This study examines the convergence rate of mean reversion by estimating the half‐lives of sectoral real exchange rates using an extensive product price panel for Japan (with the USA as the numéraire). We find that the half‐lives of sectoral real exchange rates are remarkably distorted when the grouped half‐life is measured inappropriately and the cross‐sectional dependence and potential trend breaks are ignored. After taking account of these problems, the bias‐corrected half‐life for all goods is as low as 3.00 years, close to the bottom of the consensus view of 3 to 5 years. Moreover, the bias‐corrected half‐life of mean reversion is 2.40 years for traded goods, and only approximately half that for non‐traded goods. Finally, our findings also support the view that small‐sample bias correction is critical for half‐life estimations.

No abstract