| |
 November 2007 Longer Periods vs. Recent Results Which works better when seeking future results?
For centuries philosophers have debated the strength of these inferences while scientists attempted to hone them into the scientific method. Some efforts have been better than others, yet there is some relation between the depth of research and reasoning and success. That’s why investors are always trying to find the magical combination of fundamental factors and relations that will lead to future security performance. A great deal of research has focused on the predictive ability of various factors, yet most end up concluding there is no simple answer. As a result, many investors just seek out the most recent “hot hands” on the assumption they’re most likely to maintain momentum.
Excess Return and Future Performance Others think short-term momentum doesn’t just disappear overnight. While it probably won’t persist in the long-term, it’s much more likely to continue in the short-term. If your time horizon is a year or less, a momentum strategy will be more indicative than longer-term factors.
We generally think of alpha more as a function of allocation and timing rather than an intrinsic feature of a security. Many managers claim to “add alpha” through their stock picking and timing abilities so rather than looking at stocks, we focused on mutual funds. The test was simple: we wanted to see which was more predictive of short-term performance: 10-year, 5-year, or 3-year alpha. Of course things aren’t quite that simple. First there are problems with track records and survivorship bias. As you’d naturally think, there are many more funds available today with 3-year track records than 10-year track records. Not only that, equities have enjoyed a bull market over the past three years while 10 years not only encompasses the runup into the new century but the bear market that followed. To compensate for this, we only considered funds with 10-year track records. That assured that the same funds were compared in all time periods. We also required an r-squared of at least .80. R-squared is a decimal representation of the percentage of a fund’s movement that is attributable to the movements in its benchmark index. It can range from 0.0 to 1.0. If a fund has an r-squared right around 100, it’s probably an index fund, since it’s almost identically tracking its benchmark. Returns from a fund with a low r-squared will have little to do with the index with most determined by security selection and timing. We required at least .80 because r-squared can be used to gauge the statistical significance of a fund’s beta. The lower the r-squared, the less reliable the beta. Because beta is used in determining a fund’s alpha, we relied on a relatively high r-squared to assure a meaningful comparison For the record, there were 1143 unique domestic funds in the Morningstar database meeting these standards. Had we not limited the study to “unique funds” the total number would have almost doubled to 1987. We considered “unique funds” rather than all share classes since the latter would be duplicative, differing only by the variations in their expense ratios. While this would certainly have a major bearing over the long-term, the impact is substantially diminished when considering 1-year return. For each of the three time periods, we sorted the funds based on their alphas. We then divided them into quintiles (20% groupings) and compared their returns. Although the focus was on the 1-year returns, we also got a look at other statistics such as batting average, information ratio, and up and down market capture. There weren’t many surprises, but there were a few.
What We Expected and What We Got You’d also expect that the funds with the best 10-year returns would be those with the highest 10-year alpha. Again, that’s precisely what we found. Although not charted here, the same was true for the other time periods as well: Funds with the highest 5-year alphas had the best 5-year returns, and the best 3-year alphas had the best 3-year returns. Nothing surprising there.
But we weren’t looking for the best 3, 5, or 10-year returns, we were interested in 1-year returns. Based on what we observed from the other time periods, it would seem reasonable to assume funds with the highest 3-year alphas would have the best 1-year returns of the group. Funds with the best 5 or 10-year alphas may have earned their ratings early in the period and this would be less of an issue for funds with the best 3-year numbers. As it turns out, this assumption was correct, too (see Chart 1). Yet that’s not the whole story. Take a closer look at Chart 1. Yes, the 3-year results for the top quintile are better than those for 5 and 10-years although the difference really isn’t overwhelming. Not only that, the same reasoning that led to the theory that 3-year statistics would lead to better 1-year returns would also suggest that 5-year statistics should lead to better short-term results than 10-year numbers. But not only is this not the case in the top quintile, it’s not in the other four either. In fact the 10-year results dominate the 3 and 5-year in all quintiles other than the first. That’s not what you would have expected. To be sure, these results may simply be an aberration in our limited data set. It’s quite possible they would look quite different in other 3, 5, and 10-year periods. All we can say for sure is this is how they look for this particular set. Nevertheless, it suggests shorter term statistics do not necessarily dominate when seeking short-term results. This conclusion is also supported when looking beyond return. For example, funds with the best 10-year alphas also dominated the other periods in 3-year batting average. Batting average is a measure of the frequency with which a fund’s return surpasses that of its benchmark index. True to its namesake, the number of months in the measurement period (number of times at the plate) is divided into the number of months in which the fund’s return beats the index (hits) to derive the batting average. Just as a baseball player’s batting average doesn’t distinguish between singles and homeruns, a fund’s batting average doesn’t measure how much the fund beats the index, only the number of times it does. As such, it’s more a measure of consistency than overall return. Intuitively, you would think that funds that have recently exceeded their benchmarks would be more likely to continue doing so than those that have a better long-term record. There is, after all, a certain degree of momentum not only with the funds themselves, but with their benchmarks, too. For example, when large cap growth indexes were struggling over the past several years, it was easier for large cap funds to beat them, increasing their batting averages. This wasn’t so much a function of the funds’ fantastic performance as it was the indexes’ poor results. The opposite was true ten years ago when an overwhelming majority of funds were unable to keep up with soaring large cap indexes. Given that, it’s logical to assume that funds selected on shorter-term fundamentals should have better batting averages than others selected on 10-year statistics. But that’s not what happened. Granted, our universe of funds isn’t limited to just one category of funds such as large cap growth, yet the results are still surprising. Inasmuch as the funds evaluated on 10-year alpha dominate 3-year batting average, there is some evidence that longer-term statistics can lead to funds with greater consistency. Again, this may simply be a function of this particular dataset, but the numbers are there.
Inconclusive Conclusions Limiting the study to funds with a minimum 10-year track record was a reasonable restriction for the purposes here, but may not be for investors seeking short-term results. Many of the funds we eliminated -- those with shorter track records and lower r-squareds -- may actually provide greater momentum and ability to beat the index, at least over the short-term. Obviously a much more thorough examination covering additional factors and time periods is necessary before coming to any definitive conclusions. At the very least, however, our results do suggest longer-term statistics can (and possibly should) be used when anticipating short-term fund performance. This may be particularly true for measures of consistency when longer-term results may be more indicative than shorter-term momentum. Search this site! Just enter you key word or words:
Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
|||||||||||||||||||
HEN IT COMES RIGHT DOWN to it, most security evaluation relies on the belief that the future will be like the past. It doesn’t matter if you’re running a dividend discount model or a simple screen on fundamental factors, the underlying assumption is that what worked in the past is predictive of what will occur again in the future.
