Representations of Poverty and Place PDF

Laura L Paterson
Ian N Gregory
REPRESENTATIONS
OF POVERT Y
AND PL ACE
Using
Geographical Text
Analysis to
Understand
Discourse
Representations of Poverty and Place
Laura L Paterson · Ian N Gregory
Representations
of Poverty and Place
Using Geographical Text Analysis
to Understand Discourse
Laura L Paterson Ian N Gregory
Languages and Applied Linguistics History
The Open University University of Lancaster
Milton Keynes, UK Lancaster, UK
ISBN 978-3-319-93502-7 ISBN 978-3-319-93503-4 (eBook)

https://doi.org/10.1007/978-3-319-93503-4
Library of Congress Control Number: 2018943637
© The Editor(s) (if applicable) and The Author(s) 2019

This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by
similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein
or for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Cover image: © the authors

Cover design by Fatima Jamadar
This Palgrave Macmillan imprint is published by the registered company Springer Nature
Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Acknowledgements
This work came out of a large number of collaborations and benefited

from input from many colleagues. A particular mention should go to
Dr. Andrew Hardie, from the Department of Linguistics and English
Language, Lancaster University, for his technical wizardry in helping us
manipulate such large amounts of material. Both authors would like to
acknowledge the Economic and Social Research Council (ESRC) for
supporting this work through the ESRC Centre for Corpus Approaches
to Social Science (grant number ES/K002155/1). Ian Gregory would
also like to acknowledge the contribution that the European Research
Council (ERC) made to the work described in this book through
funding from the European Union’s Seventh Framework Programme
(FP7/2007–2013)/ERC grant ‘Spatial Humanities: Texts, GIS, places’
(agreement number 283850).
v
Contents
Introduction xvii
1 Defining and Measuring Poverty 1
2 Corpus Linguistics, Critical Discourse Analysis,

and Poverty 19
3 Geographical Information Systems and Textual Sources 41
4 Conducting GTA Using Poverty as a Search Term 61
5 How to Use GTA in Discourse Analysis 95
6 Locating (Un)Employment in the National Press 123
7 Characterising Poverty in Place: Benefits Receipt

in Britain 159
8 Geography-Based Discourses of Poverty 193
vii
viii
Contents
9 Conclusions 227
245
References
Index 257
About the Authors
Laura L Paterson is a Lecturer in English Language and Applied

Linguistics at The Open University and editor of the Journal of
Language and Discrimination. She is a corpus-based sociolinguist whose
research concerns the discursive construction of UK poverty, audience
response to poverty porn, media depictions of protest, and discourses of
marriage.
Ian N Gregory is Professor of Digital Humanities at Lancaster

University. His main area of expertise is in applying Geographical
Information Science approaches in unconventional ways, particularly
through the use of textual sources. He has published six books and
nearly a hundred journal articles and book chapters.
ix
List of Figures
Fig. 2.1 Sample concordances for and poverty in the Guardian (2015) 25

Fig. 3.1 Using a combination of a spatial data and b attribute data
to represent some hypothetical census data 44
Fig. 3.2 Vector data representing a points, b lines and c polygons.
Each point, line segment, and polygon is linked
to its own attribute data, as shown in Fig. 3.1 45
Fig. 3.3 An example of raster data. Each cell is of known size
with the numeric value representing an attribute such
as height or density 46
Fig. 3.4 A fragment of geoparsed newspaper text taken from
the Daily Mail 53
Fig. 4.1 Density smoothed maps of <*poverty*> PNCs:
a Guardian b Daily Mail 69
Fig. 4.2 Statistical comparison of the PNCs from the Daily Mail
and the Guardian 70
Fig. 4.3 Carstairs scores for the UK from the 2011 census.
Polygons show local authority districts 80
Fig. 4.4 Density smoothed maps of the PNCs. Note that
for legibility individual instances have not been shown
as points: a Guardian b Daily Mail 87
xi
xii
List of Figures
Fig. 4.5 Kulldorf analysis of <*poverty*> using a 5% sample

of the as the background population. Points identified
are those from the background population that are at
risk of being in a hot or cold spot: a Guardian b Daily Mail 88
Fig. 6.1 Different definitions of unemployment using a the 2011
census definition and b workless households 2015.
Legends use quintiles 125
Fig. 6.2 Density smoothed maps of <*employ*> PNCs
in a the Daily Mail and b the Guardian 131
Fig. 6.3 Density smoothed maps of PNCs in the Employment
subset in a the Daily Mail and b the Guardian 133
Fig. 6.4 Kulldorf clusters of PNCs in the Employment sub-set
Fig. 6.5 Density smoothed maps of PNCs in the Money subset
Fig. 6.6 Kulldorf clusters of PNCs in the Money subset
Fig. 7.1 Density smoothed and Kulldorf maps of PNCs
in the Benefits subset in the Daily Mail 166
in the Benefits subset in the Guardian 175
in the Housing subset in the Daily Mail 181
in the Housing subset in the Guardian 185
Fig. 8.1 Density smoothed and Kulldorf maps for all poverty
PNCs in the Guardian 195
Fig. 8.2 Density smoothed and Kulldorf maps for all poverty
PNCs in the Daily Mail 197
Fig. 8.3 Spatial segregation analysis comparing all poverty PNCs
for the Guardian and the Daily Mail: a All PNCs b PNCs
outside London 199
List of Tables
Table 4.1 Breakdown of Daily Mail and Guardian corpora 65

Table 4.2 PNC keywords in the Guardian comparing the co-text
of <*poverty*> in London with the rest of the UK 74
Table 4.3 PNC keywords in the Daily Mail comparing the co-text
of <*poverty*> in London with the rest of the UK 75
Table 4.4 The occurrences of PNCs from the two newspapers
in local authority districts with differing levels
of deprivation 81
Table 4.5 Local authority districts with the most PNCs from
the Guardian and the Daily Mail 82
Table 4.6 Spearman’s rank correlation coefficients comparing
district-level Carstairs scores & PNCs 84
Table 5.1 Top 100 collocates for <*poverty*> 97
Table 5.2 Search terms in the Daily Mail and the Guardian 103
Table 5.3 Concordance lines containing PNCs for all search terms 107
Table 5.4 Queries with low/zero PNCs 114
Table 5.5 Queries with highest raw values of PNCs 117
Table 5.6 Subsets of search terms 121
Table 6.1 Correlation coefficients between measures of poverty,
unemployment, and worklessness at local authority
district level. Excludes Northern Ireland 126
xiii
xiv
List of Tables
Table 6.2 Top 100 lexical collocates of unemployment 128

Table 6.3 PNC keywords in the Employment subset
in the Daily Mail (n ≥ 5) 136
Table 6.4 PNC keywords in the Employment subset
in the Guardian (n ≥ 5) 143
Table 6.5 PNC keywords in the Money subset
Table 6.6 PNC keywords in the Money subset
Table 7.1 Top 25 collocates of <*benefit*> 163
Table 7.2 PNC Keywords in the Benefits subset
Table 7.3 PNC Keywords in the Benefits subset
Table 7.4 PNC Keywords in the Housing subset
Table 7.5 PNC Keywords in the Housing subset
numbers of Guardian PNCs with three measures
of poverty at local authority district level 189
numbers of Daily Mail PNCs with three measures
of poverty at local authority district level 190
Table 8.1 Frequency of poverty and place mentions
in the Daily Mail and the Guardian 194
Table 8.2 PNC Keywords for London when compared
to the rest of the UK for the Guardian
and the Daily Mail (n ≥ 5) 201
Table 8.3 PNC Keywords for the Manchester cluster. Italicised
keywords are found in both newspapers (n ≥ 5) 207
Table 8.4 PNC Keywords for the Merseyside cluster. Italicised
Table 8.5 PNC Keywords for the Birmingham cluster. Italicised
Table 8.6 PNC Keywords for the Glasgow cluster. Italicised
List of Tables
xv
Table 8.7 PNC Keywords for the Newcastle cluster. Italicised

Table 8.8 PNC Keywords for hotspots away from major urban
centres. Italicised keywords are found in both
newspapers (n ≥ 5) 217
Table 8.9 PNC Keywords for Guardian only clusters (n ≥ 5) 219
Table 8.10 PNC Keywords for Daily Mail only clusters (n ≥ 5) 220
Introduction
This book presents an entirely new, interdisciplinary method that

we use to study the geographies associated with poverty in the UK.
Bringing together corpus linguistics, critical discourse analysis (CDA),
and geographical information systems (GIS), we interrogate the dis-
courses surrounding poverty and place in two multimillion-word cor-
pora of contrasting UK national newspapers. We demonstrate that the
combination of these methods—henceforth geographical text anal-
ysis (GTA)—can be applied to large volumes of textual data, and use
our analysis of poverty and place to show how the apparent division
between quantitative and qualitative (i.e. textual) data is a false dichot-
omy. We also demonstrate the importance of considering geography—
in particular the use of toponyms—within critical analyses of language.
Researchers wishing to study poverty have the option of draw-
ing upon many different types of data. Economists, for example, are
likely to use quantitative data related to GDP, wages, inflation, and/
or wealth distribution to calculate who is (and who is not) in poverty.
Ethnographers, on the other hand, may wish to focus on the lived
experiences of those in poverty, concentrating their work on interac-
tions with people experiencing homelessness or hardship. Historians
xvii
xviii
Introduction
may work systematically through census data or workhouse records to

inform depictions of the development of poverty in a given place and/
or time, and philosophers may consider poverty in the abstract in order
to make conclusions about societal hierarchies. This range of data and
approaches indicates that poverty is a wide-ranging, multifaceted, social
issue, which can be realised in different forms and measured on differ-
ent scales. The research presented here takes language as its primary data
set and draws on techniques from linguistics in combination with tech-
niques from the spatial humanities to present an innovative method for
approaching the analysis of poverty.
Geographical approaches to poverty tend to be based on quanti-
tative data, such as statistical measures of deprivation like Carstairs
scores (Morris and Carstairs 1991), which have often been derived
from census data. Such statistical data are well-suited to use with GIS.
Quantitative analyses tend to be comprehensive, insofar as they can
cover entire countries and provide robust data on macro-level trends,
but they are less likely to show nuances within a data set. Smaller-scale
qualitative studies can also have a geographical angle, such as the Open
Society Foundations’ report on the lived experiences of six white work-
ing-class communities across Europe (OSF 2014).1 However, despite its
fairly wide geographical reach, the focus on six locations means that, as
with all small-scale studies, it is difficult to generalise from the OSF’s
data.
Significantly, linguists can work to a range of scales using a myriad of
textual sources. CDA, for example, tends to focus on the detailed anal-
ysis of a relatively small number of texts, with the aim of accounting for
the wider social context within which such texts are produced. At the
other end of the scale, corpus linguistics facilitates the analysis of large
bodies of texts spanning millions of words, which would be infeasible
to analyse using manual analytical approaches alone. To this end, there
is an established body of literature combining corpus linguistics with
(critical) discourse analysis (Conrad 2002; Orpin 2005; Baker 2006;
1One of the locations chosen by the OSF was Higher Blackley in Manchester. Based on their
findings, the authors of the report note the rise of ‘a particular negative image of white
working-class people’ in the mass media which is ‘most pronounced’ in the UK (2014: 62).
Introduction
xix
Baker et al. 2008; Mautner 2009; Baker and McEnery 2015a, etc.) indi-
cating that such methods are suitable and profitable for linguistic anal-
ysis. To date, however, existing work in corpus linguistics and CDA has
tended to largely ignore geography.
However, all texts have some (implicit) reference to place insofar
as they are constructed in a specific place and time for a given audi-
ence. Even large reference corpora, such as the British National Corpus
(BNC), which was established to provide a snapshot of British English
in the late twentieth century, are fixed in terms of the geographical loca-
tions of their texts’ authors, publishing houses, and primary audiences.
Yet, despite the implicit links between texts and their place of construc-
tion, publication, and consumption, corpus linguistics does not always
acknowledge the potential role of geography in the analysis of language.
The same criticism applies for analysis of explicit mentions of place,
such as the occurrence of place-names. Nevertheless, mentions of place
influence our understandings of the world: we may associate particu-
lar locations with being rich or poor, which may then translate to some
areas being understood to be more desirable than others. Locations
associated with poverty may also be associated with other factors,
including particular types of (low-paid) jobs, the ethnicity of the people
who live there, and/or lower life expectancies.
By combining corpus linguistics and CDA with GIS, we add a fur-
ther dimension to the academic analysis and discussion of place. Thus
we follow Cooper and Gregory’s (2011: 90) argument that ‘there is a
move towards using GIS technology to highlight the imbricated rela-
tionship between the locatedness of everyday life and the spatialities of
cultural practices’. Geographical text analysis has the capacity to aid the
critical discourse analyst in facilitating a visual representation of biases
evident in texts, as it allows the researcher to map and spatially analyse
the use of place in relation to particular themes, such as poverty. To
date, GTA has been used in humanities research, particularly in histori-
cal demography (Murrieta-Flores et al. 2015; Porter et al. 2015) and in
literary studies (Gregory and Donaldson 2016; Donaldson et al. 2017).
This book represents one of the first times that this approach has been
used to study a modern topic in the social sciences.
xx
Introduction
Our decision to draw on linguistic data to analyse poverty, as

opposed to the other data types noted above, is justified by the argu-
ment that language is a primary method of communication which
allows us to come to a shared understanding of how the world is struc-
tured. Whatever nuances of meaning each reader attaches to the word
poverty, there are some widespread interpretations that we obtain and
reproduce through society and societal norms. Uncovering these norms,
and interrogating how they came to represent common sense interpre-
tations of UK poverty, is a valuable empirical endeavour. Our choice
to focus on newspaper data stems from the accessibility and ubiquity
of such texts. Members of the public are in regular contact with the
mass media thus the messages carried have the capacity to reach and
influence the opinions of huge audiences.2 The fact that newspapers
report on poverty indicates its social significance as a topic of debate
and demonstrates its apparent newsworthiness (see Potts et al. 2015).
However, newspapers are not neutral sources of language data; in line
with their political leanings, they represent poverty in particular ways,
both reflecting and shaping how it is conceptualised by the wider pub-
lic. For example, if such outlets are shown to (re)produce negative dis-
courses about people in poverty, those consuming the media may be
misinformed and/or and may be less likely to support endeavours to
eliminate poverty proposed by institutional agents, such as the govern-
ment, charities, and food banks.
Bednarek (2006: 14) argues that readers are unlikely to assume that a
single journalist is responsible for a given article, but rather, they will see
the text as an example of an ‘institutional voice’, understanding the text
as part of a larger whole, and ‘identify the newspaper (as institution) as
the definite, or ultimate sources of what why are reading’. Such sources
‘are important objects of study because they further our understanding
of how the legitimacy of activities […] is formulated and contested’
(Breit 2010: 622). The social power of these institutional voices is
2Bednarek (2006: 12) notes that Britain has more national daily and Sunday newspapers than
any other country and that the British are the third biggest buyers of newspapers in the world.
She also notes that tabloid papers outsell broadsheets by about four to one and that over 50% of
the British population who read newspapers read a national tabloid in comparison with 13% for
broadsheets (2006: 13).
Introduction
xxi
measurable in wider society. To take one particular example, despite the

fact that UK benefit fraud (a topic closely related to poverty by the mass
media) accounted for only 0.8% of benefit expenditure in 2014/15
(DWP 2015), media depictions of benefit claimants paint welfare fraud
as a wide-ranging epidemic: in 2012, the Sun newspaper unveiled a
‘Beat the Cheat’ campaign, encouraging its readers to report suspected
benefit fraud. Furthermore, research by Ipsos MORI (2013) found that
the British public believed 24% of benefits were fraudulently claimed, a
figure thirty times higher than factual. Analysing the language used to
by the media to discuss UK poverty is thus a worthy research topic.
In the first instance, we set out to identify which place-names are
mentioned in relation to poverty in our corpora of two UK national
newspapers. Our corpora comprise news and comment articles from
the Daily Mail and the Guardian published between 2010 and 2015
(although the analysis presented herein focuses primarily on the news
sections, see Chapter 4 for more details). We establish which place-
names occur, and how frequently. We also note those locations which
are seldom, if ever mentioned and consider their relative size. We con-
sider whether place-names correlate with particular aspects of poverty
discourses, such as unemployment or benefits receipt, and we use GIS
software (ArcGIS) to investigate how media representations of place
compare to official statistics.
Our analysis of the geographies of poverty in the UK includes a wide
variety of place-names. To help guide the reader through these, Fig. 1
shows some of the most significant places referred to in subsequent
chapters. The map illustrates that the two newspapers associate many
different types of place with poverty. London and the large provincial
cities such as Birmingham, Liverpool, Manchester, Newcastle, Belfast
and Glasgow are, perhaps, predictable. There are also smaller cities
and towns in former industrial areas, such as Merthyr Tydfil in south
Wales, the Black Country (consisting of towns such as Wolverhampton
and Walsall), Nottingham and Leicester in the Midlands, and
Middlesbrough in the north-east of England. Seaside towns such as
Margate in Kent and Torquay in Devon also feature, as do port cities
such as Bristol, Hull, and Aberdeen. Finally, there are more affluent
xxii
Introduction
cities including Oxford, Cambridge, and York. In the course of the

book, we will establish how and why these places are associated with
poverty by one or both newspapers.
Fig. 1 Places that are frequently referred to in this book
To establish the terms of debate, Chapter 1 reviews how poverty is

defined and measured in different disciplines. We consider the proxy
measures that are used to estimate poverty rates, such as unemploy-
ment and overcrowding, and also review more qualitative approaches to
Introduction
xxiii
poverty that are based on the lived experiences and access to resources
of those living in poverty. The chapter also focuses on the difference
between absolute poverty and relative poverty, with UK poverty tending
to be conceptualised as the latter. As this book draws together meth-
ods and approaches from different disciplines, we have included two
chapters which explain the basics of corpus linguistics, CDA, and GIS,
in order to make GTA accessible to a range of readers from different
fields. Chapter 2 focuses on analytical approaches to large volumes of
text using corpus linguistics and CDA. Chapter 3 explores how geogra-
phy can be modelled computationally using a geographical information
system (GIS).
Chapter 4 brings these different analytical techniques together to pro-
vide a detailed description of GTA. To provide an exemplar of GTA, this
chapter is based around the use of a single search term, the word pov-
erty. Having established proof of concept for GTA at such a large scale,
we then present a much broader analysis of the geographies associated
with poverty in our chosen newspapers, which includes a comparison
between the textual data from our corpora and more-established meas-
ures of poverty, such as census and worklessness statistics. Chapter 5
describes how corpus linguistics techniques were used to identify a
much wider set of search terms that our newspapers associate with pov-
erty. These are grouped under four main themes: employment, money,
benefits, and housing. The following two chapters explore the geogra-
phies associated with these themes in more detail. Chapter 6 looks at
employment and money, while Chapter 7 moves on to benefits and
housing. It also briefly compares how the geographies of all four themes
in our corpora compare with geographies of poverty as defined using
quantitative sources.
Chapter 8 brings all of our poverty-related search terms together
to present an overall analysis of the geographies of poverty as repre-
sented in our corpora. It focuses on how different areas of the UK are
associated with poverty in different, nuanced ways by each newspa-
per. Taking a critical perspective we demonstrate that the newspapers’
choice of place, and what aspects of poverty to focus on, provides sup-
porting evidence for the overarching ideologies that they each endorse.
Our extended conclusion draws out the main findings of our analysis
xxiv
Introduction
and considers what GTA can tell us about representations of poverty.

It also addresses some of the broader implications of the methods used
and notes the limitations of this exemplar of GTA which should be
addressed in future research.
References
Baker, P., & T. McEnery (eds.). 2015. Corpora and Discourse Studies.
Basingstoke: Palgrave.
Baker, P. 2006. Using Corpora in Discourse Analysis. London: Continuum.
Baker, P., C. Gabrielatos, M. KhosraviNik, M. Kryzanowski, T. McEnery, &
R. Wodak. 2008. A Useful Methodological Synergy? Combining Critical
Discourse Analysis and Corpus Linguistics to Examine Discourses of
Refugees and Asylum Seekers in the UK Press. Discourse and Society 19 (3):
273–306.
Bednarek, M. 2006. Evaluation in Media Discourse: Analysis of a Newspaper
Corpus. New York: Continuum.
Breit, E. 2010. On the (Re)construction of Corruption in the Media: A
Critical Discursive Approach. Journal of Business Ethics 92 (4): 619–635.
Conrad, S. 2002. Corpus Linguistic Approaches for Discourse Analysis.
Annual Review of Applied Linguistics 22: 75–95.
Cooper, D., & I. Gregory. 2011. Mapping the English Lake District: A
Literary GIS. Transactions of the Institute of British Geographers 36 (1):
89–108.
Donaldson C., I. N. Gregory, & J. E. Taylor. 2017. Locating the Beautiful,
Picturesque, Sublime and Majestic: Spatially Analysing the Application of
Aesthetic Terminology in Descriptions of the English Lake District. Journal
of Historical Geography 56: 43–60.
Gregory, I. N., & C. Donaldson. 2016. Geographical Text Analysis: Digital
Cartographies of Lake District Literature. In D. Cooper D., C. Donaldson,
& P. Murrieta-Flores (eds.). Literary Mapping in the Digital Age. Abingdon:
Routledge, pp. 67–87.
Mautner, G. 2009. Checks and Balances: How Corpus Linguistics can
Contribute to CDA. In R. Wodak & M. Meyer (eds.). Methods of Critical
Discourse Analysis. London: Sage, pp 122–143.
Morris, R., & V. Carstairs. 1991. Which Deprivation? A Comparison of
Selected Deprivation Indexes. Journal of Public Health Medicine: 318–325.
Introduction
xxv
Murrieta-Flores, P., A. Baron A., I. N. Gregory, A. Hardie, & P. Rayson. 2015.

Automatically Analysing Large Texts in a GIS Environment: The Registrar
General’s Reports and Cholera in the Nineteenth Century. Transactions in
GIS 19: 296–320.
Orpin, D. 2005. Corpus Linguistics and Critical Discourse Analysis:
Examining the Ideology of Sleaze. International Journal of Corpus Linguistics
10 (1): 37–61.
Porter, C., P. Atkinson, & I. N. Gregory. 2015. Geographical Text Analysis: A
New Key to Nineteenth-Century Mortality. Health and Place 36: 25–34.
Potts, A., M. Bednarek, & H. Cape. 2015. How can Computer-Based Methods
Help Researchers to Investigate News Values in Large Datasets? A Corpus
Linguistic Study of the Construction of Newsworthiness in the Reporting
on Hurricane Katrina. Discourse & Communication 9 (2): 149–172.
1
Defining and Measuring Poverty
Any discussion of poverty and its related discourses must begin with a
definition. One positive of taking an interdisciplinary approach to the
analysis of poverty is the increase in the pool of potential resources
that we can draw upon in defining the term. We must define poverty
within the geographical boundaries of the UK and situate our definition
within the twenty-first century. Fundamentally, we must question how
to measure poverty. The very foundation of this book relies on compar-
ing measurements of poverty and/or deprivation using census data (and
its derivatives) with discursive depictions of poverty, using the tools of
Geographical Text Analysis (GTA).
Lansley and Mack (2015: 3) argue that ‘[d]efinitions of poverty
matter’ because they act as a determinant of ‘whether the incomes and
living conditions of the poorest in society are acceptable or not’. This
chapter discusses some of the many different ways of defining and
measuring poverty, both quantitative and qualitative. Section 1.1 con-
siders definitions of poverty and notes that, as there is no undisputed
way to measure poverty, any definition (used implicitly and explic-
itly) likely performs an ideological function. Section 1.2 focuses on
attempts to measure poverty both quantitatively and socioculturally,
© The Author(s) 2019 1

L. L Paterson and I. N Gregory, Representations of Poverty and Place,
https://doi.org/10.1007/978-3-319-93503-4_1
2
L. L Paterson and I. N Gregory
using measures such as census statistics and Carstairs scores. Section 1.3
considers the wider social context within which this research sits, and
summarises some of the major trends in discourses of poverty identified
in existing research.
1.1 Key Components of Poverty

The Oxford English Dictionary’s definition of poverty is ‘Destitution:
The condition of having little or no wealth or few material possessions’
(OED 2017). Far from being absolute, references to ‘little or no wealth’
and ‘few’ possessions leaves the understanding of poverty somewhat
open to interpretation. Following Chambers (2006: 3) it is important
to note that ‘What poverty is taken to mean depends on who asks the
question, how it is understood, and who responds’. Perhaps even more
importantly Chambers (2006: 3) notes that ‘Our common meanings
have all been constructed by us, non-poor people. They reflect our
power to make definitions according to our perceptions’. It is therefore
important to survey a range of definitions and be aware how the selec-
tion of definition(s) will act as a lens through which research is designed
and carried out. Furthermore, research must be informed by an aware-
ness of the social structure that facilitate non-poor people’s definitions
of poverty and which restrict the voices of those in poverty.
In their review of definitions of poverty in the late twentieth and
early twenty-first centuries, Misturelli and Heffernan (2008: 670) note
that there are seven key aspects of poverty: material, physical, economic,
political, social, institutional, and psychological. Whilst they present
an overview of developments in definitions of poverty from the 1970s
onwards, they note that geographical influences began to appear in the
1980s, with distinctions being made between ‘urban’ and ‘rural poverty’
(2008: 674). In the early 2000s, ‘concepts of poverty appeared to be
moving from defining poverty to defining the poor themselves […] the
focus was on grouping the poor into discrete categories often linked to
the poverty line’ (2008: 679). Yet, they argue that definitions of pov-
erty that employ an in/out dichotomy have been rejected as being too
simplistic; the life experiences of people close to either side of a ‘poverty
1 Defining and Measuring Poverty
3
line’ or threshold economic figure are not automatically dissimilar and,

relatedly, those on either side of the line cannot be seen as two homoge-
nous wholes. This is similar to Lee’s (1999: 174) argument that poverty
thresholds are somewhat arbitrary and are not absolute measures; there
will be some people below the poverty line not ‘experiencing poverty’
and presumably the opposite holds true for those above the line. Yet,
despite some inter-decade differences, including fluctuations in the rel-
ative weightings of the different elements of poverty, the fact that the
seven elements Misturelli and Heffernan identified repeat across their
dataset shows that the core of what constitutes poverty appears fairly
stable.
Despite this apparent stability, Gordon (2006: 29) argues that
‘there is still no official definition of poverty in the UK’. There is an
EU definition (established in 1984) which includes ‘persons, families
and groups of persons whose resources (material, cultural and social)
are so limited as to exclude them from the minimum acceptable way
of life in the Member State in which they live’ (2006: 30). By this defi-
nition, poverty is therefore conceptualised as a relative phenomenon,
not an absolute state. Gordon (2006: 32) makes a distinction between
the definition of poverty and the social realisation(s) of poverty when
he suggests that ‘Poverty is the lack of resources and deprivation is the
consequence of poverty’ and suggests that to be poor in Britain, in ‘sci-
entific terms’, people must have ‘both a low income and a low standard
of living’ (2006: 39). Lansley and Mack (2015: 3) note that the Child
Poverty Act (2010) uses 60% of median household income as a thresh-
old for measuring poverty, and thus it represents a ‘statutory recognition
that poverty is relative’.1
There is also a recognised definition of absolute poverty (see Gordon
2006: 31), endorsed by the United Nations, that references ‘a condition
characterised by severe deprivation of basic human needs, including
food, safe drinking water, sanitation facilities, health, shelter, education,
1Lansleyand Mack (2015) note this measure is sometimes reported as 60% of average household
income (taken to imply the mean), which would make abolishing (child) poverty impossible.
However, it is statistically possible for no one to live on a household income less than 60% of the
median of the UK’s household incomes.
4
and information’ (UN 1995: 57). However, one can be in poverty with-
out being in absolute poverty and, as such, some measures of poverty
noted by the UN will be largely irrelevant in countries with systematic
clean water supplies, public sanitation facilities, nationalised health pro-
vision, and free education. Lansley and Mack (2015: 3), however, argue
that foregrounding the extremes of poverty and emphasising ‘hun-
ger and homelessness’ can be used by powerful institutions ‘as a way
of underplaying the extent of poverty’. The fact that the UN’s defini-
tion of absolute poverty is not universally applicable emphasises that the
geography of poverty is significant; what is considered poverty in one
socially-defined location (such as within a country’s borders) may not be
considered poverty in other locations.
To this end, Gordon (2006: 39) argues that ‘low income and low
standard of living can only be accurately measured relative to the norms’
of individual societies, what he terms objective poverty. He proposes that
deprivation measures relating to ‘personal, physical and mental condi-
tions, local and environmental facilities, social activities and customs,
are more suitable for measuring poverty and deprivation than economic
measures of consumption expenditure (Gordon 2006: 39). In meas-
uring objective poverty, Gordon (2006: 40) notes that the variables of
income and ‘standard of living are correlated’ and acknowledges that
there ‘will always be some ambiguities near the margins about whether a
person should be defined as “poor” or not’. He suggests, therefore, that
‘it is better to conceive the poverty threshold as a band of low income
and standard of living rather than as a hard fixed line’ (2006: 40).
Despite different interpretations of poverty, its manifestations as rel-
ative or absolute, the acceptance/rejection of an in/out dichotomy, and
its geographical location, what all the definitions above have in common
is an underlying sense that poverty is something which can be observed
and scientifically measured. Language about poverty lines, margins,
thresholds, standards, minimums, income levels, etc. and yes/no
measures such as access to clean water and sanitation, treat poverty as
something quantifiable and concrete. The possibility that poverty is
unmeasurable is not considered; yet we cannot intrinsically measure
poverty as it is an abstract, socially-determined concept. In order to
address this, the following section discusses what is actually measured
when determining the boundaries of poverty.
5
1.2 Poverty by Proxy

Measurements of poverty and deprivation in human geography,
and specifically in existing research employing GIS (see Chapter 3),
tend to be based on quantitative data,2 including Carstairs scores
(Morris and Carstairs 1991), the Index of Local Conditions (Lee
1999),3 the Townsend Index (Townsend et al. 1988), and Indices of
Multiple Deprivation (Noble et al. 2006). All of these indices are cal-
culated using combinations of variables from census data or other sta-
tistical indicators, such as unemployment, overcrowded housing, or lack
of amenities (see Morgan and Baker 2006; Morris and Carstairs 1991).
However, none of these indices actually measure poverty, rather poverty
is calculated using quantifiable proxies that are presumed to be sympto-
matic of poverty. Morgan and Baker (2006: 28) note that deprivation
measures, such as Carstairs scores have been ‘constructed to act as a
proxy for data on personal/household income or wealth which have not
routinely been collected in the UK census’.
The data used to calculate Carstairs scores—unemployment, over-
crowded housing, lack of car ownership, and low social class—is
problematic. In early calculations of Carstairs scores (at least) the unem-
ployment statistics were based on male employment figures only. Wider
social factors were also ignored. For example, the ownership of a car
does not take into account the fact that cars vary greatly in both their
monetary value and level of ownership (people may have company cars,
lease their cars, purchase them with finance, own their vehicles outright,
etc.) and so the assumption that two households who each have cars
will be in some way statistically similar is questionable. Additionally,
whilst overcrowding may appear more clear cut, as it measures the ratio
2See Lee (1999) and Morgan and Baker (2006) for a discussion of statistical measures of depriva-
tion in the UK.
3The Index of Local Conditions (see Lee 1999) is based primarily on census data and includes
unemployment, poor children, overcrowding, lack of amenities, no car, children in flats, educa-
tion at 17, income support, low educational qualifications, standardised mortality rates, derelict
land, and crime.
6
of people to rooms in a house, we have to consider political factors

too, such as the Removal of the Spare Room Subsidy (Bedroom Tax)
implemented by the UK Coalition government in 2013. Government
guidelines on Housing Benefit determine that a two-bedroom home
with a combined kitchen/living space occupied by an adult couple and
two children of the same sex under 16 is appropriate housing. Thus, a
definition of overcrowding that is ‘more than one person per room’ may
need to be revisited, as the example given here leads to four people liv-
ing in effectively three rooms.
Lee (1999: 172) argues that whilst the UK census ‘is still the best
source for spatial analysis, it is limited for the analysis of poverty or dep-
rivation’ because it does not sufficiently cover variables relating to pov-
erty. He suggests that unemployment is used as a de facto measure of
poverty (1999: 174) and posits two main problems with using unem-
ployment rates as a single measure of poverty (or indeed any measures
‘which use single proxy indicators’). He states that ‘it is not possible to
read off the whole range of possible circumstances’ for individuals or
households when using a single measure and he notes that although the
relationship between a single measure and other variables may corre-
late well, nevertheless ‘the relationship between any two indicators is far
from dichotomous’ (1999: 175).
Another proxy measure for poverty that has been used in the social
sciences is social class. For example, Dorling et al. (2000) used measures
of social class, based on occupation, within local government wards
to indicate relative poverty. Here we have a three-part assumption:
occupation = social class/economic status = poverty.4 This is particularly
problematic given that definitions of social class are as fluid as
definitions of poverty (see Block 2013). Morgan and Baker (2006: 29)
4Dorling et al. (2000) estimate levels of poverty by calculating proportions of social class groups
within a ward using the midpoint of each social class—assuming a normal distribution—to cal-
culate a poverty index. Using data from Charles Booth’s study of London poverty 1889–1903
and the 1991 UK census, their results show that historically, there has not been much change in
poverty distribution in London boroughs between the nineteenth and twentieth centuries. They
note some areas that have seen lots of immigrant settlement have ‘moved down the social scale
slightly’ whilst others have been through the process of gentrification, but overall ‘affluent places
have remained affluent and poor places have remained relatively poor’ (2000: 1549).
7
note that the 2001 UK Census included the now widely-used NS-SEC
(National Statistics Socio-economic Classification) instead of the previ-
ously used ‘social class’. However, the NS-SEC still relies predominantly
on employment as a measure of socioeconomic status/groupings, and so
Lee’s (1999) criticisms (noted above) still stand.
Alternative measurements of poverty/deprivation are based on liv-
ing standards. As such, they ‘conceptualise poverty as a combination
or series of deprivations, both material and social, such that resources
are so seriously below those commandeered by the average individual
or household that they are, in effect, excluded from ordinary living pat-
terns, customs and activities’ (Lee 1999: 174). Pantazis et al. (2006)
discuss the construction and administration of the Poverty and Social
Exclusion Survey—a UK-based initiative that was founded on earlier
work by Mack and Lansley (1983/85) and the Breadline Britain Survey
(see Meinhof and Richardson 1994). By asking members of the British
public which possessions and activities they see as necessary to main-
tain a minimum acceptable standard of living, the aim of these surveys
have been to ‘try to discover whether there is a public consensus on
what is an unacceptable standard of living for Britain’ and to find out
‘who, if anyone, falls below that standard’ (Mack and Lansley 1985: 50
in Pantazis et al. 2006: 89). The results of these surveys have shown that
there is a high level of consensus (50% of participants agreeing) across
different social demographics (age, class, gender) on what items and
social activities are necessary for an acceptable living standard. UK resi-
dents without access to three or more of these essentials—which include
beds and bedding, a refrigerator, visits to friends/family, contents insur-
ance, carpets in living rooms and bedrooms, television, and a holiday
away from home—are deemed to be in poverty.
Updating the Breadline Britain surveys, Lansley and Mack (2015)
asked respondents from over two thousand households whether seventy
six different items/activities (including 30 measures relating to children)
were a necessity in twenty-first century daily life: items tested include
living in a damp-free home, possessing items such as curtains and a
dining table, eating two meals a day, and attending weddings/funerals.
Twenty-six adult items and twenty-four children’s items met the 50%
threshold for being considered necessities. Whilst Lansley and Mack
8
acknowledge that responses to the tested items were not uniform, they
point out that there was a relatively high level of consensus amongst
their participants (2015: 25ff.). They note that, in comparison to pre-
vious studies with similar methodologies, the composition of this list
of necessities has changed over time—a weekly roast dinner, for exam-
ple, is less likely to be seen as a necessity now due to changes in eating
habits, and carpets are less likely to be seen as a necessity due to the
popularity of laminate/wooden floors—and they suggest that ‘the public
accept that minimum living standards need to reflect contemporary and
not past styles of living. They believe that needs do not stand outside
society as some kind of timeless given’ (2015: 18). Thus there is con-
sensus that poverty is not fixed and absolute, but relative (at least in a
temporal sense). The items deemed necessities went beyond the basics
of food and shelter and included telephones, a leisure activity, and visit-
ing friends.
Lansley and Mack (2015: 26) argue that one of the most controver-
sial items they tested was possession of a television. They suggest that
a television, which ‘has been a near-universal possession for decades’
divided opinion amongst respondents in terms of educational back-
ground (71% of people without qualifications deemed it a necessity
compared with 43% of those with a degree) and type of occupation
(65% of manual workers to 48% of non-manual workers). Lansley and
Mack also note a correlation between television ownership being seen as
a necessity and restrictions affecting respondents’ ability to go out (such
as ill health, old age, and lack of income). However, what is telling is
that they draw on a single quote from the 1983 survey:
I watch TV from first thing in the morning till last thing at night, till the
television goes off. That’s all I’ve got: to watch television. I can’t afford to
do other things at all. (Lansley and Mack 2015: 27)
They generalise from this one participant, whom they describe as ‘a

young lone parent with a nine-month-old baby living in a damp and
decaying attic in London’ (2015: 27). Here the authors rely heavily
on social stereotypes and do not provide sufficient evidence for their
9
statements. Such generalisations run the risk of using individual expe-

riences of poverty to attribute attitudes to a generic, homogenised class
or social group. By drawing upon stereotypes Lansley and Mack (2015)
use a thirty-year-old quote to characterise the pastimes of a homoge-
nised poor.
Lansley and Mack (2015: 30) argue that their method of defining
poverty through public consensus leads to ‘a democratically defined
standard, free of value judgements by experts, officials or the govern-
ment’ which represents the views of ‘all groups in society, young and
old, rich and poor, in and out of work – and, perhaps most signifi-
cantly, across the political spectrum’. At least some of the responses
in Lansley and Mack’s study will have come from those in poverty, as
the survey covered a range of social demographics. Thus, their meas-
ure addresses the issue of poverty not being defined by the poor (cf.
Chambers 2006). Using their list of necessities, Lansley and Mack
tested over five thousand households to estimate levels of poverty across
the UK and Northern Ireland. They conclude that whilst there are high
levels of deprivation across the UK, rates are higher in ‘London, the
North-East and North-West of England, the West Midlands, Wales
and Northern Ireland, and slightly lower in Scotland and southern
and eastern England (excluding London)’ (Lansley and Mack 2015:
51). However, despite this regional variation, they argue that people
who do not possess three of the household essentials ‘form a distinct
group’ (even though they might have different combinations of missing
essentials) and, furthermore, that the deprivation experienced by this
group leads to ‘severely restricted lifestyles and opportunities’ (2015:
50). However, to homogenise those without three necessities is to
ignore potential regional and intersectional differences, such as house
prices, the variable cost of living, employment rates and the number
of available jobs available, and the health and educational profiles of
different areas. As the following analysis demonstrates, a consideration
of place in conjunction with poverty can highlight different lived expe-
riences of poverty and different economic stressors that vary according
to location.
10
1.3 Conceptualisations of Poverty

and the Poor
To conclude this chapter we focus on the language surrounding poverty,
beginning with Kress’ (1994: 29) observation that, despite its abstract
nature, poverty can be a grammatical agent and act upon its subject:
‘poverty can drive us into despair, poverty causes the breakup of fami-
lies, and so on, as the media tell us every day’. Kress (1994: 28) argues
that poverty ‘is something that you are in ’ rather than something that
you possess and, furthermore, the ‘place that you are in when you are
in poverty is an abstract place, like despair […] a kind of mental place,
an emotional state of affairs’. The use of an abstract concept as an agent
acts to obscure any human input which could be (in)directly respon-
sible for individuals experiencing poverty and performs an ideological
function insofar as the structural inequality which could be to blame
for (mass) poverty is not explicitly expressed. Further, due to its abstract
nature, poverty is difficult to oppose.
Kress also argues that the term poverty ‘is something with which you
are afflicted’ and is ‘a characteristic which acts as a description of a per-
son, a classification’ (1994: 29). By conceptualising poverty as a label for
a particular group of people we can begin to see how its use could be
socially powerful; to label someone as ‘poor’ or ‘in poverty’ is to imbue
them with a set of (negatively-loaded) characteristics that, presumably,
they cannot escape. Indeed, the small number of reports of people escap-
ing poverty that we find in our corpora (see Chapter 4) is evidence of the
perceived rarity of moving out of poverty, and goes someway to explain-
ing the apparent newsworthiness of such stories. Although it is possible
to deny that one is in poverty, any denials are determined in the first
instance by whether the person who has been labelled has a voice to pro-
test. In media discourse those who are labelled as ‘poor’ tend not to have
the right of reply.5
5Jo (2013: 516) notes how Lister’s (2004: 7) definition of poverty includes ‘shameful and cor-
rosive social relation[s] …characterized by a lack of voice, disrespect, humiliation and reduced
dignity and self-esteem’.
11
In a US context, Lorenzo-Dus and Marsh (2012: 289) analysed two

US National Security Strategy reports and found that poverty was also
referred to through metaphor, particularly metaphors relating to disease.
This is a linguistic choice where depicting ‘poverty as an illness means
classifying it as a biological entity—as a phenomenon from the natural,
rather than the social, world’. Furthermore, they note that this charac-
terisation of poverty fits with Kress’ (1994) discussion of the passivity
of poverty as a noun, as ‘natural phenomena obey natural forces, over
which societies ultimately have little control’ (2012: 289). However,
Gordon (2011, cited in Lansley and Mack 2015: 73) argues that pov-
erty ‘is not like syphilis or a biblical curse across a generation—pov-
erty is not a disease and it cannot be caught and all credible evidence
shows that it is not “transmitted” to children by their parents’ genes or
culture’.
The ideology of poverty as generational was expressed in the
‘Troubled Families’ strategy of the UK coalition government, which was
launched in 2011 and is set to run to 2020. Lansley and Mack (2015:
67) argue that families in disadvantaged households were ‘redefined by
ministers first into families that are “troubled”, and then into families
that are or cause trouble’. Unlike Kress’ analysis of poverty, such con-
ceptualisations presuppose that the poor are somehow responsible
for and/or complicit in their own poverty. Similar arguments can be
found (implied) in media reports about benefit fraud or non-working
parents with large numbers of children.6 Relatedly, Lansley and Mack
(2015: 147) argue that particular mass media outlets report that ‘there
are deprived areas of the UK with a “Benefits Street” culture based on
generations of families choosing a life on benefits’ but they note that
such claims do ‘not match the evidence’, citing research by the Joseph
Rowntree Foundation which found no families in high unemployment
6Lansley and Mack (2015: 147) claim the ‘frequency with which the tabloid press and some tel-
evision programmes feature very large families living on benefits’ is disproportionate to reality, as
reports from the DWP state that there were ‘only 180 claiming households with ten or more chil-
dren in 2010’ and 91% of households claiming benefits ‘have three or fewer children, and only
one percent have six or more’.
12
areas in Middlesbrough and Glasgow where multiple generations had

never worked. Nevertheless, negative evaluations of those in poverty,
and the related assumptions about the roles people have played in their
own poverty, carry social weight.
Pantazis et al. (2006), for example, show awareness of wider social
characterisations of the poor. They note McKay’s (2004) argument
that ‘families who cannot afford necessities often possess a number
of non-necessities’ and as a result ‘they have atypical preferences and
choose to spend their money on “luxuries” not “necessities”’ (Pantazis
et al. 2006: 114). They argue that McKay does not acknowledge that
some ‘luxuries’ may have been purchased before the onset of pov-
erty, but in any case, ‘as deprivation increases, the average number of
non-necessities a respondent possesses decreases’ (2006: 114), showing
that McKay’s criticism is not realised in actuality. However, the authors
do not consider in detail McKay’s additional claim that ‘such families
may be classified as poor using deprivation indicators, when it might
be more accurate to say that their consumption preferences deviate
from the average’ (McKay 2004: 220 cited in Pantazis et al. 2006: 114).
What McKay is tapping into here is the notion of a ‘flawed consumer’:
the ideology that the poor are responsible for their own fate because
they are unable to manage money and are ‘inadequate […] consumers’
(Bauman 2004: 38). McKay’s use of the term ‘atypical preferences’ pre-
sents those who are poor as somehow anomalous to the wider society
in which they live. Although McKay does not use the term flawed con-
sumer, this ideological assumption can be linked to blaming the poor
for their own poverty—they are just spending their money wrongly—
and does not seek to address wider social concerns, such as why people
may spend money on taboo or negatively-stereotyped items such as tele-
visions (see above), cheap frozen food, and cigarettes.7
Acknowledging that poverty is more than just economic circum-
stance, Jo (2013: 519) focuses on how shame in poverty is both exter-
nally influenced and internally experienced and sheds light on the
7See Paterson et al. (2017) for a discussion of how such items are used as indexical markers
deployed to negatively evaluate members of the working classes.

13
explicit role of the wider society’. She argues that the relationship
between poverty and shame is constructed ‘from the dominant dis-
course’—i.e. through language and cultural norms which are ‘collec-
tively assembled by multiple institutions which are governed by those
with power and influence’ (Jo 2013: 522). Such an interpretation of
how discourses of poverty are utilised supports her claim that poverty
‘cannot be fully interpreted outside of the specific social context in
which it occurs’ (Jo 2013: 519). It is clear then, that poverty is more
than a statistical measure of income or consumer expenditure. Fukuda-
Parr (2006: 7) explicitly notes that poverty relates to ‘public policy’ and
‘is now widely considered to be a multidimensional problem’ which is
manifest in ‘a complex set of deprivations’. She argues that twenty-first
century definitions of poverty ‘have refocused the concept of poverty as
a human condition that reflects failures in many dimensions of human
life’ which include not only long-held indexes of poverty such as home-
lessness, unemployment, and poor health, but also include ‘powerless-
ness and victimisation, and social injustice’ (Fukuda-Parr 2006: 7).
However, whilst including such elements within a definition of poverty
serve to illustrate its multidimensionality, Fukuda-Parr does not give
any examples of how, or indeed if (relative) poverty could actually be
measured.
1.4 Summary
One of the aims of this book is to compare the more traditional (quanti-
tative) calculations of poverty with the discursive construction of poverty
in the UK press. To this end, we compare government-endorsed statis-
tics, such as census data and unemployment figures, to the location and
representations of poverty present in our two corpora (see Chapter 4).
Each type of data presents a form of reality—it expresses a discourse of
poverty in geographical space—but neither the statistics nor the dis-
cursive representation capture the full complexity of poverty. Lorenzo-
Dus and Marsh argue that poverty ‘defies easy definition’, includes a
multitude of factors and is a ‘social, political and/or security issue’ with
links to ‘social exclusion and discrimination, economic migration, and
14
terrorism and (inter)national insecurity’ (2012: 275). They note that

poverty ‘is (re)produced through discourse practices, most of which
are enacted by groups with direct access to social, cultural and political
power’ (Lorenzo-Dus and Marsh 2012: 275).8 We interrogate just how
these different statistical and discursive depictions of poverty correlate
(or diverge) with an aim of uncovering the ideologies expressed in dis-
courses of poverty.
As poverty is multifaceted, we do not wish to provide a formal defi-
nition. Rather, we take a data-led approach to our corpora and analysis,
with the aim of investigating how the two newspapers conceptual-
ise poverty and how they locate it in space, in order to compare these
media representations of poverty with statistical datasets. In accept-
ing that poverty can be realised in different ways, we take a holistic
approach to understanding it and conceptualise it as a multifaceted
social phenomenon. For example, we note Galbraith’s (1958) claim that
poverty occurs when people’s ‘income, even if adequate for survival, falls
markedly behind their community’ (cited in Lansley and Mack 2015:
14), but we emphasise that an absence of adequate finances is not whole
story; ideologies of poverty stretch beyond mere economics. Indeed,
Galbraith notes that it is community norms which determine ‘the nec-
essary minimum for decency’ (cited in Lansley and Mack 2015: 14),
and this leads to the conclusion that to conceptualise someone as being
in poverty is, at least in part, to make a social judgement. Galbraith goes
even further by suggesting that, because of such social judgements, peo-
ple in poverty are ‘degraded’ insofar as they are conceptualised as liv-
ing beyond ‘the grades or categories which the community regards as
acceptable’ (cited in Lansley and Mack 2015: 14). These social judge-
ments are expressed in public policy, in media depictions of the poor,
and, most importantly for the present work, through language.
8Furthermore, Lorenzo-Dus and Marsh (2012: 277) note that discourse analysis undertaken in
South America has shown that there is ‘a tendency in political and media elite discourses to quan-
tify poverty, typically through statistics that deprive those being thus defied of their individual
identity’ alongside the ‘construction of “the poor” as passive, indolent and immoral’.
15
The prevalent themes discussed in Sect. 1.3 demonstrate some of the

most prominent attitudes to those experiencing poverty. There is also a
growing body of research which considers how the mass media (and the
public who engage with it) evaluate the poor and the working classes
in twenty-first century Britain. For example, in their analysis of pub-
lic responses to the television programme Benefits Street, Paterson et al.
(2017) found that members of the public drew on stereotypes of work-
ing class identities and accepted the media’s (implicit) negative evalua-
tion of particular activities and cultural norms, such as the consumption
of alcohol/tobacco. The use of similar stereotypes was also found in
van der Bom et al.’s (2018) analysis of audience response to the second
series of Benefits Street. They found that the poor were homogenised,
insofar as they were understood to look the same, talk the same, and
value the same practices (some of which were deviant behaviours asso-
ciated with crime and low morals). These findings correlate well with
Biressi’s (2011) consideration of what we would term poverty porn pro-
gramming, and Skeggs and Wood’s (2011) collection focusing on how
social class is depicted on reality television.
Similarly, Bennett (2012) analysed chavspeak—the language associ-
ated with working class identities in texts such as ‘chav dictionaries’, and
guides on how to speak chav. He concludes that by characterising chavs
(and arguably all working-class people) linguistically ‘language ideolo-
gies concerning the incorrectness of the non-standard and the inartic-
ulacy of the poor are implicated in broader social ideological trends,
discourses about not only the kinds of language that people use, but the
kinds of people that exist in British society’ (2012: 21). This homoge-
nising notion that the poor constitute a particular kind of person taps
into neoliberalism—insofar as people are considered to be responsible
for their own fate—and related characterisations of the deserving and
undeserving poor (Katz 2013). Although the deserving/undeserving
poor distinction is considered by some to be somewhat old fashioned, it
does seem able to account for certain media representations of poverty,
especially when benefits receipt is the primary focus of media texts.
Combined, this body of research points towards an understanding
of poverty that draws upon Bourdieu’s (1986, reprinted 2010) notions
of social, economic, and cultural capital, insofar as poverty relates to
16
people’s economic resources (how much money they have, where that
money comes from, and how they spend that money), their social
resources (whether they are employed, if they have access to educa-
tion, adequate health care, etc.), and cultural resources (their hobbies,
consumer habits, and local network norms). Whilst there is evaluation
of the poor in the mass media along these lines—van der Bom et al.
(2018) found that the clothing people wore on the second series of
Benefits Street was a primary site for critical evaluation, and there were
related links to flawed consumerism—there appears to be less accept-
ance of the macro-structural constraints on these forms of capital, such
as low wages, changes to government benefits (such as the Welfare
Reform Bill), the geographical location of available jobs, access to ade-
quate transport links, etc.
Our investigation of media discourses on a large scale, using our two
multi-million word corpora, focuses on the nuances of how different
aspects of poverty are deployed in relation to geography. In the two chap-
ters which follow, we introduce readers unfamiliar to corpus linguistics,
critical discourse analysis, and GIS to the fundamentals of our chosen
methods. For those who are already familiar with one or more of these
methods and their related techniques, Chapters 2 and 3 may not be
necessary. We pick up our analysis of discourses of poverty and place in
Chapter 4, which demonstrates how the combination of linguistics and
geography—in the form of GTA—can illuminate an as-yet-untapped
aspect of poverty discourses by comparing the locations of poverty
expressed by statistical measures with those referenced in media texts.
References
Bauman, Z. 2004. Work, Consumerism and the New Poor. Buckingham: Open
University Press.
Bennett, J. 2012. ‘And What Comes Out May Be a Kind of Screeching’: The
Stylisation of Chavspeak in Contemporary Britain. Journal of Sociolinguistics
16 (1): 5–27.
Biressi, A. 2011. ‘The Virtuous Circle’: Social Entrepreneurship and Welfare
Programming in the UK. In H. Wood & B. Skeggs (eds.). Reality Television
and Class. London: British Film Institute/Palgrave Macmillan, pp. 144–155.
17
Block, D. 2013. Social Class in Applied Linguistics. London: Routledge.

