Depression is one of the most common mental disorders, affecting 17% of adolescents and 8% of adults in the U.S. (NIMH, 2022). Adolescents with depression are about three times more likely to be depressed in adulthood compared to non-depressed adolescents (Johnson et al., 2018). Understanding the determinants of adolescent depression and the role of the adolescent period in shaping later-life mental health is key for informing policy interventions and treatments as well as curbing the sharp rise in both adolescent and adult depression observed over the last several decades. Motivated by a large body of evidence that suggests peer influence peaks during adolescence, I explore how adolescent peers’ underlying risk for depression impacts own mental health.

Specifically, I examine whether peers’ genetic predisposition to depression affects one’s own mental health in the short- and long-term. Peers’ genetic predisposition to depression may influence own mental health via peers’ depression as well as peers’ behaviors (e.g., substance use, interpersonal conflict).1 This study uses peers’ genetic predisposition to depression to explore broader peer effects, given the inability to explicitly disentangle these two pathways. I use data from the National Longitudinal Study of Adolescent to Adult Health (Add Health), which follows a nationally representative sample of US adolescents starting in the 1994-95 school year. The genetic data in Add Health include the polygenic risk score for major depressive disorder (hereafter, MDD score), a composite measure of genetic markers that are correlated with MDD. A higher MDD score means a higher genetic risk for depression. I define peers as same-gender grademates and exploit variation in peers’ average MDD score within schools and across grades while controlling for own genetic risk for depression.

There are three well-known challenges in identifying the causal effects of peers—the reflection problem, endogenous peer group formation, and common environments. The reflection problem arises when estimating own behavior as a function of average group behaviors because it is difficult to disentangle the effects of average group behaviors (e.g., average peer depression) and average group characteristics (e.g., peers’ average parental income) on individual behavior (e.g., own depression) since they move together in a linear-in-means model (Manski, 1993). This is not a concern in my case since genes are not affected by others’ behaviors or characteristics. Concerns about endogenous peer group formation arise because individuals tend to befriend others who have similar observed and unobserved traits. To address this issue, I rely on cohort-to-cohort variation in the average MDD score within a school (Hoxby, 2000, Hanushek et al., 2003, Angrist and Lang, 2004, Lavy and Schlosser, 2011). While parents might select a school for their children based on observed characteristics, the assignment into each grade within a school is determined by age, making the formation of grademates as good as random. Another challenge arises from the fact that peers share common environments, which may result in similar behaviors and outcomes. While school-grade correlated effects cannot be completely ruled out, I control for school and grade fixed effects and include numerous school-grade level controls to alleviate these concerns.

I find that having same-gender grademates with higher genetic risk for depression during adolescence immediately deteriorates own mental health. A one standard deviation increase in peers’ average MDD score significantly increases the likelihood of being depressed by 1.9 and 3.8 percentage points for adolescent girls and boys, respectively. These effects translate to 7.2% and 25% increases in depression relative to the sample means for females and males, respectively. To contextualize these estimates, having a mother with a college degree or higher is associated with 5% and 7% lower probabilities of being depressed among adolescent girls and boys, respectively.2 The peer effects persist into adulthood, but only for females. A one standard deviation increase in peers’ average MDD score during adolescence leads to a statistically significant 2.8 percentage point increase in the probability that a female is depressed in adulthood, a 13.6% increase. For comparison, having a mother with a college degree or higher is associated with only a 2% reduction in the probability of being depressed among adult females. These findings suggest that depression in adolescence and adulthood is influenced not only by one’s own genetic risk for depression, but also by the genes of those around us. In other words, there are important social-genetic effects in the context of mental health.

The results are robust to alternative measures of depression and specifications, and I provide various pieces of evidence in support of the main identifying assumption that within-school across-grade variation in peers’ genetic predisposition to MDD is as good as random. I also explore nonlinear effects of peers’ genetic predisposition to depression. Both in the short- and long-term, having peer groups with relatively high average MDD scores increases the likelihood of experiencing depression for females, but not males.

