Regional policy simulation and forecast model for the state of Oklahoma: A maximum entropy approach

Abstract

Scope and Method of Study: A variant of Bayesian estimation called entropy is employed to dynamically calibrate a regional general equilibrium-type model to maximize the fit to historical observations for the state of Oklahoma. It is postulated that such dynamically calibrated policy simulation model will not only be useful in policy analysis, but also for forecasting applications granting policy simulations with a time-path response to facilitate timing. Generalized Cross-Entropy estimation is employed to dynamically calibrate the system as a whole advantaging from the full set of general equilibrium constraints. Such estimation transforms the traditional econometric estimation to a non-linear math programming problem in non-linear constraints.

Findings and Conclusions: Entropic estimation allows complete systems estimation consistent with the full set of general equilibrium constraints transcending criticisms of single equation estimation. Projections of the model create forecasts that compare favorably with more traditional econometric methods for forecasting. The addition of complete market structure in estimation extends the forecasting application to policy analysis allowing for a large breadth of policy applications that illustrate not only the overall impact implications but also the timing of those implications.