The behaviour of U.K. stock prices and returns

Abstract

In this thesis I combine VAR forecasting methods with the Campbell-Shiller log-linear approximation to the present-value formula for stock prices. Four aspects of UK stock market behaviour are studied. The first analysis involves decomposing the variance of the unexpected stock return into components due to news about dividends, news about future returns, and the covariance between the two. I find that changing expectations about future returns accounts for around four times as much of the variance of unexpected returns as news about dividends, with a negligible covariance term. My second study is a detailed analysis of the links between macroeconomic risks and required stock returns. Using 27 industry-based stock portfolios, I attempt to determine the effect that a number of macroeconomic and financial factors have on expectations of dividends, real interest rates and future required returns. The results go some way to explaining why some risk factors appear not to be significantly priced in financial markets, whilst others (particularly inflation) appear to induce counter-theoretical reactions in stock prices. Given an empirical proxy for equilibrium returns, the present-value model implies a set of non-linear restrictions on the parameters of a VAR, the latter being taken as a model of investors' expectations formation. In my third analysis, I test various models of equilibrium returns using aggregate UK data, and find some support for market efficiency. In particular, in accordance with the intertemporal CAPM, I find that the well-known ability of the dividend yield to forecast stock returns can be traced to the fact that the dividend yield Granger-causes the market return variance. In the final section I test two propositions: whether rejections of the CAPM at the aggregate level can be traced to rejections in specific sub-sectors of the market; and whether investors are more skilled at eliminating mis-pricing within market sub-sectors than in the market as a whole. I find mixed support for the CAPM at the disaggregated level. Furthermore, eliminating the covariance terms from the model for sector returns has little effect on the results, providing some support for the market segmentation hypothesis.