Eight simple (too simple) reasons why I don’t like CAPM

The Capital Asset Pricing Model specifies that the expected return on an asset is a function of the market rate of return plus another factor ("Beta") for the covariance of that asset with the market portfolio.  The intuition is that pro-cyclical assets are riskier and thus they must give you higher expected return.  But I don’t buy the whole Beta bit, especially not for equity markets:

1. For the marginal investor today, the marginal utility of money doesn’t vary much across world-states.  Let’s say you expect to earn a few million dollars over your lifetime and you have access to capital markets.  How much do you care about the covariance of a single stock?

2. Tossing in any second variable will improve predictive performance of the model.  To me the broader multi-factor models just look like data mining.

3. I can see that Beta might lower the expected return to holding gold, a traditional safe harbor in tough times.  I just don’t believe Beta matters for most equity assets.  Yes construction is pro-cyclical but does this affect real world thinking about which stocks to buy?  I think views on cyclicality are dwarfed by idiosyncratic expectational factors about particular facts of the world.

4. Unlike say, profit maximization, CAPM-reasoning will not evolve in the marketplace unless people are at some level aware of the fundamental principals of the theory and take care to minimize systematic risk.  If you are ignorant of CAPM you might have lower utility but you needn’t earn less money over time.  You don’t drop out of the marketplace as a broken down beggar.

5. People compartmentalize their fears.  Insofar as you worry about systematic risk it will affect your human capital decisions and real estate decisions, not your equity investments.

6. Risk affects your equity investments by getting you to diversify.  The story ends there.  Greater fear might mean you buy more individual stocks, but you don’t look into their Betas to prefer one stock over another.

7. Did I mention that ex post Beta is not always accurate as a predictor of future Beta?

8. Fama and French have shown that the line connecting Beta and expected returns has an almost flat slope, at least if we adjust for the size of a firm relative to its book value. 

For risky equity assets in the United States, my preferred economic model is simple.  Expected return equals seven.  That is my model, "Seven."

Plus of course an random or error term.  How’s that for Occam’s Razor?


I've worked in the business of both asset management and corporate finance. What can I say from the practical point of view>

(1) CAPM is correct in its proposition that the market portfolio is the optimum for an average investor. Most actively managed portfolios underperform the broad market index over time. See Morningstar.com; note that this proposition holds for fixed-income portfolios, too.

(2) CAPM is dead wrong as a tool to compute the so-called required rate of return to be used in cash-flow discount models. Not only that most corporate finance valuation cases violate the assumptions of the CAPM - the time instability of betas make this approach about as valuable as tossing coins.

(3) I've used the Markowitz model to "reverse engineer" the assumptions that our pension fund's managers have bnbeen using. They are a bunch of trader-minded folks who don't believe anything more sophisticated than P/E. They put their intuitive projections of expected returns, risks and "optimal" asset allocation which they propose to the board. Then I compute the underlying correlation matrix they intuitively use. Sometimes their intuition looks OK. Sometimes, horrible things appear.

The CAPM is no dogma and every user must be very well aware of its pitfalls and limitations. Despite its shortcomings I maintain that everybody in the finance business should know it very well, not on the superficial MBA level.

"Seven," huh? That's one-sixth of fourty-two...

I forgot my towel today too! I'll have to steal one.

Bill- depends on what kind of manager you are, for a HF its a big deal as increased vol is going to put you lower on a FoFs optimization program. Plus higher volatility on a equal investment basis is going to lower your returns.

tom- I thought I understood this, but I can't explain it, so I guess I don't.

PK - I agree, Beta is so unstable its almost usless as a practical tool. Its my biggest problem with CAPM - practically, there is no actual Beta of an individual asset, its a huge range of values.

I am afraid this time you went too simply indeed:

(1) "How much do you care about the covariance of a single stock?"

You (are supposed to) care about the covariance of a single stock because all these tiny little covariances add up to the total risk of your portfolio (unlike these tiny little variances). You are right that that marginal utility doesn't vary much across states for the average investor, but that's exactly not the point: if you end up being the average investor, then you were lucky (at least in the U.S.) and you don't need to worry. The reason to worry is that in a down state, you are more likely that something really bad happens to you (e.g. losing your job at a time when house prices collapsed). That's the risk you are worrying about, and why you should dislike high beta stocks.

(2) "Tossing in any second variable will improve predictive performance of the model."

That's not true - the other variables which improve predictive performance are essentially all so-called scaled price variables - the price-dividend ratio, Q, market-to-book, and so on. Hence what these multi-factor models are saying is that high price assets have lower expected returns than low price assets. That could be because they have lower risk, or because they are overpriced. Given that true betas are difficult to measure and change over time, even if the CAPM was 95% true you would expect the scaled price ratios to predict future returns.

(3) - (7) are more or less well taken.

(8) "Fama and French have shown that the line connecting Beta and expected returns has an almost flat slope."

But then smaller stocks tend to have higher betas than larger stocks. Firm size may be a less noisy measure of true beta risk for many firms than measured ex-post betas.

Doesn't a pretty simple argument guarantee that returns in excess of the risk-free rate should be approximately proportional to beta?

The argument I have in mind is that, assuming that one can borrow money at the risk-free rate of return, I can *construct* a portfolio that has any beta I'd like, just by buying the market on margin. The excess return on the leveraged portfolio is proportional to beta, by construction. Therefore, one would expect that actual portfolios of securities that have a given value of beta would be priced by the market to have returns that match the constructed portfolio -- otherwise, to the extent that your "real beta" portfolio is very highly correlated with the "leveraged market" portfolio, you can arbitrage one versus the other.

In the real world, of course, no portfolio of stocks will exactly track the leveraged market portfolio, but this arbitrage argument would seem to suggest that (excess) returns will be proportional to beta, and the only the constant of proportionality will be determined by overall investor preference.


I was thinking of pension funds, which are tied to benchmarks.
But HFs hedge (in theory) and they are absolute return vehicles anyway, so volatility
shouldn't matter to them.


Not sure I buy, or perhaps not sure I follow your argument.

If you can identify (and some people believe they can) stocks that have a high alpha - returns not explained by exposure to systematic risk, then that portfolio should outperform the leveraged market portfolio for a given level of risk.

This assumes that some people are 'smarter' than the market, which I believe to be true. Of course, there are far fewer 'smarter' investors than there are people who think they're 'smarter' investors, but that doesn't mean that there aren't people with a genuine skill for stock picking. This is, in my view, the point of difference between CAPM and the way commercial investment actually works. You can start from CAPM, but a true model ought to include both a small number of brilliant investors and a large number of suckers who think that they're brilliant investors.

It's possible that I haven't followed your arbitrage argument, though - can you clarify?


Apparently Siegal does buy the generational stock collapse scenario, according to the issue of Business Week that came out today, pp. 90-91.

it presumes a sufficiently efficient market that we can't get much in the way of alpha.

Right, right. There we go. Thanks Alex, I just got a little lost in your original post.

Of course, that assumption is a big one....

Shouldn't the title of the post be "why I don't like CAPM *as a description of the world*"?

If an arbitrage argument fails as a description of the world, then either (1) transaction costs are high or (2) there's money on the sidewalk. In the latter case, it's still a useful argument.

Fama and French (8) are saying there's money on the sidewalk. I think (7) (predicting beta) is the only point arguing against this view, while the rest are about transaction costs, and thus premature.

The advice CAPM gives you under the assumption that everyone follows it may be lousy, but the advice it gives under the observation that we aren't at equilibrium is great.

Comments for this post are closed