Public Health Policy Analysis (PHPA) Paper

Cumulative Effect of Racial Discrimination on the Mental Health of Ethnic Minorities in the United Kingdom

Stephanie Wallace, PhD, James Nazroo, PhD, and Laia Bécares, PhD

Objectives. To examine the longitudinal association between cumulative exposure to

racial discrimination and changes in the mental health of ethnic minority people.

Methods. We used data from 4 waves (2009–2013) of the UK Household Longitudinal

Study, a longitudinal household panel survey of approximately 40 000 households, in-

cluding an ethnic minority boost sample of approximately 4000 households.

Results. Ethnic minority people who reported exposure to racial discrimination at 1

time point had 12-Item Short Form Health Survey (SF-12) mental component scores 1.93

(95% confidence interval [CI] =–3.31, –0.56) points lower than did those who reported no

exposure to racial discrimination, whereas those who had been exposed to 2 or more

domains of racial discrimination, at 2 different time points, had SF-12 mental component

scores 8.26 (95% CI = –13.33, –3.18) points lower than did those who reported no ex-

periences of racial discrimination. Controlling for racial discrimination and other socio-

economic factors reduced ethnic inequalities in mental health.

Conclusions. Cumulative exposure to racial discrimination has incremental negative

long-term effects on the mental health of ethnic minority people in the United Kingdom.

Studies that examine exposure to racial discrimination at 1 point in time may underesti-

mate the contribution of racism to poor health. (Am J Public Health. 2016;106:1294–1300.

doi:10.2105/AJPH.2016.303121)

Racism is a system of structuring oppor-tunity and assigning values to people and groups based on phenotypic properties that unfairly disadvantages some individuals and communities, while unfairly advantaging others.1 International evidence now docu- ments that experiencing racism, either in- stitutionally, internalized, or personally mediated, is associated with poor health.2–7

Although the large majority of the literature is from cross-sectional studies, increasing longitudinal evidence now indicates that experiences of racial discrimination predate poor health,8–16 that changes in racial dis- crimination are associated with changes in mental health,17 and that chronic exposure to everyday racial discrimination is associated with poor sleep, coronary artery calcifica- tion,18,19 and altered diurnal cortisol patterns and higher cortisol awakening response.20

Despite these novel insights on the longitudinal association between racial

discrimination and health, there is a gap in our understanding of how the accumulation of exposure to racial discrimination over time is associated with increased morbidity. Some cross-sectional studies have shown that the accumulation of exposure to racial discrimi- nation across domains (e.g., at work, in educational settings, while seeking health care) leads to a dose–response association between racial discrimination and poor health.21–25

However, to date, studies have modeled experiences of racial discrimination as epi- sodic exposures, in which racial discrimina- tion is assumed to occur at 1 point in time, and most often within a particular domain.26

Experiencing racial discrimination likely has cumulative effects on health and therefore should be conceptualized as a dynamic process that operates across time, across domains, and even across generations.26 Studies that capture exposure to racial discrimination at 1 point in time, and assess domains in isolation, are likely to underestimate the overall burden of racial discrimination on the health of individuals and its contribution to ethnic inequalities in health.27

This study addressed these limitations by examining the longitudinal association be- tween cumulative exposure to racial discrim- ination, over time and across domains, and the mental health of ethnic minority people, and assessed its contribution to ethnic inequalities in mental health in the United Kingdom.

The setting of this study was the United Kingdom, where ethnic inequalities in health have been consistently documented. For example, Black Caribbean, Pakistani, and Bangladeshi people have between 6 and 9 fewer years of disability-free life expectancy at birth than do the White British group28 and are up to twice as likely as White British people to report poor self-rated health and to have a limiting long-standing illness.29

Experiences of racial discrimination appear to be a key contributor to ethnic inequalities in health in the United Kingdom, the United States, and elsewhere,2,4,7,30,31 but given data limitations, studies to date have not been able to fully examine longitudinal effects and whether, and how, cumulative exposure to racial discrimination leads to ethnic inequalities in health.

ABOUT THE AUTHORS All of the authors are with Centre on Dynamics of Ethnicity, University of Manchester, Manchester, Lancashire, UK.

Correspondence should be sent to Laia Bécares, PhD, University of Manchester, Humanities Bridgeford Street Building 2.13P, M13 9PL, UK (e-mail: [email protected]). Reprints can be ordered at http://www.ajph.org by clicking the “Reprints” link.

This article was accepted February 4, 2016. doi: 10.2105/AJPH.2016.303121

1294 Research Peer Reviewed Wallace et al. AJPH July 2016, Vol 106, No. 7

AJPH RESEARCH

METHODS This study used data from 4 waves of

the UK Household Longitudinal Study (UKHLS), a longitudinal household panel survey of approximately 40 000 households, including an ethnic minority boost sample of approximately 4000 households. The na- tionally representative annual survey pro- vided longitudinal data on factors such as health, education, income, and social life.32

For each wave, responses were collected over a 24-month period, conducted face-to-face via computer-aided personal interview. The first wave of the survey was carried out in 2009 to 2010, with subsequent waves col- lected in 2010 to 2011, 2011 to 2012, and 2012 to 2013.

