Causal Machine Learning
In the past 60 years econometrics provided us with many tools to uncover lots of different types of correlations. The technical level of this literature is impressive (see the PEF course Advanced Microeconometrics). However, at the end of day, correlations are less interesting if they do not have a causal implication. For example, the fact that smokers are more likely to die earlier than other people does not tell us much about the effect of smoking. For example, it might just be that smokers are the type of people who face more health and crime risks for quite different (social or genetic) reasons. The same problem occurs with almost any correlation of economic or financial variables. The interesting question is always whether these correlations are spurious, or whether they do tell us something about the underlying causal link of the different variables involved?
In this course we review and organize the rapidly developing literature on causal analysis in economics and econometrics and consider the conditions and methods required for drawing causal inferences from the data. Empirical applications play an important role in this course.
Active participation of PhD students participating in this course is expected. During the second part of the course, participants will conduct their own empirical study and present their results.
General structure and rules
Active participation of the students in this course is the key to its success. Students are expected to do the following:
- Read the papers shown as ‘compulsary reading’ in the reading list BEFORE the lecture concerned with the topic.
- Each morning students will present a paper (15‑30 minutes each; depending on the number of participants) and there will be some general discussion about these papers. Students not presenting will be expected at least to sketch the papers to be able to participate in the discussion.
- Small groups of students (group size depends on number of participants) will conduct an independent empirical study (using Software of their own choice; GAUSS or STATA is recommended). In the empirical project students will show that they understood the basic concepts and are able to apply them to a ‘real life’ situation.
- Written Exam about 4 weeks after the last lecture (2 hours) (40%).
- Students’ active participation in general discussions during lectures and presentations (20%).
- Presentation of papers (20%).
- Empirical project (based on two presentations; 20%).
As defined for the econometrics specialisation of PEF.
To be published shortly before the lecture
Empirical work, literature, contents of lecture
Examination relevant literature
To be defined during the lecture