This study investigates the sensitivity of the long-term return anomaly observed in the US$/Yen currency market to sample and method bias during the period from 11 October 1983 to 21 July 1999. Initially the CUSUM statistic is employed to identify subperiods of sign shifts in the mean returns. Then the dependence in these subperiods is investigated using the Hurst [Hurst, H.E., 1951. Trans. Am. Soc. Civil Eng. 116, 770-799] and the Lo [Lo, A.W., 1991. Econometrica 59 (5), 1279-1313] rescaled range statistics. The results suggest that rejection of the random null is conditional on both the procedure and the period being tested. The Hurst test may incorrectly lead to a rejection of the null hypothesis. However, the Lo test is a more appropriate approach to the data in this study. Thus previous research based on the Hurst test may have inadvertently introduced a method bias.
|Number of pages||13|
|Journal||Japan and the World Economy|
|Publication status||Published - Dec 2000|
- Long-term memory
- Rescaled range statistic