The forward premium anomaly: statistical artefact or economic puzzle? New evidence from robust tests

Recent literature has questioned statistical inference in predictive regressi on with persistent regressors, suggesting a possible explanation for puzzles suc h as the forward premium anomaly. We therefore revisit this puzzle using three a lternative econometric methods known to provide reliable inference in the presen ce of persistent conditioning variables. While they provide less evidence agains t forward rate unbiasedness than traditional predictive regression tests, we sti ll reject using at least one method for all six currencies. Thus, while the econ ometric problems inherent in predictive regression likely play a role in this an omaly, we are left with an <i>economic</i> puzzle even after accounting for their influence.