Chain Indices in The National Accounts
Chain Indices in The National Accounts
Chain Indices in The National Accounts
Voorburg/Heerlen, 1997
NA-087
Kengetal: P30/1997-3
ISSN: 1385-1845
An earlier version of this paper was presented at the Conference on Measurement problems and
econometric modelling, Istituto Nazionale di Statistica, Rome, 13-14 January 1997.
The authors would like to thank Gert den Bakker and Steven Keuning for their useful comments
on earlier versions of the paper.
Voorburg/Heerlen, 1997
Statistics Netherlands
Voorburg
Heerlen
Address:
Prinses Beatrixlaan 428
Address:
Kloosterweg 1
Postal address:
PO Box 4000
2270 JM Voorburg
Postal address:
PO Box 4481
6401 CZ Heerlen
Internet
http://www.cbs.nl
Information desks
Telephone
Fax
(045) 570 70 70
(045) 570 62 68
infoserv@cbs.nl
Special topics
Labour and wages
Population
Construction
Consumer price index (inflation)
Culture, tourism and recreation
Manufacturing
International Trade
Agriculture
Environment
Education
Government
National Accounts
(070) 337 58 50
(070) 337 58 30
(070) 337 42 41
(070) 337 58 09
(070) 337 58 67
(045) 570 76 17
(045) 570 79 17
(070) 337 58 03
(070) 337 58 96
(070) 337 53 45
(070) 337 58 99
(070) 337 58 76
(070) 337 59 94
(070) 337 59 87
(070) 337 59 75
(070) 337 59 94
(070) 337 59 96
(045) 570 62 77
(045) 570 66 75
(070) 337 59 51
(070) 337 59 76
(070) 337 59 78
(070) 337 59 80
(070) 337 59 81
infosec@cbs.nl
infosbv@cbs.nl
infolbn@cbs.nl
infosec@cbs.nl
infokcr@cbs.nl
infolin@cbs.nl
infohih@cbs.nl
infollb@cbs.nl
infolmi@cbs.nl
infosoz@cbs.nl
infokov@cbs.nl
infopni@cbs.nl
gegevens ontbreken
voorlopig cijfer
geheim
nihil
(indien voorkomend tussen twee getallen) tot en met
het getal is minder dan de helft van de gekozen eenheid
een cijfer kan op logische gronden niet voorkomen
1996 tot en met 1997
het gemiddelde over de jaren 1996 tot en met 1997
oogstjaar boekjaar schooljaar enz. beginnend in 1996 en eindigend in 1997
boekjaar enz. 1986/'87 tot en met 1996/'97
In geval van afronding kan het voorkomen dat de totalen niet geheel overeenstemmen met de
som der opgetelde getallen.
Verbeterde cijfers in staten en tabellen zijn niet als zodanig gekenmerkt.
Explanation of symbols
data not available
provisional figure
publication prohibited (confidential figure)
nil
(between two figures) inclusive
less than half of unit employed
category not applicable
1996 to 1997 inclusive
average for the years 1996 up to and including 1997
crop year financial year school year etc. beginning in 1996 and terminating
in 1997
1986/'87-1996/'97 = book year etc. 1986/'87 up to and including 1996/'97
*
x
0 (00)
a blank
1996-1997
1996/1997
1996/'97
=
=
=
=
=
=
=
=
=
=
Abstract
Substantial part of the work on national accounts is the decomposition of value
changes of economic variables into volume changes and price changes. Nearly
all variables in the national accounts are aggregates or aggregates of
aggregates. Various methods are available to solve the problem how to weight
together the volume and price changes of individual goods and services into
price and volume changes of aggregates. A very important issue is the choice
between fixed weights and annually changing weights.
From 1980 onwards Statistics Netherlands uses annually changing weights and
chain indices derived by the multiplication of the year-to-year changes. In this
paper the history and use of chain indices in the Netherlands is briefly
discussed. In the discussions preceding the introduction of chain indices,
Netherlands Bureau for Economic Policy Analysis (CPB), the most important
user of national accounts data, played an important part making a plea for
annually changing weights and chain indices.
Four users of national accounts data, among them CPB, have been asked about
their opinion on the choice between fixed weight indices and chain indices. All of
them express that chain indices are a good base for the construction of
economic models, since changing weights guarantee a near approximation of
actual developments and the actual economic structure. However, special
attention should be paid to the tuning of the model to the characteristics of the
data and to the presentation of model results to the public.