The survey had multiple components: a representative general population sample, a general population comparison sample, an ethnic minority boost sample, and participants from the British Household Panel Survey from wave 2 onward. All ethnic minority respondents and the general population comparison sample completed, in addition to the general adult questionnaire, an extra 5 minutes of questions covering topics such as ethnic identity, migration histories, religious behavior, harassment, and employment discrimination. Further information on the UKHLS has been reported elsewhere.33

Mental Health We measured mental health with the

12-Item Short Form Health Survey (SF-12) Mental Component Summary (MCS),34

a measure of nonspecificpsychological distress that consists of 12 questions relating to the respondent’s self-reported general health, health limitations, emotional problems, pain, and feelings of depression and how they interfere with social activities. We used an algorithm to convert these items into a sin- gle mental functioning score ranging from 0 (low functioning) to 100 (high functioning), with higher values indicating better mental health. The MCS uses norm-based scoring to have a mean of 50 and an SD of 10 (see Ware et al.35 for complete scoring details).

Racial Discrimination Within the extra 5 minutes of questions,

the UKHLS included questions relating to

harassment and discrimination every 2 years, beginning in wave 1 and repeated in wave 3. These 4 questions asked respondents whether in the last 12 months they (1) had felt unsafe; (2) had avoided going to or being in several locations; (3) had been insulted, called names, threatened, or shouted at; or (4) had been physically attacked. For each domain of racial discrimination, several locations were listed for each of the questions, such as at school, at college, at work, on public transport, outside, in a public place, or at home. Respondents were asked to choose all that apply. The UKHLS adopts a 2-stage approach, whereby after responding positively to any of these items, respondents are asked the reasons that these incidents occurred. Possible attributions included their sex, age, ethnicity, sexual orientation, health or disability, nationality, religion, language or accent, or dress or appearance. We recoded these variables to indicate whether the respondent had experienced feeling unsafe, avoided a space or place, been assaulted, or been insulted because of racial discrimination based on his or her ethnicity, nationality, or religion. Because of the small number of respondents who stated that they had been physically assaulted, we combined this measure with the indicator of verbal insults.

The UKHLS also asked about discrimi- nation in the workplace within the last 12 months for those respondents who were employed during this time. Three questions asked whether the respondent had been turned down for a job, turned down for a promotion, or turned down for job-related training. The same 2-step approach was used, in which the second part of the question asked respondents the reasons that they were turned down. As with the measures of in- terpersonal racial discrimination, we grouped together the potential attributions of eth- nicity, nationality, or religion. Because of the small number of respondents stating that they had experienced racial discrimination in theworkplacewithinthelastyear,wecombined these questions into a single binary variable indicating any employment discrimination.

We created 2 cross-sectional summary variables of exposure to racial discrimination. These variables were binary and identified whether the respondent had experienced any

form of racial discrimination at wave 1 and wave 3.

To measure cumulative experiences of racial discrimination over time, we created a longitudinal summary variable that in- dicated whether the respondent reported any racial discrimination (being physically or verbally insulted, feeling unsafe, avoiding a space or place, or facing employment dis- crimination) at 1 time point or at 2 time points. Categories included no experiences of racial discrimination; experiences of racial discrimination at 1 time point (wave 1 or wave 3); and experiences of racial discrimi- nation at 2 time points (wave 1 and wave 3). We also created a summary variable that combined cumulative exposure to different domains of racial discrimination across time. This dose–response variable was summarized into 6 categories:

1. No experiences of racial discrimination, 2. Exposure to 1 domain of racial discrimi-

nation at 1 time point (wave 1 or wave 3), 3. Exposure to 2 or more domains of

racial discrimination at 1 time point, 4. Exposure to 1 domain of racial discrimi-

nationat 2timepoints (wave1 andwave3), 5. Exposure to 2 or more domains of racial

discrimination at 1 time point and 1 exposure to racial discrimination at a sec- ond time point,

6. Exposure to 2 or more domains of racial discrimination at 2 time points.

Covariates We measured ethnicity with a self-

reported variable based on the 2011 Census categories for England and Wales. Re- spondents were asked to select 1 of the 18 categories that best described their ethnic group. We have reported on the ethnic minority groups with sufficiently large sam- ples: Indian, Pakistani, Bangladeshi, Black Caribbean, and Black African. We compared these groups with the White British group.

We considered factors thought to be as- sociated with both experiences of racial discrimination and mental health in analytical models. These included sex, age (continu- ous), and equivalized household income (continuous) at wave 1. Equivalized income, a measure of socioeconomic position, was conceptualized and modeled in this study as

AJPH RESEARCH

July 2016, Vol 106, No. 7 AJPH Wallace et al. Peer Reviewed Research 1295

a consequence of the discriminatory practices experienced by ethnic minority people in a range of domains, including education, residential history, and employment. Equivalized household income was calcu- lated as the sum of the gross monthly household income divided by the modified Organisation for Economic Co-operation and Development scale. A small number of respondents (n = 70) recorded a negative income value, and thus these were recoded to 0 rather than excluding them from the sample.