Bourdieu, P. 1986/2010. The Forms of Capital. In I. Szeman & T. Kaposy
(eds.). Cultural Theory: An Anthology. Oxford: Wiley Blackwell, pp. 81–93.
Chambers, R. 2006. What is Poverty? Who Asks? Who Answers? Poverty
in Focus. International Poverty Centre: United Nations Development
Programme. http://www.ipc-undp.org/pub/IPCPovertyInFocus9.pdf. Accessed
23/9/2015.
Dorling, D., R. Mitchell, M. Shaw, S. Orford, & G. Davey Smith. 2000. The
Ghost of Christmas Past: Health Effects of Poverty in London in 1896 and
1991. British Medical Journal 321: 1547–1551.
Fukuda-Parr, S. 2006. The Human Poverty Index: A Multidimensional Measure.
Poverty in Focus. International Poverty Centre: United Nations Development
Programme. Online. http://www.ipc-undp.org/pub/IPCPovertyInFocus9.pdf.
Accessed 11/11/2015.
Galbraith, J. K. 1958. The Affluent Society. London: Penguin.
Gordon, D. 2006. The Concept and Measurement of Poverty. In C. Pantazis,
D. Gordon, & R. Levitas (eds.). Poverty and Social Exclusion in Britain.
Bristol: Policy Press, pp. 29–70.
Gordon, D. 2011. Consultation Response: Social Mobility and the Child
Poverty Review. Policy Response Series 2, PSE 2011, pp. 4–5. http://www.
poverty.ac.uk/pse-research/pse-uk/policy-response.
Jo, Y. N. 2013. Psycho-social Dimensions of Poverty: When Poverty Becomes
Shameful. Critical Social Policy 33 (3): 514–531.
Katz, M. B. 2013. The Undeserving Poor (second edition). Oxford: Oxford
University Press.
Kress, G. 1994. Text and Grammar as Interpretation. In U. Meinhof &
K. Richardson (eds). Text, Discourse and Context: Representations of Poverty
in Britain. London: Longman, pp. 24–47.
Lansley, S., & J. Mack. 2015. Breadline Britain: The Rise of Mass Poverty.
London: Oneworld.
Lee, P. 1999. Where are the Deprived? Measuring Deprivation in Cities and
Regions. In D. Dorling & S. Simpson (eds.). Statistics in Society. London:
Arnold, pp. 172–180.
Lister, R. 2004. Poverty. Cambridge: Polity Press.
Lorenzo-Dus, N., & S. Marsh. 2012. Bridging the Gap: Interdisciplinary
Insights into the Securitization of Poverty. Discourse and Society 23 (3):
274–296.
Mack, J., & S. Lansley. 1985. How Poor is too Poor? Defining Poverty. In
J. Mack & S. Lansley (eds.). Poor Britain. London: George Allen & Unwin.
18
McKay, S. 2004. Poverty or Preference: What Do ‘Consensual Deprivation

Indicators’ Really Measure? Fiscal Studies 25 (2): 201–223.
Meinhof, U. H., & K. Richardson (eds). 1994. Text, Discourse and Context:
Representations of Poverty in Britain. London: Longman.
Misturelli, F., & C. Heffernan. 2008. What is Poverty? A Diachronic
Exploration of the Discourse on Poverty from the 1970s to the 2000s. The
European Journal of Development Research 20 (4): 666–684.
Morgan, O., & A. Baker. 2006. Measuring Deprivation in England and Wales
Using 2001 Carstairs Scores. Health Statistics Quarterly 31: 28–33.
Selected Deprivation Indexes. Journal of Public Health Medicine 13(4):
318–325.
Noble, M., G. Wright, G. Smith, & C. Dibben. 2006. Measuring Multiple
Deprivation at the Small-Area Level. Environment and Planning A 38:
169–185.
OED. 2017. poverty, n. OED Online. www.oed.com/view/Entry/149126.
Pantazis, C., D. Gordon, & P. Townsend. 2006. The Necessities of Life. In
C. Pantazis, D. Gordon, & R. Levitas (eds.). Poverty and Social Exclusion in
Britain. Bristol: Policy Press, pp. 89–122.
Paterson, L. L., L. Coffey-Glover, & D. Peplow. 2016. Negotiating Stance
Within Discourses of Class: Reactions to Benefits Street. Discourse & Society
27 (2): 195–214.
Paterson, L. L., D. Peplow, & K. Grainger. 2017. Does Money Talk Equal
Class Talk? Audience Responses to Poverty Porn in Relation to Money and
Debt. In A. Mooney & E. Sifaki (eds.). The Language of Money and Debt:
An Interdisciplinary Approach. Basingstoke: Palgrave, pp. 205–232.
Skeggs, B., & H. Wood (eds.). 2011. Reality Television and Class. London:
British Film Institute/Palgrave Macmillan.
Townsend, P., P. Phillimore, & A. Beattie. 1988. Health and Deprivation:
Inequality and the North. London: Routledge.
UN. 1995. The Copenhagen Declaration and Programme of Action. World
Summit for Social Development, 6–12 March 1995. New York: United
Nations.
van der Bom, I., L. L. Paterson, K. Grainger, & D. Peplow. 2018. ‘It’s Not the
Fact They Claim Benefits but Their Useless, Lazy, Drug Taking Lifestyles
We Despise’: Analysing Audience Responses to Benefits Street Using Live
Tweets. Discourse, Context & Media 21: 36–45.
2
Corpus Linguistics, Critical Discourse
Analysis, and Poverty
The purpose of this chapter and the one that follows is to introduce
readers unfamiliar with linguistic and/or geography-based approaches to
poverty to the tools that we have chosen to use. The chapter begins with
an overview of corpus linguistics, what it is (the systematic, computa-
tionally-aided analysis of large bodies of texts using specialist software)
and what assumptions corpus analysis makes about language. Section 2.1
also describes some of the tools associated with corpus analysis, with
illustrative examples. Section 2.2 introduces and defines what we mean
when we use the term discourse and, leading on from this, we introduce
the core principles of critical discourse analysis (CDA) in Sect 2.3. The
final section of this chapter considers the benefits of combining corpus
linguistics and CDA for the analysis of social phenomena, before leading
into a consideration of the geographical aspects of our analysis of poverty
in Chapter 3.

https://doi.org/10.1007/978-3-319-93503-4_2
20
2.1 Introducing Corpus Linguistics

Corpus linguistics centres on the systematic computational analysis of
large bodies of electronically held texts—known as corpora—which are
taken to be representative of a particular text type, genre, or language
variety. The term corpus linguistics can be used to refer to a method of
analysis, but also represents a methodology which assumes particular
characteristics about language. The most basic definition of a corpus
is that it is ‘simply a collection of texts’ (McCarthy and Carter 2004:
147), which have particular characteristics in common. For example,
texts could be all of the same type, publication date, mode (spoken/
written/signed), topic, or language variety. Tognini-Bonelli (2001: 2)
stipulates that to define a dataset as a corpus it must have been created
with the express purpose of being ‘used for linguistic analysis’. Thus,
a random selection of texts, collected somewhat haphazardly, cannot
constitute a corpus. This book utilises two corpora of UK newspaper
texts, each of which comprises written British English in the form of
news articles from national sources. The two sources we have chosen to
analyse are the Guardian and the Daily Mail and our two corpora total
almost four hundred million words. For details of corpus construction,
see Chapter 4.
The method of corpus linguistics relates to the systematic application
of particular corpus tools, such as frequency counts, collocation, and
concordance analysis, many of which are explained below. In a meth-
odological sense, corpus linguistics works with some basic assumptions
about language. The ethos behind corpus linguistics is that, by utilis-
ing the processing power of computers and specialist corpus software,
the researcher can work with many more texts than would be feasible
if undertaking a manual analysis. Providing that a corpus can be said
to be representative of its designated text type or language variety, ana-
lysing many texts simultaneously allows the researcher to spot patterns
that occur beyond the scope of manual analysis. The empirical basis of
corpus linguistics is founded upon the idea that a corpus will be able
to indicate linguistic norms for a more generalisable whole. Biber et al.
(1994: 171) argue that this empirical foundation is what ‘allows us to
2 Corpus Linguistics, Critical Discourse Analysis, and Poverty
21
test assumptions [or hypotheses] about language use against patterns

found in naturally occurring discourse’. Especially where assumptions
about language and corpus data are found to be in conflict, it is imper-
ative that any claims made about the results of corpus analyses be appli-
cable to more than just the corpus itself. The analysis of large bodies of
texts can provide robust indications that linguistic patterns—in this case
relating to the discussion of UK poverty—repeat across a large dataset
and represent the systematic characterisation and evaluation of a par-
ticular social issue across wider society.
Showing that a linguistic feature or specific use of language is gener-
alisable is a key goal of using corpora for language analysis, and Stubbs
(2001: 168) notes that multiple occurrences of a phenomenon in a cor-
pus show ‘that meanings [and/or usage] are not personal and idiosyn-
cratic’ but rather, representative of wider language use. This supports
Kennedy’s (1998: 4) stance that not only do corpora provide means for
‘identifying the elements and structural patterns which make up the sys-
tems we use in a language’, they also aid in ‘mapping out our use of
these systems’. Thus, a corpus analysis of the language surrounding the
term poverty may indicate trends in how this phenomenon is conceptu-
alised in UK media discourse. For example, the terms child and fuel are
the top two L1 collocates—words occurring immediately to the left of
poverty—in our Daily Mail corpus. This finding suggests that these two
particular aspects of poverty—child poverty and fuel poverty—are con-
sidered to be more important than international poverty (which occurs
166th in the L1 collocate list). Similarly, in the Guardian corpus, the
top L1 collocates of poverty are also child and fuel (with international
occurring in 151st place), suggesting that the key elements of how pov-
erty is represented in the mass media may be similar across our chosen
sources.
This simple example begins to illustrate that corpora can be used at
all levels of linguistic analysis, from the word (or even morpheme—the
smallest contrastive unit of meaning) upwards. Indeed, some of
the simplest calculations that corpus software can perform is to count
the frequencies of each word that occurs in a given corpus. Such counts
can be used, along with other corpus tools (explained below), to com-
pile dictionaries of the language under investigation (Longman and
22
Macmillan produce dictionaries this way). Corpus linguistics can also

be used interdisciplinarily to address social issues; the ESRC Centre for
Corpus Approaches to Social Science (CASS) specialises in stretching
the boundaries of corpus linguistics (http://cass.lancs.ac.uk/).
However, corpus linguistics (as a method and methodology) is
not without criticism. One popular opposition is the claim that cor-
pus analysis provides (only) quantitative data. Such criticism is under-
pinned by the notion that counting linguistic features is no match for
their close analysis: showing that a phrase occurs twenty times cannot
tell you how that phrase is used. Relatedly, those opposing corpus data
may argue that examples taken from a corpus are decontextualised and
thus the implications of the use of particular linguistic features go unan-
alysed. However, there is simply no reason why corpus analysis cannot
also provide qualitative data. Research combining corpus linguistics
with discourse analysis—of which this book is an example—represents a
method of analytical triangulation which is continuing to grow in popu-
larity (Sect 2.4). McCarthy and Carter (2004: 148) claim that ‘linguists
have begun to see the value’ of using corpora to see if ‘local insights’
based on individual texts can be generalised across texts ‘from similar
contexts’. Additionally, Mautner (2007: 54) notes that the ‘awareness’ of
using corpora to generate qualitative data ‘has been increasing, over the
past ten years or so’ and now corpora may be used ‘for uncovering rela-
tionships between language and the social’.
Having established the core concepts of corpus linguistics, under-
standing the ethos underpinning such analyses is only the first step.
Corpus analysis is performed with specialist software and a suite of tools
designed to aid the researcher in searching large bodies of texts. We
have used corpus-software CQPweb (Hardie 2012), but there are other
pieces of software available (both commercial and free-to-use).
One of the most basic calculations corpus software can perform is
the generation of frequency lists. In its essence a frequency list is an
alphabetical list of each word in the corpus with information about how
many times that word occurs. (Note the term ‘word’ is used here for
simplicity, a frequency list can also give information about punctua-
tion, numbers, and other characters.) They can also contain informa-
tion about how many texts in a corpus the word appears in (the word’s
23
distribution across the corpus). However, whilst calculating frequency

may appear simple, consideration must be given to complex issues such
as what constitutes a word. Decisions have to be made about whether to
treat contractions, such as don’t and haven’t, as one word or two. Similar
questions arise in relation to compound nouns, such as hair brush, hair-
brush, and hairbrush, which may occur in similar, but different forms
across a corpus, and differences in spelling also need to be considered,
especially if a corpus contains different language varieties. As frequency
lists are used as the basis for other corpus calculations (keywords, dis-
persion plots, etc.) the choices made when generating them have the
capacity to influence results further on in the analysis.1
Using frequency lists as a basis, keywords can be generated. A pos-
itive keyword (usually known just as a ‘keyword’) is a word that is
statistically more likely to occur in the corpus under analysis than in
comparable general language. A negative keyword is a word which
occurs statistically less frequency in the corpus than would be expected
based on its occurrence in general language. To generate a list of key-
words, the frequency list of a corpus is compared to the frequency list
of a reference corpus. A reference corpus is designed to be representative
of a particular language variety. Keyword lists are useful for determining
which topics are dominant in your corpus. Here we modify the concept
of keywords to facilitate co-text analysis as explained in Chapter 4.
To measure the relationship between words, corpus linguists can used
collocation. Collocations occur ‘when a word regularly appears near
another word, and the relationship is statistically significant in some
way’ (Baker 2006: 95–96). They are important because the meaning
of a word is influenced by the other words around it and any repeated
and/or consistent collocations may be significant. For example, in the
Guardian corpus, the top twenty collocates of poverty include items such
as relative and absolute, which indicate that poverty is a measurable con-
cept, alongside child and fuel (as discussed above), and unemployment
1Furthermore, frequency lists can also be lemmatised, which means that all forms of a word
(e.g. swim, swam, swum, swimming ) are treated as part of one whole, rather than as separate items
in the corpus.
24
and in-work, which suggest that employment and poverty are linked
concepts in UK media texts. Similarly, for the Daily Mail, collocates
such as abject, relative, rate, and grinding also suggest that poverty can
be measured, but there is less indication that the Daily Mail presents
poverty as directly linked to working people.
In order to calculate the statistical value of a collocation Baker et al.
(2008: 279) note that three pieces of information are needed: ‘the fre-
quency of the node [the target word that you’re investigating], the
frequency of the collocates, and the frequency of the collocation’.2
Decisions must be made about how close two words need to be to
each other in discourse to constitute a pattern. There is a stronger link
between words that are immediate collocates—words occurring side-by-
side—than if the two words are only loosely collocated, such as they
occur in the same paragraph. Collocates tend to be grouped seman-
tically or grammatically to facilitate further analysis. However, it is
important not only to spot the patterns in collocates, but also to note
collocates which do not appear to fit with the others.
Moving towards more qualitative data, concordance lines are alpha-
betically ordered lists of all occurrences of a given search term in a cor-
pus within their immediate co-text (i.e. circa ten words either side).
Concordance lists are extremely useful in corpus analysis as all instances
of a search term can be sorted in terms of their surrounding text, which
makes it easier to see repeated patterns. Figure 2.1 shows a sample of
concordance lines for the search term and poverty in the 2015 section of
the Guardian corpus.
Analysing what repeated concepts are linked to poverty through the
use of the coordinating conjunction and gives an indication as to how
poverty is presented in a wider context. For example, in the ten concord-
ance lines given here, there are five references to food (food banks, food
blogger, food, dietary, and hunger ), and five references to money (bene-
fits, income, tax, minimum wage), but the two concepts do not seem to
2Collocations can be measured using different statistical calculations, such as log-likelihood and
mutual information (a calculation based on how often a word occurs and how often that word
occurs with the target node). For more information on corpus statistics, see Brezina (2018).
25
Fig. 2.1 Sample concordances for and poverty in the Guardian (2015)

26
overlap (notwithstanding the term food banks ) as they do not occur in

the same concordance lines. The separation of poverty and food from pov-
erty and money could be an element of media representations of poverty
worthy of analysis. Indeed, to support or refute such findings, more con-
cordance lines would need to be analysed. To perform discourse anal-
ysis on concordance lines, one must look for overarching patterns, but
complement this with a focus on individual examples to see how these
patters come about. Once all patterns in the data have been established,
analysis of those concordance lines that do not fit a pattern should take
place, with an aim of investigating why they might not include particu-
lar keywords or support particular discourses (see Baker 2006: 92).
The final corpus tool to be discussed here is a suite of resources
known as taggers. The purpose of a tagger is to add an additional level
of information to texts in a corpus. For example, part-of-speech (POS)
taggers assign each word in a corpus a grammatical tag (such as singu-
lar noun, personal pronoun, verb, etc.). Taggers can also work above
the level of the word, marking boundaries between sentences or other
grammatical units. Furthermore, semantic taggers can assign elements
of a text into pre-defined categories relating to semantic fields, grouping
together like terms. Such taggers are particularly useful when analysing
discourse and/or when focusing on key themes. For example, Paterson
et al. (2017) used the USAS semantic tagger provided by Wmatrix
(Rayson 2008) to analyse the semantic field of money and commerce
in focus group transcripts. They used the corpus software to group
together references to money that basic corpus query searches would
have missed (for example, their results included ‘down and outs’, and
‘well to do’ as well as the expected ‘poor’ and ‘banker’).
Whilst the selection of corpus tools will depend somewhat on
the research at hand, this overview acts to initiate those who have lit-
tle-to-no experience of corpus linguistics with some of the tools that
researchers performing corpus analyses have at their disposal. However,
corpus methods of analysis are only one part of our approach to dis-
courses of poverty. The following section includes a definition of how
we use the term discourse and provides a brief overview of CDA—an
interdisciplinary field of study which interrogates the relationship
between language and societal power.
27
2.2 Defining Discourse

The term discourse is used to mean many things both in everyday life
and within linguistics. Baker and McEnery (2015: 4) go so far as to
suggest that the term ‘can be an excluding shibboleth which does little
to make academic research accessible or relevant to people who do not
work or study in the social sciences’. Discourse can refer simply to a
stretch of text that is either spoken or written; in this sense it is being
used as a more formal term for text but the two may be used inter-
changeably. In this book we make the distinction between discourse and
text; the term text is used to denote examples of written British English
that take the form of newspaper articles from the Daily Mail and the
Guardian. Text is thus used here as a neutral label for stretches of lan-
guage, words, sentences, paragraphs, headlines, etc. which constitute a
definable whole. In this instance, a text from the Guardian constitutes a
piece of written language with a headline, sometimes a byline, a lead-in
paragraph, and a body text. In contrast, we use discourse, as discussed
towards the end of this section, to describe how language encodes,
reflects, (re)produces, and challenges social norms.
The term discourse can also be used to refer to linked ideas or over-
arching concepts, sometimes represented by labels such as educational
discourse, political discourse, religious discourse, and academic dis-
course. Each of these terms refers to a fairly stable (but somewhat malle-
able) structuring of ideas and topics around a central theme. Used in this
sense, a discourse of religion, for example, is likely to include terms such
as holy, priest, church, synagogue, Muslim, Sabbath, god, etc. Educational
discourses are likely to express idea that all children should attend school,
study core subjects, such as maths, English, and science, and are also
likely to promote the positive aspects of education. The dominance of
such educational discourses are likely to lead to mass social agreement and
the acceptance of the UK (or other) school system as the common sense
status quo. The establishment and deployment of such discourses can act
as a window on the social context surrounding a text and can allow text
authors to invoke particular (political) evaluations of a given topic.
However, such discourses are not fixed. For example, one realisa-
tion of poverty discourse could relate to starvation, homelessness, poor
28
sanitation, disease, and death, whist other poverty discourses could

include associations with government benefits, unemployment, laziness,
flawed consumerism, etc. The type of discourse evoked will depend on
the context, and, as such, discourses are impossible to document in their
entirety. This interpretation of discourse is important for the present
study as it begins to show the importance of how people use language to
represent particular events, conditions, people, etc. If a particular pov-
erty discourse can become entrenched within a given society then the
unequal distribution of resources that can lead to poverty can come to
be seen as common sense, and those who the discourse benefits/impedes
may become invisible. Thus, the use of particular discourses, and their
endorsement at the institutional level (government, schools, the mass
media) can become powerful in terms of Gramsci’s hegemony (Wodak
and Meyer 2009: 8)—or oppression by consent. From a CDA perspec-
tive (see Sect 2.3), if the status quo is not systematically challenged, and
is accepted as common sense, the social systems implicated in the (re)
production of poverty shall remain undisturbed and be replicated by
members of society, who are complicit in the reproduction of such pov-
erty discourses.
Altheide (2000: 287) argues that ‘when symbols are pervasive they
both reflect and contribute to frames and discourse for subsequent
meaning configuration’. Thus the representation of particular narra-
tives of poverty and their location in particular places by the mass media
may impact on future depictions of the poor and have wider social and
political ramifications, such as being used to influence/justify changes
in government policy. For example, there is evidence that once a place
is poor, it will remain so; Gregory (2009) showed that places in the UK
that were poor 100 years ago still tend to be poor today, and Dorling
et al. (2000) showed that between 1889 and 1991 there was very little
change to the relative poverty rates of different London boroughs. As a
result of their association with poverty, poor places may gain bad repu-
tations as areas where opportunity is limited, and this may impact upon
the economic profile of an area or the social evaluations of people who
live there (see McKenzie 2015 for an ethnographic approach to life in
poor communities).
29
Altheide takes the position that ‘[t]heoretical boundaries are breached

when mass media content and forms are part of our every-day lives and
contribute to social definitions of self, others, and social issues’ (2000:
289). This observation sits well with Breit’s (2010: 621) claims that dis-
courses ‘mediate the relationship between concrete linguistic aspects of
texts (e.g. use of certain vocabularies or formulation) and the implica-
tion of texts in wider societal or political contexts’. Of course, language
itself is not intrinsically powerful. It only becomes powerful when people
start to use it.
Language can be used to reinforce existing power in relationships,
but can also be used to alter and challenge those relationships. It can
also tell us something about the relationships between people—who is
attributed with linguistic power based on their social status, for exam-
ple. Power is a social phenomenon, not a linguistic one. Certain linguis-
tic structures can gain social significance, but nevertheless, they are only
powerful under the right circumstances: the words ‘I sentence you to
four years in prison’ only carry weight if spoken by a recognised judge;
a toddler repeating the same words does not have the same effect (this is
known as illocutionary force). Relatedly, discourse, in the sense that we
are using it, and in the sense that it is generally (although not always)
accepted to mean within CDA (see below), refers to the way that lan-
guage is used to reflect, shape, and encode societal norms and values.
Particularly, one aspect of CDA is a focus on how the norms and val-
ues of the powerful are presented through the use of particular linguistic
structures as common sense (cf. Fairclough 2001). To illustrate how lin-
guistic structures can influence the representation or characterisation of
a particular event, we can draw on our corpora.
1. The rise in poverty rates is largely due to stagnating wages and benefit
cuts (G news: society, Nov 2012).
Example (1) includes three occurrences of nominalisation—a linguis-

tic device where a verb is turned into a noun—which serve to disguise
the processes that are expressed in the sentence. For example, ‘the rise
in poverty rates’ includes the notion that ‘poverty rates are rising’, a
30
proposition which is accepted without question. Similarly, ‘stagnat-

ing wages’ and ‘benefit cuts’ presuppose that ‘wages are stagnating’ and
‘benefits are being cut (by the government)’. None of these three pro-
cesses—rising, stagnating and cutting—are actions in the sentence; they
are backgrounded and placed in noun phrases. Such nominalisations
make it difficult to challenge the claims embedded within a sentence;
it is hard to argue that ‘poverty rates are not rising’ because the notion
that ‘poverty rates are rising’ is taken to be common sense, expressed
as ‘the rise in poverty rates’. The sentence is not making the claim that
‘poverty rates are rising’ but rather it is assuming agreement that poverty
rates are rising and making a claim about the causes of such a phenom-
enon. Similarly, (1) is not arguing that ‘wages are stagnating’ or ‘ben-
efits are being cut’, but it takes these processes for granted and reports
that their end state ‘stagnating wages’ and ‘benefit cuts’ are the reason
for ‘the rise in poverty rates’. It is difficult (although not impossible)
given the use of nominalisations in this sentence to argue that ‘poverty
is not rising’, ‘wages are not stagnating’ and ‘benefits are not being cut’.
Thus, the author of this text is establishing the status quo of their argu-
ment by using a particular linguistic feature. Nominalisations are just
one of a huge number of linguistic devices that can be used to encode
one’s interpretation of a given topic within language, and their analysis
is just one tool at the disposal of the discourse analyst. By being more
aware of the power of language we can begin to recognise and question
the societal and political ramifications of language usage.3 The aim of
the next section is to introduce readers to an interdisciplinary approach
to language analysis known as CDA which aims to foreground the
social values that are encoded linguistically within texts.
3On a macro scale it might be that particular language varieties are deemed as socially more
appropriate, or acceptable, or beautiful, and so on. For example, double negatives such as ‘I didn’t
hear nothing’ have traditionally been rejected as ‘bad English’. These judgements do not make the
language variety under scrutiny intrinsically good or bad, but the social values attached to certain
language varieties are powerful.
31
2.3 Critical Discourse Analysis

The analysis of discourse, in its most basic form, refers to the close lin-
guistic analysis of structures above sentence level. Discourse analysis
focuses on how elements of a text work together to construct a (coher-
ent) whole. Whilst discourse analysis includes analysis of grammatical
forms, such as adjectives, noun phrases, clauses, etc., it recognises that
the meanings held within a text are socially constituted and cumulative,
insofar as meaning is created across (and beyond) a whole text. Taking
the word poverty as an example, written on a page, the use of a combi-
nation of seven letters to represent three syllables, this linguistic signifier
means nothing without readers drawing upon their own existing knowl-
edge of society. Poverty can conjure up images of dying children, famine,
war-torn countries, homelessness, council estates, begging, destitution,
etc., but the exact interpretation of this term each reader has will influ-
ence their response to the topic. However, human beings do not arrive
at interpretations of words in isolation.
Critical discourse analysis acknowledges that texts are read within
specific (temporal, geographical, political) contexts, and external influ-
ences can affect how a text is produced and distributed. Just as the close
linguistic analysis of one feature or one occurrence of a particular word
is incomplete, discourses also are not restricted to a particular (single)
text. They are cumulative, built up over time and across modes and out-
puts. For example, the association between benefit recipients and wel-
fare fraud is overrepresented in the mass media with extreme cases being
foregrounded (see Lundström 2013). An example from the Daily Mail
corpus, (2) demonstrates how benefit fraud is sensationalised.
2. Beggars belief: Scrounger, 34, who lay in doorways PRETENDING

to be homeless caught living at home with his MOTHER in her
230,000 village house (DM news, May 2014).
In this case the benefits claimant is labelled a ‘scrounger’ and is not

referred to by name. The newspaper’s negative evaluation of this person
is evidenced through the use of capitalisation and the listing of the value
32
of the mother’s house: not only is the homeless person living in a house,
but a relatively expensive one in a presumably-enviable village location.
Whilst this is just a single example, Lansley and Mack (2015: 142)
note that the use of scrounger in UK newspapers was four times higher
in 2010 than it had been between 1993 and 2003 (although the gap
between these dates could hide fluctuations in usage). The repetition of
this term, and others like it, are evidence of scrounger discourse, where
benefits claimants are portrayed as (unjustly) receiving high welfare pay-
ments and spending them frivolously (see van der Bom et al. 2018).
What makes discourse analysis critical is its explicit focus on language
and power; it is underpinned by the perspective that language is a social
phenomenon which reflects, (re)constructs, and rejects social (power)
norms. Wodak (1999: 186) argues that language ‘is not an isolated phe-
nomenon’ but rather it is ‘deeply social, intertwined with social pro-
cesses and interaction’. As Fairclough (2001: 21) explains:
in seeing language as discourse and as social practice, one is committing

oneself not just to analysing texts, nor just to analysing processes of pro-
duction and interpretation, but to analysing the relationship between
texts, processes, and their social conditions, both the immediate condi-
tions of the situational context and the more remote conditions of insti-
tutional and social structures.
The term critical discourse analysis does not refer to a single, formulaic
method of research or to a unified theory of society and social construc-
tion: ‘CDA has never been and has never attempted to be or to pro-
vide one single or specific theory. Neither is one specific methodology
characteristic of research in CDA’ (Wodak and Meyer 2009: 5). What
unites practitioners of CDA is their focus on the relationship between
language (as evidenced through close textual analysis) and wider social
issues. According to Scheuer (2003: 143) analyses will discuss texts ‘on
a macrosocial level, often with reference to global political movements’
and will generally discuss ideology. Ideology is another term with mul-
tiple overlapping definitions. For Fairclough this term relates to the rel-
ative acceptance of ideas presented as common sense. He argues that
ideologies (note the plural) ‘are closely linked to language, because using
33
language is the commonest form of social behaviour, and the form of

social behaviour where we rely most on “common-sense” assumptions’
(Fairclough 2001: 2). Others may look to alternative definitions of
ideology, such as the notion of symbolic violence (cf. Bourdieu and
Wacquant 2004). The common core underpinning such interpretations
of ideology (or indeed hegemony) is the notion that the oppressed are
unconsciously complicit in their oppression. As Fairclough (2001: 3)
argues, ideology ‘is the primary means of manufacturing consent’.
Breit (2010: 621) notes that ‘discourse is intrinsically linked with its
social and material context; it is socially constituted as well as socially
constituting, thus contributing to (re)constructing specific identities,
social relations, and systems of knowledge’. Focusing on discourses of
corruption, he argues that how journalists (and the social structures
surrounding them, editors, newspaper owners, etc.) choose to frame a
particular issue allows them to ‘to reduce the complexities, ambiguities,
and contradictions that often characterize activities related to corrup-
tion’ (2010: 621). Clearly these choices stretch far beyond discourses of
corruption, and we can see their effect on discourses of poverty, with
the use of scrounger discourses and the choice of the mass media to
focus on extreme cases of poverty (or indeed benefit fraud) in order to
maintain a particular caricature of the poor who can be homogenised
and negatively evaluated (see Skeggs and Wood 2011). For example,
Lansley and Mack (2015: 141) report on the Turn2Us charity’s finding
that, over 50% of tabloid articles about benefits printed between 1995
and 2011 included ‘negative vocabulary’. The systematic repetition of
particular characterisations of social groups can influence wider public
opinion and become part of the dominant discourses of poverty, class,
the deserving and undeserving poor, etc., discourses that can then be
utilised in texts as if they are common sense.
In his conceptualisation of discourse Fairclough (2001: 18–19)
argues that language is an intrinsic part of society (not something
which is external) which constitutes a socially-conditioned social pro-
cess which is influenced ‘by other non-linguistic parts of society’. The
representations of any social phenomena put forward by powerful insti-
tutions, such as the mainstream mass media, have the potential to influ-
ence public (and governmental) opinion. Fairclough (2001: 38–39)
34
suggests that power can be realised in discourse when ‘powerful partic-

ipants’ are ‘controlling and constraining the contributions/of non-powerful
participants ’ (original italics). In face-to-face interaction this can take
the form of interruptions or topic shifts, but on an institutional level
within media texts, we can ask questions about who gets to talk, how
their arguments are reported, and how people and actions are evaluated.
As Fairclough (2001: 43) notes, ‘[i]n the British media, the balance of
sources and perspectives and ideology is overwhelmingly in favour of
existing power-holders’. Indeed, in our corpora, it is rare to find direct
quotations from those experiencing poverty, and even when such quota-
tions are included in newspaper texts, they are framed by the rest of the
articles in which they appear. Thus, the newspapers hold the power over
how poverty is represented and evaluated.
2.4 Corpus-Based CDA and the Interrogation

of Poverty
The combination of corpus linguistics with CDA is not a new method
of analysis, but one which has been evolving since the end of the twen-
tieth century, as corpus tools developed and became more widely
available due to advances in computing. As discourses are cumulative,
it follows that analysing large bodies of texts would be beneficial for
the critical discourse analyst. However, corpus linguistics is not usu-
ally focused on the kind of close reading typical of CDA, nor does it
automatically address the social and political context surrounding the
texts held within a corpus. Despite these apparent differences in focus
and scale, the two methodologies can be successfully combined to pro-
duce robust triangulated analyses of the linguistic representation, rejec-
tion, and (re)construction of social phenomena. Whilst the methods of
analysis associated with corpora and CDA tend to be conceptualised as
quantitative and qualitative respectively, this is erroneous. As discussed
above, corpora can be used to inform qualitative analysis (specifically
through close reading of concordances and an in the analysis of repeated
syntactic patterns) and CDA benefits from quantification, insofar as
35
demonstrating the existence of a linguistic pattern, as opposed to an iso-

lated construction in a single text, contributes to arguments about how
ideologies are indexed and presented.4
Discussing the benefit of using corpus linguistic methods to inform
CDA Hardt-Mautner (1995: 8) notes that the analysis of concordance
lines ‘provide[s] new ways of kick-starting the analysis because it ena-
bles researchers to pursue even the most tentative leads’. She argues that
using concordance programmes within CDA means that a researcher can
‘describe syntactic and semantic properties of key lexical items exhaus-
tively rather than selectively’ and as a result they can ‘look at a large
number of occurrences rather than generalise in an undisciplined fash-
ion on the basis of a few purposely selected examples’ (1995: 23–24).
This arguably makes any claims made about the data under analysis more
robust. As Kennedy (1998: 9) notes, corpus linguistics is not just about
how many times certain constructions occur, nor is it concerned ‘only
with what is said or written, where, when and by whom’, but rather it
is also concerned with ‘how particular forms are used’. As such, corpus
linguistics can be informed by a CDA approach which focuses on the
social function(s) of the use of particular linguistic constructions, thus
answering the question ‘why’ particular forms are used within a particu-
lar context. The basic premise that analyses using corpus linguistics tools
tend not to focus on wider social factors involved in text construction is
directly addressed and critiqued when combined with the ethos of CDA.
Richardson (2006: 29) argues that CDA involves analysis and criticism
of just how social power and social norms are represented ‘both explicitly
and implicitly’ within texts. Linking to corpus analysis, Baker (2006: 19) is
careful to note that just because a particular pattern can be found in a cor-
pus does not mean that the pattern is representative of ‘mainstream ways of
thinking’. Taking a CDA-influenced perspective he notes that ‘a sign of true
power is in not having to refer to something, because everybody is aware of
it’ and argues that sometimes what is not written or spoken is actually more
important than what is (Baker 2006: 19). Indeed, Baker et al. (2008: 282)
4Baker and McEnery (2015) provide a clear introduction to corpus-based approaches to discourse
analysis. Their collection of edited papers is of interest to readers who wish to know more about
corpus-based (critical) discourse analysis, including its many associated methods, and the topics
that scholars have analysed with this mixed method.
36
also note Kress and van Leeuwen’s (2001) claim that critical analyses must
take ‘into account absences as well as presences in the data’, arguing that
a major concern with corpus analysis is that it ‘tends to focus on what
has been explicitly written’ (2008: 297). This criticism is a sound one, as
corpus software is not designed to look for the absence of data, and this
must be taken into consideration when using corpus data to inform CDA.
Nevertheless, combinatory analyses are possible. In Paterson’s (forthcoming)
analysis of discussions of UK poverty in Below the Line newspaper com-
ments, members of the public debated child poverty using references to
the responsibilities and failings of (single) mothers. In contrast, references
to (single) fathers were rare, and when (single) fathers were mentioned,
their mentions were either rejected or presented as isolated cases. Thus the
absence of (single) fatherhood in debates about UK child poverty serve
to gender the debate and frame child poverty as a problem directly asso-
ciated with poor motherhood. The quantification of references to moth-
ers and fathers—produced using corpus analysis—demonstrates a pattern
across texts, but this pattern is only understood through close analysis of
the occurrences of such terms and their subsequent relation to wider social
norms which situate women as primarily responsible for child rearing.
The systematic analysis of the gendering of poverty debates and the
use of corpus linguistics to bring such social practices to light correlates
well with CDA’s political focus. Wodak (2001: 9) notes that the critical
element of CDA refers to ‘taking a political stance explicitly’ whilst also
having distance from the data and being able to embed the data within
society and social norms. Thus, to combine corpus linguistics, which
provides a certain distance from the data, with CDA, one has to begin
corpus work acknowledging the political purpose of a particular analysis.
It is important for researchers to acknowledge their own position in rela-
tion to their object of analysis. In the present case, both authors are from
middle-class, white British backgrounds and have no first-hand expe-
rience with poverty. As such we cannot present an insider’s view. Our
knowledge of poverty comes from observing those in poverty, either first
hand or through social networks, or mediated through other channels,
such as second-hand accounts of poverty presented in the mass media. It
is on these media-controlled accounts that we focus our analysis.
37
2.5 Summary
This chapter has served to introduce readers to the theoretical under-
pinnings and applications of corpus linguistics and CDA. The benefits
of corpus linguistics include the ability to process large amounts of
data—much more than could be easily facilitated using manual anal-
ysis alone—and can lead to the identification of patterns of language
use across large datasets. For the present study, using a corpus-based
approach helped to identify the use of place-names across a total of
almost four hundred million words. The way that textual data is pre-
sented in corpora, and the ability to generate concordances lines, is
also particularly well suited to geotagging place-names (see Chapter 3)
and manipulating textual data so that it can be used in GIS. Our
focus on CDA is complementary to our use of corpora, as it facilitates
a consideration of textual (and statistical) representations of poverty
within their wider social context. Thus, not only can we plot the refer-
ences to place-names included in media texts, and thus compare them
to statistical measures of poverty, we can also take a more qualitative
look at exactly how ideologies of UK poverty are encoded within
texts.
However, whilst the final section of this chapter discussed how
corpus linguistics and CDA can be used in combination, there was no
real consideration of geography. This is because, traditionally, corpus
linguists have not explored the geographies within texts that they have
analysed. Whilst there may be some consideration about where a text
is from when compiling a corpus—for example, if the aim is to con-
struct a corpus of British English or Australian English—there tends to
be little-to-no consideration of the places mentioned within texts. In
Chapter 3 therefore, we switch our focus to explore the basics of the
ways in which Geographical information systems (GIS) represent and
analyse features in geographical space. This provides the underpin-
ning that allows Chapter 4 to bring together corpus-based and GIS
approaches to create GTA, the major methodological advance that
allows us to explore the geographical representations of poverty in the
remainder of the book.
38
References
Altheide, D. L. 2000. Tracking Discourse and Qualitative Document Analysis.
Poetics 27 (4): 287–299.
273–306.
Baker, P., & T. McEnery (eds.). 2015. Corpora and Discourse Studies.
Biber, D., S. Conrad, & R. Reppen. 1994. Corpus-Based Approaches to Issues
in Applied Linguistics. Applied Linguistics 15 (2): 169–189.
Bourdieu, P., & L. Wacquant. 2004. Symbolic Violence. In N. Scheper-
Hughes & P. Bourgois (eds.). Violence in War and Peace: An Anthology.
Oxford: Blackwell, pp. 272–274.
Breit, E. 2010. On the (Re)Construction of Corruption in the Media: A
Brezina, V. 2018. Statistics in Corpus Linguistics: A Practical Guide. Cambridge:
Cambridge University Press.
Fairclough, N. 2001. Language and Power (second edition). London:
Longman.
Gregory, I. N. 2009. Comparisons Between Geographies of Mortality and
Deprivation from the 1900s and 2001: Spatial Analysis of Census and
Mortality Statistics. British Medical Journal 339: b3454, 676–679.
Hardie, A. 2012. CQPweb—Combining Power, Flexibility and Usability in a
Corpus Analysis Tool. International Journal of Corpus Linguistics 17: 380–409.
Hardt-Mautner, G. 1995. Only Connect. Critical Discourse Analysis and
Corpus Linguistics. UCREL Technical Papers 6. http://ucrel.lancs.ac.uk/
papers/techpaper/vol6.pdf. Accessed 18/10/2009.
Kennedy, G. D. 1998. An Introduction to Corpus Linguistics. London:
Longman.
Kress, G., & T. van Leeuwen. 2001. Multimodal Discourse: The Modes and
Media of Contemporary Communication. London: Arnold.
39
London: Oneworld.
Lundström, R. 2013. Framing Fraud: Discourse on Benefit Cheating in
Sweden and the UK. European Journal of Communication 28 (6): 630–645.
Mautner, G. 2007. Mining Large Corpora for Social Information: The Case of
Elderly. Language in Society 36 (1): 51–72.
McCarthy, M., & R. Carter. 2004. Focus-on Issue: Corpus Linguistics (Part
I)—Introduction. Journal of Pragmatics 36 (2): 147–148.
McKenzie, L. 2015. Getting By: Estates, Class and Culture in Austerity Britain.
London: Polity Press.
Paterson, L. L. Forthcoming. Electronic Supplement Analysis of Multiple
Texts: Exploring Discourses of UK poverty in Below the Line Comments.
Rayson, P. 2008. From Keywords to Semantic Domains. International Journal
of Corpus Linguistics 13 (4): 519–549.
Richardson, J. E. 2006. Analysing Newspapers: An Approach from Critical
Discourse Analysis. Basingstoke: Palgrave Macmillan.
Scheuer, J. 2003. Habitus as the Principle for Social Practice: A Proposal for
Critical Discourse Analysis. Language in Society 32 (2): 143–175.
Skeggs, B., & H. Wood (eds.). 2011. Reality Television and Class. London:
Stubbs, M. 2001. Texts, Corpora, and Problems of Interpretation: A Response
to Widdowson. Applied Linguistics 22 (2): 149–172.
Tognini-Bonelli, E. 2001. Corpus Linguistics at Work. Amsterdam: John Benjamins.
Wodak, R. 1999. Critical Discourse Analysis at the End of the 20th Century.
Research on Language and Social Interaction 32 (1–2): 185–193.
Wodak, R. 2001. What CDA Is About—A Summary of Its History, Important
Concepts and Its Development. In R. Wodak & M. Meyer (eds.). Methods
of Critical Discourse Analysis. London: Sage, pp. 1–13.
Wodak, R., & M. Meyer. 2009. Critical Discourse Analysis: History, Agenda,
Theory and Methodology. In R. Wodak & M. Meyer (eds.). Methods of
Critical Discourse Analysis (second edition). London: Sage, pp. 1–33.
3
Geographical Information Systems
and Textual Sources
At its core the geographical information systems (GIS) data model

consists of a table of data in which each row is linked to a spatial loca-
tion in the form of a point, line, polygon or pixel. While this is a highly
effective way of representing some types of data it has traditionally been
thought to be incompatible with others, particularly unstructured tex-
tual data such as corpora. Solutions to this do exist, although, as cur-
rently available, they rely on converting from an unstructured text to
a table of attribute data with combined spatial data. While this is not
a perfect solution, it does allow texts to be explored and analysed geo-
graphically. This, in turn, allows us to conduct the geographical analyses
of text that form the main theme of this book.
‘Humans are spatial beings’ who ‘exist and interact in natural and
human environments that occupy space over time’ (Siebert 2000:
538). One of the major developments in IT in recent years has been
the exponential rise in the availability of digital corpora. As described
in the previous chapter, this has led to the development of corpus lin-
guistics as an analytic approach that makes use of the availability of

https://doi.org/10.1007/978-3-319-93503-4_3
42
this type of source. Space and geography have, however, largely been
neglected by corpus linguistics. This is at least in part because the tools
and approaches used in corpus linguistics are not well suited to explor-
ing location. Since the 1990s geographers have increasingly used GIS
to represent, analyse, and visualise data that can be mapped. A con-
tinuing criticism of GIS, however, is that it is primarily suited for use
with quantitative sources, usually structured in tabular form. In their
traditional form GIS are thus not well suited to the analysis of textual
sources.
To date, developments within the fields of GIS and corpus linguistics
have occurred in parallel with little overlap: GIS has largely not made
use of textual sources while corpus analysis has largely ignored geogra-
phy. Bringing the two together, to allow texts to be analysed within a
GIS framework, requires two developments: firstly the identification of
place-names within the text, and secondly the linking of place-names to
coordinates to allow the place-name to be mapped. This can be achieved
using a technique called geoparsing which uses techniques from Natural
Language Processing (NLP) to identify candidate place-names within
the text, and matches these to a gazetteer1 to allocate them to a coordi-
nate (Grover et al. 2010). This allows the place-names within the cor-
pus to be mapped and facilitates analysis of the geographies within texts
(Gregory et al. 2015). Once places have been mapped, Geographical
Text Analysis (GTA)—described in detail in Chapter 4—can be used
to analyse the text in more sophisticated ways. This chapter introduces
GIS and its constituent fields. Section 3.1 introduces the technology
as it has been developed to handle quantitative data and the limited
extent to which the qualitative have also been incorporated. Section 3.2
explores geoparsing, how it works, and why gazetteers are so fundamen-
tal to it. Section 3.3 explores some of the more conceptual issues that
problematise representing space in this way.
1A gazetteer is a directory or index of place-names and their corresponding coordinates. It may
also contain additional information about a place including, alternative spellings, population sta-
tistics, and notes on physical features.
3 Geographical Information Systems and Textual Sources
43
3.1 An Introduction to GIS

Geographical information systems are frequently thought of as a type
of computer mapping system. This over-simplifies what a GIS is able
to do. In many ways a GIS is a type of database which allows users to
query, visualise, and analyse data that are spatially referenced, in other
words, data that have a location associated with them. This is achieved
because every item of data, represented as a row in a database table,
is linked to a location on the map (see, for example, DeMers 2008;
Heywood et al. 2011; Lo and Yeung 2006). While this data model is
relatively crude, GIS-based approaches have been applied to a wide
range of topics across the Earth and social sciences. GIS has become a
fundamental tool for many geographers and has led to the development
of a field called geographical information science (GISc) which explores
how to use geographical information effectively and appropriately
(Goodchild 1992). The adoption and acceptance of GIS within geog-
raphy and beyond has been far from uncontroversial (see the essays in
Pickles 1995 and the response by Openshaw 1997). In its early stages a
wide range of criticisms were made particularly about the quantitative
nature of GIS and the perception that the field was driven by technical
developments which were adopted uncritically. Responses to these crit-
icisms led to the development of the sub-field of critical GIS in which a
more sensitive approach to using the technology is adopted (Schuurman
2004). GIS has also started to move away from its quantitative origins
with the development of qualitative GIS which has been used to study
issues such as the geographies of fear of crime or prejudice (Cope and
Elwood 2009).
The first stage in understanding what GIS is and the potential it
offers to researchers across disciplines is to stand back from it and
understand what geographical information is. In theory any infor-
mation that refers to a location on the Earth’s surface is geographical
information. In practice, however, information really only becomes geo-
graphical information when the researcher is interested in how different
locations are different from each other. One of the most obvious exam-
ples of geographical information in the social sciences is census data.
44
(a) (b)
ID Name Population Unemployment
1 Newton 4235 862
2 Sutton 1729 209
3 Wood Side 7621 1012
4 Castlehill 6980 562
Fig. 3.1 Using a combination of a spatial data and b attribute data to represent

some hypothetical census data
To take the census, the study area (usually a country) is subdivided into
precisely defined areas such as counties, districts, or output areas (OAs).
The number of people, and sub-divisions such as their sex, age, occupa-
tional status, ethnicity, and so on, are then counted for each area with
these counts being recorded in a table where there is a row for every
area and the columns provide information on the population structure,
such as total population, numbers of people in each occupation, etc.
Thus, the census combines information about location (in the form of
the areas used to count the data) that say where the data refer to, with
tabular information that says what is at each location.
In GIS parlance, the tabular data is referred to as attribute data.
Each row of attribute data is linked to a geographical representation of
its location which is called the spatial data. With census data, the spa-
tial data will be representations of the zones used to collect and cate
gorise the data. This structure is shown in Fig. 3.1 which represents the
spatial and attribute data for some hypothetical census statistics.
Spatial data can be represented in a number of different ways. The
simplest form of spatial data is a point, in which a location is repre-
sented using a single coordinate pair (x, y ) as shown in Fig. 3.2a. Points
can be used to represent a wide variety of features including individual
addresses or the locations of buildings, such as hospitals or supermar-
kets. Depending on scale, points can also be used to represent larger fea-
tures such as towns and cities. Linear features such as roads, railways, or
rivers are represented using lines which, as Fig. 3.2b shows, are created
by joining together two or more points. Areas or zones, such as cen-
sus districts or lakes, are represented using polygons. As Fig. 3.2c shows,
45
(a) (b) (c)
Fig. 3.2 Vector data representing a points, b lines and c polygons. Each point,
line segment, and polygon is linked to its own attribute data, as shown in Fig. 3.1
a polygon is created from one or more lines that completely enclose

an area. Points, lines and polygons are thus all created from coordinate
pairs with every point, line segment, or polygon having its own attrib-
ute data. In GIS parlance, points, lines and polygons are termed vec-
tor data. Vector data are good at representing discrete features such as
buildings, towns, roads, rivers, lakes and administrative areas.
An alternative type of spatial data are raster data. Rather than
sub-divide the study area into separate features, raster data attempt to
represent continuously varying features such as height or the spread of
pollution. It does this by sub-dividing the study area into small regular
pixels, usually squares, with each pixel having its own value, as shown in
Fig. 3.3. Typically raster data will be represented graphically with high
values being shaded in darker colours. The pattern that the pixels create
will be of interest in describing the way that the height of the terrain
varies or a pollution incident has spread. Raster data are less common
in examples from human geography but, as we will see, are frequently
used to simplify point data by representing the density of points across a
study area.
Data on different themes are stored in GIS layers. Each layer con-
sists of one type of data and contains both spatial and attribute data. A
layer’s spatial data will normally be georeferenced. This means that the
coordinates that it uses are in a real-world coordinate system which can
either use a map projection, such as British National Grid, or be in lati-
tude and longitude.
The data model described above, in which spatial and attribute data
are linked together is at the core of GIS. It allows the researcher to ask
46
0 0 0 1 3 4 7 8
0 0 1 2 3 4 5 6
0 0 3 4 5 3 3 4
1 2 5 7 7 5 3 3
2 5 6 9 8 7 5 3
4 7 9 9 8 6 4 2
4 6 8 8 7 5 3 1
3 5 7 8 6 4 1 0
Fig. 3.3 An example of raster data. Each cell is of known size with the numeric
value representing an attribute such as height or density
questions of the attribute data and map the results, or to select locations
on the map and query their attribute data. The first of these would ena-
ble a question such as ‘where are places with unemployment rates above
10% found?’, the second would ask ‘what percentage of people at this
place are unemployed?’.
While this is a very simple structure, it opens up much potential
for understanding geographical phenomena. Four major opportunities
can be identified: structuring of data, data integration, visualisation,
and spatial analysis (Gregory et al. 2003). Structuring simply refers to
the fact that the data model allows us to know more about the way the
data are related to each other. For example, we know what is at a par-
ticular location, what is at neighbouring locations, and so on. As all of
the data within a GIS database have been georeferenced to real-world
location, theoretically location can be used to integrate any data within
GIS. Thus, for example, if we have data on the addresses of individ-
ual shops as points, census data representing the population for poly-
gons, and data on the transport network as lines, all of which have been
taken from different sources, these can be brought together to explore,
47
for example, how shop locations compare to background populations,

taking into account likely transport routes. From a visualisation per-
spective, GIS is obviously well suited to creating and exploring maps.
Higher-tech visualisations can also be produced using GIS, including
animations to show change over time and virtual worlds to attempt to
recreate landscapes digitally.
Spatial analysis (also known as spatial statistics) is a form of analy-
sis that explicitly includes location. This can be interpreted informally
as simply any analysis in which location is included. However, there is
also a field of statistics called spatial analysis in which the coordinate
data are made an explicit part of a statistical analysis. An important fea-
ture of statistical spatial analysis is that the results of the analysis will
change if the features under study are moved. There are clear analogies
between mapping and statistical spatial analysis on one hand and, for
example, scatter plots and regression analysis in conventional statistics.
Scatter plots may suggest a relationship but techniques such as regres-
sion are needed to confirm the pattern really exists and to quantify it. In
the same way, a map may suggest a pattern or relationship, while spatial
statistics are used to confirm this and to quantify its extent.
Spatial analysis techniques vary from relatively simple techniques
which attempt to identify whether a series of points cluster or not, to
techniques that attempt to quantify how the relationship between two
or more variables varies over the study area (Fotheringham et al. 2000;
Lloyd 2011). One important feature of spatial analysis is that there
are two potential types. On the one hand there are conventional sta-
tistical summaries, known as global or whole map statistics. These say,
for example, that the points within a dataset tend to cluster across the
study area or that there is a relationship between two variables across the
study area. There is also the potential to create local statistics which can
say, for example, that the data cluster in these areas but not in others, or
that a relationship between two variables is different in different parts of
the study area (Fotheringham 1997). Local statistics are usually repre-
sented using maps.
GIS is well suited to analysing quantitative data which can be allo-
cated to clearly defined locations. GIS has been used extensively to
analyse census data and in many other types of demographic analysis
48
(Reibel 2007). Epidemiological studies of health and disease patterns

have been a rich seam for GIS (Cromley and McLafferty 2012; Rushton
2003). Events such as crimes or road traffic accidents have also been
extensively analysed using GIS as precise locations are available as part
of the recording of these events (Chainey and Ratcliffe 2005). A sig-
nificant criticism of GIS is, however, that its analyses tend to be data
led. If the required data are available with both spatial and attribute
components then GIS is well suited to analysing it. If not, it cannot be
incorporated into a GIS database and is thus likely to be excluded from
the analysis. This criticism is valid and suggests that if GIS is to spread
beyond its strongholds in quantitative social science and the Earth
sciences it needs to be able to handle a wider range of sources.
One way in which this has been done is through the development
of the field of qualitative GIS (Cope and Elwood 2009). While a tra-
ditional GIS analysis of crime, for example, would analyse the types of
crime recorded at specific locations, a qualitative GIS analysis might
explore fear of crime by asking participants where they felt safe and
unsafe. Their responses are then georeferenced, perhaps by asking the
participants to mark the relevant locations on maps, and can be ana-
lysed. One of the best examples of this type of work is Mei Po-Kwan’s
analysis of fear of crime among Muslim women in Columbus, Ohio
after the September 11th attacks (Kwan 2008).
Although qualitative GIS broadens the scope for GIS analysis it
requires the same basic technology. Attribute data are likely to be nom-
inal, ordinal, or interval data in tabular form albeit trying to quantify
something intangible such as perception. Spatial locations will still be
represented by points, lines, polygons, or pixels but these become some-
what problematic due to the precise nature of these features. While it is
sensible to represent a census district using precise polygon boundaries,
representing the places where people feel unsafe is likely to be more dif-
ficult. In some cases, such as ‘the other side of the tracks’ or in clearly
defined neighbourhoods it is possible for a person to put precise bound-
aries to areas they regard as unsafe. It is more likely, however, that they
can say where they definitely feel unsafe and definitely feel safe but that
a boundary placed between them would be highly arbitrary. GIS is thus
49
not good at handling this type of situation due to the imprecise nature
of the spatial data.
3.2 Geoparsing: The Link Between Texts

and GIS
Converting a text into a GIS layer requires three stages: first the place-
names need to be identified, second a coordinate or other spatial rep-
resentation needs to be found for the place-names, and third, a way of
appropriately converting the required co-text into tabular form needs to
be found. The first two stages of this are referred to as geoparsing, a
process by which place-names are identified in a text or corpus in the
geotagging stage and then allocated to a coordinate in the georesolu-
tion stage (Grover et al. 2010). Geotagging is a NLP task that makes
use of the fact that place-names are proper nouns. The use of Named
Entity Recognition (NER) algorithms to identify proper nouns is a
well-established part of the NLP toolkit. Once the proper nouns have
been identified, further rules can be used to produce a list of candidate
place-names and separate these from other proper nouns such as peo-
ple’s names.
The georesolution phase then involves matching the candidate place-
names with one or more gazetteers which will provide coordinates for
the place-names. In its simplest form a gazetteer is a database table that
lists place-names and their associated real-world coordinates. Frequently
gazetteers will host a range of other information also associated with
the place-name such as variant spellings, information about the admin-
istrative hierarchy, such as which district, county or state the place is
found in, and what type of feature it is, such as a settlement, moun-
tain, forest, and so on (Southall et al. 2011; Berman et al. 2016). There
are a number of publically available gazetteers, perhaps the most widely
used of which is Geonames (Geonames, n.d.). This is freely available
with global coverage and is increasingly updated using crowd-sourc-
ing. The Getty Thesaurus of Geographical Names (TGN) is another
widely used general purpose gazetteer (Getty Research Institute, n.d.).
50
Other gazetteers have more specific purposes, for example, the

Ordnance Survey’s 1:50,000 gazetteer gives the 1 km grid square in
which place-names appear on its 1:50,000 maps (Ordnance Survey
2014), while the Pleiades gazetteer was specifically created to incorpo-
rate place-names from the Ancient World (Pleiades, n.d.; Simon et al.
2016).
The accuracy of the coordinate information will also be crucially
affected by the gazetteer. Some gazetteers are very clear about what their
coordinates refer to. The OS 1:50,000 gazetteer, for example, gives the
coordinates of the centroids of the 1 km grid squares of place-names
that are included on the 1:50,000 series maps. This makes it relatively
easy to understand: place-names are included if they are included on
that series of maps and their coordinates show where the place-name
is located on the map, which is not necessarily exactly where the place
actually is. It also means that the coordinates form a regular grid of loca-
tions that are one kilometre apart in both x and y. Thus, assuming the
place can be represented by a precise point location and that the loca-
tion of the place-name on the map accurately and sensibly represents
the location of the place, the OS gazetteer is likely to locate the place
to within 500 metres in x and y of where it should be. With other gaz-
etteers, such as Geonames, it is harder to know exactly what the point
provided represents.
In theory, a geoparser will provide a coordinate for every place-name
within a text. In reality this process is complex and error prone. The
main reason for this is that place-names are far from simple and unam-
biguous and therefore are difficult for computers to handle in a fully
automated manner. The first major set of problems lies at the geotag-
ging stage. It is often difficult to determine whether a proper noun
refers to a place, person, or other feature. For example, ‘Lancaster’ is a
town in the north-west of England. However ‘Stuart Lancaster’ is a per-
son (a recent England rugby coach), the ‘Duke of Lancaster’ may be a
person or may be a public house, and the ‘Lancaster’ was a well-known
Second World War aeroplane. Rule-based systems, such as assuming
that a word is not a place-name if it is preceded by a title such as ‘Mr’,
‘Ms’, or ‘Duke of ’, or by a first name taken from a lexicon, or followed
by a word such as ‘bomber’ or ‘aeroplane’ may help to resolve this but
51
will not be fool-proof and will require manual checking. A second

problem is that place-names frequently consist of more than one word
making them both harder to geotag and to disambiguate. ‘Oxford’,
‘Oxford Street’ and ‘Oxford Road’ are entirely different places, as are
‘York’ and ‘New York’.
Another source of ambiguity is that place-names may refer to more
than one place. Many place-names that originated in Europe are also
found in North America including obvious ones such as Boston
(Massachusetts), London (Ontario), and Paris (Texas), and less obvi-
ous ones such as ‘Lancaster’ which is also found in Pennsylvania, New
Hampshire, and California. Resolving to the correct continent may
not be a major problem, however, similar ambiguities occur at more
local levels. Within Britain there are several different places called, for
example ‘Newport’, ‘Newcastle’, ‘Wellington’, ‘Barrow’, ‘Bradford’, and
‘Whitchurch’. The place-name Kensington was a source of ambiguity
in reports concerning poverty, as it referred to both the Royal borough
of Kensington and Chelsea in London—which tends to be associated
with affluence—and the inner city area of Kensington, Liverpool, which
is associated with high poverty and unemployment rates. Sometimes
full names may be used to disambiguate place-names, for example,
‘Newcastle upon Tyne’ or ‘Newcastle under Lyme’ but these are not
used universally and are often left to the reader to determine.
Spelling variations represent a further set of problems. ‘Saint Helen’s’
can be spelled in six different ways depending on whether and how
‘Saint’ is abbreviated (‘Saint’, ‘St.’ or ‘St’) and whether ‘Helens’ is given
an apostrophe. ‘Newcastle’, ‘Newcastle on Tyne’, ‘Newcastle-upon-
Tyne’ and so on represent a further set of complexities. In addition to
these genuine variations, there can also be variations caused by spelling
errors or data capture issues particularly where sources have been dig-
itised using optical character recognition (OCR) technology (Tanner
et al. 2009). Finally, there are also issues around what does and does not
constitute a place-name. Ultimately this is a decision for the researcher.
In most cases, people geoparsing a text will be most interested in the
names and locations of features such as settlements. A decision needs
to be taken about whether to include, for example, street names or
the names of local features such as pubs, hotels, and other buildings.
52
At the other extreme it may be sensible to exclude high-level fea-

tures such as countries, British counties, and US states, or at least to
not represent them as points. Beyond this, there may be further deci-
sions that a researcher needs to make. For example, we had to consider
‘Westminster’ as a special case. We chose to keep those instances where
Westminster referred to an identifiable geographical area, and remove
those where it was used metonymously to refer to the UK parliament
and/or the government.
Automated solutions can be found to help to resolve some of these
problems. Disambiguation can be assisted by using the locations of
place-names that occur nearby in the text and selecting the option that
is nearest to these. Spelling variations can be handled by having com-
mon variants in the gazetteer or through the use of fuzzy matching
techniques. While these technical solutions may help, in reality place-
names are so complex that fully geoparsing a large corpus in an entirely
automated manner is likely to leave a situation where the result is has a
considerable number of errors and where the researcher will be unaware
of what the implication of these errors is likely to be. Thus geoparsing
must be complemented by at least some manual analysis in order to
ensure (and increase) the accuracy of the results.
A way of achieving this, which we use in the present study, is termed
concordance geoparsing (Rupp et al. 2014; Gregory et al. 2015).
Rather than geoparse the entire corpus, this works on the assump-
tion that only relatively small parts need to be geoparsed at one time.
The user selects a search term and extracts a relatively wide set of
co-text around this, such as 50 words left and right. This co-text,
which comes in the form of concordance lines (see Sect. 2.1), is then
geoparsed. The results of the geoparsing can then be explored using
both the concordance lines and by mapping the results. Where errors
or ambiguities are spotted these can be corrected and, importantly, the
corrections written to a file that means that when a new search-term is
geoparsed these corrections can be applied automatically. In many cases,
corrections are likely to allow the gazetteer to be improved or updated.
In other cases, they may only be resolved as flagging, for example an
ambiguous place-name as needing manual intervention. Concordance
53
Trial re-tests of claimants in <enamex long="-2.23" lat="53.8" st_name="Burnley">Burnley</enamex> and

<enamex long="-2.1" lat="57.13" st_name="Aberdeen">Aberdeen</enamex> found 68.6 per cent either did not
have a valid claim and were unable to prove they were too ill to take a job
Fig. 3.4 A fragment of geoparsed newspaper text taken from the Daily Mail
geoparsing allows users to geoparse in an iterative manner in ways that

means that they understand the issues around how place-names are used
in the corpus and can correct them as appropriate. It also enables gaz-
etteers to be improved and enhanced. Concordance geoparsing is not
mutually exclusive from geoparsing a whole corpus. It can be used in
the first instance to gain an understanding of place-names in the corpus
and to build up corrections that can be applied when the whole corpus
is geoparsed.
Whatever form of geoparsing is used, the result will be that geotags
are added to either the entire corpus or to concordance lines derived
from it. These geotags are likely to be in XML format. They will identify
place-names and provide other information as attributes. This attribute
information will include the coordinates, and may also include addi-
tional information such as a standardised version of the place-name and
perhaps other information derived from the gazetteer. An example of a
geoparsed fragment of newspaper text is shown in Fig. 3.4. The place-
names are identified using an XML tag, ‘enamex’ whose name, by con-
vention, is surrounded by angle brackets (< >). The original spelling of
the place-name is enclosed within the tag in the form <enamex> place-
name </enamex> where the second tag, whose name starts with a slash,
indicates the end of the place-name. The first tag also includes the
attribute information which, in this case, includes latitude (lat), lon-
gitude (long), and a standardised version of the place-name’s spelling
(st_name).
The final stage in the process is to convert from this text format
to the tabular format required by GIS. Technically this is reasona-
bly straight-forward. A program identifies each geotag and converts it
into a row in a table. Each row represents a place-name instance from
the corpus. At a minimum it will include the place-name and a lati-
tude and longitude or x and y figures for it. A range of other informa-
tion is also likely to be desirable including: standardised versions of the
54
place-name, any other information from the geotag, information on

where in the corpus the instance has been extracted from, such as a file-
name and word number, and perhaps the search-term used, if concord-
ance geoparsing. Finally some co-text around the place-name is likely to
also be included to provide some context about what is being said about
the place. However, this is likely to be fairly limited as most database
formats limit text fields to a maximum of 255 characters.
Effectively, therefore, geoparsing starts with an unstructured text
and converts it into a table in which each row represents a place-name.
Converting this into a GIS layer is straight-forward, most types of GIS
software can easily convert coordinate data in a table into spatial data,
so that a table can be converted into a point layer. Fundamentally, this
manipulation of text into a new format that stresses its geography does
mean the free-flowing nature of the text has been lost. However, the
geoparsing process puts us in a position where the text can be explored,
visualised, and analysed geographically in ways that were previously
impossible, opening the potential for new forms of analysis.
3.3 Realities of Place and Space

