Agency and structure: the impact of ethnic identity and racism on the health of ethnic minority people

Authors


Abstract

Abstract To understand ethnic inequalities in health, we must take account of the relationship between ethnic minority status, structural disadvantage and agency. So far, the direct effects of racial oppression on health, and the role of ethnicity as identity, which is in part a product of agency, have been ignored. We set out to redress this balance using data from the Fourth National Survey of Ethnic Minorities. Factor analysis suggested that dimensions of ethnic identity were consistent across the various ethnic minority groups. Initially some of these dimensions of ethnic identity appeared to be related to health, but in a multivariate model the factor relating to a racialised identity was the only one that exhibited any relationship with health. These findings suggest that ethnic identity is not related to health. Rather, the multivariate analyses presented here showed strong independent relationships between health and experiences of racism, perceived racial discrimination and class.

Introduction

Stripped of its dynamic social, economic, gender and historical context, culture becomes a rigid and constraining concept which is seen somehow to mechanistically determine peoples’ behaviours and actions rather than providing a flexible resource for living, for according meaning to what one feels, experiences and acts to change (Ahmad 1996: 190).

Despite an increasing interest in ethnic inequalities in health in Britain, what constitutes being of an ethnic group, and how this is related to health, has remained largely unexplored (Nazroo 1998a, Smaje 1996). This is partly due to an assumption in epidemiological research that ethnic differentials in health are a consequence of innate characteristics related to ‘ethnic’ or ‘racial’ difference. This encourages the use of inflexible assessments of ethnicity that treat ethnic categories as reflecting undifferentiated groups. Research findings have often focused on, or at least implied, universal genetic and cultural explanations for the relationship between ethnic status and health (Bhopal 1997, Sheldon and Parker 1992).

More recently, research has begun to explore the possible contribution that structural factors, particularly differences in socioeconomic position, might make to ethnic inequalities in health (Harding and Maxwell 1997, Nazroo 1997a). Such studies have found a clear class effect in the relationship between ethnicity and health, suggesting that the processes through which ethnic minority status leads to class disadvantage are central to understanding ethnic inequalities in health. However there may be other elements of ethnicity as ‘structure’ that might also be important, and that have so far received little attention. In particular there is some evidence that experienced or perceived racism might be related to poorer health (Krieger and Sidney 1996, Krieger et al. 1993, Benzeval et al. 1992).

Of course, ethnicity represents more than a position in a social structure, it also represents an identity that is at least partly a consequence of agency. The potential importance of ethnicity as identity to ethnic differences in health has been much discussed (Smaje 2000 and 1996, Nazroo 1998a, Ahmad 1996), but has not been empirically investigated. This paper will attempt to redress this imbalance through an exploration of ethnicity as identity and its relationship with socioeconomic position, racism and health.

Structural components of ethnic differences in health

Racism may well be the key route through which structure influences ethnic inequalities in health. Racism can affect health in a number of ways (Krieger 2000). It can affect health directly, through the negative physical and psychological consequences of interpersonal racist victimisation and racial discrimination. Or indirectly, in the way that institutional racism leads to the identification of ethnic minority groups, their reification as biologically and culturally different, and the consequent exclusion and social and economic disadvantage that ethnic minority people experience (Miles 1989).

Recent studies have clearly shown both that socioeconomic gradients in mortality and morbidity exist for different ethnic groups (Harding and Maxwell 1997, Nazroo 1997a, Lillie-Blanton and LaVeist 1996, Jones-Webb and Snowden 1993), and the complexity of making adjustments for socioeconomic position when attempting to draw comparisons across different ethnic groups (Kaufman et al. 1998, Nazroo 1998a, Nazroo 1998b, Nazroo 1997a, Krieger et al. 1993). These studies suggest that conventional measures of socioeconomic position may actually serve to conceal the socioeconomic disadvantage experienced by ethnic minority groups, rather than expose it. Analysis previously undertaken on the Fourth National Survey (the dataset used here) suggests that the internal heterogeneity of traditional class groupings, based here on the Registrar General’s measure of occupational class, masks the concentration of ethnic minority people in lower income occupations, poorer quality housing and longer periods of unemployment than white people in the same class (Nazroo 1997a).

In addition, the more immediate health consequences of racism have been largely ignored (Gillborn 1995, Krieger et al. 1993). Racism has been shown to have direct physical consequences, and such oppression can become internalised, damaging self-esteem, and potentially compromising available social support, which will also have consequences for health (Krieger and Sidney 1996, Krieger et al. 1993, Benzeval et al. 1992). However studies exploring the health effects of racism have so far produced inconsistent results. Part of this inconsistency stems from the difficulties associated with assessing the extent to which individuals experience racism (Krieger 2000). In addition, the way in which an individual reacts to racism has been shown to mediate its health effects. In a study exploring rates of hypertension among black and white Americans, Krieger and Sidney (1996) found that Black Americans who said they would report and challenge racism had a lower blood pressure than those who said they would tolerate but not report it. They suggest that this negative health effect is a consequence of internalised anger, which was more likely among those that experienced but did not report racial harassment.

