5. Results 5.1. Descriptive analysis Table 1 presents descriptive associations between perceived discrimination and adolescents' mental health outcomes in the follow-up survey. Results show that adolescents who perceived more societal discrimination (i.e., whose score was above the median) were more likely to report depressive symptoms in the top 50th, 25th, and 10th percentiles of the sample when compared to adolescents who perceived less societal discrimination (i.e., whose score was at or below the median). For instance, 47% of adolescents who reported more societal discrimination scored in the top 50% of depressive symptoms, compared to 37% who reported less societal discrimination. Similarly, adolescents who reported interpersonal discrimination, representing 57% of our sample (see Table A.1 in the appendix), were also more likely to fall into the top of the distribution of the depressive symptoms when compared to adolescents who did not report interpersonal discrimination. We found a similar pattern of results for self-esteem but without statistically significant differences between groups. 5.2. Multivariate analyses Table 2 presents the results of the logit regressions for both mental health outcomes. The first three columns present regressions using the three binary cut-off values defined for depressive symptoms (see description in Section 4.2) and the last three columns present regressions using the three binary cut-off values defined for self-esteem. Panel A presents the results for perceived societal discrimination and Panel B, for perceived interpersonal discrimination. We do not present all parameters estimated due to space limitations, but they are available upon request. Table 2 presents the results for our main independent variables of interest (i.e., perceived societal and interpersonal discrimination) and SES in terms of odd ratios, along with their interaction as a ratio of odd ratios, and the baseline odds. We note that the baseline odds refer to the group of adolescents who have a value of zero on all the independent variables (hereafter “baseline group”).5 According to the first column of Panel A, within the baseline group we expect to find 0.51 adolescents reporting depressive symptoms scores in the top 50% for every one adolescent reporting depressive symptoms scores in the bottom 50%. In terms of discrimination, each one standard deviation increase in perceived societal discrimination corresponds to an increase in the baseline odds of falling in the top 50% of depressive symptoms by 1.27 (i.e., it increases 27%). The ratio of odd ratios presented for the interaction between perceived societal discrimination and SES indicates that the effect of perceived societal discrimination on depressive symptoms attenuates by a factor 0.94 (i.e., minus 6%), with each one standard deviation increase in parental SES In the second and third columns, we find similar odd ratios for perceived societal discrimination. Contrary to results in the first column, we find that the odd ratio for SES is statistically significant and positive only when predicting the top 10th percentile of depressive scores and the interaction between discrimination and SES is not significant. In columns 1–3 of Panel B, we find that adolescents who report experiencing interpersonal discrimination are significantly more likely to score in the top 50th, 25th, and 10th percentiles of depressive symptoms. As with perceived societal discrimination, we note a signifi- cant association between SES and depressive symptoms only when predicting the top 10th percentile of depressive scores, and a significant interaction between discrimination and SES only when predicting the likelihood of scoring above the median of depressive symptoms. Columns 4–6 of Table 2 present the results for the self-esteem. Panel A shows that, like findings for depressive symptoms, perceived societal discrimination is associated with the likelihood of falling below the median score of self-esteem. However, this pattern does not hold true when predicting the top 25th and 10th percentiles. SES is inconsistently associated with self-esteem in columns 4 and 6, and the interaction between perception of societal discrimination and SES is not statistically significant in any column. Panel B shows that the association between interpersonal discrimination and low self-esteem scores (in the bottom 25 and 10% of the sample) is positive and statistically significant for adolescents reporting average SES. The interaction between interpersonal discrimination and SES is generally not significant, with one exception: SES moderates the effect of discrimination on the odds of falling in the bottom 50% of self-esteem scores