As described above, GIS represents features as points or other graphic
objects on a Euclidean plane and geoparsing allows us to georefer-
ence and map a text. One key issue, however, is the exact relationship
between ‘place’ as a construct and place-names and extracted and geo-
referenced from a text. This issue concerns the notion of place as a con-
ceptual/psychological reality, but also depends on what language users
actually mean when they use a place-name. Whilst references to a geo-
graphical space do occur in our corpora, place-names can also be used
as a metaphor/hyponym for the people living within particular geo-
graphical boundaries. Thus, the locations mentioned in the corpus, even
though they can be assigned global coordinates, cannot simply be taken
to depict a particular area of landmass. It is the critical investigation of
what these terms mean, how they are used, and how they are evaluated,
which interrogates the ideological function of place-names in media
texts.
55
The location of the production of texts and the places they refer to
can index powerful strategies. For example, in 2011 South Sudan was
formally recognised as a country, but its conceptual realisation as a sepa-
rate entity in global politics would not have been possible if institution-
ally-produced texts rejected its new, independent identity by refusing to
use the signifier ‘South Sudan’ when referring to a particular geographi-
cal location/ideological entity. Similarly, the renaming of a geographical
area has ramifications for its assumed identity and implied ownership.
The history of Heraklion, Crete is a good example of this. The city’s cur-
rent name is rooted in Greek mythology, but at several points in his-
tory it has been given names that reflected its Byzantine (Kastro), Arab
(Rabdh el Khandaq/Chandax) and Venetian (Candia) rule. Thus we can
see how linguistic choices can be socially/politically powerful in refer-
ences to place.
Whilst name changes can illustrate the power of linguistic choice,
critical discourse analysts must question the most foundational concepts
within their research. Thus, we must ask what is meant by place as an
integral part of investigating how references to place are used within
wider discourse frameworks. Cresswell (2004: 1) notes that ‘no-one
quite knows what they are talking about’ given that place is ‘not a spe-
cialized piece of academic terminology’ and as such, does not have a sin-
gular definition.2 He argues that ‘[w]riting about and researching place
involves a multi-faceted understanding of the coming together of the
physical world (both “natural” and “cultural”), the processes of mean-
ing production and the practices of power that mark relations between
social groups’ (Cresswell 2004: 122). However, Porter et al. (2015: 33)
note that further work within Historical GIS may shed light on how
people use the term place and ‘could usefully examine what “place”
meant’ to the authors of texts, questioning whether place-names were
2See Cresswell (2004) for an overview of different approaches within human geography to
notions of ‘place’. He notes that geographers and philosophers such as Sack (1997) and Malpas
(1999) argue that society and geography are connected through ‘place’, but ‘the realm of the
‘social’ has no particular privilege in discussions of place’ (2004: 31). ‘Malpas and Sack are argu-
ing that humans cannot construct anything without being first in place – that place is primary to
the construction of meaning in society ’ (Cresswell 2004: 32).
56
‘primarily a reference to an environment or shorthand for the “type” of

people who lived there’. Cresswell (2004: 103) acknowledges the power
of place within wider society. Suggesting that places can help to create
notions of ‘in’ and ‘out’ or ‘us’ and ‘them’.
Questions of naming highlight potential ambiguities in noun use and
problematize the concept and processes of naming more generally. They
also illustrate the homogenising function of using place-names in dis-
course. An international news report about ‘Wales’, for example, could
construct the entire country as a uniform and tangible whole. In turn,
its population, unless specified elsewhere in the text, are all assumed to
hold the same social values, economic status, and/or political beliefs. This
feature of language is especially pertinent when place-names are used in
their possessive forms, e.g. ‘London’s property owners’, ‘Bradford’s youth’.
This effect is extended even further when abstract concepts are used and
geographical locations are personified, e.g. ‘Yorkshire’s fear’, ‘Ireland’s joy’.
As an example, Dreier (2005) writes about the asymmetrical distri-
bution of news stories across different neighbourhoods and cities in the
US, noting that particular urban (poor) areas tend to be frozen out of
media reports unless they are related to something sensational or break-
ing news. Whilst fundamentally his analysis relies on considerations of
social and physical space and place, he does not systematically look at
the geography of the places that he mentions. Such an analysis would
support his work significantly, potentially providing robust evidence for
his arguments. One flaw of Dreier’s paper is that he does not consult any
primary data, but instead draws together anecdotes from the wider liter-
ature (and elsewhere). However, his work does show the gap in the liter-
ature where a combination of CDA and GIS could provide evidence to
support or challenge hypotheses related to how often and in what capac-
ity particular geographical locations are referred to in the news media.
Issues of scale and representation may also be important. It is pos-
sible that the geographical area referred to in a text remains constant,
whilst the lexemes used to refer to it change. For example, a newspa-
per headline could include a reference to Nottingham, but within the
lead paragraph, there could be references to a smaller, more specific
location, such as West Brigford, and, further still, a later reference to an
even more specific location like Lady Bay. Here, the term ‘Nottingham’
is being used to refer to a very small area of landmass that constitutes
57
part of a suburb of the city. Thus, it is likely that, when using GIS, any
characteristics associated with Lady Bay could be mapped to, and thus
associated with, the whole of Nottingham. This argument is particularly
salient when considering ethnographic work undertaken in the St Ann’s
area of Nottingham by McKenzie (2015) which references the con-
trasts between St Ann’s and more affluent parts of the city, such as West
Brigford; plotting references to Nottingham (meaning St. Ann’s) in the
same way as references to Nottingham (meaning West Brigford) in the
same location on a map will not give an accurate picture of how the two
terms were used. This problem can also be turned on its head, insofar
as the lexemes used in a text may remain constant, but refer to differ-
ent geographical locations, e.g. references to events located in Soho and
Kensington could both be referred to using the term ‘London’. Close
analysis of the surrounding co-text within which place-names are used
will highlight if there are repeated patterns of reference like this that our
analysis must address.
Concordance analysis can help with most of these issues as the
co-text of a search term or place-name provides more information about
how exactly place-names are used. Close reading of concordances can
show which characteristics are associated with a particular geograph-
ical area and we can perform a version of keyword analysis to deter-
mine how the co-text surrounding place-names differs (see Chapter 4).
In order to address issues concerning the different scale of place-name
mentions (county, city, borough, district, etc.) we can tailor our analysis
to the different levels, aggregate place-names mentioned to defined geo-
graphical areas (such as a local authority district) and also use density
smoothing (see Chapter 4) to even out the effects of different types of
place-names.
3.4 Summary
GIS is based around a crude but effective data model in which a table of
data is provided with locational information using spatial data—points,
lines, polygons, or pixels. This model allows the researcher to ask ques-
tions of the database including where are these features located? and
58
what is at this location? This provides a highly effective framework for

understanding the geographies of the features under study. In particular,
it allows the researcher to structure the data, integrate it with other data
using location, visualise it through maps and other representations, and
to analyse it in ways that explicitly include location.
A major limitation of this data model is that it requires a tabu-
lar structure in which each feature in the database is a separate record
linked to one, and only one, spatial reference. Despite this, it is still pos-
sible to represent corpora within GIS. This requires the use of geopars-
ing in which each place-name is identified and provided with a spatial
reference. This is then used to convert the text to a table with each
place-name forming a separate row. Once a text has been geoparsed it
can quickly be converted into GIS layer for subsequent analysis. As we
will see in Chapter 4, this analysis can start with simple dot maps but,
to be effective it needs to draw on a combination of techniques from
spatial analysis to focus on geography, corpus linguistics to focus on
text, critical discourse analysis to focus on the ideologies associated with
what is being said about place, and close reading of appropriate parts
of the text. By drawing these together, the remainder of the book will
explore newspaper discourses about poverty and place in the UK.
References
Berman, M. L., R. Mostern, & H. Southall. 2016. Introduction. In M. L.
Berman, R. Mostern, & H. Southall (eds.). Placing Names: Enriching and
Integrating Gazetteers. Bloomington: Indiana University Press, pp. 1–11.
Chainey, S., & J. Ratcliffe. 2005. GIS and Crime Mapping. Chichester: Wiley.
Cope, M., & S. Elwood (eds.). 2009. Qualitative GIS: A Mixed Methods
Approach. London: Sage.
Cresswell, T. 2004. Place: A Short Introduction. London: Blackwell.
Cromley, E. K., & S. L. McLafferty. 2012. GIS and Public Health. London:
Guilford.
DeMers, M. 2008. Fundamentals of Geographical Information Systems (fourth
edition). Chichester: Wiley.
59
Dreier, P. 2005. How the Media Compound Urban Problems. Journal of

Urban Affairs 27 (2): 193–201.
Fotheringham, A. S. 1997. Trends in Quantitative Methods I: Stressing the
Local. Progress in Human Geography 21: 88–96.
Fotheringham, A. S., C. Brusdon, & M. Charlton M. 2000. Quantitative
Geography: Perspectives on Spatial Data Analysis. London: Sage.
Goodchild, M. F. 1992. Geographical Information Science. International
Journal of Geographical Information Systems 6: 31–45.
Gregory, I. N., C. Donaldson, P. Murrieta-Flores, & P. Rayson. 2015.
Geoparsing, GIS and Textual Analysis: Current Developments in Spatial
Humanities Research. International Journal of Humanities and Arts
Computing 9: 1–14.
Gregory, I. N., K. Kemp, & R. Mostern. 2003. Geographical Information and
Historical Research: Current Progress and Future Directions. History and
Grover, C., R. Tobin, K. Byrne, M. Woollard, J. Reid, S. Dunn, & J. Ball.
2010. Use of the Edinburgh Geoparser for Georeferencing Digitized
Historical Collections. Philosophical Transactions of the Royal Society A 368:
3875–3889.
Heywood, I., S. Cornelius, & S. Carver. 2011. An Introduction to Geographical
Information Systems (fourth edition). Pearson: Harlow.
Kwan, M.-P. 2008. From Oral Histories to Visual Narratives: Re-presenting
the Post-September 11 Experiences of the Muslim Women in the United
States. Social and Cultural Geography 9 (6): 653–669.
Lo, C. P., & A. K. W. Yeung. 2007. Concepts and Techniques of Geographical
Information Systems (second edition). Toronto: Pearson.
Lloyd, C. D. 2011. Local Models for Spatial Analysis (second edition). Boca
Raton, FL: CRC Press.
Malpas, J. E. 1999. Place and Experience: A Philosophical Topography.
Cambridge, MA: MIT Press.
Openshaw, S. 1997. The Truth about Ground Truth. Transactions in GIS 2:
2–24.
Ordnance Survey. 2014. The National Grid FAQs. https://www.ordnancesur-
vey.co.uk/support/the-national-grid.html. Accessed 19/2/2016.
Pickles, J. (ed.). 1995. Ground Truth: The Social Implications of Geographic
Information Systems. New York: Guilford Press.
Pleiades. n.d. Pleiades. https://pleiades.stoa.org. Accessed 11/7/2017.
60

Reibel, M. 2007. Geographic Information Systems and Spatial Data
Processing in Demography: A Review. Population Research Policy Review 26:
601–618.
Rupp, C. J., P. Rayson, I. N. Gregory, A. Hardie, A. Joulain, & D. Hartmann.
2014. Dealing with Heterogeneous Big Data When Geoparsing Historical
Corpora. Proceedings of the 2014 IEEE Conference on Big Data, pp. 80–83.
Rushton, G. 2003. Public Health, GIS, and Spatial Analytic Tools. Annual
Review of Public Health 24: 43–56.
Sack, R. 1997. Homo Geographics. Baltimore: Johns Hopkins University Press.
Schuurman, N. 2006. Formalization Matters: Critical GIS and Ontology
Research. Annals of the Association of American Geographers 96 (4): 726–739.
Siebert, L. J. 2000. Using GIS to Document, Visualize, and Interpret Tokyo’s
Spatial History. Social Science History 24 (3): 537–574.
Simon, R., L. Isaksen, E. Barker, & P. de Soto Canamares. 2016. The Pleiades
Gazetteer and the Pelagios Project. In M. L. Berman, R. Mostern, & H.
Southall (eds.). 2016. Placing Names: Enriching and Integrating Gazetteers.
Indiana University Press: Bloomington, pp. 97–109.
Southall, H. R., R. Mostern, & M. L. Berman. 2011. On Historical
Gazetteers. International Journal of Humanities and Arts Computing 5:
127–145.
Tanner, S., T. Munoz, & P. H. Ros. 2009. Measuring Mass Text Digitization
Quality and Usefulness: Lessons Learned from Assessing the OCR Accuracy
of the British Library’s 19th Century Online Newspaper Archive. DLib
Magazine 15 (7/8).
4
Conducting GTA Using Poverty
as a Search Term
Chapter 3 introduced geographical information systems (GIS) as a

special form of database designed for use with spatially referenced data.
It also explored geoparsing, the process by which the geographical com-
ponents of a text can be incorporated into the data models required by
GIS. This involves identifying the place-names in the text and allocat-
ing them to coordinates. In this way an unstructured text is converted
into a GIS point layer with attribute data in tabular form which enables
textual sources to be analysed within a GIS environment. To analyse
a geoparsed text, we can combine techniques from different fields to
determine where the text is talking about—using methods from spatial
analysis—and what it is saying about these places—using techniques
from corpus linguistics. The method resulting from the combination
of approaches is termed Geographical text analysis (GTA) (Gregory and
Donaldson 2016; Murrieta-Flores et al. 2015; Porter et al. 2015).
The aim of this chapter is primarily to describe GTA as a set of tech-
niques. It demonstrates how GTA works and provides an initial explora-
tion of the ways that the media represent poverty and place. Section 4.1
discusses how corpora suitable for analysing media discourses of pov-
erty were created. The rest of the chapter is concerned with an exemplar

https://doi.org/10.1007/978-3-319-93503-4_4
62
GTA using the single corpus query: <*poverty*>.1 This query was
designed to return all hits of poverty as well as terms such as anti-pov-
erty, poverty-stricken, etc. Using <*poverty*> as an example, a variety of
approaches to GTA are explored. First, density smoothing is used to
identify the major geographical patterns associated with poverty within
the text. Spatial segregation analysis is then used to explore the dif
ferences in the geographies of the two newspapers. These techniques
explore the basic geographies that the texts associate with the search-
term. We then move to using variations on corpus linguistics techniques
to identify the discourses that different newspapers associate with differ-
ent places. The patterns found for <*poverty*> are then compared with
other data, namely quantitative data on poverty and the background
geographies from the two newspapers, to identify possible reasons for
the patterns found. The findings of this chapter are preliminary because
it attempts to represent a complex, multi-faceted issue such as poverty
using a single search-term. This simplicity of approach also leads to
problems of small numbers once we start exploring instances in different
places. Nevertheless, the chapter provides some initial findings that we
return to in later chapters.
4.1 Constructing the Corpus

Our analysis of the representation of poverty and place is restricted to
UK national newspapers due to their strong pedigree as news sources
with large audiences. We chose to focus on two contrasting news
sources—the Guardian and the Daily Mail—over a time span of five
years (2010–2015). This period coincides with a change of government
in the UK, with the Coalition government of Conservative and Liberal
Democrat MPs coming to power to replace Labour in 2010. In 2015
there was further change when the Conservative party won an overall
1The corpus queries we used are given in angled brackets. The asterisks are known as wildcards
in corpus query syntax and are used to denote that the query term can be prefixed/followed by
zero or more characters. So the query <book*>, for example, would return hits for book, books,
bookend, booked, etc.
4 Conducting GTA Using Poverty as a Search Term
63
majority. Throughout our period Conservative MP David Cameron was

Prime Minister and his governments introduced fiscal policies within an
overarching discourse of austerity. Policies, such as Welfare Reform—an
overhaul of the UK benefits system—have impacted upon the lives of
the poor. Other changes to the benefits system, including the introduc-
tion of a benefits cap and the removal of the Spare Room Subsidy (also
known as the bedroom tax), decreased the amount of money that peo-
ple can receive.
Our first source, the Daily Mail is a right-leaning tabloid news-
paper which has consistently endorsed the Conservative party and its
austerity-based policies. Specifically, we have sourced articles from the
newspapers’ online entity Mail Online. The Daily Mail ’s circulation fig-
ures were 1,657,867 in May 2015 (Turvill 2015), with its website traf-
fic measured at 14,383,578 average daily browsers in February 2016
(Jackson 2016), making it the most popular newspaper in Britain. Our
second source, the Guardian, is a left-leaning broadsheet that respec-
tively endorsed the Liberal Democrat and Labour parties in the 2010
and 2015 UK general elections. The newspaper tends to stand in oppo-
sition to government cuts to welfare budgets and related services. Its
circulation figures for May 2015 were 178,758 copies per day (Turvill
2015) although this figure does not include website traffic,2 which was
measured at 8,872,392 average daily visitors in February 2016 by the
Audit Bureau of Circulation (Jackson 2016), making it the second
most-visited daily UK newspaper.
To include as much media coverage of poverty as possible, our corpora
include all texts that fit the following criteria: (i) texts appearing in the
online versions of the Daily Mail/Mail on Sunday OR the Guardian/the
Observer, (ii) texts published between January 2010 and July 2015 (inclu-
sive), and (iii) texts categorised by the two sources as ‘news’ or ‘comment’
(articles categorised as ‘sport’, ‘travel’, etc. were not included). As there
were no restrictions on corpus size, and to ensure maximum potential
coverage of poverty, we have chosen to use whole texts. The texts used
represent a broad genre of online newspaper articles published under the
2The Guardian does not make a branding distinction between its print and online content.
64
masthead of established news institutions. It is worth noting that we took

the newspapers’ own classification of articles without question, and thus
there could be some slight discrepancy between sources. If an article was
categorised as an example of a ‘debate’ text by the Daily Mail then we
accepted this. Whether on close inspection different readers would agree
with such a classification is beyond the scope of the present study. The
two newspapers differ in their classification systems; the Guardian texts
had more subdivisions (news: politics, news: society, news: news, and
commentisfree) than the Daily Mail texts (news, debate). Whilst such dif-
ferences are just one way in which the corpora differ, here the Guardian
news categories have been conflated and commentisfree has been taken
as comparable to the Daily Mail ’s debate section. As such, the corpora are
taken to be comparable insofar as they represent the same text type, from
UK national publications, across the same time period.
The Daily Mail corpus contains 342,154 articles in comparison with
the Guardian ’s 64,976 texts, meaning that the Daily Mail produces an
average of 168.47 online texts per day in the tested categories compared
with the Guardian ’s 31.9. Whilst this discrepancy may seem initially
surprising due to the comparable size of the two newspaper institutions,
the Daily Mail articles include the recycling of material (where a phrase/
sentence from one article is repurposed for a slightly different article)
and we have controlled for this in our close analysis. The total word
count for each corpus (324,692,331 and 56,403,306 respectively) indi-
cates that articles from the Daily Mail also tend to be longer (with aver-
ages of 948.97 and 868.06 words per article). However, the Daily Mail
website includes content that is produced in partnership with other
newspapers, particularly the Chinese news source the People’s Daily.
Removing this type of article from the corpus (as it is not comparable
with the content of the Guardian corpus), as well as removing articles
relating to the arts (as this is a separate section in the Guardian which
was not included in the corpus) leads to the following figures for the
Daily Mail: 341,431 texts, 324,239,704 words (averages 168.11 per day
and 949.65 words long) as shown in Table 4.1. Whilst we cannot claim
that the corpora are representative of the Daily Mail and Guardian per
se (they do not include sports articles, reviews, etc.) we can claim that
both the Daily Mail and Guardian corpora are representative of news
and debate journalism in both sources.
65
Table 4.1 Breakdown of Daily Mail and Guardian corpora

Total News Comment
No. of No. of No. of No. of No. of No. of
words texts words texts words texts
Daily Mail 324,239,704 341,431 308,113,484 326,774 16,126,220 14,657
Guardian 56,403,306 64,976 36,530,430 43,508 19,872,876 21,468
The texts were mined from the two newspapers’ websites. All boil-
erplate information (surrounding images, advertisements, html links,
etc.) were removed and the articles were saved in a text file format.
The articles were grouped by month and XML tagged for their source
publication, month of publication, and article type (news or debate).
They were also marked up using two types of linguistic tagging. Firstly
parts of speech (POS) tags were assigned to each word in the texts
and then semantic tags were added using the USAS tagger (Rayson
et al. 2004). When the tagging had been completed, the texts were
uploaded into the corpus software CQPweb (Hardie 2012) to facili-
tate the first stage of GTA.3 We used the whole of each corpus to gen-
erate our corpus queries (see Chapter 5), but our GTA of poverty and
its related discourses focuses on the news sections of our corpora only.
We chose to focus on those texts which represented the institutional
voice of each newspaper, as opposed to those in the comment/debate
sections which were likely to include more individual/personal opin-
ions about poverty.
4.2 Establishing the Geographies of Poverty

in the Two Newspapers
Exploring the two newspapers’ representations of poverty and place
starts with a simple linguistic query on the use of the word ‘poverty’.
Searching for <*poverty*> returned 8004 instances in the Guardian
and 9737 instances in the Daily Mail. The Daily Mail corpus is larger
3Our thanks go to Andrew Hardie for his invaluable help with this part of the process.
66
than the Guardian (see Sect. 4.1) so these figures were normalised to
facilitate comparison. Thus, with 141.9 instances of <*poverty*> per
million words (pmw), the Guardian seems significantly more interested
in poverty than the Daily Mail which only has 30.0 instances pmw.
Concordance geoparsing (see Sect. 3.2) was then used to convert these
instances and their co-text into a format suitable for GTA.
As discussed in Chapter 3, place-names are complex and ambigu-
ous. One particular issue is the importance of scale. Where place-names
refer to towns and villages they can sensibly be represented within a GIS
using a point. Larger areas, however, are more problematic. For exam-
ple, if a city such as London is represented using a point, then how does
this point relate to place-names within the city, such as Westminster or
Islington. Similar issues exist with names of countries, counties, and
other high-level administrative units. We chose to exclude generic ref-
erences to poverty in the UK, England, Wales, Scotland, and Northern
Ireland due to their lack of geographical specificity. Below this, high
level administrative areas including county names and ‘London’ were
flagged as being ‘high-level units’ so that these could be handled dif-
ferently when required in subsequent analysis. Smaller cities and low-
er-level administrative areas including districts and parliamentary
constituencies were not flagged as it was felt that within this analysis
these can realistically be represented using a point. This choice is some-
what arbitrary and its implications will be returned to below.
The resulting geo-parsed data can be used to identify place-name
co-occurrences (PNCs) which are defined as the occurrence of a place-
name within 10 tokens either side of the search-term. PNCs are the basic
unit of analysis within GTA. They consist of the search-term, the place-
name, and the co-text that surrounds them, and can be represented using
a point representing the location referred to by the place-name. Thus (1),
below, forms a PNC between poverty and Blackpool because the two
words are within 10 word tokens of each other. More than one PNC can
be created by a single instance of the search-term. Example (2) generates
two PNCs, one for Birmingham and one for Liverpool.
1. Mike Barry, once a debt adviser with Citizens Advice, and now oper-
ations director of the town’s credit union, is dismayed—both by
67
Blackpool’s worsening poverty and by the rise of the corporate mon-

eylenders4 (G news: news, July 2013).
2.
More than 20% of children experience severe poverty in
Birmingham and Liverpool (G news: society, Feb 2011).
The number of PNCs provides an indication of how closely the search-

term is associated with place. In this case, neither of the two newspapers
associates <*poverty*> with place particularly strongly. The Guardian
contains 245 <*poverty*> PNCs, while the Daily Mail has 237. This
means that only 3.06% and 2.43% of instances of the raw tokens of
<*poverty*> found in each corpus respectively result in a PNC.
The resulting PNCs can be mapped as point data. However, pat-
terns from dot maps are difficult for the human eye to perceive. For
this reason a technique called density smoothing can be used to make
the patterns within the data more understandable. Effectively, density
smoothing converts the point pattern into a raster surface in which each
raster cell’s value measures how many points are near to that cell using a
formula by which points nearer to the cell are given a higher weighting
than those further away (Lloyd 2011).5
Density smoothed maps of <*poverty*> in the two corpora are shown
in Fig. 4.1. These maps must be interpreted with caution because place-
names flagged as high-level units have been included. The densities have
been classed using z-scores. A z-score measures how far from the mean
each value lies, a z-score of 0.0 is the mean of the dataset, 1.0 is one
standard deviation above the mean, −2.0 is two standard deviations
below the mean and so on. The higher class intervals on the map use
z-scores of 1.96 and 2.58 respectively which are the absolute values that
4Bold text indicates the search-term, underlined text is the place-name.

5Key to this is the concept of a bandwidth which measures how quickly the weighting declines
with distance. If the bandwidth is too low only areas very close to points will be identified and
the density smoothing will not improve on the point pattern, if it is too high, the variations
across the study area will be smoothed out. Bandwidths can be justified using measures based on
the standard distance of the dataset, a measure similar to the standard deviation. In this case we
1/4
2
use the formula proposed by Fotheringham et al. (2000: 149) hopt = 3n σ where hopt is the
optimal bandwidth, σ is the standard distance, and n is the number of points in the dataset.
68
less than 5% and 1% of the observations should exceed assuming a two-

tailed normal distribution. These thresholds therefore provide an indi-
cation of spatial clustering although they should not be taken as formal
measures of statistical significance.
Figure 4.1 shows that the Guardian ’s coverage of <*poverty*> is
strongly concentrated on London, the only part of the country that has
z-scores above 2.58. Perhaps more surprisingly, the second most con-
centrated cluster is centred on York which is the only other place with
values above the 1.96 threshold. Working from north to south, there
are weaker clusters centred on Glasgow, the north-east of England,
Liverpool, and the M62 corridor,6 Birmingham, the south-east outside
of London, and Torbay in Devon (see Fig. 4.1 in the book introduc-
tion for the locations of these places). The association of major urban
centres with poverty is not surprising, but the major cluster at York
and the lesser cluster at Torbay are somewhat unexpected. In the den-
sity smoothed map of the Daily Mail ’s PNCs, the pattern appears even
more London-centric than the Guardian ’s as nowhere beyond London
crosses the 1.96 threshold. 48.9% of the Daily Mail ’s PNCs lie within
the London cluster compared to 37.1% in the Guardian. Beyond
London, there are no obvious clusters in York or Torbay. Much of
the remainder of the Daily Mail map looks similar to the Guardian ’s.
There are clusters in Glasgow, north-east England, the M62 area
(although this is more concentrated on Manchester and spreads less
to the east than in the Guardian ), Birmingham, and the south-east
beyond London. Two clusters that appear in the Daily Mail but not the
Guardian are Belfast and Margate, in Kent.
Comparing map patterns by eye is difficult and highly subjective so,
instead, spatial analysis can be used to conduct a more formal test as
to whether and where two patterns differ. This involves using a tech-
nique called spatial segregation analysis (Rowlingson 2015; Diggle et al.
2005)7 that compares two datasets to identify places that have signif-
icantly more points from one dataset than the other. The results of
using spatial segregation analysis to compare <*poverty*> from the two
6The M62 motorway runs east-west across the north of England from Hull to Liverpool.
7We are very grateful to Barry Rowlingson for this assistance with this.
Fig. 4.1 Density smoothed maps of <*poverty*> PNCs: a Guardian b Daily Mail
69
70
Fig. 4.2 Statistical comparison of the PNCs from the Daily Mail and the Guardian
newspapers are shown in Fig. 4.2. A global test8 on this pattern gives a
value of p = 0.01 suggesting that the two patterns do vary significantly
from each other by location. The local statistics, shown in Fig. 4.2,
8Spatial analysts distinguish between global and local statistics. A global statistic, usually
expressed as a single summary statistic such as a p-value or r2, gives the average pattern or rela-
tionship across the study area. Local statistics, by contrast, express the relationship at multiple
71
confirm that the Daily Mail ’s coverage of poverty concentrates more

heavily on London than the Guardian ’s. This is particularly true to the
east of London. The Daily Mail is also prevalent in Northern Ireland
because of the cluster in Belfast, a place that has no Guardian PNCs
for <*poverty*>. The Guardian associates poverty far more with parts
of eastern England, particularly east Yorkshire, than the Daily Mail.
This is only partly caused by the York cluster; the Guardian also seems
to concentrate more than the Daily Mail on other parts of Yorkshire,
as well as further north to Middlesbrough and south into Lincolnshire.
The Guardian also concentrates on Devon more than the Daily Mail, an
emphasis that can be explained by its Torbay cluster. Having established
where each newspaper talks about poverty, the next step is to interrogate
what each newspaper is saying about poverty in these locations.
4.3 Geographical Variations in Discourses

of Poverty
The basic finding from our spatial analysis is that, when the newspapers
mention poverty within the scope of place, they both strongly associ-
ate <*poverty*> with London. Nearly a half of the Daily Mail PNCs
and over a third of the Guardian ’s refer to London or places within it.
Beyond London, both newspapers largely concentrate on the major
urban centres. Spatial analysis, and the maps that it creates, are excel-
lent tools for identifying and describing spatial patterns. However, they
are very limited in their ability to explain why these patterns exist. To
understand the patterns and the differences between them, we must
move away from mapping and spatial analysis, and back into the cor-
pora from which the patterns emerged.
The easiest point of interest to address using the corpora is to ques-
tion what it is about the unexpected places, such as York and Torbay
locations across the study area and, as a consequence, are usually shown in map form. They show
what the pattern or relationship is at each location (Fotheringham 1997). As an example, a global
statistic might suggest that across the country there is a positive relationship between unemploy-
ment and crime. Local statistics, however, might show that this relationship is strong in some
areas, weaker in others, and is not present or even negative in others.
72
from the Guardian and Margate from the Daily Mail, that leads to
their association with poverty. The obvious first approach to this is to
closely read the concordances that create the PNCs. The Guardian ’s
York cluster is formed by seventeen PNCs all of which refer to York
itself and all of which are in the society section of the newspaper.
Significantly, eleven occur in a single article, which focuses on York’s
attempts to make itself a poverty-free city. There are also implications
within the concordances that, although York is a relatively wealthy
place, it contains a significant amount of hidden poverty, as shown in
(3–4).
3. York embodies a specifically middle-England kind of poverty

(G news: society, July 2014).
4. If work was guaranteed to drag you out of poverty, York should not
have a poverty problem (G news: society, July 2014).
The Guardian ’s relatively high concentration on east Yorkshire is only

partly driven by its emphasis on York, which accounts for seventeen of
its forty PNCs in this area (by contrast the Daily Mail has twelve). The
Guardian also talks about poverty in Bradford, particularly in relation
to race, and seems to have a range of stories across this region (5). The
Daily Mail ’s concentration on Belfast is largely explained by its name
appearing in several lists of areas with high child poverty or other dep-
rivation indicators, such as (6). These types of list rarely appear in the
Guardian.
5. Research published last month by the Joseph Rowntree Foundation

suggests that Bradford’s real problem—poverty—has been over-
looked (G news: society, Aug 2011).
6. CONSTITUENCIES WITH THE WORST LEVELS OF CHILD
POVERTY 47 per cent Manchester Central 43 per cent Belfast
West 43 per cent Glasgow North East 42 (DM news, Feb 2013).
7. Top of the ‘at-risk’ list are the ‘English Riviera’ towns of Torquay,
Paignton and Brixham. The groups experiencing poverty vary from
region to region. In Torbay (G news: society, June 2012).
73
The Torbay cluster in the Guardian and the Margate cluster in the Daily
Mail are both perhaps a little misleading. Torbay is created from only seven
PNCs of which four are repeated occurrences of (7). From the Daily Mail,
the Margate cluster is created from three PNCs, all of which are associ-
ated with celebrities. One refers to Tracy Emin who ‘grew up in poverty in
Margate’ while the other two refer to Mary Portas’ efforts to reform retail-
ing in the town which, it was claimed, was ignoring the town’s problems
with poverty. The fact that these clusters are created from so few PNCs is
indicative of the relatively low z-scores that they are given in Fig. 4.1.
Exploring the concordances themselves allows us to identify what is
driving patterns with relatively small numbers of instances. Analysing
larger amounts of text requires the use of corpus linguistic techniques.
Keyness is a corpus linguistics measure that allows one corpus or sub-cor-
pus to be compared with another to identify which words are found sig-
nificantly more/less frequently in one than in the other. These words are
known as keywords (Baker 2006, see also Sect. 2.1). A variation of keyness,
contrastive concordance analysis, can be used to compare two sets of
PNCs. The word frequencies within one set of PNCs is compared with
another set to discover which PNC keywords have statistically significant
log likelihood scores. These can be used to compare how a single news-
paper represents poverty in different places, or to compare the two news-
papers’ representation of poverty in particular places. An obvious question
generated by the maps is the extent to which each newspaper discusses
poverty in London differently from poverty in the rest of the country.
Table 4.2 identifies the PNC keywords for <*poverty*> in London
when compared to elsewhere in the country. Starting with keywords in
London, exploring the co-text reveals that outer is a keyword because
there is a perception that poverty is spreading from the parts of inner
London, that have traditionally been associated with poverty, into areas
in outer London. This also leads to a number of instances of now as
in they ‘now live in outer London’.9 Other than this, it is difficult to
discern any major trends about what is being said about London due
to small numbers and sometimes the multiple occurrences of phrases
9Italics indicate PNC keywords.

74
Table 4.2 PNC keywords in the Guardian comparing the co-text of <*poverty*>

in London with the rest of the UKa
Place Sig. Keywords
London <.01 outer, wealth, notoriously, mixed, most, compared, now
London <.05 regions, only, because, alongside, number, than, disparity,
rate, wide, reveal, where
Rest of UK <.01 region, hidden
Rest of UK <.05 life, constituency, last, five, says, happy, she, crime
aPNC keywords are found within 10 words of the search-term, are statistically
significant using log-likelihood at the level shown by ‘Sig.’, have a minimum fre-
quency of 3 and are presented in reverse order of log-likelihood score. Place-
names have been excluded.
including lists of place-names. An example of this is that notori-

ously, mixed, wealth, and most all come from (8), which gives 5 PNCs:
Kensington and Chelsea, London, Enfield, Brent, and Waltham Forest,
which account for almost all of the observed occurrences of these PNC
keywords.
8. Other good performers among poorer parts of the country are the
London boroughs of Kensington and Chelsea (where poverty and
wealth are most notoriously mixed ), Enfield, Brent, and Waltham
Forest (G news: society, June 2013).
The PNC keywords for outside London show some slightly more inter-
esting patterns. Hidden comes in part from the discourse about York
which frequently refers to ‘hidden poverty’. There are also a number
of instances of hidden poverty being found in counties, mainly in the
south-east of England. These continue to develop what was found in
York, namely that the Guardian ’s take on poverty outside London is
associated with hidden poverty in relatively wealthy places. Constituency
follows this theme, the most commonly cited being David Cameron’s
Witney constituency, which is again used to highlight poverty in rel-
atively affluent, or leafy places. The PNC keyword crime makes direct
links between crime and poverty, and she tends to refer either to indi-
vidual women who have risen from poverty in a particular place, or who
are talking about it in relation to places outside London. The findings
for region, which is distorted by the phrase ‘region to region’, happy
75
Table 4.3 PNC keywords in the Daily Mail comparing the co-text of <*pov-
erty*> in London with the rest of the UK
Place Sig. Keywords
London <.01 two, one, rate, reached, country, protest, he, Victorian,
years
London <.05 slum, caf, says, reason, cereals, imported, walked, Keely,
poorest, improved, great, compared, depression,
Facebook, home, become, during, high, rose, class,
byword, nostalgia, despite, situation, capital, broadly,
miles, men, walk, show
Rest of UK <.01 too, risks, lowest, people, left, care, we
Rest of UK <.05 level, making, road, citing, wherever, small, fuel, herself,
find, families, local, her, poorer, five, here, taking, miss,
ignoring
which comes from a single quote that occurs in multiple PNCs, five
which refers to one in five people in an area living in poverty, and says,
which is used to preface a quote, are less illuminating.
Trying to sub-divide the poverty PNCs into more detailed classes
than simply London and the rest of the UK rapidly runs into small
number problems. One interesting finding is that when the non-Lon-
don urban clusters of Liverpool, the M62, Yorkshire (excluding York),
Glasgow and Birmingham are compared with the rest, life is significant
at the p < 0.01 level and fuel at the p < 0.05 level. Life either discusses
the link between poverty and low life expectancy or people being con-
demned to a life in poverty, and 6 of the 7 instances of this are found
in these urban areas, while fuel refers to fuel poverty. This suggests that
the issues of low life expectancy, life-long poverty, and fuel poverty are
primarily associated with major urban areas in contrast to the hidden
poverty in more rural areas.
Table 4.3 shows the Daily Mail ’s PNC keywords comparing London
with the rest of the UK. Analysing the Daily Mail ’s concordance lines
shows that that the newspaper often repeats sentences or paragraphs sev-
eral times in different articles. An example of this is a story repeated in
several texts which contrasts the poverty in the borough with an expen-
sive café in Tower Hamlets that sells breakfast cereal for £2.50 or £3.20
a bowl (depending on the article). Within these stories, example (9)
occurs 6 times and (10) occurs 4 times.
76
9. Tower Hamlets has the highest rate of child poverty in the coun-
try—it reached 49 per cent this year (DM news, Dec 2012).
10. In Tower Hamlets, where one in two children live in poverty (DM
news, Dec 2014).
These two examples10 do much to explain the high instances of two,

one, rate, reached, and country. This story is also responsible for a num-
ber of PNC keywords that are significant at the p < 0.05 level namely
caf, says, reason, cereals, imported, Keely, Facebook, and nostalgia.
A number of more reliable themes also emerge from the Daily Mail ’s
keywords about poverty in London: London is associated with poverty
because of men who were born into poverty in London and rose out
of it. This explains the occurrences of he, become, and rose. Historical
stories about poverty are also frequently related to the city. Again, the
phenomenon of repeated phrasing is apparent, with 6 references to the
Jarrow marches which, although used in different contexts, are realised
similarly to (11) and lead to protest, men, walk, and walked appearing as
PNC keywords.
11. when 200 men walked to London to protest about unemployment

and poverty (DM news, July 2010).
History also emerges in other ways. Victorian and slum occur because of
discussion of London slums, particularly those in the East End in the
Victorian era, and the phrase ‘during the Great Depression ’ puts all of
these words (except ‘the’) onto the list of PNC keywords.
The Daily Mail includes some positive stories about modern pov-
erty in London. For example, there are reports about pupils who have
achieved good educational results despite living in areas with high lev-
els of poverty, which account for many instances of high and despite.
The Daily Mail also contrasts poverty in London, particularly East End
10These two examples also demonstrate the limitations of using PNC keywords for non-lemma-
tised data. Clearly (9) and (10) both refer to child poverty, but one refers to ‘children’ and the
other ‘child’. Combined it is likely that the lemma child would be a PNC keyword, but sepa-
rately they do not meet significance thresholds.
77
places such as Limehouse and Poplar, with affluence elsewhere. This

leads to compared being a PNC keyword. The Daily Mail also compares
poverty in London with the affluence of David Cameron’s Witney con-
stituency. This contrasts with the Guardian ’s perspective, where poverty
in Witney is seen as a problem, albeit a hidden one. Improved, situation,
and broadly largely come from one PNC, as in (12).
12. 42 per cent of children were below the poverty line in the con-
stituency of Bethnal Green and Bow while 41 per cent were liv-
ing in poverty in Poplar and Limehouse. The situation has broadly
improved since 2011 (DM news, Feb 2013).
Again, this story is putting a more positive slant on the situation. By

contrast, capital occurs in part because of London is described as the
‘poverty capital of Europe’. Finally, home is always used in relation to an
individual’s home rather than in relation to homelessness, while class is
usually used in relation to ‘the top of the class’ rather than social class.
Outside of London there are again issues of repeated text. Example
(13) leads to four PNCs which, between them, account for most instances
of care, we, wherever, find, and citing. Only one instance of care refers to
children being taken into care. Similarly, a list of Local Authorities occur-
ring with ‘Lowest Level of Child Poverty’ gives nine PNCs which accounts
for most instances of too, risks, lowest, level, and local.
13. Mr. Miliband said people care about inequality ‘wherever we find it’,
citing child poverty in Glasgow and Liverpool, unemployment in
Motherwell and Newcastle and (DM news, Sept 2014).
Three of the PNC keywords, herself, her, and Miss, refer typically to sto-
ries about women who were born into poverty but raised themselves out
of it, or who had come into a large sum of money. These PNC keywords
are indicative of a trend in the Daily Mail to focus on individual sto-
ries (see Chapter 6). There are also reports about women who either left
someone in poverty while spending money on themselves, or conversely,
where someone else had left a woman in poverty. Another story associ-
ated with women refers specifically to Mary Portas, who is represented
78
as ‘ignoring poverty in Margate and making it too trendy’, and contrib-

utes to the PNC keyness of making and ignoring. People is similar to the
feminine-marked terms discussed above as it typically refers to either
people who have risen from poverty or been left in it. In cases where
people have risen from poverty the article usually implies agency, while
in cases where people have been left in poverty it implies powerlessness.
In contrast to the occurrences of positive slants on poverty we see
associated with London, there does not appear to be a positive slant
on stories located elsewhere. For example, fuel and families draw atten-
tion to fuel poverty and families in poverty outside London. Small
and poorer have only three instances each and thus wider trends are
not identifiable. As with the Guardian, attempting to compare urban
areas away from London with the remaining places quickly runs into
small number problems. The only PNC keywords that are significant
at p < 0.01, have a frequency of five or over, and are not parts of place-
names or numbers, are we and have. Comparing London as depicted in
the Guardian and the Daily Mail, the latter seems far more interested
in child poverty in London than the Guardian. The Daily Mail has 37
PNCs with the word ‘child’ in their co-text compared to only 11 in the
Guardian. This is at least in part caused by the Daily Mail listing places
with high rates of child poverty, and areas of London feature heavily in
these lists. Combat for the Daily Mail also comes from this, as it is asso-
ciated with ways to combat child poverty (although this is largely based
on multiple PNCs from one instance of combat ).
In general, there are two main themes that emerge from the two
newspapers’ discussions of poverty in different places: (i) those asso-
ciated with aggregate poverty where poverty is treated as an issue
for society, and (ii) poverty associated with individual people and,
often, their rise from it (which is frequently linked to gender as well
as place). No strong theme emerges from the Guardian ’s discussions
about aggregate poverty in London other than poverty’s spread from
inner to outer London. The Daily Mail, by contrast, has much more
to say. It has a somewhat mixed message that, on the one hand, points
to improvements in poverty in London, while also describing London
as the ‘poverty capital of Europe’ and comparing poverty in inner-city
London with the wealth in other parts of the country, especially David
79
Cameron’s Witney constituency. Outside London, both newspapers

have fairly similar themes on aggregate poverty. Both talk about fuel
poverty, but the Guardian talks about the links between poverty, crime,
and low life expectancy, while the Daily Mail talks about families in
poverty in a more personalised way. The one theme not found in the
Daily Mail is the Guardian ’s discussion of hidden poverty in typically
more wealthy areas. It is worth stressing, however, that this analysis is
limited by relatively small numbers of fewer than 250 PNCs in both
cases, especially as some PNCs are caused by having multiple place-
names near to the search-term or the same story being reproduced sev-
eral times. Nevertheless, this analysis does much to shed light on the
different discussions of and perspectives on poverty that create the pat-
terns identified by the spatial analysis.
4.4 Comparing Representations of Poverty

with Quantitative Measures
So far we have shown that both newspapers’ coverage of poverty is heav-
ily skewed towards London—the Daily Mail more than the Guardian—
and then tends to concentrate on other major cities, although there
are exceptions to this such as York, Torbay, and Margate. This poses
the question of to what extent this is justified. In other words, are the
newspapers actively concentrating on areas that have high levels of dep-
rivation? As described in Chapter 1, deprivation can be measured quan-
titatively using indicators such as Carstairs scores which, in this case,
can be derived from 2011 census data by summing the z-scores from
four variables: overcrowded housing, low-skilled work, unemployment,
and lack of a car. Scores of 0.0 indicate average levels of deprivation,
higher values indicate higher levels of deprivation, with values above
4.0 indicating that on average each of the four measures is one standard
deviation above the mean (Morgan and Baker 2006).
Carstairs scores for local authority districts in the UK in 2011 are
shown in Fig. 4.3. Conceptually, comparing these to PNCs for <*pov-
erty*> is simple as both datasets are available in geo-referenced form—
the PNCs are geo-referenced using points, while the Carstairs scores are
80
Fig. 4.3 Carstairs scores for the UK from the 2011 census. Polygons show local
authority districts
for local authority districts. If all of the PNCs referred to local author-
ity districts or sub-divisions of them then comparing the two would
be simple. However, the place-names in the PNCs refer to a range of
features including settlements such as cities, towns and villages, and a
range of administrative units including London boroughs, districts,
parliamentary constituencies, and counties. One solution to this could
be to ignore higher-level units which, as described above, are already
flagged as such, and compare the remainder with Carstairs scores at
81
local authority district level. An alternative is to aggregate the Carstairs

scores to county level and compare these with all PNCs. This is more
reliable but the Carstairs scores will be averaged out across large areas
with potentially highly variable rates of poverty, particularly in London
which includes some of the wealthiest and most deprived parts of the
UK. This is part of what geographers refer to as the modifiable areal
unit problem (Openshaw 1984; Wong 2009) where relationships
between two datasets will vary according to both the scale at which the
relationship is measured, and on the arrangement of the aggregated
units. It is considered good practice to analyse data at the lowest possi-
ble level of aggregation to avoid these issues, and/or to repeat the analy-
sis at more than one scale.
Using Carstairs scores at local authority district level, and thus ignor-
ing PNCs classed as referring to high-level units, Table 4.4 compares the
number of districts with different levels of deprivation with the num-
ber of PNCs referring to them in the two newspapers. In general, this
shows that the two newspapers do concentrate their coverage of poverty
on areas that have higher levels of deprivation. The Daily Mail seems
to do this somewhat more than the Guardian: 86% of the Daily Mail ’s
PNCs are in districts with above average deprivation (a Carstairs score
of above 0.0) compared to 75% of the Guardian ’s, with similar differ-
ences at higher levels of deprivation.
Table 4.5 explores the extremes of this in more detail by looking at
the ten districts with the most PNCs for each corpus. Five places—
Tower Hamlets, Liverpool, Glasgow, Islington and Middlesbrough—are
on both lists and also have high levels of deprivation. Beyond this, the
Daily Mail seems to concentrate more than the Guardian on areas with
Table 4.4 The occurrences of PNCs from the two newspapers in local authority
districts with differing levels of deprivation
Carstairs scores Guardian Daily Mail
Count % Count %
Above 0.0 133 76.4 145 86.3
Above 2.0 105 60.3 128 76.2
Above 4.0 75 43.1 97 57.7
All 174 100.0 168 100.0
Table 4.5 Local authority districts with the most PNCs from the Guardian and the Daily Mail
82
Top 10 Guardian Districts Top 10 Daily Mail Districts Top 10 Carstairs Districts
District G DM Carstairs District G DM Carstairs District Carstairs G DM
PNCs PNCs PNCs PNCs PNCs PNCs
York 17 0 −0.77 Tower Hamlets 14 35 6.53 Newham 7.50 0 3
Tower Hamlets 14 35 6.53 Manchester 3 13 4.20 Hackney 7.27 2 6
Liverpool 12 10 5.07 Liverpool 12 10 5.07 Kingston upon 6.73 2 0
Hull
Glasgow City 9 6 4.71 Glasgow City 9 7 3.20 Tower Hamlets 6.53 14 35
Torbay 6 0 1.03 Belfast 0 6 4.71 Barking & 6.15 2 0
Dagenham
Birmingham 6 1 4.90 Hackney 2 6 7.27 Sandwell 5.90 0 2

Lambeth 5 3 5.13 City of London, 0 5 3.40 Southwark 5.85 1 3
Westminster
Bradford 5 2 3.17 Middlesbrough 4 4 5.73 Middlesbrough 5.73 4 4
Islington 4 4 4.88 Islington 4 4 4.88 Wolverhampton 5.73 0 0
Kensington & 4 1 2.08 Lambeth 5 3 5.13 Leicester 5.72 0 1
Chelsea
West 4 0 −3.25 Thanet 2 3 2.47
Oxfordshire
Middlesbrough 4 4 5.73 Southwark 1 3 5.85
Newham 0 3 6.53
Stockton-On- 0 3 4.21
Tees
83
high levels of deprivation. The lowest Carstairs score in the Daily Mail ’s
list is Thanet in north Kent with a Carstairs score of 2.47, while the
rest are all over 3.0. By contrast, the Guardian has two places—York
and West Oxfordshire—that have below average deprivation and a fur-
ther two—Torbay and Kensington and Chelsea—which have Carstairs
scores that are lower than Thanet. The Guardian thus appears to have an
interest in poverty in some places that are relatively affluent. This is not
present in the Daily Mail.
Looking at the areas with the highest Carstairs scores (Table 4.5) the
Daily Mail seems to concentrate on high deprivation areas more than
the Guardian, with twice as many of its PNCs referring to places in
these areas. In both cases, Tower Hamlets dominates the list with over
half of all of the PNCs for these ten districts, although the cereal café
story contributes significantly to the Daily Mail’s interest in this area.
There are three other east London districts with very high levels of dep-
rivation: Newham, Hackney, and Barking and Dagenham. In com-
parison to Tower Hamlets these districts attract very little attention
with only 6 and 9 PNCs in the Guardian and the Daily Mail respec-
tively. Away from London, it is noticeable that the districts with high
levels of deprivation are not big cities but are instead Kingston upon
Hull, Sandwell and Wolverhampton (both in the Black Country),
Middlesbrough, and Leicester. This may in part be a result of the mod-
ifiable areal unit problem. Deprived areas in major cities may be aggre-
gated with wealthier areas thus averaging their Carstairs scores, while
smaller places are more homogeneous. Nevertheless, with the excep-
tion of Middlesbrough, poverty in these places is largely ignored by
both newspapers. If the list is broadened to cover all thirty districts with
Carstairs scores over 4.0 we find that, within our scope of ten words
either side, the Guardian never mentions poverty in relation to 9 of
these while the Daily Mail does not mention it in 12. Neither newspa-
per associates <*poverty*> with Blaenau Gwent (south Wales), Knowsley
(Liverpool), Greenwich (London), West Dunbartonshire (outside
Glasgow), Wolverhampton or Walsall (both in the Black Country).
Table 4.6 uses Spearman’s rank correlation coefficients to compare
deprivation and numbers of PNCs for local authority districts. Overall,
the Guardian ’s PNCs correlate slightly more strongly with Carstairs
84
Table 4.6 Spearman’s rank correlation coefficients comparing district-level

Carstairs scores & PNCs
N Guardian Daily Mail
All 404 .316** .231**
Carstairs <= 0.0 217 −.016 −0.14*
Carstairs > 0.0 187 .321** .333**
Carstairs > 2.0 88 .362** .294**
Carstairs > 4.0 30 .155 .288
Daily Mail instances > 0 66 .594** .568**
Guardian instances > 0 73 .296* .522**
Daily Mail and Guardian instances > 0 27 .398* .532**
**significant at p < 0.01, *significant at p < 0.05
scores than the Daily Mail ’s but both are statistically significant. Perhaps
unsurprisingly given what we have already seen, the significance disap-
pears if only districts with very high levels of deprivation (above 4.0), or
below average deprivation, are considered. Excluding districts for which
the Daily Mail has no PNCs raises its correlations noticeably; however,
excluding areas for which the Guardian has no PNCs reduces its coeffi-
cients. This is because the Daily Mail concentrates its attention on areas
with relatively high deprivation while the Guardian ’s PNCs are more
dispersed across the range of poverty values.
In summary, these analyses seem to show that the two newspapers
generally do concentrate their coverage of <*poverty*> on areas that can
be shown quantitatively to have high levels of deprivation. The Daily
Mail does this more than the Guardian, which also draws attention to
poverty in some more affluent places. Within this, however, there are
some clear exceptions, particularly with districts with high deprivation
being largely ignored by either or both newspapers.
4.5 Comparing Representations of Poverty

with the Newspapers’ Background
Geography
Rather than focusing on deprived areas, an alternative interpretation
of the spatial analysis of <*poverty*> is that newspapers simply talk
about some places more than others. If newspapers show geographical
85
favouritism in the places that they represent overall—either as a result

of conscious editorial decisions or unconscious biases—then the geog-
raphies they associate with poverty, or indeed any other theme, will fol-
low this more general use of place-names. We therefore need to be able
to identify whether our results for <*poverty*> are more than just an
indication of how newspapers refer to place more generally. In many
ways this question is similar to one that has an established history in
spatial analysis, namely do certain types of events tend to cluster against
a geographically variable background population. A classic example of
this is whether diseases such as childhood cancers cluster in particular
places, the challenge being that any map of these diseases is likely largely
to represent the background population distribution and it will thus be
difficult to identify clusters of disease cases from background popula-
tion clusters in towns and cities. A range of statistical approaches have
been developed to identify such clusters (see, for example, Openshaw
et al. 1987; Besag and Newell 1991; Fotheringham and Zhan 1996;
and Kulldorf 1997). In our case we are using Kulldorf ’s (1997) spatial
scan statistic, a technique that considers every point location for which
a background population is recorded. It tests whether there are statisti-
cally significantly more or fewer events near to each location under the
null hypothesis that events are distributed at random among the back-
ground population. ‘Near’ is defined using circles of increasing radii
around the points that make up the background population. The most
statistically significant locations are recorded.
Rather than diseases in a population, in GTA the question becomes
whether the distribution of PNCs simply reflects the background geog-
raphy of place-names in the corpus as a whole or whether there are hot-
spots—places where more instances of PNCs are found than expected
from the background—or coldspots—where there are fewer than
expected. The ‘events’ are the locations of PNCs of the search-term, in
this case <*poverty*>. The background population is more difficult to
define. Ideally, the entire corpus would be geoparsed and the locations
of all place-names within it would provide the background geography
(Donaldson et al. 2017). Here, however, this is not practical due to the
difficulties in geoparsing corpora of this size. An alternative approach
is to test the search-term PNCs against a sample of other PNCs that
86
might be thought to provide a realistic background geography. This

is similar to testing for clusters of a disease by comparing the pattern
of disease events with patterns from diseases that are not thought to
exhibit clustering.
To generate a background population we decided to geoparse a ran-
dom sample of 5% of all the occurrences of the in our corpora. The is
the most common word in the English language and is likely to have
regular distribution across a corpus. 153,000 instances of the in the
Guardian included 9206 PNCs (located within the UK, using the same
criteria as were used for <*poverty*>), and 782,000 instances of the in
the Daily Mail gave 46,596 PNCs (normalised frequencies of 252.01
and 151.23 per million words, respectively). The normalised frequen-
cies seem to suggest that the Daily Mail tends to refer to UK place-
names much less than the Guardian does, and this should be kept in
mind throughout the analysis of poverty (and when using this method
for comparing different corpora). The locations for the PNCs, shown
in Fig. 4.4, provide what can be taken to be the background geogra-
phy of place-names within the two corpora. Kulldorf ’s scan statis-
tic can then be used to identify which places have more/fewer PNCs
than the background geography would have us expect. In this way we
are able to establish how the geography of the search-term varies from
the background geography of the corpus as a whole. This method can
be replicated for any English language corpus to provide a background
population of the place-names mentioned.
Figure 4.5 shows the results of the Kulldorf analysis. The three
major hotspots in the Guardian are east London, east Yorkshire, and
Torbay. Thus the Guardian ’s coverage of <*poverty*> in these areas
is not an artefact of general place-name distribution. The only other
two statistically significant hotspots are High Wycombe (west of
London) and Preston in north-west England. These can be disre-
garded as they are generated by 2 and 1 PNCs respectively, albeit in
areas that do not co-occur with the in the 5% sample. The coldspots
stretch from the Thames Estuary to Suffolk, and from the English bor-
der counties through southern Scotland to Fife. These are centred on
rural areas, but Newcastle-upon-Tyne is in the southern part of the
Border coldspot, and the Thames Gateway area lies in the southern
Fig. 4.4 Density smoothed maps of the PNCs. Note that for legibility individual instances have not been
87
shown as points: a Guardian b Daily Mail

