Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
1
Social Activity and Ethnicity in the UK1
Lucinda Platt, ISER, University of Essex
Paper prepared for presentation at the RC28 Montréal meeting, August 2007. Introduction
This paper explores the extent to which social activity in England and Wales
varies by ethnic group and whether social deprivation is more marked for
some groups than others. In doing this it sets out to enhance our
understanding of one particular dimension of poverty and its variation by
ethnicity, as well as empirically informing discussions of social capital
formation, in relation to informal measures of participation, and of ethnic
capital.
Lack of social participation has been a widely accepted element of the
definition of poverty since at least Townsend’s seminal (1979) study of
poverty. Opportunities for ‘normal’ social interaction and the enjoyment of
communal or community activity, including such things as celebrations of
significant religious events and being able to extend hospitality to children’s
friends (and their parents), neighbours and family are seen as core to living a
full and non-deprived lifestyle. Consequently, measures of multiple
deprivation, or of poverty conceived multi-dimensionally rather than purely in
income terms, tend to incorporate participative measures, albeit such
measures are often very limited due to the constraints of the data (Mack and
Lansley 1985; Gordon and Pantazis 1997; Nolan and Whelan 1996; Gordon
et al. 2000; Berthoud et al. 2004). And social inclusion is defined as including
‘integration with family, friends and community’ (Burchardt, 2002).
2
Participation, as a specific dimension of poverty or social inclusion has,
nevertheless, received little sustained attention in its own right independent of
it inclusion in summed multiple deprivation scores. There are clearly
arguments in favour of treating the different dimensions of poverty discretely
and empirically ascertaining their relationship with one another – and with the
default measure of poverty in income terms (Capellari and Jenkins 2007).
Lack of ‘normal’ social participation is an aspect of deprivation that is arguably
best served by being considered separately from more material forms of
deprivation, and this paper specifically examines four different aspects of
social engagement or ‘sociability’. Using the British Home Office Citizenship
Survey 2001 for England and Wales, it also explores whether the results
support the idea of some underlying propensity to lack of sociability for which
all four measures act as indicators. The four measures selected were: having
friends and neighbours round, visiting friends and neighbours, involvement in
some form of organised activity or ‘club’, and going out for a social reason.
These represent potentially distinct aspects of sociability, though in the final
analysis the chances of lacking all four forms of social engagement, or none,
are examined.
Specifically this investigation examines whether patterns of sociability are
shaped by the ethnic group of the respondent, and whether lack of social
engagement can be associated with any group more than another. There are
a number of reasons why it is particularly pertinent to examine ethnic group
variations in lack of social participation. First, income poverty rates vary
dramatically across the UK’s ethnic groups, with particularly high rates for
3
Pakistanis and Bangladeshis, but with above average rates for all minority
groups (Platt 2007). Measures of material deprivation tend to echo the
patterns of income poverty to the extent that relevant analysis exists, but lack
of suitable data have inhibited comparisons of multiple deprivation across
ethnic groups and there are no extant studies of variation in social deprivation
according to ethnicity. Given the extensive interest in ethnic differences in
poverty, there is clearly a contribution to be made by understanding the extent
to which participation echoes or diverges from other forms of deprivation. We
might expect that, as with forms of material deprivation, the poorest groups
are also those most at risk of lacking social participation. However, the
relative geographical clustering of the most disadvantaged groups may help to
reduce isolation – and indeed – the reason for the clustering may itself be to
provide some form of ‘buffer’ against negative experiences through social
contact. The analysis presented here can also help us to understand the
extent to which ethnic differences in social engagement appear to derive from
income differences between groups, by examining the association between
lack of social engagement and ethnic group both with and without taking
account of income.
Moreover, simply having some grasp of patterns of participation and
association as they vary across ethnic groups, goes some way to helping us
understand the extent to which the ethnic group categories into which
individuals are allocated are meaningful in a more general sense. It has been
argued that the crude – and much criticised – ethnic group categories
available to quantitative analysts serve their purpose at the point at which they
4
reveal substantial inequalities between groups (Platt 2002).2 It has also been
proposed that we should be looking to refine our allocation of individuals to
groups beyond the simply ticking of subjective categories by looking at actual
measures of group belonging, such as patterns of association (Fenton
forthcoming). While the data employed in this paper do not allow us to identify
who respondents are associating with,3 ascertaining whether or not there
appear to be distinctive patterns of association for the different categories is a
first step towards building up a more sensitive understanding of ‘groupness’. A
number of qualitative studies have investigated the ways in which particular
groups interact with others, and the extent to which these interactions are
group-based in some sense, as well as the meanings of those interactions,
but of necessity such research, though illuminating, cannot answer empirical
questions relating to the extent and patterns of participation (Hudson et al.
2007; Salway et al. 2007).
The empirical analysis of lack of social participation is also of particular
interest given the increasing attention in both UK policy and academy paid to
the concept of social capital (Aldridge and Halpern 2002; ONS 2001; Li et al.
2003; Edwards et al. 2006); and tangible measures of such social capital have
been identified across British social surveys (see
http://www.statistics.gov.uk/socialcapital/).4 Social capital is largely construed
instrumentally, as potentially having pay-offs for both individual and
community (Bourdieu 1997; Coleman; Putnam 1995, 2000), compared to a
focus on participation in the context of deprivation, which treats it as an end in
itself: an experience that is crucial to the current well-being of the individual.
5
Nevertheless, even though the primary concern of this paper is to understand
the extent to which individuals are deprived of social engagement, there are
clearly ways in which this analysis of social participation can inform debates
on social capital. Much of the British literature on the measurement of social
capital has focused on measures of ‘civic participation’ or associational
membership and trust (Duffy 2004; Li et al. 2002; Li et al. 2003; Warde et al.
2003; Pennant 2005). The sorts of activity that we examine here – reciprocal
visiting, participation in more organised activities such as clubs or
volunteering, as well as simply going out – can be linked to the kinds of
informal associative activity emphasised by Coleman (1988), who saw social
capital formation primarily as a consequence of activities pursued for other
purposes. These measures also, however, overlap with measures of civic
participation as more formally conceived in the existing literature and as
emphasised in Putnam’s work. In fact, Putnam’s (2000) index of social capital
includes among its 14 components two measures of informal sociability. By
providing an empirical understanding of the patterning of social participation
and of the extent to which different forms predominate across different groups
the paper examines a neglected aspect of social capital and provides a basis
for exploring more systematically current claims about community, cohesion
and isolation.
Once again, the ethnic group dimension to the analysis has particular
applicability when we consider the links with social capital formation.
Community cohesion is often taken to have a specific application to the extent
to which different ethnic groups participate across a range of different
6
dimensions of participation (Commission on Integration and Cohesion 2007).
And the search for resources within communities is pursued energetically in
policy-related research (Furbey 2006). However, participation in this
framework is conceived of individualistically and, as in much of the policy level
discussions of social capital, the potential negatives are not discussed, either
for the individual enmeshed in tight social bonds that restrict their upward
mobility or via the use of networks to promote the exclusion of others (Portes
1998; Lin 2001).
On the other hand the increasing anxiety about ‘segregation’ and the relative
concentration of certain minority groups treats co-residence as a potentially
problematic phenomenon (Community Cohesion Review Team 2001), without
consideration either of the extent to which it may foster a protective
environment for the most disadvantaged and marginalised groups (Heath and
Cheung 2006; Bajekal et al. 2004) or of the ways in which it might foster
social capital of the bridging (within group) as well as the assumed bonding
variety (Schuller, Baron and Field 2000; see also Granovetter 1973; Lin
2001). There is also an implication that geographical dispersal will also be
accompanied by both economic and social integration. There has been some
challenge to this view (Peach 2005), but the extent to which geographical and
economic integration may in fact be at odds with social integration is made
evident in the results from this paper.
By contrast work on ethnic capital has illustrated the extent to which co-
residence and tight social networks may act as a resource (Zhou 2005). Both
7
the meaning of ‘ethnic capital’ – as human or social capital, or both – and its
potential role in both aiding or inhibiting upward mobility or integration within
society have been debated (Borjas 1992; Esser 2005). Underlying much of
the discussion has been an assumption of relatively close ties within groups –
yet the extent to which this is reflected in patterns of social activity remains
open to empirical investigation. Existing investigation into ethnic ‘enclaves’ in
the UK does not suggest much evidence of positive effects (Clark and
Drinkwater 2002), but the contribution of social networks to either fostering or
inhibiting advancement is not well understood, particularly in relation to the
kinds of informal social contact treated here. This paper, then, aims to provide
a contribution to understanding how social contact does, indeed, play out for
different groups, which may itself help us to develop our understanding of
ethnicity as a resource further in future.
