For as long as there have been markets, there have been people who identify and try to exploit patterns in those markets. Some of these rules of thumb have a narrative behind them—one old thesis is that the biggest, best-performing companies are also the last to start dropping when the market turns, because investors get nervous and sell their more speculative holdings to buy the reliable ones. Or there's a psychological effect at play, like the "first five days" heuristic, which finds that performance in the first week of the year predicts how the market will do the rest of the year—in that one, the thinking is that people reset a lot of views in the new year, and the market's near-term performance has an outsized impact on whether they're looking for things that will go up or things that will go down.

But some of these rules of thumb instead require simply inhuman amounts of data, and they turn out to be more durable than those other heuristics. Factors are among them. The story of factors starts with the efficient market hypothesis, which holds that prices roughly approximate all information about companies, such that an investor can more or less choose how much risk they want to take, but can't get outsized returns without increasing risk.

That model runs into an interesting problem: if you look at enough historical data, you can find cases where some companies did reliably outperform the average, even after accounting for risk. Small-cap stocks, for example, have outperformed large ones over long periods.¹ Stocks that are cheap on a price-to-book basis have outperformed expensive ones (the “value factor”). Even more strangely, stocks that have recently outperformed tend to keep doing so—albeit slightly (this one’s called the momentum factor, naturally). And there's a whole zoo of other factors: high dividend payers tend to outperform; high-margin companies outperform; companies that reinvest conservatively outperform the ones that always double down.

Regardless, the bottom line is that it shouldn't work this way; there shouldn't be a free lunch. In an efficient market, investors get paid for taking risks, so if they're getting paid extra, there must be some risk that they’re unknowingly taking.

There have been various attempts to prove exactly that—for example, to demonstrate that the value factor (stocks that are cheap on a price-to-book basis) means buying companies that face more operational uncertainty. The thinking there is that more bad things that can happen to a tin mine than can plausibly happen to, say, Coca-Cola or Visa. But while value stocks do have sharper operational drawdowns than growth stocks, they also tend to be in industries that have been around long enough for mean reversion to be a big contributor to returns² .

If factors work—and they mostly do, though performance for some of them went through a difficult period in the last few years—then the next question to ask is: why do some fairly simple strategies generate excess returns?³

If factors work, the next question is: why? And this doesn't just mean asking why some companies are statistically more likely to produce good returns on capital, but why investors systematically underestimate them. There are decent stories for most of these:

• For quality-related factors, the basic bet is that investors go where the action is. The more consistent a company's margins, and the less it reinvests, the less the business changes. That means analysts often don't have much to do. ("Based on my extensive research, it turns out that people really like affordable and consistent food available at numerous locations, so McDonald's is still looking good.")⁴ Of course, these companies can get exciting again when they lever up, but that leverage makes them screen worse on quality, so it's still captured by a factor model.

• For momentum, the temptation is to say that people like to chase assets that have moved recently, and to an extent that's true. But that can't work forever, and if that were the main explanation then momentum stocks would experience extreme reversals that would destroy the factor's long-term record. Another explanation is that at least some investors underreact to changes in fundamentals. It's sort of like a single-asset version of Andrew Lo's adaptive markets hypothesis: if who is the smartest analyst for a given asset changes over time, then at any given time the newest smartest-analyst probably won't have the asset base necessary to push the stock all the way to where it ought to trade.

There are two general ways to use factors. One is to run your entire portfolio based on them, i.e. to build a screening tool that gets you a list of companies that fit the factors you expect to perform well, or just that fit a bunch of the classic factors, and then to buy them. There are plenty of tools for doing exactly this.

But a more interesting option is to use them as a backwards-looking tool: this page will give you factor exposure for specific stocks, or for an entire portfolio. It's useful to dump in things you own, or things you made or lost a lot on, just to see 1) which factors you tend to like, and 2) which factors like you back. There are some investors with a talent for digging through value stocks—which is basically a process of looking at fifty cheap companies to find the two or three that don't deserve to be that cheap. There are others who excel in chasing momentum. (And within that there are directional differences; I've found that my track record of buying stocks after they've doubled is a lot worse than my track record shorting them after they're down by half.)

Multi-manager/pod funds do this all the time: they'll often require managers to maintain factor-neutral portfolios (i.e. for every cheap stock you buy, you need to find an equally-cheap company you're excited to short). The funds that don't do this tend to find themselves with large factor exposures they weren't necessarily aware of; you can basically see this happening live on days when large-cap growth stocks are all down and heavily-shorted stocks are all up. You don't have to be as risk-obsessed as those funds, because you're probably not levered 6 or 8 to 1 the way they are. But it does pay to borrow some ideas they use to get the most out of portfolio managers, and to apply them to your own investing.

Read More in The Diff

The Diff has cited factor models a few times:

• This writeup of small conglomerate Loews looks at its performance as an example of cases where value underperforms ($).

• This piece looks at passive investing as its own quasi-factor.

• This piece is a way of thinking about momentum by looking at all-time highs.

• And in this one, we examine whether "value" can be done in a systematic way at all ($).

A Word From Our Sponsors

The Life-Changing Concepts newsletter gives founders and unfair advantage in business with mental models. Get a summary of 20 useful mental models when you join. It’s free, join 7,000+ founders.

1. This signal is old, but controversial, since a) it’s heavily influenced by just a few historical periods, and b) the smaller the stock, the noisier the data. Survivorship bias and high bid/ask spreads make this one tricky.

2. We've been using tin for millennia and have been electronically exchanging value for less than a century, so there are stronger historical priors that the mining business will eventually return to some reasonable level of returns.

3. Now is a good time to note that these excess returns are something on the order of 1-2 points of excess performance annualized, over long periods. The exact attribution depends on how many factors you use; many of them overlap, and some (like value and momentum) offset each other somewhat.

4. In fast food there's a thin line between deep primary research and hazing. I knew an analyst who covered the restaurant sector when KFC first launched the Double Down, and his portfolio manager made him purchase and eat one.

Help us learn more about you!

We'd like your feedback so we know what to cover, so if you haven't already, please fill out this one-minute form.

Join the discussion!