Also a sensitivity analysis is carried out on the choice of index number formulae
and weights. Alternative volume estimates are presented for the periods 19861993 and 1921-1939 applying Laspeyres and Paasche fixed weight formulae
and Laspeyres, Paasche and Fisher chain formulae.
Taking the Fisher chain indices as standard, the most important conclusions
are:
both chain Laspeyres and chain Paasche volume growth rates and indices
are good approximations of chain Fisher indices in nearly all cases;
most substantial deviations from the chain Fisher indices are found when
using fixed weight Laspeyres formulae.
Table of Contents
Abstract
Table of Contents
1. Introduction
11
13
15
17
19
21
7. Non-additivity
31
8. Concluding remarks
33
References
35
Annex
37
39
41
1. Introduction
An important purpose of the compilation of national accounts is measuring
changes in economic variables. Changes in the production and use of goods
and services are caused by a combination of two factors: a change in price and
a change in quantity and quality (in national accounts often denoted as: volume
change).
Part of the work on national accounts is the decomposition of value changes into
volume changes and price changes. The most important purpose is the estimate
of real growth rates (volume indices). The second goal is the estimate of price
changes (deflators).
An important characteristic of this work when imbedded in a national accounts
framework is that volume indices and deflators of various variables and at
different levels of aggregation are interrelated in a systematic way. This is to be
achieved by using supply and use tables or input-output tables as an integrating
and balancing framework.
Price and volume indices of aggregates are always compiled from price and
volume indices of individual goods and services. Direct observation of price and
volume changes of aggregates is - by definition - impossible. Nearly all items in
the national accounts are aggregates or aggregates of aggregates. Examples of
the latter are total household consumption expenditure, total imports and total
exports, but also total output and total intermediate consumption of industries.
So the question rises how price and volume changes of individual goods can be
added to price and volume changes of aggregates. Various methods are
available to solve this problem. Different index formulae are available with
different weighting schemes. So one has to make a choice. It is beyond the
scope of this paper to discuss in depth the theoretical and practical
considerations with respect to this choice. For a broader discussion of the
theoretical and practical considerations with respect to the choice of index
number formulae and weighting schemes the reader is referred to e.g. Allen
(1975) and to Al et al. (1986).
Besides aggregates, special attention must be paid to a number of balancing
items. Most important of course is GDP. GDP is the difference between two
aggregates: total output value and total intermediate consumption of goods and
services by all industries. Hence price and volume changes of GDP are the
result of the price and volume changes of total output and of total intermediate
consumption of all industries together. Since total output and total intermediate
consumption are aggregates whose volume and price changes are a result of
changes at a detailed level, price and volume changes of GDP are also a result
of changes at a detailed level.
11
13
15
16
17
The advantage of using chain indices and the relatively minor disadvantage of
the resulting non-additivity of constant price series has also been elaborated by
Al et al. (1986).
At present, the annual National accounts publications of Statistics Netherlands
contain the following time series: values in current prices, annual volume growth
rates, volume indices, annual price changes, indices of deflators and values in
prices of a base year (recently: 1990). The series of volume indices and
deflators are chain indices derived from multiplying the annual changes. The
values in prices of a constant base year are calculated with the help of chained
volume indices. Hence in this case, sub-series are non-additive to aggregates in
our publication.
18
19
21
Laspeyres
(weights
1986)
Laspeyres
(chain)
Fisher
(chain)
Paasche
(chain)
Paasche
(weights
t,1986)
1.4
2.6
4.6
4.0
2.2
2.0
0.8
1.4
2.6
4.4
3.9
2.2
2.0
0.8
1.4
2.7
4.6
3.5
2.0
2.0
0.7
2.7
0.8
3.5
4.2
3.1
2.5
1.0
2.7
0.8
3.5
4.2
3.1
2.5
0.9
2.7
0.8
3.5
4.1
2.9
2.4
0.9
0.9
4.5
4.9
1.6
0.2
0.6
-2.8
0.9
4.5
4.9
1.5
0.2
0.6
-2.9
0.9
4.5
4.8
1.4
-0.1
0.4
-3.0
3.9
9.0
6.5
5.3
4.7
2.8
1.4
3.7
9.0
6.3
5.2
4.6
2.8
1.4
3.7
9.1
6.2
5.0
4.7
2.9
1.4
4.0
7.5
6.7
4.2
4.1
2.0
-2.2
3.9
7.4
6.6
4.1
4.1
2.0
-2.3
3.9
7.4
6.4
4.0
3.9
2.1
-2.4
The results in the column Paasche (weights t, 1986) are obtained by deflation
with fixed base (1986) Laspeyres deflators and henceforth will be called fixed
weight Paasche volume indices.
