|
Much of the recent work on equilibrium asset pricing has sought to link both the cross-sectional and the time-series pattern of asset risk premia with the pattern of covariances between realised asset returns and consumption growth. Assets that provide a high (low) return at times when consumption is high (low) carry a high (low) risk premium because they provide a high (low) return at the time it is not needed by consumers. The risk premium for each asset is thus proportional to the covariance of the asset return with consumption. Empirical tests of a simple model of consumption and asset returns, the so-called Consumption Capital Asset Pricing Model (CCAPM), have led to strong statistical and economic rejections across a wide range of assets, and have performed poorly when used to explain the cross-sectional variation in expected stock returns. Investigators have pointed out a number of shortcomings in the original tests of the CCAPM, one of which is that the tests use consumption data, which are plagued by measurement error and time-aggregation bias. In Discussion Paper No. 1262, Research Fellow Gikas Hardouvelis, Dongcheol Kim and Thierry Wizman evaluate the ability of asset pricing models that do not use consumption data, and models that use consumption data as a proxy for true consumption, to explain the time-series and cross-sectional variation of expected returns on portfolios of stocks. Although some parameter restrictions are rejected by models that do not use consumption data, the paper finds that they provide economically meaningful estimates of the representative agent's preference parameters and fit the data slightly better than models which use consumption data to proxy true consumption. Moreover, models without consumption data are not rejected when they are augmented to account for the `size' effect. Asset Pricing Models with and without Consumption:
An Empirical Evaluation |
|