One of the Most Successful Active Managers of all Time Shows Why Active Management Doesn’t Work
A little while ago, we used the Fidelity Magellan fund as an example in an article on style drift. That got me thinking about Peter Lynch. He and the Magellan fund are a really interesting case study on investing with active managers.
There’s little question that Lynch had “it.” You can slice and dice the data any which way you want, but it’s pretty clear he was able to beat the market on a risk-adjusted basis. Often he is pointed to as “proof” that active management works.
But it’s actually the opposite: he’s the exception that proves that it doesn’t work.
Before we dive into the numbers, let’s back up and think about active management in general. For active management to work, a manager doesn’t just need to be able to predict what will happen next, they need to know if it will be better or worse than what the market expected. Active management is like betting on sports – you can’t just pick who’s going to win, you need to be able to consistently beat the spread.
Look at the market’s reaction to the monthly job report. If the market simply reacted based on whether something good or bad happened, it would go up when the economy added jobs and down when it lost them. But it doesn’t work that way.
The market moves based on what happened relative to what people thought would happen. If the consensus expectation was roughly 300,000 new jobs but “only” 200,000 jobs were created, the market would go down. If the market thought 100,000 jobs would be lost, but only 50,000 were lost, it would go up.
As you’ve probably guessed by now, being able to predict the real and expected outcomes of one jobs report alone would be tough. Being able to do so consistently (in jobs reports and beyond) is incredibly rare.
Some of the best evidence of the rarity of consistently successful active managers is subsequent returns studies. The idea here is that if there is value – even unevenly distributed value – in active management, then if you look at the funds who beat the market in one period, they should be more likely to win in the next period. If Active Manager Adam was able to beat the market last time, he should be able to do it again and again and again, right?
But that’s not what happens. We don’t see winners repeating.
In fact, the data shows that a fund beating its benchmark in one period tells us nothing about the likelihood of it doing so again. The familiar disclaimer “past performance is not indicative of future returns” gets slapped on pretty much anything investment-related. And it’s true. Just because something – a fund, a stock, a manager, etc. – beat the market in the past, that’s no reason to think it will do so again.
Now, this doesn’t mean it’s impossible for an active manager to beat the market through skill rather than luck. It just means it’s really rare. So let’s examine one of the few examples of a manager who clearly possessed that rare quality.
Peter Lynch Was a Really Good Investor
When we say Peter Lynch was a good investor, we aren’t simply saying he beat his benchmark for a while. A lot of managers are out there beating benchmarks – but that’s either because they are taking a different amount of risk than their benchmark, or they possess another rare quality (albeit more fleeting than Mr. Lynch’s): sheer dumb luck.
Our best tool for measuring a manager’s performance is regression analysis. It allows us to look at their actual returns – rather than what they say about their fund – and understand exactly what is going on from a factor exposure perspective. Normally, when we perform such an analysis on high-performing funds that everyone is excited about, we find out their performance is completely explained by the risks they’re taking. In other words, they aren’t adding any value – it’s all about how their funds are allocated. So it’s only fair to acknowledge when the models show that there is something more going on.
And in Peter Lynch’s case, there was something pretty massive going on. When you run the Magellan fund’s returns during Lynch’s tenure through a three-factor regression model, he was adding 0.78% per month. That’s almost 10% per year. And the chance that this was just noise is pretty much nil – there’s roughly a 0.0000007% (or 1 in 150 million) chance that this was random.
I guess it’s possible to argue that this was random. After all, if everyone in the US ran their own mutual fund, there would statistically have to be two people that good through pure luck. But this is grasping at microscopic straws.
The regression model I’m using didn’t actually exist in Lynch’s golden age (the early ’80s), and the fund benchmarked itself against the S&P 500 Index. So how did he do against the S&P 500?
Before we look at this, I want to make two points:
- Comparing the Magellan Fund, against the S&P 500 fund isn’t the best comparison in the world since the Magellan fund is taking more small cap risk – which means we would expect it to have higher returns on average.
- At that point, there weren’t all that many more appropriate options for mutual funds to compare themselves against. They didn’t the massive numbers of hyper-specific indexes floating around that we do today, especially that regular investors were familiar with.
A good way to compare the two funds is to look at how often the Magellan Fund beat the S&P 500 Index.
Over the twelve full years that Lynch was at the helm, he beat the S&P 500 Index all but two years (1984 and 1987), or 83% of the time. Pretty impressive, right? Especially considering that back then, most people weren’t being constantly bombarded by investment information like they are now.
Lynch ran the Magellan fund for 156 months (May 1977 to April 1990). Of those 156 months, he beat the S&P 500 97 times, meaning he beat “the market” just under two-thirds of the time. This is astoundingly good.
In short, Peter Lynch was able to beat the market.
But we’re making that statement after looking at a whole bunch of data twenty-five years after he retired, when it doesn’t do anyone any good. For information like this to be useful, we need to be able to identify superior managers while they are still managing money and we have time to throw our money at them.
Is that possible?
Let’s look at how long it takes to statistically identify that this skill exists. If it takes almost his entire tenure to identify his superiority, then active managers – even the really good ones – have a pretty big problem.