At first sight the differences between the results for different formulae seem to be
rather small. However, one must take into consideration that they concern large
macro totals. In the Netherlands every 0.1 % of GDP equals about 600 million
guilders (280 million ECU). Secondly, if the differences every year show the
same sign (- or +) in a time series, yearly differences, however small, can cause
serious long term deviations. Thirdly, only one year with a larger deviation can
cause large deviations in all succeeding years of a time series.
When discussing the results we will consider the chain Fisher volume indices as
the standard with which we compare other results. The reason is that Fisher
indices can be seen as good approximations of superlative indices that can be
seen as ideal indices from a theoretical point of view (see also section 4).
22
The sequence of the columns of the tables in this paper has been chosen such
that a certain following order of results could be expected: highest growth rates
are expected with the fixed weight Laspeyres formula and lowest with the fixed
weight Paasche formula and the results of the chain formulae are expected to be
in between.
The results in table 1 show that most substantial divergences are with Laspeyres
fixed weight indices. The Laspeyres, Fisher and Paasche chain type indices in
most cases yield comparable results and the differences between the Paasche
fixed weight and chain indices are generally speaking much smaller than
between the Laspeyres fixed weight and chain indices. A remarkable and for the
statistical practice very important conclusion which can be drawn is that, as ESA
states, the Laspeyres chain indices provide good approximations of the ideal
Fisher chain indices and that the Laspeyres fixed weight indices do less or not at
all. This is especially important for the cases where the differences between
Laspeyres and Paasche fixed weight indices and between the Laspeyres fixed
weight index and the Laspeyres chain index are substantial. See: GDP (1988
and 1993), Consumption of households (1991), Capital formation (1992 and
1993), Exports (1988, 1989, 1990 and 1993), Imports (1990, 1991 and 1993).
Generally speaking the results in table 1 also bear out the statement of SNA that
Laspeyres volume indices yield higher growth rates than Paasche volume
indices and that the differences between the Laspeyres and Paasche chain
indices are smaller than between the corresponding fixed weight indices.
Table 2 presents the time series of volume indices corresponding with the
annual growth rates in table 1. It will not be surprising that the conclusions that
can be drawn from the results in table 1 and 2 are quite similar.
23
Laspeyres
(weights
1986
Laspeyres
(chain)
Fisher
(chain)
Paasche
(chain)
Paasche
(weights
t,1986)
101.4
104.0
108.8
113.1
115.6
117.9
118.9
101.4
104.0
108.6
112.8
115.3
117.5
118.5
101.4
104.1
108.9
112.7
115.0
117.3
118.2
102.7
103.6
107.2
111.7
115.1
118.0
119.2
102.7
103.6
107.1
111.6
115.0
117.9
119.1
102.7
103.5
107.1
111.5
114.7
117.4
118.5
100.9
105.5
110.7
112.4
112.6
113.3
110.1
100.9
105.5
110.7
112.4
112.6
113.2
109.9
100.9
105.5
110.6
112.1
112.0
112.4
109.0
103.9
113.2
120.5
126.9
132.8
136.6
138.5
103.7
113.0
120.1
126.4
132.2
135.9
137.8
103.7
113.1
120.1
126.1
132.0
135.9
137.7
104.0
111.8
119.3
124.2
129.3
132.0
129.1
103.9
111.6
118.9
123.8
128.9
131.4
128.5
103.9
111.6
118.8
123.5
128.3
131.0
127.9
Tables 3.a, 3.b and 3.c give the results of alternative calculations of the annual
volume growth rates of total output, total intermediate consumption and value
added for three industries which play quite a different role in the Dutch economy.
Agriculture and horticulture
This is an industry which includes very different activities. As a result of different
volume and price changes of output and input in the underlying sub-industries
(livestock raising, arable farming and horticulture) the average production
structure of total agriculture can change rapidly. Hence, one can expect rather
substantial deviations of volume indices calculated with different index number
formulae. Especially for value added, which is, as the balancing item of output
and intermediate consumption, influenced by changes in both output and input.
