• Open Access

It's enough to make you sick: the impact of racism on the health of Aboriginal Australians

Authors


Correspondence to:
A/Prof. Ann Larson, Combined Universities Centre for Rural Health, PO Box 109, Geraldton, Western Australia 6531. Fax: (08) 9964 2096; e-mail: alarson@cucrh.uwa.edu.au

Abstract

Background: Experience of interpersonal racism has been neglected as a mechanism by which inequalities between Aboriginal and non-Aboriginal people are created and maintained.

Methods: Cross-sectional survey of randomly selected residents of a rural Australian town (n=639). Interpersonal racism was measured by two questions on experiences in the past four weeks of negative racially based treatment that evoked an emotional or physical response. Health was measured with the mental and physical health component scores of the Short-Form 12 and self-reported fair or poor general health. Linear and logistic regressions modelled the effects of interpersonal racism on health, controlling for age, sex, socio-economic status and Aboriginality.

Findings: The 183 Aboriginal respondents had lower health component scores, were more than twice as likely to report fair-to-poor general health (34% compared with 17%, p<0.001), and 2.6 to 5.0 times more likely to report negative racially based treatment. Demographic and socio-economic characteristics were not associated with reporting negative racially based treatment. After controlling for other variables, Aboriginal respondents who reported negative treatment were more likely to have poor health on all three measures. Non-Aboriginal respondents who reported experiencing negative treatment had lower mental health component scores.

Implications: Experiencing racist treatment should be recognised as a social determinant of health. Improved health care and other initiatives may not eliminate health inequalities in the absence of fundamental changes in how non-Aboriginal people behave towards Aboriginal people.

Researchers are increasingly naming racism as the cause of persistent health differences by racial or ethnic classification in the United States,1,2 the United Kingdom3 and New Zealand.4 In Australia, racism has been implicitly or explicitly named as the root cause of the extreme socio-economic and health disadvantage experienced by Aboriginal Australians.5–7 However, in explaining poor health, most research places greater emphasis on social and economic disadvantage,8 locational disadvantage,9 lack of investment in effective health interventions10 and insensitivity to cultural differences.11 In this paper, we use cross-sectional data to test the relationship between experience of racism and health for Aboriginal Australians. Our research will assist in the development of better interventions to address and eliminate racially based disparities.12

Racism and its effects are usually conceptualised as occurring on two levels: institutional and interpersonal.13 Institutional racism is expressed through economic and political systems and maintained by the policies and practices carried out by government and other institutions. Examples of institutional racism can be found in all sectors, from public housing14 to health care.15 The result, intentional or unintentional, is that Aboriginal people receive less benefit from the same policies. Interpersonal racism is the discriminatory interactions between individuals, such as demeaning comments by a health care provider or a shop assistant or the behaviour of a neighbour. Jones adds internalised racism as a third level, which refers to the adaptations made by persons who experience racial discrimination.12 Internalising negative racial stereotypes is a consequence of institutional and interpersonal racism and can result in low self-esteem, depression and hostility.16 Our study directly measures institutional and interpersonal racism, but internalised racism is not easily captured in cross-sectional studies and is not addressed.16,17

Racism and health

Institutional racism is often measured through differences in groups’ socio-economic status and there is a persistent relationship between socio-economic variables and health indicators. However, studies that statistically control for differences in education, employment or other indicators of institutional racism do not explain all of the differences in health status between racial groups.13,18

Two comprehensive reviews of the literature have found a consistent association between self-reported experience of racial discrimination and poor mental health outcomes, using indicators ranging from self-reported measures of general mental well-being to psychosis.17,19 The association between racial discrimination and somatic health is less consistent,17 but positive associations have been found with self-reported general poor health, bed-days, hypertension, blood pressure and smoking.3,4,20–22 Six studies report that the negative relationship between self-reported experience of racial discrimination and health remains after controlling for socio-economic variables, which suggests the effects of institutional and interpersonal racism are additive.2–4,18,21,23