However, while this evidence suggests that the exploration of structural factors is central to the understanding of ethnic inequalities in health, it is important to consider how they might be integrated with an understanding of agency. Here the conceptualisation of ethnicity as identity is of particular use in understanding cultural lifestyles, as well as responses to socioeconomic disadvantage and racial discrimination.

The role of agency: ethnic identity

Most commentators on the situation of ethnic minority people discuss a notion of ethnicity that reflects self-identification with cultural traditions that provide personal meaning and boundaries between groups (see, for example, the collected works in Barot 1996). Importantly, this self- and group-identification, rather than being something innate and fixed, is something that is formed and perpetually transformed in relation to representation to, and reaction by, the external audience (see the opening quote from Ahmad 1993). So, work on identity has suggested that an individual has a range of different ‘identities’, reflecting our age, gender, social class, ethnicity, etc, and from which it seems possible to choose – what Hall calls the ‘cultural supermarket’ effect (1992: 303) – and which locate us in our social context (Smaje 1996, Deaux et al. 1995).

Identity, however, is not, of course, entirely self-constructed. Jenkins (1994) defines two aspects of identity: the ‘nominal’ (the name) and the ‘virtual’ (the experience). Defining who is and what it is to be a member of a particular social group is seen to involve the consolidation of internal and external processes: the external imposition of a characterisation, for example, will affect the social experience of living with that identity and the self-image of those so defined. Further, the effect of representation is such that what it is to have a particular identity will also vary according to the external audience. It is argued, therefore, that defining who we are, both by name and in experience, is dynamic and relatively ambiguous and, while being a largely internal process, will be heavily influenced by wider society. In this way, the experience of racism, for example, can be seen actually to structure an individual’s own identity, as well as affecting the way in which someone with that identity interacts with others.

So individual decisions about who we are and our lifestyle choices, while appearing to be unbounded and, therefore, solely a consequence of agency, are, in reality, made within social constraints, what Bourdieu terms ‘habitus’(1977). The theory of habitus explores the way in which symbolic representations influence behaviour through:

a whole body of wisdom, sayings, commonplaces, ethical precepts (‘that’s not for the likes of us’) and, at a deeper level, the unconscious principles of the ethos which … determines ‘reasonable’ and ‘unreasonable’ conduct for every agent (1977: 77).

Thus Bourdieu argues that while social practice has some purpose and practical intent for the individual, these goals are located within an individual’s own experience of reality, which is related to who and what they are and is, therefore, at least partially, externally defined.

Bourdieu argues that the only means of expanding this sphere of ‘reasonable’ behaviour is through increasing the lifestyle choices available, via forms of ‘capital’, which are also delimited by social position. So, attempts by social groups to define and appropriate their own lifestyle will also be restrained and influenced by the social structure and wider society (see also Smaje 1996). For example, Hall (1992) discusses the way in which globalisation and sustained migration have lead to the pluralisation of national cultures and identities. He suggests that identities may react to this by becoming more traditional, ‘attempting to restore their former purity and recover the unities and certainties which are felt as being lost’ (Hall 1992: 309), or they may adapt. He explores the notion of a translation of culture, which occurs where people are obliged to come to terms with new cultures surrounding them, but also wish to retain strong links with their places of origin and associated traditions, and so form a new ‘hybrid’ identity with aspects of each. And, he argues, these changes are not restricted to minority groups. The strengthening of local identities in the UK, for example, coincide with ‘a revamped Englishness, an aggressive little Englandism, and a retreat to ethnic absolutism’ (Hall 1992: 308).

So, while aspects of ethnic identity may be internally defined, and therefore the consequence of agency, the scope of those choices will be restricted and affected by social structure and any consideration of ethnic identity needs to take account of this. There is also a need to consider ethnicity as a ‘hybrid’ identity (Modood 1998, Hall 1992), one that is influenced by internal and external factors, locally and globally – an identity that is not just given, but which is continually changed across contexts and over time. Importantly, this notion of ethnicity as identity emphasises the political process of ethnic affiliation at the expense of behavioural markers of ethnicity (including, of course, those behaviours relating to health) (Modood 1998). And, of course, such ethnic affiliation could provide important symbolic and material resources that are health promoting. While socioeconomic disadvantage might contribute to ethnic inequalities in health, there remains a cultural component to ethnicity which could also make a major contribution to differences in health.

The demand, then, is to build on work that has investigated the impact of structural factors on the health of ethnic minority people to explore the role of agency. In this paper we do this by including an assessment of the importance of ethnicity as identity. This is not straightforward for two reasons. First, the contextual nature of ethnic identity makes it hard to operationalise in a quantitative study. Second, the relationship between culture/identity and health is not likely to be straightforward. For example, while some customs may lead to improved social support (Kelleher 1996), sociocommunal engagement and psychological wellbeing (Halpern and Nazroo 1999), in some circumstances they may also lead to negative discrimination and isolation. Here we will describe one approach to exploring the relationship between ethnic identity, structure and health: an approach that uses quantitative techniques, but which also recognises the multi-dimensional nature of ethnic identity.