Finally, as well as considering differences in social participation according to
ethnic group, this paper pays substantial attention to health status. This stems
from the perspective that excluding the limitations that may derive from the
constraints of ill-health or caring may produce a misleading picture of group
differences. We know that rates of ill-health, disability and caring vary widely
across ethnic groups (Nazroo 1997; Erens et al. 2001). On one level this is
unsurprising given that poverty both stems from and contributes to ill-health
(Jenkins and Rigg 2004; Burchardt 2000); and rates of poverty vary greatly
across minority ethnic groups (Platt 2007). And there is an extensive literature
charting how ill-health, disability and caring may impinge on social
participation in a variety of ways (Howard 2001; Locker 1983; Parker 1993).
8
While one might posit that it is ill-health that reduces participation, the positive
aspects of social contact may also have a beneficial impact on the health of
those already suffering from a chronic condition (or their absence a negative
one). A number of studies have suggested that social participation and
engagement may have a beneficial impact on health, and that social isolation
may have a negative one (Berkman 1984; Berkman et al. 2000; Veenstra
2000); and Whelan (1993) has argued that social support protects against
chronic stress consequent on material deprivation. Thus, ethnic group
patterns of participation may both need to be understood in the context of
differential health status, and potentially be able to contribute to
understanding those differences. Current explorations of the role of social
support in relation to ethnic group health outcomes are ambiguous in their
results (Stopes-Roe and Cochrane 1990; Pollard et al. 2003) It is also
important to understand how these are mediated by income, given that those
with poor health tend to be worse off than those in good health.
In the next section I describe the data and the variables and approach in more
detail. Subsequent sections discuss the results of the analyses, examining
men and women separately. The final section draws some conclusions.
Data and methods
Data
Social engagement and its lack was analysed using the Home Office
Citizenship Survey (Home Office 2003). This is a biennial survey which is
explicitly designed to capture information about the involvement of individuals
9
in a range of community and civic activities, their child-rearing practices and
sources of information and support, and their experience of their
neighbourhood and attachment to it. This paper uses the survey for 2001
since it was the only one to include the participation variables which form the
basis of this analysis. The sampling unit was the individual (rather than the
households) and most of the questions related to the respondent’s own
experience, views and perceptions. The survey was specifically intended to
capture ethnic group differences in ‘citizenship’ and community experiences
and therefore, on top of its c. 10 000 person main sample, the design
incorporated a booster sample of c. 5 000 members of minority ethnic groups.
Survey weights that adjusted for survey design and for response probabilities
have been employed throughout the following analysis.
The focus of this paper is on adult respondents of working age (18-59/64).5
The outcomes for men and women are examined separately throughout the
paper, given the wide variation in both health status, caring, and social activity
according to sex.
Variables
To measure lack of social engagement four questions were selected to
summarise different aspects of sociability. From the grouped response
options, I constructed binary variables from the chosen questions to represent
deprivation on the measure, i.e. lack of that particular form of social
participation. The measures do not necessarily capture all forms of social
engagement. However, between them they allow for variation in types of
10
social engagement by group or circumstances and, I argue, together they
effectively cover most potential forms of social activity. The four measures
are:
� Lack of visiting: goes round to friends or neighbours less often than once a
fortnight;
� Lack of being visited: has friends or neighbours over less often than once
a fortnight;
� Lack of going out: goes out with friends or neighbours less often than once
a fortnight; and
� Lack of organised activities: is involved in ‘clubs’ less often than once a
month or not at all. ‘Clubs’ is here an inclusive term which includes
voluntary activities, activities based round a religious centre or focus, as
well as organised interest groups. It thus summarises any more organised
forms of social activity.
Health status was measured by a question on whether the respondent
suffered from a limiting long-term health problem. Caring was measured by
the response to whether the respondent cared for someone within or outside
the household with such a limiting long-term health problem. In the 2001
survey the caring question did not distinguish between whether the person
cared for lived in or outside the household.
Overall 24 per cent of the working age population in this survey had a long-
term illness and nine per cent were carers. Nearly a quarter of the carers had
a long-term illness themselves, which means that two per cent of the sample
11
were both ill and caring. Rates of illness were fairly equal across the sexes
though slightly lower for women; but women had higher rates of caring. As we
would expect, illness and caring responsibilities both became more likely with
age.
For ethnic group, a 17-level ethnic group variable was reduced to eight
groups. The eight categories were white British, Indian, Pakistani,
Bangladeshi, Black African, Black Caribbean, Chinese, and Other and Mixed.
However, the results for the Chinese are not extensively discussed in what
follows given the small counts on which they are based. The combination of
the various, numerically small, ‘mixed’ groups with ‘other’ groups results in a
residual category that is not inherently meaningful. The seven main
categories are not uncontested; nor are they assumed to be homogenous.
However, they are regularly employed in analysis as representing
aggregations with distinctive histories and displaying diverse outcomes both in
absolute terms and which often persist when a range of relevant
characteristics are held constant (see Modood et al. 1997; Platt 2007).
Whether such diversity extends also to social participation is an empirical
question that is investigated here.
Characteristics considered relevant to understanding differences across these
main variables were held constant in the multivariate analyses to allow the
comparison of differences between those with otherwise comparable
characteristics. These ‘control variables’ were identified and measured in the
following way:
12
� age: a binary variable was constructed from a banded variable to
distinguish between those aged 18-39 and those aged 40-59/64, since
rates of illness and caring are much higher at the older age ranges, and
patterns of participation might also be expected to change over the life
course.
� the presence of a child aged under 5 years, which could be expected both
to limit opportunities for social activity as well as creating opportunities for
certain, child-based activity. The presence of pre-school children has
often been found to be the crucial indicator for various forms of (non-)
participation, including labour market participation. It should be noted that
this variable is strongly associated with age and thus it predominantly
serves to distinguish the probability of lack of social participation within the
younger age group.
� presence of a partner, since being in a relationship could substitute for
forms of social participation, as well as potentially enabling access to
alternative, additional networks. Partnership covers both cohabitation and
marriage.
� work history: whether currently working; whether worked in the past but
currently not in employment; and whether never worked. Some of the
justification for the reform of incapacity benefits revolves around the notion
that being on sickness benefits is bad for the health and for social relations
(as well as for the economy) (DWP 2006). Work can also be regarded as
a form of participation in itself: it could be argued that either those in work
are more likely to have social contacts (Petersen et al. 2000), resulting in
positive effects on measures of social participation, or that work itself
13
constitutes such a form of social contact, which might result in negative
effects on other forms of social participation. It was important to
distinguish within the non-working between those who had worked at
some time – and might well return – and those who had never worked.
Those who had worked in the past might have existing work-based
networks to call upon. On the other hand the negative impacts of
unemployment on social engagement are well known. By contrast, those
who have never worked may have developed alternative sources of social
support and contact – or conversely may be among the most isolated. This
distinction was felt to be particularly important given that the proportions of
those who have never worked varies by ethnic group (for men as well as
women).
� income. Because illness and income are so closely related and because
some ethnic groups have lower average income, the analysis also controls
for income level, to check whether any observed effect of illness or of
ethnicity can in fact be attributed to the consequences of low income. In
deprivation analysis, the measure of deprivation tends to be validated in
relation to income. That is, the lack is only considered to constitute
deprivation if it occurs where income levels are also low. Here, however,
we examine the role of income in mitigating or accounting for social forms
of deprivation. A measure of household size adjusted the income bands
for households of equal size.
Methods
14
Multivariate probit models were estimated to model outcomes for all four
dependent variables simultaneously. This approach allows for greater
flexibility in associations between the independent and each of the dependent
variables: the coefficients on the regressors can vary with each outcome
variable. At the same time, it enables us to explore whether there are
correlations between the unobservable characteristics associated with each
outcome. If all the off-diagonal correlations are equal to zero then the
multivariate probit reduces to a series of univariate probits. If however there
are correlations across the equations, it suggests that they are best modelled
together. Such correlations across each pair of equations could be argued to
imply an underlying propensity to lack of sociability for which the four
measures acted as indicators.6
Separate multivariate probits were estimated for men and women including
the independent variables identified above and with two versions: one without
income and one including it.7 This was for two reasons: one practical and one
theoretical. Around 29 per cent of adult respondents did not respond to the
income question (27 per cent of men and 30 per cent of women). The non-
response also varied by ethnic group, with as many as 39 per cent of
Pakistani men and 50 per cent of Pakistani women not providing a response.
