promoting access to White Rose research papers
White Rose Research Online
eprints@whiterose.ac.uk
Universities of Leeds, Sheffield and York
http://eprints.whiterose.ac.uk/
This is an author produced version of a paper published in Applied Spatial
Analysis and Policy.
White Rose Research Online URL for this paper:
http://eprints.whiterose.ac.uk/76875/
Paper:
Norman, PD (2010) Identifying change over time in small area socio-economic
deprivation. Applied Spatial Analysis and Policy, 3 (2-3). 107 - 138.
http://dx.doi.org/10.1007/s12061-009-9036-6
1
Identifying change over time in small area socio-economic deprivation
Paul Norman
School of Geography, University of Leeds, Leeds, LS2 9JT, UK
* Correspondence to:
Paul Norman
School of Geography
University of Leeds
Leeds, LS2 9JT
Tel: +44 (0)113 34 38199
Fax: +44 (0)113 34 33308
p.d.norman@leeds.ac.uk
Please cite as:
Norman P (2010) Identifying change over time in small area socio-economic deprivation. Applied Spatial
Analysis and Policy 3(2-3) 107-138
2
Identifying change over time in small area socio-demographic deprivation
Abstract
The measurement of area level deprivation is the subject of a wide and ongoing debate regarding the
appropriateness of the geographical scale of analysis, the input indicator variables and the method used to
combine them into a single figure index. Whilst differences exist, there are strong correlations between
schemes.
Many policy-related and academic studies use deprivation scores calculated cross-sectionally to identify
areas in need of regeneration and to explain variations in health outcomes. It would be useful then to
identify whether small areas have changed their level of deprivation over time and thereby be able to:
monitor the effect of industry closure; assess the impact of area-based planning initiatives; or determine
whether a change in the level of deprivation leads to a change in health. However, the changing
relationship with an outcome cannot be judged if the ‘before’ and ‘after’ situations are based on
deprivation measures which use different, often time-point specific variables, methods and geographies.
Here, for the whole of the UK, inputs to the Townsend index obtained from the 1991 and 2001 Censuses
have been harmonised in terms of variable detail and with the 1991 data converted to the 2001 Census
ward geography. Deprivation has been calculated so that the 1991 scores are directly comparable with
those for 2001. Change over time can be then identified. Measured in this way, deprivation is generally
shown to have eased due to downward trends in levels of lack of access to a car, non-home ownership,
household overcrowding but most particularly, to reductions in levels of unemployment. Despite these
trends, not all locations became less deprived with gradients of deprivation largely persisting within the
UK’s constituent countries and in different area types.
For England, Wales and Scotland, the calculation of Townsend scores can readily be backdated to
incorporate data from the 1971 and 1981 Censuses to create a 1971-2001 set of comparable deprivation
scores. The approach can also be applied to the Carstairs index. Due to differences in data availability
prior to 1991, incorporating small areas in Northern Ireland would be challenging.
Key words: Townsend & Carstairs deprivation index; Census; Change over time; Small geographical
areas & wards; UK
3
Identifying change over time in small area socio-economic deprivation
1. Introduction
Many schemes have been devised which aim to reduce a range of socio-demographic information about
geographical areas to a measure which summarises the characteristics of areas in a meaningful way.
These schemes can be traced back to ideas developed in the 1920s by human ecologists in Chicago (e.g.
Park et al., 1925); work which was later developed through community analysis by Shevky and Williams
(1949) and social area analysis by Shevky and Bell (1955). Robinson (1998) details various approaches
including factor analysis which began to be used by geographers in the 1960s and led to methods that
cluster geographically distant places together on the basis of various socio-economic commonalities
(Burrows and Rhodes, 1998). This became the basis of geodemographics, the “classification of small
areas according to their inhabitants” (Rothman, 1989: 1). Whilst geodemographics is mainly used in
business applications, in the UK the Office for National Statistics (ONS) and it’s predecessor have
produced a series of general purpose classifications in Great Britain, including the Craig-Webber
classification (Webber and Craig, 1978), the 1991 and 1999 classifications of local government and
health authorities (Wallace and Denham, 1996; Bailey et al., 1999) and more recently, in a collaboration
with academia, a classification of Output Areas (the smallest UK census geography) and larger
administrative geographies (ONS, 2004; Vickers and Rees, 2006).
