Correcting measurement errors in transition models based on retrospective panel data

We propose in this paper a dynamic n-state transition model to correct for measurement error, that could arise for example from recall and/or design bias, in retrospective panels. Our model allows the correction of measurement errors, when very little auxiliary information is available, over a long period of time taking into consideration the conjuncture fluctuations. The technique suggested shows that it is sufficient to have population moments (for at least one point in time) to correct over- or under-reporting biases. Using a Simulated Method of Moments, one can estimate a transition- and time-specific correction matrix for the labor market flows in a biased retrospective panel. Using retrospective and contemporaneous data from Egypt, we estimate the model and show the significance and robustness of our correction. We show through a reform evaluation that neglecting measurement error in the data would have produced significantly different and misleading results.