Long Swings in Exchange Rates
检验了美元汇率是否存在长期波动，发现基于季度数据的检验结果不稳健，但使用周度数据后能显著拒绝随机游走假设，支持长期波动的存在。
Several authors have reported evidence of long swings in U.S. dollar exchange rates by rejecting the random walk in favor of a Markov regime-switching model. We show that this evidence is not robust to an extension of the sample period. One should, however, not conclude that long swings are thus absent, because the tests may have insufficient power. A possible reason for this is the low data frequency, because existing tests are based on quarterly data, and we find that the power can be increased substantially by raising the data frequency. Indeed, weekly data reject the random walk, so we conclude that long swings are in the data. This conclusion is supported by significant superiority of regime switching over random walk out-of-sample forecasts of the direction of change.