9409040 Tauchen It is quite common across many research areas in empirical economics for there to be an accumulated body of knowledge regarding the statistical characteristics of the relevant data. This knowledge is acquired from the years of experience of many investigators using statistical models to describe the data. The models are termed statistical because, with rare exceptions, the parameters have no direct behavioral interpretation. A structure model, such as an equilibrium asset pricing model or an equilibrium business cycle model, stands in sharp contract to a statistical model. A structural model typically provides a complete description of the underlying dynamics, and many parameters have behavioral interpretations. Estimating a modern stochastic structural model presents formidable computational and statistical challenges. A reasonable principle for structural estimation is to bring to bear on the problem the accumulated stock of knowledge regarding the reduced form statistical models that have been found by others to work well. The proposed project undertakes extensions and refinements of new estimators for dynamic models based on this principle. The proposed work is in two parts: a methodological section and an empirical section. The methodological work entails development of reliable methods for statistical inference, comparative work with other estimators, and visualization work to learn more about the dynamics embedded in the model and the characteristics of the objective function of estimation. The empirical work applies the new estimators to address issues in asset pricing, financial volatility, and empirical auction research. The main research tools are nonlinear estimation, Monte Carlo simulation, and the AVS visualization package. The results of the project will be of interest to empirical economists and statisticians, and in particular to researchers doing simulation work or to researchers becoming interested in interactive v isualization of structural models.