Analysis Plan To examine the burden of experiencing

racial discrimination on the mental health of ethnic minority people and to explore the longitudinal associations between cumulative exposure to racial discrimination and mental health, we fitted a series of multiple linear regression models. The first set of linear regression models examined the association between the different measures of racial dis- crimination and mental health at wave 4. Withineachofthesemodels, wecontrolled for MCS scores at wave 1, while adjusting for age, sex, and equivalized household income.

For the analyses that aimed to model the contribution of racism to the risk of mental illness for ethnic minority people, compared with ethnic majority people, we built 2 linear regression models, one using cross-sectional data and the other using longitudinal data, reflecting different ap- proaches to modeling the extent of ethnic inequality. The cross-sectional model pro- vided a more comprehensive account of the association between the markers of social and economic inequality and ethnic inequalities in mental health, because it described their potential contribution to ethnic differences as observed in the population. However, be- cause it was cross-sectional, it may have contained some element of reverse causation (e.g., mental illness leading to lower incomes) and hence overestimated causal effects. The longitudinal model was a stricter test of causal effects, but because it modeled change over 4 waves of data, it did not account for causal effects that would have occurred before the initial observation period.

We built both linear regression models in a stepwise manner. We first compared each

ethnic minority group with the White British group (the reference category), adjusted for sex and age differences across ethnic groups (step 1). Then, we included 2 individual markers of social and economic inequality that could be considered to be a consequence of living in a context where identities are racialized. As the first marker, we used equivalized household income, our measure of socioeconomic position (step 2). As the second marker, we used reports of their exposure to racial discrimination (step 3). The final step (step 4) included both equivalized household income and experi- ences of racial discrimination.

All models were fitted in Stata version 1336

and included the appropriate cross-sectional and longitudinal weights to account for the stratified sample and nonresponse.32

RESULTS Levels of psychological distress were

similar for the White British (mean = 49.6; SE = 0.89) and Indian groups (mean = 49.4; SE = 0.51). The Black African group (mean = 50.9; SE = 0.57) had significantly lower levels of distress than did the White British group, whereas the Pakistani (mean = 45.9; SE = 0.78), Bangladeshi (mean = 46.5; SE = 1.54), and Black Carib- bean groups (mean = 48.3; SE = 0.60) all had significantly higher levels of distress than did the White British group.

Table 1 shows the prevalence of racial harassment and discrimination experienced by ethnic minority groups at waves 1 and 3. All ethnic minority groups reported higher levels of racial discrimination at wave 3 compared with those reported at wave 1: more than one third of the Bangladeshi group (35%) and more than a quarter of the Indian (28%), Pakistani, (27%), and Black African (26%) groups reported that they had expe- rienced some form of racial discrimination at wave 3, compared with about 1 in 5 in wave 1. Table 1 also shows the prevalence of cumulative exposure to racial discrimination over time and across domains at wave 4. The Bangladeshi group consistently reported the highest cumulative exposure to racial discrimination across all of the domains of racial discrimination, whereas the Black Caribbean group reported the least exposure.

Table 2 shows the effects of racial dis- crimination on mental health. Compared with respondents who reported no experi- ences of racial discrimination, respondents who reported exposure to any domain of racial discrimination at 1 time point (either at wave 1 or at wave 3) had a deterioration in mental health scores (MCS) at wave 4 by 2.27 (95% confidence interval [CI] = –3.42, –1.12) points. Exposure to racial discrimi- nation at both time points reduced MCS scores by 5.78 (95% CI = –8.47, –3.10) points. Those who reported that they had previously been insulted or attacked at 1 time point (either at wave 1 or at wave 3) had MCS scores 3.38 (95% CI = –5.10, –1.67) points lower, and those who reported exposure to racial insults or attacks at both wave 1 and wave 3 had MCS scores 5.03 (95% CI = –8.36, –1.69) points lower than did those who reported that they had not been insulted or attacked because of their ethnicity, nationality, or religion. Similar associations were found for those who reported that they had felt unsafe and those who reported that they avoided places.

The final section in Table 2 shows the dose–response effects over time and across domains. Respondents who reported expo- sure to 1 domain of racial discrimination at 1 time point had MCS scores 1.93 (95% CI = –3.31, –0.56) points lower, and respondents who reported exposure to 2 domains of racial discrimination at 1 time pointhad MCS scores 2.98 (95% CI = –4.57, –1.33) points lower than did those who reported no exposure to racial discrimination. Respondents who reported exposure to 2 domains of racial discrimination at 1 time point and further exposure at a second time point had MCS scores 5.65 (95% CI = –8.90, –2.40) points lower than did those who reported no ex- posure to racial discrimination. Finally, those who reported 2 or more domains of racial discrimination at 2 time points had MCS scores 8.26 (95% CI = –13.33, –3.18) points lower than did those who reported no ex- posure to racial discrimination.