88
Fig. 4.5 Kulldorf analysis of <*poverty*> using a 5% sample of the as the background popula-
tion. Points identified are those from the background population that are at risk of being in a hot
or cold spot: a Guardian b Daily Mail
89
part of the East Anglian coldspot. The absence of Belfast and other
previously identified areas is perhaps explained by the fact that the
Guardian apparently does not tend to refer to these areas. The Kulldorf
results for the Daily Mail show two major hotspots—London and
Liverpool—and a further two minor hotspots at Wokingham (west
of London) and York that are a result of only one PNC. As with the
Guardian, much of East Anglia is a coldspot, although this stretches
further north through Lincolnshire and into eastern Yorkshire. South-
west London and Surrey form a second coldspot with Birmingham and
much of the West Country forming a third. This coldspot is perhaps
similar to the Border one from the Guardian in that it is largely rural
but has a large urban centre on its fringe, in this case Birmingham.
This coldspot also contains Bristol.
These results reveal further insights into the places that the two
newspapers associate with <*poverty*>. Although Fig. 4.1 showed
that both newspapers’ coverage of <*poverty*> is very London-centric,
Fig. 4.5 suggests that this is can largely be explained by the newspaper’s
coverage of place being London-centric. The exception to this is East
London which stands out in both newspapers as being more closely
associated with poverty than this London-centric coverage would
lead us to expect. This does correspond with the highest Carstairs
scores being found in districts in East London. Elsewhere, the sur-
prising finding is that, with the exception of Liverpool in the Daily
Mail, urban centres with high levels of deprivation, such as Glasgow,
Manchester, and Birmingham, do not receive more attention than
would be expected. Neither do smaller high deprivation areas such as
Middlesbrough, Hull, Leicester, or south Wales. The Guardian associ-
ates places such as York and Torbay with poverty in ways that would
not be expected either from the newspaper’s background geography or
the places’ levels of deprivation. Rural areas, particularly in East Anglia,
tend to be under-represented by both papers. These areas typically do
not have high levels of deprivation as measured by Carstairs scores, but
equally are not particularly wealthy.
90
4.6 Summary
This chapter has shown how GTA enables us to identify and explore
the geographies that the two newspapers associate with a particular
theme. Focusing on the single search-term <*poverty*>, it has explored
a range of techniques that allow us to identify the areas that the two
newspapers associate with this search-term, compare these two geogra-
phies to see whether and where they are different, and explore the lan-
guage in PNCs. It also compared data extracted from texts to statistical
sources. Both newspapers concentrate their coverage of <*poverty*>
on London in general and smaller areas like Tower Hamlets in par-
ticular. This is especially true for the Daily Mail. Outside London, the
other major urban centres are also associated with poverty, particularly
Liverpool and Glasgow, although both newspapers tend not to focus on
Birmingham and the West Midlands, while the Guardian also does not
consider Belfast or, to a lesser extent, Manchester. The PNC keywords
suggest a broadly similar discourse about aggregate poverty in the major
cities outside London, with issues such as fuel poverty and links with
crime and low life expectancy being raised. However, the Daily Mail
tends to talk about this in relation to individuals and families while the
Guardian does so in a more aggregate way. Within London, the Daily
Mail carries stories that emphasise improvements and relative success
stories, such as areas that have good exam results despite high levels of
poverty, but it also describes London several times as the ‘poverty capi-
tal of Europe’. The Daily Mail also draws attention to historical poverty
in London. Neither of these themes is really apparent in the Guardian.
Away from the urban centres, the Guardian focuses on ‘hidden’ poverty
in more affluent areas, including Witney, the south-east more generally,
and York. These stories are far less common in the Daily Mail.
Even this initial analysis has shown how the two newspapers are both
similar and different in their approaches to poverty and place. There are
some initial indications in the PNC co-text of the different ideological
stances of the two newspapers. Individualisation is important in the
Daily Mail, for example, which could be seen to endorse a neoliberal
attitude towards poverty and its causes. Similarly, a focus on historical
91
poverty could be used to endorse the position that poverty no longer

exists in the UK. To determine whether these initial trends are repre-
sentative of wider discourses of poverty, it is important to test more
search terms and Chapter 5 details how we set out to do this.
Comparing the pattern of PNCs with Carstairs scores and the back-
ground geographies from the newspapers shows that the two newspa-
pers broadly focus on areas with high levels of poverty when these are
in big cities (with notable exceptions, such as Birmingham, Belfast and,
to a lesser extent, Manchester). They are also patchy in their coverage of
London, particularly the East End, where Tower Hamlets gets most of
the attention and nearby districts with similarly high levels of depriva-
tion such as Hackney, Newham, and Dagenham and Redbridge, receive
far less. Outside London, areas of high deprivation away from big cities
are generally under-represented. Due to its stories about hidden poverty,
the Guardian concentrates on rural areas in the south-east in particular.
However, these areas receive less attention related to poverty than they
do overall in the newspaper suggesting that these stories perhaps reflect
an interest in these areas and the people within them. The Kulldorf
analysis reinforces this by suggesting that both newspapers concentrate
their coverage of poverty on east London in particular, over and above
what would be expected from their background geographies.
As noted above, these findings must be treated with caution because
they are based on a single search-term and rapidly run into small num-
ber problems. This is particularly true when lists of place-names cause
multiple PNCs from one instance of a search term and is exacerbated
by the Daily Mail ’s habit of reproducing exact phrasings several times
in different texts. Nevertheless, they do demonstrate the potential for
using GTA to understand the discourses around poverty and place.
Fundamentally, the preceding analysis constitutes proof of concept
of large-scale concordance geoparsing and GTA. It also reveals that
there are some clear geographies and biases in the way that this topic
is presented in the two newspapers. But discourses of poverty are not
restricted to the geographical profile of a single term. In the next chap-
ter we document how GTA can be scaled up to incorporate multiple
search terms centred on a single theme.
92
References
Besag J., & J. Newell. 1991. The Detection of Clusters in Rare Diseases.
Journal of the Royal Statistical Society Series A 154: 143–155.
Diggle, P., P. Zheng, & P. Durr. 2005. Nonparametric Estimation of Spatial
Segregation in a Multivariate Point Process: Bovine Tuberculosis in
Cornwall, UK. Journal of the Royal Statistical Society Series A 54: 645–658.
Donaldson, C., I. N. Gregory, & J. E. Taylor. 2017. Locating the Beautiful,
Fotheringham, A. S., C. Brusdon, & M. Charlton. 2000. Quantitative
Fotheringham, A. S., & F. Zhan. 1996. A Comparison of Three Exploratory
Methods for Cluster Detection in Spatial Point Patterns. Geographical
Analysis 28: 200–218.
Cartographies of Lake District literature. In D. Cooper, C. Donaldson, &
P. Murrieta-Flores (eds.). Literary Mapping in the Digital Age. Abingdon:
Hardie, A. 2012. CQPweb—Combining Power, Flexibility and Usability
in a Corpus Analysis Tool. International Journal of Corpus Linguistics 17:
380–409.
Jackson, J. 2016. Independent, Mirror, Express and Star Suffer Sharp Fall in
Traffic. Guardian 17/3/2016. https://www.theguardian.com/media/2016/
mar/17/independent-mirror-express-and-star-suffer-sharp-fall-in-traffic.
Accessed 12/5/2017.
Kulldorf, M. 1997. A Spatial Scan Statistic. Communications in Statistics:
Theory and Methods 26: 1481–1496.
Murrieta-Flores, P., A. Baron, I. N. Gregory, A. Hardie, & P. Rayson. 2015.
93

GIS 19: 296–320.
Openshaw, S. 1984. The Modifiable Areal Unit Problem. Concepts and
Techniques in Modern Geography 38. Norwich: Geo Books.
Openshaw, S., M. Charlton, C. Wymer, & A. Craft. 1987. A Mark 1
Geographic Analysis Machine for the Automated Analysis of Point Data
Sets. International Journal of Geographical Information Systems 1: 335–358.
Rayson P., D. Archer, S. Piao, & T. McEnery. 2004. The UCREL Semantic
Analysis System. Proceedings of the Beyond Named Entity Recognition
Semantic Labelling for NLP Tasks Workshop, Lisbon, Portugal, pp. 7–12.
Rowlingson, B. 2015. Spatial Segregation Analysis. Personal Communication.
Turvill, W. 2015. National Newspaper Circulations, May 2015: Mail on
Sunday Overtakes Sun on Sunday, Times Remains Only Growing Title.
http://www.pressgazette.co.uk/national-newspaper-circulations-may-
2015-mail-sunday-overtakes-sun-sunday-times-remains-biggest-growing/.
Accessed 15/1/2016.
Wong, D. 2009. The Modifiable Areal Unit Problem. In A. S. Fotheringham
& P. A. Rogerson (eds.). The Sage Handbook of Spatial Analysis. London:
Sage, pp. 105–124.
5
How to Use GTA in Discourse Analysis
Whilst the GTA of <*poverty*> in the previous chapter provided a

way into our corpus data and demonstrated that there are patterns in
how the Guardian and the Daily Mail refer to poverty in geographical
space, the linguistic aspect of the analysis so far has been restricted to
the results of one search term. Interrogation of a single term is insuffi-
cient for illuminating wider elements of the relationship between pov-
erty and place. Whilst the PNC keywords for <*poverty*> suggested
some potential ideological links between poverty and place, the analy-
sis, so far, has remained largely at the lexical level. One of the biggest
problems we face is that it is possible to talk about poverty without ever
using that specific term. For example, a text may refer to ‘the poor in
Newcastle’, ‘deprived areas of Birmingham’ or use more evaluative lexis,
such as ‘scroungers queue outside Burnley benefits office’. All of these
(albeit fictional) examples tap into a wider discourse of poverty and
place. Thus, the analysis in the preceding chapter neatly demonstrates
how GTA can be used to analyse the geographical associations of a par-
ticular search term, but it cannot tell us anything conclusive about the
wider discourses associated with poverty, and the ideologies they index.

https://doi.org/10.1007/978-3-319-93503-4_5
96
In this chapter, we document how we selected additional search

terms (Sect. 5.1), detail the processes for using GTA across multiple
search terms and large corpora, and discuss some of the decisions that
we had to make in terms of where the boundaries surrounding dis-
courses of poverty lie (Sect. 5.2). We consider the interconnectedness
of discourses and note some of the other social issues, such as crime
rates and the accessibility of the UK housing market, which were closely
linked to discourses of poverty and place.
5.1 Selecting Search Terms

In order to establish what terms were being used to talk about and
talk around poverty in our corpora, we used collocation analysis (see
Sect. 2.1) to find terms which were suitable for geoparsing. We began
by generating a list of collocate lemmas1 for <*poverty*> in both
the Daily Mail and the Guardian corpora. This was the first step in
the process of selecting suitable search terms and generating corpus
queries. Collocates had to occur within ten words either side of the
node, have a minimum frequency of 5, and a minimum collocation
frequency of 5. Ten words either side is quite a wide span for deter-
mining collocates, but we chose this distance to parallel the span
used when calculating PNCs. Furthermore, we wanted to include
not only those terms which are close collocates of <*poverty*> but
also those words which may occur slightly further away in a text but
were pivotal to expressing ideas about poverty. However, we were
mindful that the broader the span, the higher the potential num-
ber of collocates. In order to centre our analysis on lemmas which
had a strong association with poverty, as indicated by a high collo-
cation value, we chose to focus on the top 100 collocates of <*pov-
erty*> in each corpus (Table 5.1), all of which had log-likelihood
1A lemma refers to all forms of a word: e.g. the lemma BEG includes beg, begging, beggar, begged,
begs, etc.
Table 5.1 Top 100 collocates for <*poverty*>
Daily Mail Both Guardian
abject, Africa, aid, alleviation, 2020, alleviate, and, below, benefit, 60%, 300,000, absolute, action, Alison,
American, census, cent, center, cor- charity, child, define, definition, dep- breakdown, by, campaigner, cause,
ruption, country, cycle, dependency, rivation, Duncan, eradicate, extreme, CPAG, credit, cut, disadvantage,
economic, education, figure, founda- family, food, Frank, fuel, Garnham, educational, eliminate, end, exclu-
tion, ‘fuel, global, grow, help, high, government, grinding, homeless- sion, fall, field, focus, goal, group,
labour, law, many, more, nation, pen- ness, household, hunger, in, income, halve, ifs, increasing, living, Milburn,
sion, per, percent, population, porn’, increase, inequality, injustice, into, Nicolson, out, parenting, plunge,
Potok, rural, slum, southern, stricken, in-work, level, lift, line, live, low, problem, reduction, rev, root,
struggle, than, tsar, tsar’, violence measure, median, million, mobility, smith, target, taxpayers, UK, wage,
number, of, Oxfam, pensioner, peo- worklessness
ple, poor, poverty, push, rate, reduce,
relative, rise, Rowntree, social, tackle,
trap, unemployment, wealth, wel-
fare, work
5 How to Use GTA in Discourse Analysis
97
98
values well above the threshold value of 15.13, which made them sta-
tistically significant to the level p < 0.0001.
Combining collocate lists from the Daily Mail and the Guardian gave
a total of 141 potential search terms (59 terms occurred in both lists).
Using Table 5.1, the immediate conclusion that can be drawn about the
two newspapers is that they discuss poverty using some similar language
(child, extreme, family, household, government were common to the top
100 collocates in both corpora). They also draw upon similar themes or
semantic fields. For example, 24 of the top 100 collocates in the Daily
Mail refer to some form of measurement (grinding, high, level, median,
measure ) as do 22 of the Guardian ’s collocates (absolute, below, increase,
million ). This suggests that both newspapers treat poverty as something
which is measurable (high, increase ) and quantifiable (million, percent ).
Furthermore, both collocate lists include define and definition, suggest-
ing that there is some dispute over what poverty actually is and how it is
defined.
Discounting function words and terms relating to measurement, 31
collocates were unique to the top 100 in the Daily Mail and 30 were
unique to the Guardian ’s top 100. Of course, this does not mean that
these apparently unique terms never collocate with <*poverty*> in
the opposite corpora, indeed many do, but rather the differences in
Table 5.1 are indicative of which lemmas, and thus related topics, are
given prominence in the co-text of <*poverty*> in each newspaper.
For example, three of the collocates unique to the Guardian top 100
are proper nouns (Alison, Milburn, and Nicolson ), which indicates that
the Guardian ’s treatment of poverty involves reference to key individ-
uals, such as Alison Garnham, Chief Executive of the Child Poverty
Action Group, Alan Milburn, a Labour politician who headed the Social
Mobility and Child Poverty Commission, and Rev. Paul Nicolson, a rep-
resentative of Taxpayers Against Poverty. The Daily Mail collocates also
include Garnham and both sets of collocates include Frank, in reference
to Labour MP Frank Field, Oxfam, Rowntree (as in the Joseph Rowntree
Foundation), and Duncan, as in Iain Duncan Smith, a Conservative
MP who was Secretary of State for the Department of Work and
Pensions and who is credited with the introduction of the Universal
99
Credit benefits system.2 Taking a closer look at how these individuals

are referred to, whilst most are quoted or their policies directly refer-
enced, Nicolson’s name only appears in the Guardian collocates because
he has had 40 letters published. This finding demonstrates that not all
collocates are equal when determining their significance in discourses of
poverty.
It is also by looking at these unique collocates that we get a sense of
how either newspaper connects poverty to place. Of the unique collo-
cates in the Guardian top 100 only UK makes any reference to location,
but the Daily Mail collocates include Africa, American, and southern,
which on closer inspection occurs 245 times in the phrase ‘the Southern
Poverty Law Center’.3 This institution is based in Montgomery,
Alabama and investigating its occurrence in the top 100 collocates tells
us that the Daily Mail has publish 158 texts referring to it, which, along
with global, suggests a focus on poverty outside the UK. Secondarily,
this suggests that southern is probably not a suitable search term for the
present analysis.
Indeed, not all of these 141 potential search terms were viable.
Numbers like 2020 and function words like of, in, and and occur too
many times in the corpus generally to make geoparsing them worth-
while. For example, of occurs 6,794,702 times in the Daily Mail
corpus but only 6155 (0.09%) of those are within ten words of <*pov-
erty*>, with comparable figures of 1,394,935 and 5158 (0.36%) in
the Guardian. Thus, geoparsing general items like these would lead to
too many potential PNCs that are not related to poverty, which would
skew our analysis. Similarly, those collocates which suggest some meas-
urement/scale of poverty—lemmas such as high, rate, median, etc.—are
likely to be also used to describe other measurable entities. The top five
2The top Daily Mail collocates also include reference to the Labour party, but not the
Conservative party (Conservative 1638th on the list of collocates and is not significant, and Tory
is 1265th, LL value 4.039). Whilst occurrences of labour also refer to manual labour, etc. as well
as the Labour party, the fact that one political party is related to poverty more than the others is
worthy of future investigation.
3This finding also explains why the American English spelling of center occurs in the top 100
collocates as does Potok, which is in reference to Mark Potok who works at the Southern Poverty
Law Center.
100
collocates for <high> in the Guardian (using the same collocation crite-
ria as above) are court, street, level, rate, and unemployment, only one of
which—unemployment—shows any immediate connection to poverty.
Where these measurement terms are used within discourses of poverty
and place, they are likely to occur in clusters such as ‘high poverty rate’
and would be picked up during the geoparsing of <*poverty*>.
Other collocates unsuitable for geoparsing included child and fuel,
both of which were used to premodify poverty to draw readers’ attention
to particular types of poverty or to the impacts of poverty on particular
social groups:
1. An end to fuel poverty in housing by 2016 (G news: politics, Jan

2010).
2. As we discovered on Monday this week, child poverty is now grow-
ing (DM news, Jan 2010).
The labels child poverty and fuel poverty appear relatively fixed. Of the
227 times that fuel collocates with <*poverty*> in the Guardian, 196
(85%) are occurrences of the phrase fuel poverty, and there are similar
figures of for the Daily Mail: 303 (91.3%) out of 332 collocations are
fuel poverty. For child and poverty the figures are 1151 collocations with
527 (34.9%) occurrences of child poverty in the Daily Mail and 2752
collocations with 1329 (51.7%) child poverty in the Guardian. The fact
that terms such as fuel poverty are repeated and seem fixed suggests that
most references to people who are unable to afford fuel, or references
to children in financial hardship, will be phrased in this way. As such,
geoparsing all occurrences of fuel and child is unlikely to lead to lots
of PNCs relating to poverty that do not occur in the phrases fuel/child
poverty.4
A similar pattern occurs with the term ‘poverty porn’ as it is used to
describe a particular type of television programme which present the
4A similar argument could be made for pensioner poverty. However, we chose to include the query
<pension*> as it referred not only to pensioner poverty, but also to the state pension, a type of
government benefit.
101
day-to-day lives of those in receipt of government welfare payments in

a pseudo-documentary format for mass audience entertainment (see
Paterson et al. 2017). Additionally, lemmas such as alleviate, eradicate,
and target were also used primarily as immediate collocates of poverty.
However, geoparsing these terms would not likely tell us much about
poverty and place, given that we can alleviate all sorts of issues (the top
R1 collocates for alleviate in the Daily Mail were poverty, suffering, pain,
and pressure ).5 It would be interesting to geoparse terms like these to see
what issues are being alleviated or eradicated in particular geographical
areas, but this is beyond the scope of the present analysis.
Whilst all of the collocates in Table 5.1 would be useful starting point
if we were only performing an analysis of poverty, our analysis is com-
plicated by our duel focus on poverty and place. So, for example, the
individuals referred to in the Guardian top 100 collocates tell us about
the types of people whom this particular newspaper refers to when talk-
ing about poverty. The prevalence of measurement terms and quanti-
fiers in both corpora demonstrates that the abstract concept of poverty
is treated as something concrete which can be quantified. The repeated
use of the L1 position to premodify poverty (child poverty, fuel poverty,
etc.) tells us that poverty is seen as something which can be subdivided
into discrete categories, whilst the verbs alleviate, eradicate, and tackle
indicate that the newspapers are reporting on or endorsing affirmative
action to reduce poverty. Ideologically, this presupposes that:
a. poverty is fundamentally a bad thing,

b. a decrease in poverty would be for the social good,
c. human action can decrease poverty rates, and
d. there are people in positions of power who are acting or should act
to decrease poverty.
5What this does tell us is that alleviate has negative semantic prosody (Stubbs 2001), insofar as we
only seem to alleviate things that are bad. Eradicate has a similar pattern; its top R1 collocates in
the Daily Mail include polio, child, poverty, terrorism, and ISIS. Thus we can argue that poverty is
presented as a particularly bad social problem which needs to be addressed.
102
From our list of collocates, we can begin to conceptualise a discourse

or discourses of poverty as drawn upon by the UK media, and we can
see which particular lexical items are used when such ideologies are
deployed. However, none of these conclusions automatically apply to
discourses of poverty AND place.
In order to include only those references to discourses of pov-
erty which relate to places within the UK we had to choose terms to
geoparse which we thought would be most fruitful. Thus, we eliminated
proper nouns (with the exception of Iain Duncan Smith), function
words, quantifiers and measurement terms, collocates which predomi-
nantly premodified poverty (i.e. fuel, child, etc.), words for social groups
(people, group, population ), and general terms like push and live. This left
us with 24 potential search terms which appeared to be key terms in
discourses of poverty: benefit, charity, credit, cut, deprivation, Duncan,
exclusion, food, homeless, household, hunger, income, inequality, mobility,
parenting, pension, poor, poverty, social, taxpayers, unemployment, wage,
welfare, worklessness. But even with this relatively small number, as is
shown in Sect. 5.2, some search terms were more associated with pov-
erty and place than others.
To these 24 collocates we added 11 terms relating to money and
employment, including cost, earn, expense, penniless, price, salary, spend,
and redundant. We complemented homeless with beg and, as food was
too general, we chose to narrow this term to focus on a food banks and
also included starve in relation to hunger. The term social too was too
broad (its top collocates in the Guardian are care, mobility, worker, hous-
ing, and medium ), so we combined it with another potential search
term and used the phrase social exclusion. We also switched mobility for
the more general disability. Similarly, as geoparsing parenting would
have been quite broad, we chose to focus on single-parenthood. We
also added 17 search terms relating to benefits receipt, which mostly
included the names of particular benefits, such as Employment Support
Allowance (also known as ESA), as well as related terms such as aus-
terity, claimant, dole and handout; the latter being included as it is the
Daily Mail ’s preferred term for benefits.
From this list of potential search terms we generated 64 different cor-
pus queries (including <*poverty*>), as shown in Table 5.2. We chose
103
Table 5.2 Search terms in the Daily Mail and the Guardian

Daily Mail Guardian
No. of hits Freq. pmw No. of hits Freq. pmw
1 *afford* 26,494 85.86 6379 174.62
2 *allowance* 7845 25.42 3318 90.83
3 *austerit* 6217 20.15 2709 74.16
4 {beg} 8490 27.51 391 10.7
5 *benefit* 55,297 179.21 19,611 536.84
6 *charit* 45,018 145.89 11,841 323.4
7 *claim* 301,829 978.17 26,897 736.29
8 *communit* 72,368 234.53 17,829 488.06
9 *cost* 117,875 382.01 18,336 501.94
10 credit* 28,766 93.22 5309 145.33
11 cut* 112,251 363.78 28,920 791.67
12 *depriv* 4381 14.2 1948 53.55
13 *disab* 20,457 66.3 4227 115.71
14 DLA 327 1.06 265 7.25
15 dole* 2616 8.48 261 7.14
16 Duncan Smith* 3604 11.68 1919 52.53
17 earn* 37,803 122.51 5126 140.32
18 *employ* 87,550 283.73 17,042 466.52
19 *entitl* 1699 37.91 2103 57.57
20 *equali* 8049 26.09 4392 120.23
21 ESA* 751 2.43 320 8.76
22 *expens* 34,726 112.54 5383 147.36
23 fit for work 113 0.37 182 4.98
24 (fitness to work|fit- 137 0.07 15 0.41
ness-to-work| fit
for work)
25 (food bank*|- 1424 4.61 1347 36.87
foodbank*|-
food-bank*)
26 handout* 2835 9.19 198 5.42
27 hardship* 1754 5.68 430 11.77
28 *home* 482,368 1563.26 40,010 1095.25
29 hous* 304,246 986 33,468 916.17
30 (hunge*|hungr*) 9227 29.9 947 25.92
31 income* 23,655 76.66 7680 210.24
32 IDS* 672 2.18 187 5.12
33 job* 117,232 379.93 22,583 618.2
34 JSA 118 0.38 184 5.04
35 *money* 129,601 420.01 16,662 456.11
36 *paid* 70,692 229.1 9569 261.95
37 pay* 170,672 553.11 24,752 677.57
38 penn* 27,146 87.97 1247 34.14
(continued)
104
Table 5.2 (continued)
Daily Mail Guardian
No. of hits Freq. pmw No. of hits Freq. pmw
39 pension* 42,851 138.87 9081 248.59
40 personal independ- 59 0.19 76 2.08
ence payment
41 PIP 623 2.02 196 5.37
42 poor* 36,648 118.77 9675 264.85
43 <*poverty*> 8689 28.16 5288 144.76
44 price* 82,198 266.39 7428 203.34
45 redundan* 3920 12.7 1484 40.62
46 rehous* 227 0.74 144 3.94
47 rent* 22,241 72.08 4572 125.16
48 *salar* 15,107 48.96 2140 58.58
49 *sanction* 9505 30.8 1934 52.94
50 *scroung* 608 1.97 311 8.51
51 (single 183 0.59 9 0.25
father*|single-fat*)
52 (single mother* 2653 8.60 235 7.32
|single-mot*)
53 (single parent* 919 3.14 334 9.38
|single-par*)
54 social* exclu* 66 0.21 153 4.19
55 *spend* 107,417 348.12 18,357 502.51
56 starv* 6248 20.25 408 11.17
57 struggl* 43,776 141.87 5298 145.03
58 *subsidi* 4649 15.07 1221 33.42
59 *tax* 130,913 424.26 28,920 791.67
60 vulnerable* 14,527 47.08 4794 131.23
61 *wage* 17,340 56.2 5352 146.51
62 *welfare* 17,666 57.25 7726 211.49
63 work* 449,073 1455.35 75,934 2078.65
64 (work capability 160 0.52 362 9.91
assessment*|W-
CA*|work-capabili-
ty-assessment)
Total 3,344,571 535,419
to make our queries broad by using wildcards; that is we did not merely
search for the collocates as listed in Table 5.1, but included all forms of
those collocates.6 For example, the query for the collocate unemployment
6Corpus query notation used in Table 5.2: * = none or more characters, {} = all forms of a lemma,
() = search for all forms within the brackets separated by the pipe (|) symbol.
105
was <*employ*> which would return hits for employ, employee, unem-
ployment, unemployed, etc. Table 5.2 also includes information about the
raw number of tokens in the news section of each corpus and the nor-
malised frequencies (hits per million words).
In addition to the terms listed in Table 5.2, we also tested par-
ticular items of evaluative lexis, including chav, feckless, hard work-
ing, lazy, lifestyle, luxury, shirker, skiver, and useless, but none of these
returned more than five PNCs and many returned none at all. Thus,
it can be determined that even if these terms are used within dis-
courses of poverty, they are not talked about in relation to particular
places. In some cases these terms were used in the corpora to refer to
individuals, and their location was mentioned; however, references to
individuals tended to occur as part of wider discourses of crime or
benefits fraud, which, for the present analysis at least, were not con-
sidered to refer directly to poverty (see Sect. 5.2). Having selected our
search terms, run our queries, and downloaded our concordances,
the next step was to geoparse the hits for our 63 queries (excluding
poverty, see Chapter 4).
5.2 The Boundaries of Discourse

The totals column in Table 5.2 shows that there were over 3.3 mil-
lion hits for these search terms in the news sections of the Daily Mail
corpus and over half a million in the Guardian corpus. However, the
Guardian corpus is smaller than the Daily Mail corpus, and not all
of the hits in Table 5.2 will relate directly to poverty. To address the
first issue, normalising to hits per million words shows that the Daily
Mail uses the search terms on average 10,315.12 times pmw whilst
the Guardian uses them 9,492.69 times pmw, which makes their
use across corpora more comparable. The second issue, that not all
query hits will be involved in discourses of poverty, is slightly more
problematic. In order to ensure that our analysis only included occur-
rences of our search terms that were used to talk about poverty, erro-
neous/irrelevant hits had to be removed. One way to do this would be
to manually analyse the concordance lines for all 3,879,990 hits listed
106
in Table 5.2, but this was clearly infeasible. We also had the option
to take a systematic stratified sample of these almost four million hits
but, as is shown below, not all search terms were equally associated
with discourses of poverty, and thus taking samples would likely have
skewed our results.
To establish which of our search terms were most associated with pov-
erty and place, we decided to geoparse all hits listed in Table 5.2. This
had two benefits. It eliminated hits which did not occur within ± 10
words of a place-name—thus greatly decreasing the number of concord-
ance lines for manual analysis—and it also gave us a rough measure of
which of our search terms were often associated with place. The raw fig-
ures for the geoparsed queries are provided in Table 5.3, which includes
information about the number of hits for each query, the number of
potential PNCs identified by the geoparser (All PNCs ±10), and the
final total of PNCs post manual analysis which were used within dis-
courses of poverty (PNCs post analysis). Information is also provided
about the percentage values of final PNCs to total number of origi-
nal hits, the percentage value of final PNCs compared to the potential
PNCs identified by the geoparser, and the normalised frequencies of
final PNCs per million words.7
The first conclusion that can be drawn from Table 5.3 is that
geoparsing all of our query hits significantly decreased the amount of
manual analysis needed as there were 503,739 concordance lines con-
taining potential PNCs (434,930 for the Daily Mail and 68,809 for the
Guardian ) in comparison to the almost four million total query hits.
Whilst still a large number, and geoparsing such large files (e.g. 449,703
concordance lines for <work*> in the Daily Mail ) did take time, this
was the most efficient way to thin our dataset. We chose to manually
analyse all 503,739 concordance lines containing PNCs within ± 10
7The first column (% of total hits) provides a measure of just how often one of the query hits
generated a PNC used within discourses of poverty. The second (% of all PNCs) provides a meas-
ure of how often PNCs used within discourses of poverty occurred in the total number of PNCs
identified by the geoparser.
Table 5.3 Concordance lines containing PNCs for all search terms
Daily Mail Guardian
No. of All PNCs % of % of PNC No. of All PNCs % of % of PNC
hits PNCs post total all pmw hits PNCs post total all pmw
±10 anal- hits PNCs ±10 analysis hits PNCs
ysis
1 *afford* 26,494 6927 43 0.16 0.62 0.14 6379 1497 356 5.58 23.78 9.75
2 *allowance* 7845 684 42 0.54 6.14 0.14 3318 204 78 2.35 38.24 2.14
3 *austerit* 6217 1118 8 0.13 0.72 0.03 2709 368 43 1.59 11.68 1.18
4 {beg} 8490 225 40 0.47 17.78 0.13 391 54 18 4.60 33.33 0.49
5 *benefit* 55,297 213 44 0.08 20.66 0.14 19,611 1785 382 1.95 21.40 10.46
6 *charit* 45,018 9881 58 0.13 0.59 0.19 11,841 1728 118 1.00 6.83 3.23
7 *claim* 301,829 25,155 99 0.03 0.39 0.32 26,897 3991 238 0.88 5.96 6.52
8 *communit* 72,368 54,841 44 0.06 0.08 0.14 17,829 1158 58 0.33 5.01 1.59
9 *cost* 117,875 14,310 60 0.05 0.42 0.19 18,336 2165 189 1.03 8.73 5.17
10 credit* 28,766 2989 17 0.06 0.57 0.06 5309 502 58 1.09 11.55 1.59
11 cut* 112,251 958 12 0.01 1.25 0.04 28,920 3200 235 0.81 7.34 6.43
12 *depriv* 4381 855 324 7.40 37.89 1.05 1948 612 383 19.66 62.58 10.48
13 *disab* 20,457 1676 14 0.07 0.84 0.05 4227 339 90 2.13 26.55 2.46
14 DLA 327 21 2 0.61 9.52 0.01 265 7 1 0.38 14.29 0.03
15 dole* 2616 224 20 0.76 8.93 0.06 261 20 11 4.21 55.00 0.30
16 Duncan 3604 295 15 0.42 5.08 0.05 1919 164 34 1.77 20.73 0.93
Smith*
17 earn* 37,803 2907 20 0.05 0.69 0.06 5126 587 90 1.76 15.33 2.46
18 *employ* 87,550 561 38 0.04 6.77 0.12 17,042 329 223 1.31 67.78 6.10
19 *entitl* 1699 1070 9 0.53 0.84 0.03 2103 229 6 0.29 2.62 0.16
20 *equali* 8049 968 7 0.09 0.72 0.02 4392 565 87 1.98 15.40 2.38
21 ESA* 751 122 9 1.20 7.38 0.03 320 5 2 0.63 40.00 0.05
107
(continued)
Daily Mail Guardian
108

ysis
22 *expens* 34,726 5938 15 0.04 0.25 0.05 5383 668 69 1.28 10.33 1.89
23 fit for work 113 7 7 6.19 100.00 0.02 182 2 2 1.10 100.00 0.05
24 (fitness to 24 5 2 8.33 40.00 0.01 15 0 0 0.00 0.00 0.00
work|fitness-
to-work)
25 (food bank* 1424 346 99 6.95 28.61 0.32 1347 284 128 9.50 45.07 3.50
|foodbank*
|food-bank*)
26 handout* 2835 375 26 0.92 6.93 0.08 198 25 8 4.04 32.00 0.22
27 hardship* 1754 158 18 1.03 11.39 0.06 430 22 8 1.86 36.36 0.22
28 *home* 482,368 93,395 96 0.02 0.10 0.31 40,010 9475 595 1.49 6.28 16.29
29 hous* 304,246 57,913 88 0.03 0.15 0.29 33,468 6723 291 0.87 4.33 7.97
30 (hun- 9227 810 14 0.15 1.73 0.05 947 163 8 0.84 4.91 0.22
ge*|hungr*)
31 income* 23,655 2096 36 0.15 1.72 0.12 7680 799 122 1.59 15.27 3.34
32 IDS* 672 40 0 0.00 0.00 0.00 187 12 0 0.00 0.00 0.00
33 job* 117,232 13,146 623 0.53 4.74 2.02 22,583 2537 675 2.99 26.60 18.48
34 JSA 118 8 0 0.00 0.00 0.00 184 14 5 2.72 35.71 0.14
35 *money* 129,601 7197 47 0.04 0.65 0.15 16,662 1468 110 0.66 7.49 3.01
36 *paid* 70,692 4993 53 0.07 1.06 0.17 9569 1070 89 0.93 8.32 2.44
37 pay* 170,672 21,514 102 0.06 0.47 0.33 24,752 1180 96 0.39 8.14 2.63
38 penn* 27,146 5344 28 0.10 0.52 0.09 1247 332 1 0.08 0.30 0.03
(continued)
Daily Mail Guardian
ysis
39 pension* 42,851 721 4 0.01 0.55 0.01 9081 665 44 0.48 6.62 1.20
40 personal 59 9 2 3.39 22.22 0.01 76 1 0 0.00 0.00 0.00
inde-
pendence
payment
41 PIP 623 124 0 0.00 0.00 0.00 196 15 0 0.00 0.00 0.00
42 poor* 36,648 3991 264 0.72 6.61 0.86 9675 1283 307 3.17 23.93 8.40
43 *poverty* 8689 608 241 2.77 39.64 0.78 5288 745 245 4.63 32.89 6.71
44 price* 82,198 12,654 7 0.01 0.06 0.02 7428 1071 31 0.42 2.89 0.85
45 redundan* 3920 414 141 3.60 34.06 0.46 1484 97 42 2.83 43.30 1.15
46 rehous* 227 63 13 5.73 20.63 0.04 144 43 38 26.39 88.37 1.04
47 rent* 22,241 5076 56 0.25 1.10 0.18 4572 840 251 5.49 29.88 6.87
48 *salar* 15,107 1337 5 0.03 0.37 0.02 2140 194 34 1.59 17.53 0.93
49 *sanction* 9505 1136 2 0.02 0.18 0.01 1934 125 17 0.88 13.60 0.47
50 *scroung* 608 67 10 1.64 14.93 0.03 311 23 5 1.61 21.74 0.14
51 (single 183 22 1 0.55 4.55 0.00 9 2 2 22.22 100.00 0.05
father*
|single-fat*)
52 (single 2653 336 22 0.83 6.55 0.07 235 56 28 11.91 50.00 0.77
mother*
|single-mot*)
(continued)
109
Daily Mail Guardian
110

ysis
53 (single 919 78 5 0.54 6.41 0.02 334 19 7 2.10 36.84 0.19
parent*
|single-par*)
54 social* exclu* 66 11 4 6.06 36.36 0.01 153 31 6 3.92 19.35 0.16
55 *spend* 107,417 4882 11 0.01 0.23 0.04 18,357 2433 51 0.28 2.10 1.40
56 starv* 6248 377 8 0.13 2.12 0.03 408 37 3 0.74 8.11 0.08
57 struggl* 43,776 4336 88 0.20 2.03 0.29 5298 579 92 1.74 15.89 2.52
58 *subsidi* 4649 708 5 0.11 0.71 0.02 1221 202 4 0.33 1.98 0.11
59 *tax* 130,913 16,495 57 0.04 0.35 0.18 28,920 3711 99 0.34 2.67 2.71
60 vulnerable* 14,527 1453 6 0.04 0.41 0.02 4794 318 49 1.02 15.41 1.34
61 *wage* 17,340 1877 79 0.46 4.21 0.26 5352 691 260 4.86 37.63 7.12
62 *welfare* 17,666 1967 71 0.40 3.61 0.23 7726 779 132 1.71 16.94 3.61
63 work* 449,073 36,893 378 0.08 1.02 1.23 75,934 10,564 93 0.12 0.88 2.55
64 (work 160 10 0 0.00 0.00 0.00 362 2 0 0.00 0.00 0.00
capability
assessment*|
WCA*|work-
capability-
assessment)
Total 3,344,458 434,930 3659 0.11 0.84 11.88 535,419 68,809 6746 1.26 9.80 184.67
111
words of one of our search terms. Although labour intensive, we made

this decision because we wanted our analysis to include all realisations
of discourses of poverty co-occurring with PNCs for our chosen queries
and we wanted to test GTA on a large scale in order to streamline the
method for future projects
Whilst the half a million hits seems like a lot of data to manually
analyse, the number of concordance lines was reduced somewhat when
we automatically eliminated geoparser errors (i.e. PNCs which were not
actually places) and PNCs which were not within the UK (e.g. Canada,
Morocco, and France). The number of hits were further decreased by
the removal of national-level UK place-names (UK, (Great) Britain,
England, Wales, Scotland, and Northern Ireland); although these hits
do not form part of the present analysis as they are not geographically
specific enough for our purposes, we have kept this data for future work
on discourses of poverty. It is worth investigating whether ‘UK poverty’
is discussed in similar or different terms to the localised poverty we are
focused on here.
A third set of hits which could easily be eliminated related to errone-
ous query hits. As the queries were broad, there were occasions where
queries returned hits which were not directly related to our initial set
of search terms. For example, <*wage*> returned relevant hits such as
wage and wages, but also returned sewage, whilst <*tax*> returned tax-
payer(s), but also taxi. One response to these erroneous hits would be
to write more specific queries, but we wanted our analysis to be as com-
prehensive as possible and did not want to miss anything significant,
but unpredicted, in our analysis of poverty and place. In any case, as the
geoparser outputs are viewed as spreadsheets, unwanted query hits are
easily filtered out and eliminated.
We determined whether PNCs occurred within discourses of pov-
erty by looking at the wider co-text of the PNC and its related query
term. In order to keep this process consistent, one author analysed all
the concordance lines. Whilst reading the concordance lines in this way
was somewhat subjective, there was no other (computational) method
that we could use to ensure the PNCs that we would eventually plot
112
(see the following chapters) were used within discourses of poverty.8 In

any case, most of the concordance lines in the geoparsed output were
either related to discourses of poverty, as in (3–4), or clearly not about
poverty, as in (5).
3. Until recently, Bo worked behind the bar of a West London pub and
says she struggles to pay her bills ‘just like everyone else’ (DM news,
Apr 2012).
4. He was already aware of a problem in South Shields, where around
610 young people have been claiming Jobseeker’s allowance for six
months or longer (G news: politics, July 2012).
5. In Bournemouth, where private income has increased ten-fold,
the number of patients waiting too long for operations has risen
(G news: society, Nov 2014).
However, we had to make some decisions about what actually consti-

tuted discourses of poverty. Whilst we wanted to keep the scope of our
analysis broad, there were occasions where our search terms occurred
but references to poverty were implied or secondary to the aim of a
given text. As shown in (6–7), when reporting on criminal cases and
court proceedings, the Daily Mail had a tendency to state whether a
defendant was unemployed or in receipt of government benefits.
6. Prosecutor Gareth Hughes told Manchester magistrates that the out

of work hotel chambermaid spat on each car, yanked off a wing
mirror, damaged two others and bent back windscreen wipers (DM
news, July 2011).
7. Taylor smashed windows with a hammer before pulling out the knife
at Fleetwood Job Centre (pictured) in Lancashire. The jobseeker,
who has 59 previous convictions, was jailed for 42 weeks (DM news,
Mar 2015).
8It is possible to avoid this issue by only selecting texts related to the research topic at hand,
rather than working with such broad corpora.

113
In (6), the fact that the defendant is ‘out of work’ is inconsequential

to the charges against them, and similarly in (7) the jobseeker status of
the accused—which implies receipt of Jobseeker’s Allowance—bears no
causality to their actions. The explicit mention of employment status or
implicit references to benefits within a wider discourse of crime serves
to make conceptual and ideological links between crime and benefits
and portray benefits recipients as deviant. A relationship between crime
and benefits receipt was also made explicitly in the Daily Mail corpus,
through the (over)reporting of benefit fraud, as shown in (8–10). There
was also a tendency to capitalise on big-money cases by reporting on the
same case multiple times.
8. an asylum seeker from Sierra Leone fraudulently stolen (sic) more
than 400,000 in benefits while holding down two jobs in London
(DM news, Sept 2012).
9. Benefit cheat: Kelvin Kaloo, of Dunstable, Bedfordshire,9 claimed
nearly 100,000 in handouts from Brent Council and Central
Bedfordshire Council (DM news, Jan 2013).
10. The Mail regularly highlights the most outrageous abuses of an
insanely over-indulgent system, including the Afghan family of
seven, living in a £1.2 million house in Acton and costing taxpay-
ers £170,000 a year in benefits (DM news, Feb 2010).
However, benefit fraud is not automatically or necessarily directly

related to poverty; reports of benefit fraud in the Daily Mail tend to
focus on extreme examples where claimants had fraudulently obtained
many thousands of pounds. Someone with such wealth at their disposal
cannot really be considered poor or in poverty. Thus, the decision was
made to remove hits relating to crime and benefits fraud from the anal-
ysis. We consider the significance of the topics and discourses which are
related to discourses of poverty, such as crime and immigration (see 10)
in more detail in the conclusion.
9Where two place-names co-occurred in this way, we only included the most local place-name in
our analysis to avoid duplicating PNCs.
114
Table 5.4 Queries with low/zero PNCs

No. PNCs Daily Mail Guardian
0 IDS* IDS*
PIP personal indepdenence
JSA payment
(Work capability assess- PIP
ment*|WCA*|Work- (Work capability assess-
capability-assessment) ment*|WCA*|Work-
capability-assessment)
1 (single father*|single-fat*) DLA
(single parent*|single-par*) penn*
2 DLA ESA*
(fitness to fit for work
work|fitness-to-work) (single father*|single-fat*)
personal independence
payment
*sanction*
3 – starv*
4 pension* *subsidi*
social* exclu*
Some search terms were clearly more related to discourses of poverty

and place than others. As shown in Table 5.4, four queries returned no
relevant PNCs in the Daily Mail and four returned no relevant hits in
the Guardian. Similarly, there were several queries that, when geoparsed
and manually analysed, returned fewer than 5 PNCs used within dis-
courses of poverty.
All of the search terms producing small numbers of PNCs were those
that we added to the list of search terms originally generated using col-
location data. This result suggests that the method of using collocate
analysis to select query terms is robust at identifying the key terms in dis-
course which index particular ideologies. However, one issue that com-
plicates this finding somewhat is that most PNCs returned for <(single
father*|single-fat*)> were related to poverty as were 100% of the PNCs
for <fit for work> both of which were addded to the list of search terms
but did not appear in the collocate lists for either newspaper.10
10The query relating to single fatherhood also had a 100% return in the Guardian, compared to a
4.55% return in the Daily Mail.

115
11. Almost a third of people in Burnley and Aberdeen were fit for

work and 38% had the potential to work with the correct support
(DM news, Apr 2011).
12. Today Chris, an accountant from Dronfield, Derbyshire, is a single
father. He went bankrupt and lost his home paying out half a mil-
lion pounds in his fight for answers into his wife’s death (DM news,
Jan 2014).
13. Lashbrook is a single father living in south Kilburn with his baby
daughter, who is now almost two. Out of work following a con-
struction work injury in 2010, this is the fourth time he has been
forced to turn to Sufra for help (G news: socety, Apr 2014).
Thus, we can posit that when these newspapers mention place in relation
to fit for work (a type of assessment for government benefits) they over-
whelmingly do so to talk about poverty. On the other hand, however,
many of the queries in Table 5.4 are the names of particular UK govern-
ment benefits—Personal Independence Payment (PIP) and Employment
Support Allowance (ESA)—but these queries did not return high per-
centages of PNCs. This, it seems from the raw data that there is no strong
relationship between particular benefits and place, although as the analy-
sis in the following chapters shows, there were exceptions to this finding.
Taking proportion into account, what we can say about search terms
that returned low numbers of PNCs is that, fundamentally, these search
terms are not associated with particular locations in our corpora. We
can take this one step further by comparing the overall percentages for
all queries combined; only 0.11% of the Daily Mail and 1.27% of the
Guardian query hits related poverty to place. The implications of this
finding for critical discourse analysis are significant: they suggest that
poverty, which is treated as a measurable phenomenon (see Sect. 5.1), is
not locatable. Discourses of poverty may be drawn upon to talk about a
UK or global picture, and may include terms such as single-parent, social
exclusion, and sanction, but discussion on such a wide geographical scale
means that poverty is made somewhat abstract, somewhat elsewhere, and
somewhat other. Focusing on the local aspects of poverty make it less of a
social phenomenon and more of an observable, concrete reality. Because
poverty is not consistently associated with place, the use of place-names
116
in reports about poverty become salient, insofar as they are anchoring a

text to a particular location. Thus, by extracting PNCs, we can interro-
gate the significance of place-name mentions for discourses of poverty.
On the other hand, there were queries that returned relative high
numbers of PNCs. <job*> returned the highest number in both cor-
pora (623 in the Daily Mail and 677 in the Guardian ). As well as rele-
vant hits for job and jobs (which tended to refer to job losses) as shown
in (14–15), this query also returned hits for jobseeker and Jobseeker’s
Allowance (16–17) which quantified the number of people claiming
benefits, discussed the ratio between job applicants and vacancies, and
ranked the areas with the highest proportion of JSA claimants (19).
14. dozens of workers could lose their jobs, particularly in Chirk near
Wrexham and Marlbrook in Herefordshire (DM news, July 2010).
15. A section of the Lambeth website suggests claimants may need to
find a job to replace benefits (DM news, Feb 2013).
16. Figures published this week showed that 4045 18-to-24 year
olds in East Lancashire were claiming Jobseeker’s Allowance last
month (DM news, Apr 2010).
17. In Haringey, Lewisham, Waltham Forest and Hackney, people
claiming jobseeker’s allowance outnumber job vacancies by more
than 20 to 1 (G news: society, Aug 2011).
18. Many claimants in Birkenhead who have trouble with their bene-
fits end up at Mersey Advice, a welfare rights charity five minutes
walk (sic) from the jobcentre dubbed by staff “the fourth emer-
gency service in Birkenhead” (G news: society, June 2012).
19. four of the worst jobless claimant blackspots are in Birmingham
constituencies (G news: society, Jan 2012).
The top 10 queries by raw frequency are given in Table 5.5. Terms
including <*poverty*> and <poor*> are expected, given that they label
either the phenomenon of poverty or those experiencing poverty. The
occurrence of <job*>, <work*>, <redundan*>, <pay*>, and <*wage*>
indicates that terms relating to employment, and the money gained
from employment, are particularly prominent within discourses of pov-
erty and place. There is also a related set of terms concerning benefits
Table 5.5 Queries with highest raw values of PNCs
Daily Mail Guardian
Top PNCs No. % of all hits No. % of all hits
1 job* 623 0.53 job* 675 2.99
2 work* 378 0.08 *home* 595 1.49
3 *depriv* 324 7.40 *depriv* 383 19.66
4 poor* 264 0.72 *benefit* 382 21.40
5 *poverty* 241 2.77 *afford* 355 23.78
6 redundan* 141 3.60 poor* 307 3.17
7 pay* 102 0.06 hous* 291 0.87
8 *claim* 99 0.03 *wage* 260 37.63
9 (food bank*|food- 99 6.95 rent* 251 5.49
bank*|food-bank*)
10 *home* 96 0.02 *poverty* 245 4.63
117
118
receipt (<*claim*>, <*benefit*>) which suggest that poverty and place is

discussed in terms of welfare payments. Terms including those relating
to food banks and <*afford*>, and the evaluative label <depriv*> indicate
that discourses of poverty and place also concern a lack of money and a
need for basic necessities, such as food.
Finally in Table 5.5, there are four terms associated with housing,
<*home*>, which occurs in the top ten for both corpora, <hous*>,
and <rent*>. The fact that three housing-related terms occur in the
Guardian but there is only one in the Daily Mail could indicate that
discourses of poverty and place in the Guardian focus more on the topic
of homes/housing. Indeed, the two newspapers did treat these terms
differently. PNCs co-occurring with house and home in the Daily Mail
were just as likely to refer to the sale of a multi-million pound proper-
ties (20) as they were to refer to poverty. Such aspirational mentions of
housing and place did not tend to occur in the Guardian. In any case,
examples like (20) were eliminated during manual analysis as they do
not relate to poverty.
20. Or for £1.85 million they could have Perryston House, an

eight-bedroom home on Ayrshire’s ‘golf coast’ (DM news, Dec
2014).
21. It was completely unaffordable to working-class families like us.”
Tower Hamlets is one of the most deprived boroughs in London,
but the housing crisis has hit here too (G news, Feb 2012).
22. British housebuyers, particularly first-time buyers, could find them-
selves priced out of London, where house prices have risen to a
three-year high (DM news, Aug 2013).
23. A single mother with three children was placed by her local council
in a £1.5m mews house, left, in Kensington, west London (DM
news, Aug 2012).
Where the two newspapers converge is in their discussion of UK house

prices, particularly the unaffordability of houses in London. However,
whilst the Guardian takes the position that rising house prices are
excluding poorer groups of society from some areas (21), the Daily
Mail, whilst acknowledging the difficulties of getting on the property
119
ladder (22), uses the high prices of particular areas of London to sug-
gest that people claiming Housing Benefit should not be able to live
there (23). Thus, whilst UK property prices are not necessarily related
to poverty—not being able to afford a house in central London does
not automatically make one poor—references to benefits recipients hav-
ing access to properties that would be unaffordable for many imply that
perhaps those who claim benefits are receiving too much money. Of
course, what is not considered in examples such as (23) is that many
people in receipt of Housing Benefit, especially in London, are in paid
employment and, furthermore, if they were to lose Housing Benefit
(or have it capped) they may have to relocate, which might also lead to
unemployment.
5.3 Summary
This initial analysis of multiple geoparsed search terms has shown
that, despite differences between the two corpora, there are repeated
key elements which appear to sit at the core of discourses of poverty
and place: employment, money, housing, and benefits. The process of
expanding GTA to consider how multiple search terms work together
within wider discourses is fundamentally still in its infancy. However,
the method detailed above for determining suitable search terms based
on collocates of a primary node seems robust. Indeed, where addi-
tional queries were generated in response to wider reading, they tended
to be less fruitful than those terms which collocated with <*poverty*>.
One potential way to eliminate less fruitful terms in future GTA-
based discourse analysis would be to only include search terms which
were bidirectional collocates (e.g. words which are collocates of pov-
erty, but which also had poverty as a collocate). However, for analysis on
large, non-specialised corpora this may decrease the number of poten-
tial search terms too much. To decrease interference from erroneous
query hits, it would be advisable to work with lemmas (although this
depends on the availability of lemmatised corpora) or run a process of
trial and error where different query syntax is tested before geoparsing.
As this analysis is the first of its kind, we chose not to do this in order
120
to demonstrate some of the issues that arise when geoparsing multiple

search terms. Fundamentally though, the occurrence of erroneous hits
only affected the proportional measures, as reported in Table 5.3.
The most difficult part of the process is deciding which concordance
lines to accept as realisations of discourses of poverty and which to elim-
inate. This required some subjective judgement. It is possible that other
scholars would choose to draw different boundaries than the ones we
have used. For example, they might include reports of benefit fraud,
or they might exclude generic references to job losses. Indeed, not all
job losses lead directly to poverty, and cases where an individual lost an
extremely-well-paid job were not included in our analysis. However,
generic references to job cuts, job losses, and redundancies tended to
be quantified (‘800 job losses’, ‘10,000 prison and probation jobs cut’),
which suggested that a large number of people would end up unem-
ployed and not earning a wage. As such, we opted to keep them in our
analysis.
In order to structure our analysis of the different elements of dis-
courses of poverty and place, we grouped our search terms into the
subsets shown in Table 5.6. Unavoidably there were some queries that
were difficult to categorise. Terms such as <*salar*> and <*wage*> clearly
relate to employment, but they have been allocated to the Money subset
because salary and wage relate directly to the money received for employ-
ment, as does the term <*pay*>. However, <*pay*> can also relate to
the payment of benefits, as well as the payment of wages. Therefore, we
decided to keep all of these related terms together in one subset.
Having established these subsets, we can see that, proportionally at
least, the Daily Mail ’s coverage of poverty and place tends to focus on
employment and evaluation more than people and food, whereas the
Guardian hits are spread more evenly across subsets. In both corpora,
food, people, and social unity appear to be more marginal elements and,
as such, we do not look at these subsets in great detail here. The follow-
ing chapters take a closer look at what is actually written about poverty
and place, and where these different elements—employment, benefits,
money, and housing—are geographically located by each newspaper.
121
Table 5.6 Subsets of search terms

Daily Mail Guardian
No. % No. %
Employment *employ*, job*, redundan*, 1180 34.56 1036 15.92
work*
Money *afford*, *austerity*, *cost*, 514 15.06 1542 23.70
earn*, *expens*, income*,
*money*, *paid*, pay*,
penn*, price*, *salar*,
*spend*, *wage*
Benefits *allowance*, *benefit*, 446 13.06 1397 21.47
*claim*, credit*, cut*,
*disab*, DLA, dole*, Duncan
Smith*, ESA*, fit for work,
(fitness to work|fitness-to-
work), handout*, IDS*, JSA,
personal independence
payment, PIP, *sanction*,
*subsidi*, *tax*, *welfare*,
(Work capability assess-
ment*|WCA*|Work-
capability-assessment)
Food (food bank*|foodbank* 121 3.54 139 2.14
|food-bank*),
(hunge*|hungr*), starv*
Housing *home*, hous*, rehous*, 253 7.41 1175 18.06
rent*
Social unity *charit*, *communit*, 109 3.19 263 4.04
*equali*
Evaluation {beg}, *depriv*, *entitl*, 763 22.35 875 13.45
hardship*, poor*,
*scroung*, social* exclu*,
struggl*,vulnerable*
People pension*, (single father* 28 0.82 80 1.23
|single-fat*), (single
mother*|single-mot*),
(single parent*|single-par*)
Total 3414 6507
122
References
6
Locating (Un)Employment
in the National Press
Having established the key terms associated with poverty in our corpora,
this chapter focuses on two of the subsets of search terms identified in
Chapter 5. We take employment—a topic closely related to poverty—
as a case study and how show how discourse analysis and GTA can be
combined. The chapter begins by contextualising discussions of (un)
employment within their wider social context and also considers the
linguistic co-text of hits returned by the query <*employ*> (Sect. 6.1).
Section 6.2 discusses the similarities in the geographical distribution
of PNCs for the query <*employ*> and the Employment subset more
broadly. This is followed by an analysis of the language used in the
Employment subset (Sect. 6.3) and the Money subset (Sect. 6.4).
6.1 Social and Linguistic Context

of Unemployment
We begin with an overview of the wider social context within which the
texts in our two corpora were produced. After the global financial crisis in
2008, the UK unemployment rate was 6.4% and rose to a peak of 8.4%
in the final quarter of 2011. Since that point, unemployment levels have
https://doi.org/10.1007/978-3-319-93503-4_6
124
fallen, moving below 7% in the first quarter of 2014, and statistics from
the third quarter of 2016 place it at 4.8% (ONS 2016a). However, despite
the statistics suggesting positive overall trends, 4.8% equates to 1.6 million
people. During the time period covered by our corpora Lansley and Mack
(2015: 90–91) note cases where over 1700 people applied for eight jobs
in a coffee chain-store in Nottingham in 2013, 1300 applied for twenty
fixed-term Christmas positions in retail in the Midlands in 2010, and
thirty cinema jobs in West Bromwich in 2013 received 2000 applications.
Unemployment can be measured in a number of different ways.
The International Labour Force definition includes everyone aged
over 16 who is not in employment and is either available to start
work within the next two weeks, waiting to start a new job within the
next two weeks, or has been actively looking for work in the last four
weeks (ONS, n.d. (a)). This is the measurement used by the 2011 cen-
sus (ONS 2014). An alternative measure of unemployment is to look
at workless households, defined as households that contain at least
one person aged between 16 and 64 where no-one aged 16 or over is
in employment (ONS, n.d. (b)). Figure 6.1 shows these two different
measures of unemployment, at slightly different dates. The maps show
strongly similar patterns, although there is more worklessness in rural
areas, compared to unemployment which tends to be a more urban
phenomenon. Table 6.1 measures these similarities formally using cor-
relation analysis. It shows that the two measures are strongly correlated,
and both also correlate closely with Carstairs scores. Therefore, a rela-
tionship clearly exists between geography and unemployment and, by
extension, poverty, which is persistent between different definitions.
Unemployment tends to be lower in the South-East and higher in
urban areas including London, South Wales, the West Midlands, the
North-West and North-East of England, and central Scotland. These
differences are marked. For example, in 2015, Lansley and Mack (2015:
116) noted that in ‘areas such as Hartlepool in the North-East of England,
and north and east Ayrshire in Scotland, the ratio of unemployed to job
vacancies has remained three times higher than the national average’. In
addition to formally-defined unemployment, the ONS indicates that
there were a further 8.89 million people who were ‘economically inactive’
and unable to work in the third quarter of 2016 (ONS 2016b: 3). This
group includes students, people looking after family members (especially
Fig. 6.1 Different definitions of unemployment using a the 2011 census definition and
6 Locating (Un)Employment in the National Press
b workless households 2015. Legends use quintiles

125
126
Table 6.1 Correlation coefficients between measures of poverty, unemploy-

ment, and worklessness at local authority district level. Excludes Northern
Ireland
Census unemployment Workless households,
2011 2015
Carstairs Score, 2011 .949** .609**
Census unemployment – .602**
2011
**Significant at p < 0.01
children), and people who have taken early retirement, as well as those
who have long-term illnesses. Most of those categorised as long-term sick,
or caring for family members will be in receipt of some form of govern-
ment benefit, such as Child Benefit, Personal Independence Payments,
Carer’s Allowance, and/or Employment and Support Allowance (ESA).
Due to the coalition government’s cuts to the welfare budget, those in
receipt of such benefits may have seen their payments of decrease or cease
entirely. Thus, poverty does not merely relate to unemployment, it also
shares a relationship with so-called economic inactivity.
To analyse (un)employment, as depicted in the news sections of our
two corpora, we can start at the macro level and compare the collocates
of <*employ*> for each newspaper. By beginning with the corpora—not
just focusing on the immediate co-text of PNCs—we can provide evi-
dence for each newspaper’s general stance on this issue, which we can use
in close analysis of their representation of (un)employment at particular
locations. We can also, of course, compare the PNCs for <*poverty*>
and <*employ*> directly. There are 87,550 hits for <*employ*> in the
news section of Daily Mail (a normalised frequency of 283.73 pmw). In
comparison, there are 17,042 hits in the Guardian news Section (466.52
hits pmw). From this we can deduce that, overall, the Guardian uses the
lemma employ1 relatively more frequently than the Daily Mail.
However, such a broad-stroke analysis sheds light on why we cannot
reduce poverty discourses to individual search terms; the use of employ
1Small capitals denote lemmas: all forms of a word, such as employ, employee, unemployed, employ-
ment, etc.
127
does not cover all potential references to employment. Reference could

be made to particular job titles or alternative terms like job or work may
be used. Hits for the query <*work*> show that the Guardian uses this
term 2078.65 times pmw, whereas the Daily Mail uses it 1455.4 times
pmw. Similarly for job the figures are 618.2 and 379.93 respectively.
Despite the limitations of such statistics (we can ‘employ common
sense’ and ‘work out at the gym’, neither of which relate to employ-
ment, but are included in the normalised frequencies2) the fact that
the Guardian has higher normalised frequencies for all these employ-
ment-related terms is evidence that the Guardian appears more con-
cerned with (un)employment than the Daily Mail.
Taking a closer look more specifically at unemployment, we can start
to see nuances in how the term is used by each newspaper. The query
<(unemployment|unemployed)> returns 13,453 hits in the Daily Mail
(43.60 pmw) and 4357 hits in the Guardian (119.27 pmw), suggest-
ing that the latter refers to unemployment relatively more frequently. To
highlight the semantic fields associated with unemployment, we man-
ually categorised the top 100 collocates in each newspaper (Table 6.2).
There were 50 shared collocates, 35 were unique to the Daily Mail and
37 were unique to the Guardian (7 grammatical terms and numbers
were removed).
Overwhelmingly, the groups of collocates suggest that, similarly to
poverty (see Chapter 5), unemployment is discussed by both news-
papers in terms of its measurability: 24 of the 50 shared collocates
occur in this category. However, despite both the Daily Mail and the
Guardian appearing to focus on unemployment rates, statistics and levels,
the collocates unique to the Daily Mail appear to focus on falling unem-
ployment (1) whilst the Guardian tends towards collocates like mass and
risen, which emphasise increases in unemployment (2).
1. The quarterly fall in overall unemployment was mainly among 16 to