To test this, we can step through Lynch’s returns, running regressions every month until we can confidently say this is real skill, not random noise. So we need a threshold that helps us separate luck from skill. Let’s set it at 0.1%. So we want less than a 0.1% chance that Lynch’s outperformance is just random noise.
It doesn’t actually take long for Lynch to prove it. In just over four years, from May 1977 to May 1981, he proved he was adding value above and beyond his factor exposure. For comparison, it takes only five months fewer (beginning in January 1926) to prove that the S&P 500 has higher returns than one-month US Treasury Bills (largely considered the safest – and lowest returning – investment out there).
This is really impressive.
Peter Lynch Shows Why You Can’t Pick Winners
So how do I get from “Peter Lynch is a superhuman investor who can see the future” to “active management doesn’t work”? It seems contradictory, to say the least.
Let’s look at what happened after Lynch left. How good was Peter Lynch, who clearly knew what it took to beat the market, at identifying someone else who could do the same?
That’s the situation most investors are in. For you to be better off using active management, you need to identify someone who can consistently beat the market on a risk- and fee-adjusted basis. Otherwise you’d be better off with a plain, boring, passive investment strategy.
It turns out even Peter Lynch couldn’t do it. We’ll look at the numbers in a little bit, but let’s stop and think about the situation Lynch was in when choosing his successor. I compared his situation above to the situation most investors are in when choosing an active manager, but that’s not quite right. He was in a much better position than everyone else.
As you can probably imagine, Lynch was pretty well-revered by the end of his tenure. Everyone wanted to work with him. That means he had infinitely more access to any information he wanted about the people who would succeed him.
He worked with them, so he knew how they went about investing. He knew their temperament. He knew their ideas. He knew what picks they did and didn’t make and why. He knew how they were to work with – would people follow them? And beyond that – he knew what was going on in their lives. In short, he had any and all of the information he might have conceivably thought was meaningful.
And he still couldn’t pick someone who could beat the market.
Lynch chose Morris Smith as his successor. Unfortunately for our analytics, Smith left the Magellan fund just two years after he started, which makes our regression analysis difficult. To get around this, we’ll include the manager that followed Smith – Jeffrey Vinik.
Lynch may not have directly picked Vinik (though it’s hard to believe that he didn’t have a good deal of influence on the decision – it was only two years after he retired), but this would have been the investment experience that an investor in the fund would have had.
So let’s look at the same two tests that we used for Peter Lynch: how they compared to the S&P 500 Index on a monthly basis, and (the much more informative) regression analysis.
With the S&P 500 comparison, Lynch’s successors look pretty average. Smith and Vinik ran the fund for a combined seventy-three full months. In that time, the Magellan Fund beat the S&P 500 Index over thirty-nine of those months, so 53% of the time.
If we separate out the two managers, there’s a tantalizing suggestion that Lynch may have been onto something. Smith ran the fund for twenty-six total months and beat the S&P 500 Index in fifteen of those months. He outperformed 58% of the time, so he wasn’t too far off from Lynch.
But it’s too small of a sample size to do anything with. Remember, it takes nearly four years to show that stocks have better returns than one month Treasury bills. This is just way too far inside the random noise to make any sort of statement.
Vinik had a longer tenure to work with – forty-seven total months – but not quite as promising of results. Over his time as manager, he beat the S&P 500 Index twenty-four months – 51% of the time. In other words, a coin flip. That’s about what you would expect from random chance.
But these monthly observations are more anecdotal than real data. What does the regression analysis tell us?
Over the tenure of both managers – from 6/90-6/96 – there was no alpha (added value gained specifically from their guidance) to be found. The returns of the fund were explained by its factor exposure (aside from the random noise from active management).
The same thing is true if we only look at Vinik’s tenure (8/92-6/96). There was no discernable positive or negative alpha. Aside from adding some randomness from his active management, he might as well be running an index fund.
What Lessons Can We Learn From Peter Lynch?
We can learn all sorts of lessons from Lynch, but one is particularly applicable to typical investors: picking winners is hard.
Even Lynch, who was a winner himself, couldn’t do it. Or at least his pick(s) weren’t good enough for us to be able to separate their skill from the random noise of the market.
Picking winners is more than just finding a fund with a five-star rating from Morningstar and pretty good three- and five-year returns. You have to find someone who can out-predict the entirety of the market for years at a time. They need to be able to see the connections that everyone else doesn’t, and then figure out what they will mean for security prices.
They need to be able to predict the future. That’s a pretty big ask.
And even with all of the information available to him, and the skill to do it himself, Lynch couldn’t identify anyone else who could follow in his footsteps.
 For the entirety of this analysis I will be using the Fama-French 3 Factor Model when I refer to regression analysis. Ken French is my father.
 The alpha over the regression time period was -0.03, with a t-stat of -0.17. Both the point estimate and the confidence are so low that this is just a more verbose way of saying 0.
 Vinik’s alpha over the period 8/92-6/96 was -0.14 with a t-stat of -0.50. Worse than Smith’s numbers, but still very much in the “more verbose way of saying 0” category.