Navigation – Plan du site

AccueilNuméros3Sketch of a research program on t...

Sketch of a research program on the contribution of discrimination to mental health inequalities: a critical review of evidence, models and methods

Martin Aranguren


Dans les deux premières décennies du 21e siècle des centaines de travaux ont documenté une corrélation négative entre discrimination et santé mentale. Toutefois, des critiques ont pointé que cette association opère au sein de groupes sociaux donnés, laissant de côté la question de savoir si les discriminations engendrent des inégalités de santé mentale entre les groupes. L’article esquisse quelques éléments d’un programme de recherche multi-méthode sur le rôle des discriminations dans la production d’inégalités de santé mentale entre les groupes. L’analyse procède en quatre étapes. Premièrement, une revue critique est offerte de la base empirique disponible sur la corrélation entre discrimination perçue et détresse émotionnelle. Deuxièmement, le lien entre discrimination et détresse est inséré dans un modèle plus englobant ayant pour but d’expliquer les inégalités sociales de santé mentale. Troisièmement, une stratégie empirique multi-méthode visant à intégrer des déclarations rétrospectives de discrimination vécue avec des mesures objectives de traitement discriminatoire est proposée. Quatrièmement, une illustration est offerte de cette stratégie empirique à partir de deux recherches récemment achevées sur les cas des musulmans et des noirs en Europe.

Haut de page

Texte intégral

I thank Philippe Coulangeon, Emanuele Ferragina and Jocelyne Streiff-Fénart, as well as the other contributors to this special issue, for their comments on earlier versions of the manuscript.

1In the first two decades of the 21st century literally hundreds of studies were published on the link between discrimination and mental health. This body of research has been reviewed in a number of meta-analyses that differ in coverage and emphasis but converge with no exception in the estimate of an overall negative correlation between perceived discrimination and mental health (Britt-Spells et al. 2018, Carter et al. 2017, 2019, de Freitas et al. 2018, Jones et al. 2016, Lee and Ahn 2011, 2012, 2013, Paradies et al. 2015, Pascoe and Smart Richman 2009, Pieterse et al. 2012, Schmitt et al. 2014, Triana, Jayasinghe and Pieper 2015).

2While this result seems beyond doubt, critics have observed that it applies within groups, leaving unaddressed the question whether discrimination produces inequalities in mental health between groups (Schwartz and Meyer 2010, see also Williams et al. 2019). For example, a typical result has been to show that among Black Americans a higher score on a perceived discrimination scale is correlated with a lower score on a mental well-being scale (or with more symptoms of a specific psychopathology, or of nonspecific emotional distress). But this is not equivalent to demonstrating that Black Americans, on average, are more distressed than White Americans, and that the gap between these groups is (at least partly) imputable to the fact that Black Americans experience more discrimination than White Americans.

3The present article is a first attempt to sketch some elements of a multi-method research program concerned with the role of discrimination in the production of mental health inequalities between groups. The article has four aims: 1) to offer a critical review of the available evidence on the correlation between perceived discrimination and emotional distress, 2) to embed the connection between discrimination and distress within a more encompassing model for explaining social inequalities in mental health, 3) to propose a multi-method empirical strategy for integrating retrospective reports of lived discrimination with objective measures of discriminatory treatment, and 4) to illustrate this multi-method strategy drawing on recently completed research.

  • 1 In non-experimental research, a data production scheme (e.g. administering a questionnaire to, or c (...)

4The article is organized in four corresponding sections. Drawing on meta-analytical publications, the first section reviews the studies that have investigated the correlation between perceived discrimination and emotional distress. This literature is largely overrepresented by samples that are cross-sectional in design,1 from the US and in particular about Black Americans, and statistically nonrepresentative of these or other populations. These limitations notwithstanding, this corpus provides remarkably robust evidence of a nonsmall correlation between perceptions of discrimination and reports of emotional distress, making it unlikely that the association is artefactual or inconsequential.

5The second section of the article attempts an explanatory conceptualization of this robust and nonsmall correlation, with a view to explaining inequalities in emotional distress between social groups by treating discrimination as an intervening or mediating variable. To this purpose, I review two bodies of research. First, seeing discrimination as a stressor in the social environment, I draw on life events research (Hammen 2005, Mazure 1998, Paykel 2001) to identify basic templates for explaining the environmental causation of mental disorder. Second, I embed this explanation within the “social stress” model (Dohrenwend 2000, Pearlin 1989), and the “minority stress” model in particular (Meyer 2003), regarded as a two-level explanation in which “macro-social” disparities in mental health between groups result from the aggregation of “micro-social” individual confrontations with stressors, and more specifically with discrimination.

6The third section is concerned with issues of method. Research carried out within the life events and the social stress traditions usually relies on retrospective self-reports for documenting stressors, whether by use of checklists in representative surveys or of more accurate semi-structured interviews but with smaller samples. The dependence on retrospective reports raises a number of consensually acknowledged problems of reliability and validity (reviewed in Dohrenwend 2006). Some of these problems can be effectively addressed by experimentally manipulating minor stressors, but to the best of my knowledge this possibility has not been put in practice. The multi-method strategy that I propose consists in experimentally testing the prevalence of one particular stressor, namely discrimination in social interactions, and in using survey data providing mental health measures to assess the impact of that stressor on the average mental health of different social groups.

7Drawing on two recently completed studies dealing with European minorities, the fourth and last section seeks to demonstrate the heuristic value of this novel empirical strategy. As theoretically expected, these studies indicate that Muslims (Aranguren 2022a), on the one hand, and Blacks (Aranguren 2022b), on the other, are more depressed than average partly because they are more exposed to discrimination.

1. Critical review of studies on the correlation between perceived discrimination and emotional distress

Review of meta-analyses

8In order to assess the evidence for the correlation between perceived discrimination and mental health, I reviewed all the meta-analyses published in scientific journals up to March 2022. Table 1 lists them in chronological order, specifying the journal title, the target population(s), the date range of publication of the meta-analyzed individual studies, the number of individual studies that included measures of mental health, and the average strength of association between perceived discrimination and mental health across individual studies, expressed as a 95% confidence interval for the estimated Pearson’s correlation coefficient r. The latter coefficient appears on the table as it was originally reported in the meta-analytical publication. Although its sign is positive for some publications and negative for others, the conceptual meaning is consistent. More perceived discrimination correlates with higher distress (positive coefficient) or with poorer mental health (negative coefficient). The conceptual meaning can be considered to be consistent because experiencing more distress is in many respects similar to seeing one’s mental health deteriorate.

Table 1. Meta-analyses on the correlation between perceived discrimination and mental health

Table 1. Meta-analyses on the correlation between perceived discrimination and mental health

9Meta-analyses on this correlation appeared for almost ten years at a roughly constant rate, starting in 2009 and coming to a temporary halt by 2018. Most of these meta-analytical publications have found an outlet in journals of psychology. Whenever this information has been provided, most of the original studies included in the meta-analysis seem to have been published during the 2000s. Samples from the US and composed of African American respondents in particular largely dominate the pool of individual studies. Paradies and colleagues (2015) estimate at 80% the share of American samples, while the uniquely sustained focus on Black Americans can be sensed from inspecting the description of the target populations on Table 1. Further, almost 70% of the individual samples are non-representative. In 70% of the cases the sample size is below one thousand. Nearly 90% of the individual samples contribute cross-sectional data (Paradies et al. 2015). The number of studies included in the meta-analysis varies substantially from one publication to the other, with a minimum of 14 and a maximum of 328. This number depends on the scope of the analysis, with country, target group and setting (e.g. the workplace) acting as the main dimensions of focus breadth.

10At least eight different dedicated instruments have been used for measuring perceived discrimination. Together, these dedicated instruments regard nearly 40% of the samples, with no single instrument covering more than 10% of the total. A pattern in these scales is that questions are usually about mundane interpersonal events, their occurrence (yes/no) or frequency (from never to very often), and their likely cause or motive (e.g. the respondent’s race or ethnicity). For example, “How often in the last year were you treated rudely or disrespectfully because you are Black?” (Daily Life Experiences subscale of the Racism and Life Experiences Scale, Harrell 1997), “How many times in the past year/in your entire life have you been treated unfairly by strangers because you are Black”? (Schedule of Racist Events, Landrine and Klonoff 1996), or “How many times have you experienced discrimination on the street, been prevented from doing something, or been hassled or made feel inferior in a public setting because of your race, ethnicity or color?” (Experiences of Discrimination Scale, Krieger et al. 2005).

