ABSTRACT
There are two basic questions to be answered when constraint item (CI) estimation is being considered for a particular M equation modelling problem. First, how do the statistical properties of CI estimators compare to alternative estimators involving the full set of equations, and second, if CI estimation is to be used, what factors should determine the choice of the dependent variable to be left as the CI. The choice of the CI affects the ease of computation because it determines the coefficient constraints of the M-1 equation subset. The coefficient and error term constraints which were derived in the case of non-stochastic explanatory variables apply to identified simultaneous equations models. One factor in choosing the CI is the desire to minimize the number of ignored constraints and this can be accomplished by choosing the CI as that dependent variable whose equation contains the maximum number of explanatory variables.
