Modeling Household Poverty Status Using Repeated Cross-sectional Surveys

The authors propose a framework to estimate the probability of being poor in a dynamic setting based on a large information set that includes individual characteristics and macro‐economic variables. The joint inclusion of personal characteristics along with contextual factors allows separation of idiosyncratic shocks from aggregate shocks affecting poverty. The authors combine data from different cross‐sectional surveys and fit a dynamic logistic hierarchical model within a Bayesian framework using standard Markov chain Monte Carlo techniques. The authors’ approach is exemplified by estimating household poverty status in Kyrgyz Republic as a function of time, regions, country, regional level variables and household level socio‐demographic characteristics.