I. Introduction

Computational economics is developing as a central part of economic research. Economists have used computers for many years to estimate and simulate econometric models, but the use of computers is now spreading rapidly into all areas of economics and is fundamentally reshaping the way we do research. In the past, data was used by a single researcher in fitting collections of equations. Now large and complex databases are organized on servers and accessed over networks by geographically dispersed users. Environmental issues were analyzed with theoretical methods or with small models of a few equations. Now general equilibrium models with hundreds of equations are used to analyze the effects of carbon emission taxes on global warming.

Computational economics changes the way we think about problems. In economic theory, and to a lesser extent in econometrics, traditional analytical research has focused on formulating a problem in the context of an analytically tractable mathematical model, attempting to derive general 3existence2 or 3characterization2 theorems, or to derive more specialized 3closed-form solutions2 in particular cases. This nearly uniformly entails serious compromises with reality and removes many interesting problems from consideration: e.g., nonlinear constraints and functions of interest in econometrics, and virtually any substantive extension of a dynamic programming problem beyond certainty equivalence in economic theory.

Computational methods are removing these constraints. For example, in econometrics, innovations in statistical multiple integration are making many traditional and very hard analytical problems in the distribution of multiple-step-ahead forecasts obsolete, because we can compute arbitrary good approximations to exact distributions that incorporate both parametric and model uncertainty. This has immediate implications for short- and intermediate-run macroeconomic forecasting, and is one fact underlying the widespread use of VAR forecasting models. The ability to handle nonlinearities and corner solutions in dynamic programming can be applied to the failure of financial institutions. In a closely related development, the numerical solution of stopping value problems has immediate application to the design of cost-effective policies to control recruitment, retention and retirement in the armed forces.

These changes in the way we organize our research have prompted consideration of an Initiative in Computational Economics at the National Science Foundation. As a part of that initiative this report has been prepared to focus on three sets of issues:

1. Contributions

The contributions that computational economics can make to complement other research methods in analyzing major economic problems.

2. Opportunities

Areas of economics which offer opportunities where computational economics can make important contributions.

3. Infrastructure

The infrastructure developments in hardware and software needed to facilitate research in this field.

[ Next | Previous | Table of Contents | Title Page ]