Table 3 shows cross-sectional differences in mental health scores for each ethnic mi- nority group, compared with the White British group. Adjusting for age and sex in step 1 shows the significantly lower levels of av- erage mental health scores for Pakistani, Bangladeshi, and Black Caribbean people,

AJPH RESEARCH

1296 Research Peer Reviewed Wallace et al. AJPH July 2016, Vol 106, No. 7

compared with White British people. Addi- tionally adjusting for income differences in step 2 reduced the coefficients for the Paki- stani, Bangladeshi, and Black Caribbean groups substantially and to nonsignificance

for Pakistani and Bangladeshi people. Step 3 adjusted for exposure to racism and dis- crimination and similarly reduced the nega- tive coefficients for these 3 ethnic minority groups, although a significant disadvantage

remained for each of them. The final model (step 4) adjusted simultaneously for both in- come differences and racism, and in this model, substantial reductions occurred in the negative coefficients for the Pakistani,

TABLE 1—Prevalence of Racial Discrimination Among Ethnic Minority Groups at Waves 1, 3, and 4: UK Household Longitudinal Study, 2009–2013

Indian, Weighted % or No.

Pakistani, Weighted % or No.

Bangladeshi, Weighted % or No.

Black Caribbean, Weighted % or No.

Black African, Weighted % or No.

Wave 1

Any domain of racial discrimination 20.5 17.5 21.9 14.1 22.4

Verbally or physically assaulted 9.4 8.6 16.7 9.7 14.4

Felt unsafe 15.3 13.3 17.0 5.9 12.4

Avoided places 6.4 5.8 9.9 2.1 5.6

Job discrimination 1.1 0.4 0.3 2.5 2.6

Wave 3

Any domain of racial discrimination 27.6 27.2 35.1 14.4 26.0

Verbally or physically assaulted 11.2 10.7 10.6 4.8 12.3

Felt unsafe 21.0 20.1 26.1 8.3 16.1

Avoided places 13.3 15.0 21.9 5.4 11.9

Job discrimination 1.1 0.9 1.3 1.2 1.8

Wave 4

Any domain of racial discrimination

No exposure 62.5 61.3 58.1 76.3 60.8

1 event at 1 time point 26.9 32.7 26.7 19.1 29.9

2 events at 2 separate time points 10.6 6.0 15.2 4.6 9.3

Verbally or physically assaulted

No exposure 82.5 83.4 76.8 86.5 76.4

1 event at 1 time point 14.4 14.0 19.2 12.5 20.5

2 events at 2 separate time points 3.1 2.7 4.1 1.0 3.1

Felt unsafe

No exposure 70.7 70.7 68.1 86.9 76.0

1 event at 1 time point 22.4 25.2 20.7 12.1 19.4

2 events at 2 separate time points 6.9 4.1 11.2 1.1 4.6

Avoided places

No exposure 83.8 81.1 75.3 93.4 83.4

1 event at 1 time point 12.8 17.0 17.5 5.7 15.6

2 events at 2 separate time points 3.5 1.9 7.2 0.9 1.0

Dose–response

No exposure 62.5 61.3 58.1 76.3 60.8

1 event at 1 time point 16.3 17.1 12.2 14.9 18.6

‡ 2 events at 1 time point 10.6 15.6 14.5 4.2 11.3 1 event at 2 time points 2.4 0.6 4.0 1.2 1.8

‡ 2 events at 1 time point and 1 event at another time point

3.7 2.9 0.9 2.9 3.6

‡ 2 events at 2 time points 4.5 2.6 10.2 0.6 3.8

Unweighted base 846 627 417 502 510

Weighted base 508 398 294 354 329

AJPH RESEARCH

July 2016, Vol 106, No. 7 AJPH Wallace et al. Peer Reviewed Research 1297

Bangladeshi, and Black Caribbean groups. For the Pakistani and Bangladeshi groups, associations became nonsignificant, and for the Indian group, we saw a substantial increase in the positive coefficients. For the Black African group, results showed a mental health advantage, as compared with the White British group, once their economic and racism disadvantages were controlled for.

Table 4 presents findings from the model predicting longitudinal change in mental health scores. Results showed that whereas change over time in mental health among Black Caribbean, Indian, and Bangladeshi people did not differ from that of White British people, inequalities in mental health became greater over time for the Pakistani group, whose SF-12 scores decreased by 3.20 points (95% CI = –4.48, –1.93) relative to the White British group. After adjusting for socioeconomic disad- vantage (step 2), the increase in poor mental health for Pakistani people,

compared with White British people, was reduced, and this inequality attenuated even further after additionally adjusting for experiences of racial discrimination, al- though it remained statistically significant. For the Black African population, we saw an improvement in mental health over time, compared with the White British group, and this association strengthened as we adjusted for socioeconomic disadvantage and experiences of racial discrimination (see steps 2–4).

