CIREQ Lunch Seminar: Shaimaa Yassin
"Correcting Measurement Errors in Transition Models Based on Retrospective Panel Data"
(Co-Authored with François Langot)
This paper offers a method to correct for measurement errors that could arise when estimating transition rates and stocks relying on retrospective panels. A dynamic n-state transition model is built to allow for the correction of errors in both transitions and stocks, when very little auxiliary information is available, over a long period of time, and taking into consideration the macroeconomic cyclical fluctuations. The method is particularly useful for the assessment of labor market dynamics in developing countries where frequent data collection is costly and recalled information raise to be the only candidate for these research purposes. It is sufficient to have validation population moments for only one point in time to estimate the measurement errors in our model. Using a Simulated Method of Moments, a transition- and time-specific correction matrix is estimated, to correct both labor market flows and stocks, in a biased retrospective panel. Recalled and validation data on workers' flows and stocks from Egypt are used to show the significance and robustness of our correction method. A reform evaluation application is also used to show the significance of maintaining the cyclical information in the data by relying on a dynamic transition model.