Tennis, golf, or chess are all activities that require skill. In fact, we can quickly identify the skilled players in these endeavors. On the other hand, craps and roulette are pure games of chance. Skill plays no part in the outcome. But what about managing a stock portfolio? Can managers “beat the market”? If so, can we tell if they are skillful or just lucky? Can we use past performance to predict future performance? Do winners repeat?

In the context of the efficient market debate, if markets are efficient, then management may not be able to add value. Measuring performance for management results requires a benchmark. However, it’s important to use the right benchmark or we will hopelessly confuse ourselves. It’s not very useful to compare apples and oranges, or foreign and domestic stock performance. As academics and consultants have delved deeper into the performance issue, the benchmarks have necessarily become more elegantly defined.

It’s also vitally important to have “clean” data. No one wants to do a study only to find out that the data used was corrupt. The ultimate nightmare for academics is to have someone else point out that their data is corrupt. Fortunately we have a great deal of clean data available from reliable third-party sources that most of us can agree on. For example, SEI, a private consulting firm, maintains the largest database on investment performance of institutional managers. Morningstar supplies extensive data on the mutual fund industry, and the Center for Research Securities Pricing (CRSP) maintains a database on individual security pricing.

A Quick Test

Here’s a very crude test of management performance: Let’s compare the domestic-equity mutual fund performance supplied by Morningstar against the S&P 500 index for one-, three-, five-, and ten-year periods, looking back from April 30, 1995. The S&P 500 index is a fair comparison for large, domestic companies. Our results:


  • Of the 1,097 funds Morningstar covered for the one-year period, 110 beat the S&P 500, while 987 fell short. Results ranged from 46.84% to -32.26%, while the S&P 500 attained a 17.44% return.
  • During the three-year period, the S&P 500 returned 10.54%, while results in the funds varied from 29.28% to -15.02% compounded annually. Of the total 609 funds, only 266 beat the S&P 500.
  • Shifting to the five-year period, of 470 funds, 204 beat the S&P 500. Results ranged from 27.35% to -8.51%, while the index racked up 12.62%.
  • At ten years, only 56 of 262 funds managed to beat the index, and results varied from 24.77% to -4.06% compounded annually against 14.78% for the S&P 500.

If beating the S&P 500 is a valid test of management ability, then a lot of managers are clearly not worth their salt. Far fewer of them appear to be winners than we might have expected.

This Test May Not Be Entirely Precise, But…

Let me be the very first to say that while this little study makes the point and is valid, it isn’t perfect. In all cases, the average fund result fell below the index. However, the average result doesn’t take into account the size of the fund. A few small funds could throw off the average in either direction, so perhaps we shouldn’t be too concerned about the average.

Another reason we might be concerned about our little exercise is the issue of survivor bias. Funds that fail during the measurement period are not measured in the results. Mutual fund companies often make poorly performing funds “disappear” by merging them into more successful funds. Fund performance is not merged and the companies succeed in burying their mistakes. The survivors presumably have a better record than the total number that started the measurement period. Voilˆ! A little bit of marketing magic allows the fund companies to show performance better than their shareholders actually experienced. A better study would account for this distortion.

A problem that disturbs me in this type of analysis is that a single year may account for extraordinary results. If last year was the extraordinary one, then it will show up in all time periods. The results will appear to be far more consistent than they actually were. A fund that had nine average years followed by a great year will look good for the past one, three, five, and ten years! If the great year had occurred during the first year, then the ten-year result would look good, but the one-, three-, and five-year periods would look only fair. This presents a far different picture, even though the total results are the same. So we haven’t adjusted for consistency of results.

Finally, we haven’t adjusted for risk. Both the big winners and losers may have taken large risks to get where they are.

What About the Winners?

Some managers did beat the averages, and a few of them did by a very wide margin. All of them may be expected to claim superior skill and cunning. But what about them? Is it possible to conclude that they are wise men and women and that the others are fools? By extension, can the people who invested with these winners also claim to be wise? Could we have predicted which players would have become winners?

