Beta is the correlation of a company's stock price to that of the overall market. As such it gives insight in how volatile a stock's price has been in reltation to the market, thus, how risky the investment is likely to be. Beta is used in the CAPM to estimate a company's cost of capital, hence determining its market valuation.

The value of a stock is therefore based on the assumption that beta is a reliable measure of risk, in the long term.

The question I have is, how reliable is beta for long term risk? Hence, should it be used to measure the value of a company?


1 Answer 1


Though still widely taught, the unfortunate truth is that the CAPM is largely an empirical failure at predicting stock returns. It's still widely used in corporate finance and investment banks for reasons I don't fully understand. How did we get here?

  • In the 70s, the CAPM gained immense popularity because the logic was beautifully simple and empirically, stocks with higher betas appeared to have higher average returns.
  • In the 80s and 90s though, this patten essentially reversed, stocks with lower betas had higher average returns! (Search for Fama-French papers on the CAPM if interested.)

Academic, empirical asset pricing has since moved on from the CAPM to broader factor models such as the Fama-French 3 Factor model, Carhart 4 Factor model , or the Fama-French 5 Factor model. In the current literature, key factors that appear to have forecasting power:

  • Value
  • Momentum
  • Operating profitability (some notion of quality)

All these factor models follow the classic academic paradigm of writing expected returns as a linear function of covariance with some set of factors.

Another line of empirical asset pricing takes a more behavioral approach, writing expected returns as a function of firm characteristics rather than covariance with risk factors. While there's a huge conceptual difference between the two, in practice, the two aren't as different as you might think. High book value to market value stocks all tend to do well or do poorly at similar times and it isn't that different whether: (i) returns are higher for stocks that covary with a portfolio of high book to market ratios (i.e. it's about covariance) or (ii) returns are higher for stocks with high book to market ratios (i.e. it's about firm characteristics).

This doesn't mean beta is entirely useless though! It still remains true that stocks tend to move together. Companies with higher betas are more sensitive to market swings.

Points of caution:

  • Estimated betas of individual companies aren't very reliable. Estimated betas of portfolios of companies are much more reliable.
  • CAPM cost of capital calculations based upon a company's estimated beta are probably trash, but telling your boss (or your corporate finance professor) that might not go over well.
  • $\begingroup$ Thanks, I like your answer. When seeing stocks as actual companies, I think looking at firm characteristics makes the most sense to me. Alhough there is always a level of exposure to the market/economy ehich shold sone how be factored in. Do you know of any papers that look at firm characteristics rather than beta? $\endgroup$
    – Matthias
    Commented Oct 7, 2016 at 10:03
  • $\begingroup$ Are portfolio betas more reliable because that beta is an average of betas whose statistical noises are independent and get hopefully cancelled out with respect to "n", number of stocks? $\endgroup$
    – user121416
    Commented Dec 14, 2022 at 0:33
  • 1
    $\begingroup$ @user121416 That's the right idea, but the word "independent" is too strong: return residuals are almost always cross-sectionally correlated no matter what the model/specification is. This flows through and your errors in estimating beta won't be independent. Averaging (i.e. forming portfolios) is still a helpful technique for some statistical issues! It can still help. It's a pretty low-tech but simple, understandable, and robust way to get more stable estimates. $\endgroup$ Commented Dec 21, 2022 at 18:53
  • $\begingroup$ @MatthewGunn Very interesting. How do we reach the conclusion that they are correlated? (not doubting that they are, i'm interested because you have ten thousand cross sectional stocks and no obvious ordering, so what statistical test would tell that they're correlated?) $\endgroup$
    – user121416
    Commented Dec 24, 2022 at 4:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.