Do Mutual Funds Time the Market? Evidence from Portfolio Holdings

Existing literature has found no evidence of market-timing ability by mutual funds using tests based on fund returns. This paper proposes alternative market-timing tests based on observed fund holdings. The holdings-based measures are shown to be more powerful than the return-based measures, and are not subject to “artificial timing” bias. Applying the holdings-based tests, we find strong evidence of mutual fund timing ability. Our findings also suggest that market-timing funds tend to have higher returns and trade more actively. Furthermore, they seem to have market-timing information beyond those common return-predictive economic variables documented in the academic studies. Finally, we quantify the potential economic value of market-timing as a contingent claim. The magnitude of the estimated values indicates that market-timing is potentially an important investment strategy deserving more academic attention. Two strands of finance literature stand in curious contrast to each other. On the one hand, many studies have argued that aggregate stock market returns are predictable and that such predictability should have a significant impact on an investor’s optimal asset allocation.1 On the other hand, there is scant evidence that investors take advantage of such predictability in their portfolio decisions. In particular, earlier studies dating back to Treynor and Mazuy (1966) and Henriksson and Merton (1981), and recent studies such as Becker, Ferson, Myers, and Schill (1999) and Jiang (2003), all find that mutual funds on average do not exhibit significant market-timing ability.2 Fund managers are perceived as more skillful and better informed investors. If they cannot explore market return predictability, it is unlikely that anyone else could.3 The market-timing measures in existing studies are mainly based on fund returns. In this paper, we argue that these measures may not be powerful enough to detect market-timing ability for the following reasons. First, existing timing tests, such as those of Henriksson and Merton (1981) and Treynor and Mazuy (1966) (hereafter “return-based measures”), are based on nonlinear relations between fund returns and market returns. Fund returns are volatile with observations available often at low frequency and over a relatively short time period. This limits the statistical power of such tests. Second, as Jagannathan and Korajczyk (1986) point out, the return-based measures may lead to incorrect inference when fund managers engage in dynamic trading or invest in securities with option-like features, a phenomenon known as the “artificial timing” bias. To get around these problems, we use data on portfolio holdings to examine mutual fund market-timing For evidence of market return predictability, see, for example, Campbell (1987), Campbell and Schiller (1988a, 1988b), Cochrane (1991), Fama and French (1987, 1988, 1989), Fama and Schwert (1977), Ferson (1989), Keim and Stambaugh (1986), Lamont (1998), Lewellen (1999), and Pontiff and Schall (1998). For how such predictability should affect an investor’s asset allocation decisions, see, for example, Kendal and Stambaugh (1996), Barberis (2000), Balduzzi and Lynch (1999), Campbell and Viceira (1999), and Campbell, Chan, and Viceira (2003). In addition, Graham and Harvey (1996), Aragon (2003), Fung, Xu, and Yau (2002), and Blake, Lehmann, and Timmermann (1999) find no evidence of market-timing ability by investment newsletters, hedge funds, or pension funds. The only exception, to our knowledge, is Bollen and Busse (2001) who find positive timing ability for a sample of 230 domestic equity funds at daily frequency. One may argue that institutional constraints on mutual funds, such as no-short-sale, might prevent them from timing the market. However, market-timing can be achieved by adjusting portfolio holdings and the cash component without short-selling any securities.