Macroeconomic models used for understanding and forecasting GDP, employment, investment, and consumption have a lot of parameters that determine the preferences of consumers and firms, the technological constraints, labor market and investment frictions, and how prices react to monetary policy.
These parameters need to be estimated from past data. The most advanced statistical methodology for this is Bayesian estimation, but with current tools it is only feasible in practice for complex models when a linear approximation of a model is used. While for some models this provides satisfactory results, for other models, such as monetary policy models which take into account the fact that nominal interest rates cannot be negative, or models of uncertainty shocks, nonlinearities are important and need to be accounted for.
This project aims to develop a practical methodology for estimation of mid-size macroeconomics models, such as those used by central banks and forecasting institutions, in their full nonlinear form, using a global approximation combined with recent methodological advances in computational statistics. The resulting software suite, documentation, and examples will be made available under a free software license.