DISCUSSION In this study, we set out to explore

whether, and how, cumulative exposure to racial discrimination over time was associated with the mental health of ethnic minority people in the United Kingdom. In a novel contribution to the literature, we have documented the corrosive effect that the cumulative experience of racial

discrimination has on the mental health of ethnic minority people. We found a cumu- lative, dose–response relation between ex- periences of racial discrimination and the mental health of ethnic minority people, so that ethnic minority people who reported repeated occurrences of racial discrimination, over time and across domains, had a reduction of 8 points in their MCS scores, compared with their peers who did not report any experiences of racial discrimination. Fear of racial discrimination expressed through reporting feeling unsafe or avoiding spaces or places had the biggest cumulative effect on the mental health of ethnic minority people. This findingwouldsuggestthatpreviousexposureto racial discrimination over the life course, or awareness of racial discrimination experienced by others, can continue to affect the mental health of ethnic minority people, even after the initial exposure to racial discrimination. Other UK-based studies also have reported the in- creased harm of fear of experiencing racial discrimination on health,7,30 which likely captures not only the previous experiences of racial discrimination as described earlier but also the vigilance and anticipatory stress of a possible future racist encounter.

In the second part of the study, we assessed the contribution of racial discrimination to ethnic inequalities in mental health. We did this by modeling 2 different dimensions of racial disadvantage that lead to poor health: the direct experiences of racism on physio- logical changes37 and the social and economic consequences of living in a racialized soci- ety.38 We found that in the cross-sectional analyses, adjusting for socioeconomic disad- vantage and experiences of racial discrimi- nation eliminated ethnic inequalities in mental health for Pakistani and Bangladeshi people and reduced inequalities for Black Caribbean people. Findings from the longi- tudinal analyses showed that controlling for socioeconomic disadvantage and experiences of racial discrimination attenuated inequalities in mental health for Pakistani people, as compared with White British people.

Even though we analyzed longitudinal data and accounted for cumulative exposure to racial discrimination over time, and across domains, we assessed only experiences of racial discrimination that participants had experi- enced when they were sampled by the UKHLS,andthuswewerenotabletoassesstheir

TABLE 2—Longitudinal Association Between Accumulation of Reported Racial Discrimination Experienced at Waves 1 and 3 and Psychological Distress (SF-12 Scores) at Wave 4 Among Ethnic Minority People: UK Household Longitudinal Study, 2009–2013

b (95% CI)

Any domain of racial discrimination

No exposure (Ref) 1

1 event at 1 time point –2.27 (–3.42, –1.12)

2 events at 2 separate time points –5.78 (–8.47, –3.10)

Verbally or physically assaulted

No exposure (Ref) 1

1 event at 1 time point –3.38 (–5.10, –1.67)

2 events at 2 separate time points –5.03 (–8.36, –1.69)

Felt unsafe

No exposure (Ref) 1

1 event at 1 time point –3.11 (–4.52, –1.69)

2 events at 2 separate time points –6.36 (–10.08, –2.65)

Avoided places

No exposure (Ref) 1

1 event at 1 time point –2.15 (–3.62, –0.67)

2 events at 2 separate time points –8.15 (–15.50, 0.08)

Dose–response

No exposure (Ref) 1

1 event at 1 time point –1.93 (–3.31, –0.56)

‡ 2 events at 1 time point –2.98 (–4.57, –1.33) 1 event at 2 time points –1.87 (–4.90, 1.15)

‡ 2 events at 1 time point and 1 event at another time point –5.65 (–8.90, –2.40) ‡ 2 events at 2 time points –8.26 (–13.33, –3.18)

Note. CI = confidence interval; SF-12 = 12-Item Short Form Health Survey.

AJPH RESEARCH

1298 Research Peer Reviewed Wallace et al. AJPH July 2016, Vol 106, No. 7

previous experiences of racial discrimination or their lifetime exposure to social inequality. It has been argued elsewhere39 that certain measures of socioeconomic disadvantage contain significant residual confounding of an underlying concept, and this was likely reflected in our results. For example, income does not adequately reflect all dimensions of disadvantage, and, similarly, ex- posure to racial discrimination in the last 12 months cannot fully capture the effects ofracial discrimination over thelifecourse.It also should be recognized that socioeconomic

disadvantage and racial discrimination are not evenly distributed across all ethnic minority groups, and this was evidenced in our findings.

In both cross-sectional and longitudinal analyses, we found that for the Black African group, taking into account the harm created by racial discrimination actually improved their levels of mental health, as compared with the White British group. The health advantage of the Black African group relative to the White British group has been previously reported,29 so it was not surprising to see this improvement

increase once the effects of racialization were accountedfor.Like alloftheotherethnicgroups included in this study, the Black African group was heterogeneous in terms of country of origin, reasonsfor migration,anddifferencesin both the health and the socioeconomic profiles of the individual subgroups.40 Therefore, the health advantage of the Black African group reported here may be applicable to only some people within this large group.

