Journal of Economic Perspectives—Volume 10, Number 1—Winter 1996—Pages 69-85
The Computational Experiment: An Econometric Tool
Finn E. Kydland and Edward C. Prescott
In a computational experiment, the researcher starts by posing a well-defined quantitative question. Then the researcher uses both theory and measurement to construct a model economy that is a computer representation of a national economy. A model economy consists of households, firms and often a government. The people in the model economy make economic decisions that correspond to those of their counterparts in the real world. Households, for example, make consumption and savings decisions, and they decide how much to work in the market. The researcher then calibrates the model economy so that it mimics the world along a carefully specified set of dimensions. Finally, the computer is used to run experiments that answer the question. 1
Such experiments have become invaluable tools in quantitative aggregate theory. 2 They are being used, for example, to estimate the quantitative effects of trade liberalization policies, measure the welfare consequences of changes in the tax system and quantify the magnitude and nature of business cycle fluctuations induced by different types of shocks. In this paper, we review the use of the computational experiment in economics.
1 Lucas (1980), in his paper on methods and problems in business cycle theory, explains the need for computational experiments in business cycle research.
2 Shoven and Whalley (1972) were the first to use what we call the computational experiment in economics. The model economies that they used in their experiments are static and have many industrial sectors.
■ Finn E. Kydland is Professor of Economics, Carnegie-Mellon University, Pittsburgh, Pennsylvania, and Research Associate, Federal Reserve Bank of Cleveland, Cleveland, Ohio. Edward C. Prescott is Professor of Economics, University of Minnesota, and Research Associate, Federal Reserve Bank of Minneapolis, both in Minneapolis, Minnesota.