Financial and International Markets2007Assignment 1: Empirical Evidence on Capital Asset Pricing ModelThe Capital Asset Pricing Model (CAPM) developed independently by Sharpe (1964), Lintner (1965) and Mossin (1966) assumed that asset returns are normally distributed and investors have mean preferences. Hence, it is possible to estimate expected returns, and thus cash flows, by adding an asset to a diversified portfolio. The asset in question is then sensitive to systemic risk (β, beta) as well as the expected return of the market and that of a risk-free asset. This is expressed as [E(Ri) – Rf] / β = E(Rm) – Rfwhere E(Ri) is the expected risk of the asset, Rf is the risk-free asset, β is the sensitivity of the asset to the market, E(Rm) is the expected return from the market, often assumed to be the return from the market index and [E(Ri) – Rf], the difference between the expected return from the asset and that the risk-free return is the market premium.
Thus, risk from adding an asset comprises market risk, or systemic risk, or specific risk, which can be minimized by adding a large number of assets. Thus, theoretically, expected excess return may be estimated by regressing the following equation: [E(Ri) – Rf]t = αit + β [E(Rm) – Rf] - eitThe average market return is usually assumed to be the historical return from the market index.
Hence, if the coefficient to the historical risk-free asset is estimated to be zero, that is investors can borrow at risk-free rates, then the expected excess return would be estimated to be sensitive to the market risk. However, empirical tests have found that the coefficient to the risk-free asset is not zero.
Besides, the model is not directly testable since historical returns may not be the same as future returns, the real structure of the market portfolio may not be known and the market index may not also be an appropriate estimation of the market portfolio (Trandafi). The CAPM model essentially rests on the assumption of normal distribution of asset returns, so that the variance of the returns appropriately measures systemic risk. However, empirical tests have shown that returns may not be normally distributed. Further, investors are assumed to have rational expectations, capital markets perfect and that there are no asset arbitrage possibilities.
These assumptions are also suspect and empirical studies have shown that the model may need to be modified substantially when each of these assumptions fail. Fama and French (1992) found that non-market risks and book-to-market value ratios are statistically significant factors that affect asset returns. Following Banz’s (1981) result of firm size on expected returns, Fama and French (1992) found that firm size may be more significant than CAPM (cited in Shackleton & Fergal, 2004). Besides, in contrast to normal distribution of asset returns as assumed by the CAPM model, empirical evidence suggests that probability of extreme returns is higher than normal returns.
Thus, the skewness and kurtosis of the distribution of asset returns become significant (Kraus & Litzenberger, 1976 Fang & Lai, 1997, Harvey & Siddiqui, 2000 cited in Shackleton & Fergal, 2004). However, Shackleton and Fergal (2004) found from data of London International Financial Futures and Options Exchange (LIFFE) that systemic variance of returns are significant determinants of a cross section of options contracts while the effect of skewness is less significant.