The purpose of this research is to enable econometric and economic models to better represent economic behavior and, thus, provide better answers to questions confronting decision makers in the private and public sectors. The project will involve the reformulation of problems in statistical inference and optimization, inspired and made possible by recent advances in widely available computational capacity. Research methods iterate between the problems that define and motivate economic analysis and econometrics, computational experiments, and the development of new analytical tools. In economic modeling, the emphasis is on numerical mapping from decision rules to objective functions. Much of this development is in the context of particular problems, including seasonal adjustment, nonstructural representations of macroeconomic time series, cointegration models, the solution of nonlinear or rational expectations models that are generic and algorithmic, and can, therefore, be used systematically in studying a wide range of problems. This and related research by others over the next decade will enable economists to address policy problems much more directly, with emphasis shifting from the constraints of analytical approaches to a greatly enriched modeling of behavior.
