Next time someone tells you they’re great at investing, think twice…
How many times has your friend told you that they’ve made a great investment? It could be in a stock or crypto, or for those of you born to be gamblers, options or NFTs. I give credit where it’s due, but with that said, I think there are a few things that need to be cleared up when looking at fund/investor performance.
Based on chance alone, a lot of investors do well when markets are up
If we run a simulation on the one-period return for investors based on the average return and standard deviation of the ASX200 over the last 10 years (9.3%, 13.6%, respectively; Morningstar), based on a sample of 100 investors, these would have been their returns.
Accordingly, when markets are even more bullish over the short term, such as over 2021 (eg. the ASX200 returned 17%+; here), it becomes quite easy to see that due to simple randomisation, and nothing more, quite a significant portion of investors beat the market (in expectation, of course, this is simply 50%, if you assume a normal distribution).
This simple statistical demonstration, of course, is the foundation for the ‘monkey stock-picking experiment’, where a blindfolded primate was found to be a better stock picker than the average investor; both professional and amateurs. The intuition is simple — a random stock picker is subject to the same ‘random’ forces of the market that a human is, but does not suffer from the significant psychological biases that hinder us humans’ returns.
Cathie Wood, and the importance of performance attribution
When writing this, I thought immediately of Cathie Wood’s ARK Innovation Fund, which flaunted a 134% outperformance of the S&P 500 benchmark in 2020. With the crash of the tech sector, Wood’s fund is more than 50% down from its peak, now underperforming both the S&P and Nasdaq benchmarks. This illustrates the importance of attribution of performance rather than simple evaluation. To do this, you need to ask the rather complicated question:
How do I know that the investor isn’t just doing well because of luck?
In the case of Wood, her risky tech bets in Tesla, Zoom and Coinbase (just to name a few) were rewarded with enormous upside, but met with commensurate losses during the tech crash throughout 2021–22. Her fund’s performance highlights the importance of looking at fund performance over a longer time period to understand a fund manager’s ability to generate consistent alpha, rather than short-term gains that might be unsustainable.
But then, if some investors do well, and others do poorly, shouldn’t this balance out? And shouldn’t we hear about both ends of the spectrum?
The issue of survivorship bias, and why you only hear from investors when they’re up
Let’s take a quick detour from our boring investing piece.
Survivorship bias — what’s the issue?
If the red dots represent where the war plane has been hit over many years, where should you build more armour for the plane?
Logic would simply point to the wings, the tail and the centre part of the plane body, since this is where the sample indicates that the plane has been shot most. However, making this inference would be falling victim to survivorship bias, a form of sampling/selection bias. Sampling bias occurs when the probability of attaining an element/individual from a population for a ‘random’ sample is higher for certain elements of the population — thus meaning that the sample is no longer random (more here).
Applied to the war plane example, it is highly likely that the plane is hit by bullets all over (intuition would tell you that someone shooting at the plane would find it very hard to aim specifically at a part of the plane, so it should be roughly random over time) — but since most flight-critical machinery operates from the body of the plane, it is likely that planes hit in their centre would likely never make it back to base for the data to be recorded. Eddie Woo makes a great video explaining this point clearly.
Other examples of sampling bias include:
- Customers being surveyed by their service providers (eg. banks, utility companies) — if you send out an email to 100 of your customers at random, and (as an example) their satisfaction scores out of 10 are also random (say, normally distributed), will your sample report the same results? Probably not, because who is most likely to respond to a survey? Most likely is probably customers very dissatisfied, following this would be customers very satisfied, and customers in the middle range (eg. score of 5–8) will not bother replying, because of the effort.
- Voters being surveyed over the phone — you’re instantly biasing your sample by taking the results of people who pick up, but what about the ones that don’t? Are people who are less likely to pick up more likely to vote one way or another?
- ‘Random’ surveys in public, eg. shopping malls — if you ask 100 people on a Tuesday afternoon, at a shopping mall, if they go to the gym, would this be a random sample? Obviously not — people out at that time would likely be individuals who do not work, who are not in school, and might even have more time on their hands — all factors which might bias whether they go to the gym.
Long story short, it’s often very hard to get a truly random sample.
So, how does this apply to investors?
Simply put, investors have egos — they don’t like telling others when they’re down, and they love flaunting their big wins. Naturally, this gives us a false impression of how investors are truly doing in the market — outperformance or simply positive returns, irrespective of what portion of all returns they form (eg. in the simulation above, this is ~86%), are much likely to form a greater portion of what is reported, than if investors were exposed to all investor performances transparently.
There is significant academic literature to support this bias in the context of investor performance, eg.:
- Walters et. al (2021) argue that “tend to recall returns as better than achieved and are more likely to recall winners than losers”.
- Elton et. al (1996) argue that “mutual fund attrition can create problems for a researcher because funds that disappear tend to do so due to poor performance”. In essence, the worst funds’ results are usually never reported because they are dissolved before this can be done.
Of course, this can occur across different dimensions. It could occur:
- Across individuals — only the best performing individuals’ investing performance is reported (this is actually not so much of a problem if you’re looking to invest in a fund, but it is certainly a problem if the individual is telling you something like — “you need to get onto crypto dude, I’ve been making so much money!”)
- Across time — investors only flaunt their returns when they are up, but don’t talk about their performance in prior periods, or how long they have been in the market for (this goes back to the randomness point above)
- Across investments — investors only talk about their most successful investments (eg. they talk about being up 100% on one stock, but their whole portfolio is down YTD)
The rundown
In short, the two reasons we have talked about for bias in perceiving outperformance are:
- Outperformance due to randomness alone
- Outperformance is reported far more than underperformance
As an investor, how can you combat this? No financial advice here, but some practices I would try take into account might be:
- Don’t just look at what the front cover of the news is showing you and invest in it. Look at other funds, stocks and investors and how your desired investment weighs up. Particularly, you want to look at investor returns over longer time periods, and compared to risk-appropriate benchmarks.
- Try look at the investor’s methodology and see if their returns can be attributed to their stock picking skill, or randomness (eg. look at investment memos, decks). You want to know if these returns can be replicated over a longer time period
- When investing in a new asset class, don’t just listen to one person — look at how the market as a whole has been doing, and how others have been burned
Hope you all found this helpful!
