Causal Inference

This course covers the following approaches to causal inference:
• the microeconometric approach,
• the macroeconometric approach, and
• the graph-theoretic approach.
The former two of these three categories reflect the type of empirical application in which
causal inference tends to be applied in practice. For instance, in microeconometrics,
causality is frequently looked at in regression discontinuity designs or via difference-indifference
specifications while, in macro- or financial econometrics, the concept of causality
is generally interwoven with those of exogeneity, structural breaks and structural vector
autoregressions. As opposed to that, the graph-theoretic idea of causality originates in
philosophy and computer science and implements a very different structural approach to
causal inference; it has recently gained popularity in the social sciences.
The material will be presented in the lectures in a both qualitative and quantitative
manner. Computer simulations, empirical illustrations as well as exercises will supplement
informal discussions.

By the end of the course participants will have gained a sound understanding of how causal
inference can be conducted in modern statistics and econometrics. They will be able to
follow the state-of-the-art literature and apply the techniques to empirical econometric
analyses of their own.

Each participant is requested to work on a given topic over the summer months and
present his/her results in a 20-minute talk at the September session. The homework
assignment as well as the presentation will be carried out individually. A 3- to 5-page
report must be submitted one week prior to the presentation and will be disseminated
amongst participants . The topic can be theoretical, empirical or simulation-based. It
is important that both report and presentation reflect the fact that this is a course in
econometrics. That is to say, the emphasis of the treatment should lie on
• a clear exposition of the econometric model,
• a detailled description of the econometric methods, and
• a critical econometric discussion of the results.
Standard econometric notation should be used throughout. Computer code and empirical
data should be made available, e.g. by attaching appropriate files to the pdf document of the report.

Familiarity with basic probability and statistical theory is assumed, as well as of theessentials of regression and time series analysis. The text by Stock &Watson (2015) is usedthroughout this course, and a revision of its chapters 2-7 and 14 is highly recommendedas a preparation. Participants are advised to look for the keyword {\uc1\u8216*}causal effect{\uc1\u8217*} in theindex to get an idea of the contexts in which causality will play a rôle. The companionwebsite of the book contains multiple choice tests and quizzes, convenient for self-study.