Methods

Fourth National Survey of Ethnic Minorities

The findings presented in this paper are based on secondary analysis of the Fourth National Survey of Ethnic Minorities (FNS). The FNS was a representative survey of ethnic minority and white people living in England and Wales that was undertaken in 1993 and 1994 by the Policy Studies Institute and Social and Community Planning Research (now the National Centre for Social Research). A sample of 5,196 people of Caribbean and Asian origin and 2,867 white people underwent a structured face-to-face interview conducted by an ethnically matched interviewer in the language of the respondent’s choice. Respondents were allocated into an ethnic group on the basis of answers to a question on their family origins. In addition to physical (Nazroo 1997a) and mental (Nazroo 1997b) health, the questionnaire covered a comprehensive range of information on both ethnicity and other aspects of the lives of ethnic minority people, including demographic and socioeconomic factors. Included in this was a section on ethnic identity. However, only half of the ethnic minority sample and none of the white sample were asked the questions on ethnic identity. (For further details of the methods and findings, see Modood et al. 1997.)

Across the health outcomes and in terms of sociodemographic profiles and responses to individual identity measures, analysis of the FNS data showed great similarity between Pakistani and Bangladeshi people, on the one hand, and Indian and African Asian people on the other (Modood et al. 1997, Nazroo 1997a). To overcome the problem of small numbers, these groups were combined for this analysis. The sample for this analysis was: 591 Caribbean people; 903 Pakistani or Bangladeshi people; and 1,013 Indian or African Asian people. The Chinese sample (N = 107) was too small to include.

Ethnic identity, occupational class and health

To determine underlying dimensions of ethnicity that might contribute to a sense of identity, a factor analysis (Kim and Mueller 1979) of these variables was conducted. This technique is used to identify factors that can be used to represent correlations among sets of inter-related variables. We first conducted this for each ethnic group separately, then for all of the ethnic minority groups combined. The principal components method of factor extraction was used. This produces factors in sequence according to the amount of the total sample variance they account for. The total variance explained by each factor is called the eigenvalue. This analysis reports only factors with an eigenvalue of 1 or over (Kim and Mueller 1979). Principal components analysis was followed by oblique rotation to allow for correlation between the different factors identified. Cronbach’s alpha reliability coefficients give results from tests of correlation between the variables clustering under the different factors. Individual respondents were allocated a factor score for each of the factors identified, which summarised their responses to all of the questions included in the analysis.

The factor analysis concentrated on questionnaire items relating to descriptions of ancestry and ethnic affiliation, lifestyle, experience and perceptions of racism, and social and community involvement. The factor analysis included questions on: aspects that would be perceived to be important in a description of the respondent given by themselves and a white person; participating in customs and behaviours which may be seen as traditional to an ethnic group; membership of ethnically-specific organisations; experience or recognition of racist elements in British society; and the extent to which the respondents saw themselves as ‘British’ and as a member of their ethnic group. The details of these questions will be shown in the results section.

To explore the relationship between ethnic identity, class and health, a series of regression tests were undertaken using self-reported fair or poor health as the outcome variable, and the different dimensions of ethnic identity, occupational class, age and gender as independent variables. Class was assigned using the head of the household’s occupation (coded as non-manual or manual), with a third group of those with no full-time worker in the household. Where it was not clear which household member was the head of household (e.g. where there was more than one working adult), class was allocated on the basis of gender (with men’s occupations being used in preference to women’s) and age (e.g. a father’s occupation being used in preference to a son’s, if the father was below retirement age).

To explore associations between ethnic identity and specific health conditions, the analysis was repeated for probable heart disease (diagnosed angina, diagnosed heart attack or severe chest pain), diagnosed diabetes and diagnosed hypertension.

Racial discrimination and health

To investigate further the relationship between structural components of ethnicity and health, we performed a logistic regression test to explore the relationship between experience of racial harassment, perceptions of the extent of racial discrimination in Britain and reported fair or poor health. Variables included were: occupational class, gender, age and whether the respondent had been the victim of a racially motivated attack in the last 12 months (coded ‘none’, ‘verbal attack’ and ‘physical attack on one’s property or person’). This analysis was repeated adding in a question asking whether the respondent felt British employers discriminated against people on the basis of race or religion (coded ‘no’, ‘a few’, ‘some’ and ‘most’), which was asked of only half the sample.

Results

Dimensions of ethnic identity

Analysis retaining all factors with an eigenvalue of one or over produced a five factor model. Half (49.1 per cent) of the total variance in the model was explained by these five factors. The working titles given to the five factors were: nationality important for self description; ‘ethnicity/race’ important for self-description; traditional; community participation; and member of a racialised group.

The questions loading on factor 1 (nationality important for self description) were:

  • • If you were describing yourself on the phone to a new acquaintance of your own sex from a country you have never been to,
    • • would your nationality tell them something important about you?
    • • would the country your family came from tell them something important about you?
  • • If a white person who knew and liked you was describing you to another white person, would they think it important to mention:
    • • your nationality?
    • • the country your family came from?

