ABSTRACT
Despite the productive efforts of many, questions remain concerning the insecure assumptions underlying the sampling theory, likelihood, and asymptotic approaches and the usefulness of traditional multiple-equation estimation and inference procedures in helping us find order when using the partial-incomplete underlying economic data that is normally found in practice. Against this backdrop, we propose a new method of estimation in multiple-equation statistical models that is widely applicable because it does not require the specification of a parametric family for the likelihood function. The estimation rule is robust with respect to likelihood, is flexi-
ble with respect to the dynamic, stochastic, and feedback nature of economic data as well as to the introduction of prior information, and is computationally simple. Using linear and quadratic risk measures, we compare the finite-sample performance of this method to other widely used traditional estimation rules.