Paralleling these schemes and prompted by concerns over increasing social and economic inequalities in
the context of public expenditure constraints, Senior (2002: 124) finds a “remarkable, and somewhat
bewildering, growth in the use of deprivation measures.” Drawing on a wide debate about the concepts
and measures of poverty and deprivation (Townsend, 1970; Townsend, 1979), in this context, deprivation
can be defined as a state of disadvantage relative to the local community, wider society or the nation to
which an individual, family or group belongs (Townsend, 1987). People can be deprived of adequate
education, housing of good quality, rewarding employment, sufficient income, good health and
opportunities for enjoyment (Dorling, 1996). To identify relatively deprived areas in the UK, various
indexes have been devised such as: the Jarman Underprivileged Area index (UPA) (1983); the Townsend
index (1987); the Carstairs index (Carstairs and Morris, 1989); Breadline Britain (Gordon, 1995); the
Index of Local Conditions (DoE, 1983 and 1994); the Index of Local Deprivation (Noble et al., 2000);
and the Index of Multiple Deprivation (IMD, 2004; Noble et al., 2006). The UK indexes have mainly
been at the electoral ward scale (a relatively local, small area geography) and are predominantly based on
a composite of census-derived variables as indicators of relative conditions between areas although in the
recent IMD alternative geographies and input variables are used. Deprivation indexes have also been
developed in the US, Canada, New Zealand, France and elsewhere (Bell et al., 2007; Havard, 2008).
4
These deprivation measures are highly influential for the allocation of public resources (Simpson, 1996;
Brennan et al., 1999; Chatterton and Bradley, 2000; Blackman, 2006) and are regularly used as
explanatory variables in models of various outcomes including health in both the UK (Law and Morris,
1998; Senior et al., 2000; Boyle et al., 2002; Dibben et al., 2006; Diez Roux, 2005; Norman et al., 2005;)
and in other countries (Lorant et al., 2001; Tello et al., 2005; Karpati et al., 2006; Pearce et al., 2006).
Choice of which index to use is subject to debate (Mackenzie et al., 1998; Davey Smith et al., 2001) but a
high degree of correlation between schemes is found (Morris and Carstairs, 1991; Hoare, 2003).
Despite repeated calls from academics and others, unlike many other countries the UK Census has not
included an income question (Marsh, 1993; Dorling, 1999; Boyle and Dorling, 2004). Thus, one of the
regular criticisms of the construction of deprivation indexes is the use of ‘proxy’ indicators. Further
criticisms include subjectivity in the choice of input variables, which geographical scale to use, over-
complex (and thereby opaque to the general public) or inappropriate methodologies, different relevance in
urban and rural locations and the arbitrary choice of threshold scores for fund allocation. Senior (1991),
Bradford et al. (1995), Simpson (1996) and especially Senior (2002) provide very useful and detailed
reviews of indexes and their construction; various aspects of which are discussed further below. Given
that these deprivation indexes are area measures, note that the ‘ecological fallacy’ warns against assuming
that all persons and households in deprived areas are necessarily deprived (Fieldhouse and Tye, 1996;
Sloggett and Joshi, 1998) and that the results obtained and disseminated at one geographical scale will not
necessarily hold at another; the ‘modifiable areal unit problem’ (MAUP) (Openshaw and Taylor, 1981).
Despite much input from highly regarded researchers, Congdon (2004: 742) believes that “much work
remains to be done in both index construction and technique.”
A deprivation score, along with all other data derived from the decennial census, offers an insight into
local socio-demographic conditions pertaining at the time of the census. However, a deprivation score
calculated for an electoral ward for one census is not directly comparable with a score calculated for
another census. As with other work which constructs a time-series of socio-demographic data (Martin et
al., 2002; Norman, 2006), this is for several reasons. The questions on the census forms as well as the
categories by which information on people or households is disseminated may vary through time (Marsh,
1993) and as society changes, the meaning or relative importance of particular variables may change. The
geography of local government is under periodic review (Norman et al., 2007) so the wards for which
deprivation scores are calculated may vary in their population size due to boundary revisions and/or may
cease to exist. Even if both the attribute information used as index input variables and the geography
remain constant between censuses, the score recorded at one census is relative to the national situation in
that year. Thus, a ward may have the same score at successive censuses but may have become relatively
more or less deprived over time in comparison with other wards.