(Cronbach’s alpha = 0.61)

The questions loading on factor 2 (‘ethnicity/race’ important for self description) were:

  • • If you were describing yourself on the phone to a new acquaintance of your own sex from a country you have never been to,
    • • would your skin colour tell them something important about you?
    • • would the fact that you are Asian/Caribbean tell them something important about you?
  • • If a white person who knew and liked you was describing you to another white person, would they think it important to mention:
    • • your skin colour?
    • • that you are Asian/Caribbean?

(Cronbach’s alpha = 0.65)

The questions loading on factor 3 (traditional) were:

  • • How often do you wear Asian clothes/something that is meant to show a connection with the Caribbean or Africa? (Responses: ‘Never’; ‘At social events’; ‘At home’; ‘At work, or while shopping’; ‘All the time’)
  • • ‘Who do you speak to in a language other than English?’ (‘No-one’; ‘Own-age relatives’; ‘Younger relatives’; ‘Older relatives’; ‘Friends outside work’; ‘Work friends’)
  • • ‘Would you personally mind if a close relative were to marry a white person?’ (‘I wouldn’t mind’; ‘I would mind a little’; ‘I would very much mind’)
  • • Do you strongly agree, agree, neither agree or disagree, disagree or strongly disagree with these statements?:
    • • ‘In many ways I think of myself as being British’
    • • ‘In many ways I think of myself as being Asian/Caribbean’

(Cronbach’s alpha = 0.56)

The questions loading on factor 4 (community participation) were:

  • • ‘Does your voluntary work bring you mainly into contact with people of your ethnic origin, mainly white people or about equally with both?’(‘Don’t volunteer’; ‘Mainly White’; ‘Both’; ‘Mainly people from my own ethnic group’)
  • • ‘Do your activities with this organisation bring you mainly into contact with people of your ethnic origin, mainly white people or about equally with both?’(‘Am not a member of an organisation’; ‘Mainly White’; ‘Both’; ‘Mainly people from my own ethnic group’)

(Cronbach’s alpha = 0.40)

The questions loading on factor 5 (member of a racialised group) were:

  • • ‘In the last twelve months, have you been a victim of a racially motivated attack (verbal or physical abuse to the person or property)?’
  • • ‘Have you ever been treated unfairly at work or been refused a job on the basis of race, colour or your religious or cultural background?’
  • • ‘How many of the employers in Britain do you think would refuse a job to a person because of their race, colour, religion or cultural background?’(‘None’; ‘A few’; ‘About half’; ‘Most’)

(Cronbach’s alpha = 0.40)

When the factor analysis was conducted for each ethnic group separately, the results were very similar for each group.

Analysis of variance in the mean factor scores for each ethnic group showed no statistically significant difference between the ethnic group means for factors 1 (nationality important for self-description, F = 0.465, p = 0.6) or 4 (community participation, F = 0.358, p = 0.7). There was, however, a statistically significant variation between the ethnic group means for factors 2 (‘ethnicity/race’ important for self description, F = 176.2, p < 0.001, with a lower mean score for the Caribbean group, compared with the two South Asian groups), 3 (traditional, F = 406.2, p < 0.001, again with a lower mean score for the Caribbean group) and 5 (member of a racialised group, F = 132.4, p < 0.001, with a lower mean score for the Pakistani or Bangladeshi group and a higher mean score for the Caribbean group). These findings would suggest that the dimensions of ethnic identity are consistent across different ethnic minority groups, but the nature and significance of particular dimensions vary markedly. (See Nazroo and Karlsen, forthcoming, for a more detailed discussion of this and for a discussion of the relationship between these dimensions of ethnic identity and socio-demographic indicators.)

Ethnic identity and health

Reporting fair or poor health:Tables 1a, 1b and 1c show odds ratios of the relationship between self-reported fair or poor health, ethnic identity (using the dimensions of ethnic identity identified using the factor analysis) and occupational class for each ethnic group. The model was constructed in three stages: first with only self-reported health and the dimensions of identity; then with age and gender; and finally with occupational class.

