We test for the existence of poverty traps and distribution-dependent growth using a nonlinear dynamic panel data model of household incomes allowing for endogenous attrition. Our estimates for Hungary and Russia in the 1990s reveal significant nonlinearity in the dynamics, consistent with the claim that income inequality attenuates growth in mean income. However, we do not find evidence of a threshold effect at low incomes, as postulated by models of dynamic poverty traps. Our results indicate that households generally bounce back from transient shocks, though we find that the adjustment process is slower for households who are poorer in steady state.