INFORMATION THEORETIC ALTERNATIVES TO TRADITIONAL SIMULTANEOUS EQUATIONS ESTIMATORS IN THE PRESENCE OF HETEROSKEDASTICITY

Finite sampling properties of information theoretic estimators of the simultaneous equations model, including maximum empirical likelihood, maximum empirical exponential likelihood, and maximum log Euclidean likelihood, are examined in the presence of selected forms of heteroskedasticity. Extensive Monte Carlo experiments are used to compare finite sample performance of Wald, Likelihood ratio, and Lagrangian multiplier tests constructed from information theoretic estimators to those from traditional generalized method of moments.