Swinging for the Fences
Cathie Wood, Mutual Funds, Tournaments, Convex Payoffs, and Broken Clocks
A loyal reader1 points me to an interesting WSJ column about star money manager Cathie Wood. Wood is the founder and CEO of ARK Invest, which rose to fame with concentrated bets on leading tech companies, most notably Tesla. This has been very lucrative for ARK:
Cathie Wood’s firm went from about $3.5 billion under management across several funds shortly before the pandemic struck to hit the $50 billion mark a year later.
ARK’s flagship ETF, ARK Innovation ETF (ARKK), was up 351% from the beginning of the pandemic to the peak this February, vs. 64% for the S&P. Most of the growth in assets under management came from new inflows from investors.
There has been a dramatic reversal in the last three months, with ARKK down 31%, trailing the 6% return of the S&P.
The author of the article, Spencer Jakab, cites an analysis that claims that the average actual investor in ARKK is now trailing the S&P, even though ARKK is still up more than 2x over last year. The reason is that most of the current investor base jumped into ARKK after the fund put up the biggest returns, just in time for the recent fall.
The most interesting claim Jakab makes is actually that the ARK experience is part of a historical pattern:
Fund managers like Gerald Tsai in the 1960s who rode Polaroid and Xerox to stardom or various dot-com visionaries in the late 1990s wound up doing poorly for clients who discovered them after they became hotshots....Analyst Meb Faber points out that not one of the five Morningstar “fund managers of the decade” through 2010 even managed to beat the market in the next 10 years. The best of the bunch, Bruce Berkowitz’s Fairholme Fund, became the worst.
This is interesting because even if the highest performing managers had no skill at all, we would expect reversion to the mean; the best fund managers in one period would merely show average returns in the next period, like any primate in possession of a dartboard.
Jakab is claiming that perhaps top managers tend to actually underperform in the next period. This would be at odds with the typical trend where stock returns show momentum, and it would be particularly surprising because you would expect the highest performing managers to have significant skill. Is this possible?
A good starting point is that 2010 Morningstar list of “fund managers of the decade”:
Bruce Berkowitz, Fairholme Fund (winner)
Charlie Dreifus, Royce Special Equity
Don Yacktman, Yacktman
Joel Tillinghast, Fidelity Low-Priced Stock
Steve Romick, FPA Crescent
The mutual fund industry is dominated by the major fund complexes: Fidelity, T. Rowe Price, Franklin Templeton, just to name a few. However, only one of the five funds on this list, Fidelity Low-Priced Stock, belongs to a major fund group.
To understand why this is, we have to consider the incentives faced by mutual fund managers. Retail investors disproportionately put their money into the funds that performed the very best in the most recent period.2 Presumably this is because the very top performers get the most free publicity from publications like Morningstar and the WSJ.
However, they do not pull their money from the weakest performers at nearly the same rate, meaning fund flows are asymmetric. Heads you win big, tails you don’t lose much.
Fund managers earn a fixed percentage of assets under management, so inflows translate directly into fund manager profits. In economic terms, the payoffs to fund managers are convex - the very top funds, like ARK, make a ton, while funds that merely did very well only make a little.
We would therefore expect fund managers to craft their strategy to maximize their chance of being one of the top performers in any year, even if it also increases the risk of being one of the worst performers in a given year as well.
The name of the game is to have a chance to gain free positive publicity, which can lead to massive inflows and massive profits. Fund management can be thought of as a tournament with convex payoffs, where the rewards flow to a few winners at the top.
On the other hand, most fund managers are not self-employed; they work at a big company like BlackRock, for a boss that will fire them if they underperform too much in a given year. Their incentive is to not deviate too much from the index. They won’t care as much about publicity and tournaments and convex payoffs. They just don’t want to get fired, which is its own negative convex payoff. This gives rise to a phenomenon known as closet indexing, where some fund managers will partially mimic a passive index, allegedly to improve their job security.
You generally end up with two groups of fund managers: founder-managed funds, who are trying to deviate from the index to have a chance at finishing on top, and employee-managed funds, who hug the index to avoid getting fired.
