Quantifying the "Peso Problem" Bias: A Switching Regime Approach.

This paper provides an estimate of the peso problem bias in relation to the forward discount puzzle. we argue that the EMS provides a natural laboratory for studying such effects. We suggest an endogenous, data-driven procedure to classify those ocservations, that are to be deleted from the regression of the exchange rate returns on the appropriate interest rate differential in order to isolate the peso problem bias. We then apply a pooled regression technique to obtain reasonable efficient slope estimates, and find the estimate of the peso problem bias to be equal to -.6, statistically as well as economically significant.