Next, I explore mechanisms that could drive the link between peers’ genes and own depression, including friendship, substance use, educational attainment, and labor market outcomes. Friendship is a natural mechanism to consider since peers with higher underlying risk for depression may reduce the quantity and quality of social ties. I find that having peers with high genetic risk for depression worsens friendship and social connectedness both in the short- and long-term. For both adolescent girls and boys, an increase in same-gender grademates’ average MDD score decreases the probability of spending time with friends. The results also indicate that being exposed to peers with high genetic risk for depression during adolescence reduces the frequency of hanging out or communicating with friends in adulthood for both genders, the number of high school friends that females still have as adults, and the number of close friends for females in adulthood. Thus, weaker social ties may be an important channel that explains the baseline effects, especially for females.

Substance abuse is often associated with depression, and may be another channel linking peer genetic risk for depression and own depression. I find evidence that having peers with high genetic risk for depression increases substance use, particularly for females. An increase in same-gender grademates’ average MDD score increases the frequency of binge drinking among females in both adolescence and adulthood, and also generates a modest increase in female marijuana use in adulthood. There is no evidence of increased substance use among males. Finally, I study effects on socioeconomic outcomes, including college attendance, employment, and labor income. Males who had same-gender grademates with higher genetic risk for depression during adolescence are less likely to attend college, while females are less likely to be employed. The findings indicate that substance abuse and lower socioeconomic status may be additional underlying channels that explain the persistence of the social-genetic effects into adulthood.

This paper contributes to the literature in several ways. First, I add to a growing literature studying peer effects on depression. While a significant correlation between peer and own mental health is well documented in the psychology and medical literatures, only a handful of studies explore causal peer effects on mental health (Eisenberg et al., 2013, Zhang, 2019, Giulietti et al., 2022).3,4 Eisenberg et al. (2013) and Zhang (2019) find minimal or no short-term effects of peers’ mental health on own mental health using variation generated from random assignment of college roommates in the US and random assignment of junior secondary school students to classrooms in China, respectively. Different than Eisenberg et al. (2013) and Zhang (2019), I focus on the role of adolescent peers in the US context. Most similar to my work is Giulietti et al. (2022), which examines the long-term effects of peers’ depression on own depression using Add Health. They find that an increase in the share of same-gender grademates in adolescence who are depressed significantly increases females’ likelihood of being depressed in adulthood. My analysis differs from that of Giulietti et al. (2022) in several ways. First, because I focus on peers’ genetic risk for depression, I can identify the contemporaneous effects of adolescent peers on own mental health, whereas (Giulietti et al., 2022) cannot identify such effects due to the reflection problem. Second, I explore a wider range of potential mechanisms, including friendship and socialization, substance use, and socioeconomic status. My results complement those of Giulietti et al. (2022) and suggest that social-genetic effects may be an important factor underlying their findings.

This work also contributes to a growing literature on social-genetic effects. The importance of genetics in mental health has been well recognized (Abkevich et al., 2003, Greene and Vostanis, 2007, NIMH, 2020), but little is known about the indirect effects of the genetic makeup of those around us (i.e., social-genetic effects) (Baud et al., 2017, Domingue and Belsky, 2017, Cawley et al., 2019, Sotoudeh et al., 2019, Cawley et al., 2023). Most studies on social-genetic effects focus on genetically-related groups such as families and relatives.5 Recently, researchers have examined indirect genetic effects using genetically-unrelated groups such as friends and classmates (Domingue and Belsky, 2017). For instance, Sotoudeh et al. (2019) find a significant causal effect of peers’ genetic risk for smoking on own smoking behavior in adolescence. Brunello et al. (2020) explore the short- and long-term effects of peers’ genetic risk for body mass index (BMI) on own BMI. They find significant short-term peer effects on BMI for females, with no effects for males. My analysis contributes to this small but growing literature on genomic effects beyond the family. The results imply that there are significant social-genetic effects in the context of mental health, which are stronger among females, consistent with the findings in Brunello et al. (2020).

The remainder of this paper is organized as follows. Section 2 includes background information on polygenic scores. Section 3 describes the data, sample construction, mental health outcomes, and descriptive statistics. In Section 4, I discuss the empirical strategy and identifying assumptions. Section 5 reports baseline results, and in Section 6, I explore and discuss possible mechanisms underlying the baseline effects. Section 7 discusses robustness of the baseline results. Concluding thoughts are in Section 8.