Evidence across disciplines suggests that economic well-being a?ects an individual’s psychological well-being, and also that psychological disorders can have substantial negative e?ects on individual income. Together, these studies suggest a feedback loop that may trap some in poverty. However, estimating the causal links between income and psychological well-being is di?cult due to this simultaneous causality. In this paper, I overcome this endogeneity with a panel GMM approach to estimate a dynamic system of equations that identi?es both causal links. Using a nationally representative panel dataset from South Africa, I ?nd evidence of impacts in both directions. The average e?ect of changes in psychological well-being on income is mainly driven by a large e?ect among those near the threshold for clinical depression. Meanwhile, the e?ect of changes in income on psychological well-being is especially pronounced among the poor, indicating the possibility of a strong feedback loop among an especially vulnerable subgroup – the poor with low levels of psychological well-being. An impulse response function analysis indicates that this bi-directionality can nearly double the long-term impact of shocks, while simulations show that this relationship can explain prolonged poverty spells and low resilience to shocks. A formal test for poverty traps shows that individuals with low levels of psychological well-being exhibit markedly different income dynamics that suggest the existence of a multi-equilibrium poverty trap.