Long ago, I wrote my senior thesis on this topic, where I found that only about 12% of all funds were founder-managed. Then, as now, the leaderboards were dominated by this small subset of founder-managed funds, only the names were different. Back then it was Bill Gross of PIMCO; in 2010 it was Bruce Berkowitz of Fairholme; today it is Cathie Wood of ARK.
Returning to our 2010 list, Berkowitz and Yacktman both founded their own shops, and FPA is an employee-owned partnership (and the Crescent Fund is founder-managed), so 60% of the list qualifies as founder-managed. The remaining two fund managers benefited from some enhanced level of job security: Royce is a small shop still run by its founder, Charles Royce, and Joel Tillinghast an institution at Fidelity, and still actively manages money there.3
This WSJ article covering the top funds of 2020 suggests the same trend persists today. The three winners are Dennis Lynch, who manages a tech-focused fund at Morgan Stanley, Ron Baron, founder and manager of Baron Partners, and of course, Cathie Wood of ARK.
It is often said that while the baseball great Babe Ruth set the career record for home runs, what is less remembered is that he also set the career record for strikeouts.
In any field where the outcome is a product of luck and skill, we often look at the outcome and think we are seeing a noisy measure of skill, when in fact we are seeing a noisy measure of skill and approach. Babe Ruth hit a lot of home runs in part because he was willing to take a mighty hack and risk looking foolish.
Ruth’s approach created a higher variance in outcome each time he came to the plate. Sometimes he would hit a home run, but this came at a cost of fewer singles and more strikeouts.
Before Babe Ruth, the league was dominated by hitters like Ty Cobb, who would hit a lot of singles and doubles, but rarely homer or strike out. Even after Ruth arrived on the scene, these contact hitters were still quite productive - in fact, in 1925, near the peak of Babe Ruth’s career, it was the aging Ty Cobb who led the league in offensive production.4 Later on, hitters like Tony Gwynn and Ichiro would go on to have Hall of Fame careers with the contact approach.
The Babe Ruth approach became the norm throughout baseball for the players with the right skillset. Most of his imitators did not attain a Ruthian level of productivity; while the Babe would hit 50 home runs in a year and strike out 80 times, someone like 1980s journeyman Rob Deer would hit only 25 home runs and strike out 175 times over the course of a season.
If we were to look at a small sample - say, a leaderboard of the most productive hitters on a given day - we will see a lot of hitters with the high variance approach, who are able to get four bases with a single swing of the bat. Since the average hitter only comes to the plate four or five times in a day, the contact hitters are far less likely to show up in such a leaderboard, being capable of only getting one or two bases at a time.5
In any small sample, we are likely to see more Rob Deers on the leaderboard than Babe Ruths - there are very few players who can hit like Babe Ruth, and very many players who can hit like Rob Deer.6
The problem with using tournaments and leaderboards over small sample sets to identify skill is selection bias. You end up excluding almost everyone with a Ty Cobb-like consistent approach, and you end up with a surplus of mediocre hackers like Rob Deer who happened to have a lucky set of at-bats. The leaderboards tell you less about who is good at hitting, and more about who swung for the fences and got lucky yesterday.
It is only with a larger sample size that you can get a better measure of the value of a hitter, and the good contact hitters will surpass lucky power hitters on their way up the leaderboard. This is not a problem in baseball, with 162 games in a year, but is a much harder problem in the world of long-term investing, where it can take decades to get a decent measure.
The interplay of tournaments and asymmetric and convex payoffs in fields of luck and skill is everywhere around us in the world.
One interesting example is the field of economic forecasting. Harvard professor Owen Lamont produced a study showing that some economic forecasters rationally intentionally make bad forecasts to gain attention; if you predict a recession every year, eventually you will be right, and you will be lauded in the media as the person who predicted the Great Recession. He dubs this the “broken clock” strategy. More consistently accurate forecasters must be content to labor in anonymity.
