Climbing Out of Poverty, Falling Back In: Measuring the Persistence of Poverty over Multiple Spells

This paper investigates the persistence of poverty over individuals' lifetimes using a hazard rate, or spells approach. Previous research on poverty dynamics using the spells approach has been limited by its failure to take into account multiple episodes of poverty. I estimate hazard models for exiting from and for returning to poverty and use the estimated parameters to calculate distributions of total time spent in poverty over multiple spells, using longitudinal data from the Panel Study of Income Dynamics. These models incorporate observable personal and household characteristics, as well as unobserved heterogeneity. My findings emphasize the importance of considering multiple spells in an analysis of poverty persistence. For black and white individuals falling into poverty in some year, approximately 50 and 30 percent, respectively, will have family income below the poverty line in at least five of the next ten years. A single spells approach predicts comparable figures of only 26 and 13 percent. To check the robustness of these predictions I also utilize two alternative approaches -- direct tabulations from panel data and estimation of a components-of-variance model -- and compare predictions of poverty persistence based on the three methods.