Table 1a.  Ethnic identity and risk of fair or poor health – Indian or African Asian group : Odds ratios (95% confidence intervals)
 Indian or African Asian (n = 950)
 Model 1Model 2Model 3
Nationality1.02 (0.86–1.21)1.09 (0.91–1.31)1.09 (0.91–1.31)
Ethnicity1.30 (1.07–1.56)1.05 (0.86–1.29)1.08 (0.88–1.33)
Traditional1.62 (1.32–1.99)1.24 (1.01–1.53)1.20 (0.97–1.50)
Community1.10 (0.93–1.29)1.07 (0.90–1.28)1.09 (0.91–1.31)
Racialisation1.14 (0.96–1.37)1.33 (1.10–1.61)1.38 (1.14–1.67)
Gender
Male 1.001.00
Female 1.58 (1.06–2.34)1.60 (1.07–2.40)
Age 1.07 (0.99–1.15)1.08 (1.00–1.16)
Class
Non-manual  1.00
Manual  1.63 (1.04–2.56)
No full-time worker in the household  2.49 (1.56–3.97)
Chi-sq (df)30.0 (5)102.7 (8)128.0 (10)
p-value<0.001<0.001<0.001
Table 1b.  Ethnic identity and risk of fair or poor health – Pakistani or Bangladeshi group : Odds ratios (95% confidence intervals)
 Pakistani or Bangladeshi (n = 835)
 Model 1Model 2Model 3
Nationality1.02 (0.85–1.21)1.10 (0.93–1.30)1.09 (0.92–1.29)
Ethnicity1.25 (1.00–1.56)1.04 (0.81–1.33)1.03 (0.80–1.32)
Traditional1.76 (1.38–2.25)1.39 (1.04–1.84)1.34 (1.00–1.79)
Community0.94 (0.76–1.17)1.06 (0.85–1.33)1.13 (0.88–1.43)
Racialisation1.07 (0.85–1.33)1.14 (0.88–1.47)1.18 (0.91–1.52)
Gender
Male 1.001.00
Female 1.51 (0.97–2.34)1.57 (1.00–2.47)
Age 1.09 (1.00–1.19)1.12 (1.03–1.22)
Class
Non-manual  1.00
Manual  1.45 (0.74–2.84)
No full-time worker in the household  2.48 (1.35–4.58)
Chi-sq (df)26.0 (5)87.8 (8)91.8 (10)
p-value<0.001<0.001<0.001
Table 1c.  Ethnic identity and risk of fair or poor health – Caribbean group : Odds ratios (95% confidence intervals)
 Caribbean (n = 555)
 Model 1Model 2Model 3
Nationality0.77 (0.62–0.97)0.95 (0.74–1.23)0.94 (0.73–1.22)
Ethnicity1.13 (0.89–1.42)1.16 (0.90–1.49)1.16 (0.90–1.49)
Traditional1.05 (0.81–1.37)1.08 (0.82–1.43)1.06 (0.81–1.40)
Community0.91 (0.74–1.12)1.09 (0.86–1.38)1.08 (0.85–1.37)
Racialisation0.90 (0.72–1.13)0.94 (0.73–1.20)0.94 (0.73–1.20)
Gender
Male 1.001.00
Female 1.81 (1.13–2.92)1.77 (1.09–2.86)
Age 0.97 (0.87–1.08)0.99 (0.89–1.10)
Class
Non-manual  1.00
Manual  0.94 (0.52–1.70)
No full-time worker in the household  1.40 (0.80–2.45)
Chi-sq (df)5.8 (5)56.0 (8)57.3 (10)
p-value0.3<0.001<0.001

Before the inclusion of age and gender, factors 1, 2 and 3 had statistically significant associations with self-reported health for one or more ethnic groups. A high score on factor 1 (importance of nationality in a self-description) was associated with a statistically significant reduction in fair or poor health among Caribbean people, but had no relationship with health for the other two groups. A high score on factor 2 (importance of ‘ethnicity/race’ in a self-description) was associated with increased reports of fair or poor health that approached significance for the Pakistani or Bangladeshi group, and was significant for the Indian or African Asian group. A high factor 3 score (traditional) was associated with a 60 to 80 per cent greater likelihood of reporting fair or poor health for both of the South Asian groups.

After including age and gender in the models, the effects of the ethnic identity factors changed. Both the reduced risk for Caribbean people associated with higher scores on factor 1 and the increased risk for South Asian people associated with higher scores on factor 2 disappeared. The effect of factor 3 for the South Asian groups remained statistically significant, but the size of the association declined substantially. Entering age and gender into the models separately suggested that it was age that accounted for the reduction in identity effects (not shown in the tables). Interestingly, after including age in the models, the association between self-reported fair or poor health and factor 5 (racialisation) increased and became statistically significant for the Indian or African Asian group.

When occupational class was entered into the models, it appeared to have an effect for each ethnic group. For the two South Asian groups, the class effect was significant and sizeable; for example, South Asian people living in households with no full-time worker had a two-and-a-half times greater chance of reporting fair or poor health, compared with South Asian people living in non-manual-headed households. There was also an effect for the Caribbean group, but this was smaller and not statistically significant. Although the inclusion of occupational class in the model did not greatly reduce the odds ratios for the identity factors, in the full model none of the identity factors made a statistically significant contribution, with the exception of factor 5 for the Indian or African Asian group.

Specific health outcomes:Table 2a shows associations between the five dimensions of ethnic identity, age, gender, occupational class and heart disease, for each ethnic group. Factor 1 was the only ethnic identity factor which exhibited any statistically significant effect, with a 60 per cent increase in risk of heart disease among the Pakistani or Bangladeshi group. Occupational class showed a statistically significant association for both of the South Asian groups, with those from households with no full-time worker at between three and seven times greater risk of heart disease than those from households headed by a non-manual worker.

