Lesson 1 of 7
CAPM and its record
The model prices risk with market beta alone. The best known test of it, over 1963 to 1990, found little left once size and book to market were allowed for.
By Javier Sanz
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
In one minute
- The model's claim is narrow: expected return is the risk-free rate plus one company-specific multiplier times one market premium.
- The best-known test of that claim, over 1963 to 1990, found market beta had little remaining relation to average returns once size and book to market were allowed for.
- Practitioners kept using it because the model still needs some multiplier and beta is the one every data source already computes.
The capital asset pricing model is the arithmetic behind almost every discount rate in use. It is also a model whose central empirical claim was challenged more than thirty years ago and never fully recovered.
The model, as one line
What it actually claims
Three things, and they are stronger than they look.
That only the risk you cannot diversify away is priced, so everything specific to a company drops out. That the whole of that risk for one company is captured by a single number, its co-movement with the market. And that the relationship between that number and expected return is a straight line through the risk-free rate.
What the tests found
Fama and French, writing in the Journal of Finance in 1992, sorted US stocks on size and on book to market over 1963 to 1990 and found that once those two were accounted for, market beta had little remaining relation to average returns in their sample.
That is not a claim that beta is meaningless, and the paper is a measurement over one long window in one market rather than a law. It is a claim that the single-factor version of the model did not describe the cross-section of returns that were actually earned, which is the thing the model exists to describe.
Why it survived anyway
Because a valuation still needs a company-specific multiplier, and no replacement has the same properties: a beta is computable from prices anyone holds, comparable across companies, and produces a single number a spreadsheet can take. The alternatives are either multi-factor models with their own estimation problems or a judgement with no procedure behind it.
The guard rail that admits the problem
The beta in that table is not the measured one. Procter and Gamble's two-year weekly beta measured 0.24 and is held at a floor of 0.30 before it reaches the cost of equity; the bound runs from 0.30 to 3.00, and the source trail records when it was applied.
The reason the bound exists is the honest version of this lesson. A measured beta far outside that range is more often a statistical accident of a short or thin price history than a property of a business, and a model whose only company-specific input needs a clamp is telling you how much weight that input can carry.
Check yourself
What did Fama and French report about market beta in their 1992 paper?
- That beta predicted returns better than size or book to market
- That once size and book to market were accounted for, beta had little remaining relation to average returns in their 1963 to 1990 sample
- That beta should be measured over longer windows
Show the answer
Answer: B.
It is a measurement over one long US window, and it challenged the single-factor model's central empirical claim rather than the arithmetic.Why do practitioners still use the model?
- Because the 1992 result was later withdrawn
- Because a valuation needs a company-specific multiplier and beta is the one that is computable, comparable and already published
- Because regulators require it
Show the answer
Answer: B.
The alternatives are multi-factor models with their own estimation problems, or an unlabelled judgement. Beta survives on availability rather than on its record.Why is a measured beta held inside a range before it reaches a cost of equity here?
- To make companies look more similar
- Because a beta far outside the range is usually an artefact of a short or thin price history rather than a property of the business
- Because the model requires a beta above 1
Show the answer
Answer: B.
The bound runs from 0.30 to 3.00 and the source trail says when it was applied. A single input that needs a clamp is an input to lean on lightly.Sources
- Sharpe's capital asset pricing model: expected return as the risk-free rate plus beta times the equity risk premium.
- Fama and French, Journal of Finance, 1992: size and book to market against market beta over 1963 to 1990.
- The product's own stored cost of capital inputs for the two named companies, with the 10-year Treasury par yield stored 2026-09-15 and a 5.5 percent premium labelled an assumption.
What this leaves out
- The 1992 result is one sample, one market and one period. Later work disputes its interpretation, and this lesson does not settle it.
- The 5.5 percent premium used in this product is an assumption and is labelled one wherever it appears. It is not a measurement.
- No cost of equity is suggested here for any company.
- Nothing here is advice about any security.