What beta is measuring

Beta is the slope of a regression: a security's returns plotted against the market's returns over some past period. A beta above one says the price moved proportionally more than the market across that window. That is the entire content of the measure, a statistic about two price series, and it says nothing about either business.

Four choices behind any published beta

Four decisions are made before a beta exists, and all four are ordinary practice.

Publishers rarely say which four they picked, and the figure above shows why that silence matters: the four rows span 4.00 points of cost of equity for one unchanged company, and the three broad-index rows alone span 2.50.

The rate moves the rows equally; beta does not

What beta is not measuring

Beta measures how a quoted price moved with the market. Marks argues that risk is the probability of permanent loss (A loss that is realised and never recovered, as against a price that falls and later recovers.)permanent loss: A loss that is realised and never recovered, as against a price that falls and later recovers., which a price series does not observe. A company whose price stayed steady while its business deteriorated had a low beta and a rising chance of that loss. A thinly traded stock, a recent spin-off or a short listing gives the regression too few informative prices, and it still returns a number.

It is also a levered number

The coefficient measured on a share price describes the equity of whatever debt load the company carried across that window. Modigliani and Miller's 1958 result says financing alone does not change what the whole business is worth. It does change who carries the risk: more debt makes the same business riskier for its shareholders, so the beta measured on the shares rises with leverage even when the business has not changed. A company that changed its borrowing partway through the window has a beta blended across two different capital structures. Analysts unlever a beta before comparing companies with different debt loads, for the same reason.

What the record says, and what to check instead

Fama and French, writing in the Journal of Finance in 1992, found that once a company's size and book-to-market ratio were accounted for, market beta had little remaining relation to average returns in their 1963 to 1990 sample. That has not stopped practitioners using it, because the model still needs some company-specific multiplier, and beta is the one every data vendor already computes.

The more useful check reads the risk factors instead. Open Item 1A of the company's most recent Form 10-K and ask, for each risk it lists, whether a regression of weekly returns against an index could have detected it. A customer concentration or a loan covenant rarely moved the price on the days the index moved, so the answer is usually no. That list, not a better beta, is what a cost of equity leaves out. The model leaves company-specific risks out on purpose, because a diversified holder can spread them; they belong in the cash-flow scenarios and the margin of safety, not in a higher rate.

Check yourself