الگوبرداری از عملکرد اعتبارات تخصیصی پیشنهادی سهام، اوراق قرضه، و پول نقد توسط خانه های سرمایه گذاری بین المللی

We conduct performance tests of the recommended asset allocations made by a panel of international investment houses (the “Houses”) from 1982 through 2005. We compare the returns and Sharpe Ratios from the recommended-weight portfolio against those of several benchmark portfolios and to a set of 10,000 returns and Sharpe Ratios from randomly shuffled-weight and shuffled-weight change portfolios. We find that the Houses generally fail to outperform the benchmarks. The shuffled-weight change benchmark exhibits a robust “style-preserving” property in that the average portfolio standard deviation is nearly equal to the portfolio standard deviation from the actual recommended weights.

Perhaps the most fundamental investment decision is how best to allocate portfolio dollars among equities, bonds, and cash. However, there is little published research concerning advice on this basic investment decision. By contrast, many studies examine the ability of investors to make superior decisions about market timing and security selection. Blake et al. (1999) underscore the importance of asset allocation when they show that potential gains to investors from asset allocation dwarf potential gains from market timing and security selection.1 A motivating question we address in this paper is how best to measure the performance of asset allocation recommendations in the face of differing investment “styles” and their associated ex-ante risks. We introduce two new performance measures that use randomization to generate a benchmark return distribution. In these randomization methods, the recommended portfolio weights, as well as the recommended portfolio weight changes, are repeatedly shuffled to generate two distributions of portfolio returns. We can then compare the actual performance to these distributions and calculate significance levels. In our study, we examine the long-horizon performance of recommended asset allocation strategies made by a panel of international investment houses between April 1982 and July 2005. These investment houses (the “Houses”) provide recommended asset weights to equities, bonds, and cash. The recommendations come from surveys published in the Financial Report, a confidential newsletter that was purchased by The Economist Newspaper, Ltd., in 1989. In these surveys, money managers in different countries are asked to provide asset allocation recommendations for a hypothetical investor who has “no existing investments, no overriding currency considerations, and the investment objective of long-term capital growth.” These criteria give us a sample that is well-suited to a detailed examination of the performance of asset allocation recommendations over a long horizon. We look for performance attributable to style by separating the Houses into “equity champions” (Houses that recommend equity holdings that are greater than the average equity holding), “fixed-income champions” (Houses that recommend equity holdings that are less than the average equity holding), and by examining individual House performance. We find that the shuffled-weight change benchmark portfolio exhibits a robust “style-preserving” property. That is, the average portfolio return standard deviation is nearly equal to the portfolio return standard deviation from the actual recommended weights. This property is robust across time periods and return series. It is also robust as to whether the recommendations come from equity champions, fixed-income champions, individual Houses, or an overall consensus. In our sample, as a group or individually, the Houses do not exhibit much, if any, skill in shifting asset allocations among equities, bonds, and cash. Independent and simultaneous research by Annaert et al. (2005) also concludes that recommendations from these Houses are unable to outperform passive portfolio benchmarks. However, our study differs in significant ways. First, Annaert et al. (2005) do not incorporate transaction costs. In our study, we account for transaction costs using a method we develop to rebalance a portfolio to target weights when transaction costs differ among asset classes. Annaert et al. (2005) only use unconditional performance tests. In addition to the randomization methods, we measure portfolio performance using unconditional Sharpe Ratios and the Ferson and Khang (2002) conditional weight measure. We find that the Ferson and Khang (2002) conditional weight measure confirms the basic conclusion of this study. That is, the Houses exhibit little, if any, ability to make timely asset allocation shifts among equities, bonds, and cash. Our findings have important implications for investors who rely on basic asset allocation advice. The scale of their collective decisions is colossal. Mutual fund investments in particular are vast. As of the second quarter of 2006, world-wide mutual fund assets totaled $19.4 trillion, distributed among 59,385 funds across 41 countries (www.ici.org, November 6, 2006). It is possible that this enormous investment pool has benefited from knowledgeable asset allocation recommendations over time. However, asset allocation recommendations that are highly correlated with those in our sample would have had little value for investors. The paper proceeds as follows. In 2 and 3, we describe the return and weight data, respectively. In Section 4, we detail how we incorporate transaction costs into portfolio rebalancing. In Section 5, we summarize the unconditional portfolio performance results, and in Section 6, we present two robustness inquiries. Section 7 contains conditional portfolio performance and Section 8 concludes.

In this study, we introduce two new performance evaluation methods that use randomization to generate a benchmark return distribution. In these randomization methods, the recommended portfolio weights, as well as the recommended portfolio weight changes, are repeatedly shuffled to generate two distributions of portfolio returns. We also introduce a method to rebalance portfolio weights to targeted weights after accounting for transaction costs. We use the randomization method and a conditional weight-based method to examine the unconditional and conditional performance of recommended portfolio weights to equities, bonds, and cash by a panel of international investment houses (the “Houses”) from 1982 through 2005. The recommendations come from surveys published in the Financial Report, a confidential newsletter purchased by The Economist Newspaper, Ltd., in 1989. In the surveys, the Houses recommended portfolio allocations among equities, bonds, and cash, as well as equity holdings across countries. We examine the performance of portfolios constructed from the average recommendation of the participating Houses at each survey date. We also look for style performance by separating the Houses into equity champions, fixed-income champions, and by examining individual House performance. We find some differences between the performance of Houses that tend to favor equities and Houses that tend to favor bonds. Houses that favor equity outperform several benchmarks toward the end of the sample period. However, the performance of the two styles is similar. We conclude that as a group or individually, the Houses do not exhibit much, if any, skill in shifting asset allocations among equities, bonds, and cash. Blake et al. (1999) point out that “most (pension) funds would have been better off with their strategic asset allocations placed in passive index funds, and yet they purchased active management services.” If investment managers decide to index, they are still left with the question of how best to allocate portfolio dollars among equity, bond, and cash index funds. Our results suggest active management services for asset allocation among equities, bonds, and cash might also be of little value to investors. A valid comparison between Sharpe Ratios (as well as returns) is to compare an investment portfolio to a benchmark with a similar ex-ante risk. A benchmark portfolio that matches a manager's style should therefore result in a benchmark that also matches the ex-ante risk of the manager's portfolio. The shuffled-weight change benchmark portfolio exhibits a robust “style-preserving” and risk matching property. That is, the average portfolio return standard deviation is nearly equal to the portfolio return standard deviation from the actual recommended weights. This property is robust across time periods and return series. It is also robust to whether the recommendations come from individual Houses, equity champions, fixed-income champions, or an overall consensus. Further research on the properties of this benchmark are necessary. But, because a pre-specified benchmark is not needed, it is possible that this benchmark could prove useful in judging the performance of many types of money managers, including hedge funds. Additionally, inference is based on an empirical distribution so this benchmark can be used to judge the significance level of portfolio performance.