Model simulation, validation and comparison

This work package will be essential in bringing together the set of policy-focused macro-financial models to be developed by the consortium, investigating their comparative properties and setting the ground for policy design and evaluation. The steps that this involves will be enormously facilitated by the Macroeconomic Model Database, to which the models will be added. Various novel methodological challenges and substantive issues to be addressed will arise. Most of the macroeconomic models to be developed by the consortium cannot be solved and simulated using the workhorse approach in the DSGE literature, namely log-linearizing the model’s equilibrium conditions and then employing linear rational expectations model methodology. This is as higher-order moments tend to take on important roles in macroeconomic models featuring financial markets with both normal and crisis regimes, and as these models involve non-smooth nonlinearities, such as constraints that for some of the households and firms may only bind temporarily. Models of such a type require working with global approximation methods. Such methods tend to be plagued by curse of dimensionality-type problems, and various approaches to deal with these will be considered. Further methodological challenges will arise when confronting the consortium’s models with the data. As standard linear Kalman filtering techniques are not applicable anymore, it becomes necessary to employ nonlinear filtering techniques. Turning to model validation and comparison, this, too, involves new methodological challenges for highly nonlinear models, and these will be addressed on the basis of newly proposed global sensitivity analysis procedures. In terms of substantive issues, given the novelty of macroeconomic models with realistic financial sectors, little is known about their relative performance and ability to capture macroeconomic dynamics both within and outside of financial crises, and thus a thorough comparison of these models and their mechanisms will be imperative.

Key Objectives

(O.8.1) Develop effective algorithms for the solution of macroeconomic models featuring nonsmooth nonlinearities such as occasionally binding constraints or multiple regimes.

(O.8.2) Improve on existing methods for the estimation of DSGE models and extend these to allow reliable estimation involving nonlinear filtering techniques.

(O.8.3) Take advantage of and to fine-tune global sensitivity analysis procedures so as to allow for informative model validation and comparison.

(O.8.4) Understand which features of financial markets are empirically essential for policyfocused macroeconomic modelling.

Main participants