Data Snooping and Market-Timing Rule Performance

Abstract

We reassess the performance of market-timing rules when controlling for data-snooping biases. For the first time, a comprehensive set of simple and complex market-timing rules is examined and tested for statistical signi ficance, using the White (2000) 'Reality Check, the Hansen (2005) SPA test, as well as their stepwise extensions by Romano and Wolf (2005) and Hsu et al. (2009). Even though individual market-timing rules signi ficantly outperform a buy-and-hold strategy at both daily and monthly frequencies when considered in isolation, their outperformance, generally, does not remain signi ficant after correcting for data snooping. Relative to the alternative of investing in the risk-free rate, however, we find signi ficant outperformance of the best rules, even after data-snooping adjustment, when testing at a monthly timing frequency.