Thus, including income substantially reduced sample sizes. By examining
models with and without income, therefore, it was possible to identify
comparable coefficients across the two, which were insignificant in the income
models but where the larger sample sizes from non-income model rendered
them statistically significant. At the theoretical level, the interest was in
15
observing the extent to which forms of social engagement – or, rather, its
absence – were coterminous with lower incomes. That is, could lower levels
of social engagement be explained in terms of financial constraints?
Predictions were estimated for outcomes on the different measures of lack of
sociability and for the probability of experiencing all or none of them with
characteristics held constant and varying only on the focus of interest
(ethnicity, or caring and health status). These predicted probabilities illustrate
the very different chances and patterns of social contact that men and women
from different ethnic groups and with or without ill-health and caring
responsibilities can expect to experience.
Overview
Tables 1 and 2 (for men and women respectively) show how the four
measures of lack of social engagement varied by ethnic group and health and
caring status. Caribbeans and Black Africans were more likely to face
infrequent visiting and receiving of visits from friends or neighbours than
Pakistanis and Bangladeshis, with white British, Indian and Mixed and other
falling in between. On the other hand, Bangladeshi and Pakistani men as well
as women were less likely to go out or participate in organised activities than
those of white British, mixed and other and Indian ethnicity. These patterns
would suggest that there was some substitution between the measures, with
different groups favouring different routes for sociability. However, the
Caribbeans appeared to lose out in terms of social engagement overall. Both
male and female Caribbeans appeared more deprived of social engagement
16
than the white British on all four measures. This may have implications for
their wellbeing and their opportunities. On the other hand, Black Africans,
whose overall profile makes them look similar to the Black Caribbeans in
terms of sociability, had the lowest rates of deprivation on the measure of
organised activity (participation in ‘clubs’). The central role of the Church for
many Black African communities may help to explain this finding (Hudson et
al. 2007; Salway et al. 2007). Again, then, we may be seeing some
substitution in relation to available or preferred forms of activity.
Both illness and caring appeared to increase the chances of lacking social
engagement across measures. However, overall it seemed that it was caring
that put more constraints on sociability than illness, particularly for men.
[TABLES 1 AND 2 ABOUT HERE]
The question then becomes whether these distinctive patterns of social
engagement by ethnic group and health / caring status remain when other
relevant characteristics are held constant. And to what extent do these
measures, crude as they may appear, function as markers of an underlying
propensity to lack of social engagement – or self-sufficiency? These
questions are covered in the next sections.
Multivariate analysis
Tables 3 and 4 show the estimates of the multivariate probit regression
models for men and women respectively. All the variables discussed in the
17
Data section, except household income and household size, were included in
the models. In the tables, positive coefficients indicate that the characteristic
is associated with greater deprivation – or a lower chance of engaging in the
activity – while negatively signed coefficients indicate that the chances of
lacking this form of social engagement are reduced (or social engagement is
increased).
Men
Table 3 shows that older men were less likely to visit, be visited or go out. But
their involvement in organised activities was not significantly different from
younger respondents. Partnership may represent an alternative form of social
contact for respondents, as those with a partner were more likely to be
infrequent visitors of their friends and neighbours and to go out less
frequently. Having a young child seemed to constrain the types of activity
engaged in: fathers of young children were more likely to miss out on going
out and participating in organised activities relative to their peers with older or
no children. But patterns of visiting and being visited were not significantly
affected.
Qualifications reduced chances of lacking social engagement across the
board, as would be expected; and not being employed appeared to allow
greater time for visiting and being visited but restricted opportunities for going
out. This could a disguised income effect: those not in work having fewer
resources to spend on going out, with substitution between more expensive
and cheaper activities taking place. The model with income included
18
(discussed below) indicated that this was indeed the case. However, for
those who had never worked, income did not seem to make much difference
and thus lack of networks to go out might be inferred as being more important.
Among men, illness was scarcely significantly associated with any of the
measures of sociability. That is, people with a long-term health condition
were no more or less likely than those without to lack social opportunities. It
has been suggested that isolation may exacerbate or produce ill-health, but
the lack of association between sociability and long-term illness here gives
little support to that argument – at least for those of working age. The story for
caring was somewhat different. Caring rendered lack of social engagement
significantly more likely for men in relation to three out of the four measures.
Interestingly, it was involvement in organised activities that was not
significantly affected, even though it might be expected that more organised
social contact would be the hardest to combine with the demands of caring.
Looking at ethnic variations in lack of social participation, Bangladeshis were
much less likely to lack reciprocal visiting than their otherwise comparable
white British counterparts, and Pakistanis were less likely to lack opportunities
for visiting. On the other hand Pakistanis and Bangladeshis, alongside Black
Caribbeans and Black Africans, were much more likely to miss out on going
out. For Black Africans this was accompanied by higher chances of being
deprived on the visiting / visited measures, and, for Black Caribbeans on the
organised activity and (marginally) on the visiting measure. Thus, these two
Black groups appeared to face limited social opportunities compared to both
19
their white peers and those from other minority ethnic groups. Again, the
amount that can be explained by differentials in income will be illustrated from
the results of the next model.
Finally, Table 3 showed substantial variations in the cross-equation
correlations (the rhos). The correlations between lack of involvement in
organised activity and the other three outcome variables were between 0.16
and 0.28. The correlation between not visiting and not being visited was,
however, over 0.88, which shows a high level of reciprocity between these
activities (people don’t ask you round if you don’t ask them). Correlations
between these measures of visiting and not going out were above 0.5. All the
correlations were, nevertheless, statistically significant and the likelihood ratio
test clearly rejects the independence of the equations. As discussed above, if
there are correlations across the equations, it indicates that they are best
modelled together. The results suggest an underlying propensity to lack of
social engagement. That is, that these four measures act between them as
indicators of a tendency not to participate socially, call it asociability or, more
positively, self-sufficiency, that cannot itself be measured directly and that is
independent of the characteristics already controlled.
[TABLE 3 ABOUT HERE]
Women
Table 4 shows that the same pattern for the cross equation correlations were
observed for women, suggesting that these measures work as indicators of
20
such a latent propensity across the sexes. The individual correlations were
also of a similar size.
In relation to the coefficients for the independent variables, age, qualifications
and work history variables showed the same overall pattern of effects for
women as for men. Partnership also seemed to act as a substitute for social
activity for women. Those with young children were less likely to lack visitors
than those with older or no children – and this was controlling for work status.
But for women, unlike for men, having a young child was not associated with
lower participation in organised activities.
Illness had a statistically significant association with infrequently going out
among women, as well as a marginally significant effect on low levels of visits
to friends and neighbours. Women carers were, however, not especially at
risk of receiving infrequent visits; and they were more likely than non-carers to
be involved in organised activities. Nevertheless, caring increased risks of
infrequent visiting and infrequently going out, suggesting that their
responsibilities kept them predominantly at home.
Black African women showed high risks (relative to comparable white British
women) of being deprived across the first three measures, though the
coefficient was negative – albeit not statistically significant – for involvement in
organised activities. And Black Caribbean women replicated this pattern of
reduced participation on the first three measures, and much more strongly
than their male counterparts. That is, they appeared to differ more from white
21
British women than Black Caribbean men differed from white British men in
terms of social participation. Pakistani and Bangladeshi men lost out on going
out compared to their white British counterparts; but they showed no
significant differences in relation to organised activities. However, Pakistani
and Bangladeshi women were significantly more likely to lack involvement in
organised activities than white British women as well as going out less.
Similarly, where Pakistani and Bangladeshi men engaged in reciprocal visiting
more than white British men this was not evident for Pakistani and
Bangladeshi women compared to white British women. Finally, where Indian
men were insignificantly different from their white counterparts in relation to
their patterns of social participation, Indian women were significantly more
likely to lack social engagement, relative to white women, on at least two of
the measures. This is not a ‘South Asian’ effect, however, since Caribbean
women and Black African women also showed much higher risks of being
deprived on these measures than their white British counterparts. Nor is it a
simple gender effect since, clearly, these contrasts are with white British
women – who might be expected to face many of the same gender
constraints as other women, particularly when employment and family status
are controlled.
