Essays in asset pricing and portfolio decisions

This dissertation addresses possible reasons for rejection of the consumption based capital asset pricing model--CCAPM--(essays I & III) and applies an alternative approach to assess rejections of the capital asset pricing model (essay II).; In essay I, I construct a new consumption and asset returns data set for Japan and Germany, and conduct multi-country comparisons of the CCAPM using the GMM methodology. The similarities in parameter estimates for the discount factor and the coefficient of relative risk-aversion are strikingly similar across Germany, Japan and the U.S. The model is rejected based on (i) multiple assets and (ii) short term bonds across the three countries. I fail to reject the model based on stock index returns only. I report the existence of an unexplained equity premium; i.e., an equity-premium puzzle in all the three countries.; Essay 2 applies the Bayesian approach to investigate efficiency of a given portfolio. An implication of the CAPM is in terms of portfolio efficiency; i.e., that the intercept terms in particular regressions are zero. I exploit the power of Monte Carlo integration and conduct inference regarding the intercept terms ({dollar}alpha{dollar}'s) which measure inefficiency. A Highest posterior density (HPD) region is constructed for the intercept terms; it supports the inference of inefficiency of the NYSE Value Weighted portfolio. Posterior inference is conducted using two other measures of efficiency-{dollar}lambda{dollar} and {dollar}rho{dollar}, leading to the same inference. A utility-based measure of divergence from efficiency also supports inefficiency of the market portfolio.; Essay III investigates the role of the temporal aggregation bias in the CCAPM using the technique of GMM, within the context of an AR(2) data generation process. The bias in the estimates of the discount factor ({dollar}beta{dollar}) and the CRRA ({dollar}gamma{dollar}) becomes higher as the level of aggregation increases. There are over-rejections of the model at lower aggregation levels. There appears to be a need for size correction of the test. The bias and precision effects cannot be explained away by sampling variation.