Tactical asset allocation in a real-life setting

Abstract

This thesis will test one of the most popular market-timing strategies, using the longest available data set ranging from 1857 to 2012. The market-timing strategy has already proven to deliver superior results in the period 1926-2012 in a back-test. Which is why, the performance of the period pre-1926 will be compared to the post-1926 performance in a back-test. The performance of the two periods is similar, but period 1857-1925 is found to have the greatest improvements when the strategy is tested, which I find to be due to the lack of long consecutive bull markets. In both time periods the market-timing strategy is providing a small increase in returns while decreasing the volatility significantly when compared to a passive buy-and-hold strategy. In order to minimize the potential data-mining bias that all in-sample technical analysis struggle with, an out-of-sample simulation method is tested on the entire data-set and it is found that the performance is poorer than what was found in a back-test. The reason for the out-of-sample deterioration is mainly found to be due to changes of the optimal moving-average length.