[TABLE 4 ABOUT HERE]
There are clearly preferences or constraints operating to shape patterns of
social participation in ways that are distinctive across ethnicities. One such
constraint that might potentially be playing an important role is, of course,
22
levels of resources. There is only a measure of annual household income
(from all sources), so we do not know about the respondent’s individual
access to resources, and this issue of control over resources might be
particularly relevant for women’s participation – or lack of it. However,
controlling for the income level in the household should still provide some
grasp of whether financial limitations were affecting patterns of participation
and opportunities for social contact.
The effect of income
We now turn therefore to one of the original motivating issues for the analysis.
To what extent are differences in participation and in lack of social
engagement – co-terminous with income differences? The multivariate probit
models were therefore re-estimated controlling additionally for banded income
and household size (to adjust income to needs). Table 5 reports the
coefficients for illness, caring and ethnic group from this model. The full tables
can be found in Tables A1 and A2 in the Appendix. Income itself had a
statistically significant and negative effect across all four equations for both
men and women. That is (as would be expected), social participation
increased with income level, across all measures. The cross-equation
correlations (the rhos) remained very similar for the two models for both men
and women, so the argument that there is an underlying propensity to lack of
social engagement holds for these models. Given the non-response on
income, the models with and without income are based on different samples
and so cannot be directly compared. Nevertheless, if we find sizeable and
statistically significant results for ethnic group or health status following the
23
inclusion of income in the model, it suggests that that these effects are
insensitive to income. That is, it is not the differences in poverty between
groups that account for the differences in patterns of participation.
Conversely, if income differences are playing a significant part in the patterns
of social participation that we observe, then we would expect to find negligible
and statistically insignificant coefficients on ethnic group and health status,
indicating that the observed differences between groups can be attributed to
their very different experiences and characteristics, including in terms of
household resources.
For men, coefficients reduced substantially in size and became statistically
insignificant in relation to going out for carers and for Pakistani, Bangladeshi
and Black Caribbean men. This suggests that variations in patterns of going
out for these groups are heavily dependent on access to resources, as might
be expected. However, Black African men remained significantly more likely
to miss out on going out than their white British counterparts of the same
income level, as well as to lack visiting. And, male carers and Black
Caribbean and Chinese men were significantly more likely to lack reciprocal
visiting than their non-caring / white British counterparts. Access to resources
is thus important for enabling some forms of social participation, especially
going out, for men from some marginalised groups, refuting the argument that
forms of social contact are culturally rather than economically driven. On the
other hand, different patterns of neighbourly visiting appear relatively
independent of income, suggesting different access to forms of ‘ethnic capital’
and potentially the protective use of community resources.
24
For women carers, restrictions on social participation appeared relatively
insensitive to income. Among those with long-term illness, however, patterns
of going out and visiting were statistically insignificant from their healthy
counterparts at the same income level. Interestingly Pakistani and
Bangladeshi women’s participation was insignificantly different from
comparable white British women’s with the exception of participation in
organised activities. On the other hand, Indian, Black Caribbean and Black
African women were significantly less likely to participate on three out of the
four measures than their white British peers. For the Bangladeshi and
Pakistani women it could be argued that the observed differentials in Table 2
were in large part caused by the lack of resources wherewith to engage in
social participatory activities, despite the fact that it is for these groups that
cultural constraints on participation are most frequently invoked (Tackey et al.
2006; and cf. Dale 2002).
By contrast, among those Black women who do not tend to be considered so
culturally constrained, we find that their patterns of participation are
insensitive to income and, in some cases, the size of the coefficient even
seemed to increase, once income was controlled. In these instances, it is
those who are not necessarily the worst off who nevertheless experience
lower participation. If we take seriously the notion of lack of participation as a
form of poverty we should consider as deprived those who lack opportunities
for participation despite their income level. And we can see that this lack of
25
social participation could not be simply fixed by higher incomes or
participation in employment.
[TABLE 5 ABOUT HERE]
Adding up social engagement
This final section treats the question of what these different patterns of social
participation all add up to. In the preceding sections the focus has been on
the distributions of the individual measures of sociability across potentially
vulnerable groups. On the other hand, early on the question was raised of
whether there were some substitution effects, which might indicate that it was
preferences for particular types of sociability rather than constraints that
determined the patterns of social engagement. Here, predicted probabilities
for being deprived on all four measures or on none are estimated. Do the
differential risks on individual measure add up to mean that the chances of
overall social engagement or lack of social engagement are highly unevenly
distributed? To examine the marginal impact of ethnic group and of illness
and caring, other characteristics were held constant across groups for these
predictions;8 and the predicted probabilities from the models for men and
women were compared. Figures 1 and 2 show the predicted probabilities of
having ‘zeros’ on all four measures (full social engagement), or having ‘ones’
on all four measures (lacking social engagement on all counts).
Figure 1 shows the patterns contrasting those who were not ill and not caring
with those who were. This comparison is made purely for white British men
26
and women. While empirically there were few in the sample who were both ill
and caring and who fitted the baseline characteristics selected, the point is
heuristic. It shows the extreme case of being both long-term ill and caring and
the effect that has on extreme positions of sociability. For both men and
women, the combined effects of being ill and caring caused an approximately
four percentage point reduction in the chances of not being deprived on any of
the four measures. For risks of being deprived on all four measures,
however, the impact was greater for women than for men. Men who were
caring and long-term ill were only one percentage point more likely to not be
participating on any of the four measures than their well, non-caring
counterparts. For women, by contrast, the increase in risk was of the nature
of three percentage points. These effects are relatively small. The risks of
extreme social isolation are not as great as we might have anticipated for
those experiencing both ill-health and caring responsibilities. There would
appear to be some balancing between different forms of social engagement
that means that still relatively few are constrained in relation to all four
measures of sociability.
[FIGURE 1 ABOUT HERE]
On the other hand, the picture was much starker for those from certain
minority ethnic groups. Figure 2 illustrates the predicted probabilities for men
and women from six ethnic groups. (Chinese are excluded because the small
sample sizes make some of the coefficients unreliable, and Mixed and other is
27
both a residual category and differed little from the white British in all models.)
The health and caring status for Figure 2 was set as not ill and not caring.
Figure 2 shows that, for women, all the minority groups had substantially
lower probabilities than their white counterparts of not being deprived on any
measure, with the Bangladeshis, Black Caribbeans and Black Africans
clustering together at around 19 percent. They also had lower probabilities
than their male counterparts across all groups. Women from minority groups
also had higher probabilities of being deprived on all four measures than their
white British counterparts, with the position of Black Caribbean and Black
African women standing out here. Black African and Black Caribbean men
also fared badly in relation to participation by comparison with men from other
groups (though not by comparison with women from the same group). And
the situation of Pakistani and Bangladeshi men indicates the centrality of
sociability for them – other things being equal.
Comparing Figure 2 with Figure 1, we can see that the predicted probability of
limited sociability on all four measures for white women who were ill and
caring was still well below the predictions for Black Caribbean and Black
African women, which were estimated at 12 and 13 per cent respectively.
This emphasises once again the particularly deprived position of Black
Africans and Black Caribbeans, and especially women from these groups, in
relation to their constrained patterns of social engagement. If social support
and social contact is an important element of current well-being, as well as of
28
the welfare of future generations, then this result should give us cause for
concern.
[FIGURE 2 ABOUT HERE]
Conclusions
Results showed that there was indeed variation in patterns of social
participation by both illness/caring and ethnic group. However, the impact of
illness on opportunities for social participation was not substantial once
relevant factors were held constant. That does not mean that, in absolute
terms, those with a long-term illness are not more isolated than those without,
but this would appear to have less to do with the illness than with other
associated characteristics (such as age, income and qualifications). For
carers, particularly male carers, their responsibilities did seem to create more
constraints on their social activity. And only some of this can be attributed to
income: caring responsibilities in their own right appeared to divorce people
from opportunities for extensive social engagement. However, there seemed
to be some substitution between different forms of social engagement, and
carers, even ill carers, were not much more likely to be deprived on all four
measures than non-carers.
Turning to or main focus on ethnic group, there was clear variation in patterns
of social engagement by ethnic group. In particular there were distinctively
high levels of reciprocal visiting among Bangladeshis and Pakistanis – and
particularly men from these groups. Conversely, Black Africans were more
29
likely to engage in more organised activities and less in the more informal
social activities. Black Caribbeans showed more differences from the white
majority in their patterns of social participation given that they are the minority
group which shows the highest overall levels of geographical and economic
integration and analysis frequently reveals their comparability with the white
majority. Given that these differences tended to be in terms of deficits – that is
lower levels of social engagement, they in a sense echo the analysis of the
economic and labour market experience of Caribbeans which shows overall
compability but with some clear deficits and disadvantages – particularly in
unemployment rates, that suggest on-going processes of racialised exclusion.
