Two theories have been proposed to explain the observed association between depression and poverty, namely social causation and social drift. Little is known regarding the relative importance of social causation and social drift in low and middle-income countries, where poverty is more severe and where most of the world's depressed individuals live. Methods We analysed nationally representative longitudinal data from the National Income Dynamics Study in South Africa and simultaneously tested social causation and social drift hypotheses using structural equation modelling across three waves. Results Worse individual economic status at time 1 and 2 was independently associated with worse depression two years later at time 2 (standardised linear regression coefficient ß = -0.110, Standard Error (SE): 0.024) and four years later at time 3 (ß = -0.113, SE: 0.025) respectively. Conversely worse depression at time 1 and time 2 was independently associated with worse economic status at time 2 (ß = -0.037, SE: 0.016) and time 3 (ß = -0.028, SE: 0.012) respectively. In addition, the "effect" of depression on future assets was stronger among people with less baseline assets. Limitations The time span between data rounds is relatively short (four years); response rates are unequal across ethnic, age and sex groups; and the measure of depression is based on self-report. Conclusions Social causation and social drift act simultaneously in this population, reinforcing poverty/depression cycles. Multi-sectoral policies are required that both prevent depression by addressing its economic determinants, and provide evidence-based treatment to mitigate the economic impact of depression.