Table 3.a shows that, with some exceptions for 1991 and 1993, deviations in
output and intermediate consumption are rather modest. Differences in the
results are more substantial with value added. For 1987-1991 chain indices give
24
Laspeyres
(weights
1986)
Laspeyres
(chain)
Gross output
1987
-1.4
1988
1.9
1989
3.3
1990
3.8
1991
2.2
1992
2.0
1993
0.5
Intermediate consumption
1987
0.7
1988
-0.5
1989
-0.0
1990
-1.3
1991
1.2
1992
0.0
1993
-0.5
Value added (basic prices)
1987
-4.2
1988
5.3
1989
7.8
1990
10.0
1991
3.2
1992
4.2
1993
1.6
Fisher
(chain)
Paasche
(chain)
Paasche
(weights
t,1986)
-1.4
2.0
3.4
3.5
2.0
2.4
0.8
-1.6
1.9
3.2
3.4
1.6
2.1
0.4
-1.8
1.8
3.0
3.2
1.3
1.8
-0.1
-1.8
1.7
2.9
3.5
1.1
1.8
0.6
0.7
-0.4
0.2
-1.6
1.2
0.1
-0.6
0.6
-0.5
0.1
-1.5
1.3
0.0
-0.6
0.5
-0.5
0.1
-1.5
1.3
-0.1
-0.7
0.5
-0.4
0.2
-1.4
1.4
0.1
-0.3
-4.2
5.2
7.5
9.5
2.8
4.9
2.5
-4.5
5.1
7.1
9.2
2.0
4.4
1.5
-4.7
5.0
6.6
8.9
1.2
4.0
0.6
-4.7
4.4
6.4
9.5
0.8
4.6
3.2
Petro-chemical industry
This industry is rather sensitive to fluctuations on the international markets.
Because it absorbs for the greater part semi-manufactured products from oil
refineries, it is also relatively sensitive to substantial price fluctuations.
25
Laspeyres
(weights
1986)
Laspeyres
(chain)
Gross output
1987
4.8
1988
8.9
1989
1.7
1990
3.1
1991
-4.1
1992
-2.2
1993
-0.8
Intermediate consumption
1987
6.2
1988
12.5
1989
-0.9
1990
0.9
1991
-4.6
1992
-2.0
1993
-3.1
Value added (basic prices)
1987
1.2
1988
-0.6
1989
9.6
1990
9.0
1991
-2.8
1992
-2.8
1993
5.1
Fisher
(chain)
Paasche
(chain)
Paasche
(weights
t,1986)
4.8
9.7
0.6
3.3
-5.3
-2.6
-1.8
3.9
9.7
0.8
2.9
-5.4
-2.6
-1.8
3.1
9.7
0.9
2.6
-5.4
-2.6
-1.9
3.1
9.7
0.2
3.2
-5.3
-3.3
-2.2
6.2
11.5
-1.3
1.5
-4.3
-2.9
-2.9
5.9
11.3
-2.3
1.7
-4.4
-3.0
-3.0
5.7
11.2
-3.4
1.8
-4.5
-3.2
-3.1
5.7
11.5
-4.1
1.6
-4.6
-3.5
-3.4
1.2
5.6
4.0
6.7
-8.2
-1.2
2.7
-0.8
6.3
7.3
5.8
-8.3
-0.9
2.8
-2.7
7.0
10.8
4.9
-8.4
-0.1
2.9
-2.7
8.1
8.1
8.0
-6.7
-2.1
2.8
The consequence is that the resulting volume growth rates from different index
number formulae show larger deviations (see: table 3.b). It should be noted that
the fixed weight Paasche indices as well as the chain Laspeyres and the chain
Paasche indices sometimes also show larger deviations. However, table 3.b
shows that larger deviations again most frequently occur with the fixed weight
Laspeyres indices. As could be expected the most substantial deviations are
found with value added (a balancing item).
Construction of buildings and dwellings
This industry shows strong fluctuations in volume growth rates. However price
movements and changes in production structure are rather gradual. This means
that the estimates of volume growth rates are less dependent on the type of
index number formula.