These distinctive patterns of social participation do give some support to
notions of groupness associated with the ethnic group categories, as
discussed in the introduction – and this notion of groupness is further
supported by the fact that patterns of participation tended to be more similar
between the sexes within a group than within the sexes. Patterns of
participation on their own did not closely match the wel-attested ranking of
income or material deprivation across the groups, suggesting that this form of
deprivation – of social engagement – is best considered in its own right.
Multivariate analysis enabled consideration of the extent to which the absolute
differences could be attributed to differences in characteristics and
circumstances. Among Bangladeshi men and Pakistani women, the contrast
between visiting and going out was clear even when most characteristics
were controlled – but the lower tendency to go out did seem to be associated
with income constraints. In the final analysis, Bangladeshi and Pakistani men
30
appeared to have a greater propensity towards (or opportunities for) social
engagement and were less at risk of social isolation than other groups. This
may suggest the important role of social contact in ameliorating otherwise
very disadvantaged circumstances – and indicate the potentially positive role
of co-residence in local communities in facilitating that. The results indicate
that a focus on improving material circumstances of these groups would have
pay-offs in extending their range of social contacts through different forms of
activity. Indian men appeared little different from comparable white British
men; however Indian women were much more likely to experience low
participation across activities than comparable white women from the same
income bracket, suggesting that rather different opportunities across the
sexes within this group and possibly access to household resources. Indian
women have not experienced the policy attention paid to Bangladeshi and
Pakistani women, whose low levels of labour market participation (often,
thought not necessarily accurately or informatively attributed to ‘cultural
barriers’) are often regarded as the key to improving the welfare of the groups
overall. But these findings suggest that in terms of social contact, there are
constraints facing Indian women that go beyond the material circumstances of
the household.
However, it is when turning to the experience of the Black African and Black
Caribbean groups that the most distinctive finding emerges. The Black
Caribbean group is the longest-standing of Britain’s post-war minority groups
and, as noted, is the most ‘integrated’ according to conventional measures of
labour market participation, geographical distribution, inter-ethnic unions and
so on. In particular Caribbean women’s high labour market participation rates
31
and the greater tendency than among the white majority for Caribbean lone
parents to combine work and childcare are noted alongside the high levels of
qualifications among women from these groups. While young Caribbean
men’s unemployment rates are a cause for substantial concern, the group as
a whole is experiencing upward mobility and increasingly its profile is
becoming comparable to that of the population as a whole (Platt 2005). Black
Africans tend to be more recently arrived and more highly concentrated in
London, but also often bring high levels of skills and qualifications, which
might be expected to increase opportunities for participation. But these results
show how that both these groups experience distinctively higher risks of not
engaging in opportunities for social participation, and that these are not
associated with characteristics or family or employment status (which can
place extensive demands on women in particular (Reynolds 2005). In
particular we observe that the highest probabilities of not participating in any
of the four forms of activity are reserved for Black African and Black
Caribbean women. As mentioned, much has been made of the potential
cultural and religious constraints on the activity of women from South Asian
groups. And it is important to remember the large absolute differences for
South Asian women. However, if we are concerned about the well-being
offered by social contact and the deprivation suffered by its lack more
attention should be paid to supporting the networks and combating the
exclusionary processes faced by Black women; and how to build social capital
within groups which are often considered well-integrated into labour markets
and subject to relatively little policy attention. These women may have
alternative forms of informal social contact not measured here, but it is hard to
32
imagine what they would look like, and the finding invites further investigation.
Meanwhile, it serves as an important caveat about the potential costs of
‘integration’ in a stratified and racialised society.
Notes
1 Acknowledgments: I would like to thank the Home Office /Communities and Local Government for the use of these data and the Data Archive at the University of Essex for making them available to me. I am, of course, solely responsible for the analysis and for the interpretation made of them. I would also like to thank CASE at the LSE for hosting me while I worked on the first version of this paper, and Howard Glennerster for his positive feedback on the working paper version. The University of Essex allowed me leave during which I worked on this paper (among other projects). Thanks go to the British Academy, which funded my presence at the RC28 Montreal meeting where I presented the current version of the paper. I have received helpful comments from participants at seminars at Loughborough University, Institute of Education, University of London, University of Essex, Stratification Seminar Cardiff meeting, comments which I have aimed to work through in what is presented here. I am grateful to colleagues on the Joseph Rowntree Foundation funded project on Long-term ill health, poverty and ethnicity, a project which prompted my work on this paper, and to Sarah Salway for her detailed comments on an early version of the paper. 2 The categories typically employed in secondary analysis (and available to analysts) are now the ONS 2001 Census categories. There are typically 15 such categories in data covering England and Wales. In this paper these have been reduced to 8 since the ‘other’ categories are not considered independently meaningful and the ‘mixed’ categories are not considered meaningful in aggregate and are too small individually to allow separate analysis. 3 In 2003 and 2005 versions of the survey questions on ethnic group of friendship networks was asked. However, it was not asked in 2001, and in the later surveys, the questions on social practices were absent. 4 ONS have defined five dimensions of social capital: views about the local area; civic participation; social networks and support; social participation; reciprocity and trust. The approach adopted here does not fit clearly into one of these dimensions, but it links social networks and support and social participation. 5 The lower age of 18 rather than 16 was chosen as few 16 and 17 year olds are in fact in employment or can be considered fully independent. While 16 is the minimum school leaving age, 18 is the age of majority. 6 It could be argued that such correlations are picking up characteristics that are potentially measurable that we have failed to include in the equation rather than ‘unobservables’, i.e. an unmeasurable latent propensity for sociability or, conversely, self-sufficiency. However, we include a wide range of independent variables and have tested for a number of others, which were found not to contribute to the explanatory power of the model and were therefore excluded. 7 The multivariate probit model was estimated using the –mvprobit– program for Stata software (Capellari and Jenkins, 2003). Each model was run with 30 draws and using antithetic acceleration in the interests of robustness of results. Predicted probabilities were derived using the companion post-estimation program –mvpred–. I am grateful to Stephen Jenkins for his advice on the use of mvprobit. 8 Characteristics were set to: younger, partnered, no child under five, in employment, with level three qualifications and with annual household income of between £30 000 and £34 999
33
References
Aldridge, S., Halpern, D. with Fitzpatrick, S. 2002 Social Capital: A discussion
paper, London: Performance and Innovation Unit.
Attwood. C., Singh, G., Prime, D., Creasey, R. and others 2003 2001
Citizenship Survey: People, families and communities, Home Office
Research Study 270, London: Home Office.
Bajekal, M., Blane, D., Grewal, I., Karlsen, S. and Nazroo, J. 2004 ‘Ethnic
differences in influences on quality of life at older ages: a quantitative
analysis’, Ageing and Society 24:709-728.
Berkman, L.F. 1984 ‘The relationship of social networks and social support to
morbidity and mortality’ in S. Cohen and S.L. Syme (eds) Social
Support and Health, San Diego: Academic Press.
Berkman, L.F., Glass, T., Brissette, I. and Seeman, T.E. 2000 ‘From social
integration to health: Durkheim in the new millennium’, Social Science
& Medicine 51: 843-857.
Berthoud, R., Bryan, M.L. and Bardasi, E. 2004 The Dynamics of Deprivation:
the relationship between income and material deprivation over time,
London: Department for Work and Pensions.
Borjas, G. 1992 ‘Ethnic capital and intergenerational mobility’, Quarterly
Journal of Economics 107: 123-50.
Bourdieu, P. 1997. ‘The Forms of Capital’ in A. Halsey, H. Lauder, P. Brown
and A. Stuart Wells (eds) Education: Culture, Economy, Society,
Oxford University Press.
Burchardt, T. 2000 Enduring Economic Exclusion: Disabled people, income
and work, York: Joseph Rowntree Foundation.
34
Burchardt, T. 2002 ‘Measuring social exclusion’ in J. Hills, J. LeGrand and D.
Piachaud (eds) Social Exclusion. Cambridge: Cambridge University
Press.
