38th Congress of the European Regional Science Association: "Europe Quo Vadis? - Regional Questions at the Turn of the Century", 28 August - 1 September 1998, Vienna, Austria
The residual in an estimated equation is composed by measuring errors and omitted variables. If an omitted variable is autocorrelated it can to some degree be included in the model by the lagged autocorrelated residual. It can be shown that in a variable coefficient built after the principle of the expansion method a variable autocorrelation coefficient is likely to appear. The purpose of this article is first to discuss how many unknown and excluded variables can be traced in the residual. Second to compare the efficiency of the method used on ordinary equations with the efficiency of the method used on equations created by the expansion method. Finally the method is tried on data for an urban model.