ABSTRACT
Assuming that the first term dominates, the constraint item (CI) should be chosen such that the first term is minimized and this means that variable whose variance is largest relative to the variance of the sum of the remaining possible CI choices. Since no equation contains every explanatory variable, any choice of the CI will result in an M-1 equation system in which the coefficients of certain explanatory variables are known to sum to zero over the M-1 equations. In the context of the preceding analysis, it would seem wise to restrict the possible choices of the CI to those equations containing the most explanatory variables. The parameter estimates and forecasts of an individual equation will vary according to whether or not that equation is chosen as the CI. In choosing the CI, the search should be for an equation with a large error variance relative to other equations, and a minimum of excluded explanatory variables.
