedefop (2011a).

Figure 1:4 Future job opportunities by occupation, EU-27+

Legislators, senior officials and managers

Technicians and associate professionals

Service, shop and market sales workers

Skilled agricultural and fishery workers

Craft and related trades workers

Plant and machine operators and assemblers

Source: Cedefop (2011a).

qualifications is projected to decrease. Those jobs requiring intermediate level

Figure 1:5 Future job opportunities by broad level of qualifications, EU-27+

Source: Cedefop (2011a).

Future labour supply will be influenced by two important facts:

Figure 1:6 Labour supply by level of qualifications (EU-27+)

Source: Cedefop (2011a).

1.3.2. Alternative scenarios

It is important to separate the short-term and long-term effects. In the short

Table 1:1 Main impacts on employment in the different scenarios

1.4. Conclusions and the way forward

Many technical alternatives to the current modelling approach have been

As part of the new Europe 2020 strategy, Cedefop is mandated to produce

2.2. National forecasts

(4) Available from Internet:

The Italian model, presented by Massimiliano Tancioni, is a macro (dynamic

2.3. Comparison and evaluation studies

2.5. Presentation of skill supply and demand forecasts

2.6. Technical responses to the presentations and

2.6.1. Historical data

Figure 2:1 Overview of (Cedefop) existing approach

2.6.2. Forward-looking inputs

2.6.3. Expert judgment

pan-European approach is not able to make use of local information, expected

2.7. Limitations of the modelling