26
Laspeyres
(weights
1986)
Laspeyres
(chain)
Gross output
1987
5.1
1988
12.2
1989
4.6
1990
0.5
1991
0.5
1992
-2.2
1993
-4.4
Intermediate consumption
1987
6.7
1988
13.4
1989
5.6
1990
0.6
1991
1.7
1992
-1.6
1993
-3.0
Value added (basic prices)
1987
1.4
1988
9.5
1989
2.1
1990
0.2
1991
-2.5
1992
-3.7
1993
-8.1
Fisher
(chain)
Paasche
(chain)
Paasche
(weights
t,1986)
5.1
12.2
4.6
0.5
0.5
-2.2
-4.4
5.1
12.3
4.6
0.5
0.5
-2.2
-4.4
5.1
12.3
4.6
0.5
0.5
-2.2
-4.4
5.1
12.3
4.6
0.5
0.6
-2.2
-4.4
6.7
13.4
5.6
0.6
1.7
-1.6
-3.1
6.7
13.4
5.6
0.6
1.7
-1.6
-3.2
6.7
13.4
5.5
0.6
1.7
-1.7
-3.2
6.7
13.4
5.5
0.5
1.8
-1.8
-3.3
1.4
9.6
2.2
0.1
-2.4
-3.8
-7.4
1.4
9.6
2.2
0.1
-2.3
-3.7
-7.2
1.4
9.7
2.3
0.2
-2.3
-3.6
-7.0
1.4
9.7
2.3
0.2
-2.3
-2.8
-6.6
Table 3.c shows that gross output and intermediate consumption yield only very
small (or no) differences between the results of different formulae. The same
holds in most cases for value added. However for 1993 both fixed weight indices
show substantial deviations from the three chain indices.
Period 1921-1939
The interwar figures provide a very interesting data set for testing the differences
between growth rates arising from the use of different index number formulae
and weighting schemes. The reason is that the interwar period is characterised
by large fluctuations in economic growth rates and price movements. During the
years 1921-1939 periods of economic growth and decline alternated.
Table 4.a gives the annual growth rates of GDP during the interwar period
estimated with different index number formulae. The results are derived from an
earlier study of den Bakker (1991).
27
Laspeyres
(weights
1921)
Laspeyres
(chain)
Fisher
(chain)
Paasche
(chain)
Paasche
(weights
t,1921)
1922
1923
1924
1925
1926
1927
1928
1929
.5
2.7
8.1
3.9
7.5
5.1
5.3
2.3
6.5
2.1
7.2
3.6
6.8
4.6
4.9
2.4
6.1
2.0
7.3
3.3
6.7
4.6
4.8
2.3
5.6
1.9
7.5
3.1
6.6
4.5
4.8
2.3
5.6
1.7
7.8
3.1
6.0
4.2
4.7
2.9
1922-29
5.2
4.7
4.6
4.5
4.5
1930
1931
1932
1933
1934
-1.6
-5.4
-1.2
-0.2
-0.2
-1.2
-4.6
-0.9
0.4
-1.1
-1.1
-4.5
-1.1
0.4
-1.3
-1.1
-4.4
-1.3
0.3
-1.4
-1.7
-5.4
-2.3
0.6
-1.5
1930-34
-1.7
-1.5
-1.5
-1.6
-2.1
1935
1936
1937
1938
1939
4.4
6.4
7.2
-4.4
9.1
2.7
5.5
5.7
-3.2
8.1
3.0
5.3
5.8
-2.9
8.3
3.2
5.2
5.9
-2.7
8.6
2.9
5.6
7.0
-2.9
9.8
4.5
3.8
3.9
4.0
4.5
1935-39
The years 1921/29 and 1934/39 are periods of growth. Table 4.a shows that for
those years (except 1929) the fixed weight Laspeyres indices give higher volume
growth rates than the Paasche indices and also higher growth rates than the
chain Laspeyres indices. Furthermore it is quite clear that, just as for 1987-1993,
the chain Laspeyres indices give a much better approximation of the ideal
chain Fisher indices than the fixed weight Laspeyres indices. Besides it is
remarkable that for 1924, 1936 and 1937 the fixed weight Paasche index as well
as the fixed weight Laspeyres index is higher than all three chain indices.
The years 1929/34 (except 1933) are a period of economic decline. For those
years the difference between the fixed weight Laspeyres and Paasche indices is
not systematic. However, in most cases fixed weight Laspeyres indices show
sharper declines than the chain Laspeyres indices. Again a very important
conclusion is that the chain Laspeyres gives a much better approximation of the
chain Fisher than the fixed weight Laspeyres indices. Also for this period the
differences between the chain Laspeyres, Fisher and Paasche indices are relatively - small.