Limitations Even though this study was able to take

advantage of longitudinal and multidimen- sional data, it was limited in some respects. First, the UKHLS did not ask respondents about exposure to racial discrimination over their life course. Therefore, we were unable to consider any of the processes or experiences of racial discrimination before their first interview.

Second, even though we were able to examine experiences across various domains of racial discrimination, the domains explored do not represent the full range of circum- stances and places where racial discrimination could be experienced; thus, results pre- sented here may have underestimated the prevalenceofracialdiscriminationexperienced by ethnic minority people in the United Kingdomand its association with mental health.

Third, we observed higher levels of racial discrimination at wave 3 than we observed at wave 1, indicating possible measurement error. In a previous study, people initially stated on a questionnaire that they had not experienced racial discrimination; however, during an in-depth interview later, they said that they had experienced racial discrimina- tion but found it too difficult to discuss.29 In this case, perhaps respondents were more willing to report experiences of racial dis- crimination at the second interview after having been alerted to the content of the questionnaire at the previous interview.

Conclusions Several longitudinal studies showed that

racial discrimination predates poor health and reinforces ethnic inequalities in health.31 In this study, we confirmed the longitudinal effects in a large population-based study and additionally showed that cumulative expo- sure to racial discrimination over time

TABLE 3—Ethnic Inequalities in the Cross-Sectional Association Between Experiences of Racial Discrimination and Psychological Distress (SF-12 Scores): UK Household Longitudinal Study, 2009–2013

Step 1, b (95% CI) Step 2, b (95% CI) Step 3, b (95% CI) Step 4, b (95% CI)

White British (Ref) 1 1 1 1

Indian 0.08 (–0.52, 0.69) 0.10 (–0.49, 0.70) 0.37 (–0.22, 0.96) 0.39 (–0.19, 0.97)

Pakistani –1.04 (–1.76, –0.31) –0.59 (–1.31, 0.14) –0.77 (–1.50, –0.05) –0.33 (–1.05, 0.39)

Bangladeshi –1.22 (–2.11, 0.32) –0.79 (–1.69, 0.11) –0.95 (–1.85, –0.05) –0.52 (–1.43, 0.38)

Black Caribbean –1.08 (–1.73, –0.43) –0.94 (–1.59, –0.29) –0.84 (–1.50, –0.19) –0.71 (–1.36, –0.51)

Black African 0.27 (–0.41, 0.96) 0.59 (–0.09, 1.28) 0.54 (–0.14, 1.23) 0.86 (0.19, 1.54)

Constant 50.51 (50.20, 50.83) 49.18 (48.81, 49.55) 50.52 (50.20, 50.83) 49.18 (48.81, 49.55)

Note. CI = confidence interval; SF-12 = 12-Item Short Form Health Survey. Step 1 controls for ethnicity, sex, and age; Step 2 controls for ethnicity, sex, age, and equivalized household income; Step 3 con- trolsfor ethnicity, sex, age, and exposureto racial discrimination; and Step 4 controlsfor ethnicity, sex, age, equivalized household income, and exposure to racial discrimination. Range of unweighted SF-12 scores: White British group (0–77.11); Indian group (12.08–70.46); Black African group (10.29–68.65); Pakistani group (7.95–70.45); Bangladeshi group (4.89–69.18); and Black Caribbean group (8.22–70.53).

TABLE 4—Ethnic Inequalities in the Longitudinal Association Between Experiences of Racial Discrimination at Waves 1 and 3 and Psychological Distress (SF-12 Scores) at Wave 4: UK Household Longitudinal Study, 2009–2013

Step 1, b (95% CI) Step 2, b (95% CI) Step 3, b (95% CI) Step 4, b (95% CI)

White British (Ref) 1 1 1 1

Indian 0.20 (–0.70, 1.10) 0.24 (–0.65, 1.14) 0.70 (–0.23, 1.64) 0.74 (–0.18, 1.66)

Pakistani –2.15 (–3.44, –0.85) –1.86 (–3.15, –0.57) –1.66 (–2.96, –0.37) –1.38 (–2.68, –0.09)

Bangladeshi –1.51 (–4.66, 1.64) –1.27 (–4.44, 1.90) –1.01 (–4.10, 2.08) –0.48 (–0.65, 1.53)

Black Caribbean 0.08 (–1.01, 1.16) 0.15 (–0.93, 1.24) 0.37 (–0.73, 1.47) 0.41 (–0.69, 1.50)

Black African 2.17 (1.03, 3.31) 2.38 (1.24, 3.51) 2.71 (1.58, 3.83) 2.90 (1.80, 4.03)

Constant 23.85 (22.82, 24.88) 23.27 (22.22, 24.31) 23.89 (22.86, 24.92) 23.31 (22.26, 24.36)

