In a cross-sectional analysis we used data from two nationally representative studies of aging: ELSA in England and the HRS in the US. These studies were designed to be comparable and closely matched in sampling and questionnaire content. Harmonized data files from wave 5 (2010) of ELSA and wave 7 of HRS (2010) were obtained from the Gateway to Global Aging (g2aging.org). Analyses were constrained to 2010 as perceived discrimination was assessed in ELSA at this time point only.

ELSA and HRS cohorts have been described in detail elsewhere (25, 26). Analyses were restricted to participants aged ≥ 50 years who provided perceived discrimination data, resulting in analytic samples of N = 8,671 (ELSA) and N = 7,927 (HRS), Figure 1.

ELSA was approved by the London Multicentre Research and Ethics Committee (MREC/01/02/91). Approval for HRS was obtained from the University of Michigan Institutional Review Board (https://hrs.isr.umich.edu/publications/biblio/9048).

Perceived discrimination was assessed in both cohorts using the same items based on the frequency of discrimination in five situations: “In your day-to-day life, how often have any of the following things happened to you (1) you are treated with less respect or courtesy; (2) you receive poorer service than other people in restaurants and stores (ELSA: shops); (3) people act as if they think you are not clever; (4) you are threatened or harassed; (5) you receive poorer service or treatment than other people from doctors or hospitals” (almost every day/at least once a week/a few times a month/a few times a year/less than once a year/never). As previously reported (27), the data were skewed, with most individuals reporting “never” experiencing discrimination, we created a binary variable to indicate whether participants had experienced discrimination in the past year (a few times or more a year vs. less than once a year or never), except for the fifth item which was dichotomized to indicate whether respondents had ever experienced discrimination from doctors or hospitals (never vs. all other options) as most participants reported “never” experiencing discrimination in this setting. A follow-up question asked participants to attribute the discriminatory experience to one or more reasons: physical disability, financial status, race, sex, sexual orientation, weight, and age. This measure has good validity for the assessment of discrimination (28, 29) and has been widely used in research investigating associations between discrimination and health in ELSA and HRS (5–8, 11, 12, 30).

Age was modeled as a categorical variable (52–59; 60–69; 70–79; 80+ years old). Sex (female/male) was modeled in binary. Wealth (excluding pension wealth) is the most relevant indicator of SEP in these cohorts (25, 26). This measure is based on detailed assessments of socioeconomic resources (e.g., financial wealth, including housing costs, assets, earnings, debts; and physical wealth) (i.e., land and jewelery). We modeled wealth as a continuous and categorical variable based on a cohort-specific median split (below median = low SEP, above the median = high SEP).

Analyses were carried out in R. We first assessed whether the HRS and ELSA samples differed in age, sex, and wealth (our key covariates) using a series of t-tests for normally distributed continuous variables (age, wealth) and χ² (chi-square) tests for binary variables (sex and SEP level).

We assessed unadjusted cross-national differences in the prevalence of perceived discrimination attributed to disability, financial status, sex, race, sexual orientation, and weight using multiple logistic regression models and χ²-tests. Analyses of perceived disability discrimination were restricted to those living with a longstanding physical limitation (defined as an impairment in basic or instrumental activities of daily living or impaired mobility). Perceived race discrimination analyses were restricted to ethnic minority participants. Perceived sex discrimination analyses were limited to female participants. perceived weight discrimination analyses were restricted to obese participants (Body Mass Index: BMI > 30 kg/m²). We choose to restrict all analyses to improve the precision of the estimates so that we compare the prevalence of discrimination in non-dominant groups (e.g., women) who do and do not perceive discrimination.

Sexual orientation was assessed using item: “Which statement best describes your sexual desires over your lifetime? Please include being interested in sex, fantasizing about sex or wanting to have sex”: (1) entirely for women, (2) mostly for women, but some desires for men, (3) equally for women and men, (4) mostly for men, but some desires for women, (5) entirely for men, and (6) no sexual desires in lifetime. We categorized participants with reported desires entirely for the opposite sex as heterosexual, entirely for the same sex as gay and those reporting desires equally for both sexes, mostly for the same sex, or some desires for opposite sex as bisexual. The number of lesbian, gay, or bisexual (LGB) individuals was small in both ELSA (n = 139) and HRS cohorts (n = 9). Therefore, sexual orientation discrimination analysis was not restricted to LGB individuals. As in previous work (22), data were unweighted as we combined two subsamples of HRS and ELSA respondents, which had different weights.

