Poverty Dynamics in Ethiopia: State Dependence and Transitory Shocks

Abstract

This paper focuses on the persistence of poverty in rural and urban households in Ethiopia by estimating dynamic probit models. The model allows for unobserved heterogeneity, first order state dependence and serially correlated error components. The dynamic probit model of poverty that controlled for household heterogeneity and serial correlation performed better in explaining the dynamics of poverty in Ethiopia. In rural areas, the effect of controlling for heterogeneity and serial correlation was, typically, to increase the coefficient of the true state dependence one fold. In urban areas controlling for transitory shocks brought out, more strongly, the effects of differences in towns of residence on the incidence of poverty, while it reduced the importance of exogenous household attributes such as ethnicity, age and family background. Transitory shocks also contributed to poverty persistence in two additional ways. First, the persistence of urban poverty increased dramatically once we controlled for transitory shocks. Secondly, intrinsic risk of falling into poverty also declined substantially implying only a tiny fraction of the urban population would be at risk of falling into poverty were it not for transitory shocks.