Mullins [1982, p. 110, and 112] described the usefulness of CAPM for both security analysis and capital budgeting. He emphasised that although some of the assumptions are unrealistic, the CAPM provides a useful measure of risk premia and that, in the spirit of Friedman

'... the true test of the CAPM, naturally, is how well it works.'

and proceeds

' ... much progress has been made in the development of richer asset-pricing models. As of yet, however, none of these more sophisticated models has proved clearly superior to the CAPM.'

He also stressed that the CAPM should not be used as a precise algorithm, but should be used as part of an overall strategy. This section describes some of the more common applications of the CAPM.

Portfolio Selection

Sharpe & Cooper [1972] applied the theory of the CAPM to portfolio selection, by constructing a portfolio strategy for investing in equities. Their emphasis was on using the basic principles of the CAPM to derive a strategy to yield enhanced performance and predictability. They used the CRSP database of NYSE equity returns from 1926 to 1968 to assess a strategy, using βs based on a equity's performance over the previous five years. The proxy used for the market portfolio was the Fisher Index. As opposed to Black, Jensen & Scholes [1972], Sharpe & Cooper only included equities that alwayshad a five-year history. Their adjustments were also annual, rather than monthly. In addition, they did not use a true β based on appreciation plus dividends, but a market sensitivity measure based solely on appreciation. They used this measure as they considered dividend yields to be fairly stable, and the correlation between β and the (appreciation only) market sensitivity measure was high (ρ = 0.95). Equities were grouped into ten risk-return classes based on their βs, and each portfolio strategy entailed investing in equal market-value weighted amounts within each portfolio class. Adjustments of risk class and market value were calculated annually.

Sharpe & Cooper found that most portfolios did not change their risk grouping more than one step from year to year. They therefore concluded that past βs are a good indicator of future βs, with relative returns of each risk-return class conforming with the expectations of CAPM. This supported the application of the CAPM principlesto portfolio strategy, using historical covariance with the market to predict the future risk and expected return of an equity. However, it should be noted that their analysiswas confined to the construction of equity portfolios. They did not consider a wider definition of the market.

Gentry & Pike [1970] tested the risk-return hypothesis on portfolios of equities, held by 34 life insurance companies, with data from 1956 to 1967. Their findings supported the positive linear relationship between risk and return. While not full tests of the CAPM, these studies were important because they promoted the use of the CAPM to describe the relative risk of numerous types of investment decisions.

Ward & Saunders [1976] applied the CAPM to British data, specifically the performance of UK unit trusts, to determine if the risk-return measures implied are applicable to the London International Stock Exchange ("ISE"). The risk-free rate used was the twelve-month local authority deposit rate, and the market rate was that on the Financial Times 650.19 Regressions were performed based on both the annual re turns and the excess returns. Both methods gave similar results for the estimated β's (ρ2 = 0.9967). Three performance measures were calculated from the data on absolute returns, respectively Jensen's, Treynor's and Sharpe's. These were reviewed in Chapter 4.

All three measures ranked the funds similarly, notwithstanding that the Sharpe measure is the only one which takes into account the total risk of an investment. As in Jensen [1968] they examined whether any significant management effect (Jensen's α) could be found. Ward & Saunders found that no unit trust achieved a superior rate-of-return, but that 21% performed worse than expected by the CAPM, as opposed to the 12% of unit trust under-performers in Jensen. Their proposed hypotheses to explain these results were that:

  • the CAPM does not adequately explain the structure of the UK securities market; and/or
  • the index used to represent the market portfolio was incorrect; and/or
  • management and trading expenses are higher for British unit trusts than their American equivalent.

Although the funds' βs did not fully describe their excess returns, Ward & Saunders point out that there was an active readjustment in the valuation of unit trusts within the time period which could account for that. The index used was the widest definition available at the time, thus an alternative could not be proposed. That management and trading expenses are higher for British than American unit trusts was supported by previous studies.

Ward & Saunders defended the usefulness of the model by grouping unit trusts into five categories - income, international, balanced, specialist and capital. They then calculated average estimated βs, returns and the variance of each category, finding that the expectations by group conformed with the observed βs. For example, capital trusts had the highest average β and variance. They concluded that investors should use the estimated βs as a measure of risk on which to base portfolio strategy. Although they mentioned the zero-β model, they did not explicitly consider it in their study.

In the United States, Brigham & Crum [1977] questioned the enormous increase in the use of the CAPM in the mid-1970's. Their particular concern was its use in public utility rate cases. They cited eight State and several Federal commissions which had considered arguments based on the model. The problems, highlighted in previous studies, were fairly broad:

  • unrealistic assumptions;
  • uncertainty as to the appropriate risk-free rate;
  • inability to measure market risk premia; and
  • non-stationarity of βs and uncertainty as to the predictability of future βs.

The problems associated with the estimation of the relevant variables, especially in periods of uncertainty, was demonstrated via simulation. Using a simulated company with a true original β of 0.75 and subjecting it to an increase in risk and cost of capital, they found that the calculations of β and the true cost of capital were underestimated by most practitioners. They demonstrated the anomalous effect of this in the market with five US businesses (two real estate investment trusts and three others) which all failed, the calculated βs of which declined as they neared bankruptcy. They related these results to the circumstances surrounding the US utility companies in 1977. Although undergoing rising costs and consequently risk, the calculated βs of these utility companies remained unchanged. They concluded that although the CAPM is elegant, it is still the subject of a large amount of research and should be used with caution.


19Now called the FT-SE 650.