Executive Summary:
Beating the Market: Harder Than It First Appears
Many recent research papers report evidence that trading strategies based upon historical data are able to earn abnormally high returns. These trading strategies are put forward as evidence that stock markets are not efficient. In the framework of the efficient markets hypothesis, the ability to consistently generate abnormal returns is anomalous. Something may be missing from the research, though. In their research, Professor Doug Hanna of SMU Cox and co-author Mark Ready of University of Wisconsin-Madison incorporate transactions costs into the equation and find that the predictive abilities of some popular trading strategies are overstated, or nonexistent.
Background
This research directly examines the well known Haugen and Baker strategy purporting stock market predictability. The efficient markets hypothesis generally states that the market price will quickly impound all publicly available information and, therefore, it is impossible to use historical information to beat the market. Haugen and Baker report that a long-short stock selection strategy based on more than 50 measures of accounting information and past return behavior would have generated excess returns of approximately 3% per month, or 36% annually. Hanna and Ready examined data during the time periods of the original studies as well as time periods afterward to validate the trading strategies and also examined after-transaction cost returns to the Haugen and Baker strategy.
In addition to the Haugen and Baker strategy, Fama and French’s book-to-market strategy and Jegadeesh and Titman’s momentum strategy were examined. Fama and French (1992, 1993) developed the three-factor model, where returns are predicted as a function of three factors: market returns, a firm’s book-to-market value, and firm size. The Fama/French results are easily replicated in both the original study’s time period (1979-1990) and also in a subsequent time period 1991-2001, after their original study. They interpret their findings as evidence of risk factors. The momentum strategy of Jegadeesh and Titman (1993) suggests that returns are consistent in 3-, 6- and 12-month intervals. That is, firms that perform the best in those time frames keep performing well, and the returns of the worst performing firms continue to be poor. They consider this phenomenon an anomaly. Hanna commented, “Many of the controversies in the financial literature are matters of interpretation—classifying empirical regularities as either anomalies or risk factors is really a matter of taste or, possibly, belief.”
At Issue
Looking at the performance of mutual funds, few outperform the market by 36% a year. Hanna commented, “It seemed to us that the returns of the Haugen/Baker strategies would be difficult to attain. And, when we layered transaction costs onto their data, in fact, the returns were significantly reduced.” He continued, “The impact of transaction costs will depend on the intensity of trading within a particular strategy. The fact that some of these strategies that promise high returns have high turnover means that those high returns are unattainable after considering transaction costs.” Implementing the Haugen/Baker strategy would be prohibitively costly given the frequent buying and selling necessary to continually rebalance the portfolios.
One possible explanation for identifying patterns in stock return data comes from data snooping or data mining. Given any historical data, one can find patterns even if the data was generated by a random source. To test whether a pattern really exists, Hanna and Ready first take the proposed strategies and test them using independent time periods (see Table 3 in paper). When using later time periods, the results of Fama and French and Jegadeesh and Titman are largely robust. The returns to the Haugen and Baker strategy were mitigated when using a new sample period.
In their second set of tests, Hanna and Ready examine the effect of including transaction costs for the three trading strategies discussed above. Historical return data are often quoted at the midpoint of the bid/ask spread or the closing price for a day. Any new order will likely trade at some price other than that shown in the data. Buy orders will execute closer to the asking price and sell orders will likely be executed closer to the bid price. Hanna and Ready use actual intra-daily data to get realistic estimates of the prices that investors would face if they were to implement any of these trading strategies. The Haugen and Baker strategy suffered the most from the inclusion of transaction costs due to the high frequency of trading necessary to periodically rebalance the required portfolios. After transaction costs, the abnormal returns to the Haugen and Baker strategy are not significantly different from zero. The Fama and French strategy is largely robust to transaction costs because there is little trading required – it is primarily a buy and hold strategy. The Jegadeesh and Titman return momentum strategy is intermediate, the after transaction costs returns to this strategy are reduced somewhat but continue to be statistically robust.
As part of their examination of transaction costs, Hanna and Ready develop models that provide a better way to include transaction costs in the empirical research. A logarithmic functional form for the transaction costs was developed—the log of the transaction costs and the log of the firm size seem to be linearly related and should be taken into account when researchers are deciding how to estimate transaction costs in empirical tests. A method to estimate transaction costs was also developed; coefficients are provided for researchers to use in estimating transaction costs, making the inclusion of transaction costs more ‘available’ according to author Hanna.
Hanna and Ready also considered how investors might use these strategies simultaneously. If offered the ability to invest in any of four portfolios, what combination of investments would offer an optimal strategy? The resulting optimal portfolio (after considering transaction costs) included a significant investment in the Fama and French strategy (high book-to-market); a smaller investment in the 6-month Jegadeesh and Titman momentum strategy; selling short the low book-to-market portfolio of Fama and French; and the balance of the investment in the market benchmark (the Russell 3000 in this case). The Haugen and Baker strategy did not play a role in the optimal portfolio when these other strategies were already available to investors. See Hanna and Ready’s Table 7 for after transaction cost optimal portfolios.
Conclusion
Research has frequently ignored transaction costs when looking for potentially profitable trading strategies. In the real world, however, these costs cannot be avoided. Further, different strategies will be affected more or less depending on how much trading activity is required to implement the strategy. Hanna concludes, “Asset managers may be tempted to follow proposed strategies that promise high abnormal returns. Custodians of investment portfolios must factor in the transaction costs likely to be associated with a particular strategy.”
Many intelligent people are running mutual funds, Hanna mentions, “and there are few funds that are able to consistently beat the market.” He also pointed to the rise in popularity of index funds, which generally offer a reasonable return and require little active trading (i.e., low transaction costs). “I think it important to align academic research with the real world or else we end up with all kinds of spurious results or inappropriate conclusions, which may send people on costly adventures,” Hanna stated. “Transaction costs are important, we believe, in differentiating between various investment strategies.”
“Profitable Predictability in the Cross Section of Stock Returns” by Doug Hanna and Mark Ready is forthcoming in the Journal of Financial Economics.
Additional reading:
Fama, E., French, K., 1992. The cross-section of expected stock returns. Journal of Finance, 47, 3-56.
Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33, 3-56.
Fama, E., French, K., 1996. Multifactor explanations of asset pricing anomalies. Journal of Finance, 51, 55-84.
Haugen, R., Baker, N., 1996. Commonality in the determinants of expected stock returns. Journal of Financial Economics 41, 401-440.
Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: implications for stock market efficiency. Journal of Finance, 48, 65-91.
Jegadeesh, N., Titman, S., 2001. Profitability of momentum strategies: an evaluation of alternative explanations. Journal of Finance 56, 699-720.
|
