Adolescent Mentoring and Adult Depressive Symptoms: A Propensity Score Weighted Analysis of Add Health Data

Introduction: Mentoring relationships are widely viewed as important developmental resources for adolescents, with prior research linking mentoring to improvements in academic engagement, psychosocial adjustment, and behavioral outcomes. However, less is known about whether mentoring has lasting implications for mental health into adulthood. In addition, access to mentoring is not evenly distributed across young people. Adolescents’ likelihood of having a mentor may be shaped by individual, family, school, and neighborhood characteristics, raising important questions about selection bias in estimating the long-term effects of mentoring. This study examines the association between adolescent mentoring and adult depressive symptoms using a nationally representative longitudinal dataset and a propensity score weighting approach designed to strengthen causal inference.

Methods: Data were drawn from the National Longitudinal Study of Adolescent to Adult Health (Add Health). Mentoring was measured retrospectively at Wave III, based on whether respondents reported having an adult who made a positive impact in lives after age 14. Adult depressive symptoms at Wave IV served as the primary outcome. To address nonrandom selection into mentoring, this study estimated multilevel propensity scores using logistic regression models with a random intercept for school, considering the school-cluster sampling strategy of Add Health. Inverse probability of treatment weighting (IPTW) was then applied to create a weighted analytic sample balanced on a broad set of baseline covariates, including demographic characteristics, academic performance, school attachment, self-esteem, early depressive symptoms, family socioeconomic resources, and neighborhood conditions. After weighting, weighted linear regression models were estimated to examine the association between mentoring and adult depressive symptoms.

Results: Access to mentoring was patterned by both individual and contextual characteristics. Indicators of adolescents’ socioeconomic status at both the family and community levels were strong determinants of reporting mentoring relationship. Race was also a significant determinant, with Latinx and Asian Americans being less likely to report a mentoring relationship even after controlling for socioeconomic factors, whereas the disparity for African American adolescents disappeared in multivariate models. After IPTW, balance across observed covariates improved substantially, and mentoring was significantly associated with lower depressive symptoms in adulthood. Interaction analyses further indicated that the association between mentoring and depressive symptoms did not vary by race or immigration status.

Discussion and Implication: This study contributes to social work research in several ways. It extends the mentoring literature by focusing on adult mental health rather than short-term adolescent outcomes alone and highlights the importance of addressing selection into mentoring when evaluating its effects. Also, by incorporating both individual- and community-level characteristics into the propensity score framework, the study underscores that mentoring opportunities are shaped not only by youth characteristics but also by institutional context. Although the association between mentoring and mental health appears consistent across racial and immigration groups, access to mentoring is not equally distributed, and disparities in access may reproduce broader inequalities in mental health and life chances. From a social work perspective, expanding access to mentoring relationships, particularly for youth facing structural disadvantage, may serve the common good by promoting more equitable pathways into adulthood mental health.