Interpersonal racism is more frequently hypothesised to cause poorer health through biological pathways involving prolonged heightened stress. The anger, frustration and humiliation such behaviour provokes results in a range of biological responses including greater release of cortisol.24 Wyatt and colleagues reviewed the theoretical, experimental and population-based research on racism and physiological functioning related to cardiovascular diseases including acute and chronic heightened blood pressure, sodium excretion and neurochemical processes inhibiting immune functions.16 Chronic exposure to interpersonal racism is more likely to result in long-term somatic health problems than one-off or acute experiences. Among African Americans, persistent, repeated occurrences of everyday, negative, racially based treatment is more strongly related to poor physical health than obvious experiences of discrimination.21

Racism research in Australia

Australian research on racism has focused on the attitudes and behaviours of the perpetrators rather than the effects on those who are the targets of discrimination.25 It has found that racist attitudes and behaviour are relatively common. In Western Australia, 52% of urban residents and 69% of residents of a regional centre revealed prejudice against Aboriginal Australians.26 A rare study focusing on Aboriginal people's experience of racism found racist attitudes and behaviour were ubiquitous.25 An unpublished survey of Indigenous people in Darwin uncovered that the experience of interpersonal racism was common across many settings and that these experiences were associated with a range of mental and physical health indicators.27

In this article, we investigate if the experience of interpersonal racism has a measurable effect on the health of Aboriginal Australians. We hypothesise that Aboriginal people experience interpersonal racism more frequently than non-Aboriginal people and that the poorer health of Aboriginal people is positively associated with those self-reported experiences. Further, we hypothesise that interpersonal racial discrimination has a cumulative effect that may explain some of the health differences between Aboriginal and non-Aboriginal people.

Methods

Sample

The data come from a survey conducted in late 2003 in an isolated rural Australian town. The University of Western Australia and the Western Australian Aboriginal Health Information and Ethics Committee granted ethics approval. At the 2001 Census, the town had almost 6,000 usual residents, with about 1,000 identifying as Aboriginal. Compared with the non-Aboriginal population, Aboriginal households had lower weekly median household income ($500-$599 compared with $700-$799) and larger mean household size (3.4 compared with 2.6). The Aboriginal unemployment rate was 17% compared with 5% for non-Aboriginal people.

Eligible respondents were over 18 years old and had lived in the town for at least 12 months. Using a spatial database of property street addresses, 57 blocks of 25 adjacent addresses were randomly selected. The probability of selecting blocks was twice as high in the four Census Collection Districts (CCDs) where at least 25% of usual residents were Aboriginal. This was to allow meaningful statistical analysis of Aboriginal respondents.

Trained and supervised local interviewers were assigned to each block. They initially mapped all addresses in the selected blocks to identify occupied residential dwellings. Then each occupied dwelling was approached to determine if there were residents who met the selection criteria. If there were Aboriginal people living in the household, all eligible residents were invited to participate. If there were no Aboriginal people living in the household, only the adult in the house whose birthday was due next was asked to complete a questionnaire.

Variables

Several strategies have been used to measure the experience of interpersonal racism in surveys.13,28 Unlike other measures of racist behaviour in the literature,21,22 we explicitly asked about treatment the respondent considered to be racially based. Our questions did not specify the type of behaviour or the context of that behaviour.20 Specifically, we used two questions from the 2002 Behavioral Risk Factor Surveillance System of the Centers for Disease Control and Prevention (HTTP://www.cdc.gov/brfss/about.htm):

  • 1“Within the past four weeks, have you felt emotionally upset as a result of how you were treated because of your race (for example angry, sad, or frustrated)?”
  • 2“Within the past four weeks, have you experienced any physical stress or symptoms as a result of how you were treated because of your race (for example a headache, an upset stomach, tensing of your muscles, or a pounding heart)?”

These questions required respondents to make their own judgement about whether the behaviour or attitude was racially based and did not limit the nature or context of the treatment. We controlled for the severity of the perceived negative treatment by asking only about those that invoked a particular emotional or physical response. Each question was answered by a simple yes or no and referred to a period of four weeks to minimise recall bias. In the regression models, these two variables were combined into a single measure of any experience of negative treatment.

