Each year, 3–5% of individuals suffer from depression. Approximately 50% will experience persistent disorder; however, factors associated with depression persistence in the population are not yet fully understood. We tested the hypothesis that long-term depression course is predicted by socioeconomic position. Data come from the Gazel cohort study, which began in 1989 (n = 20 624,35–50 ans, 73% of men). We included 13,142 participants who reported symptoms of depression (CES-D) in 1996, 1999, 2002, 2005 et 2008. These repeated measures were dichotomized (high symptom level : yes vs no) and we identified four longitudinal trajectories (no depression, decreasing symptoms, intermediate/increasing symptoms, persistent depression). Socioeconomic position was measured by occupational grade in 1996 (manual worker/clerk, administrative associate/technician, vs. manager). Analyses were conducted using multinomial regression models, stratifying on sex and controlling for socio-demographic characteristics, negative life events, health behaviors, and preexisting health problems. The probability of being depressed during follow-up followed a socioeconomic gradient. For instance, compared to managers, administrative associates/technicians were more likely to have persistent depression (fully adjusted ORs: men: 1.28, 95% CI 1.05–1.56; women: 1.86, 95% CI: 1.05-3.29) as were manual workers/clerks (fully adjusted ORs: men: 2.32, 95% CI 1.76–3.06; women: 2.63, 95% CI 1.41-4.92). This socioeconomic gradient is consistent with a social causation explanation of persistent depression. The implication for clinicians and public health decision makers is that efforts aimed to reduce the burden of depression should take into account mental health needs across the population, rather than solely focus on high-risk groups.