11As for mental health, the most frequent measures appear to be depression (reported in 37.2% of articles), followed by self-esteem (24.3%), psychological stress (21.3%), distress stricto sensu (18.3%) and anxiety (14.4%). The most elaborate measures of mental health rely on a battery of questions such as the Center of Epidemiological Studies-Depression scale (Radloff 1977). The latter includes questions such as “Please tell me how much of the time during the past week you felt depressed/felt that everything you did was an effort/you sleep was restless/you were happy/you felt lonely/you enjoyed life/you felt sad/you could not get going?”. The response levels refer to varying frequencies, for instance “None or almost none of the time”, “Some of the time”, “Most of the time”, “All or almost all of the time”.

12In spite of considerable diversity on these and yet other dimensions, the estimated association across meta-analyses is remarkably consistent in sign and size. Perceived discrimination correlates with poorer mental health, or higher emotional distress, roughly at = .2. By the standards of psychological research (Bosco et al. 2015), this association appears to be moderate in size (that is, non small). As an anchor of effect size, it is useful to mention the much smaller correlation between perceptions of discrimination and physical health, estimated at r = -.09 (Paradies et al. 2015). The most commonly used indicators of physical health are blood pressure, hypertension and overweight.

A robust discrimination-distress association but a problematic correlation between discrimination and self-esteem. How come?

13It is instructive to contrast the highly consistent pattern of findings concerning the discrimination-distress association with the more controversial correlation between perceiving discrimination against the self and having a lower self-esteem. Crocker and Major (1989) reviewed a vast literature on the subject concluding that the available evidence did not sufficiently support the theoretical expectation that felt discrimination should lower self-esteem. To restore intelligibility, they proposed a set of “strategies” with which members of stigmatized groups respond to perceived discrimination in an effort to protect or enhance the sense of self-worth. One of these protective strategies consists in attributing negative outcomes (e.g. not being taken for a job or promotion, meeting an unfriendly stranger, being overly sanctioned for a minor transgression) to the prejudiced views that others hold against the individual’s stigmatized group.

14The authors further suggested that one moderator predicting the stigmatized individual’s inclination to impute negative feedback to others’ prejudice is the centrality of membership in the stigmatized group to the individual’s self-concept –e.g. the centrality of being Black for a Black person’s self-image. In other words, the more identification with the stigmatized group is salient to the individuals’ definition of themselves, the higher the probability of having resort to self-protective strategies such as attributions to prejudice. Countering this claim (but for the rest in overall agreement Crocker and Major’s position), Branscombe et al. (1999) proposed an alternative scheme, known as the rejection-identification model, in which the salience of group membership for the individual’s self-definition increases as a result of perceptions of discrimination, and not the other way around (see Ramos et al. 2012 for supporting evidence based on a longitudinal survey).

15Beyond this unresolved dispute on the direction of causation, at first sight self-esteem and mental well-being (or its reverse, emotional distress) seem to be closely related constructs. Why, then, a robust discrimination-distress association but mixed support for the correlation between discrimination and self-esteem? Crocker and Major (1989) offer a plausible answer to the question, as they consider potentially damaging collateral effects of using the self-protective mechanisms. They point out that the tendency to attribute negative outcomes to prejudice may lead to “universal helplessness”, as the individual grows accustomed to accounting for their own predicament in terms of an enduring, pervasive and uncontrollable cause. The authors observe that this global feeling of being helpless in the face of external adversity appears to be associated with the motivational and cognitive deficits of depression but, interestingly enough, not with self-esteem. If this reasoning is correct, it would imply that the coping strategies used for buffering the damaging effects of perceived discrimination on self-esteem may not be effective in protecting the stigmatized individual from depression and emotional distress more generally. What is more, by colaterally inducing a permanent sense of helplessness, they may even exacerbate the impact of discrimination on mental health.

Some limitations of this literature

16The main strength of this corpus is to consistently document a moderate association between perceived discrimination and degraded mental health. The limitations concern aspects of empirical research, on the one hand, and the type of question that this research is capable of answering, on the other.

17The first empirical limitation is the unknown generality of the association. Does perceived discrimination and emotional distress correlate moderately outside of the US? Do all stigmatized groups around the world respond as Black Americans do? Although the surveyed meta-analyses point to positive answers, the evidence from non-American samples or concerned with groups other than Black Americans remains too scarce to warrant any firm conclusions.

18The second empirical limitation is actually a set, and covers problems of measurement validity. The first of these has to do with the relationship between discrimination as a lived experience vs. an objective fact. Do retrospective reports of perceived discrimination offer a good indicator of past occurrences of discrimination? Accurately reporting objective past events of discrimination against the self depends at least on three conditions: 1) at the time of perceiving, the perception must correspond to fact; 2) at the time of responding to the researcher, the subject must be able to accurately recall this accurate perception, and 3) again at the time of responding, the subject must be willing to share the accurately recalled accurate perception with the researcher. These three conditions are difficult to satisfy, and the degree to which they are met is unfortunately unknown.

  • 2 “Confounding” refers here to a problem of measurement. A measure of perceived discrimination may be (...)

19The second measurement issue points to the problem of confounding.2 Empirically, no temporal ordering is implied in the correlation between perceived discrimination and emotional distress. However, in most cases researchers interpret this co-occurrence as a sequence, placing perceived discrimination first, as the cause, and emotional distress second, as the effect. This interpretation is apparent in the common practice, that I have so far reproduced, of consistently naming first perceived discrimination and second emotional distress, suggesting that the order of words in syntax mimics the order of things in the world.

20To correctly interpret a correlation between two measures in causal terms, both measures must be mutually independent. However, there are strong indications (reviewed in Dohrenwend 2006) that retrospective reports of perceived discrimination and ratings of the present state of mental health may be confounded. The emotional mood at the time of responding may influence the retrospective report of perceived discrimination, reversing the direction of causation that is most commonly assumed. Perceived discrimination, that is, may as well be a symptom of emotional distress as an antecedent cause of it. The cross-sectional nature of most of the data in this domain does not allow to settle for one interpretation or the other.

21Beyond problems of measurement, critical reviews have emphasized that this research corpus is inadequate to one of the central goals of a sociology of mental health, namely that of accounting for the unequal distribution of emotional distress across population strata (Schwartz and Meyer 2010, see also Williams et al. 2019). The critics have emphasized that the moderate correlation holds only within stigmatized groups, observing that evidence of this within-group correlation is of no help for explaining between-group disparities in mental health as a result of unequal exposure to discrimination. To illustrate: that perceived discrimination is associated with depression among Black Americans is no proof that Black Americans are more depressed than average and that this surplus depression results from greater exposure to discrimination. As a matter of fact, previous studies do not generally support a mental health disadvantage for Black Americans compared to White Americans (cf. Schwartz 2002).

22To recapitulate: between the 1990s and 2020 hundreds of published studies have investigated the association between perceived discrimination and mental health. These studies rely mostly on cross-sectional data from unrepresentative samples in which the US and Black Americans are largely overrepresented. The reliability and accuracy of measures of perceived discrimination as an indicator of objective occurrences of discrimination are unknown. The cross-sectional nature of most of the data makes it impossible to exclude that measures of perceived discrimination are confounded with measures of emotional distress. These studies are well-suited to examining the correlation between perceived discrimination and degraded mental health within specific groups, but not to the more classically sociological task of uncovering inequalities between groups and imputing them to differential exposure to discrimination.

23Despite these limitations, this body of research has suceeded in establishing a robust correlation of moderate size between perceived discrimination and various measures of non-specific distress. The robustness of the correlation, as well as its size, makes it unlikely that the association is artefactual or inconsequential.