Cappellari, L. and Jenkins, S.P. 2003 ‘Multivariate Probit Regression using
Simulated Maximum Likelihood’, The Stata Journal 3(3): 278-94.
Cappellari, L. and Jenkins, S.P. 2007 ‘Summarising Multiple Deprivation
Indicators’ in S.P. Jenkins and J. Micklewright (eds) Poverty and
Inequality Re-examined, Oxford: Oxford University Press.
Clark, K. and Drinkwater, S. 2002 ‘Enclaves, neighbourhood effects and
economic outcomes: Ethnic minorities in England and Wales’, Journal
of Population Economics 15: 5-29.
Coleman, J. 1988 ‘Social capital in the creation of human capital’, American
Journal of Sociology 94: S95-S120.
Commission on Integration and Cohesion 2007 Our Shared Future: Final
Report of the Commission on Integration and Cohesion. London:
Communities and Local Government.
Community Cohesion Review Team 2001 Community Cohesion: A Report of
the Independent Review Team Chaired by Ted Cantle. London: Home
Office.
Dale, A. 2002 ‘Social exclusion of Pakistani and Bangladeshi women’,
Sociological Research Online 7 (3).
Department for Work and Pensions [DWP] 2006 A New Deal for Welfare:
Empowering people to work, CM 6730, London: The Stationery Office.
Duffy, B. 2004. ‘Life Satisfaction and Trust in Other People’, London: MORI.
Erens, B., Primatesta, P. and Prior, G. 2001 The Health Survey for
35
England. The Health of Minority Ethnic Groups 1999, London: The
Stationery Office.
Edwards, R. Franklin, J., Holland, J. 2006 Assessing Social Capital: Concept
Policy and Practice. Newcastle: Cambridge Scholars Press.
Esser, H. 2004 ‘Does the ‘New’ Immigration Require a ‘New’ Theory of
Intergenerational Integration?’, International Migration Review 38(3):
1126-59.
Furbey, R., A. Dinham, R. Farnell, D. Finneron, G. with Wilkinson, C.
Howarth, D. Hussain, and S. Palmer. 2006 Faith as Social Capital.
Bristol: The Policy Press.
Gordon, D. and Pantazis, C. 1997 Breadline Britain in the 1990s, Aldershot:
Ashgate.
Gordon, D., Adelman, L., Ashworth, K., Bradshaw, J., Levitas, R., Middleton,
S., Pantazis, C., Patsios, D., Payne, S., Townsend, P. and Williams, J.
2000 Poverty and Social Exclusion in Britain, York: Joseph Rowntree
Foundation.
Granovetter, M. 1973 ‘The Strength of Weak Ties’, American Journal of
Sociology 78 (4): 1350-80.
Green, H., Connolly, H. and Farmer, C. 2004 2003 Home Office Citizenship
Survey: People, Families and Communities, Home Office Research
Study 289, London: Home Office.
Heath, A. and Cheung, S.Y. 2006 Ethnic Penalties in the Labour Market:
Employers and discrimination, Department for Work and Pensions
Research Report 341. Leeds Corporate Document Services.
36
Home Office Communities Research and BMRB International 2003 Home
Office Citizenship Survey 2001 [computer file], Colchester, Essex: UK
Data Archive [distributor], November 2003, SN: 4754.
Howard, M. 2001 Paying the Price: Carers, poverty and social exclusion,
London: Child Poverty Action Group.
Hudson, Maria, Joan Phillips, Kathryn Ray, and Helen Barnes. 2007 Social
cohesion in diverse communities. York: Joseph Rowntree Foundation.
Jenkins, S.P. and Rigg, J.A. 2004 ‘Disability and Disadvantage: selection,
onset and duration effects’, Journal of Social Policy 33 (3): 479-501.
Li, Y., Savage, M. and Pickles, A. 2003 ‘Social Capital and Social Exclusion in
England and Wales, 1972-1999’, British Journal of Sociology 54 (4):
497-526.
Li, Y., Savage, M., Tampubolon, G., Warde, A. and Tomlinson, M. 2002
‘Dynamics of social capital; trends and turnover in associational
membership in England and Wales, 1972-1999’, Sociological Research
Online 7 (3): U97-132
Lin, N. 2001 Social Capital: A theory of social structure and action,
Cambridge: Cambridge University Press.
Locker, D. 1983 Disability and Disadvantage: The consequences of chronic
illness, London: Tavistock.
Loizos, P. 2000 ‘Are Refugees Social Capitalists?’ in S. Baron, J. Field and T.
Schuller (eds) Social Capital: critical perspectives, Oxford University
Press.
Mack, J. and Lansley, S. 1985 Poor Britain, London: Allen & Unwin.
37
Murphy, R., Wedlock, E. and King, J. 2005 Early Findings from the 2005
Home Office Citizenship Survey, Home Office Online Report 49/05.
Modood, Tariq, Richard Berthoud, J. Lakey, J. Nazroo, P. Smith, S. Virdee,
and S. Beishon. 1997 Ethnic minorities in Britain: diversity and
disadvantage. London: Policy Studies Institute.
Nazroo, J. Y. 1997 The Health of Britain's Ethnic Minorities, London: Policy
Studies Institute.
Nolan, B. and Whelan, C. 1996 Resources, Deprivation and Poverty, Oxford:
Clarendon Press.
Office for National Statistics [ONS] 2001 Social Capital: A Review of the
Literature, London: ONS.
Parker, G. 1993 With This Body: caring and disability in marriage,
Buckingham: Open University Press.
Peach, C. 2005 ‘Social integration and social mobility: segregation and
intermarriage of the Caribbean population in Britain’ in G.C. Loury, T.
Modood, and S.M. Teles (eds) Ethnicity, Social Mobility and Public
Policy: Comparing the US and UK. Cambridge: Cambridge University
Press.
Pennant, R. 2005 Diversity, Trust and Community Participation in England,
Home Office Findings 253, London: Home Office.
Petersen, T., Saporta, I. and Seidel M.-D. L. 2000. ‘Offering a job: meritocracy
and social networks’, American Sociological Review 106 (3): 763-816.
Platt, L. 2002 Poverty among Britain’s Ethnic Minority Groups. London:
CPAG.
Platt, L. 2007 Poverty and Ethnicity in the UK. Bristol: The Policy Press.
38
Pollard, T. M., Carlin, L. E., Bhopal, R., Unwin, N., White, M. and Fischbaher,
C. 2003 ‘Social Networks and Coronary Heart Disease Risk Factors in
South Asians and Europeans in the UK’, Ethnicity & Health 8(3): 263-
75.
Portes, A. 1998 ‘Social Capital: It’s origins and applications in modern
sociology’, Annual Review of Sociology 24: 1-24.
Putnam, R.D. 1995 ‘Bowling alone: America’s declining social capital’,
Journal of Democracy 61: 65-78.
Putnam, R.D. 2000 Bowling Alone: The collapse and revival of American
community, New York: Simon Schuster.
Salway, S., L. Platt, P. Chowbey, and K. Harriss. 2007 Long-term Illness,
Poverty and Ethnicity: Understanding the experiences of sufferers and
their families. Bristol: The Policy Press.
Schuller, T., Baron, S. and Field, J. 2000 ‘Social Capital: a review and critique’
in S. Baron, J. Field and T. Schuller (eds) Social Capital: Critical
Perspectives, Oxford University Press.
Stopes-Roe, M. and Cochrane, R. 1990 ‘Support Networks of Asian and
British Families: comparisons between ethnicities and between
generations’, Social Behaviour 5: 71-85.
Tackey, N.D., Casebourne, J., Aston, J., Ritchie, H., Sinclair, A., Tyers, C.,
Hurstfield, J., Willison, R., Page, R. 2006 Barriers to employment for
Pakistanis and Bangladeshis in Britain. Department for Work and
Pensions Research Report No 360. Leeds: Corporate Document
Services.
39
Townsend, P. 1979 Poverty in the United Kingdom: A survey of household
resources and standards of living, London: Penguin.
Veenstra, G. 2000 ‘Social Capital, SES and Health: an individual-level
analysis’, Social Science and Medicine 50 (5): 619-629.
Warde, A., Tampubolon, G., Longhurst, B., Ray, K., Savage, M. and
Tomlinson, M. 2003 ‘Trends in Social Capital: membership of
associations in Great Britain’, British Journal of Political Science 33:
515-525.
Whelan, C.T. 1993 ‘The role of social support in mediating the psychological
consequences of economic stress’, Sociology of Health and Illness 15
(1): 86-101.