Table 2a.  Ethnic identity and risk of heart disease* : Odds ratios (95% confidence intervals)
 Indian or African Asian (n = 388)Pakistani or Bangladeshi (n = 278)Caribbean (n = 237)
  • * 

    diagnosed angina, diagnosed heart attack or severe chest pain. Model also controlled for age and gender.

Nationality1.15 (0.79–1.66)1.62 (1.13–2.30)1.13 (0.69–1.86)
Ethnicity0.74 (0.51–1.09)0.75 (0.50–1.13)1.12 (0.76–1.65)
Traditional1.01 (0.68–1.50)1.05 (0.64–1.71)0.87 (0.47–1.61)
Community1.25 (0.87–1.79)0.82 (0.48–1.42)1.29 (0.84–1.99)
Racialisation1.32 (0.82–2.14)0.84 (0.54–1.31)0.90 (0.53–1.54)
Class
Non-manual1.001.001.00
Manual1.47 (0.48–4.46)4.98 (1.16–21.32)0.92 (0.25–3.36)
No full-time worker in the household3.06 (1.30–7.22)6.68 (1.74–25.66)0.98 (0.27–3.47)

None of the dimensions of ethnic identity have a statistically significant association with diagnosed diabetes (Table 2b). Although the odds ratios are not statistically significant, there is a suggestion of a class effect for the two South Asian groups.

Table 2b.  Ethnic identity and risk of diabetes diagnosis* : Odds ratios (95% confidence intervals)
 Indian or African Asian (n = 950)Pakistani or Bangladeshi (n = 834)Caribbean (n = 553)
  • * 

    also controlled for age and gender.

Nationality0.94 (0.66–1.33)1.03 (0.72–1.48)0.98 (0.60–1.61)
Ethnicity1.07 (0.69–1.64)1.32 (0.84–2.08)1.09 (0.72–1.65)
Traditional0.84 (0.59–1.21)1.21 (0.79–1.85)0.86 (0.35–2.09)
Community1.19 (0.90–1.57)1.34 (0.96–1.86)1.03 (0.73–1.45)
Racialisation0.99 (0.66–1.48)0.99 (0.64–1.54)0.85 (0.53–1.35)
Class
Non-manual1.001.001.00
Manual1.42 (0.55–3.64)1.95 (0.50–7.67)0.53 (0.15–1.93)
No full-time worker in the household1.65 (0.68–4.01)2.02 (0.56–7.26)0.95 (0.27–3.33)

Factor 5 (member of a racialised group) was significantly associated with diagnosed hypertension for the Indian or African Asian group (Table 2c). Among the Pakistani or Bangladeshi group, living in households headed by someone in a manual occupation was significantly associated with a five times greater risk of hypertension, and those from households with no full-time worker had an almost seven times greater risk than those in non-manual occupations. There was also a suggestion of a class effect for this outcome for the Indian or African Asian group.

Table 2c.  Ethnic identity and risk of hypertension diagnosis* : Odds ratios (95% confidence intervals)
 Indian or African Asian (n = 946)Pakistani or Bangladeshi (n = 834)Caribbean (n = 551)
  • * 

    also controlled for age and gender.

Nationality1.06 (0.76–1.47)0.74 (0.55–1.00)1.00 (0.74–1.34)
Ethnicity1.04 (0.76–1.42)1.28 (0.92–1.77)0.85 (0.62–1.16)
Traditional0.80 (0.58–1.10)1.15 (0.77–1.73)0.78 (0.54–1.13)
Community1.13 (0.86–1.49)0.96 (0.62–1.48)1.06 (0.82–1.36)
Racialisation1.56 (1.14–2.13)1.16 (0.80–1.66)1.07 (0.79–1.44)
Class
Non-manual1.001.001.00
Manual2.07 (0.95–4.52)5.40 (1.55–18.82)0.95 (0.44–2.03)
No full-time worker in the household1.30 (0.60–2.85)6.78 (2.11–21.79)1.89 (0.90–3.97)

The role of racism in the reporting of fair or poor health

To explore further the health effects of racialisation that were indicated by the findings for factor 5, we undertook a regression analysis to explore the relationship between self-reported fair or poor health, occupational class and perceived or experienced racism (Table 3). Initially models were constructed for each ethnic group separately, but the similarity of findings across the three ethnic minority groups allows us to present data for all ethnic groups combined. There was a statistically significant relationship between occupational class and self-reported fair or poor health: with those from households headed by a manual worker three-fifths, and those from households with no full-time worker two-fifths more likely to report fair or poor health than those from households headed by someone in a non-manual occupation.

Table 3.  Racial discrimination, occupational class and risk of fair or poor health (all ethnic minority groups combined)* : Odds ratios (95% confidence intervals)
 All ethnic minority groups
 Harassment (n = 4106)Harassment and perceived discrimination (n = 2067)
  • * 

    also controlled for age and gender.