2. Interpreting the correlation between perceived discrimination and emotional distress: discrimination as an event responsible for mental health inequalities between groups

24The aim of the present section is to reinterpret the robust and nonsmall discrimination-distress association as part of a broader explanatory model. The latter model must be capable of describing a causal process where differences in social position (e.g. by socioeconomic status, gender, race) act as the predictor, differences in mental health between social positions as the outcome, and discrimination as the mediator or intervening variable. The envisaged conceptual construction proceeds in two steps.

25First, drawing on life events research (Hammen 2005, Mazure 1998, Paykel 2001) the correlation is reinterpreted causally, putting discrimination in the role of the cause and distress in that of the effect. The connection between the cause and the effect is ensured by the assumptions that stress in general causes distress, and that discrimination represents a particular source of stress or “stressor”.

26Second, drawing on research in the “social stress” paradigm (Dohrenwend 2000, Pearlin 1989), the stress-distress process is embedded within a more encompassing explanation meant to account for the impact of social position on mental health. The key assumption here is that differences in social position are reflected in differences in exposure to stressors. In more recent developments within this tradition, discrimination has been regarded as a stressor typical of minority position. The resulting “minority stress” hypothesis (Meyer 2003) is that individuals placed in minority positions (e.g. by race/origin or by sexual orientation) have poorer mental health than those who are not because their minority status overexposes them to stressors such as discrimination.

The environmental causation of mental illness: life events research

27The classical stress model originates in the work of the Vienna-born biologist Hans Selye (1936). Simplified to the extreme, the basic story can be told as an organism-environment transaction. The organism has a preference for homeostasis: it seeks equilibrium and dislikes disequilibrium. When an environmental event provokes disequilibrium in the organism, the outer occurrence is said to represent a “stressor”. The organism’s struggle to restablish equilibrium is known as the “stress” response or syndrom. In general, the syndrom (that is, stress) effectively prepares the organism to cope with the challenge; in such cases, the organism’s stress reaction is said to be “adaptive”. But if the required effort exceeds the organism’s resources, the stress response becomes “maladaptive” and the organism yields to “exhaustion.”

28Initiated in the 1960s, the strand known as “life events research” can be seen as an application of these concepts to the problem of the environmental (vs. constitutive or biological) causation of mental illness. Briefly stated, the main assumption is that life events in the social environment, in their capacity of stressors, provoke stress in the organism, leading in case of excessive demand to psychological exhaustion in the form of psychopathology (see Dohrenwend 1961 for an elaborate translation of Selye’s biological model into the social psychological concepts that are relevant to the problem of the environmental causation of mental illness). Life events denote discrete occurrences that are challenging for the average person, such as “death of spouse”, “divorce”, “personal injury or illness” or “fired at work” (Holmes and Rahe 1967). The magnitude of the stress that life events arouse has been conceptualized in terms of a variety of dimensions characterizing the event, among which disruptiveness, undesirability, and severity of threat. In turn, the disorders considered in the psychiatric literature on life events include depression, schizophrenia and suicide attempts (Paykel 2001). In sociological applications of this framework within the “social stress” tradition (of which more shall be said below), instances of deviant behavior (e.g. antisocial conduct, alcoholism or drug abuse) have also been treated as instances of the “maladaptive” response in need of explanation.

29Leaving other psychiatric illnesses temporarily aside, one important finding from this body of research is that individuals suffering from diagnosed depression are much more likely than their non-depressed counterparts to have experienced a severe adverse life event. Using the same instrument for the assessment of life events (namely the Life Events and Difficulties Schedule, Brown and Harris 1978), a series of community samples with women shows that depressed respondents were over 80% to have recently confronted a major life event, whereas the percentage was around 30% for the non-depressed (Mazure 1998).

30Four basic explanatory models have been used to account for the relationship between events and mental illness (Mazure 1998). These are stylizations and are surely too crude and simplistic to accurately describe most concrete applications. But they effectively capture some important contrasts.

31An early model, in disuse today, states that adverse mental health oucomes directly results from the accumulation of stressful events in a short span of time or from the impact of a single but fateful event. This model was quickly abandoned because it became apparent that most people who confront major life events actually do not get depressed (Hammen 2015).

32The acknowledgement of the latter fact, in combination with the continued desire to impute psychopathology to environmental events, provides the founding stone of the “stress-diathesis” model of mental illness that has oriented much of the psychiatric literature on the social causation of depression. The term “diathesis” designates vulnerability or risk factors that predispose the person in their possession to mental disorder. Simply put, the stress-diathesis hypothesis is that stressful occurrences eventuate in psychiatric disorder but only for those who possess the diathesis or prediposition.

33Over the years, a rich array of variables have been shown to moderate the influence of life events on depression, providing instances of diatheses (reviewed in Hammen 2018, Mazure 1998, Paykel 2001). These include demographic characteristics such as female gender, younger age, unmarried status, and poverty; biographic details such as medical disability, prior episodes of depression, a history of depression within the family, and early exposure to childhood adversity; personal dispositions such as coping style or traits such as sociotropy/dependence (especially when the adverse event involves loss of a valued interpersonal relationship) or neuroticism; and features of the person’s social environment such as lasting discord in the family (e.g. with spouse and/or children) or the availability of social support in case of difficulty, for example in the form of a confidant. The probability of stressful events eventuating in psychopathology is hypothesized to depend on the level of these moderating diatheses. In a regression analysis framework, this hypothesis leads to the specification of models with interaction terms.

34In the alternative “additive burden” model, the “diatheses” or vulnerability factors are reinterpreted as independent predictors of equal standing as life events, and mental illness is seen to result from the added influences of all these independent predictors. In fact, many of the previously mentioned diatheses can be regarded as stressors on their own right, for instance sustained medical disability or prolonged marital conflict. Although they were not originally named “stressors”, early research did recognize their demand character, as attested by expressions such as “ongoing difficulties” (Brown and Harris 1978) and “chronic strains” (Pearlin et al. 1981). The label “chronic stressors” was put in circulation in the 1980s in an attempt to unify the field of the phenomena that were increasingly felt to cause stress, and therefore qualified as stressors, yet did not conform to the definition of life events.

35Last, a birectional or transactional model posits that the presence of depression causes life events, which in turn perpetuate or intensify depressive symptoms or episodes. For example, a series of studies comparing individuals with and without histories of depression found that rates of independent events, in the sense of not possibly caused by the individual, did not differ between groups, whereas the rate of dependent events was higher for those with depression histories (reviewed in Hammen 2005). These dependent events, understood as occurrences to which individuals are likely to have contributed, often refer in these studies to ongoing interpersonal conflict within the family. That is, depressed individuals, once depression sets on, tend to create or exacerbate discord in intimate relationships, locking themselves within a vicious circle of stress and depression. More common in the sociology of mental health, the idea of “stress proliferation” (Pearlin 1989) is nevertheless very similar in spirit. It refers to the observation that stress and its distressing consequences affect not only individuals but also the role sets of which they are part. For example, a married person experiencing difficulties at work is likely to expose his or her spouse to stress as a consequence, propagating, as it were, stress from the incumbent of one role in the set to the other. But propagation from one individual partner to the other is not the only way in which stress comes to couples. Some stressors are better regarded as being directly “couple level”, for instance the fact of being in a socially marginalized relationship such as that between same-sex persons in contemporary Western culture (LeBlanc and Frost 2020).

36Which of these basic explanatory templates can we use to reinterpret the correlation between perceived discrimination and emotional distress? This association is usually regarded through the lens of the “additive burden” version of the stress-distress causal connection, most often as a matter of course and without consideration of the alternatives. In keeping with this established practice, I shall also embrace this model. But I have endeavored to offer a more panoramic overview of the theoretical landscape to clearly expose what this choice leaves out.

37In the “additive burden” approach discrimination is given the status of a stressor which, along with other stressors, is thought to make an independent contribution to the total stress that individuals experience; stress, in turn, is assumed to eventuate in distress. This results in a three-component process in which discrimination in the social environment causes stress in the person, which in turn causes distress in the same person. The emerging picture is that discrimination undermines mental health because it imposes to the discriminated individual an additional life difficulty to cope with, over and above more or less independent life events, persisting strains, and other everyday challenges. Figure 1 presents in graphical form this particular causal reinterpretation of the correlation between discrimination and distress.

