Some specification tests of uncovered interest parity

Recent exchange rate theorising has relied increasingly on some version of the uncovered interest parity (UIP) condition. This reliance reflects a belief that the UIP condition isa better building block for an exchange rate model than purchasing power parity, which has demonstrably not held for the recent floating period1. But is UIP a better building block? The empirical evidence on UIP is somewhat mixed - some researchers find in favour of the hypothesis whilst others argue that it is not supported by the data. Clearly the validity of UIP is of considerable interest and importance since a number of celebrated results, such as the Dornbusch (1976) overshooting story, depend on it. In this paper we re-examine the UIP condition for four major currencies using weekly data. Our study has a number of novel features. Thus we conduct specification tests using residuals from the LUS class in addition to the OLS class and implement error orthogonality tests which, in contrast to many of the extant empirical efficiency tests, use a specific selection criterion. We also consider whether or not errors which arise in the market for one currency have an information role in the markets for other currencies, this is assessed through semi strong error orthogonality tests