Poverty decomposition by regression
An application to Tanzania
We develop a poverty decomposition method that is based on a consumption regression model. Because this method uses an integral of the partial derivatives of a poverty measure with respect to time, the resulting poverty decomposition satisfies time-reversion consistency and sub-period additivity. Unlike the existing poverty decomposition methods, it allows us to ascribe the observed change in poverty to various covariates of interest collected at a disaggregate level. This method is applied to two datasets from Tanzania to assess, among others, the short- and long-term impacts of infrastructure and market access on poverty.