Figure 1. The discrimination-distress association reinterpreted causally, with stress acting as the mediator

Figure 1. The discrimination-distress association reinterpreted causally, with stress acting as the mediator

The social stratification of stress exposure and distress: the “social stress” approach

38As said, in the “additive burden” model the demands independently posed by different stressors sum up to determine the total environmentally-induced stress that the person faces. This template underlies much of the work that has been conducted under the auspices of the “social stress” tradition in the sociology of health and illness (Dohrenwend 2000, Pearlin 1989). The “additive burden” model is in fact assumed in what has been presented as the sociological paradigm for the study of stress (Aneshensel, Rutter and Lachenbruch 1991).

39Drawing on the logic of mechanism-based explanations (Hedström and Ylikoski 2010), the “social stress” paradigm can be regarded as a two-level account. At the higher-order level of populational distributions, position in a social hierarchy (by occupation, by gender, by ethnicity, etc.) accounts for average differences in mental health between the occupants of those positions: the lower the status of a social location, the poorer the average mental health of its incumbents. But why should inequalities in social position eventuate in disparities in mental health between groups?

40The “additive burden” model provides the desired intervening process: at the analytically lower level of the individual, independent stressors sum up to determine the amount of stress to which individuals are exposed, with more stress leading to poorer mental health. The key assumption connecting distributions in the population (the “macro” level) with the stress that individuals experience (the “micro” level) is that lower status implies greater exposure to stressors. If stress exposure increases as social status decreases, and stress exposure (regardless of social status) undermines mental health, by simple aggregation it follows that, all else equal, the ordering of mental health averages across social positions should mirror the hierarchy of status. Figure 2 represents this two-level explanation in which status differences lead to a gradient in mental health at the population level via aggregation of the “additive burden” explanation at the individual level.

Figure 2. Two-level explanation of between-group disparities in mental health. The boxes placed above refer to the populational or “macro-social” level of analysis; those below to the individual or “micro-social” level. The “macro-social” antecedent and consequent are connected by a dashed arrow to indicate that the connection is still in need of clarification. The “micro” links describe an individual-level mechanism whereby differences in status between groups lead to inter-group inequalities in mental health outcomes

Figure 2. Two-level explanation of between-group disparities in mental health. The boxes placed above refer to the populational or “macro-social” level of analysis; those below to the individual or “micro-social” level. The “macro-social” antecedent and consequent are connected by a dashed arrow to indicate that the connection is still in need of clarification. The “micro” links describe an individual-level mechanism whereby differences in status between groups lead to inter-group inequalities in mental health outcomes

41An important difference between the “additive burden” model as used in psychiatry, on the one hand, and as embedded in the more typically sociological “social stress” model, on the other, concerns the mental health outcome. Whereas in psychiatry the paradigmatic outcome is a specific mental illness (e.g. depression), in sociological studies the effects of stress are usually assumed to be non-specific, in the sense of not limited to any particular disorder. Hence the more common use of generic terms such as “emotional distress.” In this sense, the sociological use of mental health indicators is neutral, or even agnostic with regard to clinical diagnoses of mental illness. Rather than dichotomously distinguishing psychiaric “cases” from “non-cases”, the sociologist is interested in rank-order in mental health outcomes between social groups.

42Sharon Schwartz and Ilan Meyer (2010) observe that most of the literature on the correlation between perceived discrimination and mental health outcomes has in a way or another interpreted this association using some version of the “social stress” paradigm. They contend, however, that the evidence is not in full adequacy with the explanation that the paradigm promotes, and in particular a more specific extension of the “social stress” framework to the situation of stigmatized groups known as the “minority stress” model (Meyer 2003). In this model, minorities are assumed to be exposed to unique stressors that, all else equal, explain the distress surplus that members of minority groups collectively exhibit compared to their non-minority counterparts. These stressors include objective events in the social environment such as being the target of discriminatory behavior (the focus of this article), but also personal dispositions such as expectations of rejection or internalized stigma, and psychological states such as heightened vigilance. For example, relying on a French student sample, Dambrun (2007) showed that women, compared to men, are more likely to perceive discrimination, and this difference in perceptions of discrimination partly explains why women’s mental health is not as good as men’s.

43Now, the bulk of the studies that look at the association between perceived discrimination and mental health do so within a particular group. Among Black Americans, for instance, those who report poorer mental health tend to be those who also recount more experiences of discrimination; the association, that is, holds within the group of Black Americans. Schwartz and Meyer argue that a within-group association provides no evidence of a disparity between groups, which is precisely what the “social stress” model predicts. That is, showing that perceived discrimination is associated with poorer mental health among Black Americans gives no clue as to whether Black Americans, on average, are more distressed than members of other racial groups, for instance White Americans.

44The example of Black Americans is revealing because by and large this group has been the focus of most of the studies that support the robust and nonsmall correlation between perceived discrimination and emotional distress. While the within-group correlation is beyond doubt, to date there has been no general demonstration of a mental health disparity between Blacks and Whites in the US that could be imputed to unequal exposure to discrimination (Williams et al. 2019).

45It is open to empirically scrutiny whether, beyond the particular case of Blacks in the US, other stigmatized groups in other geographical contexts exhibit a disadvantage in average mental health that may be due to discrimination.

3. Problems of measurement: Are measures of stressors confined to retrospective self-reports? The possibility of manipulating minor events experimentally

Retrospective self-reports of events: checklists vs. interviews

46How have studies in the“social stress” tradition operationalized stress? Most investigations in this area rely on verbal reports provided by respondents who accept to participate in a survey. The dominant procedure for measuring stress on the basis of self-reports involves the use of a checklist of relevant life events. In its classical form (Holmes and Rahe 1967), the respondent is presented with an inventory of events and asked to check those that have occurred to him or her in a recent period. Each event has a fixed amount of stress attached to it. By summating these fixed quantities, or alternatively by simply counting the number of events selected, the researcher obtains a summary measure of the respondent’s level of stress for a given period.

47For example, Aneshensel and colleagues (1991), in a research meant to exemplify what they call the sociological paradigm for studying stress, measured the latter construct by asking respondents to check those events from a list of twelve that had occurred to them during the six months preceding the survey. The events pertain to different areas, including work (e.g. fired), family (e.g. divorce), and crime victimization (e.g. robbed). Each respondent’s total stress score was the number of events checked, each event counting as one. Note that this procedure does not distinguish the stressor, i.e. the event, from the stress it causes to the person. Stressor and stress are treated as the same thing.

48Checklist inventories, and the vast body of reasearch that relies on them, have been the target over the years of a number of methodological criticisms (reviewed in Dohrenwend 2006). The consensually recognized problems concern both the reliability and the accuracy of the resulting measures. One problem of reliability is simple forgetting: the longer the temporal distance from the experienced event, the higher the probability that the event will not be checked. Another related problem has to do with the potential influence of the mental health outcome under study (e.g. depression) on the respondent’s ability to recall and check the events in the list, also known as “recall bias”. Yet another related problem concerns the time frame of the recalled events. For example, based on a US student sample, Strack, Schwartz and Gschneidinger (1985) showed that participants exposed to positive events do not report the same level of happiness and life-satisfaction depending on whether the time frame of recall is set to the near or the distant past. When the near past was the time frame, recalling positive events led to reporting higher well-being. But when the distant past played that role, recounting positive events correlated with lower well-being at the time of responding.