28
Laspeyres
(weights
1921)
Laspeyres
(chain)
Fisher
(chain)
Paasche
(chain)
Paasche
(weights
t,1921)
1922
1923
1924
1925
1926
106.5
109.4
118.3
122.9
132.1
106.5
108.7
116.5
120.7
128.9
106.1
108.2
116.2
120.0
128.1
105.6
107.7
115.8
119.3
127.3
105.6
107.4
115.8
119.4
126.6
1927
1928
1929
1930
1931
138.8
146.1
149.4
147.0
139.0
134.9
141.4
144.8
143.1
136.5
133.9
140.4
143.7
142.1
135.7
132.9
139.3
142.5
141.0
134.8
131.9
138.1
142.1
139.7
132.1
1932
1933
1934
1935
1936
137.3
137.0
136.7
142.7
151.9
135.3
135.9
134.3
138.0
145.6
134.2
134.6
133.0
136.9
144.2
133.1
133.5
131.6
135.9
142.9
129.1
129.9
128.0
131.7
139.1
1937
1938
1939
162.9
155.7
169.9
153.8
148.9
161.0
152.5
148.1
160.4
151.3
147.2
159.8
148.9
144.6
158.8
Table 4.b shows the time series of volume indices of GDP for the years 1921 to
1939. Despite of the strong price and volume fluctuations during that period the
chain Laspeyres and Paasche indices show much similarity with the ideal
chain Fisher indices. The highest deviation is 1.4 (see 1936). In contrast, the
series of fixed weight Laspeyres indices yields very high deviations from the
Fisher indices for many years. Highest deviations are for 1937 (9.1) and for
1939 (9.5). The series of fixed weight Paasche indices shows higher deviations
than the chain Paasche indices (highest for 1935: 5.2), but the deviations are
less than for the fixed weight Laspeyres.
29
7. Non-additivity
When using chain indices to construct time series of values in constant prices of
a certain base year, the problem arises that the deflated parts of an aggregate
do not add up to the deflated aggregate itself. The magnitude of this
mathematical discrepancy varies with the way the aggregate is split up in
different sub groups. In the Dutch National Accounts the mathematical
discrepancies in constant price tables are not eliminated, but these tables are
left non-additive. In footnotes the non-additivity is explained to the users.
In the publication National Accounts of the Netherlands (see: CBS, 1995)
constant price tables are not presented separately, but always in conjunction
with other tables and in a fixed sequence: current prices, volume changes, chain
volume indices and values at constant prices (by applying these volume
indices). Price changes and chain indices of deflators are given as well.
To give an impression of the order of magnitude, table 5 presents mathematical
discrepancies for the supply and disposition of goods and services of the years
1986 to 1993 in prices of 1986. It appears that, especially within final
expenditure, differences are quite substantial. Further, mathematical
discrepancies happen to be positive as well as negative.
1986
1987
1988
1989
1990
1991
1992
1993
437860
444051
455664
476987
496589
507866
518153
522104
2 Imports (cif)
205440
213989
230257
245783
256104
266656
272161
266465
-2
-169
-305
-313
-448
-460
38
643300
658038
685752
722466
752380
774074
789854
788607
67940
69729
70720
71809
72985
74103
75347
76475
260230
267349
269603
279005
290786
299795
307364
310453
Mathematical discrepancy
3 Disposable for final expenditure (gross)
4 Final consumption expenditure
general government
households
5 Fixed capital formation (gross)
enterprises
78160
78718
82202
86784
87573
87471
87109
84451
general government
11190
11435
12018
12075
12893
13204
14183
14023
enterprises
3830
-118
-57
-1019
-1252
-939
-577
767
8 Exports (fob)
221950
230924
251652
268387
282708
295997
304507
309155
-387
5425
6687
4443
1922
-6717
643300
658038
685752
722466
752380
774074
789854
788607
6 Changes in inventories
Mathematical discrepancy
9 Total final expenditure (gross)
31
Mathematical discrepancy
Mathematical discrepancy
not eliminated
eliminated
1992
1993
growth rate
1992
1993
growth rate
518.2
522.1
+0.8
517.9
522.1
0.8
2. Imports
272.2
266.5
-2.1
272.0
266.5
-2.0
-0.2
789.9
788.6
-0.2
Mathematical discrepancy
3. 1 + 2
-0.5
+0.0
789.9
788.6
75.3
76.5
+1.5
75.5
75.8
+0.4
307.4
310.5
+1.0
308.1
307.8
-0.1
87.1
84.5
-3.1
87.3
83.7
-4.1
14.2
14.0
-1.1
14.2
13.9
-2.2
6. Changes in stocks
-0.6
+0.8
-0.6
+0.8
304.5
309.2
+1.5
305.2
306.6
-0.2
789.9
788.6
7. Exports
Mathematical discrepancy
8. 4a. - 7
32
+1.9
-6.7
789.9
788.6
+0.4
-0.2
8. Concluding remarks
In this paper we discussed the use of chain indices in the Netherlands. In Dutch
practice chain indices are applied from 1980 onwards. In the discussions
preceding to the introduction of chain indices, CPB, the most important user of
national accounts data, plays an important role in making a plea for annually
changing weights and chain indices. CPB still has a strong preference for
chaining one-period growth rates based on annually changing weights. They call
their experience with the use of chain indices excellent.
