TESTS OF 3 PARITY CONDITIONS - DISTINGUISHING RISK PREMIA AND SYSTEMATIC FORECAST ERRORS

This study provides evidence that uncovered interest differentials bet ween currencies cannot be attributed to either risk premia or systemat ic forecast errors alone. The evidence is based on developing joint te sts of three parity conditions: uncovered interest parity, purchasing power parity, and real interest parity. The study tests parameter rest rictions based on knowing that risk premiums only affect nominal and r eal interest differentials, but not inflation differentials, while sys tematic errors in forecasting exchange rates only affect nominal inter est differentials and inflation differentials, but not real interest d ifferentials. The study finds a variety of cases where deviations from some or all of the parity conditions can be systematically related to variables currently known to investors and borrowers. The most intrig uing evidence is that for some currencies, these deviations from parit y are perfectly correlated as if they were driven by a common factor. (C) 1997 Elsevier Science Ltd.