Can the Limitations of Panel Datasets be Overcome by Using Pseudo-Panels to Estimate Income Mobility?

This paper analyzes whether pseudo-panels are suitable substitutes for true panels for estimating income mobility. We obtain evidence using Chilean panel data for the period 1996-2006 and constructing pseudo-panels treating each round of the panel as if it were an independent cross-section survey. We consider three different pseudo-panel methods: the mean-based approach that identifies cohorts and follows cohort means over time, the method developed by Bourguignon, Goh and Kim (2004) that was designed to estimate vulnerability-to-poverty measures, and the method of Dang, Lanjouw, Luoto and McKenzie (2011) that predicts a lower and upper bound for the joint probabilities of poverty status in t=1 and t=2. The empirical evidence leads us to conclude that pseudo-panel methodologies do not perform well in this task. Our results indicate that pseudo panels fail in two respects when trying to predict the income mobility pattern observed in Chile. First, they do not give good results for the mobility concept each pseudo-panel method seeks to measure. Second, they also perform poor in predicting a broader set of income mobility measures. We complete the analysis making a final point about the calculation of poverty transition rates using pseudo-panels. Email addresses: [email protected] (Guillermo Cruces), [email protected] (Gary Fields), [email protected] (Mariana Viollaz).