Why Do Bad Models Persist?

History is cruel to theories. But few theories have so thoroughly failed the test of reality as the Capital Asset Pricing Model.

Fischer Black, Michael Jensen, and Myron Scholes offered sufficient evidence to reject major parts of the model in 1972. By 2004, in a review of 40 years of evidence, Nobel prize winner Eugene Fama and his research partner Ken French declared, “despite its seductive simplicity, the CAPM’s empirical problems probably invalidate its use in applications.”  Vanguard has over a billion dollars in a minimum volatility fund that bets every day against the predictions of the theory.

The core prediction of the model — that stocks with higher price volatility should produce higher returns — has failed every empirical test since 1972. Here is a graph from Fama and French showing CAPM’s predictions relative to reality.

Figure 1: Average Annualized Monthly Return vs. Beta for Value Weight Portfolios Formed on Prior Beta 1998–2003

Source: Fama and French, The Capital Asset Pricing Model: Theory and Evidence

Yet the CAPM is the centerpiece of most business school finance courses and surveys suggest it is used by over 70% of CFOs for capital budgeting (ever wonder why the track record of corporate M&A is so poor?). The language of alpha and beta, of weighted average cost of capital and expected returns, permeates the field of investing.

Why has this bad model survived?

The model is useful to academics because it provides a theoretical justification for diversification. Markowitz starts his paper on portfolio selection by presenting the conundrum that CAPM was meant to solve: “The hypothesis (or maxim) that the investor does (or should) maximize discounted return must be rejected. If we ignore market imperfections the foregoing rule never implies that there is a diversified portfolio which is preferable to all non-diversified portfolios. Diversification is both observed and sensible: a rule of behavior which does not imply the superiority of diversification must be rejected both as a hypothesis and a maxim.”  But this is coming to the right answer the wrong way.

The model is useful to practitioners because the model’s workings are complex and reliant on assumptions — and can thus be tweaked to provide a pseudo-scientific justification for almost any answer. Your project doesn’t have a high enough NPV?  Choose a different beta that makes the project look lower risk. Investment outside of the core competency?  CAPM shows diversification is better than concentration.

Dogmatic reliance on CAPM is at best useless and at worst dangerous but, either way, has no place in making investment decisions. Its persistence both in academia and in practice — despite clear evidence of failure — is concerning.

Yet many insist that without bad models like CAPM, we’d be left with no way to make decisions and that some model is better than no model at all. How would we know that it was important to diversify rather than putting all our eggs in one basket? How else would we price assets? Won’t we be left in the dark with nary an Excel model to guide us?

This is where the Verdad approach begins. Start with the assumption that nobody knows what’s going to happen. Start from that premise of absolute skepticism and research proven methods that actually do work for making good investment judgments. From these solid foundations, build a larger process that relies only on forecasting methods and investment models that have proven empirically true over long periods of time. There will be systematic error to exploit in the markets until managers stop relying on this 1970s' era thinking.