Mental and physical health were measured using the physical component summary score and the mental component summary score of the Short-Form 12 (SF-12), an internationally validated scale.29 Each score ranged from 0 to 100, with a standardised mean of 50 and a standard deviation of 10. Higher scores indicate better health. General health status was another dependent variable, using the single question on general health status that forms part of the SF-12, dichotomised as 0=excellent, very good or good health and 1=fair or poor health. This measure has been used in other studies of the health impact of self-reported racial discrimination.3,21,23 Global self-assessed health questions have been found to be valid measures of health for Aboriginal Australians whose main language is English.30

As in other research, socio-economic variables were used as indicators of the effects of historical and current institutional racism.13 The socio-economic variables were individuals’ level of education and employment. Preliminary analysis found that these variables had a positive and statistically significant relationship with physical and mental health for the total sample. To ensure adequate cell sizes, the socio-economic variables were collapsed into two or three categories. An ‘other’ category for employment included home duties, studying, unemployed, retired and unable to work.

Analysis

We report the distribution of the main variables for Aboriginal and non-Aboriginal respondents and logistic regressions to show the association of age, sex, education and employment with self-reported experience of racially based treatment. These were calculated for all respondents and separately for Aboriginal and non-Aboriginal respondents.

Linear regression models were estimated using the SF-12 scores as dependent variables to test if the experience of racially based treatment and socio-economic status explained poorer health status of Aboriginal respondents. Adjusting for age and sex, the association of socio-economic variables on physical and mental health scores was examined first and then the racial discrimination variables were added. This was done separately for Aboriginal and non-Aboriginal respondents. The extent that institutional and interpersonal racism explained the differences in health between Aboriginal and non-Aboriginal people was tested in a model with all respondents. The analysis was repeated using logistic regression to predict self-reported fair-to-poor general health.

We tested for interaction between Aboriginality and experience of negative racially based treatment in all regression models using the total sample. The terms were not significant and were not included in the final models.

All of the statistical results have been produced using the survey methods features of Stata 9, statistical software that enabled the specification of complex survey designs. Confidence intervals were calculated using the Taylor series linearisation procedure, taking into account the clustering into residential blocks, stratification and different sampling proportions.

Results

In total, 639 residents completed questionnaires, representing a response rate of 67% of households and 75% of individuals. This included 183 people identifying as Aboriginal. Less than 2% of the non-Aboriginal respondents referred to themselves as Asian-Australian or a specific Asian ancestry. The remainder described their ancestry as Australian (65 respondents), English (89 respondents), Scottish (11 respondents), New Zealand (10 respondents), and other European backgrounds. Aboriginal respondents were younger, much less likely to have more than Year 10 education and less likely to be employed full-time (see Table 1). Recent experience of racially based treatment resulting in an emotional upset was reported 2.6 times (p<0.01) as often by Aboriginal respondents and treatment that was reported to result in physical symptoms of stress was reported 5.0 times (p<0.001) more often. Aboriginal respondents were more than twice as likely to describe their general health as fair or poor (34% compared with 15%, p<0.01). Aboriginal respondents also had significantly lower mental and physical health scores than non-Aboriginal respondents.