In the field of innovation, the distribution of returns strongly favors those who swing for the fences, as the downside is limited, but the upside is uncapped. As Jeff Bezos observed in his 2015 annual letter, in baseball the most runs you can get in one swing is four, but in business, “every once in a while, when you step up to the plate, you can score a thousand runs”.
This payoff structure is one factor that can allow under-resourced startups to surpass established incumbents. Employees do not benefit from scoring a thousand runs in a single swing; they do, however, run the risk of getting fired if they are associated with a failure. Venture capitalists seek to better align incentives by sharing in the upside of any successes and funding founders who have failed before.
Amazon Unbound profiles the efforts of Jeff Bezos to encourage innovation by publicly backing executives who were part of bold projects that did not succeed. As he stated in his 2018 letter, “Amazon will be experimenting at the right scale for a company of our size if we occasionally have multibillion-dollar failures”.
The world is also full of tournaments with asymmetric and convex downside payoffs. Twitter is often described as a game where there is a main character every day, and your goal is not to be it. A high variance strategy is inadvisable there, as the upside is imaginary internet points, and the downside is cancellation.
The most interesting games are tournaments with payoffs with asymmetric and convex upside payoffs. Increasingly, outsiders who enter politics are realizing that a high variance strategy is the best way to break through the clutter and get attention - just make a lot of outrageous statements to see if anything resonates.7 It may be suboptimal for our country to choose our leaders from a pool of the loudest, craziest members of our society, but it’s never boring.
A high variance approach is neither inherently good nor is it inherently bad. It is highly dependent on the nature of the game that is being played.
Baseball is a good example of this. The effectiveness of home run hitters is sensitive to changes in the rules of the game. In 1918, Babe Ruth’s first year as a regular hitter, he led the league with only 11 home runs, and the game was dominated by contact hitters. In 1920, changes to the ball made it easier to hit, ushering in the “live ball” era; Ruth hit 54 home runs that year, triple the total of the runner up.
In the field of active money management, variance from the benchmark index is often thought to be a good thing, to a certain point. For one thing, higher variance from the benchmark is not necessarily synonymous with risk. A simple example would be if you put your entire portfolio in cash. Your portfolio would become risk-free, but your returns will now deviate greatly from the S&P 500.
Another factor is the fee structure. Managers charge a fee on the whole amount managed, even if they passively invest part of it. It is better for you to put half your money in a low-cost index fund and give half to an active manager, than to give all of your money to an active manager who effectively puts half your money in a passive index and then charges a fee on the whole amount.
The question then is one of how exactly the high variance is being generated. One possibility is that the money manager is a sector expert and stays within that narrow sector. This is a perfectly valid way to invest, and a potentially optimal one if that sector is lucrative enough. At least one person in history has made the Forbes 400 strictly through passive investment in public equities in one narrow sector.
The more concerning approach is when variance results from a manager building a concentrated portfolio in a variety of sectors. If venture capital is a field where you can score a thousand runs in a single swing, then public equity investing is a field where you can make a hundred outs in a single whiff. Mature companies have a much lower ceiling, but can still go to zero.
One illustration of this is the career of Bill Miller, who beat the S&P every year for fifteen years between 1991 and 2005 by investing in an eclectic mix of traditional value stocks and tech stocks. In 2008, he doubled down on financial stocks as the financial crisis gained steam, and today he is perhaps best remembered for his appearance at a conference on March 14, 2008, where he boasted he was investing more in Bear Stearns even as it was collapsing that day. All money managers make some bad investments; Miller had the unique misfortune of having his worst investment immortalized in an extended scene in a highly popular, Oscar-winning movie.
It is counterintuitive that a money manager can have nearly two decades of outstanding performance completely wiped away in a matter of months, but such is the brutal math of investing through big bets. When he finally agreed to step down in 2016, his funds sported long-term performance records that lagged the benchmark index.
Bill Miller is far from a unique case. Here is an analysis of the downfall of Bruce Berkowitz, our manager of the decade from 2010, who rose to prominence through successful concentrated investments in banks and insurers, and fell to earth after loading up on Sears, Fannie and Freddie, and St. Joe. And here is a discussion of how storied money manager Ruane Cunniff & Goldfarb put a third of their fund into Valeant, and Oakmark Select bet 15% of their fund on Washington Mutual.
