Past Performance Is Not Indicative Of Future Results
Who would have thought a boilerplate compliance disclaimer would contain some of the wisest and least heeded investment advice? Despite the warning, historical performance is still overemphasized in the selection of top quality active investment managers. From media publications like Barron’s to investment research databases like Morningstar, past performance, especially that of the last 1, 3, and 5 years, is used as the primary descriptive measure of an investment fund’s quality, and by extension their future potential for outperformance. Combined with a natural psychological predisposition to want to be with the “best,” investors continue to rely on this measurement to chase the manager with the strongest recent performance. It turns out this is statistically the worst time to move to such a manager.
Investors who succumb to the siren song of past performance find themselves on a treadmill of short-termism often leading to lower returns than even an average manager would achieve over time.
But why is it misleading to look at past performance in the first place? And if not performance, what do you look at to select the right manager?
The main problem with past performance is that it bears a weak or nonexistent causal relationship to future performance. Most investors comprehend this logically, but don’t understand it intuitively. An illustrative (if imperfect) analogy is choosing “heads” simply because a coin flip has resulted in heads 5 times in a row. We know, of course, that flipping heads does not change the probability of the next flip turning up heads. In fact, those 5 tosses hardly provide any useful information for predicting the outcomes of future flips. The same rule applies to manager selection.
Say the coin lands on heads 100 times in a row. Now you would probably want to bet on heads on the next flip. At this point (and probably well before) it is reasonable to ask yourself: “Why is the coin coming up heads each time?” or “Why am I still flipping this damn coin?” Since we know the outcome of the 99th flip doesn’t cause the outcome of the 100th flip, there must be other factors at play—I’ll get back to this in a bit.
This second part of the analogy hints at the other main problem with past performance, which is that, in contrast to coin flipping, we simply don’t have enough data to robustly evaluate managers based on return statistics alone. As Cliff Asness, Managing and Founding Principal of AQR Capital Management, puts it:
Nobody, including me in this essay, wants to deal with the very big problem that we often do not have enough applicable data for the investing decisions we make. We evaluate strategies, asset classes, managers, and potential risk events using histories the statisticians tell us are too short or too picked over. These histories are generally insufficient and very vulnerable to such things as data mining, ex post selection of winners who don’t repeat…and simple randomness.
I’ll call this the “But Warren Buffett” dilemma. How many outperforming funds that began 50-plus years ago still exist today materially unchanged in terms of strategy, philosophy, process and senior staff? “But Warren Buffett!” Exactly. His celebrity, and the continued success of Berkshire Hathaway prove the rule, as the circumstance is so uncommon.
It’s worth noting that even if a manager does have a long enough track record, it still does not guarantee future performance. Take, for example, an adaptive markets hypothesis view, which would stipulate that the fund’s superior historical performance induces other investors to copy their strategy, increasing competition for the same assets and thus driving down future returns. Simply put, just looking at past returns, however strong they may be, can be problematic in a market environment prone to change and adaptation.
One number is not enough
Surely third party research providers know all of this. So why do they continue to overuse historical returns? For one, they’re simple. Everyone understands what returns mean (sort of), and everyone recognizes them as the holy grail. Their cleanliness and elegance is compounded by their objectivity and quantitative nature. Return data can be used in models and screens and turned into, say, a star rating. But it’s lazy, and its simplicity renders it unhelpful, or even harmful, in hiring and firing investment managers. If I were more cynical, I might even say the star ratings are designed to give average investors a false sense of security when making investment decisions, without regard to the underlying methodology. I recognize that it might be unfair to criticize the data providers here. After all, they provide data they believe investors will want to consume—at Bridgewater Wealth, we certainly use them as part of our investment analysis. Ultimately it’s up to us, the investors, to utilize their information in a responsible manner.
Investment performance varies by strategy
Remember those “other factors at play” in the coin flip analogy? The confounding variables that cause investment performance vary by strategy. A factor important in evaluating a US small cap equity mutual fund may not matter for an illiquid alternatives strategy and visa-versa. The commonality, however, for all active managers is the importance of evaluating the decision-making capability of the manager. In fact, that’s the essence of active management, and what distinguishes it from its passive counterpart. Passive is defined by the absence of decisions; it’s a replication of the market as is, without substitutions, deletions, or additions. In order to understand how an active manager can beat their passive benchmark, you have to understand how their decisions will deviate from the benchmark in the future. That starts with studying everything—reasoning, process, frequency, timing, personnel—behind the manager’s past decisions, and understanding how those factors may change going forward.
Now that I’ve imbued in you a sense of skepticism for historical returns, and at the risk of sounding like a hypocrite, historical performance is key to understanding those past decisions. The singular number itself, as I’ve mentioned, is of little use. But the breakdown of that number into its component parts—also known as performance attribution—is an important avenue for understanding the “how” and “when” aspects of making decisions to keep or change managers.
Though this framework applies generally to all active managers, the multitude of specific methods in which managers of various strategies should be evaluated could be the subject of a lengthy book. Despite all that nuance, a clearly defined investment philosophy and process are at the heart of all repeatable value-add strategies.
In his 2017 Superbowl halftime address, as the New England Patriots were getting walloped 21-3, Coach Bill Belichick reportedly told his team, “The score is not important, focus on the process.” I can’t think of a better way to sum up my thoughts on manager selection. In what is now considered one of the all-time great Superbowl comebacks, the Patriots came out of halftime and scored 31 unanswered points to win the game. I can only hope that following Coach Belichicks advice will lead to similar investing success.