A study offers lessons for investors striving to add value through fund selection.
This article first appeared in the April/May 2011 issue of Morningstar Advisor magazine. Get your free subscription today!
We think that one of our biggest responsibilities as an investment manager is to properly set expectations. We want the advisors with whom we work to have confidence in our abilities yet also to understand the attributes of our investment philosophy that can have an impact on our short-term performance. This helps ensure that advisors share our long- term orientation, which helps us to execute the strategies on their clients' behalf.
In the course of having these discussions with advisors, the subject sometimes turns to what returns might be possible. Specifically, advisors seek to estimate a portfolio's "excess returns" over a market cycle, which they define as the portfolio's return less the benchmark return, often irrespective of risk.
In this commentary, we present a framework to estimate the potential excess returns of various asset classes, as well as a hypothetical 65% equity/35% bond portfolio. We also draw on the findings of our study, as well as our perspective managing portfolios, to offer recommendations on how to improve the odds of selecting outperforming funds.
There are a number of ways to estimate the excess returns of a diversified portfolio. In this study, we opt for an approach that we think is familiar to investors--percentile rankings.
Percentile rankings are easy to understand and uniformly apply. However, it's not always clear how a given level of outperformance, expressed as a percentile ranking, might translate to excess returns. Thus, we set out to devise a framework for estimating the potential excess returns associated with various mutual fund percentile rankings.
To that end, we compiled monthly returns for all U.S. open-end mutual funds in the following asset classes: domestic large cap, domestic mid-cap, domestic small cap, developed foreign stock, emerging-markets stock, investment- grade bond, high-yield bond, foreign bond, real estate, and absolute return (excluding leveraged funds). To eliminate survivorship bias, we added the returns of merged and obsolete funds back to the dataset. The study covered the 20-year period from January 1990 to December 2010, encompassing more than 6,000 unique funds in all. Using the monthly data, we calculated rolling one- and three-year annualized returns for every fund in each asset class. We then calculated the 10th, 25th, and 50th percentile return breakpoints for every rolling period of each asset class. For example, within the domestic large-cap group, the 25th percentile return for the rolling one-year period ended July 2007 was 18.2%. For each class with a complete dataset, we calculated percentile returns for all 241 rolling one-year periods and 217 rolling three-year periods that make up the 20-year span.
From there, we compared the 10th, 25th, and 50th percentile returns of each rolling period with the relevant index return. Returning to the previous example, we compared the 25th percentile domestic large-cap return for the rolling one-year period ended July 2007 with the Russell 1000 Index's one-year return during the same period and calculated the difference. We repeated that exercise for every rolling period of every asset class, the only exceptions being rolling periods in which there were too few funds to calculate a meaningful percentile return.
Finally, using the return data of the aforementioned asset classes, we calculated the excess return of a hypothetical 65%/35% portfolio for every rolling period for which complete data was available (1996 forward). To simulate the portfolio's rolling excess returns at the 10th, 25th, and 50th percentiles, we multiplied the rolling excess returns we'd calculated for each asset class by weightings we assigned. (The weightings are based on asset-allocation guidelines set by Ibbotson Associates.)
On the surface, what we found is unsurprising--the better the relative performance in an asset class, the more likely one was to achieve positive excess returns and the larger those excess returns, if any. This is a truism.
But there was also nuance in the data. For example, the magnitude of excess returns varied widely between asset classes (Exhibit 1). The excess returns tended to be larger at all three percentile ranks in more-volatile, or more-diverse, asset classes, such as emerging- markets stocks, where the median rolling three-year excess return at the 25th percentile was a hefty 3.1%. The excess returns were typically smaller in less-volatile or homogeneous areas like investment-grade bonds, where the median rolling three-year excess return at the 25th percentile was negative 0.01%.
In addition, excess returns moderated when measured over rolling three-year periods. For example, whereas the median 25th percentile excess rolling return of domestic mid-cap stock funds was a robust 4.1% over rolling one-year periods, it shrunk to 2.4% over rolling three-year periods. Why? Mean-reversion is a powerful force, but it tends to be more apparent over longer periods rather than shorter ones. The rolling one-year returns at the 10th and 25th percentiles often capture a manager's upswing (or, had we measured 75th or 100th percentile returns, swoon), but not the subsequent downswing (or, in the case of underperformers, rebound). By contrast, a rolling three-year window more fully captures the arc of performance, explaining why excess returns tend to be more muted at that interval.
This ties into another subtlety--mean-reversion implies that past performance that diverges from the norm will not persist into the future. Thus, a fund's percentile rankings are likely to fluctuate widely with the passage of time, meaning that the top-quartile or top-decile funds of yesteryear might be the cellar dwellers of the next rolling period. This is especially evident in shorter periods, when fund rankings fluctuate to a greater extent. To cite one example, only 29 of the 78 top-quartile domestic large-cap funds over the one-year ended December 1990 were also in the top quartile of the nearest nonoverlapping one-year period ended December 1991.
That's a good segue to the 65% equity/35% bond hypothetical portfolio. We charted the portfolio's rolling one- and three-year excess returns (Exhibit 2). For simplicity, we show the excess returns an investor would have reaped by investing in the 25th percentile fund of every asset class in every rolling period. The median rolling one-year 25th percentile excess return was 2.8%; the median rolling three-year excess return was 1.3%.
To state the obvious, it's very difficult to pick winners so consistently, making our simulation something of a best-case scenario, at least at the 25th percentile. (The median rolling one- and three-year excess returns were 6.4% and 3.5%, respectively, at the 10th percentile.)