Table 1.  Characteristics of Aboriginal and non-Aboriginal respondents.
 Total (n=624) %Aboriginal (n=183) %Non-Aboriginal (n=441) %Designed based F-ratio (p value)
Demographic variables
Age   4.2 (0.02)
 18-3427.536.824.9 
 35-5445.045.244.7 
 55+27.518.030.4 
Sex   2.4 (0.12)
 Male41.436.142.5 
Female58.663.957.5 
Socio-economic variables
Employment   8.0 (<0.01)
 Full-time40.827.244.6 
Part-time/casual/work without pay23.522.224.0 
Other35.750.631.3 
Highest qualification   48.9 (<0.01)
 Year 10 or less49.276.641.2 
 Year 12/Trade/TAFE/Uni50.823.458.8 
Interpersonal racism variables
Emotionally upset   30.5 (<0.01)
 Yes18.735.413.3 
 No81.364.686.7 
Physically upset   87.0 (<0.01)
 Yes13.635.87.1 
 No86.465.292.9 
Either emotional or physical upset   42.2 (<0.01)
 Yes21.742.115.2 
 No78.357.984.8 
Health variables
Self-reported general health   25.6 (<0.01)
 Excellent, Very good, Good82.665.685.1 
 Fair, Poor17.434.414.9 
 Mean (Linearised 95% CI)Mean (Linearised 95% CI)Mean (Linearised (Linearised 95% CI) 
Mental health48.743.050.4 
Component(47.7-49.7)(40.9-45.2)(49.4-51.3) 
Score    
Physical health49.146.250.1 
Component(48.1-50.1)(44.5-48.0)(48.9-51.1) 
Score    
Note:
Statistical tests compared distribution of characteristics between Aboriginal and non-Aboriginal respondents.

After controlling for other variables, the odds of Aboriginal people reporting racially based negative treatment were 3.6 times greater than the odds for non-Aboriginal people (see Table 2). None of the demographic or socio-economic variables were significantly related to experiencing negative treatment.

Table 2.  Logistic regression models of adjusted odds ratios (linearised 95% confidence intervals) of self-reporting negative racially based treatment by demographic and socio-economic characteristics and Aboriginality. Odds ratios in bold are significantly different from 1.0.a
 All respondentsAboriginal respondentsNon-Aboriginal respondents
  1. Note:

  2. (a) Odds ratios adjusted for sampling design.

Aboriginality   
 Non-Aboriginal   
 (reference group)   
 Aboriginal3.6 (2.3-5.8)  
Sex   
 Female   
 (reference group)   
 Male1.2 (0.8-2.0)1.1 (0.6-2.1)1.3 (0.7-2.5)
Age   
18-341.5 (0.9-2.5)0.9 (0.4-2.0)2.0 (1.0-3.9)
35-54   
 (reference group)   
 55+0.6 (0.3-1.1)0.4 (0.1-1.1)0.7 (0.3-1.6)
Employment   
 Not in workforce   
(reference group)   
Full-time0.7 (0.4-1.2)0.5 (0.2-1.4)0.7 (0.4-1.4)
Part-time/casual/work without pay1.1 (0.6-1.9)1.1 (0.4-2.8)1.0 (0.5-2.3)
Highest qualification   
 Year 10 or less   
(reference group)   
Year 12/Trade/TAFE/Uni1.0 (0.5-1.7)1.3 (0.6-3.0)0.9 (0.4-1.8)

Aboriginal people had significantly lower self-reported physical health component scores than non-Aboriginal people after controlling for demographic and socio-economic variables (see Table 3). There was a strong association between full and part-time employment and good physical health for the total sample and for Aboriginal and non-Aboriginal respondents separately. For Aboriginal people, negative racially based treatment was significantly associated with poor physical health. Those who reported negative treatment had, on average, a physical health component score of only 42.6. The measures of interpersonal racism were not significantly associated with physical health for non-Aboriginal respondents or the total sample.

Table 3.  Unstandardised linear regression coefficients (95% confidence intervals) predicting physical health scores for Aboriginal and non-Aboriginal people separately and total sample (Model 1 includes demographic and socio-economic variables, Model 2 adds racial discrimination variables). Coefficients in bold are significantly different than 0.a
 Aboriginal respondentsNon-Aboriginal respondentsAll respondents
Independent variablesModel 1Model 2Model 1Model 2Model 1Model 2
  1. Note:

  2. (a) Coefficients are adjusted for sample design, age, sex, and all variables shown in the model.

