Abstract

This dissertation investigates the impact of information on the Capital Asset Pricing Model (CAPM). Under the traditional paradigm of rational expectations information is complete, known to agents and consequently fully reflected in asset prices. This conclusion has been challenged by a large number of empirical studies showing that there is predictability in asset prices both at the cross-section and in the time series dimensions. In this dissertation, we develop a framework that reconciles these apparent pricing anomalies with the traditional one parameter CAPM.

Chapter 2 explores the effects of the introduction of parameter uncertainty in asset pricing. The rational expectations assumption is relaxed and it is assumed that investors do not have perfect information on assets. Specifically, investors do not know a parameter from the dividend distribution but are able to formulate prior beliefs. They then use a Bayesian updating rule to revise their beliefs as new information becomes available. These investors are fully rational and it is not required, as it is the case in behavioral finance, to assume that they suffer from some kind of cognitive deficiency. The deviations observed between prices set by these investors and those predicted by a true rational expectations model are solely due to the lack of information about the true value of the parameter. Furthermore, asset prices under parameter uncertainty converge towards the rational expectations prices as information about the asset increases. Empirical tests on prices resulting from this learning process will wrongly reject market efficiency. Several cross sectional variations in stock prices, which the data suggests to be at odds with the CAPM, are shown to be explainable by parameter uncertainty. The momentum, the size and the book to market effects all have empirical patterns that are explainable by parameter uncertainty.

Chapter 3 adds a new "age" factor to the standard Fama-French multifactor asset pricing model. Time series and cross sectional tests show the age factor to be a significant pricing variable which in some instances even eclipses the traditional size and book to market factors. Tests on age/size sorted portfolios, rather than the more common size/book-to-market portfolios, reveal an interesting reversal in the size factor. I argue that age captures a new source of risk not included in the standard beta, namely, a Knightian parameter ambiguity risk. This risk stems from the limited information available for younger firms which makes the distribution of their returns difficult to estimate from a small sample of observations.

In chapter 4 we show that information based factors are more successful at pricing industry portfolios than any of the common factors found in the literature. The percentage of days during which an asset did not trade in any given year is used as a proxy for information diffusion on the market. This measure is then used to construct a pricing factor which is shown to be significant in time series regressions. When pricing 30 industry portfolios the no-trade factor outperforms any other multifactor pricing model available in the literature and succeeds at fitting a pricing equation for all but 6 of the portfolios.