Three other users of national accounts data have been asked about their opinion
on the choice between fixed weight indices and chain indices. They all agree that
chain indices are a good base for the construction of economic models.
Further:
it is necessary that model builders carefully tune the specification of their
model to the characteristics of the data;
non-additivity need not be a real problem, if researchers give special
attention to the presentation of the data to the public;
for policy makers it is important that chain indices have a strong
correspondence to actual developments.
We have carried out a sensitivity analysis of the choice of index number
formulae and weights in the deflation of national accounts data. For the period
1986-1993 alternative estimates are presented for the volume growth rates of
GDP, private consumption of households, gross fixed capital formation, imports
and exports, and for the output, intermediate consumption and value added of
three industries. For the period 1921-1939 alternative estimates are given for the
volume growth rates of GDP.
Considering the chain Fisher indices as a standard, most important conclusions
that can be drawn from those results are:
in nearly all cases both chain Laspeyres and chain Paasche volume growth
rates and indices prove to be good approximations of chain Fisher indices.
This affirms the statement in ESA-1995 (section 10.64) that chain Laspeyres
volume indices (in combination with chain Paasche price indices) provide
acceptable alternatives to chain Fisher indices;
the results show that most substantial deviations from the chain Fisher
indices are found with the fixed weight Laspeyres formula. This is very
important since fixed weight Laspeyres volume indices often are used in
constant price estimations of national accounts data;
fixed weight Paasche indices generally speaking behave better than fixed
weight Laspeyres indices, but in some cases they can also lead to severe
deviations from the chain indices.
33
References
Al, P.G., B.M. Balk, S. de Boer and G.P. den Bakker, The use of chain indices
for deflating the national accounts, Statistical Journal of the United Nations,
volume 4, number 4, July 1987 p. 347-368.
Allen, R.G.D., 1975, Index numbers in theory and practice, Aldine Publishing
Company, Chicago.
Bakker, G.P. den, 1991, The choice of index number formulae and weights in
the national accounts (a sensitivity analysis based on macro-economic data for
the interwar period), National Accounts Occasional Paper nr. NA-044 (Statistics
Netherlands, Voorburg).
Balk, B.M. and C.J.J. Eijgenraam, 1979, Aspecten van een statische en
dynamische definitie van de prijsindex, Advisory Committee for the Price
Statistics of Statistics Netherlands.
Boer, S. de and G.A.A.M. Broesterhuizen, 1986, The simultaneous compilation
of current price and deflated input-output tables, National Accounts Occasional
Paper nr. NA-013 (Statistics Netherlands, Voorburg).
CBS, National Accounts of the Netherlands, 1995, Statistics Netherlands,
Voorburg.
Diewert, W.E., 1978, Superlative Index
Aggregation; Econometrica, vol.46, no 4.
Numbers
and
consistency
in
35
Annex
1) Fixed weights
qt
pq q
=
pq
0
p0 q t
p0 q 0
pt q t
pt q 0
pq
q
pq q
t
Laspeyres * Paasche
p
p
t 1 t
t 1 t 1
p t 1q t 1
qt
qt 1
p t 1q t 1
pq
pq
t
t 1
pt q t
q
pt q t t 1
qt
Laspeyres * Paasche
37
38
P-2
75,00
P-19
32,50
P-6
15,00
P-14
49,50
150,00
A-1
125,00
P-11
67,50
10,00
A large number of the above mentioned publications can be consulted in the libraries of
Statistics Netherlands at Voorburg (direct line +31 70 337 5151) or Heerlen (direct line +31 45
570 71 87).
39
40
41
42
Statistics Netherlands