49Further, the researcher lacks a criterion to assess the accuracy of the events that the respondent checks. This problem is particularly vexing when measures of perceived discrimination are used as indexes of objective events. If the respondent checks an item relying so heavily on attributions of intent to others such as “not being taken seriously” or “having your ideas ignored” (Harrell 1997), how can the researcher verify that this is more than the respondent’s impression? Another problem of accuracy concerns the ability of the inventoried events to reflect genuine environmental demands, that is circumstances that temporally precede the respondent’s mental health state and are not possibly expressions of it. Consider, for instance, the event “being observed or followed while in public places”. Depending on the context, checking this item may be less the description of an outer event than a symptom of distress, confounding the measurement of environmental demand (the posited antecedent) with that of mental health (the posited consequent). In addition, the fact that checking the same item may represent very different types of actual experience to different respondents, or to the same respondent at different points in time, further illustrates the more general problem of “intracategory variability”. Last, another general limitation arises when attention is drawn to subtle stressors such as minor forms of everyday discrimination. A classical review of psychological research (Nisbett and Wilson 1977) indicates that people sometimes lack a conscious register of the existence of both the environmental stimulus (e.g., subtle discrimination) and their own response (e.g. stress or negative emotion), and they tend to be equally unaware that the stimulus affected the response (e.g. that the subtle discriminatory act caused them stress).

50Beyond amendments to traditional checklists, the major alternative approach that has been developed to remedy some of the shortcomings of inventories involves intensive interviewing with a view to eliciting detailed narratives of life events, after which trained judges rate the severity of the recounted occurrences. The British sociologist George W. Brown pioneered the development of the first instrument of this type (Brown and Harris 1978), followed by others. In psychiatry in particular, the interview method is regarded as the methodological gold standard (Hammen 2005, Mazure 1998, Paykel 2001). This method, however, is more labor-intensive, implying that with fixed resources it yields samples of smaller size than could be built with more expedite checklists. It is important to note that in life events research intensive interviewing has been classicaly coupled with a method of probabilistic sampling (that is, respondents are randomly selected for inclusion in the study), resulting in measures that are at the same time accurate enough and representative of a well-defined target population (for example, women aged 18 to 50 within a restricted geographic area).

51If the emphasis is on quality of causal inference, the interview method may provide the best approach. If the researcher is willing to sacrifice accuracy in the name of larger representativity, checklists of discrete events may offer a more attractive alternative. It is important to emphasize, however, that for all their important differences checklists and interviews are united in their common dependency on retrospective reports.

Beyond retrospetive reports: minor events as experimental treatments

52When the aim of the investigator is to explain mental health inequalities in terms of differential exposure to stressors such as discrimination in interpersonal encounters, a further alternative method is to examine the prevalence and impact of such discrete and observable events experimentally in real-life settings. But to the best of my knowledge to date this possibility has not been pursued.

53The advantages of adopting a field experimental methodology as an alternative or supplement to the checklist/interview approach are considerable. By measuring the occurrence of stressful events on the basis of the observation of behavior in response to controlled manipulations, field experiments dissolve one after the other the problems of reliability and validity that creates the dependence on verbal reports. Nothing to be remembered, and so no problems of recall. Direct observation of the environmental demand or “stressor”, and so no need for a criterion of occurrence. Control over the environmental demand as an experimental manipulation, and so no room for confounding the stressor with the mental health outcome. While the advantages seem obvious, there are also important limitations, which may explain why this approach has not been adopted as a general solution.

54The fundamental drawback from which the others follow is that it is morally unjustifiable to expose participants to severe stress. The very idea of experimental manipulation is morally shocking if the researcher’s focus is on major, fateful occurrences. But things look differently if the researcher’s interest goes to minor incidents like “not being taken seriously”, because what makes these incidents damaging is not their isolated occurrence, as it often the case for major life events, but rather their sustained repetition over time. What is unethical is not the manipulation of stressors per se (Selye famously used to say that “life is stress”), but the fact that in doing so the experimenter may deliberately provoke distress (e.g. depression) in participants. In isolation, as they can be in an experimental procedure, the minor forms of stress involved in more mundane events are incapable of precipitating distress. Thus the researcher can produce experimental evidence of the incidence of a candidate minor stressor for a given population (e.g. subtle discrimination against Muslims in everyday interaction) and rely on observational survey data to measure the level of distress (e.g. depression) within that population.

4. Muslims and Blacks in Europe through the lens of minority stress

  • 3 The manuscript was selected finalist for the Rinus Penninx best paper award at the 2022 annual conf (...)

55I have adopted this multi-method empirical strategy for looking at the cases of European Muslims (2022a) and European Blacks (2022b) from the standpoint of the minority stress model. Although both studies are similar in logic, method and results, I shall concentrate on the one dealing with Muslims because at the time of writing it has been reviewed by a wider variety of near, distant and anonymous colleagues, is freely available online, and has even received a distinction.3 The exposition is deliberately non-technical.

The minority stress model applied to the case of Muslims in Europe

56In countries of Christian tradition such as Western European states, Muslims represent a minority. There can be disagreement about whether Muslims’ minority status has to do more with the marginalized position of Islam in the European landscape of religions and more generally of cultural traditions, or rather with a host of signs of social subordination that cluster together in Muslims regarded as a group of people. But all we need at this stage is to recognize European Muslims as a minority, which seems less controversial.

57The minority stress model, applied to the case of European Muslims, and using depression as the indicator of emotional distress, yields the following hypothesis: all else equal, European Muslims are more depressed than European non-Muslims because (or partly because) they are more exposed to discrimination than European non-Muslims. This complex hypothesis is known as a mediation model.

Logic of the mediation model

58A mediation model is a three-variable system (MacKinnon et al. 2007). In general, the mediation hypothesis is that the “predictor” variable produces changes in the “outcome” variable through the action of the “mediator” variable. Put otherwise, the mediation model states that the effects of the predictor on the outcome are channeled by the mediator, partly or entirely. Everyday life offers many intuitive examples of mediation. The shocking effect of an email on the recipient is mediated by the recipient’s interpretation of the email. The heating effect on water of pressing the “on” button of my kettle is mediated by energy travelling through the coil within the kettle. The protective effects of taking a vaccine are mediated by the vaccine’s triggering an immune response within the body. Seen as a particular application of this schema, the minority stress model states that the negative effect of minority position on mental health is (at least partly) mediated by discrimination. It is useful at this point to decompose the complex mediation hypothesis into simpler components that can be directly subjected to empirical testing.

Empirical testing of the mediation model: Hypotheses

59Hypothesis 1. The first of the three simpler statements is the expectation that European Muslims, all else equal, are on average more depressed than European non-Muslims. Put otherwise, individuals who are identical on all important variables (e.g. same gender, same age, same material conditions of life –these are called “control predictors” or “controls”) but differ only by the fact that some of them are Muslims while some of them are not, are expected to exhibit a difference in depression such that Muslims will be penalized by a “surplus” in depressive symptomatology. That is, when all control variables (gender, etc.) are held at the same level, the average level of depression among Muslims is expected to be higher than among non-Muslims.

60Hypothesis 2. The second hypothesis is that, as much as being Muslim (vs. non-Muslim) predicts a depression surplus, being Muslim (vs. non-Muslim) predicts a discrimination surplus. That is, Muslims who are similar to non-Muslims on many important characteristics (gender, etc.) nevertheless are more likely to face discrimination than their non-Muslim counterparts.

61Hypothesis 3. The third and last component hypothesis of the mediation model is that the difference in depression between Muslims and non-Muslims (the depression surplus) should decrease or disappear when the Muslims and non-Muslims being compared are not only equal on the “control predictors” (gender, etc.) but also on discrimination. That is, Muslims and non-Muslims that are similar on important characteristics (gender, etc.) and also on the level of discrimination that they face are expected to differ little or even not differ at all in depression.

62If Hypotheses 1, 2 and 3 are in accordance with the data, the mediation model is usually considered to be supported.

Testing the model using survey and experimental data

63The empirical strategy is to combine two different but interlocking studies. These are based on differently produced data sets, one resulting from a survey and the other from experiments on social interactions in public places. The survey study relies on the European Social Survey (ESS), 7th round, whereas the field experiments involve scripted face-to-face contacts between a confederate (i.e. an accomplice to the researcher) and unacquainted strangers in a real-life setting.