Aboriginality      
 Non-Aboriginal (reference group)    -3.1-2.8
 Aboriginal    (-5.1 – -1.2)(-4.7 – -0.8)
Employment
 Not in workforce (reference group)
 Full-time8.98.43.93.84.94.7
 (5.3 – 12.4)(5.2 – 11.7)(1.1 – 6.8)(0.9 – 6.7)(2.4 – 7.4)(2.2 – 7.3)
 Part-time3.53.82.72.63.02.9
 (-0.1 – 7.1)(0.2 – 7.4)(0.2 – 5.3)(-0.1 – 5.2)(0.8 – 5.2)(0.6 – 5.2)
Education
  Year 10 or less (reference group)
Year 12/Trade/TAFE/Uni-3.33.22.12.11.21.2
 (-7.5 – 0.9)(-7.3 – 1.0)(-0.3 – 4.6)(-0.4 – 4.7)(-1.2 – 3.7)(-1.2 – 3.7)
Self-reported negative racially based treatment   
 No (reference group)
Yes -3.6 -0.01 -1.4
  (-6.4 – -0.7) (-3.4 – 3.4) (-3.8 – 0.9)
R20.200.230.170.160.170.17

Compared with the physical health measure, the mental health component scores have a stronger association with the experience of racially based negative treatment (see Table 4). After introducing the racially based treatment variables, there was a small decline in the gap between predicted mental health scores for Aboriginal and non-Aboriginal people. There was no significant association between education and mental health scores and the association with full-time employment was inconsistent.

Table 4.  Unstandardised linear regression coefficients (95% confidence intervals) predicting mental health scores for Aboriginal and non-Aboriginal people separately and total sample (Model 1 includes demographic and socio-economic variables, Model 2 adds racial discrimination variables). Coefficients in bold are significantly different than 0.a
 Aboriginal respondentsNon-Aboriginal respondentsAll respondents
Independent variablesModel 1Model 2Model 1Model 2Model 1Model 2
  1. Note:

  2. (a) Coefficients are adjusted for sample design, age, sex and all variables shown in the model.

Aboriginality      
 Non-Aboriginal (reference group)      
 Aboriginal    -5.5-4.0
     (-8.1 – -3.0)(-6.5 – -1.6)
Employment
 Not in workforce (reference group)
 Full-time7.46.31.30.42.31.4
 (3.8 – 10.9)(3.0 – 9.6)(-1.0 – 3.5)(-1.8 – 2.7)(0.3 – 4.2)(-0.4 – 3.2)
 Part-time0.51.31.00.10.90.3
 (-4.5 – 5.5)(-3.0 – 5.5)(-1.3 – 3.2)(-2.1 – 2.4)(-1.3 – 3.1)(-1.7 – 2.4)
Education
 Year 10 or less (reference group)
Year 12/Trade/TAFE/Uni-4.2-3.81.81.80.80.9
 (-8.6 – 0.2)(-7.8 – 0.1)(-0.2 – 3.8)(-0.9 – 3.8)(-1.1 – 2.8)(-0.8 – 2.7)
Self-reported negative racially based treatment   
 No (reference group)
Yes -9.2 -4.5 -6.3
  (-13.3 – -5.1) (-8.0 – -0.9) (-9.1 – -3.6)
R20.120.250.040.070.100.16

In Table 5, we used logistic regression models to predict self-reported poor general health. As with the other measures of health, the experience of racially based negative treatment was significantly associated with self-reported fair or poor health among Aboriginal people and the total sample. In the total sample, Aboriginal people still had more than two times the odds of reporting poor health, after controlling for demographic and socio-economic variables. When the racially based treatment variable was included, the disparity between Aboriginal and non-Aboriginal people's likelihood of reporting fair or poor health declined slightly but remained statistically significant. Among both Aboriginal and non-Aboriginal people, full-time employment was protective against poor health.

Table 5.  Logistic regression models of adjusted odds ratios (95% confidence intervals) predicting self-reported general health status as fair or poor by socio-economic and interpersonal racism variables. Odds ratios in bold are significantly different from 1.0.a
 Aboriginal respondentsNon-Aboriginal respondentsAll respondents
Independent variablesModel 1Model 2Model 1Model 2Model 1Model 2
  1. Note:

  2. (a) Odds ratios are adjusted for sample design, age, sex, sampling strata and all variables shown in the model.