Note. CI = confidence interval; SF-12 = 12-Item Short Form Health Survey. Step 1 controls for Mental Component Summary (MCS) score at wave 1, ethnicity, sex, and age; Step 2 controls for MCS score at wave 1, ethnicity, sex, age, and equivalized household income; Step 3 controls for MCS score at wave 1, ethnicity, sex, age, and exposure to racial discrimination; and Step 4 controls for MCS score at wave 1, ethnicity, sex, age, equivalized household income, and exposure to racial discrimination. Range of unweighted SF-12 scores: White British group (0–78.08); Indian group (12.08–71.28); Black African group (11.14–75.11); Pakistani group (9.82–75.83); Bangladeshi group (12.43–68.65); and Black Caribbean group (12.33–68.18).

AJPH RESEARCH

July 2016, Vol 106, No. 7 AJPH Wallace et al. Peer Reviewed Research 1299

significantly worsens mental health. By making full use of new longitudinal data, we have been able to show how repeated ex- posure to racial discrimination over time, and accumulation of exposure across domains, affects the psychological distress of ethnic minority people in the United Kingdom and contributes to persistent ethnic inequalities in mental health. Studies that assess the cross- sectional association between racial discrim- ination and health, or examine exposure at 1 point in time, underestimate the harm of racial discrimination on the mental health of ethnic minority people and its contribution to ethnic inequalities in health.

CONTRIBUTORS All authors conceptualized the study and contributed to the interpretation of results and to the final version of the article. S. Wallace conducted the analyses guided by J. Nazroo and L. Bécares. L. Bécares wrote the first draft of the article.

ACKNOWLEDGMENTS This work was partially funded by a UK Economic and Social Research Council (ESRC) grant to J. Nazroo (ES/ K002198/1). L. Bécares was supported by a Hallsworth Research Fellowship and an ESRC Future Research Leaders grant (ES/K001582/1).

HUMAN PARTICIPANT PROTECTION Ethicsapprovalwasnotsoughtforthisworkbecauseitused publicly available, anonymized data.

REFERENCES 1. Jones CP. Confronting institutionalized racism. Phylon. 2002;50(1/2):7–22.

2. Paradies Y. A systematic review of empirical research on self-reported racism and health. Int J Epidemiol. 2006;35: 888–901.

3. Pascoe EA, Smart Richman L.Perceived discrimination and health: a meta-analytic review. Psychol Bull. 2009; 135(4):531–554.

4. Williams DR, Mohammed SA. Discrimination and racial disparities in health: evidence and needed research. J Behav Med. 2009;32(1):20–47.

5. Williams DR, Neighbors HW, Jackson JS. Racial/ ethnic discrimination and health: findings from com- munity studies. Am J Public Health. 2003;93(2):200–208.

6. Lewis TT, Cogburn CD, Williams DR. Self-reported experiences of discrimination and health: scientific ad- vances, ongoing controversies, and emerging issues. Annu Rev Clin Psychol. 2015;11:407–440.

7. Karlsen S, Nazroo J. Relation between racial dis- crimination, social class, and health among ethnic mi- nority groups. Am J Public Health. 2002;92(4):624–631.

8. Barnes LL, de Leon CF, Lewis TT, Bienias JL, Wilson RS, Evans DA. Perceived discrimination and mortality in a population-based study of older adults. Am J Public Health. 2008;98(7):1241–1247.

9. Brody GH, Chen Y-F, Murry VM, et al. Perceived discrimination and the adjustment of African American youths: a five-year longitudinal analysis with contextual moderation effects. Child Dev. 2006;77(5):1170–1189.

10. Brown TN, Williams DR, Jackson JS, et al. “Being Black and feeling blue”: the mental health consequences of racial discrimination. Race Soc. 2000;2(2): 117–131.

11. Gee G, Walsemann K. Does health predict the reporting of racial discrimination or do reports of dis- crimination predict health? Findings from the National Longitudinal Study of Youth. Soc Sci Med. 2009;68: 1676–1684.

12. Jackson JS, Brown TN, Williams DR, Torres M, Sellers SL, Brown K. Racism and the physical and mental health status of African Americans: a thirteen year national panel study. Ethn Dis. 1996;6(1-2):132–147.

13. Luo Y, Xu J, Granberg E, Wentworth WM. A longitudinal study of social status, perceived discrimina- tion, and physical and emotional health among older adults. Res Aging. 2012;34(3):275–301.

14. Seaton EK, Neblett EW, Upton RD, Hammond WP, Sellers RM. The moderating capacity of racial identity between perceived discrimination and psychological well-being over time among African American youth. Child Dev. 2011;82(6):1850–1867.

15. Kwate NO, Goodman M. Cross-sectional and lon- gitudinal effects of racism on mental health among resi- dents of Black neighborhoods in New York City. Am J Public Health. 2015;105:711–718.