We then conducted logistic regression analyses adjusting for covariates that were significantly associated with each discrimination type as in previous work (22). We implemented generalized linear models (binomial) with perceived discrimination as the independent variable and country (US/England) as the dependent variable. Likelihood ratio tests for two nested models were conducted to assess differences in perceived discrimination between the countries. We also assessed these relationships looking at each of the five discriminatory situations separately.

We performed secondary analyses to compare the role of SEP in the prevalence of perceived discrimination in England and the US. Firstly, we fitted multiple binomial regression models to compare the moderating effect of continuous wealth on each type of perceived discrimination across the two countries. We also re-ran logistic regression analyses stratified by SEP (low/high), removing adjustment for this variable. Secondly, removing the stratification by SEP, we assessed the interactive effects of SEP and country on the prevalence of perceived discrimination attributed to each characteristic separately. A log-likelihood ratio test was used to test for interaction. As in previous work (20), data were unweighted as we combined two subsamples of HRS and ELSA respondents, which had different weights.

We carried out the cross-national comparison stratified by SEP operationalized using education level. Education was included as binary variable (no higher degree and higher degree), describing the highest educational qualification attained. No higher degree included no formal education, GCSE, O-Level A-Levels, or equivalent in ELSA; and no formal education, education below high school or completed high school in HRS. Higher degree included university degree or higher in ELSA and college to post-college in HRS. We also included education as a covariate in addition to age, sex, and wealth.

The analysis concerned within-country comparisons (low SEP vs. high SEP and association between wealth and perceived discrimination) was also weighted to test the impact of adjusting for selection bias. We performed weighted logistic regression to account for selection bias in HRS and ELSA separately when assessing the association between wealth and discrimination.

This study included 8,671 individuals from ELSA [M = 66.57 (SD = 9.03) years old; 4816 (55.54%) female] and 7,927 individuals from HRS [M = 67.28 (SD = 10.80) years old; 4,583 (57.82%) female] who responded to the perceived discrimination survey (Table 1).

We observed significant differences in age, sex and SEP between ELSA and HRS samples. Participants in HRS were on average older [t₍₁₎ = 4.57, p < 0.001], more likely to be female (χ² = 8.62, p < 0.01), and less wealthy [t₍₁₎ = −9.26, P < 0.001] than those in ELSA. There were no significant differences in number of individuals in low SEP between HRS (n = 7,068, 89%) and ELSA (n = 7,581, 87%) cohorts, χ² = 0.61, p = 0.44.

In unadjusted analysis (Table 2), perceived discrimination attributed to disability [OR = 3.05, 95% CI (1.42; 6.55), p < 0.01], financial status [OR = 3.31, 95%CI (2.75; 4.00), p < 0.001], sexual orientation [OR = 4.61, 95%CI (2.48; 8.57), p < 0.001], and weight [OR = 1.41, 95% CI (1.14; 1.75), p < 0.01] was more frequently reported in England than the US. Race- [OR = 0.62, 95%CI (0.42; 0.93), p < 0.05] and sex-based discrimination [OR = 0.35, 95%CI (0.30; 0.40), p < 0.001] were more prevalent in the US than in England.

In adjusted analyses, significant cross-national differences in financial status and sex-based perceived discrimination remained. Specifically, perceived financial status discrimination [OR = 1.10, 95%CI (1.09; 1.12), p < 0.001] was more common in England than in the US (6.65 vs. 2.14%), adjusting for age, sex, and wealth. Sex discrimination was more prevalent in the US than in England [12.42 vs. 9.07%; OR = 0.87, 95%CI (0.86; 0.89), p < 0.001] adjusting for age and wealth).

We also compared perceived discrimination in five discriminatory situations separately (Supplementary Table 4). In unadjusted analyses in England, being treated with less respect was the most prevalent type of discriminatory experience regardless of the attributed cause. In adjusted analyses, being treated with less respect was more often attributed to financial- [OR = 1.45, 95%CI (1.30; 1.63), p < 0.01], sex- [OR = 1.87, 95% CI (1.78; 1.96), p < 0.001], race- [OR = 1.57 95% CI (1.34; 1.84), p < 0.001], and weight discrimination [OR = 1.74, 95% CI (1.54; 1.97), p < 0.001], in England than in the US.

