Transcript Slide 1

Isolation and ethnicity:
long-term illness and patterns of
participation
Lucinda Platt
University of Essex, [email protected]
Kaveri Harriss
London School of Hygiene and Tropical
Medicine
The project (1)
The research presented here is part of a project (joint with colleagues at the
University of Sheffield and the London School of Hygiene and Social Action
for Health) on long-term illness and poverty and how the relationship
between the two -- and the strategies in dealing with them -- vary by ethnic
group.
This project engages in a complex area that is receiving increasing research
and policy attention in relation to such issues as caring, extra costs of
disability, the relationship between sickness and work, the role of social
security benefits and the interface between work and benefits, and the impact
of long-term illness on other household members.
However, many gaps still remain, and some current research could benefit
from further development and this project hopes to explore a number of the
gaps and contribute to evidence base and understanding of patterning of
illness across households, specifically contributing to our understanding of
how it varies by ethnic group.
The project (2): details
Details of the research project:

Limiting illness and poverty: breaking the vicious
cycle, January 2005-June 2006

Funded by the Joseph Rowntree Foundation

Research team
 Sarah Salway, University of Sheffield (project leader)
 Punita Chowbey, University of Sheffield
 Kaveri Harriss, London School of Hygiene and
Tropical Medicine
 Lucinda Platt, University of Essex
 Elizabeth Bayliss, Social Action for Health

The project (3): contribution

Specific contribution of the project:





Exploring ethnic group differences in rates of illness,
coping with illness and in the relationship between illness and
various indicators of poverty (including worklessness)
Looking at and within households, considering the whole unit
and the interplay between different household members.
Focusing on social relationships and social participation
among those with long-term illness and those with caring
responsibilities and other members of their households
Looking at use of benefits by those with long-term illnesses
Using the livelihoods approach to examine coping strategies,
and how people use their strengths
The project (4): framework
Approach:
 Predominantly qualitative project (focused on
ethnographic work in the East End of London)
supplemented by quantitative exploration of some of
the issues (including what is being presented in this
paper), but integrated approach.
 Focusing on four main groups: Ghanaians, Punjabi
Pakistanis, Bangladeshis and White English.
Issues
 In
this part of the project, we wanted to
explore:
whether it makes sense to consider lack of social
participation as poverty
 What the relationship is between long-term illness or
caring and social participation
 What the relationship is between ethnic group and
social participation – and whether this varies by
whether the individual is long-term ill or a carer

Background (1)
 Since
Townsend’s 1979 work on Poverty in the
United Kingdom (and arguably before) it has
become standard to think about poverty in
terms of either (definitionally) or (causally –
resulting in) lack of participation in ‘normal’ life
including social life. However, measures of
social participation have only tended to be
accepted if ‘validated’ by low income.
Background (2)
 Long-term
illness is associated with higher rates
of income poverty. This is partly but not wholly
a consequence of the characteristics of those
who are more likely to be long-term ill, which
make them more vulnerable to poverty (e.g.
lower average qualifications levels). Given the
potential protective aspects of social interaction
/ social support on recovery from illness, it may
be important to consider how social
participation and illness relate and the role of
income in this.
Background (3)
 We
know that minority ethnic groups have
higher rates of morbidity, and particular types
of morbidity, than the population as a whole. In
particular, older Bangladeshi and Pakistani men
have particularly high rates of long-term illness.
It has not, to our knowledge, been explored in
any detail how different rates of within-group
illness affect the impact of that illness or its
relationship with other factors (work,
participation, benefit receipt etc).
Ethnicity and health:
poor reported health
Men
General population
Women
Source: Health Survey for
England, 1999, Department of
Health, in Focus on Ethnicity:
http://www.statistics.gov.uk/do
wnloads/theme_social/social_fo
cus_in_brief/ethnicity/ethnicity.
pdf
Irish
Indian
Pakistani
Bangladeshi
Black Caribbean
Chinese
0.0
1.0
2.0
3.0
4.0
Background (4)
 Minority
ethnic groups have higher rates of
income poverty than the population as a whole.
Those groups that have the highest rates of
income poverty are those that also have the
highest rates of long-term illness (Pakistanis and
Bangladeshis). However, Black Africans also
have high rates of income poverty but they have
low rates of long-term illness. How do we
understand the intersection between ethnicity,
poverty and long-term illness?
Low income and ethnicity
80
60
Before housing costs
Notes:1. Low income
household is defined as
having less than 60 per
cent of the median
disposable income.
After housing costs
40
20
0
White
Indian
Pakistani/
Black
Bangladeshi Caribbean
Black Non
Caribbean
Other
Source: Households
Below Average Income,
Family Resources
Survey, 2000/01,
Department for Work
and Pensions, in Focus
on Ethnicity:
http://www.statistics.gov
.uk/downloads/theme_s
ocial/social_focus_in_bri
ef/ethnicity/ethnicity.pdf
Contribution of this paper (1)
 Explores
a range of measures of social
participation – separately and together
 Explores relationship between ethnic group and
social participation controlling for long-term
illness / caring (and vice versa)
 Explores whether interactions between ethnic
group and illness/caring can be identified, i.e. if
the impact of illness / caring on social
participation varies by ethnic group
Contribution of this paper (2)
 Explores
whether income explains most / all of
ethnic group and illness effects: i.e. attempts to
decouple lack of participation from income
poverty
 Explores if analysis supports the existence of a
latent isolation propensity on the basis of the
multiple measures selected.
Data
The Home Office Citizenship Survey (or People,
Families and Community Survey), 2001
 A biennial survey, first carried out in 2001, with focus
on people’s local networks, their participation in civic
and social activities, including volunteering, their
approach to child-rearing and use of advice and
support, their experience of and attitudes to racisms
and their views about their local environment.
 The 2003 data were released in 2005, but inexplicable
results for the Bangladeshis raised doubts about the
quality of these data so using the 2001 data