Probability theory would account for a number of winners and losers in any random series of events. If a million people each attempted to toss heads with a coin for several rounds, after each round we could reasonably predict the number of winners. For instance, after ten rounds we would expect 976.563 survivors. Each of them came up heads ten times in a row. Since it’s random, we would not expect exactly 977 survivors, but we could consult with a statistician and predict a very tight range for the number of survivors. In an event that involves no skill at all, we can, with some confidence, predict that after ten rounds there will be survivors, and have a fair idea how many there should be. Should one of our survivors become convinced that his skill contributed to his success, we might have a difficult time shaking him from his delusion.

One way to determine whether skill contributed to the outcome would be to see if there were significantly more winners than probability would have allowed. Suppose that instead of about 977 winners, we ended up with 5,000 or 10,000. Then we might have to concede that an element of skill was involved.

If markets are efficient, we should expect to see a random distribution of results. When we study mutual fund performance, we should expect to see some winners. Probability theory demands it. We would be very disappointed and concerned if an occasional Magellan Fund (the most famous and successful fund in the history of the universe) didn’t turn up. But what we find is far fewer than a random distribution would predict. However, if we adjust the fund results by about 2% to add back in average costs of management and trading, then we get just about the bell-shaped curve that we would expect for performance distribution.

Since we have fewer rather than more winners than we would have expected, it is very difficult to support the argument that the winners got there by superior skill and cunning rather than through pure dumb luck. This is a powerful but not totally conclusive argument. Like our deluded coin tosser, Peter Lynch (former manager of the Magellan Fund) will never agree with that premise.

If It Was Good Yesterday, Will It Be Great Tomorrow?

What about track record? If management skill adds value, can past performance give us an indication about future performance? Do winners repeat? How successful will I be if I only buy the funds with the best past five-year track record?

A recent study examined mutual fund performance by category over several five-year time periods. Funds were divided into quartiles by past total performance, and then followed for an additional five years. The results were enough to blow your mind! A top quartile fund had just less than a 50% chance of being in the top half during the following five years. A bottom quartile fund had just slightly more than a 50% chance of being in the top half during the following five-year period. Similar studies with similar results were completed by a large pension fund on the performance of their managers, and by a large consulting company on the results of the managers whose performance they tracked. In other words, we can’t count on either winners or losers to repeat!

Again, this isn’t a conclusive argument. We can’t say for certain that a top- or bottom-performing fund won’t repeat, just that it doesn’t appear to be more likely to continue its performance than random chance might dictate. These types of arguments take on near religious intensity on Wall Street. I don’t expect them to be resolved in my lifetime. I do think that the overwhelming weight of the evidence suggests that markets are efficient, and that management has a rather small chance of reliably exploiting inefficiencies. Given the egos and profits involved, you can expect further spirited debate.

Dissenting Voices

To further complicate the issue, two heavyweight thinkers who might be expected to support efficient markets have just published studies which show that in the very short term (less than two years), winners may tend to repeat. Both Roger Ibbotson and William F. Sharpe have international reputations in finance, and Sharpe has a Nobel Prize in economics. So they speak with some authority. They both recently made similar observations on short-term performance. Sharpe in particular goes out of his way to point out that the data may be ambiguous.

Personally, I believe that factors other than skill and cunning can extend a fund’s winning or losing streak over a short multi-year period. For instance, during the early 1990s, a large overweighting in health care stocks would have resulted in significant over-benchmark performance for several years. Several mutual funds built reputations based on that one call alone. Since the decline of health-care-sector stocks, most of those funds have descended into a disappointing level of mediocrity. I’m not sure that chasing last year’s winner does anything other than position you with next year’s loser.

It’s easy to pick last year’s winner; it’s difficult to pick next year’s. A number of magazines routinely make mutual fund recommendations. Perhaps the most sophisticated publication among the popular business press is Forbes. One would suspect that if it can be done, they have the resources to do it. They have for years published their “Honor Roll of Mutual Funds.” If you had invested steadily in the Forbes funds, you would have had very disappointing results. This underperformance is so consistent and widely known in the industry that many mutual fund wholesale sales representatives I know consider it the kiss of death.

