Estimation in the Linear Decision Model

Using statistical decision theory with reference to a linear decision model with a quadratic welfare function in the endogenous variables, it is shown that (1) the loss function is different than the usual loss functions implied in prediction models; (2) under the Bayesian assumption that a prior distribution of the unknown parameters exists and under usual data assumptions, the minimum-risk decision implies a certain class of "optimal" estimates of the parameters, which are different from the usual existence; (3) the optimal estimates require some knowledge on the part of the estimating statistician of the decision-maker's welfare function.