The First Crack in the CAPM: the Size Effect

In the late 1970s the CAPM was not a model. It was the model. One equation, one risk measure, one clean story: a stock's expected return depends on its beta and nothing else. Anything else you might know about a company, its industry, its history, its size, was supposed to be irrelevant once you knew its beta.

Rolf Banz's doctoral work broke that. In 1981 he published evidence that a piece of information anyone can read off a screen, how big the company is, predicted returns that beta could not account for. It was the first major documented anomaly, and it cracked open the door that Fama and French, and everyone after them, would walk through.

The problem: a variable that should have been irrelevant, wasn't

Banz's test was straightforward, which is part of why it was so damaging. He looked at US stocks over 1936 to 1975, sorted them by total market value (share price times shares outstanding), and asked a simple question: after adjusting for beta, do the small ones and the big ones earn the same?

They did not. Small firms had, on average, significantly larger risk-adjusted returns than large firms. Over that forty-year window, the small end of the market beat the large end by more than the CAPM said it should.

This should not have been possible. Market value is the most public number about a company. If small firms genuinely offered better risk-adjusted returns, everyone would buy them, prices would rise, and the advantage would vanish. That it persisted for four decades was a direct challenge to both the CAPM and the efficient market hypothesis.

The key idea via analogy: the anomaly that ate its own model

There is a subtlety in what Banz actually proved, and it is worth being careful about, because it is the source of a debate that continues today.

Imagine you have a scale that is supposed to measure weight, and you find that objects painted red consistently weigh more than the scale predicts. There are two possible conclusions:

1. Red paint really does add weight. Colour is a genuine, previously unknown source of mass.
2. Your scale is broken. It systematically under-measures red objects, and colour is just revealing the flaw.

Banz's result has exactly this ambiguity, and he said so. Either small size is a real, separate source of expected return (something about small firms genuinely earns more), or the CAPM's beta is simply a broken measuring instrument that fails to capture the true risk of small companies. The data cannot tell you which. Banz was explicit that he had no theoretical explanation for why size should matter, and he treated that as a genuine problem with his own finding rather than glossing over it.

That honesty is what makes the paper good science, and it also frames the next thirty years of argument: is size a risk factor, a mispricing, or a measurement failure?

He also noted, importantly, that the effect was not linear. It was not that returns rose steadily as firms got smaller. The effect was concentrated in the very smallest firms, the tiniest end of the distribution. That detail turned out to be prophetic.

Why it mattered

• It was the first major anomaly. Before Banz, the CAPM stood essentially unchallenged. After Banz, hunting for variables that beta could not explain became a legitimate, and eventually enormous, research programme. Every factor paper that followed owes something to this one.
• It became a factor. Twelve years later, Fama and French built the size premium directly into their three-factor model as SMB, small minus big. It is one of the two new ingredients they added.
• It created the small-cap industry. Small-cap funds, small-cap indices and the entire style-box grid that organises fund management are downstream of the idea that firm size is a meaningful investment dimension.
• It taught the field to ask "compared to what?" The realisation that a benchmark can be wrong, and that apparent alpha might be a defect in your yardstick rather than skill in your manager, is one of the most important ideas in performance measurement, and this paper is where it landed hard.

The honest limitations

• The premium has been feeble since publication. This is the big one. The size effect, measured plainly, has been weak, inconsistent or absent in the decades after Banz wrote. It has done poorly enough that many serious researchers now question whether a standalone size premium exists at all. It is the leading exhibit for the argument that publishing an anomaly kills it.
• It hides in the least tradeable corner of the market. Because the effect concentrates in the very smallest firms, it lives precisely where bid-ask spreads are widest, liquidity thinnest and trading costs highest. A large fund cannot own those stocks in meaningful size. Much of the paper premium is not reachable.
• The data has traps. Databases of tiny companies from the 1930s to the 1970s suffer from survivorship problems and delisting complications. Some of the measured premium may be an artefact of how failed small companies were recorded.
• It may not be size at all. Later work argues that once you control for quality (screening out the small firms that are simply unprofitable junk), a size premium reappears. If that is right, size was never the real story: it was a noisy proxy for something else.

The one-line takeaway

Banz showed that small companies had earned more than their beta could justify, which was the first serious crack in the CAPM, though the premium he found has been so weak since he published it that size is now the cautionary tale of factor investing rather than one of its triumphs.