24-year-olds, with unemployment in this age group down by 42,000
to 917,000 (DM news, July 2011).
2Complex queries could reduce some of this noise, but the normalised frequencies are merely a
litmus test for each newspapers’ coverage of employment.
Table 6.2 Top 100 lexical collocates of unemployment
Shared collocates Unique to DM 100 Unique to G 100
Measurement
128
falling, figures, growth, high, fall, fallen, falls, fell, median, average, increasing, mass, risen
higher, highest, increase, more, overall, per cent, per-
inflation, level, levels, long- cent, record
term, low, lower, lowest,
million, number, quarter, rate,
rates, rise, rising, rose, soaring,
statistics
Work employment, job, jobs, unem- Hiring placements, programme,
ployed, unemployment, work, scheme, schemes, vacancies,
workers working
Benefits allowance, benefit, bene- claimants, jobcentre, joblessness,
fits, claiming, dole, jobless, jobseekers, welfare, workfare

jobseeker
Money poverty, wages debt, income, prices incomes, insurance, low-paid,
paid, pensions, unpaid
Policy cuts, economic, economy, austerity, government, recovery
recession
People/ people, young, youth Americans, graduate, graduates, disabled, over-50s, under-25s
Demographics Labor, Obama, population
Locations country, Eurozone, Spain, UK areas
Time Months currently, seasonally, since year
Homes Housing Household homelessness
Action Tackle Hits find, help, helping, hit, living
Evaluation crisis, struggling, stubbornly deprivation, poor, scarring,
sickness
129
2. We are the undiscovered generation. With the number of unem-

ployed under-25s rising to almost 1 million this year (G news: soci-
ety, Mar 2010).
3. Mr. Kirby said the recent fall in unemployment was a ‘hiatus rather
than a turning point’, forecasting that nearly 450,000 workers will
lose their jobs over the next two years (DM news, Feb 2010)
This does not mean that the Daily Mail ignores increases in unem-
ployment rates or long-term trends, as is evidenced by the newspapers’
shared collocates and (3). However, analysing the collocates in this way
emphasises that, whilst the newspapers are working with similar lexical
resources, the discourses surrounding unemployment in both corpora are
somewhat different. For example, unemployment collocates with more
terms relating to benefits (claimants, jobcentre, etc.) in the Guardian,
whilst the Daily Mail links unemployment to government policies (aus-
terity, government, recovery ). A subset of unique collocates in the Daily
Mail includes Americans, Obama, Eurozone, and Spain, which suggest
consideration of unemployment in different nation states (similar to
the findings in Chapter 5). In order to focus more closely on the geo-
graphical aspect of poverty discourses in both corpora, we turn to the
PNCs for employment-related search terms in the Daily Mail and the
Guardian.
6.2 Poverty and Place in the Employment

Sub-set
Beginning with <*employ*>, there are only 223 and 38 PNCs in the
Guardian and Daily Mail, respectively, which relate to poverty. Thus,
the overwhelming conclusion is that neither newspaper systematically
associates the term employment and poverty with particular locations
in the UK. Figure 6.2 shows the distribution of <*employ*> PNCs for
both newspapers. It is clear that both newspapers are heavily concen-
trated on London and the south-east: 32% of the Daily Mail ’s PNCs
are in London rising to 61% in the south-east. Despite having far
130
more PNCs, the Guardian is even more London-centric with 72% of

its <*employ*> PNCs in London and 75% in the south-east as a whole.
The Guardian does, however, have around 5% of its PNCs in both
Scotland (mainly Edinburgh) and Northern Ireland (all in Belfast, which
did not occur for <*poverty*>) whereas the Daily Mail has no coverage
of <*employ*> in either of these areas. Neither newspaper pays much
attention to Wales; there are two Daily Mail PNCs in Merthyr Tydfil
and one Guardian PNC in Cardiff. Within England, both newspa-
pers have minor subsets in the north-east; the Daily Mail has six PNCs
(16%) in and around Newcastle, while the Guardian has a cluster of
fourteen PNCs centred on Hull.
Comparing the PNCs to the measures of unemployment reported in
Fig. 6.1, it is clear that there is little relationship between the newspa-
pers’ geographical coverage of (un)employment, and quantitative meas-
ures of its distribution. None of the ten districts with the highest rates
of worklessness (see Fig. 6.1) are in London. The Guardian has only
three PNCs these ten districts, two in Glasgow and one in Liverpool,
while the Daily Mail has two, both in Merthyr Tydfil. Using the sta-
tistics for unemployment the emphasis on London is perhaps slightly
more justifiable as three of the ten districts (outside Northern Ireland)
with the highest unemployment rates are in London. Nevertheless,
there is little to justify the newspapers’ concentration on London and
their apparent lack of interest in other areas.
However, focusing on <*employ*> alone cannot tell us the whole
story. To address wider discourses of (un)employment, we turn to the
Employment subset, which consists of the results for four related que-
ries (<*employ*>, <job*>, <redundan*>, and <work*>). As discussed
in Chapter 5, all PNCs returned by these queries, but not used within
discourses of poverty were eliminated. For example, <job*> occurred
in unrelated hits such as ‘paint job’ or ‘did a good job’. By manually
removing these hits, we thinned our dataset so that we only plotted
PNCs where there was a three-way relationship between the search
terms, poverty, and place. As a result, we removed noise from our data
and provide a nuanced picture of the geographical aspect of discourses
of poverty and employment in the UK media. The PNCs for the
Employment subset are shown in Fig. 6.3.
Fig. 6.2 Density smoothed maps of <*employ*> PNCs in a the Daily Mail and b the Guardian
131
132
The Daily Mail has a total of 1185 PNCs comprising 360 unique
place-names, whilst the Guardian has 1038 PNCs comprising 272
unique place-names. The fact that there are repeated PNCs (39 for
Birmingham in the Guardian, for example) suggests that particular
places are more associated with discussions of (un)employment than
others. Figure 6.3 shows that the pattern for the Guardian is quite sim-
ilar to that for <*employ*> in Fig. 6.2, being dominated by London
(41% of PNCs), although other urban centres such as Birmingham,
Liverpool and Manchester also emerge. The pattern for the Daily Mail
is, however, significantly more complex. Although still London-centric,
only 21% of the Daily Mail ’s PNCs are in London and there are also
major clusters around Birmingham, across the north of England, and
in Glasgow. This pattern of more geographical complexity in the Daily
Mail is also found when a Kulldorf ’s spatial scan statistic is used to iden-
tify places with more (hotspots) or fewer (cold spots) PNCs than would
be expected from the background geography found in the two corpora
(Fig. 6.4). As described in Chapter 4, this background geography was
determined using a 5% sample of PNCs for <the> in each corpus.
The Guardian has hotspots around Liverpool, Humberside (cen-
tred on Hull), and in the north-east of England, particularly in
Middlesbrough and Sunderland. Interestingly, London is not a hot-
spot, suggesting that the PNCs in the Employ subset actually repre-
sent the background geography of the Guardian corpus rather than
any particular association between London and unemployment. Much
of the south of England and of Scotland are cold spots suggesting that
the Guardian does not associate these places with unemployment.
The Daily Mail does have a hotspot in London but it is in north-east
London, whilst much of the rest of London is a cold spot. Elsewhere,
there are clusters stretching from Birmingham to Nottingham, one in
Liverpool stretching north through much of Lancashire, another in
areas east of Barnsley and Wakefield to Hull and Scunthorpe, and in
Middlesbrough and Newcastle, and Glasgow. There are also smaller
clusters associated with a single place-name: Swindon is associated
with redundancies at the Honda car plant there, and Aberdeen is asso-
ciated with a pilot scheme for Incapacity Benefit reassessment.
Fig. 6.3 Density smoothed maps of PNCs in the Employment subset in a the Daily Mail and b the
Guardian
133
134
Fig. 6.4 Kulldorf clusters of PNCs in the Employment sub-set in a the Daily Mail and b the
Guardian
135
6.3 Locating the Language of Unemployment

In order to understand how the locations referred to were used within
discourses of poverty (and employment) we began with PNC keyword
analysis—a form of contrastive concordance analysis—as described in
Chapter 4. PNC keywords are a robust way to ascertain how the news-
papers are internally different in relation to their coverage of London
and the rest of the UK. Comparing the PNC keywords for each news-
paper can also act as a test of how similar they are in their coverage of
(un)employment. However, as noted in Sect. 6.1, although the news-
papers may draw on the same lexical resources, they do not necessarily
represent a given topic in the same way. To compare the newspapers in
detail and facilitate triangulation, we analysed samples of concordance
lines, following the method proposed by Baker (2006). This involved
taking a random sample of 50 concordance lines3 from the Employ sub-
set for each newspaper and undertaking manual analysis to identify key
themes, trends, categories, etc. Following the initial analysis a further
randomly-selected batch of 50 concordance lines were tested against the
patterns found in the original 50. The analysis proceeded in batches of
50 concordances until no further patterns were found. This procedure
was repeated for all subsets (Employ, Money, Benefits, Housing, and
Evaluation) which are discussed across the following chapters. In most
cases 250 concordance lines was sufficient to identify the major trends
in each subset and also highlighted some more minor trends in the data.
Table 6.3 shows the PNC keywords for the Employ subset in the
Daily Mail. To maintain consistency in the analysis, the PNC keywords
have been split into those relating to London and those relating to the
rest of the UK. However, as the Kulldorf analysis has shown, there was
not always such a dichotomy in the data. In order to avoid homogenis-
ing data for the rest of the UK, particular places of interest are explored
in the following close analysis.
3Arandom number generator was used to generate samples. No concordances were analysed
more than once.
136
Table 6.3 PNC keywords in the Employment subset in the Daily Mail (n ≥ 5)

Place Sig. PNC keywords
London <0.01 months, without, who, end, or, can
London <0.05 I, west, live, outside, back, hundreds, east, go, week,
into, families, Tesco, headquarters, more, that, than
Rest of UK <0.01 plant, at, per, police, lost, fit, near, had, capable, cent,
third, midlands, bae, said, losing
Rest of UK <0.05 factory, biggest, department, town, planning, claimants,
ten, sites, systems, even, redundant, Tata, vacancy,
places, refinery, former, Honda, hall, blow, hours, led,
tests, products, work
In London, PNCs that included Tesco referred to unpaid internships

and related to campaigns against companies who advertise vacancies
but only offer expenses and/or Jobseeker’s Allowance (4). References to
such vacancies are not restricted to London, and they also occur in the
Guardian, which reports on legal proceedings relating to unpaid place-
ments. The Daily Mail reports job losses across many different sectors in
London: terms like retail, firefighter, bankers, Deutsche bank, Homebase,
Argos, BBC, fitters, Ford, UBS, parcel, showrooms and HQ were PNC
keywords which met the threshold for statistical significance, but did
not occur 5 or more times (cf. criteria for Table 6.3). Many of the job
cuts/redundancies reported are actual. However, there are also reports
of threats to jobs and proposed redundancies (as is also found in the
Guardian, see below).
4. Right to work campaigners disrupted Tesco ’s Westminster branch

after the company advertised for workers in exchange for expenses
and Jobseeker’s Allowance (DM news, Feb 2012).
5. Meanwhile, in Tower Hamlets, the poorest borough in London and
arguably the most deprived in Britain, 58 employees have job titles
which contain the words ‘climate change’ or ‘global warming’ […]
You couldn’t make it up (DM news, Mar 2010).
6 Beggars belief: Despite hundreds of jobs being axed, Westminster
Council is recruiting an extra 50 traffic wardens. One of Britain’s
largest councils is to spend nearly £2million to increase its squad of
traffic wardens—while axeing (sic) more than 500 jobs (DM news,
Nov 2010).
137
Despite the Daily Mail reporting on job losses, which it generally eval-
uates negatively, the converse positive evaluation for job creation does
not hold true; the Daily Mail negatively evaluates particular types of
job, especially those relating to climate change (5) and the recruitment
of new traffic wardens (6). The use of ‘you couldn’t make it up’ in (5)
indicates a negative stance towards jobs relating to climate change. By
mentioning poverty in relation to these particular jobs, the Daily Mail
positions this type of job creation as opposed to or irrelevant to address-
ing poverty in Tower Hamlets. By implication, poverty is somehow a
‘real’ issue, whilst climate change is seen as less important. However,
there is no evidence that the climate change jobs mentioned in (5) were
in any way related to anti-poverty initiatives; the Daily Mail has made
this faux-connection. Similarly, the use of ‘axed’ in (6), as opposed to
more a neutral verb such as ‘cut’, positions the Daily Mail as opposed
to job cuts, but jobs for new traffic wardens are somehow unsatisfactory
(see the use of ‘despite’) and the phrase ‘spend nearly 2million’ implies
such jobs will cost too much or that money is better spent elsewhere.
PNC keywords for the rest of the UK are dominated by references
to job cuts; terms like plant, lost, products, near, losing, planning, sys-
tems, sites, Tat, BAE, Honda, refinery, places, and police all refer to real
or planned redundancies. Locating these in space, Nissan is associated
with creating jobs in Sunderland, whilst Vauxhall (Luton, Elsmere
Port), BMW (Longbridge), Honda (Swindon), BAE Systems (Brough),
Tata (Scunthorpe), Kraft/Cadbury (Somerdale, Bristol, York), Twinings
(North Shields) and Imperial Tobacco (Nottingham) are reported to be
cutting jobs across England. The Daily Mail states that, before losing
their jobs, Twinings workers in North Shields were expected to ‘train up
their Polish replacements’, noting that the workers ‘have accused bosses
of “rubbing salt into the wound”’. The Daily Mail thus takes an implied
negative stance on the relocation of jobs to Poland; the use of the quota-
tion from unnamed workers allows the Daily Mail to express dissatisfac-
tion that jobs in Britain are being moved elsewhere, but they never need
to state this explicitly.
The PNC keywords per, cent, ten, even, and vacancy refer to the num-
ber of people receiving government benefits, particularly Jobseeker’s
Allowance, or the number of vacancies in a given area. The Daily Mail
138
focuses on job availability by reporting that Aberdeen is the place where

there are ‘the greatest number of jobs available in construction’. In this
instance, reporting on locations where jobs are available (and focusing
on just one industry) works within the wider ideology that jobs are
available if people want them. However, no consideration is given to
the difficulties of relocating for work or the fact that not all jobseekers
are trained in construction. The focus on available jobs in Aberdeen is
contrasted with a set of PNC keywords associated with the reassessment
of recipients of Incapacity Benefit.4 Fit, capable, third, biggest, claimants,
test, and work are used to discuss the numbers of people being declared
fit for work after reassessment (7), particularly in Aberdeen and Burnley
where pilot studies were conducted (8).
7. Of 14,640 claimants, 5180 were fit. THE TOP 20 ‘WORKSHY’

AREAS Birmingham 5180 Glasgow City 3950 Liverpool 3280
Manchester 3030 County Durham 2970 Leeds 2570 Bradford 2430
Sheffield 2180 Stoke-on-Trent 1900 Wakefield 1880 Kirklees 183
(DM news, May 2013).
8. A pilot trial reassessing claimants who took up incapacity benefit
before 2008 found that nearly 30 per cent were fit for work. New
tests of claimants in Aberdeen and Burnley found that 29.6 per cent
should not be on sickness benefits at all (DM news, Feb 2011).
The use of ‘workshy’ to refer to all who have been assessed and deemed
fit for work homogenises people having benefits removed, independent
of the reasons for which they were reassessed as fit to work. The naming
and shaming of areas, all of which are notably in Scotland or the north
of England, does not take into account the overarching employment
and health profiles of these regions, and further implies that (large num-
bers of ) people in receipt of Incapacity Benefit are somehow devious.
4Incapacity Benefit was replaced by Employment Support Allowance (ESA) and existing claim-
ants were reassessed from 2010 onwards to determine if they met the criteria for ESA. The reas-
sessment process included a paper-based and/or face-to-face Work Capability Assessment (WCA)
with a third-party contractor.
139
The Daily Mail ’s focuses on the number of people without a job

(9) or claiming benefits is supported by concordance analysis, and it
has a tendency to rank locations (10). Areas in the north of England
are labelled as ‘jobless ghettos’ and there is even a list of ‘BENEFIT
BLACKSPOT AREA WORK-AGE ADULTS ON BENEFITS’ which
is capitalised to draw readers’ attention. The Daily Mail includes a list
of the ‘worst ten places to find work’ but this is part of the same article
about construction jobs in Aberdeen (see above). Thus, the overarching
effect is that the Daily Mail states that there are jobs available for those
who want them in places like Aberdeen (which it describes as ‘one of
the three happiest places in the UK’). It endorses the behaviour of those
who use unusual techniques to attain jobs (see (12) below), whilst rank-
ing and negatively evaluating those who receive government benefits.
9. Glasgow had the highest proportion of workless families—with

28.6 per cent of households without a job. Nearby Ayrshire is in
second—with 28.5 per cent of families in the area out of work.
Without students Nottingham sits in third place—with a jobless
rate of 27.5 per cent (DM news, Nov 2014).
10. Birmingham is listed among the five British cities suffering the
most from the blight of worklessness. It is placed fourth in the new
ONS worklessness table (DM news, Sept 2013).
11. Jeff Randall reported from unemployment blackspot Merthyr
Tydfil, where he discovered people who were so accustomed to liv-
ing in a dependency culture that they couldn’t even be bothered to
catch a bus a few miles to Cardiff, where jobs were available (DM
news, Mar 2015).
Example (11) shows these elements working together in relation to

location. The premodification of Merthyr Tydfil as an ‘unemployment
blackspot’ is combined with the homogenisation of benefits recipients
as people who are ‘accustomed to living in a dependency culture’. These
people are negatively evaluated because they ‘couldn’t even be bothered’
to travel for work. To support this position, and reinforce the implied
notion of laziness, the commute to Cardiff is presented as ‘a few miles’,
though the actual distance is 24 miles and the journey takes over an
140
hour by public transport. Thus the framing of this issue by the Daily
Mail does not take into account whether individuals could actually
make the journey to Cardiff (people may be restricted by caring duties,
for example), nor do they address the underlying issue that there are
apparently not enough jobs in Merthyr Tydfil. Furthermore, (11) is
an example of reference to generic benefits recipients; Sky reporter Jeff
Randall met ‘people’ who are never named, and the use of ‘discovered’ is
sensationalist, as it implies that Randall has found something or some-
one which has been hidden.
The Guardian also reports on Merthyr Tydfil and employment in
Cardiff, but presents the same story reported by the Daily Mail in a
different way. The Guardian positions the story as Iain Duncan Smith’s
denial of having a ‘get on yer bike moment’, which intertextually refers
to the famous quotation from Conservative MP Norman Tebbit in the
1980s, that his father had ‘got on his bike and looked for work, and
kept looking ‘til he found it’. This statement was controversial as unem-
ployment rates reached 12% in the 1980s and Tebbit’s view was taken
to endorse the neoliberal notion that those who were unemployed were
held back not by socioeconomic constraints, but their own lack of ini-
tiative. By referencing Tebbit, the Guardian makes links between the
Conservative policies of the 1980s and coalition policies on employ-
ment and welfare from 2010 onwards.
Concordance analysis also brought to light instances where the Daily
Mail focused on the employment status of individuals, with multiple
reports of people committing suicide because they lost their jobs. In
these cases, job loss is presented as causing suicide with no contextual
factors noted; the Daily Mail quotes the partner of one suicide victim as
saying ‘I’ve no doubt in my mind the reason Michael took his own life
was because he lost his job’. Reports of individuals posting job requests
on sandwich boards (12), billboards, hand-held signs, and Facebook
are evaluated positively. In the article (12) is taken from, the jobseeker
is described as ‘dedicated’, ‘determined’, someone who ‘has gone the
extra mile’ and who wants to ‘make an honest crust’. He is also directly
quoted as saying ‘I don’t want to claim benefits’. These examples make a
direct link between one’s employment status and identity (especially in
the case of the reported suicides) and sit well within capitalist ideologies.
141
12. Desperate: Jobless Martin Bailey from Nottingham who has

resorted to wearing with a sandwich board (sic) to try and attract
potential employer (DM news, Jan 2013).
13. But at Burnley Crown Court jobless Lumb was spared jail again
after claiming he was stealing six steaks to feed his new pet dog as
his state benefits had been cut (DM news, Sept 2014).
14. These are the 11 children of an unemployed mother who has pro-
voked outrage after it was revealed a taxpayer-funded £400,000
council home is being specially built for her family. The picture was
released as it was also claimed that jobless Heather Frost, 37, of
Churchdown, Gloucestershire, has paid for her partner to have fly-
ing lessons (DM news, Feb 2013).
Despite its occurrence in (12) the use of ‘jobless’ to premodify a name

or occupation is rarely used positively in the Daily Mail, and there are
repeated links between this naming technique and crime (13). For
example, a mother receiving an antisocial behaviour order is described
as ‘Jobless and single’, even though her employment and relationship
statuses are irrelevant to the legal proceedings reported. There are also
examples of a ‘jobless decorator’, an ‘out of work hairdresser’, and
a ‘jobless factory worker’, the latter of whom is the victim of assault.
Jobless is used in this position to refer to criminals, dads, ‘a squatter’, a
‘homeless drifter’, and there are multiple references to the wedding of a
‘jobless couple’. The use of this naming technique in (14) also taps into
wider discourses about benefits recipients; there is an implicature here
that someone who is unemployed, parent to a large number of children,
and living in council-owned accommodation should not be in a posi-
tion to buy flying lessons. Her status as a ‘jobless’ woman and ‘unem-
ployed mother’ positions her as a ‘flawed consumer’ (Bauman 2004)
and the purchase of flying lessons by someone in a ‘taxpayer-funded’
home taps into discourses of the undeserving poor (Katz 2013). The
term jobless is not used in this way in the Guardian, where it occurs
only seven times in premodifying position and tends to refer to generic
‘jobless households’ in Liverpool, Nottingham, and Glasgow.
Thus, interrogation of the language used by the Daily Mail to discuss
unemployment, redundancies, and the links that it makes (implicitly or
142
not) between job availability and benefits receipt suggests that the news-
paper’s treatment of such issues is underpinned by ideologies of neoliber-
alism—insofar as benefits recipients and the unemployed are responsible
for their own lot, independent of wider social factors such as job availa-
bility. Furthermore, and in direct contradiction to the individualisation
required by its apparent neoliberal stance, the Daily Mail ’s homogeni-
sation of those who are out of work and/or in receipt of government
benefits through the construction of so-called ‘workshy areas’ facilitates
the negative evaluation of entire groups of people without considera-
tion of individual circumstance. When individuals are mentioned, they
tend to be extreme (newsworthy) examples, such as parents to large
families or people claiming large amounts of benefit (see Sect. 7.2)
which are used to justify the newspaper’s stance on such issues. Such
individuals are not treated as exceptions to the norm (which statistically
they are—most people do not have 11 children), but are given as index-
ical examples of ‘the unemployed’ or ‘benefits claimants’. This allows the
Daily Mail to both ignore heterogeneity and to present welfare receipt
as a single problem, which is associated more with deviance than with
wide-spread redundancies and low vacancy rates.
In considering how the Guardian treats (un)employment, poverty,
and place (Table 6.4), there is more of a focus on the Living Wage
Campaign, which is discussed in detail in Sect. 6.4. However, repeated
references to minimum wage thresholds ‘in London’ and ‘outside
London’ increase the number of PNCs relating to this issue. Such labels
(inside/outside London) show how the geoparser cannot differentiate
between direct references to a place and references to elsewhere which
use a place-name as a starting point. The use of ‘outside London’ rein-
forces the need for close analysis to supplement geoparsing and fore-
grounds the apparent dichotomy between London and elsewhere.
Close reading of concordance lines indicated that there were also
references to job losses in the NHS and the London fire brigade and a
focus on the number of jobseekers for each available vacancy in particu-
lar areas, with the Guardian concentrating on those area where the ratio
is high (15), which contrasts with Daily Mail ’s focused on the availabil-
ity of jobs in Aberdeen.
143
Table 6.4 PNC keywords in the Employment subset in the Guardian (n ≥ 5)

London <0.01 living, if, outside, north, you, who
London <0.05 his, against, people, south, from
Rest of UK <0.01 centres, losses, face, loss, council, households, union,
plant, constituency, county, sector, trade, biggest, told,
programme, applied, public, greater, unison, beneath,
our, variation, large, labour, cut, economy, how, today,
redundancies, photograph, lab, midlands
Rest of UK <0.05 authorities, unemployed, plans, even, chasing, Sodexo,
expects, seeking, transferred, notices, applying, risk,
year, every, include, cuts, over, residents, managers,
department, john, policy, PWC, following, without,
maintenance, moment, himself, chief, shed, Pfizer, val-
leys, announced, been, call, regional, yes, AR, vacancy,
next, government, top
15. In Hackney, there were fewer than 500 vacancies for more than
11,000 claimants, while in Lewisham just over 10,000 people were
after 487 available jobs (G news: society, Aug 2011).
Outside of London, job cuts and redundancies are associated with the
most PNC keywords (losses, council, face, sector, economy, Pfizer, plant,
biggest, redundancies, midlands, announced, include, notices, Sodexo, call,
expects, chief, shed, seeking, risk, PWC, next, top, transferred ) but focus
more on police work, the MoD, and council-based positions than the
industrial occupations noted in the Daily Mail (see above). In particu-
lar, cuts and changes to the NHS 111 service (a non-emergency tele-
phone line) are localised to Bristol, Sheffield, Wakefield, Nottingham,
Hull, Stafford, Chelmsford, and Newcastle, which also accounts for
centres, loss, union, large, Unison, and trade. A further cluster of PNCs
concern unemployment rates and workless households (including vari-
ation, regional and today ) and involve repeated comparisons between
Liverpool, Nottingham, and Glasgow (where the number of workless
households is reported to be 1/3) and average rates of 1/9 elsewhere.
The Guardian focuses particularly on the number of people chasing every
vacancy and applying unsuccessfully for large numbers of jobs. Thus
they are embodying the neoliberal attitude that the Daily Mail seems
to expect, but the Guardian notes that they are unsuccessful because of
144
the lack of jobs, not their lack of effort. A similar theme is drawn upon
when claimants of Jobseeker’s Allowance are broken down by constitu-
ency: Birmingham, Belfast, Middlesbrough, and Liverpool are associated
with particularly high rates of people claiming this benefit.
Whilst the Guardian acknowledges job cuts in the private sector,
it is more likely than the Daily Mail to report on cuts and redundan-
cies in the public sector—in particular council-based employment and
the NHS—and it is also more likely to mention unions. However, as
shown in the Guardian ’s general treatment of employment (Sect. 6.1),
whilst the Guardian does report on actual job losses (16), many ref-
erences to job cuts are only proposed (17) or ‘could’ occur. Jobs are
quantified and described as ‘at risk’, or ‘under threat’, whilst cuts
are ‘expected’ or ‘planned’, and ‘redundancy estimates’ are made.
The Guardian has a tendency to report redundancies at the point of
announcement, not at the point where redundancies are made and
they are not as likely as the Daily Mail to state explicitly that jobs will
be ‘axed’, but rather draw on epistemic modality and/or conditionals
(see ‘unless a buyer can be found’ in 18) to avoid explicitly stating that
cuts will occur.
16. Job losses so far include Sunderland housing provider Gentoo,

axing 275 of its 1800 staff; 1400 jobs after the collapse of Leeds-
based housing maintenance firm Connaught last year (G news:
society, Mar 2011).
17. Beneath “Are job cuts proposed?” it is a yes in Sunderland, yes in
Darlington, yes in Durham, yes in Gateshead, yes in Hartlepool
and yes, yes, yes, yes in Newcastle, Northumberland, Redcar and
Cleveland (G news: politics, May 2010).
18. The company that boasts the world’s best-selling brand of haggis
will close its meat processing plant near Edinburgh with the loss of
1700 jobs unless a buyer can be found (G news: news, Oct 2012).
The Guardian also reports on potential job creation, both generically

(19) and using specific examples, such as (20), which relates to the con-
struction of the HS2 train line. Most examples of job creation refer to
London (or particular areas of London), whilst reports of job creation in
145
the rest of the UK tend to revolve around HS2 and/or the manufacture
of new trains. Again epistemic modality is used, insofar as the ‘economy
could generate’ jobs, and reports of job creations are ‘claimed’ by third
parties such as ‘council leaders’.
19. Over the next four years, London’s economy could generate

385,000 new jobs (G news: news, Jan 2011).
20. Council leaders have claimed a 20-minute cut in journey times
between Leeds and Manchester would be worth £6.7bn and create
nearly 30,000 jobs across the north of England (G news: society,
Nov 2014).
The Guardian reports on the number of people claiming government ben-

efits, particularly JSA (21) and Housing Benefit. As well as noting which
areas have high rates of benefits claimants, the Guardian contextualises such
reports by stating the number of available vacancies in areas of high unem-
ployment. Furthermore, example (22) debates the accuracy of such figures
by reference to the newspaper’s own research—‘even these figures mask the
true competition for work’—and reports that high jobseeker-to-vacancy
rates are a UK-wide phenomenon, not restricted to London or urban cen-
tres. The Guardian ’s position on unemployment rates and the number of
benefit claimants is thus in direct contradiction to the Daily Mail ’s position
that there are jobs if people want them, and that relocating/commuting (to
Aberdeen or Cardiff, for example) would decrease unemployment rates.
21. In Haringey, Lewisham, Waltham Forest and Hackney, people

claiming jobseeker’s allowance outnumber job vacancies by more
than 20 to 1 (G news: society, Aug 2011).
22. The worst-affected areas are spread all across the country:
Clackmannanshire in Scotland has 35 jobseekers for every
vacancy; the Isle of Wight has 21; Haringay (sic), London, 19; and
Inverclyde 18. Research by the Guardian suggests even these figures
mask the true competition for work (G news: society, Apr 2012).
The Guardian’s treatment of unemployment sits within their wider posi-

tion on benefits (see Chapter 7); there are no reports in the Employ
146
subset of people abusing or fraudulently claiming benefits, nor are

there links to flawed consumerism.5 Rather the Guardian reports on
government policy and alleged targets for the number of benefit recip-
ients expected to be sanctioned. It uses direct quotation (23) or sum-
marised testimony (24) to justify reporting on the alleged targets. The
use of ‘ordered’, ‘threatened’, and ‘disciplinary action’ position staff who
oppose sanctions as both vulnerable to management and implicitly sym-
pathetic to benefits claimants, whilst the description of (a subset of )
benefits claimants as ‘weak and vulnerable’ places them in direct con-
trast to the negatively evaluated ‘unscrupulous staff’. Thus, overall the
Guardian ’s position on (un)employment, poverty, and place, acknowl-
edges neoliberalism but rejects it in favour of foregrounding structural
inequalities and other social constraints; this is to be expected given the
left-leaning nature of the publication.
23. Further new evidence of targets is revealed, which includes: An

email from a Jobcentre manager in Derbyshire to say sanction
referrals are top priority. “By end Sept we should have been at 5.5%
(G news: society, Mar 2013).
24. A former personal adviser at a Leicester Jobcentre, says he was
ordered by managers to send more claimants for sanction, and was
threatened with disciplinary action when he questioned the policy.
Unscrupulous staff would target weak and vulnerable claimants for
sanctions, he states (G news: society, Jan 2015).
The PNC keyword and concordance analysis highlights the topics each
newspaper focuses on in relation to poverty, place, and employment.
However, it can also show what the newspapers do not do. With the
exception of the Guardian ’s focus on the Living Wage Campaign, there
are no explicit links between the Employment subset and money. Very
few of the query terms that comprise the Money subset occurred in the
PNCs for the Employment subset, suggesting little overlap in terms of
5Whilst potential PNCs concerned only with benefit fraud were eliminated (as discussed in
Chapter 5) there were some remnants of references to benefit fraud where it was mentioned
alongside other issues relating to poverty, such as homelessness, loss of employment, and people
being unable to pay their household bills.
147
how these issues are drawn upon within discourses of poverty and place.
Given that the terms in the Employ subset and the Money subset were
all collocates of <*poverty*>, and that employment is a primary way
of obtaining money, this absence of overlap is somewhat unexpected.
Nevertheless, it demonstrates that the different subsets may be used to
contribute to newspaper discussions of poverty in different ways. The
final section of this chapter focuses specifically on the Money subset.
6.4 Analysing Money Talk

The Money subset contains 14 queries6 but resulted in fewer PNCs
than the Employment subset. The Guardian has 885 PNCs and the
Daily Mail has 378, the geography of which is quite similar to the
Employ subset. The density smoothed and Kulldorf maps for the
Money subset are shown in Figs. 6.5 and 6.6. The Guardian is particu-
larly London-centric with 60% of its PNCs in or near London while
the Daily Mail is somewhat dispersed, also having clusters around
Birmingham and Manchester.
In both cases, inner London is a hotspot when compared to the
newspapers’ background geography. With the Guardian, this is the
only major hotspot, but the Daily Mail also has a significant hotspot
in the north-west of England that includes Manchester, Liverpool, and
Burnley. In both cases much of Scotland and Northern Ireland are cold
spots suggesting that money terms relating to poverty are not associated
with these places. In the Guardian, much of England is also a cold spot
which emphasises just how London-centric its coverage of this theme is.
To understand how the distributions in Figs. 6.5 and 6.6 relate to
discourses of poverty, we turn again to the PNC keywords. To keep the
analysis consistent we conducted a comparison between London and
the rest of the UK. However, as shown in Fig. 6.6, the Money subset
6<*afford*>, <*austerity*>, <*cost*>, <earn*>, <*expens*>, <income*>, <*money*>, <*paid*>,
<pay*>, <penn*>, <price*>, <*salar*>, <*spend*>, <*wage*>.
148
Fig. 6.5 Density smoothed maps of PNCs in the Money subset in a the Daily Mail and b the Guardian
Fig. 6.6 Kulldorf clusters of PNCs in the Money subset in a the Daily Mail and b the Guardian
149
150
Table 6.5 PNC keywords in the Money subset in the Daily Mail (n ≥ 5)

London <0.01 living, hour, be, wage, per, an
London <0.05 –
Rest of UK <0.01 left, after, from, her, biggest, told, save, beggar, nurse,
loan
Rest of UK <0.05 home, feminist, fines, awkwardly, university, went,
failed, claimed, t-shirt, having, such, she, just
Table 6.6 PNC keywords in the Money subset in the Guardian (n ≥ 5)

London <0.01 living, hour, wage, the, that, an, pay, than, workers, can,
minimum
London <0.05 is, do, we
Rest of UK <0.01 different, council, west, charities, from, city, town,
claimed, strong, labour, recording, never, claimants,
springing, trustees, commuter, little, ‘minimum, cal-
culating, response, millions, county, midlands, unem-
ployed, took, five, extra, jobseekers, up
Rest of UK <0.05 years, like, measures, advice, rather, growth, pounds,
average, week, about, spend, coach, and, give, some,
also, another, combined, respectively, face, MP, hun-
dreds, campaigns, increases, already, place, conference
is not always realised as such a geographical dichotomy. Furthermore,

there were more similarities than differences between the newspapers.
Thus the analysis is structured by theme, rather than by source. The
PNC keywords for the Daily Mail are given in Table 6.5 and for the
Guardian in Table 6.6.
The primary theme in the PNC keywords is a focus on the Living
Wage Campaign, which encourages employers to pay their workers
a minimum hourly wage to support an acceptable standard of living.
Figures for the living wage currently stand at £8.45 across the UK and
£9.75 in London. In the Daily Mail it is referred to using living, hour,
per, and wage, and is associated particularly with Labour, then Labour
leader Ed Miliband, and Conservative MP Boris Johnson, who was
Mayor of London at the time. The living wage is reported on extensively
by both the Daily Mail and the Guardian, and although there is a focus
on London (amplified somewhat by the repetition of ‘in London’ and
151
‘outside London’), there are references to other UK locations. In the

Guardian, hour, living, minimum, pay, wage, campaigns, different, min-
imum, never, response, springing, and up all relate to the Living Wage
Campaign, which is described as moving beyond London with action:
‘springing up in other cities—Glasgow, Leeds, Oxford, Preston, Norwich’.
The Daily Mail PNCs also refer to firefighters ‘who are paid 28,000 a
year plus the 5021 London weighting, [and] claim they must supplement
their income’, and supermarkets who pay low wages which are topped up
by benefits.7 Thus, an increase in wages ‘would reduce the need for in-work
benefits’ and reduce taxpayer costs (25). Although the Daily Mail notes that
‘half of people living in some areas earn less than living wage’, references
to such locations is generic (‘areas’). When specific places are mentioned,
focus is on those areas where the highest percentage of people earn the liv-
ing wage (26). This is similar to the focus on available jobs in Aberdeen
and Cardiff noted in Sect. 6.3. It is also in direct contrast to the Guardian ’s
reporting on the number of people not being paid the living wage (27).
25. Citizens UK is calling for London living wage to be spread to rest of

Britain. It claims rolling out £9.15 an hour wage would reduce ben-
efits by £6billion (DM news, Apr 2015).
26. Just 5.6 per cent of people working in Poplar and Limehouse
(East London) earn less than the living wage followed by 5.8 per
cent in Runnymede and Weybridge (Surrey), 7.3 per cent in South
Cambridgeshire and also 7.3 per cent in Islington South (DM news,
Apr 2014).
27. Fiona Twycross, Labour’s economy spokeswoman in the London
Assembly, said: “Poverty pay is a growing problem in London. In
2007 there were 420,000 people paid below the London Living
Wage (G news: news, Mar 2014).
The strong focus on the living wage in both newspapers suggests a direct
relationship between poverty and minimum hourly pay. The premodifi-
cation of wage by living taps into the measures of poverty that Lansley
7In the Guardian, pay includes references to retail workers, particularly those employed by
Sainsbury’s. Thus, there are parallels between the Guardian and the Daily Mail ’s coverage of retail
wages.
152
and Mack (2015) attempted to capture in their Breadline Britain sur-

veys (see Chapter 1) and are underpinned by a presumption that there
is a relative, society-defined acceptable standard of living. The impli-
cations of the term ‘living wage’ are that those who do not receive a
particular hourly rate of pay are somehow unable to live within UK
society’s accepted norms.
Both newspapers are pointing in the same direction in relation to
the Living Wage Campaign. But they differ in who they present as
the overall beneficiaries; the Daily Mail frames it as positive for tax-
payers and a way to reduce the total welfare bill, whilst the Guardian
labels it ‘A campaign to help the low-paid’ and focuses on the work-
ers who will benefit from higher wages, which they equate to a higher
standard of living. Although the Guardian does note that ‘Employers
in Brent could save as much as 5,000 if they pay all workers at least
9.15’. The ideology underpinning the Living Wage Campaign is that
all workers, independent of the type of job they do, should have access
to a minimum standard of living and that such access can be achieved
through setting minimum hourly rates of pay. Furthermore, the exist-
ing minimum wage is considered too low (the Guardian claims that
the national minimum wage is ‘a poverty wage in London’) and has
not kept up with the cost of living. This means that wages below the
living wage threshold(s) are conceptualised as directly contributing to
poverty.
Concordance analysis highlights that a concept closely related to the
living wage is the cost of living (in London). In the Guardian, terms like
rent and report relate to increasing London rents, benefits recipients pay-
ing a percentage of market rents, and people being evicted because of
the benefits cap. High rent is also directly to blame for poverty: ‘There
are families in London with a high level of rent, who just do not have
enough money left’. The Daily Mail relates rent to the benefits cap
(see Sect. 7.3), evictions, and people paying market rents, and reports
extreme cases such as (28). The construction of (28) as ‘worrying’ is eval-
uative, but the agent who performs the action of worrying is obscured.
Furthermore, by using ‘worrying’ as a premodifier not as a subject com-
plement, the argument that the case is worrying is difficult to dispute.
153
28. Last Sunday, this newspaper reported the worrying case of a family

living in a £2million Kensington townhouse, the £8000-a-month rent
being paid directly though housing benefit (DM news, July 2010).
29. Families in well-heeled suburbs have been warned to brace them-
selves for an influx of thousands of benefits claimants (DM news,
Oct 2010).
Words like claimants and will also refer to the benefits cap as well as
the living wage, and the Daily Mail reports that ‘Experts’ claim ‘cuts
“will force 200,000 benefits claimants out of London and into the
suburbs”’. However, the Daily Mail ’s stance on this issue does not
relate to the difficulties that families may face if they need to relocate,
but rather frames the issue as in (29). Here we see that the problem is
not that families may have to move because they cannot afford to live
in London, and/or are experiencing financial hardship, but rather the
families themselves are the problem for encroaching on ‘well-heeled
suburbs’ where it is implied they do not belong. The framing of the
Daily Mail ’s report as a warning for people of relatively higher soci-
oeconomic status to ‘brace themselves’ presents benefits claimants as
undesirable, and the use of ‘influx’ draws on the negative semantics
of water metaphors identified in discourses of immigration by Baker
et al. (2008).
Migrants are mentioned explicitly in relation to low-pay and a higher
welfare bill (the family in (28) are labelled as ‘former asylum seekers’).
Repeated references to a report that ‘the 20 per cent lowest paid [in
London] had seen wages fall by 15 per cent on average’ are immediately
followed by one of the sentences in (30), published across two different
years.
30a. It added that schools in some areas of the country had seen a
‘marked increase in the number of migrant pupils’ (DM news,
July 2014).
30b. Earlier this month, official figures showed that the number of
Romanians and Bulgarians working in the UK had increased by a
third in a single year to 173,000 (DM news, May 2015).
154
30c. Driving down the wage of natives pushes up the overall welfare bill
since they are likely to be paid more in tax credits and other hand-
outs to compensate (DM news, Nov 2014).
In each of these examples, a causal link is made between low wages and
immigration (more specifically migrants, particularly from Bulgaria
and Romania). In (a) and (b) this link is implied through reference
to migrant children (or presumably their parents) and migrant work-
ers, whereas in (c) the link is implied through the use of ‘natives’. The
‘natives’ are not positioned as ‘driving down’ their own wages, but rather
an unnamed force—contextually understood to be migrant workers—is
performing this action. No other potential causes for declining wages
are considered. Migrants are blamed not only for low wages, but their
presence ‘pushes up the overall welfare bill’ because ‘natives’ must ‘com-
pensate’ using ‘tax credits and other handouts’. Thus, clear links are
made here between poverty and its causes and the immigration of for-
eign nationals into the UK. This link represents endorsement of the
Daily Mail ’s generally-anti-immigration stance.
Outside London, the Daily Mail PNC keywords are more difficult
to cluster. Beggar, awkwardly, feminist, and t-shirt are associated with a
campaign where politicians were asked to wear a t-shirt with the logo
‘This is what a feminist looks like’ and an occasion where Ed Miliband
gave money to a beggar while at the Labour Party conference. Failed
refers to repossession after defaulting on mortgage payments. She and
her link to stories about individual women and the benefits they claim:
White Dee (who appeared in Channel 4’s Benefits Street ) is mentioned,
as is someone in Wythenshawe where ‘Shameless, the Channel 4 com-
edy about feckless families living on benefits’ was filmed. There are also
mentions of individual women committing benefit fraud, shoplifting,
choosing a future university, and a woman ‘suing the Government’
because she was given a voluntary work placement. This focus on stories
about particular women again supports the notion that the Daily Mail
is more likely to associate poverty with individual circumstances than
the Guardian, with further implications that women behaving badly
(shoplifting, etc.) are bringing poverty upon themselves.
155
There are also a cluster of PNC keywords below the threshold of

5 occurrences which refer to payday loans (biggest, hotspot, search,
showed, Wonga.com ), with a particular focus on Altringham in Greater
Manchester (31). Although loan itself was more likely to refer to crime,
particularly as a catalyst for armed robbery. The Guardian too focuses
on payday loans (32). Both newspapers take the stance that such loans
are negative and neither single out any individual as responsible when
discussing this topic.
31. Google’s data showed Altrincham was the biggest hotspot for the
search terms ‘payday loans’ and ‘Wonga.com’ (DM news, July 2014).
32. In Plymouth alone, the council estimates 5000 people are using
payday lenders (G news: news, July 2013).
In the Guardian, city refers specifically to Birmingham and its his-

tory of moving from ‘a high-wage, low-unemployment city to a low-
wage, high-unemployment city’ between the 1970s and 1980s. James
Turner Street (Benefits Street ) is also mentioned, as are reported cuts
at Birmingham city council, and reports of compensation for under-
paid female staff. In the case of the female staff, however, the story was
not presented as individual narratives, but rather reported on wider
action by the council to deal with complaints about low pay from a
large group of women. Council (and county ) locate discourses of pov-
erty in space, with references to Birmingham city council, pay cuts by
Southampton city council, the prosecution of Torbay council for not
paying the minimum wage, and Shropshire county council’s employ-
ment practices (33).
33. A Conservative-led council has sent letters of dismissal to its

entire workforce, telling them they will be re-hired the next day
only if they agree to a pay cut. Shropshire county council gave its
6500 staff notice of their dismissal on 30 September, but offered
them immediate re-employment if they accepted a 5.4% pay
cut as well as changes to their sick-pay arrangements (G news:
society, July 2007).
156
Councils are thus portrayed by the Guardian as the agents behind

the causes of poverty (low wages, the gender pay gap, and job cuts),
although in (33) an overarching social actor is implied through refer-
ences to the council being ‘Conservative-led’. The Guardian ’s focus on
public sector cuts is again evidence of their opposition to Conservative
and coalition-led policies to cut council budgets under the guise of aus-
terity. Thus, the shrinking of the state is the implied cause of poverty
and, as such, the ability to decrease poverty is not in the hands of the
individuals who are experiencing it.
The locations mentioned above are contrasted with more positively-
slanted stories such as the creation of jobs involving Corby council
and Essex county council, and the City of York council’s pledge to pay
the living wage. Recording, strong, growth, increases, and respectively are
PNC keywords because they occur in repeated Guardian reports about
increases in disposable income in Hull, Sheffield, Bradford, Shropshire,
Staffordshire, Herefordshire, Worcestershire, and Warwickshire, which
are compared to decreases in Manchester and York. Similarly average
refers to the lowest average wage (34) and is linked to four locations in
the north of England.8 References to the north-west are also linked to
transport and the concept of a ‘rust belt’ of traditionally industrial loca-
tions (35) which is somewhat cut off from the rest of the UK.
34. The four constituencies with the lowest average wage are in north-
west England; those working in Blackley and Broughton, Preston
and Middlesbrough earn an average salary of between £323 and
£330 per week (G news: society, June 2012).
35. Meanwhile, public investment in job-creating transport is needed
by the workless of the north-western “rust belt ” to access better-paid
jobs in Preston, Manchester, Liverpool and Leeds (G news: politics,
July 2010).
The focus on the lowest paid, presented as weekly wages (34) as

opposed to annual salary, complements the Guardian ’s focus on the
8The reference to Middlesbrough in (34) is misleading as it is in the north-east, not the north-
west as stated.
157
highest jobseeker to vacancy ratios (Sect. 6.3). It is another example of

contrast between the Guardian and the Daily Mail and evidence that
the Guardian presents a bleaker view of poverty in the UK by focusing
on the lowest paid areas, whilst the Daily Mail foregrounds more pos-
itive statistics and, by extension, downplays links between poverty and
place.
6.5 Summary
Whilst the analysis of the single query <*poverty*> in Chapter 4
demonstrated the core concepts of geoparsing and GTA and illustrated
the nuanced differences in how the Guardian and Daily Mail use a par-
ticular lemma, this chapter has begun to show that analyses of individ-
ual terms is only the starting point for discourse-level GTA. Whilst a
focus on unemployment facilitated a three-way comparison between the
Guardian, Daily Mail, and official statistics, it is clear that poverty dis-
courses are made up of more than individual terms or concepts which
are directly measurable.
By taking a closer look at how subsets of our queries are used by each
newspaper we have demonstrated that although the Guardian and the
Daily Mail (perhaps unsurprisingly) report on similar topics, their use
of the query terms and PNC keywords differs. The Daily Mail reports
on job losses, focusing particularly on locations associated with manu-
facturing (particularly the automotive industry). But it also foregrounds
references to locations where jobs are available. The Guardian reports
on those places where the jobseeker to vacancy ratio is particularly high,
and despite the PNCs clustering in London, unemployment is not pre-
sented as a particular problem restricted to the capital. Job losses are
reported across England (but less so in Scotland, Wales and Northern
Ireland), with a particular focus on public sector redundancies. By ana-
lysing the language choices of each newspaper, and the stances that
such choices indicate, we can begin to detail the Daily Mail and the
Guardian ’s orientation to wider discourses of poverty and place.
In the Money subset, the dominance of coverage of the Living Wage
Campaign suggested that it is a key issue within discourses of poverty.
158
But each newspaper used this campaign in slightly different ways. In

the Guardian it was seen as a way to tackle poverty and low pay and
increase workers’ access to a minimum standard of living, whereas in the
Daily Mail it was seen as a way to decrease the amount of money paid
in in-work benefits and reduce the burden on taxpayers (although there
were no explicit links to tax cuts). In both approaches, there are links to
welfare payments and the identities of benefits recipients. We consider the
Benefits subset, and the related Housing subset, in the following chapter.
References
273–306.
University Press.
University Press.
London: Oneworld.
ONS. 2014. 2011 Census Glossary of Terms. https://www.ons.gov.uk/ons/
guide-method/census/2011/census-data/2011-census-data/2011-first-
release/2011-census-definitions/2011-census-glossary.pdf. Accessed 24/2/
2017.
ONS. 2016a. Unemployment Rate: Sourced from the Labour Market Statistics
Time Series Dataset. https://www.ons.gov.uk/employmentandlabourmarket/
peoplenotinwork/unemployment/timeseries/mgsx/lms. Accessed 2/12/2016.
ONS. 2016b. Statistical Bulletin: UK Labour Market: Nov 2016. https://
www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employ-
mentandemployeetypes/bulletins/uklabourmarket/november2016. Accessed
2/12/2016.
7
Characterising Poverty in Place:
Benefits Receipt in Britain
The previous chapter explored two of our four major poverty-related

subsets: terms related to Employment and Money. This chapter focuses
on the other two major subsets: Benefits and Housing. Section 7.1 pro-
vides an overview of the wider context of benefits receipt in the UK.
This is followed by analysis of PNCs relating to the Benefits subset in
the Daily Mail (Sect. 7.2) and the Guardian (Sect. 7.3). As shown in
Chapter 5, references to benefits and the cost of housing are core com-
ponents of discourses of poverty and place. Whilst the former may have
been expected, given the apparent relationship between poverty and the
welfare state, the centrality of housing and rent was not necessarily pre-
dicted. It was somewhat expected that links would be made between
poverty and homelessness, but, as shown in Sect. 7.4, the relationship
between poverty and housing more often concerns the general cost of
living and high household costs. Section 7.5 summarises how the dif-
ferent subsets analysed in Chapter 6 and this chapter relate statistically
to each other and to quantitative measures of poverty including the
Carstairs, worklessness, and unemployment data introduced in earlier
chapters.

https://doi.org/10.1007/978-3-319-93503-4_7
160
7.1 The Social, Political, and Linguistic Context

of Benefits Receipt
In line with the analysis of employment in Chapter 6, it is important
to note the wider context of our corpora. In 2010—the start of the
period covered by our corpora—the total number of people claiming
out of work benefits was estimated to be 1,496,400, but by 2015 this
figure had dropped to 797,800 (ONS 2017). Conversely, there were
4,651,106 Housing Benefits claimants in January 2010, which rose to
4,894,992 in January 2015 (DWP 2018). Official statistics from the
Family Resources Survey 2013/14 show that 55% of ‘UK working age
adults who are in receipt of state support’ are in some form of employ-
ment (DWP 2016). Discussing trends in poverty from the 1980s to the
present day, Lansley and Mack (2015: 81) note that ‘the most striking
shift has been the rise in poverty among those in work ’ and claim that
those experiencing child poverty are likely to ‘have at least one person in
the household in full-time work’ (2015: 87).
Research on the portrayal of benefits receipt by the media shows the
topic is controversial. Multiple arms of the mass media make associa-
tions between benefits receipt and fraud, poor work ethic, scrounging,
and a money for nothing attitude. For example, Lundström (2013)
found evidence of anti-benefits sentiment in the UK press, whilst
Paterson et al. (2016) showed how members of the UK public used
the television programme Benefits Street to make negative generalisa-
tions about people who receive government benefits. Similar sentiments
were also found by Baker and McEnery (2015) and van der Bom et al.
(2018) in their analyses of the discourses drawn upon by people tweet-
ing about Benefits Street. Importantly, the representations of benefits
receipt presented in the mass media (as seen in programmes such as
Benefits Street, On Benefits and Proud, and Benefits Britain ) are used to
justify government actions that potentially impact all benefits recipients.
The example below is taken from the Hansard record and includes an
exchange in the House of Commons between Conservative MP Phillip
Davies and Secretary of State Iain Duncan Smith:
7 Characterising Poverty in Place: Benefits Receipt in Britain
161
Philip Davies (Shipley) (Con): Has the Secretary of State managed to

watch programmes such as “Benefits Street” and “On Benefits & Proud”?
If so, has he, like me, been struck by the number of people on them who
manage to combine complaining about welfare reform with being able
to afford to buy copious amounts of cigarettes, have lots of tattoos, and
watch Sky TV on the obligatory widescreen television? Does he under-
stand the concerns and irritation of many people who go to work every
day and pay their taxes but cannot afford those kinds of luxuries?
Mr. Duncan Smith: My hon. Friend is right: many people are shocked
by what they see. That is why the public back our welfare reform package,
which will get more people back to work and end these abuses. All these
abuses date back to the last Government, who had massive spending and
trapped people in benefit dependency. (Hansard 2014)
The position set out in this exchange is that benefits recipients illegit-
imately spend their money on tattoos and televisions, which causes
‘irritation’ for working people who do not have the money to buy
such items. Direct contrast is made between benefits recipients and
hard-working, taxpayers who are ‘shocked’ by Benefits Street, despite
the fact that many benefits recipients are in work. The language used is
highly evaluative; their (presumably expensive) televisions are described
as ‘obligatory’ and their consumption of cigarettes is ‘copious’. It is ulti-
mately implied that those receiving benefits should not be able to afford
such ‘luxuries’. The apparently flawed consumerism of benefits recipi-
ents, along with their ‘complaining’, are described as ‘abuses’ of a system
which will be fixed by Conservative-led Welfare Reform. This is, thus, a
clear example of how media depictions of benefits receipt (and poverty
more widely) can be used as supporting evidence for government policy.
Having established the socio-political background within which
the texts in our corpora were produced, we can also take an overview
of how the Guardian and the Daily Mail discuss government benefits
more generally. As a litmus test <*benefit*> returns 27,159 hits in the
Guardian (481.51 pmw) and 61,165 hits in the Daily Mail (188.38
pmw), showing that the former is 2.5 times more likely to use the term.
Of course, not all references to benefits will correspond to government
welfare payments; the different forms of benefit can be used much
162
more generally, as in ‘reap the benefits’, ‘feel the benefit’ and to ‘ben-
efit from’. Nevertheless, the collocates of <*benefit*> suggest that it is
predominantly used in reference to welfare payments. Table 7.1 lists the
top 25 lemmatised collocates of <*benefit*> in both corpora (including
the debate sub-corpora) calculated using log-likelihood, a span of ±5,
minimum frequency of 5 and a minimum collocation frequency of 5.
The Daily Mail collocates include references to particular types of bene-
fits (child, tax, incapacity, disability, housing, credit, unemployment, pension,
allowance, out-of-work ) but also include cheat and fraud (see example 1).
The Guardian, whilst also collocating <*benefit*> with particular types of
payment (housing, incapacity, tax, child, credit, disability, universal, out-of-
work, in-work, allowance, sickness, working-age ), does not collocate <*bene-
fit*> as highly with fraud. Fraud and cheat occur in the Guardian collocates,
but are ranked in 28th and 30th place. In contrast to the Daily Mail, the
top Guardian collocates also include references to sanction (2), which occurs
in 496th place in the Daily Mail ’s collocate list. The Guardian also makes
more distinctions between the employment status of benefits recipients
(out-of-work, unemployment, in-work, working-age ) than the Daily Mail,
which includes only unemployment and out-of-work in its top collocates.
1. During the recession, benefit fraud boomed—with the amount paid

to cheats up £300 million (DM news, Feb 2010).
2. A whistleblower said staff at his jobcentre were given targets of three
people a week to refer for sanctions, where benefits are removed for
up to six months (G news: politics, Apr 2011).
The different lenses of the two newspapers are also evidenced by their
number one collocates, with the Daily Mail most closely associating
<*benefit*> with claim. Analysing a random sample of 50 concordance
lines, the collocation of claim and <*benefit*> tends to refer to changes
in the rules for claiming benefits, fraudulent claims and/or prosecu-
tion, the amount of benefits individuals have claimed, the number of
people claiming benefits, and the television programme Benefits Street.1
1For analysis of public attitudes to Benefits Street, see Paterson et al. (2016, 2017), and van der
Bom et al. (2018).