Racially motivated attack
No attack1.001.00
Verbal1.56 (1.17–2.08)1.60 (1.09–2.35)
Physical/property2.27 (1.38–3.73)2.23 (1.14–4.37)
British employers are racist
None 1.00
A few 1.15 (0.84–1.58)
Some 1.52 (1.07–2.15)
Most 1.66 (1.17–2.37)
Occupational class
Non-manual1.001.00
Manual1.65 (1.38–1.99)1.58 (1.22–2.05)
Unemployed1.41 (1.02–2.00)1.39 (0.85–2.25)
Chi-sq (df)300.6 (7)176.2 (10)
p-value<0.001<0.001

There was a statistically significant association between reported fair or poor health and experience of interpersonal racial harassment. Those who reported having experienced racially-motivated verbal abuse were 60 per cent more likely to report having fair or poor health compared with those who said they had experienced no racial harassment. Those who reported experience of racially-motivated assault or property damage were over twice as likely to report fair or poor health. There was also a statistically significant association between a perception of British employers as discriminating against people on the basis of race or religion and self-reported fair or poor health. Those reporting that they believed ‘some’ or ‘most’ British employers to be discriminating were between 50 and 66 per cent more likely to report fair or poor health compared with those who believed a few, if any, British employers were racist. (See Karlsen and Nazroo 2002 for further details of the relationship between racism, class and specific health outcomes.)

Discussion

In terms of ethnic identity, these findings are consistent with earlier work suggesting that ethnic identity has nominal and virtual components and is influenced by internal and external definitions of ethnicity (Jenkins 1994). Of the dimensions of ethnic identity derived from this factor analysis, factors 1 and 2 (based on nationality, country of origin and skin colour) appear to be more closely related to the nominal components of ethnic identity. Of course, giving oneself a name also requires a perception of what it is to be that name, a perception that at least partly depends on the virtual component of ethnic identity, to which we would argue factors 3, 4 and 5 appear to be related.

Factor 3, traditional, which combined items relating to clothes, language, attitudes to mixed marriage and perception of oneself as British and Asian/Caribbean, may operate as a boundary of inclusion, providing an internal sense of identity. The elements of it that involve presentation of a public image through particular behaviours and participation in customs could be similar to what Smaje (1996) terms ‘unreflective ethnicity’, i.e. that there are aspects of behaviour that typify a particular ethnic group, while not being culturally marked by them. If this public behaviour is unreflective, yet is significant for identity, then it may involve some internalisation of external attitudes and the unconscious routinisation of behaviour, reflecting the dynamic nature of what Bourdieu (1977) calls ‘bodily hexis’, where ‘political mythology is realised and embodied and turned into a permanent disposition’ (1977: 94). However, while this would suggest some imposed internalisation in the nature of this aspect of ethnic identity, it is also likely that activities which are seen as traditional to an ethnic group will have been negotiated within the group, as well as externally influenced (see Hall’s 1992 discussion of the maintenance or revitalisation of tradition as a response to globalisation, summarised in the introduction to this paper). While aspects of this dimension of ethnic identity may be internally defined, and therefore the consequence of agency, the scope of those choices will be restricted and affected by the social structure.

Factor 4, community participation, could also be considered to reflect a boundary of inclusion. Compared with factor 3, however, the perceived need to establish ethnically-identified groups might be more likely to reflect both a response to exclusion by wider society and a positive celebration of ethnic group membership, a kind of politicised identity. This is one aspect of ethnicity as identity which could be construed as a ‘new social movement’ (Scott 1990).

The way in which an ethnic group may develop a form of politicised or essentialist identity as a racialised group in reaction to social constraint has been explored by Hall (1992). It is this dimension of identity which would appear to be reflected in factors 4 and 5. Factor 5, which incorporates perceptions and experience of racism, is the most obvious indicator, among those dimensions explored here, of external influences on ethnic identity. It could be argued that those who score highly on this factor will have recognised their ethnic status as one that has been racialised by the ethnic majority. This factor may, therefore, carry connotations similar to the term ‘blackness’ used in the 1970s and 1980s to describe the ‘expression of a common experience of exclusion and of a common political identity forged through resistance to that exclusion’ (Miles 1994: 7, see also Modood 1988). Although there was a statistically significant difference in the distribution of scores on this factor for the different ethnic groups, the fact that it consistently emerged for each, and the degree of overlap between distributions, suggests that, at least in terms of the measures used here, there is a broad similarity in the effects of experience of racism in relation to ethnic identity (Nazroo and Karlsen, forthcoming). This would also be consistent with work by Deaux et al. (1995) who described ‘ethnic/religious groups’ and ‘stigmatised groups’ as two distinct types of social identity. This suggests that there may be similarities in identity developed as a member of a stigmatised group, despite differences in (other) aspects of ‘ethnic/religious’ identity.

