Empirical Likelihood as an alternative to GMM estimation in the area of asset pricing
Abstract
This paper proposes and analyses Owens (1998, 1990, 1991) Empirical Likelihood (EL) as an
alternative to the General Method of Moments (GMM) within the Capital Asset Pricing Model
(CAPM). We concentrate on the
finite-sample properties, size and power, of their over identifi
cation
tests. Our simulation evidence shows that there are no clear advantages in terms of size when the
GMMs over identfi
cation tests based on two-step and continuously updated estimators are
compared to that based on the Empirical Likelihood Ratio (ELR) within a Mean-Variance and
Three-Moment setting. The three tests have moderate size distortions. However, our
findings
illustrate that the ELR over identifi
cation statistic is more powerful in detecting deviations from
the null under the alternatives that we analyse.