Copenhagen, Denmark

Bayesian Econometrics

This course provides an introduction to modern Bayesian methods in econometrics.

The first part of the course presents the fundamentals of the Bayesian approach, from the derivation of Bayes' theorem to its practical application to econometric models. It introduces basic concepts such as prior, posterior and predictive distributions, before presenting essential tools based on simulation methods: Markov chain Monte Carlo methods, including the Gibbs sampler and the Metropolis-Hastings algorithm. Common econometric models students are already familiar with will be revisited from a Bayesian perspective (e.g., linear regression model, binary/discrete variable models).

The second part of the course dives into more specific and technical topics. It presents some selected econometric models where Bayesian methods are particularly useful, such as latent variable models and random coefficient models (relying on data augmentation methods). It also discusses some problems that can affect standard simulation methods (e.g., slow convergence, bad mixing), and explains how these problems can be successfully overcome using recent developments in statistics.

Bayesian methods can be applied to any field of economics. The examples and exercises offered during the summer school will be drawn from various topics, including micro- and macro-econometrics, and finance.

The main goal of this course is to provide students with practical skills to apply Bayesian methods to a specific problem. Therefore, it should be of particular interest to students planning on writing a Master's thesis or preparing for a PhD programme.

Target group

Master

Course aim

Knowledge:

- Understand Bayes' theorem and how it can be applied in econometrics.

- Have a grasp of simulation methods, understand their principle and how they can be used to make inference.



Skills:

- Demonstrate an ability to select the most appropriate method for a given estimation problem.

- Be able to implement Markov chain Monte Carlo methods such as the Gibbs sampler and the Metropolis-Hastings algorithm, both theoretically (analytical derivation of the algorithm) and practically (programming).

- Demonstrate technical skills in writing code to implement Bayesian methods. Be able to develop a computer program with the R programming language or use publicly available packages to carry out their own empirical analysis.



Competencies:

- Be able to conduct a full Bayesian analysis: (1) formulate an economic model, (2) organize prior knowledge and ”beliefs” about the model (prior), (3) use relevant data to express the observed information in the model (likelihood), (4) use Bayes' theorem to update beliefs (posterior), (5) derive an appropriate algorithm to compute the posterior distribution, (6) write code to implement the algorithm, (7) interpret the results and criticize the model.

Fee info

DKK 2625: EU/EEA citizens

You do not pay tution fee if you are enrolled in a Danish University and have a preapproval for credit transfer.
EUR 1275: Non-EU/EEA citizens

You do not pay tution fee if you are enrolled in a Danish University and have a preapproval for credit transfer.