Aboriginality      
 Non-Aboriginal (reference group)      
 Aboriginal    2.62.1
     (-1.6 – 4.4)(1.2 – 3.5)
Employment
 Not in workforce (reference group)
 Full-time0.10.10.20.20.20.2
 (0.04 – 0.5)(0.04 – 0.5)(0.1 – 0.5)(0.1 – 0.5)(0.1 – 0.4)(0.1 – 0.4)
 Part-time0.50.50.60.60.50.5
 (0.2 – 1.1)(0.2 – 1.1)(0.3 – 1.1)(0.3 – 1.1)(0.3 – 0.9)(0.3 – 0.9)
Education
 Year 10 or less (reference group)
Year 12/Trade/TAFE/Uni1.31.20.50.50.70.6
 (0.6 – 2.9)(0.5 – 2.6)(0.3 – 1.1)(0.3 – 1.0)(0.4 – 1.2)(0.4 – 1.1)
Self-reported negative racially based treatment   
 No (reference group)
Yes 3.2 2.1 2.5
  (1.3 – 8.2) (0.9 – 5.2) (1.3 – 5.0)

Discussion

As we hypothesised, Aboriginal Australians were significantly more likely than non-Aboriginal people to report that they had been physically or emotionally upset by negative racially based treatment in the last four weeks. In fact, recent experience of interpersonal racial discrimination was so common that more than 40% reported it. Comparable studies have shown that African Americans report higher levels of perceived discrimination than whites2,23 and that Maori peoples report higher levels than Asians, Pacific peoples or Europeans in New Zealand.4

None of the potentially explanatory demographic or socio-economic variables were significantly associated with Aboriginal people's reported negative racially based treatment. The lack of difference in Aboriginal respondents’ experience of racially based treatment by sex, education or employment status is consistent with findings for African Americans and suggests that all Aboriginal people equally experience perceived racism as part of daily life.2,20,22

In this study, as in so many other studies of Aboriginal health, Aboriginal respondents reported poorer physical and mental health. They were more than twice as likely to describe their general health as fair or poor and, after controlling for age, sex, employment and education, Aboriginal respondents still had significantly lower physical and mental health scores.

Our study supports the hypothesis that the experience of interpersonal racism is associated with poorer health. Our finding that the mental health score was sensitive to acute experiences of racially based treatment replicates numerous studies in the United States that also showed a strong relationship between the experience of racial discrimination and measures of mental health such as psychological and psychiatric stress.19 This association was significant for Aboriginal and non-Aboriginal respondents as well as the total sample.

The research cited in the introduction, which has been able to demonstrate a relationship between physical health and self-reported experience of racial discrimination, used measures of lifetime exposure to racially based treatment. Other studies have failed to find an independent relationship with physical health because, like ours, their measure of interpersonal racism captured recent acute events and not chronic exposure.20 However, we did find that the physical health component score was significantly associated with the interpersonal racism measures for Aboriginal respondents. The more holistic measure of poor health status was also related to interpersonal racism for both Aboriginal respondents and the total sample.

Our study partly supports the hypothesis that the everyday experience of negative racially based treatment contributes to persistent health differences. These experiences clearly contributed to the poorer health of Aboriginal respondents. However, although adding the experience of racially based treatment weakened the association between Aboriginality and the health measures, in all cases the confidence intervals of the estimates overlapped.

Our socio-economic variables captured some consequences of institutional racism. The consistently lower levels of education and employment reported by Aboriginal respondents were undoubtedly a consequence of current and past discriminatory policies. Full-time employment emerged as a significant predictor of better health for Aboriginal people and this effect was not diminished when experience of interpersonal racism was added to the model. This association needs to be viewed with caution as poor physical or mental health can be a reason for being out of the workforce. The apparently limited protective role of education needs to be replicated in other settings. It is important to note that controlling for socio-economic status does not ‘explain’ all differences in health. Narrowing the socio-economic gap will reduce but not eliminate health inequalities.

