Do Institutional Investors Have Market Timing Skills? Evidence from ETFs

Abstract

Using quarterly institutional holdings and a large proprietary database of institutional trades in exchange-traded funds (ETFs), we investigate whether institutional investors have market and sector timing skills. The question of whether institutional investors have superior market timing skills has been intensely debated ever since Treynor and Mazuy’s (1966) seminal study. Studies that use non-linear regressions to investigate market timing generally find little evidence of skill. Furthermore, recent literature is often critical of this regression framework. We add to this literature by using an innovative approach to directly measure the ex-post performance of institutional investors’ market timing strategies. That is, we test whether institutional investors exhibit superior market or sector timing skills based on their holdings and trades of ETFs. In contrast to the return-based measures, our ETF-based measures do not suffer from biases induced by passive timing effect or by funds following dynamic trading strategies. ETFs offer institutional investors unique opportunities to instantly establish, increase, or decrease exposure to broad or a segment of equity markets. Unlike index mutual funds, ETFs are exempt from short-sale constraints and are traded continuously during market hours. The proliferation of ETFs over the last decade has also led to excellent liquidity. According to the Investment Company Institute, the total net assets of the U.S. ETF market exceed $600 billion at the end of 2007, making it possible for institutional managers to take sizable positions. Indeed, ETFs have become a preferred investment vehicle by which portfolio managers implement directional bets on broad market movements. If institutional investors are able to time the market, we expect to observe a positive correlation between changes in institutional holdings (or trades) of ETFs and the subsequent performance of ETFs that track market indexes. We investigate this relation in both univariate and multivariate settings. We conduct statistical inferences using a bootstrap approach. In addition to market return, we also test whether changes in ETF holdings predict subsequent ETF volatility. Market volatility influences performance measures such as Sharpe ratio. Therefore, portfolio managers can improve risk-adjusted performance by timing market volatility.