Table 7.1 Top 25 collocates of <*benefit*>
Daily Mail Collocate No. of LL value Guardian Collocate No. of LL value
freq. texts freq. texts
1. Claim 5833 2971 17091.46 1. Cut 2022 1405 5997.733
2. Tax 3118 1847 12337.39 2. Cap 1010 547 5930.261
3. Child 5259 2284 11276.97 3. House 1719 983 5671.677
4. Cheat 1694 1002 10755.33 4. Housing 1137 694 5021.859
5. Incapacity 836 418 9829.756 5. Incapacity 483 287 4968.502
6. Cut 2364 1514 7592.809 6. Claimant 670 518 4109.355
7. Cap 1263 669 7497.109 7. Tax 1511 1027 4021.37
8. From 8131 6449 7354.443 8. Child 1835 927 3765.821
9. Disability 1070 625 7029.107 9. From 3249 2726 3219.434
10. Claimant 826 616 6425.181 10. Credit 688 519 2659.03
11. Housing 1159 723 6325.772 11. Disability 524 363 2395.079
12. System 1716 1210 5477.948 12. Claim 1065 789 2317.219
13. Fraud 1126 614 5395.454 13. System 879 681 1888.258
14. Receive 1783 1298 4628.842 14. Payment 455 372 1773.977
15. Credit 1083 684 4471.915 15. Welfare 560 483 1686.019
16. Unemployment 773 501 4451.995 16. Universal 373 271 1674.223
17. Benefit 1294 521 4021.045 17. Out-of-work 168 140 1613.822
18. Payment 839 710 3489.682 18. Unemployment 391 278 1571.237
19. Welfare 797 652 3346.398 19. Receive 577 458 1466.983
20. House 2205 1279 3314.952 20. Sanction 315 207 1339.746
21. Pension 896 692 3313.436 21. In-work 157 113 1301.04
22. Allowance 611 470 3308.073 22. Allowance 334 297 1284.783
23. Pay 1919 1452 3253.795 23. Sickness 161 116 1122.974
24. Will 4102 2916 3144.391 24. People 1675 1274 1118.954
25. Out-of-work 287 211 3064.144 25. Working-age 141 105 1043.226
163
164
In contrast, the Guardian ’s top collocate is cut and an analysis of a ran-

dom sample of 50 concordance lines indicates a focus on Housing
Benefits, the coalition government’s actions to decrease the amount of
benefits paid to individuals (and thus the total welfare bill), and the con-
trast between tax cuts for high earners and benefits cuts for the poor. The
opposite analysis (cut in the Daily Mail and claim in the Guardian ) shows
that the former focuses on Housing Benefit, cuts to child benefit for peo-
ple earning over £60,000 (which it notes has strong public support), cuts
in Greece and America, and the reported speech of MPs such as Phillip
Hammond (then Defence Secretary) and Nick Clegg (then Deputy
Prime Minister), see (3). The Guardian ’s use of claim refutes arguments
that EU nationals arrive in the UK en masse only to claim benefits (4),
and that claiming benefits is a lifestyle choice (5). It focuses on Housing
Benefit and notes benefit fraud, but to a lesser extent than the Daily Mail.
3. Deputy Prime Minister Nick Clegg said the decision to cut child
benefit was ‘excruciatingly difficult’ Nick Clegg, the Deputy Prime
Minister, suggested yesterday that many of those losing child benefit
do not see themselves as well off (DM news, Oct 2012).
4. The European commission (sic) is also growing increasingly frus-
trated with the UK government for claiming that “benefit tourism”
is a massive problem while not supplying any verifiable data to sup-
port the claim (G news: news, Oct 2013).
5. Osborne will be called to the Commons to answer an urgent ques-
tion from Bob Russell, a Liberal Democrat who objected to Osborne’s
assertion that he would reduce the number of people who claim wel-
fare benefits as a “lifestyle choice” (G news: news, Sept 2010).
However, focusing just on <*benefits*> may be somewhat misleading, as

the Daily Mail also systematically refers to welfare payments as ‘handouts’.
There are 326 hits of <*handout*> in the Guardian (5.78 pmw) and 3132
(9.65 pmw) in the Daily Mail. Its top collocates are food, photograph, state,
cash and relying in the Guardian and state, welfare, photo, undated, and
claiming in the Daily Mail, suggesting that both newspapers use the term
to refer to benefits, but that the Daily Mail does this proportionally more
often. Thus, as discussed in Chapter 5, a single search term cannot tell us
the whole story. To this end, we turn to the Benefits subset (cf. Table 5.7).
165
7.2 Benefits and Geography in the Daily Mail

The Benefits subset contains 22 queries and includes terminology relat-
ing to benefits receipt and the names of different of government benefits
(excluding pensions): <*allowance*>, <*benefit*>, <*claim*>, <credit*>,
<cut*>, <*disab*>, <DLA>, <dole*>, <Duncan Smith*>, <ESA*>, <fit for
work>, <(fitness to work|fitness-to-work)>, <handout*>, <IDS*>, <JSA>,
<personal independence payment>, <PIP>, <*sanction*>, <*subsidi*>,
<*tax*>, <*welfare*>, and <(Work capability assessment*|WCA*| Work-
capability-assessment)>. Despite containing the most queries, the Benefits
subset returned only the fourth-highest number of PNCs for any subset
in the Daily Mail (446), and the second-highest in the Guardian (1397)
(cf. Table 5.7). Given that the Guardian corpus was much smaller than
the Daily Mail corpus, we can deduce that the Guardian discusses bene-
fits in relation to place much more often than the Daily Mail. There are
322 different locations in the Guardian data and 164 in the Daily Mail.
Figure 7.1 shows the density smoothed pattern and the Kulldorf map for
the Daily Mail, whilst Fig. 7.2 (Sect. 7.3) shows the corresponding maps
for the Guardian. The spatial pattern for the Daily Mail is quite similar
to that for the Money subset (Sect. 6.4), in that it has clusters in London,
Birmingham, north-west England and Glasgow. The Kulldorf map empha-
sises these clusters for the Benefits subset more than for the Money subset.
To interrogate the different geographical distributions of the Benefits
subset, we again began with the PNC keywords and conducted con-
trastive concordance analysis combined with close reading of a random
sample of concordance lines. The PNC keywords have been split into
London and the Rest of the UK to maintain consistency, but this distinc-
tion is rather artificial for the Guardian in particular (Sect. 7.3). Table 7.2
shows the PNC keywords for the Benefits subset in the Daily Mail.
Cuts refers to the government attempting to ‘balance the books’ and
co-occurs with payments and credits (in particular working tax credits ).
There are also references to the Welfare Reform Bill (2012), and support
relates to the amount of (financial) services that workless families can claim.
The Daily Mail contrasts speeches made by Conservative George Osbourne
(who was Chancellor of the Exchequer) and Labour’s Ed Miliband (then
party leader), reporting on Osbourne’s aim to restrict the amount that peo-
ple can claim and Milliband’s ‘pledge’ to increase tax credits.
166
Fig. 7.1 Density smoothed and Kulldorf maps of PNCs in the Benefits subset in the Daily Mail
167
Table 7.2 PNC Keywords in the Benefits subset in the Daily Mail (n ≥ 5)

London <0.01 cuts, credits, housing, to, a, payments
London <0.05 welfare, support, an
Rest of UK <0.01 cent, per, residents, of, claiming, town, abolish, receive,
his, one, drug, country
Rest of UK <0.05 incapacity, existing, number, suggest, show, new, also,
left, tests, near, office, numbers, on, job, life, stopped,
figures, biggest, ‘on, ‘we, alcohol, around, showed,
lived, future, only, highest, him, market, car, single,
pilot, problems, third, capital, trial, addictions, came,
should
However, there is little variation in the topics associated with the

London PNCs; they are dominated by references to Housing Benefit
and the affordability of London. Housing refers almost exclusively
to Housing Benefit and the amount of this benefit that people living
in London are claiming. Whilst the Daily Mail notes that the capi-
tal is unaffordable for many (especially in terms of rent), it focuses on
extreme examples of people claiming large amounts of Housing Benefit,
and/or being able to live in expensive properties due to Housing
Benefit.
6. Nasra Warsame and seven of her children were living in a 1.8 million
house in Westminster at a cost to taxpayers of £1600 a week. Her
husband Bashir Aden and her eighth child were living in an ‘over-
spill’ property, also on housing benefit (DM news, Jan 2010).
Although example (6) was published in an article in 2010, making

it eligible for inclusion in our corpus, it is heavily based on an article
published in November 2009, which was headlined ‘Taxpayers pay
£1600-a-week for family of ex-asylum seekers to live in luxury five-sto-
rey home’ (Ballinger 2009). This is just one example of the Daily Mail
recycling its content and repeating sensational stories about benefits
claimants. Whilst the Daily Mail never takes an explicit stance on the
family discussed within the article or their receipt of benefits, close anal-
ysis indicates a particular preferred reading of the text. The agents in the
168
headline are the taxpayers (perhaps short hand for the Daily Mail ’s tar-
get readership and thus, by extension, an example of synthetic personal-
isation) who are positioned as providing for a family in need of Housing
Benefit. However, there is a sense that taxpayers actually have no say
in how taxes are spent, that £1600 is too much to pay in benefits (it is
sensational enough to warrant a headline), and that the family’s ‘ex-asy-
lum seeker’ status is somehow significant to their receipt of Housing
Benefits. Their case is presented as newsworthy through the foreground-
ing of amounts of money that would be unaffordable for many (£1600
a week in rent, 1.8 million house, annual rent of £83,200).
The number of children the couple have is made central to the story,
as the family were moved from a previous property, which the coun-
cil deemed too small. It is also noted that the family is split between
two houses, both paid for by Housing Benefit. They are repeatedly
referred to as ‘ex-asylum seekers’, ‘A Somali family’, and ‘the former
asylum seeker’, who ‘fled unrest in Somalia in 1991 and claimed asy-
lum in Britain’, although they have received citizenship and ‘all of their
children were born here’. As the family are UK citizens, there is no
direct relationship between their route to citizenship and their claiming
Housing Benefits, yet their Somali heritage is repeatedly referred to in
order to link their benefits claim to a wider context. A context which
is further illuminated towards the end of the article, when a seemingly
unrelated story is also reported (7).
7. Last year it was revealed Afghan single mother Toorpakai Saiedi and
her seven children were given a £1.2 million property complete with
100 ft. garden by Ealing council in West London. Mrs Saiedi, 35,
received £170,000 a year in benefits. Some £150,000 of that is paid
to a private landlord for the seven-bedroom house (Ballinger 2009).
Here an unrelated example, discussing a different family, from a dif-

ferent country, living in a different area of London, is included in the
story about the family in central London. No contextual information is
provided as to how these two examples may be linked; it is left for the
reader to fill in the blanks. A correlation between asylum seekers and
large claims for Housing Benefit is therefore implied, and furthermore,
169
the article draws upon the wider discourse that asylum seekers are ‘ben-
efits tourists’ taking advantage of the UK system (a position opposed
by the Guardian ). We see many of the same techniques used to
describe the woman in (7) as were used to discuss the Somali family:
she is labelled an ‘Afghan’, her number of children is foregrounded, as
are large sums of money. Furthermore, she was reportedly ‘given’ an
expensive house. This is factually inaccurate as she does not own the
house, rather she has been provided with Housing Benefit to pay rent.
Nevertheless, there is an underlying implication that she is somehow
not deserving of the home she has.
To situate the original article within wider discourses of poverty,
descriptions of the Somali family’s house paint it as extremely desira-
ble: it is ‘luxury’, ‘fully-furnished’, ‘within walking distance of the West
End’; it has ‘two leather sofas, a flat screen television and a glass coffee
table’ and ‘a large glass sculpture situated in the middle of a courtyard’.
Similarly the property referred to in (7) was ‘complete with [a] 100ft
garden’. These descriptions draw on discourses of the deserving and
undeserving poor, with the implication that, if someone is poor enough
to need Housing Benefit, they should not be able to afford to live in
such a property and, significantly for discourses of poverty and place,
in such a prime location. But what neither of these examples consider is
that the families mentioned do not represent average benefits claimants,
rather they are at the extreme end of the spectrum for benefits receipt.
Yet these are the examples that the Daily Mail presents to its readers
to justify its wider stance on welfare reform and the endorsement of
Conservative-led policies.
Whilst only one article, the naming of particular (individual) bene-
fits recipients who are unrepresentative of wider benefits claimants is a
tactic which has been noted by Lundström (2013) as typical of reports
on benefits recipients in the tabloid press (and in the Daily Mail in par-
ticular). Other extreme examples include (8), which is underpinned by
the notion that people receiving benefits should not reside in ‘exclusive’
areas alongside celebrities. The notion of an ‘exclusive’ area is of great
relevance to discourses of poverty and place as it presupposes that one’s
economic standing can grant access to particular locations which are (or
should be) off limits to those who have less (8).
170
8. Behnaz and Emmanuel Scotts received £2670 a month in housing

benefit, which helped to pay for their four-bed apartment in the
exclusive Chelsea Harbour. The couple received jobseekers’ allow-
ance of £64.30 a week each—which Hammersmith and Fulham
Council said meant it was forced to pay them housing benefit. They
were able to live in the same blocks as singer Robbie Williams and
sir (sic) Michael Caine (DM news, Mar 2012).
9. London councils have acquired rental properties in Luton,
Northampton, Broxbourne, Gravesend, Dartford, Slough, Windsor,
Margate, Hastings, Epping Forest, Thurrock and Basildon in prepa-
ration for welfare cuts, the Guardian has found (DM news, Nov
2012).
Ideologies concerning who should (not) live in a particular place are

linked to multiple locations: links are made between government wel-
fare policies and councils preparing to move their tenants to areas with
cheaper rent (9). Thus, the relationship between welfare cuts and geog-
raphy extends beyond London. However, whilst the London PNC key-
words tended to cluster around welfare reform and Housing Benefit,
elsewhere the picture was more mixed.
Abolish refers to campaigns against the so-called Bedroom Tax (a term
used for the coalition government’s Removal of Spare Room Subsidy
policy which decreased the amount of Housing Benefits available to
people seen to be under occupying their properties—see Sect. 1.2). In
particular the bedroom tax appears to be associated with Scotland, but
this is due to the multiple PNCs generated by repeated reports, includ-
ing (10), which comes from a speech made by Ed Miliband to Scottish
voters before the Scottish Independence Referendum in 2014.
10.
He will claim: ‘We will abolish the bedroom tax. Abolish it
in Dundee, in Glasgow, but also in Carlisle, Newcastle and
Nottingham’ (DM news, Sept 2014).
This example demonstrates the importance of considering PNCs in

their wider context. The Daily Mail is not making explicit links here
between the bedroom tax and these particular locations, rather they are
reporting the words of a third party who does make these connections.
171
Where the Daily Mail does take an explicit stance is in the report-
ing the number of people receiving benefits. Per cent, existing, incapacity,
number, tests, third, trial, and residents refer to the jobless rates in differ-
ent areas, the number of people receiving benefits, and the outcomes
of fitness-to-work (re)assessments. The Daily Mail shows a tendency to
list the ‘best’ and the ‘worst’ areas for benefits receipt or worklessness
(11–14). Notably, the places with the highest rates are in the north of
England, whilst the areas with fewer claimants are all relatively afflu-
ent areas in the south of England (see 11). Additionally, the Daily Mail
claims that towns ‘claiming the most sickness benefit are in Wales and
County Durham’ in contrast with places like Uttoxeter (14) where
employment rates are high.
11. The news was particularly grim for young people in the north,
with 10 per cent of 16–24-year-olds in Hull, Grimbsy and
Middlesbrough on the Jobseekers Allowance. This compared with
less than two per cent in Cambridge and Oxford and less than four
per cent in Reading, Southampton and Bournemouth (DM news,
Jan 2010).
12. Manchester has 71,000 who have never been in work and
Liverpool has 52,000. Bradford is next with 50,000, then Sheffield
on 46,000, Tower Hamlets has 42,000 and Cardiff 41,000.
Taxpayers’ Alliance boss Matthew Sinclair told The Sun: ‘These are
truly shocking figures which underline the importance of reforming
the welfare system in order to make work pay once and for all (DM
news, Feb 2013).
13. The UK’s welfare capital is the Isle of Wight but, when payments to
the elderly are excluded, Liverpool is the biggest claimer of benefits
(DM news, Sept 2013).
14. A record surge in employment means in one town almost no-one is
on the dole. In the Staffordshire market town of Uttoxeter just 34
people are claiming Jobseekers (DM news, July 2014).
To take another extreme example, the Daily Mail claims that 99.1% of
working age people on the Cottsmeadow Estate in Birmingham (15)
‘live off state handouts’, with only one person in a full-time job, whilst
the rest are described as in (16).
172
15. The Cottsmeadow Estate is one of five communities where more

than 90 per cent of adults are on some form of benefits. The
Egerton Street area of Oldham, where 110 out of 113 (or 97.3 per
cent ) of residents claim state benefits, was second on the list (DM
news, Jan 2011).
16. Figures from the Department for Work and Pensions reveal that
among residents, 50 are listed as on the dole, 30 are signed off sick
and get incapacity benefits, 20 pocket lone-parent payments and
five claim other handouts (DM news, Jan 2011).
The choice of verbs in (16) is particularly illuminating in terms of the

Daily Mail ’s evaluation of benefits recipients; the negatively-loaded
‘pocket’, which could have implications of criminality, and the
non-standard form ‘on the dole’ alongside the use of ‘handouts’ (dis-
cussed above) are used within the scope of something that the Daily
Mail (via the DWP) ‘reveal’. Furthermore, the article is framed under
the headline ‘Britain’s most workshy estate’ despite the fact that, as
stated in the article, many of those receiving benefits were lone parents
caring for children (and thus doing unpaid work), and people who were
too sick to work. The percentage number of claimants is ‘staggering’ and
the statistics are presented as evidence which will ‘heap further pres-
sure on ministers to stamp out the something-for-nothing culture that
spiralled under Labour’. Here there is a presupposition that a ‘some-
thing-for-nothing culture’ exists in (particular locations in) Britain, and
a further implication that such a culture is realised by benefits receipt.
The blame for this culture is placed on previous Labour governments;
thus a high number of benefits claimants is seen as a sign of an overgen-
erous state.
The so-called workshy and the Labour party are not the only groups
the Daily Mail holds accountable for the number of people receiving
benefits. The article that (11) is taken from includes a hyperlink to a
second article headlined ‘Migrants ARE driving down wages of the
poor, as Equalities watchdog blames East European influx’ (Slack 2010).
Thus, whilst there is no blame (other than individual laziness) in the
original article for why there are more people claiming Jobseekers’
Allowance in some areas over others, blame is implied through links
173
to other articles which are hyperlinked from the first. This is another
aspect that needs to be considered when using corpora of online texts.
The Daily Mail makes (implied) causal links between immigration and
poverty using hyperlinks. A corpus analysis would not normally focus
on such practices; it is only through the wider interrogation of context
that the significance of hyperlinking can be teased out and examined.
Across the other PNCs, the Daily Mail makes repeated references to
the television programme Benefits Street (the first series of which was
filmed in Birmingham, accounting somewhat for it being a hotspot
on the Kulldorf map, Fig. 7.1). It reports on proposals for geographi-
cally means-tested benefits that are calculated based on the cost of liv-
ing where you live (a policy which did not become part of the Welfare
Reform Act, but can be seen in a limited sense in the different levels of
benefits cap inside and outside London). It also highlights which areas
of the UK have high levels of drug and alcohol dependency: drug, alco-
hol, addictions, highest, lived, and problems all relate to this topic and are
given as reasons for people claiming benefits in particular geographical
locations (17).
17. Bournemouth has the biggest proportion of claimants of ESA for

people with alcohol and drug addictions, with around one in 14 peo-
ple in the Dorset town who are handed the benefit abusing alco-
hol, while one in 20 have drug addiction problems. Other areas with
a high percentage of claimants with drug and alcohol addictions
include Edinburgh, Glasgow, Aberdeen and Bristol. Meanwhile,
the biggest proportion of claimants due to stress is South Hams
in Devon, closely followed by Tamworth in Staffordshire and
Richmond in North Yorkshire (DM news, Mar 2015).
In devoting a report to the reasons for Employment Support Allowance

claims, the Daily Mail groups together those claiming benefits for
addictions, as shown above, and other conditions such as stress, and
obesity. Those claiming ESA for these reasons are described as ‘being
handed £100 a week’ and their claims (and health conditions) are some-
what delegitimised. ESA is described as being ‘awarded to those who
have an illness or disability that affects their ability to work’ but this is
174
contrasted with the claim that ‘Now, according to new figures… 56,000
people with alcohol problems receive ESA, while 36,650 drug addicts
are also claimants’. The use of ‘now’ as a qualifier suggests that alcohol-
ism, drug addiction, obesity, and stress, do not meet the (Daily Mail ’s)
definition of ‘illness or disability’. Thus, whilst there is nothing ille-
gal about claiming ESA for these conditions—and, as such (17) is not
reporting on benefit fraud—the claims are seen as somehow illegitimate
and the Daily Mail is positioned as making moral judgements about
claimants with these issues and conditions.
As part of the article (17) is taken from (Newton 2015), the Daily
Mail provides a map of the ‘booze’ and ‘drug hotspots’ and a table rank-
ing the top ten worst offenders. This ‘name and shame’ approach homog-
enises people in two ways. It groups together all benefits claimants with
particular health issues and treats them as interchangeable, and does the
same for all people living in a particular area. Associating an area with
alcoholism or drug taking contributes to an overall negative character-
isation of that location; not everyone in Bournemouth is accused of
being an alcoholic, but Bournemouth is categorised as a place where
alcoholics live. The Daily Mail does not attempt to offer potential solu-
tions to problems associated with alcoholism (although stopping benefits
is implied) nor does it consider the wider social factors, health profiles,
employment opportunities, etc. which may go some way to explaining
why certain places may be more (or less) affected by alcoholism, etc.
7.3 Benefits and Geography in the Guardian

The density smoothed and Kulldorf maps for the Benefits subset
in the Guardian (Fig. 7.2) show a very different pattern to the same
maps for the Daily Mail (Fig. 7.1). While its Employment and Money
subsets are highly London-centric (see Chapter 6), the pattern for the
Benefits subset is much more dispersed, with clusters around London,
Birmingham, north-west and north-east England, and Glasgow. These
clusters, as well as Aberdeen, are also apparent as hotspots on the
Kulldorf map.
Fig. 7.2 Density smoothed and Kulldorf maps of PNCs in the Benefits subset in the Guardian
175
176
Table 7.3 PNC Keywords in the Benefits subset in the Guardian (n ≥ 5)

London <0.01 cap, housing, will, week, benefit, through, families, not,
total, cuts, many
London <0.05 their, these, borough, coalition, to, that, across, against,
reforms, homes, make, than, impact, so, changes
Rest of UK <0.01 street, city, incapacity, lowest, party, constituency, began,
council, lab, county, service, centre, estate, jobs, near,
Iain, end, visit, testing, addresses, highest, and, former,
all, worst, since, claims, support, jobseeker, welfare,
unrest, go, July, channel, Liz, minimum, starts, existing,
IB, igniting, forces, DWP, increased, photograph, years,
he, huge, staff
Rest of UK <0.05 seats, no-go, bounds, involved, leaders, looking,
trial, chasing, town, economically, mid, son, tests,
office, needed, project, northern, warning, jobcen-
tre, Scotland, national, allowance, system, at, time,
tomorrow, my, losses, away, lost, jobcentres, plans, UK,
its, according, care, health, one, universal, man, posts,
employment, off, authority, must
The PNC keyword benefit (Table 7.3) refers to benefit cuts (particu-
larly to Housing Benefit) and the benefits cap.2 Even though across
occurs as a PNC keyword in the London PNCs, it most often occurs
in the cluster ‘across the country’ (as well as in ‘across London’) and is
used to demonstrate the wide ranging impact of the benefits cap, par-
ticularly in relation to Housing Benefit, the bedroom tax, and high rent
prices. The benefits cap is described as affecting people across England
and Wales but the Guardian also presents it as a local issue which will
disproportionately affect London (18). The cap is ‘forcing people out
of London’, with the implication that homelessness is the only alter-
native, because people cannot ‘keep a roof over their heads’. There are
also reports of private landlords pre-emptively evicting tenants receiving
Housing Benefit to protect against arrears resulting from the cap.
2The benefits cap refers to the limit of state benefits payable to most working age adults.
Although there are exceptions—recipients of particular benefits may be exempt—current figures

stand at £20,000 per year, per couple or single parents, and £13,400 per single childless per-
son. The figures for people living in Greater London are slightly higher, standing at £23,000 and
£15,410 respectively (gov.uk, n.d.).
177
18. The acute housing shortage in the capital means market rents out-
strip benefit cap levels in cheaper outer London boroughs includ-
ing Haringey, Waltham Forest, and Barking and Dagenham (G
news: society, Nov 2012).
19. Ministers were accused last night of deliberately driving poor peo-
ple out of wealthy inner cities as London councils revealed they
were preparing a mass exodus of low-income families from the capi-
tal because of coalition benefit cuts (G news: politics, Oct 2010).
20. Bromley, Croydon, Enfield and Haringey—will have overall ben-
efits capped at £26,000 a year, with the scheme becoming nation-
wide in December. Opponents have warned that it could cause a
mass movement of poorer families out of parts of London or other
expensive areas, disrupting children’s education (G news: society,
Mar 2013).
Changes also refers to Housing Benefit and a ‘mass exodus’ of families

from inner to outer London, for which the coalition government are
held responsible (19). The Guardian focuses on the potential impacts
of the benefits cap on families, as opposed to claimants or recipients. This
naming technique, which is reinforced by references to the number of
school children who may have to move (20), presents benefits recipients
as part of wider family and social networks.
Against is used to report on protests opposing cuts to disability pay-
ments in particular (and austerity more widely). The Guardian reported
on these protests during their organisation—‘Thousands of people
will demonstrate in Westminster on Wednesday against cuts’—and
after the event. The choice to report on the planned peaceful protest
(labelled as ‘civil disobedience’) serves a dual purpose; it is a report of a
planned political event, but it also acts to publicise the planned protest
to its readership, who presumably share the Guardian ’s anti-cuts (and
anti-austerity) stance. Furthermore, by foregrounding protests opposing
disability cuts, the Guardian highlights government cuts to services for
those who are presented as most vulnerable (21). Thus, the amount of
sympathy for benefits recipients can be maximised. Such reports con-
trast with the Daily Mail ’s reports of people receiving sickness benefits
for conditions such as alcoholism, obesity, and stress (see 17).
178
21. Far from protecting the vulnerable, these cuts are bearing down
disproportionately on those with disability.” Wheelchair user Tony
Vanterpool, 53, from Brampton in Cambridgeshire, said he had
come to protest at the cutting of his mobility allowance of £17 a
month (G news: society, May 2011).
Outside London, allowance refers to JSA and housing allowance.

However, the number of PNCs is greatly increased by (22), which
includes multiple place-names, and also accounts for the occurrence
of Lab as a PNC keyword. In contrast to the absence of Belfast in the
<*poverty*> PNCs (Chapter 4), highest refers to repeated reports that
Derry and West Belfast have the highest number of out of work benefits
recipients (with 23 including an uncharacteristic use of the term dole ).
22. 11 Birmingham, Hall Green (Lab) 12.5%; 12 Belfast North

(DUP) 12.4%; 13 Kingston upon Hull North (Lab) 12.4%; 14
Nottingham North (Lab) 12.3%; 15 Wolverhampton South East
(Lab) 12.3%. The lowest jobseeker ’s allowance (G news: society, Jan
2012).
23. Derry is one of the UK’s most impoverished (and youngest) cities.
Nearly 10% of its working population are on the dole and, apart
from west Belfast, the city has the highest numbers on welfare bene-
fits (G news: news, May 2013).
The Benefits subset includes some mentions of benefit fraud (within the
wider context of poverty) but these are treated as anomalous individual
cases—there are no references, for example, to the total welfare bill or
fraud rates. The Guardian also reports on Benefits Street, which it terms
‘controversial’ and ‘the current crucible of Britain’s social conscience’.
However, the programme is not sensationalised or held up as represent-
ative of wider trends in benefits receipt. Estate, he, Iain, Centre, photo-
graph, and visit are responsible for the PNCs in the Easterhouse area
of Glasgow; they refer to Iain Duncan Smith visiting the area in 2002.
Began, existing, health, IB, incapacity and testing refer to benefits reassess-
ment in Burnley and Aberdeen (see Sect. 6.3). Testing also refers to the
roll out of Universal Credit, a key component of the Welfare Reform
Act; as shown in (24), the Guardian reports on initial trials of Universal
179
Credit, but there is little direct evaluation of the policy in the PNCs. By
contrast, other elements of welfare reform, particularly the bedroom tax,
are evaluated negatively.
24. Universal credit went live on 29 April with just one jobcen-

tre—Ashton under Lyne—accepting clams for universal credit and
three other jobcentres, Wigan, Warrington and Oldham, testing the
system (G news: society, Jul 2013).
25. The council estimates that the money spent by the taxpayer in mit-
igating the effects of the bedroom tax in Newcastle far outstrip the
£3 m a year the DWP saves in housing benefit (G news: society,
Mar 2015).
26. Pauline Jones, 54, was also hit with the double blow of council tax
and bedroom tax after she had downsized from a three-bedroom
house to a two-bedroom in South Ockendon, Essex. Struggling
before the £22 monthly council tax bill and the £14 weekly bed-
room tax reduction, she’s resorted to using food banks to keep her-
self afloat (G news: society, Jan 2014).
Example (25) reports that the bedroom tax is economically inefficient,

and (26) shows how the bedroom tax affects individuals. In this case,
a direct link is made between welfare reform and the use of foodbanks.
Furthermore, the woman in (26) has ‘won an appeal’ against the bed-
room tax because one of her rooms was ‘too small to be classed as a
bedroom’. The implication here is that the bedroom tax is based on
unsuitable criteria and is thus not sound policy. The Guardian further
criticises policy in (27) suggesting that welfare reform penalises the
poorest areas. Similarly, (28) argues that housing allowance policy is
inadequate because rates are set centrally by the government. Criticism
of the government is also evidenced by council, county and services which
invariably refer to councils taking money away from people, and cutting
jobs (Sect. 6.3) and, ultimately, services.
27.
An analysis by Birmingham city council that shows the more
deprived an area, the deeper the cut. According to the Birmingham
analysis, Liverpool is the region worst affected by the government’s
cuts (G news: society, May 2012).
180
28. On the Mersey, 21,000 people collecting local housing allowance

will only be able to afford 12,000 homes in Liverpool. Because wel-
fare is set at Westminster, the cuts will also be felt in Scotland (G
news: society, Jan 2012).
29. The Homeless Link survey suggests that 41% councils in England
are making “disproportionate cuts” to homelessness services. This
includes Tory-controlled Cornwall county council, which has
imposed cuts of 40% on local homeless charities (G news: society,
Mar 2011).
To support its anti-cuts stance, the Guardian uses individual examples

of hardship (26) and makes links between benefits cuts and food pov-
erty, but it also draws upon research from independent bodies. Whilst
the research in (29) may support the Guardian ’s general position it is
used specifically to reference ‘Tory-controlled’ councils. There is clearly
no ambiguity about where the Guardian lays the blame for (benefits-
related) poverty and hardship.
7.4 Housing Costs and Relative Poverty

The Housing subset shows one of the biggest differences between the
two newspapers. The Daily Mail has relatively little geographical inter-
est in housing and poverty with only 253 PNCs, but this is of course
mitigated by the preceding discussion of Housing Benefits. The
Guardian, by contrast, has 1175 PNCs. The Daily Mail ’s geographical
spread (Fig. 7.3) broadly follows the newspaper’s background geogra-
phy as shown on the Kulldorf map. The only clusters occurring are rel-
atively small and are found around Birmingham (19 PNCs), Hull (14),
Bournemouth (12), Brighton (8), and Orpington and Sidcup (6) to the
south-east of London. The cluster in south-west Wales can be ignored as
it has only one PNC.
Of these, the Birmingham cluster is driven, at least in part, by dis-
cussion about Benefits Street (James Turner Street are all PNC key-
words). The Bournemouth cluster concentrates on examples of families
claiming to want better housing to accommodate their large families
Fig. 7.3 Density smoothed and Kulldorf maps of PNCs in the Housing subset in the Daily Mail
181
182
Table 7.4 PNC Keywords in the Housing subset in the Daily Mail (n ≥ 5)

Place Sig. Keywords
London <.01 –
London <.05 council
Rest of UK <.01 she, other, street, three-bedroom, association
Rest of UK <.05 month, works, million, damp, James, into, moved, mother,
across, no-one, turner, seen, inside, and, from
(cf. Sect. 7.2). In Brighton and Hull, articles concentrate on problems

with homelessness, with the Hull cluster in particular being caused by
reports of families being displaced out of London by the benefit cap.
The Sidcup/Orpington cluster focusses on a particular story, this time
sympathetic, of a homeless family who fail to meet the criteria for local
authority housing despite having lost their income as a result of redun-
dancy (30).
30. Homeless: Carole Karawaiez and her husband Norman stand next

to the car they are sleeping in with their son and daughter. The
couple had their home in Sidcup, Kent, repossessed 10 years ago
(DM news, Aug 2013).
The limited scope of the stories covered in relation to these clusters is

also reflected in the low number of PNCs in the Daily Mail Housing
subset (Table 7.4).3
As the pattern of PNCs seems to largely follow the Daily Mail ’s back-
ground geography, this results in a strong bias towards London: 41.1%
PNCs in the Housing subset are in London. The London PNC key-
words cannot tell us much, but the concordance analysis showed that
London was associated with the affordability of housing, the difficul-
ties of paying rent or housing costs (31), and/or no longer being able to
afford such costs (32). There are also references to the role of councils
and the responsibility to house families. Thus, there is a much overlap
between the Benefits subset and the Housing subset. These stories are
3For the Housing subset in the Daily Mail it was only necessary to closely read 150 concordances
due to the relatively small number of hits in this subset (cf. Sect. 6.3).
183
typically sympathetic to the plight of people struggling to pay for hous-

ing in London, especially when the working status of the individuals
concerned is mentioned (see ‘young professionals’ in example 31).
31. One of thousands of young professionals forced to still live with his

parents because he can not afford to rent in London (DM news,
Apr 2015).
32. Many of the tenants have been living in the building for over 10
years and claim they will have to find somewhere to live outside
London as they can no longer afford to pay Brixton rent prices
(DM news, Apr 2015).
Outside London the PNC keywords are somewhat more varied. She
is associated with stories about women in poverty. The occurrence of
feminine pronouns and female-gendered nouns occurs elsewhere (cf.
Chapter 4) and suggests that the Daily Mail has more of a tendency to
write stories about individual women than the Guardian. In this subset,
reports are sometimes sympathetic, especially when their subjects are
seen as hard-working women (33), but the Daily Mail is far less sympa-
thetic when poor women have large numbers of children (34).
33. I owe money to everyone including £800 to the housing associa-

tion for the rent which is one of the highest in Exeter. I am a sin-
gle working mum and I am not entitled to any financial help (DM
news, May 2011).
34. A single mother with eight children who receives £2000 a month in
state handouts has complained she is struggling to make ends meet.
Marie Buchan, 31, whose children range in age from 12 to two
months, lives in a three-bedroom housing association house in Selly
Oak, Birmingham, but says it’s not big enough and has applied for
a bigger one (DM news, Dec 2013).
The mother in example (33) is described as a ‘Hostage in her own

home’ because she wants to move out of a property that is too big for
her, but cannot, despite ‘pleading’ with the council. Much of the article
consists of direct quotation, where she notes that ‘red tape and a lack
184
of common sense’ means she cannot move but is unable to afford her
rent. Whilst the mother in (34) is also directly quoted, a larger propor-
tion of the article is given to the Daily Mail ’s description of her circum-
stances: they mention that she has been in court for rent arrears and is
unable to pay fines for using a bus lane on ten different occasions, and
note the amount of benefits that she receives. The same woman is also
mentioned in another Daily Mail article which reports that a 37-year-
old mother of eleven ‘who never worked and claimed around £60,000
a year in benefits, appeared in a documentary series on Channel 5
called On Benefits and Proud’. Thus, two independent stories are used
together to facilitate the Daily Mail ’s contrast between deserving, work-
ing mothers, and undeserving, entitled mothers who have too many
children. This tendency to append unrelated stories, as also shown
with the Somali and Afghan families in Sect. 7.2, does not occur in the
Guardian PNCs.
As well as associating housing with poverty far more than the Daily
Mail, the Guardian also has a much stronger geographical patterns,
as shown in Fig. 7.4. There is a major cluster of PNCs in London,
which is statistically significant against the corpus’ background geog-
raphy, and contains 57.0% of the PNCs in the Housing subset.
The West Midlands is the second-densest area in terms of density
smoothing but this contains only 4.3% of PNCs. The Kulldorf maps
shows a number of small, localised hotspots outside London includ-
ing Newcastle-upon-Tyne (22 PNCs), Luton (12), Caerphilly (11),
Slough/Windsor (60), and Solihull (6). The dominance of cold spots,
however, suggests just how polarised the pattern is, with most of the
country receiving very little attention in terms of housing (with the
exceptions of the clusters noted above, the M62 corridor, and Kent
and East Sussex).
The PNC keywords for London (Table 7.5) reveal that the
Guardian ’s coverage of poverty and housing in the capital is very much
associated with unaffordability, high, rising, or ‘soaring’ rents, and the
related roles of the benefit cap, private landlords, and the authorities.
There are also multiple PNCs concerning people being forced to move
out of London to cheaper areas (as was the case in the Daily Mail ),
homelessness (35), and (the attempted abolition of ) soup runs (36).
Fig. 7.4 Density smoothed and Kulldorf maps of PNCs in the Housing subset in the Guardian
185
186
Table 7.5 PNC Keywords in the Housing subset in the Guardian (n ≥ 5)

Place Sig. Keywords
London <.01 private, would, central, will, benefit, their, housing, are,
of, households, that, the, move, sector
London <.05 out, if, north, rent, high, accommodation, affordable, can
Rest of UK <.01 as, had, workless, far, include, acquired, areas, Facebook,
Victorian, empty, like, project, guardian, take, receiving,
behalf, shows, county, wrong, Pintrest, deprived, aspire,
framework, green, services, town, per, away, properties,
welfare, its, research
Rest of UK <.05 charity, terraced, also, end, well, support, which, residents,
association, young, years, house, been, nearby, got, did,
twitter, employment, going, recipients, heard, inform-
ing, councils
35. Others will be unable to pay their rent, and will end up in arrears,
then homeless or forced to move out of London to low-rent areas of
the country (G news: society, Apr 2013).
36. Westminster council says soup runs provide a magnet for homeless
people and encourage crime, begging and antisocial behaviour. It
tried to ban soup runs in 2007 (G news: society, Mar 2011).
The Guardian associates homelessness with place (particularly London)

more than the Daily Mail. It shows a stronger tendency to cover this
topic at all: normalised frequencies for <homeless*> are 54.43 and 39.39
(pmw) in the Guardian and the Daily Mail respectively. Nevertheless,
the key theme in the Housing subset for the Guardian is the unafforda-
bility of London and the possibility that councils will house people out-
side of their boroughs because there is no suitable local accommodation.
Outside London the PNC keywords are less focused on particu-
lar topics. There are reports of comments made about images on social
media (Facebook, Pinterest and Twitter), as well as stories about work-
lessness (37), (un)employment, and deprivation (38). Housing charities
(Aspire and Framework ) are also associated with places outside London.
37. The headline workless numbers include student households, and

excluding them from the figures pushes Nottingham into third
place total to 27.3%, while Glasgow moves into first place at
28.6% (G news: society, Nov 2011).
187
38. Manchester, where almost seven in 10 households suffer depri-

vation in terms of employment, education, health and housing (G
news: politics, Apr 2015).
Whilst the Guardian reports on household unemployment statis-

tics, noting (as does the Daily Mail ) the areas with the highest rates,
it does not accept these figures in the same way that the Daily Mail
does. Rather than evaluating the people living in workless households,
the Guardian questions the validity of workless statistics which include
students.
One of the key themes that emerges when exploring the co-text
around the non-London hotspots is the importance placed on coun-
cils re-housing people outside London (noted above). This contributes
to hotspots near London, such as Slough and Luton, and those much
further away including Hull and Newcastle. (The Caerphilly hotspot is
an exception, caused in part by reports of a murder in a homeless hos-
tel.) However, reports of high-rents and a housing crisis in London are
contrasted with reports of dilapidated, empty housing in the north and
midlands of England (39).
39. Residents in England’s poorest communities are finding themselves

trapped in streets filled with demolished or boarded-up houses after
an ongoing £5bn Whitehall housing renewal project was cancelled.
Residents living in mostly Victorian terraced homes in parts of
Birmingham, Salford, Teesside, Merseyside, Lancashire and South
Yorkshire (G news: society, Feb 2011).
Examples like (39) situate poverty geographically by focusing on the

relationship between government—situated in London—and poverty
in the north of England. The absence of adequate social housing for
‘England’s poorest communities’ is presented as both the responsibil-
ity of government and the result of (Conservative-led) policy decisions.
The communities are described as ‘trapped’ by policy decisions over
which they, presumably, have no control. This is another example of
the Guardian calling (implicitly) for more state intervention to prevent
social deprivation and poverty.
188
7.5 Comparisons with Quantitative Indicators

This chapter and the previous one have explored the geographies of
themes related to poverty as represented in the two newspapers using
four subsets of terms, those related to: Employment (see Sects. 6.2 and
6.3), Money (Sect. 6.4), Benefits (Sects. 7.2 and 7.3), and Housing
(Sect. 7.4). These have been shown to have different geographies from
each other, across the two newspapers. Section 4.4 explored compar-
ing PNCs from the search term <*poverty*> with a quantitative geog-
raphy of poverty as measured using Carstairs scores at local authority
district level. In this section we return to this approach to explore how
closely the two newspapers’ representation of the four sets of poverty-re-
lated themes relates to quantitative measures. There are several issues
that this analysis explores: firstly, the extent to which newspapers seem
to concentrate their coverage of poverty on areas that have high rates
of poverty. Second, whether the newspapers concentrate their coverage
on areas with high rates of unemployment, which would seem to sug-
gest that they associate poverty with people being out of work, or on
areas with high rates of poverty, which suggests more of an acceptance
that in-work poverty also exists. To explore this theme, as well as com-
paring poverty to Carstairs scores, a relatively broad measure of poverty
(see Sect. 4.4), we also use two measures of lack of work, namely the
census definition of unemployment as recorded in the 2011 census, and
worklessness, households that contain people of working age and where
nobody is in employment, as measured in 2015. These two measures are
shown in Fig. 6.1. Table 6.1 showed that all three measures are closely
correlated but the maps reveal that there are some differences in the pat-
terns. Our third topic is the extent to which the two newspapers differ
in both respects.
Comparisons between the Guardian’s four subsets and the three
measures of poverty are shown in Table 7.6. Note that in all cases PNCs
referring to high-level units have not been included (see Chapter 4).
In each case three comparisons are made: with all PNCs and all local
authority districts, only PNCs in districts with above average poverty
as measured by a Carstairs scores of greater than 0.0, and only dis-
tricts that contain at least one PNC. The table shows that all four PNC
189
Table 7.6 Spearman’s rank correlation coefficients comparing numbers of

Guardian PNCs with three measures of poverty at local authority district level
Carstairs 2011 Worklessness 2015 Unemployment 2011
All Employment PNCs .231** .092 .145**
Carstairs > 0.0 .090 −.022 −.031
Employment PNC > 0 .255 .181 .187
All Money PNCs .299** .184** .220**
Carstairs > 0.0 .239** .089 .122
Money PNC > 0 .246** .134 .148
All Benefit PNCs .388** .273** .342**
Carstairs > 0.0 .458** .239** .350**
Benefit PNC > 0 .407** .295** .349**
All Housing PNCs .379** .229** .323**
Carstairs > 0.0 .410** .117 .255**
Housing PNC > 0 .498** .180* .427**
**Significant at p < .01
subsets consistently correlate more strongly with the Carstairs measure

of poverty rather than either of the two measures of unemployment.
Areas with high deprivation scores are most closely correlated with the
Benefits and Housing subsets, with Money next and Employment being
the weakest. The Benefits subset has a stronger correlation than Money
except when districts with no housing PNCs are excluded, where the
relationship between housing and Carstairs, at 0.498, gives the highest
of any of the coefficients. This suggests that when the Guardian talks
about districts in relation to housing, it pays more attention to (some,
but not all) districts with higher rates of poverty.
Table 7.7 repeats this analysis using PNCs from the Daily Mail. As
with the Guardian, correlations are strongest with Carstairs scores than
the other two measures. However, the two measures are statically signif-
icant more commonly than in the Guardian. In direct contrast to the
Guardian, the Daily Mail ’s Employment subset has the strongest cor-
relation with Carstairs, in the Guardian this was the weakest. Housing
and Benefits have similar correlations of around 0.29, while the Money
subset has the lowest correlation.
The two tables suggest a very different emphasis from the two news-
papers on their attitudes to districts with high rates of poverty. The
higher a district’s rate of poverty the more emphasis the Guardian gives
190
Table 7.7 Spearman’s rank correlation coefficients comparing numbers of Daily

Mail PNCs with three measures of poverty at local authority district level
Carstairs 2011 Worklessness 2015 Unemployment 2011
All Employment PNCs .419** .335** .357**
Carstairs > 0.0 .264** .251** .167*
Employment PNC > 0 .345** .317** .295**
All Money PNCs .234** .211** .183**
Carstairs > 0.0 .214** .201** .077
Money PNC > 0 .309** .318** .234**
All Benefit PNCs .288** .248** .264**
Carstairs > 0.0 .191** .222** .165*
Benefit PNC > 0 .185 .294** .127
All Housing PNCs .296** .225** .217**
Carstairs > 0.0 .223** .175* .092
Housing PNC > 0 .347** .098 .294*
**Significant at p < .01
it in relation to housing and benefits. The Daily Mail associates these

districts with terms related to employment. Both newspapers’ PNCs
correlate with Carstairs scores more strongly than measures of unem-
ployment suggesting that they do understand that poverty is more com-
plex than simply being out of work. This is, however, less pronounced in
the Daily Mail suggesting that it pays less attention to in-work poverty.
7.6 Summary
Unlike the Employment and Money subsets discussed in the previous
chapter, the Benefits and Housing subset show much more overlap in
terms of theme. Although different in their geographies, the dominance
of Housing Benefit in the Benefits subset draws on very similar topics to
the Housing subset. There is a strong focus on affordability of housing
and housing costs, particularly in London, although the two newspapers
position themselves differently in terms of this broad issue. The Daily
Mail foregrounds extreme Housing Benefit claims in order to question
whether or not people should be able to live in an area supported by
Housing Benefits which would be unaffordable to most people. It con-
structs a dichotomy between Housing Benefit recipients and ‘hardwork-
ing people’, without acknowledging that this benefit is not restricted to
191
those who are out of work. By referring to those not receiving Housing
Benefit as ‘hardworking’ and those receiving benefits as ‘workshy’ the
Daily Mail is endorsing a neoliberal ideology which positions people
as able to achieve the social status and financial security they desire if
only they work hard enough. It implies that those who are receiving
Housing Benefit, and benefits more widely, are merely inadequate work-
ers. Relatedly, there is little sympathy for addicts, or people receiving
sickness benefits because they are out of work due to stress. By contrast,
the Guardian ’s coverage of the affordability of housing and dependence
on Housing Benefits does not encourage a neoliberal agenda. It focuses
on how government policy is exacerbating hardship due to cuts to the
benefits system. The Guardian focuses on the social responsibility of
the government to provide for those in food poverty and those who are
homeless, whilst also noting that certain areas (in London) are unaf-
fordable for most.
Our final chapter looks at the overall patterns in our data and so we
combine all the PNCs for each of our subsets. Instead of focusing on
different themes, the following chapter centralises geography and looks
at the different aspects of poverty that each newspaper associates with
particular locations. These range from large urban centres—especially
London—to smaller cities and larger towns; there is no real pattern
of rural poverty evident in our data. The maps we produce represent
snapshots of the geographies of poverty in the UK in the twenty-first
century as reported in the largest newspapers in the UK. Thus, they are
indicative of how people talk about poverty, who is classed as being in
poverty, and thus indicate the dominant ideologies of poverty associated
with a particular geographical location.
References
Baker, P., & T. McEnery. 2015. Who Benefits When Discourse Gets
Democratised? Analysing a Twitter Corpus Around the British Benefits Street
Debate. In P. Baker & T. McEnery (eds.). Corpora and Discourse Studies:
Integrating Discourse and Corpora. Basingstoke: Palgrave, pp. 244–265.
Ballinger, L. 2009. Taxpayers Pay £1,600-a-Week for Family of Ex-asylum
Seekers to Live in Luxury Five-Storey Home. Mail Online 20/11/09. http://
192
www.dailymail.co.uk/news/article-1231795/Taxpayers-pay-1-600-week-
family-ex-asylum-seekers-live-luxury-storey-home.html. Accessed 22/1/2018.
DWP. 2016. Freedom of Information Requests 2016-398/2026-397. https://
www.gov.uk/government/uploads/system/uploads/attachment_data/
file/502566/389-397-2016.pdf. Accessed 13/11/2016.
DWP. 2018. Housing Benefit caseload statistics. https://www.gov.uk/govern-
ment/uploads/system/uploads/attachment_data/file/659362/housing-bene-
fit-caseload-data-to-august-2017.ods. Accessed 23/03/2018.
GOV.UK. nd. Benefit Cap. https://www.gov.uk/benefit-cap. Accessed 23/11/2017.
Hansard. 2014. Topical Questions, 13 January 2014, Vol. 573. https://han-
sard.parliament.uk/Commons/2014-01-13/debates/14011313000007/
WorkAndPensions. Accessed 4/1/2018.
London: Oneworld.
Newton, J. 2015. More than 100,000 Benefit Claimants Across the UK Are
Being Handed £100 a Week After Being Assessed as Unable to Work
Because of Addiction, Obesity and Stress. Mail Online 22/05/15. http://
www.dailymail.co.uk/news/article-3006248/More-100-000-benefits-
claimants-UK-handed-100-week-assessed-unable-work-addiction-obesity-
stress.html. Accessed 14/3/2018.
ONS. 2017. Statistical bulletin: UK labour market Jan 2017. https://www.ons.
gov.uk/employmentandlabourmarket/peopleinwork/employmentandem-
ployeetypes/bulletins/uklabourmarket/jan2017. Accessed 23/03/2018.
27 (2): 195–214.
Slack, J. 2010. Migrants ARE Driving Down Wages of the Poor, as Equalities
Watchdog Blames East European Influx. Mail Online 18/01/10. http://www.
dailymail.co.uk/news/article-1244060/Migrants-ARE-driving-wages-poor-
equalities-watchdog-blames-East-European-influx.html. Accessed 15/1/2018.
8
Geography-Based Discourses of Poverty
In this final chapter, we bring together all of the PNCs for each of our
queries to establish overarching patterns of how each newspaper deals
with poverty and place. References to poverty in this chapter, therefore,
should be taken to refer to our entire dataset (including all of our que-
ries and all PNCs directly related to poverty, as shown in Table 5.7).
The analysis of the linguistic similarities and differences between the
Guardian and the Daily Mail begins with PNC keyword analysis. As
the preceding chapters have focused on particular subsets of queries, we
do not intend to repeat data here. For this final chapter we draw on all
PNCs to investigate how each newspaper discursively constructs poverty
and place, and how particularly-salient places are related to poverty by
each publication. Before focusing on which places the two newspapers
concentrate on in their discussion of poverty, we first explore the extent
to which the two newspapers discuss poverty and place overall (Sect. 8.1).
Having established the overarching geographies of poverty and place
in our corpora, we explore the topics that the two newspapers associ-
ate with different places. Our analysis begins with London (Sect. 8.2)
before focusing on major urban areas (Sect. 8.3) and other locations
(Sect. 8.4).

https://doi.org/10.1007/978-3-319-93503-4_8
194
8.1 Poverty and Place in the Two Newspapers

Table 8.1 shows how frequently instances of poverty and place occur in
the two corpora (normalised per million words to facilitate direct com-
parison). PNCs for the, calculated using a 5% sample of each corpus (as
discussed in Chapter 4), have been used to give an indication of how
frequently the two papers use place-names generally.
References to poverty are found more frequently in the Guardian
than in the Daily Mail; the Guardian has 35% more hits for our que-
ries. The second column in Table 8.1 shows that the Guardian also uses
place-names about 14% more than the Daily Mail. When the number
of poverty PNCs is compared (column three, Table 8.1), the dispar-
ity between the two newspapers becomes much greater: the Guardian
associates poverty with place, as measured by the number of PNCs,
more than fifteen times more often than the Daily Mail. As described
below, the Guardian strongly associates poverty with London and places
within it. However, even when London-based PNCs are removed, the
Guardian still associates poverty with place eleven times more than the
Daily Mail (column four, Table 8.1).
Figure 8.1 shows the distribution of all poverty instances for the
Guardian. The density smoothed map reveals a highly London-centric
pattern; London is the only cluster that scores above the z > 2.58
threshold (see Sect. 4.2) and 49.8% of the poverty PNCs are found
in this cluster. Birmingham is the only other place to cross the z >
1.96 threshold. However this cluster only has 267 PNCs compared to
3407 in London. The Kulldorf map—generated when poverty PNCs
Table 8.1 Frequency of poverty and place mentions in the Daily Mail and the
Guardian
All poverty PNCs for the All poverty Non-London
instances (nor- from 5% sam- PNCs (pmw) poverty PNCs
malised pmw) ple (pmw) (pmw)
Daily Mail 10,855.0 143.6 12.0 8.3
Guardian 14,656.8 163.2 187.1 93.9
Ratio (G: DM) 1.35: 1 1.14: 1 15.64: 1 11.32: 1
Fig. 8.1 Density smoothed and Kulldorf maps for all poverty PNCs in the Guardian
8 Geography-Based Discourses of Poverty
195
196
are compared to the background geography of their newspapers

(calculated using a sample of PNCs for the, see Chapter 4)—com-
plicates the idea of London as a hotspot somewhat. Areas of north,
east, and south London are revealed as hotspots but London itself,
represented by a point in the centre of London, is actually a cold
spot. This suggests that the Guardian concentrates on poverty in spe-
cific parts of London (such as Hackney, Islington, Haringey, Tower
Hamlets, Brixton, Lambeth, and Southwark) more than on London
as a generic whole. Elsewhere there are hotspots around Merseyside,
stretching east as far as Wigan and Warrington, and others around
Newcastle, and slightly further south around Middlesbrough. There
are smaller hotspots around York (46 PNCs), north of Merseyside in
West Lancashire (46 PNCs), Kent and Maidstone (40 PNCs), and a
much smaller one in Morecambe Bay (6 PNCs). The Guardian has
no hotspots in Wales, Scotland, or Northern Ireland. The overwhelm-
ing finding in Fig. 8.1 is the extent of the cold spots. With a small
number of exceptions, particularly in the south-east, almost all of the
UK is a cold spot. This suggest that the Guardian strongly associates
poverty with a small number of mainly urban places in England and
largely ignores it elsewhere.
Figure 8.2 shows all of the poverty PNCs for the Daily Mail. There
is also a London-centric pattern but noticeably less so than for the
Guardian. Almost exactly a third of the Daily Mail ’s 3687 PNCs are
in London but there are other clusters in Birmingham, Liverpool, and
Manchester that lie above the z > 2.58 threshold. The Kulldorf map
shows major hotspots in London, particularly north and east London,
the West Midlands, Merseyside, Greater Manchester (stretching
north to include Rochdale, Blackburn, and Burnley), Middlesbrough,
Newcastle, and Glasgow. All of these have over 50 PNCs. Smaller hot-
spots are found around Nottingham (40 PNCs), Hull (38), Aberdeen
(31), Kent and Maidstone (28), Clacton (23), Stoke-on-Trent (17),
Keynsham (16), Okehampton (14), Ebbw Vale in South Wales (12),
and West Berkshire (10). As with the Guardian, cold spots dominate
much of the map suggesting that the geography of areas associated with
poverty is highly polarised.
Fig. 8.2 Density smoothed and Kulldorf maps for all poverty PNCs in the Daily Mail
197
198
Figure 8.3 uses the spatial segregation analysis point comparison tech-
nique described in Sect. 4.2 to compare the distribution of all pov-
erty PNCs from the Guardian and the Daily Mail. Figure 8.3a shows
that, as expected, the area around London has significantly more
Guardian PNCs than Daily Mail ones. Much of the rest of the map
has more Daily Mail instances than Guardian ones. In particular, the
Daily Mail concentrates more on an area stretching from Birmingham and
Nottingham (in the midlands) to Liverpool and Manchester (in the north).
This pattern is potentially problematic, however, because the Guardian’s
overwhelming concentration on London means, almost by definition,
that its PNCs are under-represented elsewhere. For this reason, Fig. 8.3b
repeats the analysis but with PNCs within Greater London excluded. This
shows a noticeably more even pattern between the two newspapers outside
London, but the differences are still statistically significant.
Taking the analyses from Fig. 8.3 together suggests that, beyond
London, the major cities do not receive significantly more attention
from either newspaper, with the exception of Manchester which has
statistically significantly more Daily Mail PNCs at the p < 0.05 level.
The Guardian associates the south Midlands (including Oxfordshire,
Aylesbury, Milton Keynes, and Chipping Norton), York, Sheffield,
and Cornwall with poverty significantly more than the Daily Mail.
Other areas with significantly high Guardian PNCs include Belfast,
Consett (just north of Newcastle), east Glasgow, south Cumbria,
and Weymouth, but these are represented by only small numbers of
PNCs—35 in Belfast, fewer than twenty in all other cases. Thus, they
can tell us less about the newspapers’ representation of poverty than
might initially seem possible. The Daily Mail has a complex pattern that
includes clusters around Coventry (79 PNCs), northern East Anglia
(74), Swindon and West Berkshire (36), Scunthorpe to the south of
Hull (11), Okehampton in Devon (17), and south-west Scotland (5).
Considering Figs. 8.1–8.3 together reveals the following:
• Both newspapers strongly associate London with poverty, but the

Guardian does this to a far greater extent than the Daily Mail.
• Both newspapers also associate Merseyside, Birmingham, Newcastle,
and Glasgow with poverty. Arguably, the Daily Mail associates these
with poverty more than the Guardian but this is driven, at least in
Fig. 8.3 Spatial segregation analysis comparing all poverty PNCs for the Guardian and the Daily
Mail: a All PNCs b PNCs outside London

199
200
part, by the Guardian ’s over-emphasis on London. Of these city

areas, Merseyside is most associated with poverty based on the num-
ber of PNCs in its Kulldorf hotspots in both newspapers.
• Manchester reveals a contrast between the two newspapers, being a
density smoothed cluster and Kulldorf hotspot in the Daily Mail but
a cold spot in the Guardian. It has weakly significantly more instances
in the Daily Mail than the Guardian even after London has been
removed from the comparison. This is a somewhat surprising finding,
given the Guardian ’s history as a Manchester-based newspaper.
• Away from the major cities, Middlesbrough, Hull, and Kent and
Maidstone attract attention from both newspapers.
• Cornwall and, to a lesser extent, Sheffield, the south Midlands, and
Belfast are associated with poverty by the Guardian more than the
Daily Mail but, according to the Kulldorf analysis, they are not sig-
nificant, suggesting that the Guardian does not strongly associate
them with poverty.
• Similarly, the Daily Mail gives more attention to northern East
Anglia and Coventry than the Guardian does, but these are not hot-
spots so, again, they do not receive a significant amount of interest
from the Daily Mail.
• York is an area that the Guardian associates with poverty but the
Daily Mail does not, while Aberdeen, and Swindon and West
Berkshire are associated with poverty by the Daily Mail but not the
Guardian.
8.2 Discourses of Poverty and Place in London

