Abstract

Although the literature on leveraged exchange-traded funds (LETFs) concurs with the idea that they are short-term investment tools, recent studies offer some investment strategies for them that are also profitable in the long term. These strategies, however, are typically only tested on a limited number of highly traded LETFs. This study uses different types of LETFs to examine various portfolios with different combinations of bull and bear LETFs, to find the best investment strategy in the long run. It then uses different machine learning techniques to analyze which factors define the best investment strategies, with portfolios being rebalanced on a quarterly and annual basis. The sample of this study consists of 44 pairs of LETFs from 2012 to 2020 that have different underlying assets and leverage levels. The results reveal that short-selling the combination of both bull and bear LETFs does not yield a significant positive return compared to the market, however, the return generated from short-selling a portfolio with only bear ETFs can significantly beat the market, especially when the market is bullish. The quarterly and annual results are consistent and show that short-selling a full bear portfolio is the winning strategy in both of these intervals. Moreover, the results show that as the correlation of LETFs with their underlying index increases, the return of short-selling both bull and bear LETFs decreases. At the same time, an increase in the net asset value of bull LETFs results in an increase in the return of short-selling bull LETFs and a decrease in return of short-selling the bear LETFs.