A Comparison of Dynamic 1-0 Multipliers for Kansas with Parallel Econometric Multipliers

Abstract

This paper tests a non-survey dynamic I-O model against parallel econometric models using reduced form equations. In particular, we estimate regional steady state multipliers from the two methodologies for 13 sectors, and compare them.

The input-output multipliers are derived from a dynamic intersectoral model of the state of Kansas. The model is export­-driven, and endogenizes consumption, regional government, and investment. The model includes expectational variables, capacity constraints and disinvestment constraints; it is solved to yield steady state export multipliers, assuming perfect foresight.

Conceptually similar total income multipliers are estimated econometrically from a model using reduced form equations, SUR, AR(l), and constrained positive multipliers. Some versions of the model use export proxies, which lead to econometric estimates of Kansas regional exports. The estimated multipliers are found to be relatively insensitive to the lag structure.

Multipliers derived from the two approaches are found to be modestly similar. The mean absolute differences are of the order of 35% to 50%. However, the multipliers are significantly correlated across methodologies and have similar group distributions. Chi-square tests cannot reject the hypothesis that all differences can explained by three sources of error: (1) estimation errors in the econometric model; (2) 20-25% random errors in the non-survey model; (3) a downward bias in some of the econometric estimates using proxies.

Suggestions are made for improving the robustness of the econometric model, and also for improving the proxies and the resulting econometric estimates of state exports.