This is an introductory course on "Global solution methods" to solve dynamic economic models, held at the Cowles Foundation, Department of Economics at Yale University in January/February 2019.
Prerequisites:
Undergraduate calculus and statistics. Experience in any programming language such as Python is a plus. Participants are encouraged to bring their research projects to be discussed.
Modern dynamic economic models such as DSGE models are extremely rich to capture all the effects of interest: they contain large stochastic shocks that lead to highly non-linear policies, or many agents that lead to a high-dimensional state space, to name a few. To this end, standard solution techniques such as log-linearization often fail to deliver reliable results across the domain of interest. The latter method may be useful to describe an economy that operates in normal times, but fails in the presence of strong nonlinearities such as occasionally binding constraints, among other types of salient features of the economic reality, that the policymaker considers relevant for taking policy actions and therefore the modelers would like to capture appropriately in their models.
This course is intended to provide students in economics with a self-contained introduction to the extensive and broad topic of "global solution methods."
In particular, we consider how dynamic (stochastic) economic models with substantial but finite heterogeneity can be solved numerically either by dynamic programming---that is, value function iteration or by iterating on the first order conditions-that is, time iteration.
- Simon Scheidegger (HEC, University of Lausanne)
| Date | Time | Main Topics |
|---|---|---|
| 01.14.2018 | 09:00 - 10:20 am | Introduction to Global solution methods and a crash course to Python |
| 01.16.2018 | 09:00 - 10:20 am | Introduction to Sparse Grids and Adaptive Sparse Grids |
| 01.18.2018 | 09:00 - 10:20 am | Dynamic Programming and Time Iteration with Sparse Grids |
| 01.23.2018 | 09:00 - 10:20 am | Introduction to Machine Learning (supervised and unsupervised machine learning), and basics on Gaussian Process Regression |
| 01.28.2018 | 09:00 - 10:20 am | Gaussian Process Regression (part II), Gaussian Mixture Models. |
| 01.30.2018 | 09:00 - 10:20 am | Dimension-reduction with the active subspace method. Solving dynamic models on high-dimensional, (irregularly-shaped) state spaces |
The lectures will take place at ROOM 106 at 28 Hillhouse