1. Social identity, relative deprivation
and patterns of minority partisanship
Anthony Heath
Steve Fisher
David Sanders
Maria Sobolewska
2. Aims
• To understand high levels of support for Labour
among ethnic minorities at the 2010 British
general election.
• To explore differences between the main
minorities – Indian, Pakistani, Bangladeshi,
Black Caribbean and Black African in levels of
support for Labour
• To explore class differences within minorities –
does ethnicity trump class?
3. Previous research in Britain
• We cannot explain minority support for
Labour in terms of standard socio-
demographic variables such as class
(Heath et al 1991, Heath et al 2011)
• Or in terms of standard issues and values
(Sobolewska 2005)
• Large and significant ethnic coefficients
remain in all standard regression models.
4. Relative deprivation
• Previous scholars have suggested that shared ethnic
group interests that cut across class may be important
(Studlar 1986, Heath et al 1991)
• Evidence from the labour market shows strong evidence
of ‘ethnic penalties’ at all levels of educational attainment
(Cheung and Heath 2007)
• Discrimination and prejudice are one source of these
penalties (although other forms of exclusion are also
likely)
• This suggests that Runciman’s (1966) concept of shared
feelings of relative deprivation
5. Group identity and normative
reference groups
• A second key element is group identity and a sense of
social solidarity (which may unite members in different
social classes)
• Merton’s (1957) concept of normative reference group
may be helpful (or later developments such as social
identity theory)
• Supporting evidence from Dancygier and Saunders 2006
using 1997 EMBES data.
• Our approach has some similarities with Dawson’s
theory of ‘linked fate’ but also has some crucial
differences.
6. The data
• The 2010 Ethnic Minority British Election Survey
(EMBES)
• Thanks to the ESRC for their generous funding of the
study
• To the Electoral Commission for their support and
partnership
• To TNS-BMRB (Nick Howat, Oliver Norden, Emily
Pickering) for their work on design and fieldwork
• To our Advisory Board - Irene Bloemraad, John Curtice,
Harry Goulbourne, Chris Myant, Maajid Nawaz, Lucinda
Platt, Peter Riddell, Shamit Saggar, Will Somerville,
David Voas
7. Design 1
• Stand-alone survey rather than a booster to the
main BES (ie separate sample design etc)
• Nationally-representative probability sample
• Clustered, stratified design with over-sampling in
high EM density areas and exclusion of lowest
density areas (< 2% EM)
• PAF used as sample frame
• LSOAs were the PSUs (unlike main BES)
• Initial screening of addresses
8. Design 2
• 30,000 addresses issued for screening
• In 620 PSUs
• £20 conditional incentive offered to participants
• 50 minute questionnaire, administered by CAPI
with a self-completion module for confidential
items
• Around half items exact replications of those in
main BES
• Short mailback questionnaire
9. Outcome
• 2787 respondents in total (including some
from mixed and other backgrounds who
had been indicated as belonging to one of
the 5 target groups at screening)
• Response rate of 58 – 62% (depending on
method of treating those with unknown
ethnicity from the screening exercise)
• Poor response to mailback – 975 returned
10. Sample characteristics
EMBES BES
White British 0 3126
Other white 0 57
Mixed 113 32
Indian 587 52
Pakistani 668 17
Bangladeshi 270 8
Black Caribbean 598 31
Black African 525 38
Other 26 59
11. Party ID
Lab Cons LD Other/none
White British 30 29 12 29
Indian 55 16 10 19
Pakistani 55 8 15 22
Bangladeshi 57 9 9 26
Black Caribbean 68 5 5 21
Black African 71 5 5 20
All EM 61 9 9 22
12. Class differences in Labour ID
Middle class Working class
White British 24 41
Indian 52 66
Pakistani 46 64
Bangladeshi 58 60
Caribbean 67 72
Black African 70 72
All EM 57 66
13. % with great deal in common with own
ethnic group
Middle class Working class
Indian 40 44
Pakistani 33 48
Bangladeshi 31 52
Caribbean 53 53
Black African 61 56
NB large generational differences on this question
14. % with all or most friends from same ethnic
background
Middle class Working class
Indian 43 57
Pakistani 51 66
Bangladeshi 58 71
Caribbean 43 44
Black African 48 51
NB large generational differences on this question
too
15. % who agree there is often a large gap
between what ethnic group expects and
gets
Middle class Working class
Indian 43 51
Pakistani 53 47
Bangladeshi 60 40
Caribbean 67 69
Black African 63 61
Few generational differences on this question
16. % who agree Labour is best party to
improve life for ethnic minorities
Middle class Working class
White British 36 34
Indian 49 58
Pakistani 54 57
Bangladeshi 53 47
Caribbean 57 61
Black African 70 71
Consistent with labour’s track record of introducing equality
legislation
17. Main conclusions so far
• Black groups have higher levels of subjective
solidarity, but not social involvement, and
relative deprivation than South Asians
• No class differences among Black groups in
subjective solidarity or social involvement
• Substantial class differences among South
Asian groups in social involvement tho’ not
relative deprivation
Suggests that group processes likely to be
important part of the explanation
18. Strength of Labour ID by proportion of co-
ethnic friends
All most half few
Indian 1.3 1.2 1.1 0.9
Pakistani 1.2 1.1 0.9 1.1
Bangladeshi 1.3 1.1 1.1 1.0
Caribbean 1.6 1.5 1.3 1.3
Black African 1.7 1.5 1.3 1.5
Supports theory of normative reference groups
19. Modelling the data
• Purely individualistic models don’t explain the ethnic
differences (even if we include measures of solidarity
and relative deprivation)
• Need to introduce measures of group solidarity
(‘contextual effects’)
• Ie multilevel model with the ethnic group as level 2
• 5 ethnic groups not sufficient for this strategy, but can
sensibly distinguish 14 ethno-religious groups
• 3-level model including local area measures might also
be worth exploring
20. An individual-level model
Model 1 Model 2
Indian (ref) 0
Pakistani -0.07 -0.10
Bangladeshi -0.01 -0.04
Caribbean 0.33* 0.30*
African 0.44* 0.40*
MC -0.44* 0.44*
MC*Black 0.37* 0.34*
Model 2 includes controls for closeness (NS), social involvement (NS),
relative deprivation**, individual discrimination (NS). Interactions
with relative deprivation NS
21. A multilevel approach
Model 1 Model 2
Closenessi 0.08* 0.09*
Social involvementi 0.07 0.07
Relative deprivationi 0.18** 0.15**
% Closenessj 0.51***
% Social involvementj -0.18*
% Relative deprivationj 0.24***
Level 2 variables entered singly
22. Conclusions
• Early days in our analysis, but analysis gives
some grounds for taking group processes
seriously
• Evidence consistent with normative reference
group theory
• Preliminary indications of importance of relative
deprivation (at both individual and group levels)
• Preliminary indications of importance of social
identity/solidarity – especially at the group level
– and cutting across class lines among Blacks
23. Further research
• Need to look at role of organizational
involvement, especially in co-ethnic
organizations
• Also need to look at exposure to (ethnic)
media
• And must take account of important
generational differences