16. Schulz AJ, Gravlee CC, Williams DR, Israel BA, Mentz G, Rowe Z. Discrimination, symptoms of de- pression, and self-rated health among African American women in Detroit: results from a longitudinal analysis. Am J Public Health. 2006;96(7):1265–1270.

17. Rosenthal L, Earnshaw V, Lewis T, et al. Changes in experiences with discrimination across pregnancy and postpartum: age differences and consequences for mental health. Am J Public Health. 2015;105:686–693.

18. Lewis TT, Everson-Rose SA, Powell LH, et al. Chronic exposure to everyday discrimination and coro- nary artery calcification in African-American women: the SWAN Heart Study. Psychosom Med. 2006;68(3): 362–368.

19. Lewis TT, Troxel WM, Kravitz HM, Bromberger JT, Matthews KA, Hall MH. Chronic exposure to everyday discrimination and sleep in a multiethnic sample of middle-aged women. Health Psychol. 2013;32(7): 810–819.

20. Adam EK, Heissel JA, Zeiders KH, et al. De- velopmental histories of perceived racial discrimination and diurnal cortisol profiles in adulthood: a 20-year prospective study. Psychoneuroendocrinology. 2015;62: 279–291.

21. Harris RB, Cormack DM, Stanley J. The relationship between socially-assigned ethnicity, health and experi- ence of racial discrimination for M�aori: analysis of the 2006/07 New Zealand Health Survey. BMC Public Health. 2013;13:844.

22. Harris R, Cormack D, Stanley J, Rameka R. In- vestigating the relationship between ethnic consciousness, racial discrimination and self-rated health in New Zealand. PLoS One. 2015;10(2):e0117343.

23. Harris R, Cormack D, Tobias M, et al. The pervasive effects of racism: experiences of racial discrimination in New Zealand over time and associations with multiple health domains. Soc Sci Med. 2012;74(3):408–415.

24. Harris R, Tobias M, Jeffreys M, Waldegrave K, Karlsen S, Nazroo J. Racism and health: the relationship between experience of racial discrimination and health in New Zealand. Soc Sci Med. 2006;63(6):1428–1441.

25. Harris R, Tobias M, Jeffreys M, Waldegrave K, Karlsen S, Nazroo J. Effects of self-reported racial discrimination and deprivation on M�aori health and inequalities in New Zealand: cross-sectional study. Lancet. 2006;367(9527):2005–2009.

26. National Research Council. Measuring Racial Dis- crimination: Panel on Methods for Assessing Discrimination. Blank RM, Dabady M, Citro C, eds. Washington, DC: National Academies Press; 2004.

27. Williams DR, Neighbors H. Racism, discrimination and hypertension: evidence and needed research. Ethn Dis. 2001;11(4):800–816.

28. Wohland P, Rees P, Nazroo J,Jagger C. Inequalities in healthy life expectancy between ethnic groups in England and Wales in 2001. Ethn Health. 2015;20(4):341–353.

29. Bécares L. Which ethnic groups have the poorest health? In: Jivraj S, Simpson L, eds. Ethnic Identity and Inequalities in Britain. The Dynamics of Diversity. London, England: Policy Press; 2015:123–140.

30. Bécares L, Stafford M, Nazroo J. Fear of racism, employment and expected organizational racism: their association with health. Eur J Public Health. 2009;19(5): 504–510.

31. Paradies Y, Ben J, Denson N, et al. Racism as a de- terminant of health: a systematic review and meta- analysis. PLoS One. 2015;10(9):e0138511.

32. Knies G, ed. Understanding Society – The UK Household Longitudinal Study: Waves 1-4, 2009-2013, User Manual. Colchester, UK: Institute for Social and Economic Research, University of Essex; 2014.

33. Lynn P. Sample Design for Understanding Society. Colchester, UK: Institute for Social and Economic Research, University of Essex; 2009.

34. Karlsen S, Nazroo J. Fear of racism and health. J Epidemiol Community Health. 2004;58:1017–1018.

35. Ware J, Kosinski M, Turner-Bowker D, Gandek B. User’s Manual for the SF-12v2 Health Survey With a Supplement Documenting SF-12 Health Survey. Lincoln, RI: QualityMetric Inc; 2009.

36. StataCorp. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP; 2013.

37. Clark R, Anderson N, Clark V, Williams D. Racism as a stressor for African Americans: a biopsychosocial model. Am Psychol. 1999;54(10):805–816.

38. Black D, Morris J, Smith C,Townsend P. Inequalities in Health: Report of a Research Working Group. London, England: DHSS; 1980.

39. Nazroo JY. Genetic, cultural or socio-economic vulnerability? Explaining ethnic inequalities in health. Sociol Health Illn. 1998;20(5):710–730.

40. Aspinall PJ, Chinouya M. Is the standardised term “Black African” useful in demographic and health re- search in the United Kingdom? Ethn Health. 2008;13(3): 183–202.

AJPH RESEARCH

1300 Research Peer Reviewed Wallace et al. AJPH July 2016, Vol 106, No. 7