Indeed, in terms of the exploration of patterns of these dimensions of ethnic identity, the key finding is the similarity of factor loadings across the three ethnic groups (see Nazroo and Karlsen, forthcoming). Although the distribution of factor scores in some cases differed between the groups, this similarity of factor loadings across the ethnic groups would suggest that the basic nominal and virtual components of ethnic identity among different ethnic minority groups in Britain may be broadly similar. Indeed, differences for the factor that showed the largest variation between the Caribbean and South Asian groups, factor 3 (traditional), may be a result of Caribbean people not having the same types of ‘opportunity’ as South Asian people to present themselves as members of an ethnic minority group (or rather to score highly on the particular aspects of ‘traditional’ ethnic identity measured in this analysis): for example, only a fifth of Caribbeans compared to over 90 per cent of South Asian groups in this sample spoke a language other than English (Modood et al. 1997).

In terms of the relationship between the identified dimensions of ethnic identity and health, initial models that did not include age or occupational class might have suggested that, for the South Asian groups, higher scores on the more nominal and ‘unreflective’ elements of ethnic identity (factors 2 and 3) were related to increased risk of reporting fair or poor health. From this it could be hypothesised that, for more ‘unreflective’ aspects of ethnicity at least, this is related to culturally-determined health behaviours (such as diet, exercise and smoking). But, when age was included, these effects were substantially reduced and in the full models, which also included occupational class, they were no longer statistically significant. Dimensions of ethnic identity, as measured by the FNS, also had little association with the specific conditions explored in this analysis: heart disease, diagnosed diabetes and diagnosed hypertension.

It seems therefore that the initial effect was spurious – largely a consequence of the correlation between age and both identity and health. While this could be a cohort effect, earlier analysis using the FNS to explore the effect of country of birth and migration status on health indicators and health-related behaviours suggested the existence of a cohort effect was unlikely (Nazroo 1997a). So, while the existence of a cohort effect is impossible to rule out, it is likely that this is an age effect. Here, it is important to acknowledge that an apparent relationship between traditional elements of ethnic identity and health, that might be seen in a practice setting for example, may simply be confounded by age.

Findings from the model exploring the relationship between racism, class, and self-reported health suggested that experienced harassment, perceiving British employers to be discriminatory and occupational class all have independent effects on self-reported fair or poor health. Independent of social position, as measured by our indicator of occupational class, those reporting any experience of racial harassment had between 55 and 125 per cent greater risk of reporting fair or poor health compared with those who had not. Similarly, those who perceived the persistence of racist attitudes in over half of British employers had almost a 70 per cent increased risk of fair or poor health. These findings support earlier work suggesting a positive association between experience of racism (in various forms) and experience of poorer physical and mental health (Krieger and Sidney 1996, Krieger et al. 1993, Benzeval et al. 1992, James et al. 1987, for further discussion see Karlsen and Nazroo 2002). They would also seem to be supported by other studies suggesting a positive association between experiences of assault in general and heart disease (Williams et al. 1994) and psychological distress (Williams and Hunt 1997).

It is important to note the limitations of this study. As we pointed out earlier, racism and ethnic identity are multidimensional and historically located concepts which can be only partially captured by a cross-sectional quantitative survey. Some of the factors included a small number of outcomes, which is particularly problematic when undertaking factor analysis. These limitations can be seen in the relatively low levels of total sample variance explained by our five factors, and the relatively low Cronbach’s alpha reliability coefficients for the key variables clustering under some of the factors. Although the eigenvalues give some reassurance as to the validity of our findings, these limitations would suggest a need for caution when interpreting them.

Conclusion

The evidence from the data presented here and elsewhere (Nazroo and Karlsen forthcoming, Karlsen and Nazroo 2002) suggest that ethnicity as identity, in terms of self-description, self-presentation and behaviour, membership of ethnic minority organisations, and perceptions and experiences of racism, can be clearly and consistently identified across ethnic minority groups. However, our findings suggest that ethnicity as identity does not appear to influence health; rather ethnicity as structure – both in terms of racialisation and class experience – is strongly associated with health for ethnic minority people living in Britain. The analysis presented here showed independent relationships between perceptions of racial discrimination, experience of racial harassment, class indicators and health, with all three independent variables having a considerable impact on health. This highlights the limitations of current measures of structural effects used in research exploring ethnic inequalities in health: not only do we need measures that adequately account for the different forms of social disadvantage experienced by ethnic minority groups, we also need to explore the various ways in which racism itself can impact on physical and mental health. Although few studies have attempted to explore the role of racism in the health experience of people from ethnic minority groups, this would appear to be an important aspect of their daily lives in Britain, and one that needs to be incorporated into strategies to address ethnic inequalities in health.

Address for correspondence: Saffron Karlsen, Department of Epidemiology and Public Health, University College London, 1–19 Torrington Place, London WC1E 6BT e-mail: s.karlsen@public-health.ucl.ac.uk

Acknowledgements

This paper draws on research funded by the ESRC (L128251019) under the Health Variations Programme. The data used are drawn from the Fourth National Survey and thanks are due to the funders of the survey (particularly the Department of Health), advisory groups, colleagues at the Policy Studies Institute and the National Centre for Social Research, and, most importantly, the thousands of respondents who gave of their time. We would also like to thank the anonymous reviewers for their very helpful comments.

Ancillary