In the US, individuals perceived disability (86%) and sexual orientation (75%) discrimination most often in medical settings. Being harassed was the most prevalent discriminatory experience reported for financial- (73%), sex- (67%), race- (73%), and weight-based (75%) discrimination. In adjusted analyses, being harassed was attributed to financial- [OR = 0.61, 95% CI (0.55; 0.67), p < 0.001], race- [OR = 0.59, 95% CI (0.50; 0.69), p < 0.01] and weight discrimination [OR = 0.62, 95% CI (0.54; 0.71), p < 0.01] more frequently in the US than in England (Supplementary Table 4).

Cross-national differences stratified by SEP are reported in Table 3. Figure 2 illustrates the probability of perceived discrimination as a function of wealth as estimated from unadjusted logistic regression models (95% CIs are plotted in gray). In adjusted analyses, for the low SEP groups, discrimination due to financial status remained significantly higher in England than in the US for the low SEP groups but the cross-national differences for disability and weight were no longer significant (Table 3).

Looking in England alone, there was a significant inverse wealth gradient in disability (ß = −0.61, p < 0.001), financial status (ß = −0.90, p < 0.001), and weight discrimination (ß = −0.46, p < 0.001), whereby discrimination was more likely to be reported by lower than higher SEP groups (Table 4; Figure 2). This wealth gradient was not significant in the US sample. Findings were mostly similar when wealth was modeled as a categorical variable (Supplementary Table 5).

Perceived race discrimination was more common in the US than in England in low SEP groups but not in the high SEP groups, in unadjusted analyses. However, in SEP-stratified analyses adjusted for sex, there was no longer a significant cross-country difference in perceived race discrimination. There was no significant wealth gradient for the probability of perceived race discrimination in either England or the US (Table 4; Figure 2).

Sex discrimination was more commonly reported in the US for both high and low SEP individuals. Looking in the US alone, there was a positive wealth gradient (ß = 0.10, p < 0.01) in the probability of perceived sex discrimination, suggesting higher SEP individuals were more likely to perceive discrimination (Table 4; Figure 2). This wealth gradient for sex discrimination was not evident in the English sample. In adjusted analyses stratified by SEP, cross-national differences in sex discrimination remained significant for both the low SEP group and high SEP group (Table 3). Sex discrimination was significantly more prevalent in the US than in England in low SEP group as well as high SEP group when adjusting for age.

When testing the interactive effects of SEP and country on the prevalence of perceived discrimination attributed to each characteristic separately, there was a statistically significant interaction effect between SEP and country on the prevalence of perceived financial status discrimination among the overall sample (ß = −0.05, SE = 0.02, p < 0.01) but no other discrimination types (Supplementary Table 6).

The cross-national comparison stratified by SEP operationalized using education level did not differ from the main results (Supplementary Table 7). When education was included as a covariate, the differences in perceived sex discrimination were attenuated to non-significant (Supplementary Tables 8, 9). No other differences between this sensitivity analysis and the main findings were observed.

The weighted estimates, accounting for the selection bias in ELSA and HRS separately did not deviate from the main results, when assessing the association between wealth and perceived discrimination (Supplementary Table 10).

We found financial status discrimination to be more widespread in England than in the US, independent of age, sex, and wealth. Our stratified analyses suggest this result is driven by low SEP groups. This finding aligns with earlier cross-national research in these cohorts, which found higher rates of perceived age discrimination in low SEP groups in England than in the US (22). Wealth inequality is high in both England (31) and the US (26). The history of a hierarchically organized society in England (16) and the common belief that the US is more economically egalitarian (17) may partially explain the difference.