What if successful fund selection followed a random walk? We simulated this scenario as well by randomizing the percentile ranks for each asset class in every rolling period. Not surprisingly, this simulated portfolio's excess returns are lackluster (Exhibit 3). The portfolio's median rolling one-year excess return was negative 1.1%, and its median three-year excess return was negative 1%.
This reflects the erratic relative performance of the funds selected; sometimes they're stalwarts, other times they're laggards. The random-walk simulation isn't a worse-case scenario, but its randomness represents the opposite extreme of the earlier 65%/35% simulation at the 25th percentile, which implied uncanny consistency.
For investors like us who strive to add value through fund selection, results are likely to fall somewhere in between the two extremes simulated for the hypothetical portfolio. That is, we aim to get excess returns with greater consistency than pure chance might confer but have no illusions that we'll always succeed.
So, the question becomes, how do we improve the odds of picking outperformers? Here are a few lessons that we can draw from the data, as well as our experience managing mutual fund portfolios for clients:
1. Avoid the Short Term
Basic statistics hold that we have far better odds of picking a top-quartile or top-decile fund over a longer period, such as five years, than we do of repeatedly picking elite funds from one year to the next. For example, we'd have a 1-in-4 chance of randomly picking a top-quartile fund at the beginning of a given five-year period, versus 1-in-1,024 odds of randomly picking a top-quartile fund in each of the five one-year periods. Remember, there's no "tell" in fund performance given the lack of persistence in returns and rankings.
2. Don't Try to Shoot the Lights Out
The path to standout success is usually winding and seldom linear. Consider, for example, that there were 27 top-decile large-cap funds for the five-year period ended December 2010. Of those funds, 63 suffered a bottom-half showing in one or more of the five calendar years that encompassed that five-year period. Conversely, there were 200 other large-cap funds whose returns ranked in the top decile of at least one of the five calendar years, but not the full five-year period.
3. Focus on Expenses
When it comes to relative performance, expenses are one of the few persistent advantages a fund can count on. A recent study, "How Expenses and Stars Predict Success," by Russel Kinnel, Morningstar's director of fund research, bears this out. Over a five-year period, Kinnel found that funds in the cheapest quintile of five broad categories (domestic equity, international equity, balanced, taxable bond, municipal bond) were about twice as likely to finish in the category's upper half as funds in the priciest quintile. This was even more pronounced in homogeneous areas where returns fluctuate to a lesser extent, such as bonds. True, a paltry expense ratio only lowers the hurdle a fund faces in trying to surpass its index. But it also can denote other salutary traits, such as a management team that is willing to forgo short-term enrichment (from a higher expense ratio) in pursuit of long-term success (by producing good results for clients, which a low expense ratio facilitates).
4. Swing at the Fat(ter) Pitches
As our study makes clear, excess returns tend to abound in more-diverse, volatile asset classes like small-cap and emerging-markets stocks. It's quite possible that these excess returns aren't really "excess" at all but rather the payoff for taking on more risk than the indexes concerned. Further, given the volatility of these areas, investors could see jarring changes in relative performance, which courts its own set of risks (namely, that one will make purchases, or sales, at inopportune times, denting returns). But as our study seems to underscore, excess returns are more plentiful in these riskier areas, and scarcer in bonds.
5. Accept No Imitations
We charted the rolling one- and three-year excess returns of domestic large-cap funds from 1990 to 2010 (Exhibit 4). The bulge in the middle of the chart roughly coincides with the inflating and bursting of the technology bubble. Large-cap managers who loaded up on growth stocks, which came to dominate benchmark indexes in 1999 and 2000, ruled the roost in those years, only to have their heads handed to them amid the ensuing sell-off. (Only five of the 226 large-cap funds whose one-year returns ranked in the top quartile as of February 2000 repeated that feat in the one-year period ended February 2001.) By contrast, value managers, who had suffered the ignominy of watching the tech craze from the sidelines while flouting large-cap indexes, mopped up when the bubble burst. The moral of the story? Mimicking the benchmark index is often an unforgiving business, making it better to opt for funds that show the fortitude to go their own way.
6. Be Contrarian
As mentioned previously, mean reversion is a powerful force. The implications are clear--investors chase performance into hot-performing funds at their peril. Conversely, given the cyclical nature of fund performance, investors are well advised to regularly rebalance, which mechanizes contrarian investing.
7. Remember the Other "R"
One of the clearer takeaways of our study is that fixed-income excess returns tend to be modest, even among standouts. For example, the median rolling three-year 10th percentile excess return of investment-grade bond funds was a measly 0.69% for the 1990 to 2010 period we studied. While this doesn't necessarily mean that investors should resign themselves to a benchmark return, it suggests that a greater focus on risk management is warranted and, further, that bids for equitylike excess returns are ill-advised.
Jeffrey Ptak, CFA, is president and chief investment officer of Morningstar Investment Services.
For all graphs in this article, calculations assume all dividends and capital gains distributions are reinvested, are net of underlying fund's fees and expenses, and are based on the underlying fund's net asset value as of close of trading on the New York Stock Exchange at the last business day of a month. Performance returns were calculated using a time-weighted, geometrically linked rate of return formula. Returns for periods over one year are annualized. Return data shown is based off past performance, and past performance is not indicative of future results. The results of the study are provided for informational and educational purposes only and should not be viewed as investment advice.