Zhou, M. 2005 ‘Ethnicity as social capital: community-based institutions and
embedded networks of social relations’ in G.C. Loury, T. Modood, and
S.M. Teles (eds) Ethnicity, Social Mobility and Public Policy:
Comparing the US and UK. Cambridge: Cambridge University Press.
40
Figures and Tables Figure 1: Predicted probabilities for proportions experiencing no lack of social engagement on any of four measures and those experiencing lack of social engagement on all four measures, by sex and caring/health status
Source: Home Office Citizenship Survey 2001, author’s analysis Notes: weights are used to estimate these proportions. Unweighted Ns: men=5 406; women=6 310. Horizontal axis represents the probability.
41
Figure 2: Predicted probabilities for proportions of those experiencing no lack of social engagement on any of four measures and of those experiencing lack of social engagement on all four measures, by sex and ethnic group, 2001
Notes as for Figure 1
42
Table 1: Percentage with limited forms of social engagement according to four measures, by ethnic group: men aged 16-64
Ethnic group (N -unweighted)
Infrequent visits
Infrequent visiting
Infrequent going out
Low contact
with clubs
Not socially
deprived on any of
the 4
Socially deprived
on all 4
White British (2842)
37 41 29 46 30 10
Indian (535)
34 38 31 44 32 11
Pakistani (401)
32 34 45 48 24 10
Bangladeshi (255)
23 31 43 49 28 11
Caribbean (322)
40 42 36 57 22 12
Black African (260)
39 50 44 42 20 14
Chinese (58)
54 57 24 56 18 19
Mixed and other (730)
35 39 34 45 28 10
Long-term ill (910)
41 50 40 49 20 11
Caring (423)
48 60 43 45 21 14
ALL men (5403)
37 41 30 46 29 10
Source: Home Office Citizenship Survey 2001, author’s analysis Notes: percentages are based on weighted counts and have been rounded up.
43
Table 2: Percentage with limited forms of social engagement according to four measures, by ethnic group: women aged 16-59
Ethnic group (N -unweighted)
Infrequent visits from friends or
neighbours
Infrequent visiting to friends or
neighbours
Infrequent going out
Low contact
with clubs
Not socially
deprived on any
measure
Socially deprived
on all four measures
White British (3271)
34 36 36 47 27 10
Indian (590)
36 39 45 54 23 16
Pakistani (414)
28 38 55 64 16 17
Bangladeshi (289)
29 30 56 68 11 11
Caribbean (456)
48 52 49 47 15 13
Black African (393)
45 54 56 40 19 15
Chinese (80)
45 54 32 46 26 8
Mixed and other (844)
36 37 35 49 30 11
Long-term ill (986)
37 42 48 54 20 12
Caring (683)
39 45 48 46 22 15
ALL women (6307)
35 37 36 48 27 10
Notes: as for Table 1
44
Table 3: Multivariate probit regression estimates of the effects of various characteristics on measures of lack of social engagement, 2001: Men [1]
Infrequent visits from friends or neighbours
[2] Infrequent visits to friends or neighbours
[3] Infrequent going out
[4] Irregular/ no involvement in organised activities
Older age group .593 (.055)*
.607 (.054)*
.543 (.057)* .052 (.055)
Partnered .071 (.058) .349 (.059)*
.571 (.061)* -.103 (.078)†
With child under 5 -.088 (.080)
.095 (.077) .378 (.077)* .201 (.077)*
Long-term ill -.042 (.070)
.130 (.075) .108 (.070) -.014 (.076)
Caring for someone .217 (.090)*
.389 (.093)*
.193 (.092)* -.052 (.092)
Ethnic Group Indian -.011
(.080) -.048
(.084) .073 (.075) -.010 (.075)
Pakistani -.075 (.094)
-.201 (.093)*
.348 (.087)* -.043 (.091)
Bangladeshi -.355 (.140)*
-.282 (.126)*
.329 (.125)* -.113 (.124)
Black Caribbean .154 (.103 .180 (.106)†
.318 (.101)* .216 (.101)*
Black African .288 (.106)*
.508 (.114)*
.630 (.109)* .006 (.110)
Chinese .688 (.192)*
.699 (.199)*
-.029 (.223) .337 (.201)
Other and mixed .052 (.097) .052 (.101) .191 (.094)* .036 (.098) Qualifications level 1 -.130
(.121) -.263
(.121)* .006 (.127) -.235 (.120)*
level 2 -.291 (.086)*
-.323 (.086)*
-.155 (.087)† -.419 (.085)*
level 3 + apprenticeships
-.285 (.074)*
-.287 (.074)*
-.237 (.075)* -.441 (.075)*
higher / higher diplomas
-.407 (.075)*
-.413 (.074)*
-.277 (.074)* -.662 (.075)*
other -.216 (.179)
.053 (.171) -.177 (.174) -.090 (.178)
Work history Not currently in employment
-.173 (.073)*
-.145 (.070)*
.271 (.070)* -.005 (.073)
Never worked -.501 (.149)*
-.293(.153)
†
.151 (.166) -.366 (.203)†
Constant -.395 (.081)*
-.558 (.080)*
-1.219 (.085)*
.343 (.081)*
45
Rho eq. 2-4 .205 (.031)* Rho eq. 1-4 .158 (.030)* Rho eq. 3-4 .283 (.029)* Rho eq. 1-2 .882 (.011)* Rho eq. 2-3 .550 (.025)* Rho eq. 1-3 .527 (.027)* Likelihood ratio test of rho14=rho24=rho34=rho13=rho23=rho12=0
Chi2 (6)=359.448*
Source: Home Office Citizenship Survey 2001, author’s analysis Notes: reference categories for the categorical variables are younger (aged 18-39); single; no child under 5; not long-term ill; no caring responsibilities; white British; no qualifications; in employment. Standard errors are given in brackets. *= statistically significant at at least the 5% level; †= statistically significant at the 10% level. Survey weights are used. Unweighted number of observations=5300.
46
Table 4: Estimates from a multivariate probit indicating the effects of various characteristics on measures of lack of social engagement, 2001: Women [1]
Infrequent visits from friends or neighbours
[2] Infrequent visits to friends or neighbours
[3] Infrequent going out
[4] Irregular/ no in-volvement in organised activities
Older age group .400 (.053)* .507 (.052)*
.448 (.053)*
-.041 (.052)
Partnered .249 (.051)* .278 (.050)*
.289 (.049)*
.070 (.050)
With child under 5 -.155 (.067)* -.039 (.070)
.469 (.065)*
-.030 (.064)
Long-term ill .067 (.067) .118 (.069)†
.164 (.066)*
.097 (.067)
Caring for someone .066 (.074) .149 (.074)*
.182 (.075)*
-.135 (.064)†
Ethnic Group Indian .100 (.076) .151
(.076)* .246 (.074)*
.123 (.076)
Pakistani -.040 (.106) .205 (.109)†
.313 (.107)*
.413 (.104)*
Bangladeshi .066 (.166) .081 (.147) .264 (.138)†
.424 (.138)*
Black Caribbean .483 (.084)* .528 (.085)*
.446 (.084)*
.073 (.085)
Black African .513 (.098)* .728 (.096)*
.616 (.097)*
-.111 (.096)
Chinese .439 (.154)* .675 (.179)*
.084 (.164) .088 (.180)
Other and mixed .138 (.098) .139 (.095) .045 (.088) .134 (.090) Qualifications level 1 -.145 (.094) -.114
(.095) -.225 (.092)*
-.363 (.091)*
level 2 -.114 (.073) -.113 (.071)
-.314 (.072)*
-.495 (.072)*
level 3 + apprenticeships
-.212 (.080)* -.218 (.083)*
-.414 (.079)*
-.704 (.080)*
higher / higher diplomas
-.228 (.071)* -.256 (.071)*
-.571 (.069)*
-.926 (.070)*
other -.459 (.236)† -.366 (.233)
-.207 (.232)
-.679 (.230)*
Work history Not currently in employment
-.257 (.057)* -.217 (.060)*
.200 (.055)*
-.027 (.057)
Never worked -.170 (.131) -.245 (.128)
.400 (.120)*
-.037 (.121)
Constant -.552 (.077)* -.609 -.658 .454
47
(.075)* (.076)* (.076)* Rho eq. 2-4 .212 (.029)* Rho eq. 1-4 .167 (.029)* Rho eq. 3-4 .203 (.028)* Rho eq. 1-2 .863 (.011)* Rho eq. 2-3 .466 (.025)* Rho eq. 1-3 .477 (.026)* Likelihood ratio test of rho14=rho24=rho34=rho13=rho23=rho12=0
chi2(6) = 679.47*
Notes: as for Table 3. Unweighted number of observations=6167.
