Measurement of Chronic and Transient Poverty: Theory and Application to Pakistan

This paper investigates how to characterize each person's poverty status when his/her welfare level fluctuates and how to aggregate the status into chronic and transient poverty measures. The contribution of the paper is to clarify the sensitivity of relative magnitudes of chronic versus transient poverty to the choice of a poverty measure. We show this by theoretically re-examining Ravallion's (1988) decomposition of the expected value of a poverty measure into chronic and transient components. The examination covers major poverty measures including those developed by Foster et al. (1985), which are used extensively in the existing studies. Our analysis shows that the chronic-transient decomposition using the squared poverty gap index might be too sensitive to the poverty line and that the index is justified only if we accept that the welfare cost of consumption fluctuation is independent of the depth of chronic poverty. If we instead believe that the decomposition should not be too sensitive to the poverty line and that the welfare cost of risk is more severe when an individual's chronic poverty is deeper, other poverty measures such as suggested by Clark et al. (1981) are useful. We also investigate how empirically different are the relative magnitudes of chronic versus transient poverty, depending on the choice of a poverty measure. Based on a two-period household panel dataset collected in Pakistan, we show that the difference is substantial even when the poorest experienced only a small fluctuation in their consumption.