64Occurring first in logical order, the survey study tests Hypotheses 1, 2 and 3. The results are then used to inform the design of the field experiments, whose role is to reexamine with a different data set the hypothesis of anti-Muslim discrimination, already tested in the survey study. The survey study provides the “big picture” and guidance for the field experiment. More restricted in focus, the field experiment retests the hypothesis of anti-Muslim discrimination with data that are more attuned to the causal claims entailed in the minority stress model. For details on measures and statistical analyses the interested reader may consult the original report.

65The survey study. The results of the survey study are based on a variant of regression analysis known as the “multilevel model”, which tests the mediation hypothesis at the European level and at the same time within more specific subsamples by country, gender and Muslim religiosity (high vs. low on a religiosity scale). The first task was to confirm the mediation model at the aggregate, European level. Supporting Hypotheses 1, 2 and 3, the analysis shows that in Europe overall, as expected, the average level of depression is higher among Muslims than non-Muslims, and that this surplus depression is partly imputable to Muslims’ overexposure to discrimination. In preparation for the field experiment, the second task was to identify the group by country, gender and religiosity that best confirmed the pattern found at the pan-European level. Candidate groups were, for example, German Muslim men of high religiosity, British Muslim women of low religiosity, and so on. The results very clearly indicated that the group that stands out as the best exemplification of the pattern observed at the European level are French Muslim women of high religiosity. In other words, the survey study showcased highly religious Muslim women residing in France as the most clear confirmation of a surplus in depression imputable to discrimination.

66Moving from the survey study to the field experiments. At this point, all the problems of method discussed in section 3 (see above) come to the stage. How can we be sure that these women are not reporting more discrimination precisely because they are more depressed (the problem of “confounding”)? If they were depressed for any other reasons, but depression caused them to over-report discrimination, the mediation analysis would be totally invalidated. To which must be added the more generic uncertainties that haunt the reliability and accuracy of self-report measures of discrimination.

67Conducting a field experiment to retest the hypothesis that highly religious Muslim women are discriminated against in France offers a solution to these problems of method. On the one hand, it makes the measurement of discrimination independent from that of depression, dissipating the danger of confounding. On the other, it provides direct measures of discriminatory behavior, eliminating the need for self-reports.

68The field experiments. The task of the experiments is to retest the finding from the survey study that highly religious women are discriminated against in France. The concept of discrimination (see the introduction to this issue) requires 1) a difference in treatment that 2) results in a disadvantage, penalty or lower gain for the discriminatee. The particular form of penalty we are after is stress. The first experiment addresses requirement 1; the second experiment examines requirement 2. In spite of this difference in goal, both experiments employ the same procedure: a confederate, an accomplice to the researcher, follows a script to interact face-to-face with a sample of randomly selected passengers (one at the time) on a set of platforms of the Paris metro. During the interaction, passengers are lead to believe that the confederate is another passenger who, like them, just happens to be there and then to take the metro (they are told the truth shortly afterwards).

69Experiment 1. The first field experiment used the Islamic headscarf or hijab as a public sign of high Islamic religiosity. This is not to assume that women who use the hijab wish to publicly express their deep commitment to Islam. The assumption is simply that most people in France see the hijab as a public sign of Islamic commitment, which is all we need. A professional actress served as the confederate in this experiment. In the “treatment” condition, she wore an Islamic headscarf. In the “control” condition, she appeared with uncovered hair. The rest of the clothing was identical. She followed the same script when interacting with all passengers. Replicating the finding of a previous experiment (Aranguren et al. 2021), the study uncovered that the headscarf causes men to change their pattern of gaze as they interact with the actress. More specifically, in the hijab condition men avert gaze more often, they look away more frequently, diminishing the intensity of eye contact.

70The transition to the second experiment is enabled by the following question: what is the average effect, for the recipient, of being looked at in this way? Is the “hijab-gaze” stressful? Note that the second experiment is not interested in men’s motivation for looking at the hijab-wearing confederate in the way they did. They might have been expressing regard, avoidance, embarrasssment, or anything else, but this is not essential for the present investigation. The focus here is on the consequence for the recipient, whatever the good, bad, or confused intentions of the sender.

71Experiment 2. The second experiment thus uses the “hijab-gaze” as the experimental treatment, versus a “normal” (but strictly defined) gaze for the control condition. The confederate is this time a male actor, because the “hijab-gaze” was only shown by male passengers, and the participants in the experiment are all female because male passengers produced the “hijab-gaze” when interacting with a female. The experimental procedure, the sampling technique and the social setting are the same as in Experiment 1. Does the “hijab-gaze” cause stress to its recipient? The results show that being looked at in this manner causes strong negative emotions, a classical indicator of stress. Women who are gazed at in this more evasive way by the male confederate report much stronger feelings of emotional discomfort. So yes, the “hijab-gaze” is stressful for the recipient.

72Integrating the survey study and the experiments. The survey study concluded that European Muslims’ surplus depression is partially imputable to discrimination, showcasing highly religious Muslim women from France as the subsample that best confirmed this pan-European pattern. The experiments focused on the latter group and retested the hypothesis of anti-Muslim discrimination, looking at stress as the adverse consequence (because depression is hypothesized to result from excessive stress). It was found that highly religious Muslim women in France are overexposed to a particular pattern of gaze on the part of men that functions to cause them stress.

73In spite of profound differences in method, the results of both studies are convergent and reinforce one another. What other more or less subtle stressors do European Muslims, and highly religious Muslim women from France in particular, encounter in everyday life? How frequent are these confrontations with stressors? Open questions for future work.


74More than two decades of intense investigation provide robust evidence of a nonsmall association between perceived discrimination and emotional distress. However, this literature has not examined whether and to what extent discrimination explains inequalities in mental health between groups. Moreover, it has kept a sustained focus on Blacks in the US, devoting less attention to other minorities and national contexts. Methodologically, it has relied entirely on retrospective self-reports of perceived discrimination.

75The program I have sketched builds on this literature but also seeks to address some of its weaknesses. In this sense, the multi-method studies on European Muslims and Blacks conducted under this program leave space for a degree of optimism. These studies show that it is possible and fruitful to examine the role of discrimination in the production of mental health inequalities, that this process is at least partly responsible for the deteriorated mental health of European minorities, and that survey and experimental data can be integrated to strengthen the validity of research findings.

76One possible avenue for future work could be to supplement this survey-experiment mix with the interview method, an entry unique in its capacity to prompt detailed and vivid narratives of concrete episodes of confrontations with discrimination, promising much greater accuracy than is conceivable in the context of a survey. Such an approach could be pursued in the usual “qualitative” way, but there is no ultimate reason why the semi-structured interview could not go hand-in-hand with a method of probabilistic or random sampling, as was the case of the cited classic studies in life events research. Marrying the semi-structured interview and probabilistic sampling, given the time-consuming tasks of recruiting and interviewing participants, is nevertheless likely to affect the size of the samples to be built within a single study. The classical studies from the life events tradition solved this problem by randomly sampling respondents from small territorial units (e.g. towns, the French commune) that other sources pointed as paradigmatic cases. The results are unbiased but extend only to the selected territorial unit. As with testings or any other punctual interventions, the extrapolability of these local results can be assessed by replicating the same type of study in a variety of settings.

77Understandig the influence of everyday discrimination on mental health inequalities is the sort of enterprise that cannot be optimally pursued with a single method. The survey excels in the ability to represent a national population, but provides measures of discrimination of unknown reliability and accuracy, and requires many untestable assumptions to warrant causal claims (for example, that discrimination causes depression and not the opposite). The interview method is unsurpassable as a window into the details of past situations to which the researcher was not present. But at the same time its high logistical cost reduces the size of the population to which the results can be legimately extended (assuming probabilistic sampling). The experimental method, in turn, cannot be used to randomly decide which participants will repeatedly face discrimination in everyday life for a time period, so as to then examine whether this turns out to affect their level of depression. This is simply unethical. But the experiment can be used for testing the causal hypothesis, intrinsic to the minority stress model, that the trait of interest (e.g. race), and not anything else, is what causes the people with that trait to be overexposed to minor forms of discrimination in daily life. These minor forms, in isolation, are just small hassles or turndowns with little intrinsic impact. Their damaging effects on mental health come rather from frequent repetition.