48
Table 5: Multivariate probit regression estimates of the effects of ethnic group and health status on measures of lack of social engagement controlling for income. [1]
Infrequent visits from friends or neighbours
[2] Infrequent visits to friends or neighbours
[3] Infrequent going out
[4] Irregular/ no involvement in organised activities
MEN Long-term ill -.111
(.080) .052 (.084) .016 (.078) -.101 (.090)
Caring for someone .256 (.102)*
.355 (.102)*
.136 (.104)* -.051 (.104)
Ethnic Group Indian .074 (.099) -.009
(.099) -.077. (.099) .058 (.095)
Pakistani .020 (.129) -.220 (.124)
.163 (.119) -.101 (.124)
Bangladeshi -.407 (.162)*
-.256 (.160)
-.008 (.160) -.010 (.157)
Black Caribbean .234 (.121)†
.209 (.122)†
.191 (.120) .119 (.128)
Black African .169 (.120) .311 (.125)*
.487 (.130)* -.029 (.125)
Chinese .580 (.230)*
.812 (.257)*
-.425 (.301) -.125 (.289)
Other and mixed .069 (.110) .119 (.121) .210 (.107)* .120 (.114) WOMEN Long-term ill .028 (.076) .058 (.078) .079 (.077) .067 (.078) Caring for someone .215
(.085)* .164 (.085)†
.168 (.087)† -.124 (.086)
Ethnic Group Indian .186
(.101)† .153 (.010)*
.225 (.096)* .269 (.100)*
Pakistani .025 (.134) .007 (.142) .108 (.145) .409 (.144)* Bangladeshi .192 (.195) .084 (.177) -.133 (.169) .598 (.167)* Black Caribbean .519
(.097)* .474 (.097)*
.344 (.100)* .065 (.102)
Black African .552 (.109)*
.662 (.109)*
.503 (.109)* -.063 (.113)
Chinese .502 (.188)*
.529 (.210)*
-.123 (.207) .201 (.229)
Other and mixed .067 (.114) .073 (.112) -.077 (.105) .063 (.105) Notes: As for Table 3. Unweighted number of observations= 3904 (men) and 4369 (women). Full results in Appendix: Tables A1 and A2
49
Table A1: Multivariate probit regression estimates of the effects of various characteristics including income on measures of lack of social engagement, 2001: Men [1]
Infrequent visits from friends or neighbours
[2] Infrequent visits to friends or neighbours
[3] Infrequent going out
[4] Irregular/ no involvement in organised activities
Older age group .589 (.063)*
.543 (.063)*
.532 (.065)* -.015 (.062)
Partnered .167 (.077)*
.371 (.079)*
.655 (.083)* -.001 (.079)
With child under 5 .011 (.093) .096 (.092) .301 (.091)* .221 (.092)* Long-term ill -.111
(.080) .052 (.084) .016 (.078) -.101 (.090)
Caring for someone .256 (.102)*
.355 (.102)*
.136 (.104)* -.051 (.104)
Ethnic Group Indian .074 (.099) -.009
(.099) -.077. (.099) .058 (.095)
Pakistani .020 (.129) -.220 (.124)
.163 (.119) -.101 (.124)
Bangladeshi -.407 (.162)*
-.256 (.160)
-.008 (.160) -.010 (.157)
Black Caribbean .234 (.121)†
.209 (.122)†
.191 (.120) .119 (.128)
Black African .169 (.120) .311 (.125)*
.487 (.130)* -.029 (.125)
Chinese .580 (.230)*
.812 (.257)*
-.425 (.301) -.125 (.289)
Other and mixed .069 (.110) .119 (.121) .210 (.107)* .120 (.114) Qualifications level 1 -.095
(.145) -.250
(.149)† .005 (.152) -.365 (.141)*
level 2 -.241 (.100)*
-.252 (.101)*
-.075 (.101) -.461 (.100)*
level 3 + apprenticeships
-.177 (.088)*
-.171 (.089)†
-.127 (.089)* -.413 (.090)*
higher / higher diplomas
-.324 (.092)*
-.324 (.094)*
-.049 (.092)* -.687 (.094)*
other -.172 (.204)
.005 (.202) -.172 (.221) -.010 (.214)
Work history Not currently in employment
-.224 (.085)*
-.157 (.087)†
.042 (.083) -.077 (.089)
Never worked -.105 (.192)
.173 (.198)†
.467 (.249)† -.140 (.209)
Income band -.026 (.011)*
-.025 (.012)*
-.080 (.012)* -.026 (.011)*
Household size -.062 -.038 .043 (.028) -.067 (.030)*
50
(.028)* (.029) Constant -.169
(.113) -.299
(.120)* -.860 (.121)* .667 (.122)*
Rho eq. 2-4 .182 (.035)* Rho eq. 1-4 .132 (.035)* Rho eq. 3-4 .285 (.034)* Rho eq. 1-2 .885 (.012)* Rho eq. 2-3 .549 (.028)* Rho eq. 1-3 .522 (.030)* Likelihood ratio test of rho14=rho24=rho34=rho13=rho23=rho12=0
Chi2 (6)=164.004*
Notes: As for Table 3. Unweighted number of observations=3904.
51
Table A2: Estimates from a multivariate probit indicating the effects of various characteristics including income on measures of lack of social engagement, 2001: Women [1]
Infrequent visits from friends or neighbours
[2] Infrequent visits to friends or neighbours
[3] Infrequent going out
[4] Irregular/ no in-volvement in organised activities
Older age group .398 (.061)* .493 (.061)*
.467 (.061)*
-.025 (.061)
Partnered .346 (.076)* .399 (.067)*
.316 (.067)*
.228 (.069)*
With child under 5 -.056 (.081)* -.118 (.082)
.356 (.078)*
.017 (.077)
Long-term ill .028 (.076) .058 (.078) .079 (.077) .067 (.078) Caring for someone .215 (.085)* .164
(.085)† .168 (.087)†
-.124 (.086)
Ethnic Group Indian .186 (.101)† .153
(.010)* .225 (.096)*
.269 (.100)*
Pakistani .025 (.134) .007 (.142) .108 (.145) .409 (.144)*
Bangladeshi .192 (.195) .084 (.177) -.133 (.169)
.598 (.167)*
Black Caribbean .519 (.097)* .474 (.097)*
.344 (.100)*
.065 (.102)
Black African .552 (.109)* .662 (.109)*
.503 (.109)*
-.063 (.113)
Chinese .502 (.188)* .529 (.210)*
-.123 (.207)
.201 (.229)
Other and mixed .067 (.114) .073 (.112) -.077 (.105)
.063 (.105)
Qualifications level 1 -.234 (.111)* -.125
(.113) -.223 (.109)*
-.403 (.108)*
level 2 -.142 (.087) -.094 (.087)
-.203 (.087)*
-.426 (.087)*
level 3 + apprenticeships
-.229 (.100)* -.164 (.104)
-.317 (.098)*
-.626 (.099)*
higher / higher diplomas
-.270 (.093)* -.178 (.091)†
-.375 (.089)*
-.791 (.089)*
other -.553 (.243)* -.387 (.116)*
-.105 (.262)
-.653 (.239)*
Work history Not currently in employment
-.338 (.068)* -.281 (.071)*
.132 (.067)*
-.094 (.069)
Never worked -.362 (.158)* -.294 (.149)*
.606 (.151)*
.167 (.147)
52
Income band -.005 (.012) -.032 (.012)*
-.053 (.011)*
-.037 (.011)*
Household size -.088 (.026)* .001 (.028) .096 (.026)*
-.062 (.025)*
Constant -.301 (.109)* -.437 (.108)*
-.631 (.109)*
.684 (.109)*
Rho eq. 2-4 .191 (.034)* Rho eq. 1-4 .121 (.035)* Rho eq. 3-4 .165 (.034)* Rho eq. 1-2 .867 (.013)* Rho eq. 2-3 .466 (.029)* Rho eq. 1-3 .471 (.031)* Likelihood ratio test of rho14=rho24=rho34=rho13=rho23=rho12=0
chi2(6) = 4753
Notes: As for Table 3. Unweighted number of observations=4369.