High Frequency Interest Rate Differentials and Long Memory Property in Forward Premium Anomaly

This paper considers high frequency time series for the US and German 30 day Eurobond interest rate differential. The data are used to address the issue of the order of integration of the forward premium, which is at the heart of the controversy concerning the forward premium anomaly. Extraction of the intraday periodicity in the high frequency forward premium presents unusual difficulties due to the very persistent autocorrelation in both the first two conditional moments. Parametric ARFIMA-FIGARCH long memory model in both the first two conditional moments is found to represent high frequency forward premium quite well, and are similar to results from semi parametric local Whittle estimation. There is strong evidence of long memory effects in both the first two moments of the forward premium at very short spans of high frequency data. The empirical results are consistent with long memory being an intrinsic property of the markets. The results indicate that the “lack of balance” in the Fama type forward premium regression and the consequent rejection of uncovered interest rate parity (UIP) may well be an inherent feature of the DM-$ currency markets.