Data (cont.)
 Design:
a nationally representative sample of c.
10,000 individuals accompanied by a booster
sample of 5,000 minority ethnic group
members. This facilitates analysis by ethnic
group, though for sub-groups of smaller
minority groups (e.g Black Africans), numbers
remain small.
Acknowledgement:
I am grateful to the Home Office for use of the data and to the UK Data Archive for making them available.
Neither the Home Office or the UK Data Archive, however, bear and responsibility for the analysis or
interpretation offered here.
Crown copyright material is reproduced with the permission of the Controller of HMSO and the Queen's
Printer for Scotland
Measures of (lack of)
social participation
 Infrequent
(< once a fortnight) visits to friends
/neighbours (lonbvis)
 Infrequent (< once a fortnight) receipt of visits
from friends / neighbours (lovisit)
 Infrequent (< once a fortnight) going out
socially with friends (logoout)
 Infrequent (< once a month) attendance at /
involvement with clubs (irregclb)
Results
 Overview
of rates of long-term illness and
caring
 Rates of lack of social participation
 Ordered logits on probability of isolation
controlling for other factors
 MV probit on probability of different
components of isolation controlling for other
factors
Rates of long-term illness
and caring by age band, 2001
25
Has a long-term illness
Cares for someone inside or outside the home
20
15
%
10
5
0
18-39
40-59/64
All adults
Age
Source: Home Office Citizenship Survey 2001, authors’ analysis
Long term illness and
caring by age band and sex, 2001
25
18-39
40-59/64
20
All adults
15
%
10
5
0
men
women
Long-term illness
men
women
Carer
Source: Home Office Citizenship Survey 2001, authors’ analysis
Long term illness and caring by
ethnic group, 2001
18
16
Long-term ill
Carer
14
12
10
%
8
6
4
2
0
white British
Pakistani
Bangladeshi
Black African
Ethnic group
Source: Home Office Citizenship Survey 2001, authors’ analysis
Long-term illness among men and
women by age band and ethnic group, 2001
70
60
50
white British
Pakistani
Bangladeshi
Black African
%
40
30
20
10
0
All ages
Aged 18-39
Both sexes
Age 40-59/64
All ages
Aged 18-39
Men
Age 40-59/64
All ages
Aged 18-39
Age 40-59/64
Women
Source: Home Office Citizenship Survey 2001, authors’ analysis
% with limited social
participation, by ethnic group
Infrequent
visits
Infrequent
visiting
Infrequent
going out
Low
contact
with clubs
White British
36
39
32
46
Pakistani
30
36
49
55
Bangladeshi
26
30
49
58
Black African
42
52
50
41
Source: Home Office Citizenship Survey 2001, authors’ analysis
Lack of social participation
by illness / caring and age band 2001
Infrequent visitors
Infrequent visiting
Infrequent going out
Low contact with clubs
70
60
50
40
%
30
20
10
0
All ages
18-39
Ill
40-59/64
All ages
18-39
40-59/64
All ages
Caring
Source: Home Office Citizenship Survey 2001, authors’ analysis
18-39
All
40-59/64
Lack of participation by
ethnic group, 2001
Infrequent visitors
80
Infrequent visiting
70
Infrequent going out
Low contact with clubs
60
50
% 40
30
20
10
0
white British
Pakistani
Bangladeshi
Black
African
white British
Pakistani
Ill
Source: Home Office Citizenship Survey 2001, authors’ analysis
Bangladeshi
Caring
Black
African
Isolation (lack of social
participation)
No isolation
28.1
Isolated on 1 element
25.9
Isolated on 2 elements
19.3
Isolated on 3 elements
16.8
Isolated on all 4 elements
Source: Home Office Citizenship Survey 2001, authors’ analysis
8.9
Isolation by ethnic group
100%
90%
80%
70%
60%
50%
40%
30%
20%
Isolated on all 4 measures
Isolated on 3 measures
Isolated on 2 measures
Isolated on 1 measure
Not isolated
10%
0%
white British
Pakistani
Bangladeshi
Source: Home Office Citizenship Survey 2001, authors’ analysis
Black African
Modelling isolation
 Ordered
logits controlling for
Model 1: Age group (18-44/ 45-59/64); sex;
presence of a child under 5; and illness or caring and
ethnic group
 Model 2: As above but with work history (in work,
not in work but worked in past, never worked)
included
 Model 3: As above with income bands and
household size included
 Also models were rerun controlling for qualifications
and type of area additionally