78In this, as in many other fields in the social sciences, the future lies perhaps in taking the best from each method and in compensating its weaknesses with the complementary strengths of the others.

Haut de page


Aneshensel, Carol S., Carolyn M. Rutter and Peter A. Lachenbruch. 1991. “Social Structure, Stress, and Mental Health: Competing Conceptual and Analytic Models.” American Sociological Review 56(2):166-178. doi: 10.2307/2095777.

Aranguren, Martin. 2022a. “Interactional discrimination partly explains why Muslim women are more depressed than average: integrating survey data with field experiments in the study of minority stress.” Pre-print available at

Aranguren, Martin. 2022b. “In Europe Blacks Are More Depressed than Average Partly Because They Are Discriminated against: Combining Survey and Experimental Data to Explain Mental Health Disparities as a Consequence of Unequal Treatment.” Unpublished manuscript.

Aranguren, Martin, Francesco Madrisotti, Eser Durmaz-Martins, Gernot Gerger, Lena Wittmann and Marc Méhu. 2021. “Responses to the islamic headscarf in everyday interactions depend on sex and locale: a field experiment in the metros of Brussels, Paris, and Vienna on helping and involvement behaviors.” PLOS ONE 16(7). doi: 10.1371/journal.pone.0254927.

Bosco, Frank A., Herman Aguinis, Kulraj Singh, James G. Field and Charles A. Pierce. 2015. “Correlational Effect Size Benchmarks.” Journal of Applied Psychology 100(2): 431-449. doi: 10.1037/a0038047.

Britt-Spells, Angelitta M., Maribeth Slebodnik, Laura P. Sands and David Rollock. 2018. “Effects of Perceived Discrimination on Depressive Symptoms Among Black Men Residing in the United States: A Meta-Analysis.” American Journal of Men’s Health 12(1): 52-63. doi: 10.1177/1557988315624509.

Branscombe, Nyla R., Michael T. Schmitt and Richard D. Harvey. (1999). “Perceiving pervasive discrimination among African Americans: Implications for group identification and well-being.” Journal of personality and social psychology 77(1): 135-149. doi: 10.1037/0022-3514.77.1.135.

Brown, George W. and Tirril Harris. 1978. Social Origins of Depression. A Study of Psychiatric Disorder in Women. Oxford (England): Routledge.

Carter, Robert T., Veronica E. Johnson, Katherine Kirkinis, Katheryn Roberson, Carrie Muchow and Corinne Galgay. 2019. “A Meta-Analytic Review of Racial Discrimination: Relationships to Health and Culture.” Race and Social Problems 11(1): 15-32. doi: 10.1007/s12552-018-9256-y.

Carter, Robert T., Michael Y. Lau, Veronica Johnson and Katherine Kirkinis. 2017. “Racial Discrimination and Health Outcomes Among Racial/Ethnic Minorities: A Meta-Analytic Review.” Journal of Multicultural Counseling and Development 45(4): 232-259. doi: 10.1002/jmcd.12076.

Clark, Rodney, Norman B. Anderson, Vernessa R. Clark and David R. Williams. 1999. “Racism as a Stressor for African Americans: A Biopsychosocial Model.” American Psychologist 54(10): 805-816. doi: 10.1037/0003-066X.54.10.805.

Crocker, Jennifer and Brenda Major. (1989). “Social stigma and self-esteem: The self-protective properties of stigma.” Psychological review 96(4): 608-630. doi: 10.1037/0033-295X.96.4.608.

Dohrenwend, Bruce P. 1961. “The Social Psychological Nature of Stress: A Framework for Causal Inquiry.” The Journal of Abnormal and Social Psychology 62(2): 294-302. doi: 10.1037/h0040573.

Dohrenwend, Bruce P. 2000. “The Role of Adversity and Stress in Psychopathology: Some Evidence and Its Implications for Theory and Research.” Journal of Health and Social Behavior 41(1): 1-19. doi: 10.2307/2676357.

Dohrenwend, Bruce P. 2006. “Inventorying Stressful Life Events as Risk Factors for Psychopathology: Toward Resolution of the Problem of Intracategory Variability.” Psychological Bulletin 132(3): 477-495. doi: 10.1037/0033-2909.132.3.477.

Eikemo, Terje A., Clare Bambra, Tim Huijts and Rory Fitzgerald. 2017. “The First Pan-European Sociological Health Inequalities Survey of the General Population: The European Social Survey Rotating Module on the Social Determinants of Health.” European Sociological Review 33(1): 137-153. doi: 10.1093/esr/jcw019.

de Freitas, Daniela Fonseca, Maria Fernandes-Jesus, Pedro D. Ferreira, Susana Coimbra, Pedro M. Teixeira, Andreia de Moura, Jorge Gato, Susana C. Marques and Anne Marie Fontaine. 2018. “Psychological Correlates of Perceived Ethnic Discrimination in Europe: A Meta-Analysis.” Psychology of Violence 8(6): 712-725. doi: 10.1037/vio0000215.

Greene, Jennifer C., Valerie J. Caracelli and Wendy F. Graham. 1989. “Toward a Conceptual Framework for Mixed-Method Evaluation Designs.” Educational Evaluation and Policy Analysis 11(3): 255-274. doi: 10.3102/01623737011003255.

Hammen, Constance. 1991. “Generation of Stress in the Course of Unipolar Depression.” Journal of Abnormal Psychology 100(4): 555-561. doi: 10.1037/0021-843X.100.4.555.

Hammen, Constance. 2005. “Stress and Depression.” Annual Review of Clinical Psychology 1(1): 293-319. doi: 10.1146/annurev.clinpsy.1.102803.143938.

Hammen, Constance. 2018. “Risk Factors for Depression: An Autobiographical Review.” Annual Review of Clinical Psychology 14(1): 1-28. doi: 10.1146/annurev-clinpsy-050817-084811.

Hammen, Constance L. 2015. “Stress and Depression: Old Questions, New Approaches.” Current Opinion in Psychology 4: 80-85. doi: 10.1016/j.copsyc.2014.12.024.

Harrell, Shelly P. 1997. “The Racism and Life Experiences Scales (RaLES): Daily Life Experiences (Racial Hassles).”

Hedström, Peter and Petri Ylikoski. 2010. “Causal Mechanisms in the Social Sciences.” Annual Review of Sociology 36(1): 49-67. doi: 10.1146/annurev.soc.012809.102632.

Holmes, Thomas H. and Richard H. Rahe. 1967. “The Social Readjustment Rating Scale.” Journal of Psychosomatic Research 11(2): 213-218. doi: 10.1016/0022-3999(67)90010-4.

Jones, Kristen P., Chad I. Peddie, Veronica L. Gilrane, Eden B. King and Alexis L. Gray. 2016. “Not So Subtle: A Meta-Analytic Investigation of the Correlates of Subtle and Overt Discrimination.” Journal of Management 42(6): 1588-1613. doi: 10.1177/0149206313506466.

Krieger, Nancy, Kevin Smith, Deepa Naishadham, Cathy Hartman and Elizabeth M. Barbeau. 2005. “Experiences of Discrimination: Validity and Reliability of a Self-Report Measure for Population Health Research on Racism and Health.” Social Science & Medicine 61(7): 1576-1596. doi: 10.1016/j.socscimed.2005.03.006.

Landrine, Hope and Elizabeth A. Klonoff. 1996. “The Schedule of Racist Events: A Measure of Racial Discrimination and a Study of Its Negative Physical and Mental Health Consequences.” Journal of Black Psychology 22(2): 144-168. doi: 10.1177/00957984960222002.

LeBlanc, Allen J. and David M. Frost. 2020. “Couple-Level Minority Stress and Mental Health among People in Same-Sex Relationships: Extending Minority Stress Theory.” Society and Mental Health 10(3): 276-290. doi: 10.1177/2156869319884724.

Lee, Debbiesiu L. and Soyeon Ahn. 2011. “Racial Discrimination and Asian Mental Health: A Meta-Analysis.” The Counseling Psychologist 39(3): 463-489. doi: 10.1177/0011000010381791.

