Testing asset pricing models using market expectations

Abstract

We investigate the use of market-based expectations to test the CAPM and the conditional CAPM using a generalised method of moments framework. This method is valid under much weaker distributional assumptions and provides the procedure with robustness that commonly employed tests lack. Expected returns are derived from projected price levels of individual securities that are supplied in the form of twelve{u00AD}month consensus (median) target price forecasts. The annual forecasts, updated each month, are combined with dividend expectations to calculate the necessary time series of continuous expected returns. As such, we are able to avoid the use of instrumental variable models that, we argue, are likely to suffer from overfitting data concerns. In fact, we find that expected returns estimated from analyst data, while certainly not perfect, provide a better fit in comparison to the existing instrumental variable models. In considering the testable implication of the model via a vector of orthogonality conditions, we find that using market-based expectations to test the CAPM directly leads to a rejection. Overall, the CAPM tends to underestimate returns, producing pricing errors that are large, positive and statistically significant. Our results link in with the existing asset pricing literature that also attempts to apply forward-looking data derived from analyst forecasts. The conditional CAPM, a model that benefits from time-varying parameters that are updateable in accordance with changes to the information set, is also rejected. Market based expectations are used to parameterise the marginal rate of substitution. While the results of our tests of the conditional CAPM indicate that the model is able to perform better than those reported in previous studies, it continues to consistently underestimate returns in contravention to the null hypothesis. This indicates that the market, as the sole risk factor of the model, is not enough to explain the variation of returns across assets. While beta-risk may be priced, the CAPM may not account for all priced risk factors.