Results for illness and ethnic group
Model 1
Long-term ill
Ethnic group (base
is white British)
Indian
Pakistani
Bangladeshi
Black Caribbean
Black African
Chinese
Mixed and other
groups
Model 2 (including
work history)
.233 (.061)*** .234 (.066)***
Model 3 (adding income
band and household size)
.077 (.078)
.125 (.076)
.366 (.082)***
.330 (.111)**
.491 (.086)***
.585 (.099)
.596 (.167)***
.081 (.098)
.079 (.113)
-.098 (.133)
.042 (.166)
.423 (.120)***
.323 (.123)**
.362 (.229
.098 (.125)
.103 (.078)
.374 (.092)***
.333 (.119)**
.507 (.088)***
.579 (.105)***
.632 (.172)***
.079 (.099)
Results for caring and ethnic group
Model 1
Model 2
Model 3
.269 (.083)***
.258 (.084)**
.265 (.094)**
.116 (.076)
.091 (.078)
.041 (.104)
.361 (.082)***
.359 (.091)***
.029 (.121)
.331 (.111)**
.321 (.120)**
.007 (.158)
Black Caribbean
.500 (.087)***
.514 (.089)***
.419 (.110)***
Black African
.592 (.010)***
.576 (.105)***
.376 (.113)***
Chinese
.588 (.168)***
.617 (.174)***
.363 (.220)
.088 (.098)
.081 (.099)
.091 (.112)
Caring
Ethnic group (base is
white British)
Indian
Pakistani
Bangladeshi
Mixed and other
groups
The effect of income (1)
 Including
income bands (even though a rather
crude measure resulted in illness ceasing to have
a significant association with isolation but did
not make a difference to the size or significance
of the coefficient for caring. Thus while the
isolation associated with illness can be attributed
to reduced resources, that for caring cannot.
Interestingly, work history was only significantly
related to isolation once income was included.
The effect of income (2)
 Including
income rendered non-significant the
ethnic group effects for Pakistanis, Bangladeshis
and Chinese. This suggests that it is lack of
income that affects ability to participate for
these groups, but it is interesting that this is the
case when illness / caring are already being
controlled for. Ill Bangladeshis, Pakistanis and
Chinese would thus appear to have fewer
resources available for social participation than
their long-term ill white comparators.
Adding in qualifications and area type
Living in a rural as opposed to an urban area decreased
the probability of isolation, but did not have much
effect on the other coefficients
 Qualifications (at all levels) reduced the probability of
isolation compared to having none and also when
qualifications were controlled for illness had a weaker
association with isolation. Qualifications retained their
significant negative association with isolation even
when income was controlled for, indicating that
education makes a difference to ability to participate
regardless of income.