It is easy to see how an ordinary concentrated money manager can hit a few home runs and be seduced into thinking he is no mere Rob Deer, but the Sultan of Swat himself. It takes a few more trips to the plate, or perhaps a bear market, for the truth to reveal itself. Perhaps the overconfidence that results from early success discourages managers from evolving their approach, resulting in later underperformance.
It is useful to contrast Warren Buffett, who has a sixty year track record, with the average “star” manager. Buffett has a deep knowledge of the dynamics of a wide spectrum of quality businesses accumulated over the decades; he puts on a four hour show every year where he demonstrates this. He is as comfortable making $100 billion on Apple today as he was making millions on Sanborn Map six decades ago. Most importantly, he is an astute risk manager, never making a concentrated bet in a risky business.
Extending the baseball analogy, Warren Buffett is Yogi Berra - a unicorn who hits home runs and rarely strikes out.8 In baseball, that gets you three MVP awards and ten World Series rings; and investing, that makes you almost one of a kind.
I would conclude that we have no reason to believe that top performing fund managers are particularly skilled; what sets them apart is their approach of making concentrated bets. Since top performing funds also charge high fees, we should not be surprised that they underperform in later periods.
The most recent academic study I could find suggests that top managers neither outperform nor underperform after receiving an award. We are working with small samples here, so it is difficult to be conclusive.
The most important point to remember is that we are not limited to looking purely at historical outcomes; we can look at the process that generates those outcomes as well, as we should always be doing when we are looking at games that mix luck and skill.
Everyone had a chance to read Bruce Berkowitz’s communications to investors a decade ago, and everyone can read Cathie Wood’s presentations today. Money managers speak publicly about their process all the time, and any prospective client can examine it and determine whether it is adequate. If you are too busy to do that, well, no problem, there’s always index funds!
Addendum: Byrne Hobart points out that this suggests we should see more high variance strategies in electoral politics than we currently do. Those of you that are sports fans know that it is common to see very high variance strategies from a team that is behind at the end of the game and has nothing to lose. For example, in hockey and soccer, the losing team will remove their goalie in search of the game-tying goal, and in (American) football, the losing team will start to go for it on 4th down and hold out for a touchdown instead of a field goal. This even applies at the beginning of games where one team is a big underdog - Malcolm Gladwell wrote about teams with nothing to lose employing the full-court press in basketball.
This would imply that having “nothing to lose” will be highly correlated with high variance strategies. Startups can flout regulations more easily than established companies because of limited liability (as long as the strategy is not TOO illegal) - this is one of the reasons the development of limited liability was so important. In politics, we would expect established candidates to employ low variance strategies, and outsiders and underdogs with nothing to lose to employ high variance strategies. I believe this is a known feature in Congressional elections - incumbents have a massive advantage, so any challenger that manages to win is likely to be a little crazy.
Specifically, my colleague, Abel Osorio.
Sirri and Tufano (1998). It is worth noting that there are other factors that influence fund flows, such as marketing and distribution, but performance is a big factor.
There is definitely a bit of a spectrum - you have founder managed funds, you have a small number of employee-owned partnerships, you have independent funds operated independently but owned by an investment company like AMG, and you funds where the firm is still run by the founder but there are a small number of portfolio managers with high job security. As far as this essay is concerned, the important factor is that all of these structures confer high job security and independence, and they make up a small minority of all funds.
In fairness to the Babe, he supposedly came into the season massively overweight from all of his offseason partying, and missed some playing time due to an ulcer. It happens to the best of us.
This is not meant to insult Rob Deer, who was actually a slightly above average hitter over his career. He struck out a lot, but it turns out that strikeouts are not that bad in baseball, and home runs are really good. I’m sure, however, that even he would agree that he was not exactly Babe Ruth.
Yogi Berra hit 28 home runs and struck out only 12 times in 1950, and somehow only finished third in the MVP vote.