Toward Better Benchmarks

We can build a benchmark for just about any market or portion of a market. For example, suppose we divided all the publicly listed stocks in the U.S. into ten different sizes by market capitalization on one axis, and ten different segments based on book-to-market ratio on the other axis. We now have 100 different possible submarkets. We could call each submarket an investment style, and each style could have its own index or benchmark. If we studied the performance of each style, we would find that they are sharply different from each other. Each style would have distinct rates of return and exhibit different risk or standard deviations. Each would also have different correlations from the other. Each style could go through a market cycle with dramatically different results for each time period. In other words, there isn’t just one domestic market, but many.

Most managers, but not all, confine themselves to a distinct style. Very few operate in all parts of the market or switch from one part to another. For instance, they may be large-cap value, mid-cap growth, or small-cap market. This is the area of the market they claim to know best, think has the greatest potential, or perhaps were hired to manage. In any event, over time most of the performance they obtain may simply be attributable to where in the market they invest. It wouldn’t be fair to compare a small-cap-value manager with the S&P 500, which is basically a very-large-cap, mostly-growth index. To test whether this is so, we can compare their results with the index matching their area of investment. This type of benchmark is much more precise than just arbitrarily choosing an index like the S&P 500.

These benchmark designs can become very precise and elaborate. One large consulting firm examines the unique style of a manager within the market, builds an index of all the stocks within the style, and then has a computer construct 1,000 hypothetical portfolios from the index. They then average the results of the hypotheticals to create a benchmark measuring the manager. In the vast majority of cases, managers are unable to demonstrate that they add value against the benchmark.

The conclusion that investment style is much more important than management within a style has become harder and harder to ignore. Even when a manager beats his benchmark, we are left with the problem of determining whether he was good or just lucky. The revolving door, and large institutions’ continual search for managers that can add value, adds credence to the belief that beating a benchmark can’t be done consistently. The migration from active management to indexing or passive management indicates that many large institutions have concluded that either it can’t be done or isn’t worth trying.

Taking Big Bets Against the Benchmark

Even within a carefully defined style, investors are still faced with a wide – even alarming – variation of results in both the short and long term. Looking again at the ten-year result for domestic equity funds, there is a surprisingly large variation in outcomes. Part of this is attributable to style differences within the markets. But a large amount of the variation can also be attributed to sector or timing “bets” by managers. When a manager decides to over-weight or under-weight the firms or sectors in his style group, he expects to improve results. He might decide that General Motors will do better than Ford. Or he might decide that cars will do better than banks. Or he might decide that cash will do better than stocks. From my perspective, there is a chance that he will be wrong. If so, he will not make even the benchmark return.

Most of us are risk-adverse. If at the beginning of a ten-year period we were given a choice of a sure return of 14.78% or one that might run from 24.77% to -4.06%, most investors would go for the sure return. Looking back, most individuals never came close to the benchmark and wish they had chosen it. The benchmark would have been better than all but 56 of the 262 funds or top-quartile results. A totally passive approach to selection and a policy of no market timing would have delivered very satisfactory returns. And we don’t have to be either skillful or lucky to get them.

Based on managers’ dismal record to outperform benchmarks, we have to take very seriously the argument that markets are efficient. While we will never be able to prove our case to the satisfaction of everyone, the evidence is pretty strong. This evidence is supported by studies of markets worldwide. Even if some other markets around the world are not as efficient as ours, they still are pretty efficient. If information in foreign markets isn’t as good as what we’re used to here, at least all the players are being equally deceived.

When we go about building our investment strategy, a benchmark, style, or passive approach may be very viable. After all, what’s wrong with top-quartile results?

In my own practice, I maintain a very heavy weighting in institutional index funds wherever they are available. I think that approach gives us the highest probability of a successful outcome with the lowest risk. I do hedge a little: actively managed funds have a minority position. All other things being equal (they never seem to be), when given a choice between actively managed funds, I will go for the one with the lowest cost, widest diversification, and lowest turnover. To the extent possible, I want to see predictable results. I hate underperforming the benchmark more than I would enjoy overperforming. That makes me pretty much like my clients: risk-adverse.

So we return to the thesis that asset allocation is much more important than focusing on a particular stock, timing, or manager. If it’s more critical to be in the right market or style than any other factor, how do we choose the markets? What do we know about style-investing results that will help us construct our own portfolios? In the next chapter, we’ll focus on how a firm’s size affects returns, and on the debate over growth or value styles.