A strength of this study is that it included Aboriginal and non-Aboriginal people.21 Who were these non-Aboriginal people who experienced emotional upset or physical stress as a result of perceived negative treatment because of their race? We know from additional information collected in the survey that they were not ethnically or racially different from the other non-Aboriginal respondents. Nor was racial identity more salient for this group; these individuals were not more likely to report thinking regularly about their race. However, they were much more likely than other non-Aboriginal respondents to say that Aboriginal people could not be trusted. Our regression analysis did not reveal any demographic or socio-economic associations between non-Aboriginal respondents’ experiences of negative racially based treatment, but further research could focus on those individuals who reacted physically to such perceived treatment. It is our tentative conclusion that although some non-Aboriginal people also experience negative treatment as part of living in a racially constructed society, unlike Aboriginal people their physical health does not suffer as a result of that behaviour. On the other hand, non-Aboriginal people experiencing such treatment have poorer mental health than other non-Aboriginal people.

Limitations

Despite the careful random selection of property addresses and the personal contact between interviewers and respondents, the final sample was not representative of the 2001 Census description of the statistical unit of which the town comprised approximately 90%. The sample had a higher proportion of females than enumerated as usual residents in the Census (53% of the Aboriginal population in the Census compared with 64% in the sample and 46% of the non-Aboriginal population in the Census compared with 58% in the sample). Aboriginal and non-Aboriginal respondents in the middle-age bracket (35–54 years) were over-represented. However, the proportion unemployed was almost identical in the Census and the sample. Controlling for demographic and socio-economic factors meant that the relationships explored were unlikely to be affected.

A significant limitation is the measure of interpersonal racism we used. In addition to the short reference period of four weeks, both variables used the respondents’ response to an event as the indicator. On one level, it is not surprising that people who had an emotionally upsetting or physically stressful recent experience have poorer self-reported health. We argue that our results are still valid despite this apparent tautology. First, this paper has been able to demonstrate that more than 40% of Aboriginal people in our study reported treatment in the recent past that was so severe as to produce a strong emotional or physical response. This is an important finding in its own right. Second, the associations we found between our measures of racist treatment and health are very similar to those of other studies that used different measures of interpersonal racism such as the self-reported experience of specific behaviours without reference to a response to that behaviour.19 A more general limitation is that cross-sectional associations are inevitably weaker than measures of effects over time; research on the effects of interpersonal racism over the life course is needed.

Unfortunately, there is no consensus on how to measure the experience of racial discrimination. Krieger, a leader in this field, calls for increased effort to develop valid and reliable self-reported measures.31 Our findings also highlight the need for more research on multidimensional chronic and acute measures of the experience of racial discrimination that are valid for Aboriginal Australians. The recent development of the Measure of Indigenous Racism Experiences items derived in part from qualitative research and piloted in a survey of Indigenous people in Darwin is a much-needed contribution.27

The obvious policy implication of our study is that the behaviour of non-Aboriginal people needs to change. While there are some interventions that have been proven to change racist attitudes of members of dominant groups, there is little or no research on changing behaviour that is perceived by the subordinate group as racist.32 Efforts in this area should be supported by employing effective policies in mainstream institutions such as schools and health services.33 Further work is also needed to understand the pathways by which internalised racism affects health and to identify factors that protect or exacerbate the effects of racial discrimination. Such research would explore the role of a positive Aboriginal identity,34 and supportive or destructive social and cultural Aboriginal networks.35

Conclusion

Racial discrimination needs to be recognised as an upstream determinant of health.23 Increasing our understanding of the ways that Aboriginal people experience racism and the pathways through which those experiences have an impact on health is essential if there are to be any lasting improvements. Without fundamental changes in how members of the dominant Australian culture behave towards Aboriginal people, initiatives to improve health services, educational and employment opportunities may have limited impact on health inequalities.

Acknowledgements

The study was funded by Healthway and by the Department of Health and Ageing through their support of Combined Universities Centre for Rural Health. The funders had no role in the collection, analysis, and interpretation of data or in the writing or the decision to submit the paper for publication. Data management and statistical assistance was provided by Jessica Scott and Belynda Wheatland. The authors thank the interviewers and local stakeholders for their support.

Ancillary