Measuring Half-Lives Using A Non-Parametric Bootstrap Approach

Abstract

In this paper we extend the Murray and Papell (2002) study by using a non-parametric bootstrap approach which allows for non-normality, and focusing on quarterly real exchange rate in twenty OECD countries in the post-1973 floating period. We run Augmented Dickey-Fuller (ADF) regressions, and estimate the half-lives (and confidence intervals) from the corresponding impulse response functions. Further, we use an approximately median-unbiased estimator of the autoregressive parameters, and report the implied point estimates and confidence intervals. We find that accounting for nonnormality results in even higher estimates of the degree of persistence of PPP deviations, but, as in Murray and Papell (2002), the confidence intervals are so wide that no strong conclusions are warranted on the existence of a PPP puzzle.