Because London was so prominent in the PNCs we treated it sepa-
rately (as in previous chapters). Table 8.2 shows the PNC keywords for
London for each newspaper (the Guardian ’s London PNCs were com-
pared with the Daily Mail’s London PNCs). We have included PNC
keywords occurring a minimum of 5 times (numbers, place-names, and
punctuation have been omitted). There were 241 PNC keywords for the
Guardian and 139 for the Daily Mail.
201
Table 8.2 PNC Keywords for London when compared to the rest of the UK for
the Guardian and the Daily Mail (n ≥ 5)
Newspaper Sig. PNC Keywords London
Guardian <0.01 housing, rents, affordable, private, we, councils, outer, if,
move, that, new, local, inner, moving, will, government,
other, homes, benefit, properties, property, impact,
shelter, planning, both, residents, rented, cap, means,
changes, buy, huge, transport, third, neighbourhoods,
high, hostel, example, build, demand, outnumber,
receive, Livingstone, officer, create, better, hotel, you,
average, cheaper, not, building, the, paying, started,
schemes, advice, overall, provided, foundation, fair,
Amelia, sector, community, landlords, are, side, growth,
low-income, with, social, what, cancelled, develop-
ment, gentleman, editor, statutory, taking, would,
market, these, have, services, April, such, decent, soup,
legal, priced, wage, my, research, numbers, funding,
one-bedroom, works, director, larger, eviction, whereas,
specialist
<0.05 friends, offer, fund, out, than, additional, letter, apply,
shrink, assembly, regeneration, either, allowances, issue,
might, supporting, strategy, whom, guardian, Patrick,
included, evict, infrastructure, PCTS, by, rights, leave,
central, only, minimum, dependent, appeal, Randeep,
block, am, Wednesday, Ramesh, approach, email,
Monday, economically, bite, double, cleaners, sell, is,
shortage, it, places, of, claiming, provision, comes,
institute, mixed, lot, aid, analysis, universal, unless, but-
ler, wards, cope, substantial, Saturday, developments,
mothers, commission, neighbouring, accounts, inten-
tion, factory, housebuilding, contract, conservative,
claimant, short, deepest, unaffordable, should, recent,
increasingly, renting, action, so, be, put, place, prob-
lems, hit, very, fifth, soaring, few, wales, drop, decision,
earnings, driving, event, wide, movement, twitter, evi-
dence, standards, forces, age, extended, immediately,
limits, budgets, outstrip, squeezed, mansions, allow,
JSA, seeing, failing, failure, department, related, con-
servative-run, providing, vast, funds, pushing, two-bed,
federation, postcodes, accepted, win, increases, cities,
vacancies, temporary, accommodation, according, now,
people, least, result
(continued)
202
Newspaper Sig. PNC Keywords London
Daily Mail <0.01 cent, per, Mr, child, constituencies, enlarge, pictured,
combat, sustainably, yesterday, today, Mrs, BBC, poor-
est, cafe, taxpayers, district, updated, streets, Keely,
store, Romanian, pupils, street, reached, baker, Ms,
repossessed, slum, months, smith, men, vouchers, area,
without, was, constituency, he, details, Blair, meaning,
prepares, Muslim, walked, exclusive, arrived, features,
Tories, blocked, Victorian, map, speaking, never, areas,
worst, stage, fallen, website, house, children, claim,
speech, Iain, pound, successful, benefits, jobless, mail,
economy, during, forcing, top, poverty, pub, seven-bed-
room, located, pledge, father, eat, culture, part-time,
gang, reveals, grew, shops, Duncan, Tesco, open,
heavily, believed, queue, illegal, credits, Europe, show,
homeless, days, Ed, her, Miliband, while
<0.05 lives, fact, crime, jubilee, stewards, daughter, interview,
worldwide, asylum, restaurant, summer, payments,
his, unemployment, revealed, in, suffering, recently,
woman, labour, business, ago, she, English, daily,
millions, search, sending, figures, money, hundreds,
after, were, use, evictions, residential, regularly, bussed,
husband
Table 8.2 shows that the Guardian PNC keywords with the highest
log likelihood scores are very strongly associated with Housing bene-
fits and the benefits cap (housing, rents, affordable, private, move, mov-
ing, homes, benefit, properties, property, impact, Shelter, residents, rented,
cap ). Taken together, with the Guardian ’s wider focus on cuts, benefits,
and wages, these PNC keywords are used to express the newspaper’s
overarching argument that high housing costs, low wages, and the gov-
ernment’s cap on benefits mean that people are being forced to move
away from London to cheaper places—a process sometimes referred
to as ‘social cleansing’ (move, moving, out, to, force(d), cleansing, leave,
evict ). Relatedly, political organisations and individual politicians, such
as London mayor Boris Johnson, Chancellor of the Exchequer George
Osborne, and the London Assembly are all prominent, either for their
role creating welfare-related policies (George Osborne is strongly associ-
ated with the benefits cap), or objecting to them as ‘Kosovo-style social
cleansing’—a phrase attributed to Boris Johnson.
203
Thus, the overarching discourse of poverty in London in the

Guardian is that poverty is relative: London housing costs are more
expensive than elsewhere and wages are not high enough to provide a
decent standard of living (see the Living Wage Campaign, Sect. 6.3).
The Guardian also presents a causal link between London prices and
benefits receipt: as a result of high relative house prices people turn to
Housing benefit, which is now capped, meaning that the unaffordabil-
ity of London ultimately increases.
1. Before 2025 rents on most two-bedroom properties in the south will

become unaffordable to those claiming local housing allowance. Within
15 years, much of London’s commuter belt will become too expensive
for the state to pay for the poor to live in (G news: society, Nov 2010).
2. Towns such as Chelmsford, Newbury, Bath and Maidstone would be
no-go areas for those on benefits and all of Hertfordshire would be
out of bounds. The capital would be unaffordable within a decade (G
news: society, Nov 2010).
3. tight new housing benefit caps that do not increase in line with the
rising rents, and benefit reform, has made central London increasingly
inaccessible to people on low incomes (G news: society, Nov 2013).
Example (1) is framed by the claim that ‘Large swaths of southern

England will become off limits to Housing benefit recipients in a lit-
tle more than a decade because of the government’s proposed plans to
cut welfare bills’ (Ramesh and Sparrow 2010). Here the causality is
made explicit through the use of ‘because’. Furthermore, benefit cuts
are deemed responsible for ‘triggering a huge migration of the poor to
the north’ (Ramesh and Sparrow 2010), with the Guardian providing
lists of places which it claims will become ‘no-go areas’ for people on
lower incomes (2). The Guardian takes the position that those claim-
ing Housing benefit should not be excluded from areas where rents are
high (by government action); at its extreme this can be expressed as the
proposition that people should not be restricted in their choice of place
to live based on their economic circumstances (see Sect. 7.3). Housing
benefit is positioned as a means to ensure that those who cannot afford
London rents are not excluded from the city (where they are likely to
204
work or go to school). But the benefits cap, imposed by the govern-

ment, is discursively constructed as directly restricting access to high-
rent areas (3) and actively contributing to relative poverty.
The Guardian therefore conceptualises poverty as a condition associ-
ated with the removal of opportunity from groups of people based on
their economic circumstance. Individual people are not to blame for their
poverty, rather poverty is the effect of structural inequality influenced by
market forces, capitalism, and government endorsement of neoliberalism.
The Guardian positions itself as pro-welfare state and implicitly draws on
socialism by presenting the argument that poverty, within London in par-
ticular, would decrease with the removal of the benefits cap and the intro-
duction of a government-mandated living wage (see Sect. 6.4).
By contrast, the PNC keywords in the Daily Mail do not cluster so
neatly into overarching themes. The Daily Mail makes frequent use of
statistics (per, cent, statistics, map, economy, spokesman ), which suggest
a macro-level conceptualisation of poverty as something which can be
measured, endorsing the ideology that poverty is (easily) quantifiable.
Despite taking poverty as something which is quantifiable, reports of large
numbers in poverty and/or claiming benefits are contrasted with individ-
ual (and sometimes sensationalist) examples which are taken as represent-
ative of homogenous groups (Mr, Mrs, Ms, he, woman, her, she, husband ).
Not all individuals mentioned in relation to poverty and place are eval-
uated negatively (see Sect. 6.3 and the discussion of ‘jobless’) as actions
which are seen as an embodiment of neoliberal principles are presented in
a positive light. However, such examples are used contrastively with sto-
ries of individuals whose circumstances and actions are used to endorse
particular negative stereotypes about the poor. As shown in (4) and (5),
associations are made between receipt of benefits and being ‘jobless’,
‘depraved’, taking drugs, committing crime, and dealing in stolen goods.
4. Jobless Lee Miller, who has demanded a bigger council house for his
enormous family, claims he would turn to petty crime because he has
‘bills to pay’. The 40-year-old said he would travel to London and
target the wealthy to provide for his children (DM news, Oct 2010).
5. The pair led a depraved lifestyle, funding their mutual crack addic-
tion through benefits and flogging stolen goods. Jobless Spence
205
would regularly crash as Efremi’s north London pad, smoking drugs

for days at a time (DM news, Nov 2014).
Other trends in the London PNC keywords for the Daily Mail include
combating child poverty (child, sustainably, combat ) and the impor-
tance of schooling (pupils, schools, school ), in among more general PNC
keywords associated with money and employment (poorest, vouch-
ers, jobless, pledge (the failure of Tony Blair’s ‘pledge on poverty’), part-
time, unemployment, suffering, richest ), and housing (repossessed, house,
forcing, homeless, evictions, residential, mortgage ). There is opposition
to the cereal café in Shoreditch (café, Keely, restaurant ), which shows
some sympathy for those who live in one of the poorest boroughs in
the UK (see Sect. 4.3). Politics and politicians are again mentioned
(Smith, Blair, Duncan, Cameron, Tories, Iain, Ed, Miliband, Labour )
but there are also groups of PNC keywords which have no parallels in
the Guardian, such as those relating to minorities (Romanian, Muslim,
asylum ), crime (illegal, gang, crime), and history (Victorian, slum ). One
example which neatly contrasts the newspapers’ attitudes to poverty in
London is the occurrence of one-bedroom and seven-bedroom as PNC
keywords in the Guardian and Daily Mail respectively. This reveals
the former’s preference for stories about people living in overcrowded
accommodation because they cannot afford to move and the latter’s ten-
dency to report stories of (ex-migrant) benefit claimants living in large
properties at the taxpayers’ expense (see Sect. 7.2).
The Daily Mail ultimately endorses the benefits cap and other welfare
reform policies, and conceptualises poverty as caused by the (in)action
of (groups of ) individuals who are responsible for and/or complicit in
their own misfortune. The occurrence of Romanian as a PNC keyword
relates to multiple reports of migrants sleeping rough (6), begging, and/
or making children beg for them. A similar theme occurs in references
to Polish migrants (7) and is generalised to ‘groups of EU nationals’, as
in (6). Thus, at least some homelessness in London is blamed on migra-
tion rules and the free movement of EU citizens.
6. Police from Westminster regularly target Romanian beggars and

rough sleepers in London Mrs May will also say there is a ‘recurring
206
problem’ with groups of EU nationals who set up camps in public

areas in London beg and steal from tourists (DM news, June 2013).
7. The problem of Polish rough sleepers in the (sic) London has become
so bad that last year a homeless charity in the capital produced a
Polish-language video warning of the dangers of moving to the city
without a job or means to support themselves (DM news, Feb 2014).
In terms of the Daily Mail ’s overarching ideological position on pov-

erty, the conclusions here must be more tentative than those posed for
the Guardian. The Daily Mail does not systematically make direct links
between government policies and the causes of poverty and, as such, it
rejects a socially-oriented interpretation of poverty. This is evidenced
through reports of individual people’s circumstances, and a focus
on job cuts in the private, as opposed to the public, sector: the Daily
Mail focused more on job losses relating to declining industry than the
Guardian, who reported more on council job losses (see Chapter 6).
Ultimately, whist the Daily Mail does recognise that many people in
London are in poverty, it treats poverty as if it is an individual charac-
teristic. The direct causes for poverty are not high house prices and gov-
ernment cuts, but are implied to be the failure of government(s) to stop
alleged abuse of the benefits system (by individual scroungers and, more
generally, immigrants) and to increase the incentives for people to work.
8.3 Poverty in Urban Centres

Outside London, Manchester showed the biggest contrast in geographical
attention between the two newspapers. Although it received similar num-
bers of PNCs, 232 in the Guardian and 241 in the Daily Mail, it is a clus-
ter and hotspot in the Daily Mail but a cold spot in the Guardian. Spatial
segregation analysis reveals that there are significantly more PNCs for
the Daily Mail at p < .05 even with London removed from the compar-
ison. Table 8.3 compares the PNC keywords for the two newspapers. It
includes a comparison between Manchester and the rest of the UK (excl.
London) for each newspaper, and a direct comparison between the news-
papers. Including the former allows us to see where the two newspapers
are similar, whilst the direct comparison highlights the differences.
207
Table 8.3 PNC Keywords for the Manchester cluster. Italicised keywords are
found in both newspapers (n ≥ 5)
Manchester Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 incapacity, conference, city,
Facebook, last, police, as
<0.05 back, to, far, they, against, you,
Britain, losses
Guardian vs Daily Mail <0.01 council, as
<0.05 city, have
Daily Mail vs UK (excl. London) <0.01 feminist, awkwardly, t-shirt, giving,
conference, tests, pilot, market,
beggar, hospital, a, medical,
scheme, anywhere, plan, get,
begin, area, to, central, town
<0.05 before, rent, top, recently, found,
estate, under, fit, say, cent
Daily Mail vs Guardian <0.01 cent, per, town, Mr, child, near, pilot,
fit, made, his, left, help, she, also,
feminist, anywhere, awkwardly,
beggar, begin, hospital, t-shirt,
plan, close, rent
<0.05 being, market, area, found, money,
estate, number, top, a
It is clear from the sparsity of PNC keywords that the Guardian has
little to say about poverty in and around Manchester. The only themes
that seem to emerge as distinctive for Manchester compared to the rest
of the country are incapacity benefit and public sector cuts, which in
this case are associated particularly with the police. Council is a PNC
keyword when compared with the Daily Mail and is primarily associated
with council cuts, while conference refers to political party conferences
being held in the city, and they frequently refers to people protesting or
campaigning about issues including benefit cuts, homelessness provi-
sion, and the removal of the Education Maintenance Allowance (pro-
vided to young people who remained in education post-16).
The Daily Mail does not just emphasise Manchester but is also con-
cerned with towns nearby, such as Burnley and Rochdale. This emphasis
is in part revealed by town appearing as a PNC keyword when comparing
the Daily Mail to the Guardian. The most talked about of these towns is
Burnley, which is discussed in relation to the trial of new fit-for-work assess-
ments (tests, pilot, medical, fit, found ) relating to the replacement of incapac-
ity benefit with ESA (see Sects. 6.2 and 6.3). Rochdale is labelled (passively)
208
as ‘the most deprived area of England’ and is described as a recession-hit

market town, struggling due to the numbers of asylum seekers it has taken.
8. In the Falinge area, dubbed the ‘sick-note capital of Britain’, the

majority of the population is on sickness benefits. Moreover, almost
a fifth of Rochdale’s population now comes from migrant commu-
nities. In nine of the town’s schools, 70 per cent of the pupils speak
English as a foreign language (DM news, May 2010).
In (8) a link is made between the number of people receiving sickness

benefits and the number of migrants living in the area, but there is no
evidence presented to suggest a direct correlation between these two varia-
bles. The Daily Mail leaves the link between benefits receipt and migrants
implicit and further implies that there is some relationship between the
languages spoken by children in the area and its relative deprivation.
Whilst never stated explicitly, the underlying assumption here is that
somehow migrants are to blame for the high numbers of people receiving
benefits, despite the fact that, elsewhere, the Daily Mail notes that there
are no jobs available in Rochdale. The importance of adjacency has already
been noted in Chapter 7 and analysing adjacent or appended stories
linked to poverty though their proximity (such as the languages spoken in
schools in example 8) would be an interesting endeavour. However, such
analysis is beyond the scope of current discussions of poverty and place.
Away from stories specifically referring to these two towns, places
in this cluster are associated with deprivation, with schemes to help
deprived areas and references to child poverty. Plan is used in relation to
planned cuts, and rent and estate are associated with housing. As is often
the case, the Daily Mail refers to statistics in relation to place (per, cent,
number, top ) far more than the Guardian. There are also stories associ-
ated with individual people (Mr, he, she ), who are sometimes politicians
but often ordinary members of the public.
Similar to Manchester, Merseyside, Birmingham, Glasgow, and
Newcastle are all urban areas that have similar relative importance in the
two newspapers. Table 8.4 shows the PNC keywords for Merseyside.
For the Guardian, benefits are again a theme (universal, credit(s), test-
ing, system, tax ). However, in contrast to London, the benefits men-
tioned concern the rollout of universal credit, and tax credits, but there
209
Table 8.4 PNC Keywords for the Merseyside cluster. Italicised keywords are
Merseyside Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 universal, credits, Anderson,
addresses, common, harder, riots,
level, deprived, held, in, affected,
highest, testing, combined, urban,
able, claimed, study, system
<0.05 tax, credit, most, workless, house-
hold, life, these, other, years, how
Guardian vs Daily Mail <0.01 universal, other, credits, social, and,
addresses, affected, says, system,
tax, over, claimed, before, while,
sector
<0.05 will, life, common, public, authori-
ties, councils, able, housing, local,
council, that, under, testing, house-
hold, do, harder, riots, region,
held, set, urban, credit, combined,
service, spending, at
Daily Mail vs UK (excl. London) <0.01 highest, workless, households, areas,
poorest, sharp, communities, most,
deprived, homes, rate, list, number,
where, suggest, them, mother,
cent, incapacity, need, benefits, of
<0.05 receive, per, nearly, out, is, many,
figures, no, children
Daily Mail vs Guardian <0.01 cent, where, poorest, per, communi-
ties, households, Mr, number, list,
incapacity, sharp, receive, suggest,
nearly, mother, worst
<0.05 benefits, losing, jobless, rate, today,
areas, his, highest, top, places
are no repeated mentions of Housing benefits. There is also deprivation

(riots, deprived, urban—often in the sense of urban dereliction, workless,
life—usually expectancy or chances) and a strong impression of things
getting worse (harder, highest, level, affected, most ). Politics is also a
theme, with Anderson, the mayor of Liverpool, opposing cuts to councils.
The Daily Mail ’s PNC keywords also include a focus on benefits
(incapacity, benefits, receive
) with incapacity benefits being primary.
There is also a story about a Mr. Sharp whose attempts to go to a job
interview were thwarted by the rules around benefits (9). Here the Daily
210
Mail positions government bureaucracy as restricting this individual’s

access to work and depicts Mr. Sharp as someone who is being proac-
tive and looking for a job, thus embodying the neoliberal principles the
Daily Mail values.
9. he claims he had to pull out of the interview thanks to bizarre rules

that meant he could not miss a ‘Back To Work’ class on the same day.
Jobcentre staff warned Mr. Sharp, 33, from Huyton, Merseyside,
that if he missed the seminar he would lose his benefit money (DM
news, Oct 2010).
As with London, the Daily Mail draws on statistics (homes—used in

‘1 in 3 homes has no…’, rate, list, number, cent, per, figures ). The PNC
keywords also evidence themes of deprivation (deprived, workless, chil-
dren, no—in the sense of no one working) and reports on things being
bad or getting worse (communities, most, poorest, highest )—similar
themes to those found in the Guardian. Thus, whilst the two newspa-
pers appear to disagree about the levels and causes of poverty in London
(Sect. 8.2) they appear to have more in common when associating pov-
erty with Liverpool.
By contrast, the PNC keywords for Birmingham (Table 8.5) reveal
perhaps the biggest differences between the newspapers in all of the four
major non-London clusters. Analysing the PNC keywords make it clear
that the Daily Mail ’s coverage of poverty in Birmingham is dominated by
references to three high-profile stories. The first of these relates to the tel-
evision programme Benefits Street (10), which first aired in 2014 (Street,
James, Turner, show, lives, channel, residents, benefits ). The second is a
series of stories about a woman called Marie Buchan (see Chapter 7) who
has eight children and receives benefits (says, cap, three-bedroom, months ).
The evaluation of Buchan is similar to that identified in Sect. 7.2
but without the emphasis on the recipient being foreign-born. The
third involves a woman named Cait Reilly who took the government to
court claiming that an unpaid work placement in the shop Poundland,
which she had to do in order to keep her benefits, breached her human
rights (11). These three stories contribute to her being a PNC keyword,
and state is often used in relation to either Benefits Street or the Marie
211
Table 8.5 PNC Keywords for the Birmingham cluster. Italicised keywords are
Birmingham Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 street, channel, city, council, biggest,
the, since, increase, local, already,
nearly, claiming, economy, con-
ference, over, paying, workers,
disabled, today
<0.05 cut, its, planning, rents, to, number,
was, man, more, times, tory
Guardian vs Daily Mail <0.01 council, local, workers, housing, this,
as, constituency,
<0.05 last, year, economy, announced,
plans, under, that, now, affordable,
health, minimum, rents, region,
paying, before, authority, authori-
ties, times, so, some, than
Daily Mail vs UK (excl. London) <0.01 street, turner, James, Poundland,
says, cap, show, lives, channel,
three-bedroom, months, inner-city,
state, her, residents, benefits, capa-
ble, where, city, first, over, set
<0.05 seen, receive, cut, work, them
Daily Mail vs Guardian <0.01 cent, per, James, turner, street, town,
show, capable, months, benefits,
where, lives, her, residents, inner-
city, she, work, state, Poundland,
them, three-bedroom
<0.05 his, around, family, receive, in, says
Buchan story. There are also references to the fitness for work (capable,
receive, cut, work ) of benefits recipients in Birmingham.
10. Residents in a deprived Birmingham road were filmed making false

welfare claims and growing cannabis in their homes (DM news, Jan
2014).
11. Mr. Johnson also criticised unemployed graduate Cait Reilly who
claimed earlier this month that she is suing the Government for
making her taken an unpaid work placement at Poundland in
Birmingham. He said she ‘sneered’ at hardworking Britons by say-
ing the work was ‘forced labour’ (DM news, Jan 2012).
212
12. Workshy map of Britain revealed: Thousands of incapacity benefit

claimants found to be capable of working Birmingham has most
claimants capable of work (DM news, May 2013).
The people appearing on Benefits Street are accused of ‘false’ claims and
drug taking (10), Cait Reilly was described as sneering at ‘hardworking
Britons’ by the Mayor of London, and benefits claimants in Birmingham
are labelled ‘workshy’ and ‘capable of work’ (12). What unites these
stories is their negative evaluation of (predominantly) able-bodied peo-
ple who are represented as able to work but choose not to do so. Their
receipt of benefits, therefore, is seen as illegitimate—a position which
depends on an underlying ideology that benefits claimants can be easily
divided into the deserving and undeserving poor. This understanding of
poverty is reinforced by the contrast made between stories such as those
in (10–12) and more sympathetic stories, such as (13), which concerns a
National Lottery-backed food bank in Birmingham.
13. Earlier this year, reports from Birmingham’s Citizens Advice Bureau
revealed residents unable to afford their electricity bills were turning off
their fridges, leading to cases of food poisoning (DM news, Sep 2011)
Here there is no explicit negative evaluation of those who have caught

food poisoning, rather they are described in the article (13) comes from
as ‘The crippling human cost of the recession’ and ‘desperate families
starving in poverty’ (Watson 2011). Anomalously for the Daily Mail
those accessing the foodbank are not presented as architects of their
own misfortune, but rather ‘the recession’ is blamed for them not being
able to pay their bills and buy food. Thus, those visiting the foodbank
represent the deserving poor—those who face hardship through no
personal failings—who are in direct contrast to benefits cheats and the
‘workshy’.
The Guardian’s reports on poverty in Birmingham also include some
PNC keywords related to Benefits Street (Street, Channel ). However, it
is more concerned with cuts, particularly those associated with local
government (council, local, already, workers, cut, planning ); indeed the
Birmingham clusters is, in part, responsible for the focus on public
213
sector cuts discussed in Sect. 6.4. Benefits are also a theme (claim-
ing ), but the focus is more on Jobseeker’s Allowance than other bene-
fits. There is also a story about a man setting himself on fire outside a
Jobcentre when his benefits were cut. Housing is present again (rents)
and so are wages (paying ), particularly in relation to disabled people
earning the minimum wage. As with Merseyside, but in a less pro-
nounced manner, there is again a theme of things being bad or getting
worse, with number often being used to note the extremes of poverty, to
endorse the position that times are hard/tough.
With 99 PNCs, Glasgow is a hotspot for the Daily Mail but not for
the Guardian, which has 139 PNCs in the cluster. This difference is not
statistically significant using spatial segregation analysis once London is
removed. Despite the numerical similarity, the different focuses of the
two newspapers are revealed by the PNC keywords in Table 8.6. The
Guardian is largely interested in the city because of Iain Duncan Smith’s
visit to the Easterhouse estate (see also Sect. 7.3), which is connected
to his proposals for welfare reform (14) and the creation of the Centre
for Social Justice (Iain, Duncan, Smith, visit, estate, his, Centre, Social ).
Social sometimes also occurs in relation to social deprivation or social
Table 8.6 PNC Keywords for the Glasgow cluster. Italicised keywords are found
in both newspapers (n ≥ 5)
Glasgow Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 Iain, Duncan, visit, Smith, estate, life,
family, when, how
<0.05 as, now, his, new
Guardian vs Daily Mail <0.01 Iain, a, I, when, social, visit, for,
Duncan, smith
<0.05 estate, there, city, you, centre
Daily Mail vs UK (excl. London) <0.01 independent, lost, be, defence, ship-
yard, yards, another, capable, such,
are, at, of, figures
<0.05 some, would, this, being, including,
cent, work, food
Daily Mail vs Guardian <0.01 lost, per, cent, at, be, capable,
defence, would, independent,
including, shipyard, find, yards
<0.05 are, such, figures, being
214
enterprises, a deprivation theme that also accounts for life, as in low life
expectancy, and family, which is used in terms of family breakdown.
14. The occasion became known as the Easterhouse epiphany and in

2003, Duncan Smith announced he wanted the Conservatives to
become the “party for the poor” (G news: politics, Jun 2012).
The Daily Mail ’s PNC keywords for Glasgow draw on common dis-
courses of benefits claimants being capable for work, as found in relation
to the other cities discussed above (work is also used in the context of
looking for work or the Department of Work and Pensions). Similarly,
there is a focus on statistics (figures, per, cent ). The Daily Mail also asso-
ciates Glasgow with food banks, but this PNC keyword is accounted
for by multiple instances of ‘1 million people are being fed from food
banks such as this one in Glasgow’. Two further themes are distinctive
to Glasgow. The first concerns job losses in shipyards (lost, shipyard,
yards ). The second relates to the debate about Scottish independence
and its relationship to defence, particularly jobs in the defence industry
(defence, independent ). We do not find mentions of the 2014 independ-
ence debate in relation to other locations.
The final urban centre with similar geographical characteristics is
the area around Newcastle, which includes Gateshead, Tyneside, and
Sunderland. This area has 129 Guardian PNCs and 98 Daily Mail
PNCs, making it a hotspot for both newspapers with no significant dif-
ference between them. As with Glasgow, this masks noticeable differ-
ences between the two newspapers (Table 8.7).
The Guardian PNC keywords come from a range of different sto-
ries and cover themes such as benefits (paid, benefit, claimed, bedroom,
housing, unemployed ), politics (Labour, councils ), and charities and food
banks (bank, food, charity ). She is used to refer to individual women
who lost their jobs in the public sector or were affected by the bedroom
tax. The Daily Mail has very few PNC keywords for this cluster and
does not draw on the same themes as the Guardian for this geographical
area; food banks, for example, were not associated with Newcastle by
the Daily Mail. However, one PNC keyword (staff ) breaks this pattern,
215
Table 8.7 PNC Keywords for the Newcastle cluster. Italicised keywords are
Newcastle Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 bank, far, had, paid, benefit, food, a,
go, told, claimed, against, labour,
take
<0.05 set, I, charity, bedroom, councils,
workers, week, no, she, where
Guardian vs Daily Mail <0.01 a, far, councils, labour, housing, I,
benefit
<0.05 bank, area, living, take, so, charity,
set, unemployed, claimed, park,
paid, had, some, year, where
Daily Mail vs UK (excl. London) <0.01 Staff
<0.05 his, left, by, last, no
Daily Mail vs Guardian <0.01 –
<0.05 Was
as it tended to be used to report on redundancies among public sector

workers, a trend more commonly found in the Guardian (see Sect. 6.3).
The topics covered in each cluster discussed above tend to be differ-
ent to those used in relation to London (Sect. 8.2). Given the domi-
nance of London in our dataset, analysing all PNCs at the macro-level
(as in the previous two chapters) can hide the nuances of place’s impor-
tance in debates about poverty. The dominance of a particular location
may not occur in all datasets analysed using GTA, but the preceding
discussion shows the importance of analysing PNCs for a range of geo-
graphical clusters and conducting analyses at different levels of abstrac-
tion. In the final section of this chapter, we focus on the smaller, yet still
significant, clusters in our data.
8.4 Poverty Away from Urban Centres

The three other areas that appear as hotspots in both newspapers, but
which are away from major urban centres, are centred on Middlesbrough
(including Cleveland, Redcar, and Stockton-on-Tees), Hull, and Kent
216
and Maidstone. PNC keywords for these places are shown in Table 8.8.
In each case slightly different overarching discourses are revealed.
The Middlesbrough area is used by both newspapers as an exemplar
of poverty but the way this is done plays out differently in each publi-
cation. In the Guardian there is a discourse of supremacy for negative
reasons, with terms like highest, areas and where being used to describe
Middlesbrough as one of the areas with the highest levels of cuts or
unemployment, or being an area where unemployment is high and/or
life expectancy is low (cf. Glasgow and the Daily Mail, above). Unlike
many other places, however, the Guardian makes no clear links between
Middlesbrough and public sector cuts. This is anomalous and is per-
haps explained (in contrast to the Guardian’s treatment of Cornwall, see
below) by the fact that Middlesbrough council is held by Labour.
The Daily Mail focuses on Middlesbrough as an exemplar at a far
more localised level. In particular, it focuses on Limetrees Close (15)
and singles out this street as a place with problems of high crime and
unemployment (Limetrees, Close, community, situated, edge, near, area ).
15. Limetrees Close is situated in the deprived community of Port

Clarence, on the edge of Middlesbrough. Port Clarence suffers
from a very high crime rate, and very high unemployment rate, and
has suffered high levels of arson, fly-tipping and anti-social behav-
iour (DM news, Dec 2014).
The parallelism between ‘very high crime rate’ and ‘very high unem-
ployment rate’ serves to indicate a perceived relationship between these
two issues. Furthermore, the repetition of ‘suffers/suffered’ also suggests
links between high unemployment and ‘arson, fly-tipping and anti-so-
cial behaviour’, as they are created by the same process of suffering
(although the direction of any implied causality is not expressed). The
links between these issues are not made explicitly, but rather by men-
tioning them in the same sentence, the Daily Mail is presenting them
as linked; why mention these issues together if they are not (intended to
be) connected in some way. Mentions of Limetrees Close are also linked
to the second series of Benefits Street that was filmed nearby in Stockton
(series, average, residents ).
217
Table 8.8 PNC Keywords for hotspots away from major urban centres. Italicised
keywords are found in both newspapers (n ≥ 5)
Middlesbrough Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 are, under, highest, areas, where,
while, average
<0.05 Such
Guardian vs Daily Mail <0.01 are, there, highest, under, housing,
their, from, he
<0.05 it, to, people
Daily Mail vs UK (excl. London) <0.01 Limetrees, situated, close, edge,
Tata, products, losses, division,
community, on, series, long, aver-
age, is, followed, across, near, area,
as, job
<0.05 child, residents, such, worst, an, per
Daily Mail vs Guardian <0.01 –
<0.05 close, Limetrees, losses, area, Tata,
products, edge, cent, on, long, divi-
sion, near, situated, per, followed,
series, job, child, at
Hull Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 far, into
<0.05 in been
Guardian vs Daily Mail <0.01 Been
<0.05 into, he, at, was
Daily Mail vs UK (excl. London) <0.01 like, find, places, to
<0.05 Per
Daily Mail vs Guardian <0.01 per, find, places
<0.05 by, to, like
Kent* Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 Town
<0.05 homes, from, county
Guardian vs Daily Mail <0.01 –
<0.05 has, be, council, people
Daily Mail vs UK (excl. London) <0.01 cancel, crime, month, money, I, asso-
ciation, police, couple, moved, last
<0.05 house, had, to, no
Daily Mail vs Guardian <0.01 money, police
<0.05 Mr, find, couple, cancel, crime, up,
moved, I
*For the Kent cluster PNC keywords are included where n ≥ 3
218
There are two other themes associated with the Middlesbrough clus-
ter which are frequently found in the Daily Mail. The first is concerned
with particular industries, in this case job losses (and occasionally gains)
at Tata Steel (Tata, products, losses, division, long, followed, across, job ),
which echoes the Daily Mail ’s focus on industrial redundancies noted
in Sect. 6.3. The second, as is a commonly-used feature of Daily Mail
reports on poverty, is associated with statistics (child, poverty, such, worst,
per, cent ). This is not to say that the Guardian does not use statistics
when reporting on poverty and/or specific topics such as benefits receipt
or employment, but it does not use statistics as consistently as the Daily
Mail appears to. The overarching approach employed by the Daily
Mail appears to be to use individual (extreme) stories/examples, such as
Limetrees Close, and contextualise them as if they are representative of
wider statistical trends.
Both newspapers have stories about Hull being a place where peo-
ple are forced to move to because of changes to Housing benefits
in London: far is used in the Guardian in the context of having to
move as far as Hull, and like is used by the Daily Mail in the context
of rehousing people from London to places like Hull. So the distance
between Hull and London is emphasised to reinforce the injustice of
forced movement. Thus, for both newspapers, Hull perhaps reflects the
emphasis on housing and benefits in London; Hull is present in debates
about poverty precisely because it is not London. The Daily Mail also
has its emphasis on statistics which, in this case, includes stories about
Hull being one of the top ten worst places to find a job.
The Kent cluster does not have many PNC keywords, suggesting that
perhaps discussions of poverty in this area are fairly typical of discus-
sions of poverty elsewhere (outside London). That is, it is not overly
associated with a particular aspect of poverty. We reduced the minimum
threshold for PNC keywords for Kent to three to investigate whether
there were any more minor trends in the data. However, it is only asso-
ciated with familiar themes, such as the effects of council cuts in the
Guardian (warned, county ) and references to housing (town, homes ). The
Daily Mail also has stories about cuts, focusing more on the impact on
crime and policing (crime, police ), and has a story about the abuse of the
benefits system, focusing on an unemployed family with five children
219
who have been moved into a four-bedroomed housing association house

(association, couple, moved, house ).
In contrast to the apparent similarities in how the newspapers repre-
sent poverty in Kent, we conclude this section by focusing on those areas
where they differ the most. Beginning with the Guardian (Table 8.9)
we consider the locations which were hotspots for only one newspaper.
York is the clearest example of a place that the Guardian associates
with poverty that is largely ignored by the Daily Mail. The reason for
this was largely identified in Chapter 4, namely the Guardian ’s focus on
York attempting to make itself a poverty-free city. Chapter 4 focussed
simply on PNCs around the search-term <*poverty*>. When all of
our search terms are included in the analysis, the PNC keywords in
Table 8.9 reveal that whilst poverty and city are particularly associ-
ated with the anti-poverty initiative, bite and cuts are associated with
the impact of cuts and living references the living wage. Cornwall is
another area that only the Guardian associates with poverty. Its PNC
keywords reveal that the links made between poverty and place in this
location are largely associated with the Conservative-controlled county
council making cuts (16). The mention of which political party con-
trols the council, combined with the choice of the verb ‘imposed’,
makes clear who is making decisions in this situation and who, it is
implied, is being unfairly acted upon (‘local homeless charities’ and
ultimately ‘rough sleepers’).
16. This includes Tory-controlled Cornwall county council, which has

imposed cuts of 40% on local homeless charities at a time when
Table 8.9 PNC Keywords for Guardian only clusters (n ≥ 5)

York Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 bite, poverty, living, any, city, at, not,
we
<0.05 have, says, over, cuts, year
Cornwall Sig. PNC Keywords
Guardian vs UK (excl. London) <0.01 county, average, cuts, year, national,
last
<0.05 where, councils, like, I
220
official statistics show it has the highest number of rough sleepers in

England after Westminster (G news: society, Mar 2013).
There are a number of places of interest away from major urban cen-
tres that the Daily Mail associates with poverty but the Guardian does
not. These include smaller urban centres such as the clusters centred on
Coventry (72 PNCs), Aberdeen (31 PNCs), and Swindon and West
Berkshire (31 PNCs), which are all hotspots for the Daily Mail with sig-
nificantly more poverty PNCs than the Guardian, and a larger, more
rural cluster in East Anglia stretching from Cambridge to Norwich
which, although it has 63 PNCs, is a cold spot in the Daily Mail but
still has significantly more PNCs than the Guardian. The keywords for
these four clusters are shown in Table 8.10.
Each of these places seems to have its own narrative and it is hard
to identify common threads between them. The attention on Aberdeen
is dominated by the pilot scheme for ESA that also led to Burnley
being prominent in the Daily Mail. Almost all of the PNC keywords,
Table 8.10 PNC Keywords for Daily Mail only clusters (n ≥ 5)

Coventry Sig. PNC Keywords
Daily Mail vs UK (excl. London) <0.01 advert, city, foodbank, more, with,
unemployed, free, jobs, for,
vouchers, firm, make, staff, charity,
school, lose, to, than, office, a,
food, people
<0.05 Their
Aberdeen Sig. PNC Keywords
Daily Mail vs UK (excl. London) <0.01 tests, per, fit, if, cent, found, work,
benefit, that, were, to,
<0.05 and, will
Swindon/West Berkshire Sig. PNC Keywords
Daily Mail vs UK (excl. London) <0.01 Honda, car, aged, hit, between,
losses, factory, month, at, be, plant,
last
<0.05 for, will
East Anglia Sig. PNC Keywords
Daily Mail vs UK (excl. London) <0.01 advertised, plus, jobseeker, only,
allowance, job, pay, having, for, no
<0.05 last, at, were, work
221
including tests, per, fit, if, cent, found, work, benefit, that, were, and to
relate to this story. Contrastingly, there are several separate themes asso-
ciated with Coventry indexing a range of attitudes to poverty. The first
(advert, city, more, with, unemployed, jobs ) involves negative evaluation
of unemployed people within the city who are characterised as not look-
ing for work. Example (17) is repeated in three different texts. The use
of ‘air quotes’ questions the validity of the phrase ‘looking for work’—
it implies that people who are unemployed are not actually performing
this action—whilst the mention of 10,000 people contrasted with the
capitalised ‘TWO’ serves to reinforce the difference between the num-
ber of unemployed people and applications for a particular job. This
example is taken as indicative of wider trends in the number of jobs
applied for in Coventry (even though no context is provided about the
types of job on offer) and serves to support ideologies of the lazy and/or
idle poor.
17. Workshy Coventry? Advert for 20 jobs in a city with more than

10,000 unemployed ‘looking for work’ gets just TWO replies (DM
news, Aug 2011).
There are also more sympathetic stories, one of which concentrates on

schools helping parents get food from foodbanks (foodbank, free, vouch-
ers, charity, school, food ), suggesting that the Daily Mail does not apply
the same level of agency to children experiencing poverty as they do
to adults. Indeed, the existence of the term child poverty (discussed in
Chapter 4) and the Daily Mail ’s reports on the measurable number of
children not bringing food to a Coventry school serves to reify the lived
experience of poverty amongst a certain demographic, whilst simulta-
neously setting it apart from other, more abstract notions of poverty.
Child poverty appears to be a phenomenon in and of itself, and investi-
gating how this term is used in wider debates about UK poverty (inde-
pendent of place) is worthy of further research. In Coventry, there are
also references to redundancies, especially after the firm CityLink went
into administration (firm, staff, lose, office ), and mentions of a report by
Coventry University into the white working class.
222
Redundancies are also a cause of the Swindon and West Berkshire

cluster, particularly at the Honda car plant in Swindon (Honda, car, hit,
losses, factory, plant, will ). However, akin to the discussion of Aberdeen
in Sect. 6.3, there are also more positive stories about the availability of
apprenticeships for young people (aged, between, month, last ). Finally,
although the East Anglia cluster is quite diverse, the PNC keywords
concentrate on two stories. The first concerns protests against a work-
fare scheme where jobs are offered for only Jobseekers’ Allowance plus
expenses (advertised, plus, jobseeker, only, allowance, job, pay, no, work ).
Whilst the Daily Mail does not take a clear stance on this scheme, pro-
testers are described as a ‘pressure group’ and the newspaper notes that
the jobs on offer were allegedly posted ‘in error’. The second concerns a
television programme about (migrant) fruit pickers (18).
18. A Panorama programme where the long-term unemployed were

taken to farms in East Anglia to pick crops. The unemployed
couldn’t bend over to pick the vegetables, looking and sounding
more like vegetables themselves, and others who were interviewed
outside a JobCentre said they couldn’t be bothered to get up for the
wages on offer (DM news, Feb 2012).
The explicit negative evaluation ‘sounding more like vegetables’ and

reference to the ‘workshy’ stereotype implied by ‘couldn’t be bothered’
supports the Daily Mail ’s overarching position on benefits claimants.
The apparent physical inability of those concerned to pick crops, and
their complaints about low wages, are not seen as legitimate. However,
the article (18) is taken from (Atkinson-Small 2012) is actually an anti-
immigration piece where, despite the negative evaluation of the people
in (18), the social group actually blamed for workelessness are migrant
workers: ‘Immigration and youth unemployment go hand in hand’. In
contrast to the position taken in (18), the author suggests that there are
‘hundreds of thousands of British kids willing to step into these roles
[fruit picking] and the tragedy is that they can’t because of uncontrolled
immigration’. This example highlights the importance of looking at how
UK poverty is depicted within a wider context. For example, we have
already seen that poverty and life expectancy are linked (cf. Glasgow,
223
above). Here, Atkinson-Small states a causal link between immigration

and unemployment rates of British-born nationals (see also the discus-
sion of the languages spoken in Rochdale schools, above). However, this
argument is never fully evidenced, as it is unclear how the reduction of
migrant workers would, presumably, decrease the apparent laziness and
increase the physical abilities of the people referred to in (18).
The thematic analysis of PNCs in previous chapters showed the
major trends in how the two newspapers discuss poverty and place.
However, the geographically-focused approach taken in this chapter
has shown that the dominance of London in the PNCs somewhat over-
shadowed the nuances of the relationship(s) between poverty and place
drawn upon in each newspaper. The clearest example of the different
relationships established by the newspapers is evidenced by their focus
on different types of benefit in particular areas. Not only do the two
newspapers report on different benefits, their focus on particular ben-
efits is filtered through place; discourses relating to benefits and benefit
claimants are thus localised. Ultimately poverty is conceptualised as rel-
ative, with the newspapers portraying different areas of the UK as expe-
riencing poverty in different ways. The preceding analysis is also further
justification for the argument that critical analysis of media texts would
benefit from a consideration of place (at multiple levels of abstraction).
8.5 Summary
There is an interesting paradox in much of what this chapter has
revealed. On the one hand, the Guardian has more instances of both
poverty search terms and place-names than the Daily Mail, and its
interest in poverty and place (as numerically measured) far exceeds that
of the Daily Mail. On the other hand, the Daily Mail seems to have a
wider agenda both in terms of the places that it discusses and the dis-
courses that it associates with those places. The Guardian ’s coverage of
poverty and place is dominated by London with a focus on housing
and its apparent unaffordability in the light of low wages and benefit
cuts. This comes together in stories about people being forced out of
the capital due to high housing costs. Outside London, the Guardian ’s
224
stories often concentrate on local government cuts. Housing remains

a theme, as do benefits, although it is interesting that Housing benefit
seems to be seen as a London-based issue while other places are associ-
ated with other types of benefits (e.g. Incapacity Benefit in Manchester
and Universal Credit in Merseyside). Charity and food banks are also
important, but the Guardian most strongly associates these with
Newcastle.
The Daily Mail ’s construction of relationship(s) between poverty
and place are more fragmentary both geographically and thematically.
There are however some common threads. The Daily Mail concen-
trates on private companies far more than the Guardian, particularly,
but not always, in relation to job losses. Examples include Tata Steel
in Middlesbrough, Honda in Swindon, CityLink in Coventry, and the
shipyards in Glasgow. A second theme that has a clear impact geograph-
ically is the pilots for the new ESA fitness-to-work tests, with the asso-
ciated finding that many claimants are being assessed as able to work.
This leads to large amounts of attention on Burnley and Aberdeen.
More generally, the Daily Mail seems to associate poverty and place for
two sets of reasons. Either there is a focus on stories about individu-
als or streets such as Benefits Street (James Turner Street) in Birmingham
or Limetrees Close in Middlesbrough, or an area is discussed in very
aggregate terms. The aggregation of poverty statistics and the choice
to report on multiple individuals with similar circumstances (from the
same street or area) serves to homogenise representations of those who
are poor and/or in receipt of benefits. Stories of individuals are often
negative, particularly when associated with Housing benefit and work-
less families with large numbers of children. London, Birmingham, and
Middlesbrough are particularly associated with these stories.
There are also stories about individuals in poverty who are attempting
to make the best of the situation despite difficulties which may, like in
the case of Mr. Sharp in Liverpool (Sect. 8.3), be presented as caused by
the inflexibility of the benefits system; in this case an individual’s agency
(and related neoliberal agenda) is being impeded by state-controlled
conditions of welfare. By contrast, aggregate stories tend to concentrate
on identifying particular places as having particular problems, which are
generalised to the whole areas, and child poverty is seen as a particular
225
concern. A similar practice can be seen in the Daily Mail ’s reporting of

alcohol dependency hotspots (cf. Chapter 7).
It is notable that some places are associated with particular issues
in both newspapers, while others are treated differently. Both newspa-
pers seem to regard Merseyside as one of the most deprived parts of the
country, with workless, deprived, highest, and most featuring in the PNC
keywords for both (Table 8.4) alongside references to different kinds of
benefits (Universal Credit and Tax Credits in the Guardian, Incapacity
Benefit in the Daily Mail ). Middlesbrough also appears to be associated
with deprivation by both newspapers, although in the Guardian this is
more aggregate while the Daily Mail focuses on Limetrees Close in par-
ticular. At the other end of the scale, Glasgow has completely different
associations in the two newspapers. The Guardian ’s focus is dominated
by Iain Duncan Smith’s visit and its consequences, whilst the Daily Mail
associates Glasgow with jobs and Scottish independence. Thus, the for-
mer endorses a position which sees Glasgow as particularly deprived
and on the political radar for welfare reform, whilst the latter minimises
issues of welfare.
Whilst we established in Chapters 4 and 5 that, in terms of raw
hits of our queries, there was no overwhelming systematic association
between poverty and place by either newspaper (although there was a
stronger trend in the Guardian ), what we have shown here is that when
poverty, and its collocates, are associated with place, they tend to be
used in a way which facilitates the conceptualisation of poverty as an
urban phenomenon. This may be due to population density and the
fact that more people living in a given area means more people could
possibly experience poverty. It could also be due to the relative news-
worthiness of reporting large redundancy/unemployment figures, or
rates of benefits receipt in densely-populated areas, as opposed to focus-
ing on (relatively) smaller numbers for such measures in rural areas.
Nevertheless, we have shown that the links made between poverty and
place are highly localised. The individualisation of poverty and place is
evidenced by those stories which are repeated in a given newspaper and
lead to multiple PNCs for a particular area. Although a focus on indi-
viduals was characteristic of the Daily Mail (a finding similar to that of
Lundström (2013), see Chapter 1) the Guardian also reported on indi-
viduals to a lesser extent.
226
What we can therefore conclude is that there is no one discourse of

poverty and place that is UK-wide. Rather, discourses of poverty change
in relation to place and, of course, in relation to the underlying ideol-
ogies of the newspapers reporting on such places. The strong associa-
tion the Guardian makes between poverty and York is not present in
the Daily Mail because the latter tends to focus on areas of high pov-
erty rates and York has a relatively low poverty rate. The Daily Mail also
focuses on benefits receipt and uses extreme examples of benefits claim-
ants to endorse an underlying support of Conservative-led policies to
shrink the state, reflecting an adherence to ideologies of neoliberalism
and capitalism. The poverty reported in York does not seem to fit this
pattern. By contrast, the Guardian does not focus on extreme examples
of benefits receipt and it does not seem to acknowledge that some fam-
ilies do claim large amounts. Nor does it make links between poverty
and crime and/or migration. It focuses on public sector redundancies,
benefits sanctions, and high house prices in order to oppose government
cuts and to argue for a larger state, positions which are in keeping with
its left-wing stance.
References
Atkinson-Small, J. 2012. It’s Time to Pull Up the Drawbridge: Ban Immigration
for Five Years and Give Our Youth a Chance. Mail Online 21/2/2012. http://
www.dailymail.co.uk/debate/article-2104253/UK-ban-immigration-5-years-
youth-chance.html. Accessed 13/3/2018.
Ramesh, R., & A. Sparrow. 2010. Housing Benefit Cuts Will ‘Push Poor Out of
South’, Experts Warn. Guardian 8/11/2010. https://www.theguardian.com/
society/2010/nov/08/housing-benefit-north-south-divide. Accessed 22/2/2018.
Watson, L. 2011. Lotterty Hands Over £425,000 for Charity Food Bank in
Birmingham. Mail Online 22/9/2011. http://www.dailymail.co.uk/news/
article-2040024/Lottery-hands-425-000-charity-food-bank-Birmingham.
html. Accessed 21/2/2018.
9
Conclusions
In the preceding chapters we have presented a new interdisciplinary

method of analysis which draws together techniques from corpus lin-
guistics, critical discourse analysis (CDA), and geographical information
systems (GIS). Taking the discursive construction of UK poverty as a
case study, we have provided a clear method which demonstrates how
geographical text analysis (GTA) can bridge the gap between quanti-
tative and qualitative data. We have shown how GTA facilitates com-
parison between the textual sources found in our corpora and official
statistics associated with poverty and, by focusing on two newspapers,
we have also demonstrated how GTA can be used to contrast the geog-
raphies of multiple corpora. Our analysis of place, in the form of place-
names, has provided a new lens through which to analyse the language
used by the mass media to discuss UK poverty.
In this final part of the book, we summarise our findings and eval-
uate what they can tell us about the role of place-names in discourses
of poverty and place. In Sect. 9.1 we consider the similarities between
the two newspapers’ treatment of UK poverty, including their shared

https://doi.org/10.1007/978-3-319-93503-4_9
228
lexical resources, their London-centric nature, and their unexpected

focus on housing. Section 9.2 focuses on the nuances of the Daily
Mail ’s approach to UK poverty, whilst Sect. 9.3 does the same for the
Guardian. In Sect. 9.4 we focus specifically on place and the patterns in
the place-names used, before evaluating GTA in Sect. 9.5. We also note
the limitations of the state of the art of GTA and propose some poten-
tial avenues for further research.
9.1 Trends in Newspaper Discourses of Poverty

Given their political leanings and established public images, it is no sur-
prise that our analysis has shown that the Daily Mail and the Guardian
take different stances on poverty. It was also expected that newspaper
reports on UK poverty would include references to employment and
benefits receipt. What is novel here is that we have shown how the two
newspapers endorse their position on UK poverty by choosing to focus
on particular topics as realised in particular geographical locations.
Ultimately, the Daily Mail takes a right-of-centre stance that supports
capitalism and neoliberal ideologies. One instance where this position
is realised through allusion to place is in the example where the Daily
Mail implied that unemployed people in Merthyr Tydfil should travel
further to find work, as there were jobs available in Cardiff (Sect. 6.3).
By contrast, the Guardian takes a left-wing view and advocates for a
larger state and the common good, rather than profit, as shown in its
treatment of public sector cuts in Birmingham (Sects. 6.4, 8.3) and its
focus on the benefits cap in London (Sect. 7.3).
Whilst political differences such as these were to be expected, there
were similarities between the two newspapers, perhaps the most promi-
nent of which was the fact that both newspapers’ coverage of poverty was
London-centric. There are also similarities in the language used by the
two newspapers when discussing poverty. Each had similar collocates for
poverty (Sect. 5.1) and largely comparable numbers of query hits over-
all (although the Guardian did use the search terms more to talk about
poverty and place). Furthermore, there was no evidence that highly
evaluative terms, such as scrounger or chav, were systematically used to
9 Conclusions
229
negatively evaluate the poor. Nor does either newspaper tend to draw on
particular indexes of the working classes—such as smoking, wearing par-
ticular clothing, drinking alcohol, etc.—found elsewhere in public dis-
courses of poverty (cf. Paterson et al. 2017, van der Bom et al. 2018).
Thus, neither newspaper engages primarily with UK poverty and
place in the same way it has been portrayed through other media out-
lets. For example, poverty porn programming like Benefits Street, with
its evaluative narration and editorial choices to focus on deviant behav-
iour, reinforces stereotypes about the undeserving poor in particular
locations in England.1 Such media representations and use of ‘scrounger
discourses’ (see van der Bom et al. 2018) are socially powerful; ref-
erences to such programmes have been used to support government
positions on welfare reform (see the Hansard extract in Sect. 7.1). The
newspapers’ coverage of poverty is somewhat more highbrow and less
sensationalist (with the potential exception of the Daily Mail ’s use of
extreme individual examples). The two newspapers do draw on stereo-
types of the poor to a certain extent, but mostly this is implied, rather
than comprising the explicit labelling of someone as a scrounger or the
overt evaluation of particular behaviours and/or consumer practices.
This is because the stereotypes attached to the UK poor are already well
established; the newspapers assume that their readers have been exposed
to such stereotypes and thus, to an extent, they have become reified and
accepted as common sense.
Another unexpected finding is that reports on poverty—London-
centric as they are—are dominated by references to the high cost of liv-
ing and high house prices, especially in the Guardian. Not being able
to afford (to own) a property in a major capital city does not automat-
ically make someone poor, and there seems to be a wide level of pub-
lic consensus on this; in the Breadline Britain surveys (cf. Lansley and
Mack 2015) home ownership was not widely seen as an essential for
having a good quality of life in twenty-first century Britain. It was not
1A similar trend can be found in fictional caricatures of the working classes. People tend to be
represented as either feckless, as in sitcoms such as Shameless and the sketch show Little Britain
(see Tyler 2008) or hardworking victims of circumstance at the mercy of government policies and
an ineffectual welfare state (see Ken Loach’s I, Daniel Blake, 2016).
230
expected that the affordability of housing would be such a core theme

in discourses of poverty and place. However, references to housing
and the cost of living (especially rents) were extremely prevalent in the
co-text of the PNCs and therefore could not be ignored in our anal-
ysis. The occurrence of housing as a core theme in our data also acts
as evidence for the merits of data-led research; by generating the search
terms in the way that we did (using collocation to select suitable query
terms, Sect. 5.1) we did not merely impose our own ideas about what
concepts would be linked to poverty in language. In a similar vein, the
relative absence of popular political terms such as austerity and recession,
neither of which tended to occur in the PNC keywords, were surprising
because we would have expected them to be drawn into debates, given
the close association between government policies, welfare reform, and
the overarching discourse of austerity which became popular after the
2008 recession.
The correlation of housing with other (more expected) terms relat-
ing to poverty, such as those in the Benefits, Employment, and Money
subsets, begs the question about what it is to be poor in the UK. Whilst
we find reports of extreme poverty in both of our corpora—and in most
cases these are treated with sympathy and are underpinned by notions
of the deserving poor—the majority of PNCs focused on relative pov-
erty, particularly in terms of benefits receipt. The prevalence of reports
about relative poverty supports the overarching notion that there is no
such thing as a hard poverty line or threshold by which poverty can be
objectively measured, yet both newspapers treated poverty as if it is (eas-
ily) measurable and quantifiable. The Daily Mail in particular endorses
this position by its use of statistics as a matter of rule rather than excep-
tion. Despite the trend towards quantification, the discourses of pov-
erty identified herein are evidence of its conceptualisation as a relative
phenomenon, characterised not necessarily by different amounts of pov-
erty, but different types. People are poor in London because they can-
not afford expensive houses and the cost of living, people are poor in
Liverpool because they are unemployed, people are poor in Birmingham
because there are lots of council cuts to services, welfare, and jobs, peo-
ple are poor in Rochdale because they are sick (as evidenced by high
numbers of Incapacity Benefit claimants), and people are poor in
9 Conclusions
231
Glasgow because of sink estates, low life expectancy, and poor health
profiles (see Chapter 8).
Similarly, different social groups’ experiences of poverty were evalu-
ated differently. For example, children had no agency in their own pov-
erty and the label child poverty appears to refer to a particular type of
poverty (a topic worthy of further investigation). Nevertheless, children
were used discursively by the newspapers (the Daily Mail in particular)
to index a particular stance on the causes of poverty: references were
made to the number of children that people had and to the different
languages spoken by migrant children in Rochdale (Sect. 8.3). Again,
the newspaper did not need to explicitly set out the positions that peo-
ple should be able to financially support the children they conceive
or that migrants are to blame for fewer jobs and a higher welfare bill.
These notions are implied because they are expressed elsewhere (within
other media or within other articles in the Daily Mail ) and there is an
assumption that knowledge of such positions will act as a filter through
which the newspaper’s readership understand the preferred reading and
the subtext of any given article concerning poverty.
Further supporting evidence for the position that poverty is some-
thing relative, qualitative, and abstract, as opposed to something funda-
mentally economic, was the fact that money terms did not always play
a central role in the newspapers’ reports on poverty. There was little ref-
erence to people counting the pennies or earning less money, suggesting
that a conceptualisation of poverty based solely on economic resources
is not representative of how the newspapers present this phenomenon
to the wider public. When numbers are brought into reports, they tend
not to relate to the specific price of goods or services, or the amount of
money people have (with the exception of reports about the amounts
received for Housing Benefit reported in the Daily Mail ), rather they
tended to focus on poverty in aggregate terms: the number of people
claiming benefits in a particular geographical area, the employment
rates for different cities, etc.
Money tended to be referred to implicitly, through references to ben-
efits or jobs (which implies wages). When money was explicitly talked
about, reports tended to focus on the Living Wage Campaign, which
both newspapers ultimately endorsed albeit for different reasons: the
232
Guardian suggested that it would improve people’s quality of life, whilst

the Daily Mail took the position that higher wages would reduce the
overall welfare bill, because it would make work more attractive than
benefits receipt and would decrease the number of people drawing on
Housing Benefit and particular government credit schemes. Ultimately,
what underpins both newspapers’ arguments is the reasoning that pov-
erty can be alleviated (simply) by increased employment and higher
wages. Thus, there are clearly points in discourses of poverty where the
two newspapers agree on both its causes and its remedies. However,
as was expected, there are also instances where the two newspapers are
wholly dissimilar and one way their different stances on poverty is evi-
denced is through their choice to focus on particular aspects of poverty
in particular geographical locations.
9.2 Poverty and Place in the Daily Mail

