Unbiased estimation of the half-life to PPP convergence in panel data

Three potential sources of bias introduce complications for panel data estimation of the half-life of purchasing power parity deviations. They are bias induced by inappropriate cross-sectional aggregation of heterogeneous coefficients, small-sample estimation bias of dynamic lag coefficients, and bias induced by time aggregation of commodity prices. All of these biases have been addressed individually in the literature. In this paper, we address all three biases in arriving at our estimates. Using an annual panel data set of CPI-based real exchange rates for 21 OECD countries from 1973 to 1998, we obtain a point estimate of the unbiased half-life of 3.0 years with a 95% confidence interval of 2.3-4.2 years.