Timing the US Stock Market Using Moving Averages and Momentum Rules: An Extensive Study

Abstract

In this thesis we investigate the performance of moving average and momentum strategies by simulating returns, both in-sample and out-of-sample, while simultaneously taking into account important market frictions. We do so for two stock indices and four stock portfolios, at daily and monthly frequency, in the period from 1928 to 2015. This is carried out in order to examine if the active strategies outperform the passive benchmark on a risk-adjusted basis, and to see if the trading rules pro table when tested in-sample also are pro table out-of-sample. In addition, and for the rst time, we examine the relevance of data frequencies in out-of-sample testing. A stationary block bootstrap methodology is adopted in order to evaluate the statistical signi cance of the risk-adjusted performance, measured by the Sharpe ratio. We nd that in-sample pro table trading rules perform poorly when tested out-of-sample. However, we are able to nd statistically signi cant outperformance when trading in small-cap stocks; yet, the outperformance disappeared in recent past. Moreover, we investigate how the performance depends on the split point between the in- and out-of-sample period and the length of the in-sample period. We nd that the performance of an out-of-sample test highly depends on the choice of split point as well as in-sample period length. Consequently, the out-of-sample testing procedure is not a complete remedy for the \data mining bias". Finally, we are not able to nd conclusive evidence suggesting any bene t of trading more frequently. Key words: technical analysis, market timing, moving averages, time-series momentum, out-of-sample simulations, trading frequency