In the Daily Mail, there is a discourse of the idle poor; poverty is
understood to be cause by individual laziness, and is underpinned by
the notion of the undeserving poor. The newspaper takes the neolib-
eral stance that it is possible to get oneself out of poverty through hard
work alone (and indeed there are multiple reports of people escaping
poverty—see Chapter 4). Following Lansley and Mack (2015: 203) this
position implies that ‘inequality merely reflects differences in talent and
endeavour’ and, by the same token, means that economic and power
differences ‘with rewards at the top overstating economic contribution,
and those at the bottom understating it, can be conveniently ignored’.
The Daily Mail ’s discussion of foodbanks in Birmingham (Sect. 8.3)
does buck this trend somewhat and it is one of the only times the Daily
Mail acknowledges that external forces, in this case ‘the recession’, have
played a role in poverty rates.
The newspaper’s position is realised in language though reports of
individual examples, which are repeated across multiple texts. The most
newsworthy examples tended to be the most extreme, such as families
with large (legitimate) benefit claims. Whilst benefit fraud as a topic in
its own right was not considered in this study, general deviant behaviour
9 Conclusions
233
was implied in many reports about large benefit claims. Examples of a

discourse of deviance include articles in the Daily Mail reporting that
homeless people in London were part of migrant begging schemes
(Sect. 8.2) and an article about a homeless person who lived in their
mother’s relatively-expensive house (Sect. 2.3).
The newspaper positively evaluates individuals who are seen to
embody neoliberal principles. There are multiple reports of people
going to extreme lengths to attract potential employers—such as wear-
ing a sandwich board to advertise their availability for work (Sect. 6.3).
In this example, the tone of the article is not sympathetic to the person
because they are using desperate measures to find employment and/or
because walking around wearing a sandwich board could be perceived
as an embarrassing thing to do. Rather they are praised and presented as
the antithesis of the idle poor. The Daily Mail positions punitive bene-
fits policies as standing in opposition to such entrepreneurial exercises,
as seen in the example where a job applicant was unable to attend a
job interview because of a scheduled appointment at the Jobcentre
(Sect. 8.3).
The Daily Mail takes the position that, if you have a welfare sys-
tem for the poor, it will breed laziness and/or complacency and it will
be abused. This position is evidenced in the reports of large Housing
Benefit claims, the ranking of areas with the highest number (and low-
est number) of benefits claimants, and reports on claiming disability
benefits for stress and alcoholism (Sect. 7.2). The latter example also
relates to the notion of ‘luxuries’ and links to flawed consumerism, by
referring to the consumption of alcohol. The newspaper ultimately sup-
ports Conservative-led welfare reform policies and takes the position
that cuts to benefits will be an incentive for recipients to work. It notes
that there are jobs available in particular locations, such as Aberdeen,
and reports that unemployment figures are falling, with the implication
that jobs are available if people are willing to be flexible and/or travel
(see the Cardiff example, discussed above). However, there is little real
consideration of where jobs are available, and the fact that not everyone
is able to move around the country for work.
The overarching notion that the benefits system is systematically
abused allows the Daily Mail to construct UK taxpayers (who it assumes
234
to be its readers) as the ultimate victims of poverty and the related

need for the welfare state; they are the ones paying for it, while others
are benefiting. Such a position supports the right-wing politics of the
Daily Mail and the examples presented in the newspaper are given as
evidence of the need for a smaller state and, relatedly, decreased taxa-
tion. This position does not, however, take into account the fact that a
large majority of benefit payments relate to the state pension, and there
is little to no acknowledgement in the Daily Mail of in-work poverty.
Rather, a dichotomy between benefits recipients and ‘hardworking peo-
ple’ is established.
The Daily Mail also presents UK poverty as associated with and
implicitly cause by immigrants. Ultimately the blame is placed on the
immigration policies of previous (Labour) governments and the UK’s
acceptance of immigrants from around the EU. The Daily Mail’s posi-
tion on the relationship between poverty and immigration is reflected
in their reporting of high welfare receipt in areas with large immigrant
populations, reports of children speaking a range of different languages
at school (here the children are implicitly indicative of a wider ‘prob-
lem’), and (ex)migrants with high Housing Benefit claims. The claims
reported in our PNCs are legitimate—we were focused on poverty, not
criminality. However, the Daily Mail ’s references to benefit fraud and
other criminal activities in texts concerning poverty highlights that dis-
courses do not exist in isolation. Rather, they are deployed in combina-
tion with other discourses—such as discourses of deviance—to imply
(and sometimes explicitly state) links between concepts. Examining
these links is beyond our focus on discourses of poverty and place, but
understanding what other discourses were drawn upon in our data
(discourses of criminality, anti-immigration discourses, etc.) gives an
indication of the wider scope of discussion within which discourses of
poverty sit.
Acknowledging these related discourses also demonstrates scope for
future research. As Fairclough and Wodak (1997: 276) rightly argue,
discourses ‘are always connected to other discourses which were pro-
duced earlier, as well as those which are produced synchronically and
subsequently’. No discourse, indexing an ideological position, is held in
a vacuum independent of other semiotic resources and/or language use.
9 Conclusions
235
For example, discourses relating to race can link to discourses of migra-

tion, which may overlap with discourses of poverty and the undeserv-
ing poor. Whilst Fairclough and Wodak’s position concerns abstract or
conceptual links between discourses, GTA shows that there can also be
measurable physical links between the deployment of particular ideolo-
gies and their use in relation to physical space. Thus, repeated references
to poverty (or any social phenomenon) can be used to encapsulate and
perpetuate particular ideologies about a particular locale.
9.3 Poverty and Place in the Guardian

In contrast to the Daily Mail, the Guardian takes the position that
poverty is caused by societal structures, not individuals’ faults or per-
sonal failings. The poor have drawn the short straw in an unequal sys-
tem; they are not lazy or entitled. Fundamentally, taking this position
means accepting that there is an unequal system of distribution and,
relatedly, it must be accepted that to change the system would allevi-
ate poverty. The Guardian positions its readers and the wider public as
opposed to welfare reform, hence its coverage of anti-cuts protests. In
many cases, the newspaper reports on planned demonstrations against
government policy, which has the added result, not only of making their
readers aware of such protests, but of publicising these demonstrations.
The Conservative-led government’s welfare reform policies are blamed
for poverty and they are linked explicitly to place by the repetition of
reports of Iain Duncan Smith visiting the Easterhouse area of Glasgow
(see Sect. 9.4). Here a specific place is strongly associated with poverty
and government policy.
Benefits statistics are not sensationalised to angle for welfare cuts and
poverty is depicted as clustering in particular areas due to lack of oppor-
tunity, jobs, and access to resources—a discourse of disadvantage. The
government is also blamed for poverty clusters in London, where the ben-
efits cap is presented as making (parts of) London no-go areas for many
people (Sects. 6.4, 7.3). Thus, it is implied that people should be allowed
equal opportunity to travel, live, and work anywhere and that government
policy is effectively restricting their geography (and their movement).
236
This is in direct contrast to the Daily Mail ’s treatments of benefits claim-

ants living in affluent areas (Sect. 6.4).
The Guardian also seems to associate the public sector—particularly
Jobcentres and councils—with poverty prevention/alleviation, and it
sees government initiatives, such as welfare reform, fitness to work reas-
sessments, and changes to the benefits system as compounding poverty,
rather than alleviating it. This is evidenced somewhat through repeated
references to public sector cuts; more cuts equals fewer jobs equals more
unemployed people (Sect. 6.3). However, it is worth considering why
the Guardian focuses so strongly on cuts to council jobs, as opposed to
the private sector job losses reported by the Daily Mail. The newspaper
seems particularly focused on the poverty experienced (or threatening)
its (largely) middle class readership, many of whom potentially work in
the public sector. The Guardian ’s related focus on poverty in relatively
affluent areas was also unexpected. This focus, most clearly shown in its
focus on anti-poverty initiatives in York (Sects. 4.3, 8.4), contributes to
the conceptualisation of poverty as a relative, qualitative, phenomenon.
It also connects poverty to shame (cf. Jo 2013) through the use of the
term ‘hidden poverty’. This term suggests that poverty is not what peo-
ple think it is (i.e. absolute poverty or extreme relative poverty in areas
of high unemployment and benefits receipt), nor is it actually what is
measured by official statistics. The Guardian seems open to accept that
people can be poor without meeting particular thresholds and, further-
more, people can be poor in areas of relatively high employment.
The Guardian does not draw on negative stereotypes of the poor
and/or the working class when discussing poverty. However, it explic-
itly acknowledges negative representations of the working class in other
media outlets (such as in poverty porn programming, and in right-wing
newspapers) and argues that the so-called ‘benefits culture’ or ‘benefits
lifestyle’ does not exist. It is sympathetic to those who use foodbanks and
those receiving benefits (particularly Jobseekers Allowance and Incapacity
Benefit). It rejects the idea that migrants are ‘benefit tourists’ who also
contribute to poverty rates by taking jobs, etc. (see Sect. 7.1). The
newspaper’s reporting on poverty and place does not tend to mention
illegitimate benefits claims; there is silence on fraud for the most part.
9 Conclusions
237
Additionally, no strong links are made between poverty and place and
areas of high rates of migrant settlement.
Having established the ideologies drawn upon, the discourses used,
and the different assumptions held implicit by each newspaper, we can
argue that—based on our analysis—place is used to shape how each
newspaper presents poverty to their readers. The Daily Mail endorses
a neoliberal position and paints the poor as predominantly idle and
undeserving. There are some realisations of flawed consumerism—par-
ticularly in reports about expensive London houses paid for by Housing
Benefit (especially for (ex)migrant families) and in reports about peo-
ple receiving benefits who are planning their wedding and buying fly-
ing lessons (Sect. 6.3)—but this was not a strong trend in the data. By
contrast, the Guardian takes a more socialist position—advocating for
a more equal distributions of wealth, but it also seeks to reinforce the
concept of relative poverty, positioning it as a problem not only for the
working classes, but also for those who are employed in traditionally
middle class professions.
9.4 A Spotlight on Place

Both newspapers show geographical biases in their reporting of poverty
that are over and above the geographical biases found in their report-
ing more generally (calculated using the newspapers’ background geog-
raphies, Sect. 4.5). Around half of the Guardian ’s reporting that links
poverty with place focuses on places in London; for the Daily Mail this
is around a third. Part of the reason for this is that the news agenda
in both newspapers is generally London-centric—as evidenced by the
background geography of each newspaper. However, even once this has
been taken into account, places within London are associated with pov-
erty more than would be expected. Outside London, there are also some
clear biases. Major urban centres also tend to be particularly associated
with poverty, particularly Liverpool.
Other places with high rates of poverty, be they urban peripheries,
former industrial towns, or rural areas, tend to be overlooked unless
there is a clear reason to talk about them. Burnley, a deprived former
238
mill town, receives large amounts of attention because of the trial of

Incapacity Benefit reassessment in the town. Middlesbrough (another
deprived former industrial town) and its surrounding area are associ-
ated with poverty, at least in part, because of the second series of Benefits
Street, and Margate, a seaside town, is linked to poverty because of Mary
Portas’ work there. Other locations with similar economic and poverty
profiles as these three places are frequently ignored by both newspapers.
It is also interesting that even within London, places with similar
poverty profiles are treated very differently. From their different per-
spectives, both newspapers focus on Tower Hamlets while neighbour-
ing Newham, with similar levels of poverty and a similar demographic
profile, is largely overlooked. One possible explanation for this is that
Tower Hamlets consists of much of London’s traditional East End and
therefore has a long tradition of being associated with poverty. A second
possibility is that Tower Hamlets is where wealthy London (including
the City and Canary Wharf ) is adjacent to deprived London, and thus
the poverty is more visible.
The newspapers’ focus on poverty in London, at the expense of the
rest of the UK, may have significant implications for our understand-
ing of poverty and the measures required to alleviate it. London’s econ-
omy is thriving so the city has low unemployment but a serious lack
of affordable housing. This means that people in London experience
poverty in a very different way to people in other parts of the country
where poverty is more commonly associated with industrial decline or
stagnation leading to unemployment and a lack of opportunity. London
is seen as the exception—it is the site of a housing crisis and it includes
areas of extreme poverty—but it is reported as the rule. The media dis-
course, from both left- and right-wing perspectives, is dominated by
London and London’s problems. Added to this, decision makers, par-
ticularly politicians, are based on London and may have little experience
of poverty as realised elsewhere, and the policy makers are not immune
to the representations of poverty presented by the mass media. This
suggests that much of the perception of poverty and responses to it are
driven by London-based agendas that have limited relevance to the rest
of the UK. Such a claim is evidenced by reports that when Iain Duncan
Smith visited Easterhouse in Glasgow he was so shocked by what he saw
9 Conclusions
239
that it led to significant government policy initiatives that were very dif-
ferent to many of the existing debates around poverty.
9.5 Evaluating Geographical Text Analysis

As well as providing a new way to interrogate the discursive construc-
tion of UK poverty, this book also has major methodological implica-
tions which have the potential to impact the analysis of any topic in
human geography or other subjects where there is an interest in the
geographies within texts. The methods used here allow us to identify
place-names and the themes associated with them in very large cor-
pora. While the idea of ‘distant reading’ of large digital corpora is not
new (Moretti 2005), there has been a backlash against it which argues
that such sources cannot be understood without a detailed knowledge
of the source, determined by a close reading of at least parts of it (e.g.
Hammond et al. 2016).
In this book, we have demonstrated that using GTA means that these
two approaches—distant reading and close reading—are not incom-
patible. Indeed mapping allows and encourages a ‘multiscale approach’
(Taylor et al. in press) in which the aggregate analyses map and describe
the broad patterns of where the text (or corpus) is talking about in rela-
tion to a particular theme. These maps then drive questions about why
the texts in a corpus refer to these places, what is being said about them,
and what places are not mentioned in the corpus. Subsequent analyses,
including the close reading of concordance lines, allows researchers to
answer these questions and thus develop a detailed understanding of
the geography of the topic, theme, or (social) issue under study, which
includes understanding how and why there are differences between
places, sources, and, potentially, time points. This approach can be
applied to any theme for which suitable digital corpora are available.
Given the dramatic rise in the availability in digital material (Risen and
Lichtblau 2013), much of it in textual form, this opens up huge poten-
tial for new forms of analysis.
As well as human geography, our approach also has significant impli-
cations for corpus linguistics and CDA. Previously, these fields have had
240
little interaction with geography. GTA allows the researcher to first map
the place-names within the text and then to conduct more sophisticated
analyses based on location. The similarities and differences between the
two newspapers set out above would arguably have been made visible
by other methods of analysis, such as CDA or corpus-based discourse
analysis, but what the present study has shown is how these underlying
ideologies are promoted and endorsed by the choice of each newspaper
to report on particular places. Place, something which may be seen as
relatively innocuous and easily passed over in linguistic analysis, plays
an important part in media-endorsed conceptualisations of what pov-
erty in twenty-first century Britain is, what it looks like, its causes, and
who is affected. As such, and as we have shown here, critical studies of
social phenomena, such as poverty, can benefit greatly from a close con-
sideration of place.
Whilst the availability of geoparsers suitable for this type of analysis
is currently somewhat restricted, this is testament to the state of the art.
In the first instance, we would encourage researchers interested in GTA
to work interdisciplinarily with scholars who have access to such tools.
Secondarily, it is hoped that as the method becomes more developed
and widely-known more geoparsers (and related tools) will become
available for widespread use. Additionally, it is hoped that future work
describing and documenting the concordance geoparsing process will
make the method more accessible to those without backgrounds in GIS
and/or corpus linguistics.
Methodologically, this book provides two particular innovations. The
first is the use of collocation analysis, with some manual intervention,
to establish which search terms would be fruitful for the exploration of
a particular theme (in a given corpus). The analysis of collocates—across
multiple corpora in this instance—enables the researcher to gain a
greater understanding of the language associated with a particular topic,
based on a combination of prior knowledge and a systematic analysis
of the corpus itself. In this book the generation of lexical collocates for
<*poverty*> led to us adding a significant number of search terms that
we had not previously considered as relevant to the theme of poverty
and place. The second innovation is the establishing of background
geographies for our corpora using a concordance geoparsed sample of
9 Conclusions
241
instances of the. This allowed us to generate an understanding of the

geography of each corpus as a whole without needing to geoparse them
in their entirety. This, in turn, enables us to identify places that were
more or less strongly associated with a particular theme than would be
expected given the background geography. These two methodological
innovations can be applied to any corpus for the purposes of GTA at
the point of data selection (the collocation analysis) and analysis (the
comparison with background geography).
There are a number of areas where our methods could be developed
further. The first concerns ongoing development of the geoparser, which
could lead to a more robust and trainable tool. This would decrease
the need for manual checking of concordance lines. A more signifi-
cant development would be to move beyond what it is currently pos-
sible to geoparse automatically. In terms of the current state of the art,
a geoparser identifies place-names and thus, any subsequent analysis
of geography is solely the analysis of the geography of named places.
Assigning locations (and ultimately coordinates) to uses of deictic terms,
such as location markers like here and there, and generic references,
such as in the city, the town, or this place, is beyond the remit of exist-
ing geoparsers (and automatic geoparsing techniques). This potentially
means that apparent differences in geography could be influenced by
differences in writing style. For example, if texts from one source, show
a tendency to use a proper noun, such as Newcastle, at the beginning
of an article, but switch to deictic references throughout the rest of the
article, the frequency of mentions of Newcastle would be different when
compared with another source that uses Newcastle consistently through-
out. The use of deictic terms may influence the frequency or density of
place-name mentions, and this needs to be kept in mind when viewing
maps generated using place-names only. Developing the ability to go
beyond geoparsing place-names and to be able to handle deixis would
provide valuable new abilities to GTA and would illuminate our under-
standing of both geographies and writing styles.
A final limitation to consider is that GTA does not take into account
additional semiotic resources beyond text. Whilst this is unlikely to
influence the maps generated using GTA it could have a significant
impact on the analysis of discourse. As noted above, for the most part
242
we found no trend suggesting the explicit demonization of the poor

in our corpora, and we have argued that the newspapers could leave
particular evaluations implicit because of their readers’ wider aware-
ness of media representations of the poor. However, implicit references
within a text are not the only way that allusions to particular ideolo-
gies about poverty and the poor can be made. For example, a report
about high welfare receipt in a particular area does not need to reference
what benefits recipients spend their money on explicitly if the article
is accompanied by an image—usually displayed prominently near the
headline—which includes someone drinking alcohol, smoking a ciga-
rette, or wearing a particular brand of clothing. Readers of online news-
paper texts do not get the meaning of an article solely from the text
alone. This points to a limitation of using textual data only. Whilst we
have adapted textual data for use in GIS using tools from corpus lin-
guistics, multimodal corpora are still in their infancy and, whilst we
could geotag images, their content would have to be manually analysed
in order to determine whether there was a relationship between images
and place. This is an area for further research, but it currently lies
beyond the boundaries of corpus linguisticsand GIS. Therefore, analyses
using GTA must be aware that the text—although primary—is not the
only semiotic resource which can index ideologies.
There is, therefore much potential for future research. Some of this is
technical. As well as improved geoparsing and the ability of geoparsers
to automatically cope with deixis, there are also areas where the meth-
ods used to analyse the data, from both spatial and corpus perspectives,
could be further developed. More important, however, is the wider
application of GTA beyond the boundaries of human geography, CDA,
and applied linguistics. As stated above, the methods demonstrated
herein can be applied to any research areas where there is a corpus of
digital texts and an interest in the geographies they contain. This means
that there is the potential for a step-change in our understanding of the
geographies within texts. GIS-based analyses are quantitative by nature
but, as described in this book, GTA offers the ability to apply GIS
and digital methods to textual sources. This challenges the notion that
such methods and tools are merely quantitative in nature, and bridges
the gap between quantitative analysis and traditionally qualitative close
reading approaches.
9 Conclusions
243
9.6 Summary
Through our exemplar analysis of UK poverty in two corpora of
national newspaper texts we have demonstrated the benefits of using
GTA for analysing the linguistic representation of a social phenom-
enon. Fundamentally, we have shown how a consideration of geogra-
phy can illuminate how different aspects of a given issue are highlighted
in relation to place, to concretise abstract concepts such as poverty by
anchoring their lived experience in geographical space. We accepted the
argument that the mass media plays a role in the construction of British
society: media outlets inform us who to vote for, who to cheer for, who
to vilify, and who to laugh at, newspaper articles draw our attention
to stories of crime, deviance, and violence, etc. Following Fairclough
(2001: 30) we note that the ‘constant doses of “news” which most peo-
ple receive each day are a significant factor in social control, and they
account for a not insignificant proportion of a person’s average daily
involvement in discourse’ (referring here to the production, distribu-
tion, and consumption of different texts). We have shown how the read-
ers of each of our newspapers are being presented with a different view
of what UK poverty is, where it is, what causes it, and what can alleviate
it. Part of the role of a critical discourse analyst is to interrogate lan-
guage in use and ‘make what may be non-obvious evident’ (O’Halloran
2012: 92). Our analysis of discourses of poverty and place has shown
how each newspaper presents poverty but, more importantly, how their
representation of poverty is filtered through the lens of place—a topic
rarely explicitly and systematically considered within CDA. Ultimately
the versions of UK poverty presented by each newspaper may influence
their readers’ opinions about the poor, their stance on poverty, and how
(or indeed if ) they believe UK poverty should be tackled, and by whom.
References
Fairclough, N. 2001. Language and Power (second edition). London: Longman.
Fairclough, N., & R. Wodak. 1997. Critical Discourse Analysis. In T. A. van
Dijk (ed.). Discourse as Structure and Process: A Multi-disciplinary Approach.
London: Sage, pp. 258–284.
244
Hammond, A., A. J. Brooke, & G. Hirst. 2016. Modeling Modernist

Dialogism: Close Reading with Big Data. In S. Ross & J. O’Sullivan (eds.).
Reading Modernism with Machines: Digital Humanities and Modernist
Literature. Basingstoke: Palgrave, pp. 49–78.
London: Oneworld.
Moretti, F. 2005. Graphs, Maps, Trees. Verso: London.
O’Halloran, K. 2012. Electronic Deconstruction: Revealing Tensions in the
Cohesive Structure of Persuasion Texts. International Journal of Corpus
Linguistics 17 (1): 91–124.
Risen, J., & E. Lichtblau. 2013. How the U.S. Uses Technology to Mine More
Data More Quickly. New York Times 8/6/2013.
Taylor, J. E., I. N. Gregory, & C. Donaldson. in press. Combining Close
and Distant Reading: A Multiscalar Analysis of the English Lake District’s
Historical Soundscape. International Journal of Humanities and Arts
Computing.
Tyler, I. 2008. ‘Chav mum, chav scum’: Class Disgust in Contemporary
Britain. Feminist Media Studies 8 (1): 17–34.
References
Altheide, D. L. 2000. Tracking Discourse and Qualitative Document Analysis.

Poetics 27 (4): 287–299.
Atkinson-Small, J. 2012. It’s Time to Pull up the Drawbridge: Ban
Immigration for Five Years and Give Our Youth a Chance. Mail Online
21/2/2012. http://www.dailymail.co.uk/debate/article-2104253/UK-ban-
immigration-5-years-youth-chance.html. Accessed 13/3/2018.
273–306.
Baker, P., & T. McEnery (eds). 2015a. Corpora and Discourse Studies.
Baker, P., & T. McEnery. 2015b. Who Benefits When Discourse Gets
Democratised? Analysing a Twitter Corpus Around the British Benefits
Street Debate. In P. Baker & T. McEnery (eds.). Corpora and Discourse
Studies: Integrating Discourse and Corpora. Basingstoke: Palgrave,
pp. 244–265.
© The Editor(s) (if applicable) and The Author(s) 2019 245

https://doi.org/10.1007/978-3-319-93503-4
246
References
Ballinger, L. 2009. Taxpayers Pay £1,600-a-Week for Family of Ex-asylum

Seekers to Live in Luxury Five-Storey Home. Mail Online 20/11/09. http://
www.dailymail.co.uk/news/article-1231795/Taxpayers-pay-1-600-week-
family-ex-asylum-seekers-live-luxury-storey-home.html. Accessed 22/1/2018.
University Press.
Bednarek, M. 2006. Evaluation in Media Discourse: Analysis of a Newspaper
Corpus. New York: Continuum.
Bennett, J. 2012. ‘And What Comes Out May Be a Kind of Screeching’: The
Stylisation of Chavspeak in Contemporary Britain. Journal of Sociolinguistics
16 (1): 5–27.
Berman, M. L., R. Mostern, & H. Southall. 2016. Introduction. In M. L.
Berman, R. Mostern, & H. Southall (eds.). Placing Names: Enriching and
Integrating Gazetteers. Bloomington: Indiana University Press, pp. 1–11.
Besag J., & J. Newell. 1991. The Detection of Clusters in Rare Diseases.
Journal of the Royal Statistical Society Series A 154: 143–155.
Biber, D., S. Conrad, & R. Reppen. 1994. Corpus-Based Approaches to Issues
in Applied Linguistics. Applied Linguistics 15 (2): 169–189.
Biressi, A. 2011. ‘The Virtuous Circle’: Social Entrepreneurship and Welfare
Programming in the UK. In H. Wood & B. Skeggs (eds.). Reality Television
and Class. London: British Film Institute/Palgrave Macmillan, pp. 144–155.
Block, D. 2013. Social Class in Applied Linguistics. London: Routledge.
Bourdieu, P. 1986/2010. The Forms of Capital. In I. Szeman & T. Kaposy
(eds.). Cultural Theory: An Anthology. Oxford: Wiley Blackwell, pp. 81–93.
Bourdieu, P., & L. Wacquant. 2004. Symbolic Violence. In N. Scheper-
Hughes & P. Bourgois (eds.). Violence in War and Peace: An Anthology.
Oxford: Blackwell, pp. 272–274.
Breit, E. 2010. On the (Re)construction of Corruption in the Media: A
Brezina, V. 2018. Statistics in Corpus Linguistics: A Practical Guide. Cambridge:
Cambridge University Press.
Chainey, S., & J. Ratcliffe. 2005. GIS and Crime Mapping. Chichester: Wiley.
Chambers, R. 2006. What Is Poverty? Who Asks? Who Answers? Poverty
in Focus. International Poverty Centre: United Nations Development
Programme. http://www.ipc-undp.org/pub/IPCPovertyInFocus9.pdf. Accessed
23/9/2015.
Conrad, S. 2002. Corpus Linguistic Approaches for Discourse Analysis.
Annual Review of Applied Linguistics 22: 75–95.
References
247
Cooper, D., & I. Gregory. 2011. Mapping the English Lake District: A
Literary GIS. Transactions of the Institute Of British Geographers 36 (1):
89–108.
Cope, M., & S. Elwood (eds). 2009. Qualitative GIS: A Mixed Methods
Approach. London: Sage.
Cresswell, T. 2004. Place: A Short Introduction. London: Blackwell.
Cromley, E. K., & S. L. McLafferty. 2012. GIS and Public Health. London:
Guilford.
DeMers, M. 2008. Fundamentals of Geographical Information Systems (fourth
edition). Chichester: Wiley.
Department for Communities and Local Government. 2017. Policy Paper:
2010 to 2015 Government Policy: Support for Families. https://www.
gov.uk/government/publications/2010-to-2015-government-policy-sup-
port-for-families/2010-to-2015-government-policy-support-for-families.
Accessed 2/11/2017.
Diggle P., P. Zheng, & P. Durr. 2005. Nonparametric Estimation of Spatial
Segregation in a Multivariate Point Process: Bovine Tuberculosis in
Cornwall, UK. Journal of the Royal Statistical Society Series A 54: 645–658.
Donaldson C., I. N. Gregory, & J. E. Taylor. 2017. Locating the Beautiful,
Dorling, D., & D. Fairburn. 1997. Mapping: Ways of Representing the World.
Harlow: Longman.
Dreier, P. 2005. How the Media Compound Urban Problems. Journal of
Urban Affairs 27 (2): 193–201.
DWP. 2015. Fraud and Error in the Benefits System. https://www.gov.uk/gov-
ernment/uploads/system/uploads/attachment_data/file/575407/fraud-and-
error-stats-release-2015-16-final-estimates.pdf. Accessed 13/8/2016.
DWP. 2016. Freedom of Information Requests 2016-398/2026-397. https://
www.gov.uk/government/uploads/system/uploads/attachment_data/
file/502566/389-397-2016.pdf. Accessed 13/11/2016.
DWP. 2018. Housing Benefit Caseload Statistics. https://www.gov.uk/govern-
ment/uploads/system/uploads/attachment_data/file/659362/housing-bene-
fit-caseload-data-to-august-2017.ods. Accessed 23/3/2018.
248
References
Fairclough, N. 2001. Language and Power (second edition). London:

Longman.
Fairclough, N., & R. Wodak. 1997. Critical Discourse Analysis. In T. A. van
Dijk (ed.). Discourse as Structure and Process: A Multi-disciplinary Approach.
London: Sage, pp. 258–284.
Fotheringham, A. S., & F. Zhan. 1996. A Comparison of Three Exploratory
Methods for Cluster Detection in Spatial Point Patterns. Geographical
Analysis 28: 200–218.
Fotheringham, A. S., C. Brusdon, & M. Charlton. 2000. Quantitative
Fukuda-Parr, S. 2006. The Human Poverty Index: A Multidimensional Measure.
Poverty in Focus. International Poverty Centre: United Nations Development
Programme. Online. http://www.ipc-undp.org/pub/IPCPovertyInFocus9.pdf.
Galbraith, J. K. 1958. The Affluent Society. London: Penguin.
Geonames. n.d. Geonames Homepage. http://www.geonames.org. Accessed
11/7/2017.
Getty Research Institute. n.d. Getty Thesaurus of Geographic Names. Online.
http://www.getty.edu/research/tools/vocabularies/tgn. Accessed 11/7/2017.
Goodchild, M. F. 1992. Geographical Information Science. International
Journal of Geographical Information Systems 6: 31–45.
Gordon, D. 2006. The Concept and Measurement of Poverty. In C. Pantazis,
D. Gordon, & R. Levitas (eds.). Poverty and Social Exclusion in Britain.
Bristol: The Policy Press, pp. 29–70.
Gordon, D. 2011. Consultation Response: Social Mobility and the Child
Poverty Review. Policy Response Series 2, PSE 2011, pp. 4–5. http://www.
poverty.ac.uk/pse-research/pse-uk/policy-response.
Gov.UK. n.d. Benefit Cap. https://www.gov.uk/benefit-cap. Accessed 23/11/
2017.
Gregory, I. N. 2009. Comparisons Between Geographies of Mortality and
Deprivation from the 1900s and 2001: Spatial Analysis of Census and
Mortality Statistics. British Medical Journal 339, b3454: 676–679.
Cartographies of Lake District literature. In D. Cooper, C. Donaldson, &
P. Murrieta-Flores (eds.). Literary Mapping in the Digital Age. Abingdon:
References
249
Gregory, I. N., C. Donaldson, P. Murrieta-Flores, & P. Rayson. 2015.

Geoparsing, GIS and Textual Analysis: Current Developments in Spatial
Humanities Research. International Journal of Humanities and Arts
Gregory, I. N., K. Kemp, & R. Mostern. 2003. Geographical Information and
Historical Research: Current Progress and Future Directions. History and
Grover, C., R. Tobin, K. Byrne, M. Woollard, J. Reid, S. Dunn, & J. Ball.
2010. Use of the Edinburgh Geoparser for Georeferencing Digitized
Historical Collections. Philosophical Transactions of the Royal Society A 368:
3875–3889.
Hammond, A., A. J. Brooke, & G. Hirst. 2016. Modeling Modernist
Dialogism: Close Reading with Big Data. In S. Ross & J. O’Sullivan (eds.).
Reading Modernism with Machines: Digital Humanities and Modernist
Literature. Basingstoke: Palgrave, pp. 49–78.
Hansard. 2014. Topical Questions, 13 January, Vol. 573. https://han-
sard.parliament.uk/Commons/2014-01-13/debates/14011313000007/
WorkAndPensions. Accessed 4/1/2018.
Hardie A. 2012. CQPweb—Combining Power, Flexibility and Usability
in a Corpus Analysis Tool. International Journal of Corpus Linguistics 17:
380–409.
Hardt-Mautner, G. 1995. Only Connect. Critical Discourse Analysis and
Corpus Linguistics. UCREL Technical Papers 6. http://ucrel.lancs.ac.uk/
papers/techpaper/vol6.pdf. Accessed 18/10/2009.
Heywood, I., S. Cornelius, & S. Carver. 2011. An Introduction to Geographical
Information Systems (fourth edition). Pearson: Harlow.
Ipsos MORI. 2013. Perceptions Are Not Reality. https://www.ipsos.com/ipsos-
mori/en-uk/perceptions-are-not-reality. Accessed 15/7/2016.
Jackson, J. 2016. Independent, Mirror, Express and Star Suffer Sharp Fall in
Traffic. The Guardian 17/3/2016. https://www.theguardian.com/media/2016/
mar/17/independent-mirror-express-and-star-suffer-sharp-fall-in-traffic.
Accessed 12/5/2017.
Jiang, H., & J. R. Eastman. 2010. Application of Fuzzy Measures in Multi-
criteria Evaluation GIS. International Journal of Geographical Information
Science 14 (2): 173–184.
250
References

University Press.
Kennedy, G. D. 1998. An Introduction to Corpus Linguistics. London:
Longman.
Kress, G. 1994. Text and Grammar as Interpretation. In U. Meinhof & K.
Richardson (eds.). Text, Discourse and Context: Representations of Poverty in
Britain. London: Longman, pp. 24–47.
Kress, G., & T. van Leeuwen. 2001. Multimodal Discourse: The Modes and
Media of Contemporary Communication. London: Arnold.
Kulldorf, M. 1997. A Spatial Scan Statistic. Communications in Statistics:
Theory and Methods 26: 1481–1496.
Kwan, M-P. 2008. From Oral Histories to Visual Narratives: Re-Presenting the
Post-September 11 Experiences of the Muslim Women in the United States.
Social and Cultural Geography 9 (6): 653–669.
London: Oneworld.
Lee, P. 1999. Where Are the Deprived? Measuring Deprivation in Cities and
Regions. In D. Dorling & S. Simpson (eds.). Statistics in Society. London:
Arnold, pp. 172–180.
Lister, R. 2004. Poverty. Cambridge: Polity Press.
Lo, C. P., & A. K. W. Yeung. 2007. Concepts and Techniques of Geographical
Information Systems (second edition). Toronto: Pearson.
Lorenzo-Dus, N., & S. Marsh. 2012. Bridging the Gap: Interdisciplinary
Insights into the Securitization of Poverty. Discourse and Society 23 (3):
274–296.
Mack, J., & S. Lansley. 1985. How Poor Is Too Poor? Defining Poverty. In J.
Mack & S. Lansley (eds.). Poor Britain. London: George Allen & Unwin.
Malpas, J. E. 1999. Place and Experience: A Philosophical Topography.
Cambridge, MA: MIT Press.
Mautner, G. 2007. Mining Large Corpora for Social Information: The Case of
Elderly. Language in Society 36 (1): 51–72.
Mautner, G. 2009. Checks and Balances: How Corpus Linguistics Can
Contribute to CDA. In R. Wodak & M. Meyer (eds.). Methods of Critical
Discourse Analysis. London: Sage, pp. 122–143.
References
251
McCarthy, M., & R. Carter. 2004. Focus-on Issue: Corpus Linguistics (Part
I)—Introduction. Journal of Pragmatics 36 (2): 147–148.
McKay, S. 2004. Poverty or Preference: What Do ‘Consensual Deprivation
Indicators’ Really Measure? Fiscal Studies 25 (2): 201–223.
Meinhof, U. 1994. From the Street to the Screen: Breadline Britain and the
Semiotics of Poverty. In U. Meinhof & K. Richardson (eds.). Text, Discourse
and Context: Representations of Poverty in Britain. London: Longman, pp.
65–92.
Misturelli, F., & C. Heffernan. 2011. The Language of Poverty: An
Exploration of the Narratives of the Floor. Sustainable Development 19 (3):
206–222.
Moretti, F. 2005. Graphs, Maps, Trees. Verso: London.
Selected Deprivation Indexes. Journal of Public Health Medicine 13 (4):
318–326.
Murrieta-Flores, P., A. Baron A., I. N. Gregory, A. Hardie, & P. Rayson. 2015.
GIS 19: 296–320.
Newton, J. 2015. More Than 100,000 Benefit Claimants Across the UK
Are Being Handed £100 a Week After Being Assessed as Unable to Work
Because of Addiction, Obesity and Stress. Mail Online 22/5/15. http://
www.dailymail.co.uk/news/article-3006248/More-100-000-benefits-
claimants-UK-handed-100-week-assessed-unable-work-addiction-obesity-
stress.html. Accessed 14/3/2018.
Noble, M, G. Wright, G. Smith, & C. Dibben. 2006. Measuring Multiple
Deprivation at the Small-Area Level. Environment and Planning A 38:
169–185.
O’Halloran, K. 2012. Electronic Deconstruction: Revealing Tensions in the
Cohesive Structure of Persuasion Texts. International Journal of Corpus
Linguistics 17 (1): 91–124.
OED. 2017. Poverty, n. OED Online. www.oed.com/view/Entry/149126.
Accessed 2/11/2017.
252
References
ONS. 2014. 2011 Census Glossary of Terms. https://www.ons.gov.uk/ons/

guide-method/census/2011/census-data/2011-census-data/2011-first-re-
lease/2011-census-definitions/2011-census-glossary.pdf. Accessed 24/2/2017.
ONS. 2016a. Unemployment Rate: Sourced from the Labour Market Statistics
Time Series Dataset. https://www.ons.gov.uk/employmentandlabourmarket/
peoplenotinwork/unemployment/timeseries/mgsx/lms. Accessed 2/12/2016.
ONS. 2016b. Statistical Bulletin: UK Labour Market: Nov 2016. https://
www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employ-
mentandemployeetypes/bulletins/uklabourmarket/november2016. Accessed
2/12/2016.
ONS. n.d.(a) Unemployment. https://www.ons.gov.uk/employmentandlabour
market/peoplenotinwork/unemployment. Accessed 2/1/2018.
ONS. n.d.(b) Households by Combined Economic Activity Status of
Household Members by Local Authority: Table A1 LA. https://www.ons.
gov.uk/employmentandlabourmarket/peoplenotinwork/unemployment/
datasets/householdsbycombinedeconomicactivitystatusofhouseholdmem-
bersbylocalauthoritytablea1la/current. Accessed 24/2/2017.
Openshaw, S. 1984. The Modifiable Areal Unit Problem. Concepts and
Techniques in Modern Geography 38. Norwich: Geo Books.
Openshaw, S. 1997. The Truth About Ground Truth. Transactions in GIS 2:
2–24.
Openshaw, S., M. Charlton, C. Wymer, & A. Craft. 1987. A Mark 1
Geographic Analysis Machine for the Automated Analysis of Point Data
Sets. International Journal of Geographical Information Systems 1: 335–358.
Ordnance Survey. 2014. The National Grid FAQs. https://www.ordnancesur-
vey.co.uk/support/the-national-grid.html. Accessed 19/2/2016.
Orpin, D. 2005. Corpus Linguistics and Critical Discourse Analysis:
Examining the Ideology of Sleaze. International Journal of Corpus Linguistics
10 (1): 37–61.
OSF. 2014. A Report on Six European Cities: Europe’s White Working Class
Communities. New York: Open Society Foundations. http://www.open-
societyfoundations.org/sites/default/files/white-working-class-over-
view-20140616.pdf. Accessed 19/6/2014.
Pantazis, C., D. Gordon, & P. Townsend. 2006. The Necessities of Life. In C.
Pantazis, D. Gordon, & R. Levitas (eds.). Poverty and Social Exclusion in
Britain. Bristol: The Policy Press, pp. 89–122.
Paterson, L. L. Forthcoming. Electronic Supplement Analysis of Multiple
Texts: Exploring Discourses of UK Poverty in Below the Line Comments.
References
253

27 (2): 195–214.
Pickles, J. (ed.). 1995. Ground Truth: The Social Implications of Geographic
Information Systems. New York: Guilford Press.
Pleiades. n.d. Pleiades. https://pleiades.stoa.org. Accessed 11/7/2017.
Potts, A., M. Bednarek, & H. Cape. 2015. How Can Computer-Based
Methods Help Researchers to Investigate News Values in Large Datasets?
A Corpus Linguistic Study of the Construction of Newsworthiness in
the Reporting on Hurricane Katrina. Discourse & Communication 9 (2):
149–172.
Ramesh, R., & A. Sparrow. 2010. Housing Benefit Cuts Will ‘Push Poor Out
of South’, Experts Warn. The Guardian 8/11/2010. https://www.theguard-
ian.com/society/2010/nov/08/housing-benefit-north-south-divide. Accessed
22/2/2018.
Rayson, P. 2008. From Keywords to Semantic Domains. International Journal
of Corpus Linguistics 13 (4): 519–549.
Rayson P., D. Archer, S. Piao, & T. McEnery. 2004. The UCREL Semantic
Analysis System. Proceedings of the Beyond Named Entity Recognition
Semantic Labelling for NLP Tasks Workshop, Lisbon, Portugal, 2004, pp.
7–12.
Reibel, M. 2007. Geographic Information Systems and Spatial Data
Processing in Demography: A Review. Population Research Policy Review 26:
601–618.
Richardson, J. E. 2006. Analysing Newspapers: An Approach from Critical
Discourse Analysis. Basingstoke: Palgrave Macmillan.
Risen, J., & E. Lichtblau. 2013. How the U.S. Uses Technology to Mine More
Data More Quickly. New York Times 8/6/2013.
Robinson, A. H., J. L. Morrison, P. C. Muehrcke, A. J. Kimerling, & S. C.
Guptil. 1995. Elements of Cartography (sixth edition). Chichester: Wiley.
254
References
Rowlingson, B. 2015. Spatial Segregation Analysis. Personal Communication.

Rupp, C. J., P. Rayson, I. N. Gregory, A. Hardie, A. Joulain, & D. Hartmann.
2014. Dealing with Heterogeneous Big Data When Geoparsing Historical
Corpora. Proceedings of the 2014 IEEE Conference on Big Data, pp. 80–83.
Rushton, G. 2003. Public Health, GIS, and Spatial Analytic Tools. Annual
Review of Public Health 24: 43–56.
Sack, R. 1997. Homo Geographics. Baltimore: Johns Hopkins University Press.
Scheuer, J. 2003. Habitus as the Principle for Social Practice: A Proposal for
Critical Discourse Analysis. Language in Society 32 (2): 143–175.
Schuurman, N. 2006. Formalization Matters: Critical GIS and Ontology
Research. Annals of the Association of American Geographers 96 (4): 726–739.
Siebert, L. J. 2000. Using GIS to Document, Visualize, and Interpret Tokyo’s
Spatial History. Social Science History 24 (3): 537–574.
Simon, R., L. Isaksen, E. Barker, & P. de Soto Canamares. 2016. The Pleiades
Gazetteer and the Pelagios Project. In M. L. Berman, R. Mostern, & H.
Southall (eds.). Placing Names: Enriching and Integrating Gazetteers.
Bloomington: Indiana University Press, pp. 97–109.
Skeggs, B., & H. Wood (eds). 2011. Reality Television and Class. London:
Slack, J. 2010. Migrants ARE Driving Down Wages of the Poor, as Equalities
Watchdog Blames East European Influx. Mail Online 18/1/10. http://www.
dailymail.co.uk/news/article-1244060/Migrants-ARE-driving-wages-poor-
equalities-watchdog-blames-East-European-influx.html. Accessed 15/1/2018.
Southall, H. R., R. Mostern, & M. L. Berman. 2011. On Historical
Gazetteers. International Journal of Humanities and Arts Computing 5:
127–145.
Tanner, S., T. Munoz, & P. H. Ros. 2009. Measuring Mass Text Digitization
Quality and Usefulness: Lessons Learned from Assessing the OCR Accuracy
of the British Library’s 19th Century Online Newspaper Archive. DLib
Magazine 15 (7/8).
Taylor, J. E., I. N. Gregory, & C. Donaldson. in press. Combining Close
and Distant Reading: A Multiscalar Analysis of the English Lake District’s
Historical Soundscape. International Journal of Humanities and Arts
Computing.
Tognini-Bonelli, E. 2001. Corpus Linguistics at Work. Amsterdam: John
Benjamins.
References
255
Townsend, P., P. Phillimore, & A. Beattie. 1988. Health and Deprivation:

Inequality and the North. London: Routledge.
Turvill, W. 2015. National Newspaper Circulations, May 2015: Mail on
Sunday Overtakes Sun on Sunday, Times Remains Only Growing Title.
http://www.pressgazette.co.uk/national-newspaper-circulations-may-
2015-mail-sunday-overtakes-sun-sunday-times-remains-biggest-growing/.
Accessed 15/1/2016.
Tyler, I. 2008. ‘Chav mum, chav scum’: Class Disgust in Contemporary
Britain. Feminist Media Studies 8 (1): 17–34.
UN. 1995. The Copenhagen Declaration and Programme of Action. World
Summit for Social Development, 6–12 March. New York: United Nations.
Unwin, D. 1995. Geographic Information Systems and the Problem of ‘Error
and Uncertainty.’ Progress in Human Geography 19: 549–558.
Watson, L. 2011. Lotterty Hands Over £425,000 for Charity Food Bank in
Birmingham. Mail Online 22/9/2011. http://www.dailymail.co.uk/news/
article-2040024/Lottery-hands-425-000-charity-food-bank-Birmingham.
html. Accessed 21/2/2018.
Wodak, R. 1999. Critical Discourse Analysis at the End of the 20th Century.
Research on Language and Social Interaction 32 (1–2): 185–193.
Wodak, R. 2001. What CDA Is About—A Summary of Its History, Important
Concepts and Its Development. In R. Wodak & M. Meyer (eds.). Methods
of Critical Discourse Analysis. London: Sage, pp. 1–13.
Wodak, R., & M. Meyer. 2009. Critical Discourse Analysis: History, Agenda,
Theory and Methodology. In R. Wodak & M. Meyer (eds.). Methods of
Critical Discourse Analysis (second edition). London: Sage, pp. 1–33.
Wong, D. 2009. The Modifiable Areal Unit Problem. In A. S. Fotheringham
& P. A. Rogerson (eds.). The Sage Handbook of Spatial Analysis. London:
Sage, pp. 105–124.
Wood, H., & B. Skeggs (eds). 2011. Reality Television and Class. London:
Index
A Benefits cap 63, 152, 153, 173, 176,

Absolute poverty 3, 4, 236 177, 202, 204, 205, 228, 235
Addictions 167, 173 Benefits claimants 32, 142, 145,
Affluent areas 90, 171, 236 146, 153, 160, 167, 169, 172,
Aggregate poverty 78, 79, 90 174, 212, 214, 222, 226, 233,
Alcoholism 174, 177, 233 236
Asylum seekers 153, 167–169, 208 Benefits receipt xxi, 15, 102, 113,
Austerity 63, 102, 121, 128, 129, 116, 142, 159, 160, 165, 169,
147, 156, 177, 230 171, 172, 178, 203, 208, 218,
225, 226, 228, 230, 232, 236
B
Background geography(ies) 62, C
84–86, 89, 91, 132, 147, 180, Carstairs scores xviii, 2, 5, 79–81,
182, 184, 196, 237, 240, 241 83, 84, 89, 91, 124, 188–190
Bedroom tax 6, 63, 170, 176, 179, Census xviii, xxiii, 1, 2, 5–7, 13,
214 43, 44, 46–48, 79, 80, 97,
Benefit fraud xxi, 11, 31, 33, 113, 124–126, 188
120, 146, 154, 162, 164, 174, Child poverty 3, 21, 36, 72, 76–78,
178, 232, 234 98, 100, 101, 160, 205, 208,
221, 224, 231
© The Editor(s) (if applicable) and The Author(s) 2019 257

https://doi.org/10.1007/978-3-319-93503-4
258
Index
Collocate(s) 24, 96, 163 Foodbank(s) xx, 103, 108, 117,

Concordance geoparsing 52–54, 66, 121, 179, 212, 220, 221, 232,
91, 240 236
Concordance(s) 20, 24, 25, 34, 35, Fraud xxi, 31, 105, 113, 160, 162,
57, 72, 73, 75, 105, 107, 139, 163, 178, 236
140, 146, 152, 182 Fuel poverty 21, 75, 78, 79, 90, 100,
Conservative 62, 63, 98, 99, 140, 101
150, 155, 156, 160, 165, 169,
187, 201, 219, 226, 233, 235
Contrastive concordance analysis 73, G
135, 165 Gazetteer 42, 49, 50, 52, 53
Corpus linguistics xvii–xix, xxiii, 16, Geographical information systems
19, 20, 22, 26, 34–37, 41, 42, (GIS) xvii–xix, xxi, xxiii, 5,
58, 61, 62, 73, 227, 239, 240, 16, 37, 41–49, 53–58, 61, 66,
242 227, 240, 242
Correlation coefficients 83, 84, 126, Geographical text analysis (GTA)
189, 190 xvii, xix, xxiii, xxiv, 42, 61,
Council cuts 207, 218, 230 62, 65, 66, 85, 90, 91, 95, 96,
Crime 5, 15, 43, 48, 71, 74, 79, 90, 96, 111, 119, 123, 157, 227, 228,
105, 113, 141, 155, 186, 202, 235, 239–243
204, 205, 216–218, 226, 243 Geoparsing 42, 49, 51–54, 58, 61,
Critical discourse analysis (CDA) 85, 96, 99–102, 106, 119,
xvii–xix, xxiii, 19, 28–32, 120, 142, 157, 241, 242
34–37, 56, 227, 240, 243
H
D Homelessness xvii, 4, 13, 27, 31, 77,
Density smoothing 57, 62, 67, 184 97, 128, 146, 159, 176, 180,
Deserving poor 212, 230 182, 184, 186, 205, 207
Disability 102, 162, 163, 173, 174, Housing benefit 119, 145, 153,
177, 178, 233 164, 167–170, 176, 177, 179,
190, 191, 203, 224, 231–234,
237
F
Fitness to work 103, 108, 114, 121,
165, 236 I
Flawed consumerism 16, 28, 146, Immigration 113, 153, 154, 173,
233, 237 222, 223, 234
Index
259
Incapacity benefit 132, 138, 207, O

212, 224, 225, 230, 236, ONS 124, 139, 160
238
P
J PNC keywords 73–78, 90, 95,
Jobcentre 116, 128, 129, 146, 162, 135–138, 143, 147, 150,
176, 179, 210, 213, 222, 233 154–157, 165, 167, 170, 176,
Jobless 116, 128, 139, 141, 171, 180, 182–184, 186, 200–202,
202, 204, 205, 209 204–220, 222, 225, 230
Private sector 144, 236
Public sector 144, 156, 157, 207,
K 212, 214–216, 226, 228,
Kulldorf spatial scan statistic 86, 132 236
L R
Layers 45 Raster data 45, 46
Laziness 28, 139, 172, 232, 233 Reassessment 132, 138, 178,
Life expectancy 75, 79, 90, 214, 216, 238
231 Redundancies 120, 132, 136, 137,
Living wage 142, 146, 150–153, 141–144, 157, 215, 218, 221,
156, 157, 203, 204, 219, 231 222, 226
Relative poverty xxiii, 6, 28, 180,
204, 230, 236, 237
M Rent(s) 104, 109, 117, 118, 121,
Migrant(s) 153, 154, 172, 208, 222, 152, 153, 159, 167–170, 176,
231, 233, 236 177, 182–184, 186, 201–203,
Minimum wage 24, 142, 152, 155, 207, 208, 211, 213, 230
213 Rural poverty 2, 191
N S
Necessities 7–9, 12, 118 Scrounger 31–33, 228, 229
Neoliberal, neoliberalism 15, 90, Smith, Iain Duncan 160, 178, 213,
140, 142, 146, 191, 204, 226, 225, 235, 238
228, 232, 233, 237 Spatial humanities xviii
260
Index
T W
Taxpayers 97, 98, 102, 113, 152, 158, Wage(s) xvii, 16, 29, 30, 97, 102, 104,
167, 168, 171, 202, 205, 233 110, 111, 116, 117, 120, 121,
128, 150–156, 172, 201–203,
213, 222, 223, 231, 232
U Welfare reform 16, 63, 161, 165,
Unaffordable 118, 119, 167, 168, 169, 170, 173, 178, 179, 205,
190, 191, 201, 203 213, 225, 229, 230, 233, 235,
Undeserving poor 15, 33, 141, 169, 236
212, 229, 232, 235 Workless households 124–126, 143,
Universal credit 98, 178, 179, 208, 187
224, 225 Worklessness xxiii, 97, 102, 124,
Urban poverty 2, 56, 68, 75, 90, 126, 130, 139, 159, 171, 186,
124, 191, 196, 206, 215, 220, 188–190
225, 237 Workshy 138, 142, 172, 191, 212,
221, 222
V
Vacancy(ies) 116, 124, 128, 136,
137, 142, 143, 145, 157, 201

Representations of Poverty and Place PDF

Uploaded by

Copyright:

Available Formats

Representations of Poverty and Place PDF

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

Representations of Poverty and Place PDF

Uploaded by

Copyright:

Available Formats

Laura L Paterson

ISBN 978-3-319-93502-7 ISBN 978-3-319-93503-4 (eBook)

Library of Congress Control Number: 2018943637

© The Editor(s) (if applicable) and The Author(s) 2019

Cover image: © the authors

This work came out of a large number of collaborations and benefited

1 Defining and Measuring Poverty 1

2 Corpus Linguistics, Critical Discourse Analysis,

3 Geographical Information Systems and Textual Sources 41

4 Conducting GTA Using Poverty as a Search Term 61

5 How to Use GTA in Discourse Analysis 95

6 Locating (Un)Employment in the National Press 123

7 Characterising Poverty in Place: Benefits Receipt

8 Geography-Based Discourses of Poverty 193

Laura L Paterson is a Lecturer in English Language and Applied

Ian N Gregory is Professor of Digital Humanities at Lancaster

Fig. 2.1 Sample concordances for and poverty in the Guardian (2015) 25

Fig. 4.5 Kulldorf analysis of <*poverty*> using a 5% sample

Table 4.1 Breakdown of Daily Mail and Guardian corpora 65

Table 6.2 Top 100 lexical collocates of unemployment 128

Table 8.7 PNC Keywords for the Newcastle cluster. Italicised

This book presents an entirely new, interdisciplinary method that

may work systematically through census data or workhouse records to

Our decision to draw on linguistic data to analyse poverty, as

measurable in wider society. To take one particular example, despite the

cities including Oxford, Cambridge, and York. In the course of the

Fig. 1 Places that are frequently referred to in this book

To establish the terms of debate, Chapter 1 reviews how poverty is

and considers what GTA can tell us about representations of poverty.

Murrieta-Flores, P., A. Baron A., I. N. Gregory, A. Hardie, & P. Rayson. 2015.

© The Author(s) 2019 1

1.1 Key Components of Poverty

line’ or threshold economic figure are not automatically dissimilar and,

1.2 Poverty by Proxy

of people to rooms in a house, we have to consider political factors

They generalise from this one participant, whom they describe as ‘a

statements. Such generalisations run the risk of using individual expe-

1.3 Conceptualisations of Poverty

In a US context, Lorenzo-Dus and Marsh (2012: 289) analysed two

areas in Middlesbrough and Glasgow where multiple generations had

deployed to negatively evaluate members of the working classes.

terrorism and (inter)national insecurity’ (2012: 275). They note that

The prevalent themes discussed in Sect. 1.3 demonstrate some of the

Block, D. 2013. Social Class in Applied Linguistics. London: Routledge.

McKay, S. 2004. Poverty or Preference: What Do ‘Consensual Deprivation

© The Author(s) 2019 19

2.1 Introducing Corpus Linguistics

test assumptions [or hypotheses] about language use against patterns

Macmillan produce dictionaries this way). Corpus linguistics can also

distribution across the corpus). However, whilst calculating frequency

Fig. 2.1 Sample concordances for and poverty in the Guardian (2015)

overlap (notwithstanding the term food banks ) as they do not occur in

2.2 Defining Discourse

sanitation, disease, and death, whist other poverty discourses could

Altheide takes the position that ‘[t]heoretical boundaries are breached

Example (1) includes three occurrences of nominalisation—a linguis-

proposition which is accepted without question. Similarly, ‘stagnat-

2.3 Critical Discourse Analysis

2. Beggars belief: Scrounger, 34, who lay in doorways PRETENDING

Fig. 2.1 Sample concordances for and poverty in the Guardian (2015) 25

Fig. 4.5 Kulldorf analysis of <poverty> using a 5% sample

Table 4.1 Breakdown of Daily Mail and Guardian corpora 65

Table 6.2 Top 100 lexical collocates of unemployment 128

Table 8.7 PNC Keywords for the Newcastle cluster. Italicised

Fig. 1 Places that are frequently referred to in this book

1.1 Key Components of Poverty

1.2 Poverty by Proxy

1.3 Conceptualisations of Poverty

2.1 Introducing Corpus Linguistics

Fig. 2.1 Sample concordances for and poverty in the Guardian (2015)

2.2 Defining Discourse

2.3 Critical Discourse Analysis

2.4 Corpus-Based CDA and the Interrogation

3.1 An Introduction to GIS

Fig. 3.1 Using a combination of a spatial data and b attribute data to represent

3.2 Geoparsing: The Link Between Texts

3.3 Realities of Place and Space

4.1 Constructing the Corpus

Table 4.1 Breakdown of Daily Mail and Guardian corpora

4.2 Establishing the Geographies of Poverty

4.3 Geographical Variations in Discourses

Table 4.2 PNC keywords in the Guardian comparing the co-text of <poverty>

11. when 200 men walked to London to protest about unemployment

4.4 Comparing Representations of Poverty

Table 4.6 Spearman’s rank correlation coefficients comparing district-level

4.5 Comparing Representations of Poverty

Whilst the GTA of <poverty> in the previous chapter provided a

5.1 Selecting Search Terms