Lee, Debbiesiu L. and Soyeon Ahn. 2012. “Discrimination Against Latina/Os: A Meta-Analysis of Individual-Level Resources and Outcomes Ψ.” The Counseling Psychologist 40(1): 28-65. doi: 10.1177/0011000011403326.

Lee, Debbiesiu L. and Soyeon Ahn. 2013. “The Relation of Racial Identity, Ethnic Identity, and Racial Socialization to Discrimination–Distress: A Meta-Analysis of Black Americans.” Journal of Counseling Psychology 60(1): 1-14. doi: 10.1037/a0031275.

MacKinnon, David P., Amanda J. Fairchild and Matthew S. Fritz. 2007. “Mediation analysis.” Annual review of psychology 58: 593-614. doi: 10.1146/annurev.psych.58.110405.085542.

Mazure, Carolyn M. 1998. “Life Stressors as Risk Factors in Depression.” Clinical Psychology: Science and Practice 5(3): 291-313. doi: 10.1111/j.1468-2850.1998.tb00151.x.

Meyer, Ilan H. 2003. “Prejudice as Stress: Conceptual and Measurement Problems.” American Journal of Public Health 93(2): 262-265. doi: 10.2105/AJPH.93.2.262.

Meyer, Ilan H. 2003. “Prejudice, Social Stress, and Mental Health in Lesbian, Gay, and Bisexual Populations: Conceptual Issues and Research Evidence.” Psychological Bulletin 129(5): 674-697. doi: 10.1037/0033-2909.129.5.674.

Paradies, Yin, Jehonathan Ben, Nida Denson, Amanuel Elias, Naomi Priest, Alex Pieterse, Arpana Gupta, Margaret Kelaher and Gilbert Gee. 2015. “Racism as a Determinant of Health: A Systematic Review and Meta-Analysis.” PLOS ONE 10(9): e0138511. doi: 10.1371/journal.pone.0138511.

Pascoe, Elizabeth A. and Laura Smart Richman. 2009. “Perceived Discrimination and Health: A Meta-Analytic Review.” Psychological Bulletin 135(4): 531-554. doi: 10.1037/a0016059.

Paykel, Eugene S. 1997. “The Interview for Recent Life Events.” Psychological Medicine 27(2): 301-310. doi: 10.1017/S0033291796004424.

Paykel, Eugene S. 2001. “The Evolution of Life Events Research in Psychiatry.” Journal of Affective Disorders 62(3): 141-149. doi: 10.1016/S0165-0327(00)00174-9.

Pearlin, Leonard I. 1989. “The Sociological Study of Stress.” Journal of Health and Social Behavior 30(3): 241-256. doi: 10.2307/2136956.

Pearlin, Leonard I., Elizabeth G. Menaghan, Morton A. Lieberman and Joseph T. Mullan. 1981. “The Stress Process.” Journal of Health and Social Behavior 22(4): 337-356. doi: 10.2307/2136676.

Pieterse, Alex L., Nathan R. Todd, Helen A. Neville and Robert T. Carter. 2012. “Perceived Racism and Mental Health among Black American Adults: A Meta-Analytic Review.” Journal of Counseling Psychology 59(1): 1-9. doi: 10.1037/a0026208.

Ramos, Miguel R., Clare Cassidy, Stephen Reicher and S. Alexander Haslam. 2012. “A longitudinal investigation of the rejection–identification hypothesis.” British Journal of Social Psychology 51(4): 642-660. 10.1111/j.2044-8309.2011.02029.x.

Schmitt, Michael T., Nyla R. Branscombe, Tom Postmes and Amber Garcia. 2014. “The Consequences of Perceived Discrimination for Psychological Well-Being: A Meta-Analytic Review.” Psychological Bulletin 140(4): 921-948. doi: 10.1037/a0035754.

Schwartz, Sharon. 2002. “Outcomes for the Sociology of Mental Health: Are We Meeting Our Goals?” Journal of Health and Social Behavior 43(2): 223-235. doi: 10.2307/3090198.

Schwartz, Sharon and Ilan H. Meyer. 2010. “Mental Health Disparities Research: The Impact of within and between Group Analyses on Tests of Social Stress Hypotheses.” Social Science & Medicine 70(8): 1111-1118. doi: 10.1016/j.socscimed.2009.11.032.

Selye, Hans. 1936. “A Syndrome Produced by Diverse Nocuous Agents.” Nature 138(3479): 32-32. doi: 10.1038/138032a0.

Tajfel, Henri. 1982. “Social Psychology of Intergroup Relations.” Annual Review of Psychology 33(1): 1-39. doi: 10.1146/

Triana, María del Carmen, Mevan Jayasinghe and Jenna R. Pieper. 2015. “Perceived Workplace Racial Discrimination and Its Correlates: A Meta-Analysis.” Journal of Organizational Behavior 36(4): 491-513. doi: 10.1002/job.1988.

Williams, David R., Jourdyn A. Lawrence, Brigette A. Davis and Cecilia Vu. 2019. “Understanding How Discrimination Can Affect Health.” Health Services Research 54(S2): 1374-1388. doi: 10.1111/1475-6773.13222.

Zautra, Alex J., Charles A. Guarnaccia and Bruce P. Dohrenwend. 1986. “Measuring Small Life Events.” American Journal of Community Psychology 14(6): 629-655. doi: 10.1007/BF00931340.

Haut de page


1 In non-experimental research, a data production scheme (e.g. administering a questionnaire to, or conducting interviews with, a sample of respondents) is said to follow a cross-sectional design when the time of measurement is the same for all the measured variables. For example, one individual respondent fills the questionaire in a single session occurring the same day. Similarly, data are said to be cross-sectional when they result from a cross-sectional data production design. This can be contrasted with a longitudinally designed data production process, in which the same units of observation (e.g. respondents to a survey) provide measures at different points in time. For example, the respondent fills the questionnaire one day of year 1, one day of year 2 and one day of year 3. One desirable property of measures derived from a longitudinal design is that it enables the analysis of changes in time within the same respondent (for example, a respondent reports no symptoms of depression in year 1 but feels depressed in year 2).

2 “Confounding” refers here to a problem of measurement. A measure of perceived discrimination may be confounded with one of mental health, e.g. depression, because reports of perceived discrimination may be influenced by the cognitive and motivational deficits of underlying depression at the time of responding. I feel depressed, and this may lead me to perceive discrimination where there is not, or to fail to perceive discrimination where there actually is. The same can be said of the influence of current depression on the ability to recall, or the willingness to report, past events of discrimination.

3 The manuscript was selected finalist for the Rinus Penninx best paper award at the 2022 annual conference of the International Migration Research Network (IMISCOE) held in Oslo.

Haut de page

Table des illustrations

Titre Table 1. Meta-analyses on the correlation between perceived discrimination and mental health
Fichier image/png, 2,6M
Titre Figure 1. The discrimination-distress association reinterpreted causally, with stress acting as the mediator
Fichier image/png, 82k
Titre Figure 2. Two-level explanation of between-group disparities in mental health. The boxes placed above refer to the populational or “macro-social” level of analysis; those below to the individual or “micro-social” level. The “macro-social” antecedent and consequent are connected by a dashed arrow to indicate that the connection is still in need of clarification. The “micro” links describe an individual-level mechanism whereby differences in status between groups lead to inter-group inequalities in mental health outcomes
Fichier image/png, 224k
Haut de page

Pour citer cet article

Référence électronique

Martin Aranguren, « Sketch of a research program on the contribution of discrimination to mental health inequalities: a critical review of evidence, models and methods »Appartenances & Altérités [En ligne], 3 | 2023, mis en ligne le 01 mars 2023, consulté le 14 juin 2024. URL : ; DOI :

Haut de page


Martin Aranguren

Chargé de recherche au CNRS
Centre de Recherche sur les Inégalités Sociales, Sciences Po

Articles du même auteur

Haut de page

Droits d’auteur


Le texte seul est utilisable sous licence CC BY-NC-ND 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont « Tous droits réservés », sauf mention contraire.

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search