Comparison of three methods for developing models of resource demand

Abstract

This paper compares a new rather general modeling procedure with "static" and "dynamic" econometric procedures. The resulting models can all be used to make projections of demand for natural resources.

If such models are combined with supply models, one can estimate the Long-term impact of resource policy. The particular resource discussed in this paper is natural gas in the Pacific Gas and Electric (PG&E) service area of California (large portions of the north and central sections of the state).

Of the three modeling procedures discussed, the most general is emphasized in this paper. This method, described in Section 3, is related to the modeling procedure proposed by Forrester ^10,11 combined with parameter identification.^2,19,30 A precise state ment and critique of the method is provided by Mitchiner et al.¹⁹ The other two modeling proce dures, chosen for purposes of comparison, are stan dard econometric methods. The first of these is a modification of "static" (algebraic) linear regres sion modeling. The second is a "dynamic" linear regression procedure wherein a lagged value of demand is used as one of the input variables.^1,3,5

The proposed general methodology was not inferior to the two "standard" procedures in any case investigated and was clearly superior in the analysis of firm in dustrial demand for natural gas. This result was expected because the standard procedures are special cases of the proposed methodology.