Computational Methods in Econometrics

Course Objective

In statistics and econometrics, we often base our test statistics and confidence intervals on quantities with unknown distributions. The
typical solution to deal with this is to either derive the distribution through hard analytical work, or derive its limit distribution and use
it as an approximation. In this course, we will study how to perform tests using alternative ways based on computer simulation. The main
examples of these methods are Monte-Carlo testing and bootstrap, which can lead to much more accurate inferences than the traditional methods in certain situations.

The main goal of this course is twofold: (1) to understand the theory behind Monte-Carlo and bootstrap methods and (2) to be able to apply them in practice using your computer.

Course Content

In this course, we discuss numerical and simulation-based estimation methods and their use in econometrics and data science. We start with a small recap of statistics (in particular, estimators, test statistics and their distributions). In the second part, we discuss the assumptions
made for these results and introduce a new simulation-based hypothesis testing method called Monte Carlo testing. In the third part, we move to a more complex setting where less assumptions are made, and we discuss the foundations of bootstrap testing.

We illustrate how these methods are used in practice for a variety of econometric models including nonlinear models, models allowing for heteroskedasticity and autocorrelation.

Additional Information Teaching Methods

Lectures (4 hours per week) to introduce the new methods and concepts; Work groups (2 hours per week) to work on exercises and, later in the course, to work on homework assignments.

Method of Assessment

Written exam (percentage P), Assignments (percentage 100-P), where P is in [50,70].