Modern Causal Analysis in the Social Sciences (MAD)
Course: | Time: Tue 10-12 | Room: DIGITAL | Term: Summer 2021
What is the effect of education on income? Has a job creation scheme created jobs? Does unemployment increase the probability to abstain from voting?
These questions are causal questions: does a change in X cause a change in Y?
To identify causation, experiments with randomized control and treatment groups are regarded as the gold standard. Oftentimes in social science, only observational data is available. This data poses obstacles to causal analysis and as one learns in statistics, correlation does not imply causation. But what does imply causation? in this case? The seminar will cover concepts and methods of modern causal analysis that are trying to give an answer to that question. In particular, the following topics are discussed:
1. the concept of causality based on counterfactuals and directed acyclic graphs (DAGs)
2. two methods based on selection on observables: regression adjustment and propensity score matching
3. two methods based on selection on unobservables: fixed effects and regression discontinuity
The individual topics are presented in an accessible way not relying on mathematical knowledge. Presented methods are applied using real-world examples and applications are carried out in R.
Participants will be able to a) think causally and create DAGs, b) critically discuss methods of causal analysis c) and apply causal analysis to answer own research questions.