Interactions between
ethnic group and illness / caring
 We failed to identify any interactions between ethnicity
and long-term illness in any of our models.
 However, when exploring caring interacted with
ethnicity, we found that for Caribbeans and Chinese,
the interaction between caring and ethnic group
produced a significant negative effect that outweighed
or at least equalled the positive main effects of caring
and ethnicity. That is, while being Caribbean or Chinese
made isolation more likely among those not caring, and
caring made isolation more likely for the white group,
Caribbean and Chinese carers were no more likely than
white non-carers to be isolated.
Modelling components
of isolation simultaneously
 Multivariate
probits were run to test
simultaneously the effect of the explanatory
variables on the different components of
‘isolation’ – the lack of participation variables.
 Correlations between the error terms across
these equations lend support to the view that
there is a latent continuum of isolation (or
propensity to be isolated) captured by these
repeated observed measures (cf. measurement
of deprivation).
Predicted probabilities from
the multivariate probits

Predicted probabilities for different sets of characteristics were
estimated on the basis of the regression results for a models with
and without income. The ‘individuals’ for whom probability of
single measures of isolation, all measures together and no
isolation were estimated were:






1 (typical values) younger white man with no children under 5 not long-term ill
and in employment
2 (contrast case) older Bangladeshi woman, long-term ill with child under 5, never
worked and no qualifications
3 Younger Caribbean woman, not long-term ill, child under 5, in work, level 3
qualifications
4 Older Pakistani man, long-term ill, not currently in work, no child under 5, level
1 qualifications
5 Younger Caribbean man, no child under 5 not long-term ill, in employment,
level 4 qualifications
6 Older white woman, long-term ill not currently in work, level 1 qualifications
Predicted probabilities: no or
max isolation (without income and area)
45
40
35
No isolation
Maximum isolation
30
25
%
20
15
10
5
0
person 1 person 2 person 3 person 4 person 5 person 6 average
for
sample
Predicted probabilities: no or max isolation
(with average income and in urban area)
40
35
30
No isolation
Maximum isolation
25
% 20
15
10
5
0
person 1 person 2 person 3 person 4 person 5 person 6 average
for
sample
Predicted probabilities: single
indicators (without income and area)
average for sample
going out
friends round
visits friends
clubs
person 6
person 5
person 4
person 3
person 2
person 1
0
10
20
30
40
50
%
60
70
80
90
Conclusions (1)




There are ethnic differences in levels and types of social
participation
Most minority ethnic groups are more likely to be isolated than
their white counterparts of the same age group sex and family
status. However, for Pakistanis, Bangladeshis and Chinese this is
more to do with income than ethnicity per se.
Similarly the association of long-term illness with isolation seems
to be to do with economic resources rather than illness per se.
But this is not the case with caring, which is isolating at each
income level.
For Black Caribbeans and Black Africans, higher probabilities of
isolation are not affected by income levels.
Conclusions (2)
There is little evidence that the relationship between
illness and isolation varies by ethnic group. However,
there is evidence that the relationship between caring
and ethnicity does so for some ethnic groups. For
Chinese and Black Caribbeans, those caring do not
appear to experience the greater risks of isolation
generally associated with caring or their ethnicity.
 Unobservable characteristics are correlated across
equations for components of isolation, suggesting a
latent ‘propensity to isolation’ captured by the repeated
measures (assuming that a sufficient range of observed
characteristics have been incorporated in the equations)

Next steps
 To
conduct analyses separately for men and
women (sample sizes permitting) rather than
simply controlling for sex
 To formulate the relationship between income,
illness and participation more clearly
 To examine whether any other measures of or
refinements to measures of participation might
be relevant.
 To incorporate income measure into mvprobits
 Other?