A growing literature suggests that LGB adults have poorer health and wellbeing than their heterosexual counterparts (32). We investigated rates of sexual orientation discrimination in lesbian, gay, and bisexual (LGB) adults and observed higher reports of sexual orientation discrimination in England than in the US. We were unable to conduct adjusted analyses due to small numbers identifying as LGB, so caution is needed in interpreting our findings. However, our findings add to the limited research on middle-aged and older LGB populations (14). Longitudinal work in ELSA indicates that perceived discrimination is associated with poorer wellbeing in LGB participants, particularly when the discrimination is attributed to sexual orientation (14). This suggests the age-related burden of poor health and wellbeing in LGB groups may be compounded by discrimination, especially in England. Older adults in England grew up at a time when homosexuality was classified as a mental illness and sex between two men was illegal (33, 34). Middle-aged adults grew up at a time when schools were banned from teaching about homosexuality (35) and there was the onset of the AIDS epidemic. These and other negative experiences may have had lasting effects on participants' perceptions of discrimination. However, the social context was also challenging in the US and the cross-national difference needs further research.

In unadjusted analyses, the prevalence of disability discrimination was higher in England than in the US. The English sample included more individuals living with disability compared to the US sample. This is in keeping with international estimates showing that a slightly greater proportion of the population of the UK is in receipt of disability payments than in the US (36). Similarly, there were more individuals who perceived disability discrimination in England than in the US. This finding was patterned by SEP, as the association for high SEP groups was attenuated to the null in stratified analyses. In the UK, approximately one in five people self-report disability (37) and 4 million older adults (36% of people aged 65–74, and 47% of those aged 75+) live with a limiting long-standing illness (38). The practical difficulties associated with disability may be aggravated by discrimination. Previous work in ELSA (11, 12) and other UK cohorts (6) suggests that perceived discrimination has compounding adverse effects on wellbeing in those with disability. Further, in England, reports of perceived discrimination were inversely related to wealth. Therefore, perceived disability discrimination, if unaddressed, could place an additional burden on marginalized aging English adults.

In unadjusted analyses, individuals with obesity perceived weight discrimination more frequently in England than in the US. This difference was socially patterned, as the association remained for low SEP groups alone in stratified analyses. There was no significant difference after adjustment for age and sex. Obesity is common in both countries, particularly in low SEP groups, though rates are consistently higher in the US (39, 40). The “normalization” of obesity may have resulted in reduced perceived discrimination based on this attribute in the US (27). In ELSA, weight discrimination has been prospectively linked with poorer wellbeing (13), and cross-national comparisons suggest that the impact of obesity on wellbeing is partially mediated by discrimination (9, 41). Weight-based perceived discrimination is associated with poor health behaviors (42) which may further harm health, particularly for marginalized groups, as there is a recognized wealth gradient in health behaviors (43).

Women in the US perceived sex discrimination more frequently than women in England, independent of age and wealth. This discrepancy may mirror differences in gender inequality between the countries, as in 2010 and 2020, the US ranked twice as high on the global gender inequality index as the UK (44), based on economic, educational, and political disparities. Further research is needed to evaluate the impact of political under-representation and economic and employment disempowerment on reports of sex discrimination in the population. We observed a wealth gradient in sex discrimination in the US. This is in keeping with earlier work in younger samples (7). It is unclear why high SEP women report more encounters with sexism. One possibility is that they may recognize sexism more readily, for example by being able to distinguish it from wealth-based discrimination. More research is needed to investigate this possibility.

In unadjusted comparisons, perceived racial discrimination was more common in the US than in England. This difference did not remain in adjusted analyses. Caution is needed in interpreting these findings due to the low proportion of ethnic minorities in both samples (25, 26). However, this was the most reported form of perceived discrimination in both English and US cohorts. This is concerning considering the increased recognition of the impact of racial discrimination on mental health (45, 46) as well as emerging evidence on physical health impacts (5, 47).

This study contributes to the evidence concerning the prevalence of distinct types of perceived discrimination in England and the US, using nationally representative samples of adults aged 50 and over. We have employed a harmonized measure perceived discrimination cross-nationally. In addition, cross-national differences were assessed stratifying by SEP, which may have improved precision and also elucidated a significant wealth gradient in perceived discrimination in middle-aged and older adults. However, the study is not without limitations. The analyzed sample had few ethnic minorities and LGB participants, reducing the generalizability of the findings. Unweighted estimated for the cross-national comparisons were used because ELSA and HRS are weighted differently. Study results reflect self-reported perceptions of discrimination rather than objective encounters with discrimination. In addition, as in other observational studies, the findings may be prone to the bias introduced by unmeasured confounding. Finally, future research is encouraged to include additional types of perceived discrimination (e.g., attributed to transgender and gender non-conforming status).