Alongside description and prediction, causal inference is one of the central aims of quantitative empirical social research. In research practice, often only non-experimental data are available, which makes causal inference difficult due to non-random selection. This workshop will provide a basic introduction to non-experimental methods of causal inference that explicitly address the problem of non-random selection and are increasingly used in social science research. As an alternative to workshops that specialize in one method, this workshop will offer an introductory general overview of different methods of causal inference. Firstly, the counterfactual model and directed acyclic graphs (DAGs) will be introduced as overarching theoretical frameworks and practiced using examples. The implications for regression analysis with regard to the selection of control variables and model building will be explained. Building on this, the methods of regression adjustment, propensity score matching, (coarsened) exact matching, entropy balancing, inverse probability weighting, instrument variable estimators, regression discontinuity design, and difference-of-differences estimators will be presented in an application-oriented introduction. The methods will be practiced with the statistics program Stata using data and simple examples from social sciences research.

 

The course will be taught live (no recordings) via Zoom. The course will bedivided into input and exercise units of equal duration. The input units will be lecture-based, with opportunities for Q&A and discussion. During the exercise units on the counterfactual model and causal graphs, you will be divided into groups to solve small problems together, and the solutions will be jointly discussed. During the hands-on empirical exercises, you will learn how to implement the different methods of causal inference in simple empirical exercises using the statistical program Stata and simple data and examples from social science research. Joint discussions will focus on the proper application of each method and the interpretation of the results. Given the introductory and overview character of the course and the expected heterogeneity of participants, it is not intended that you work on your own research examples. However, you should be able to apply the knowledge gained to their own research after the course.

You will find the course useful if you  are doctoral students and scientists in the field of social sciences who would like to get an application-oriented introductory overview of the foundations and methods of causal inference.

By the end of the course, you will be able to:

 

You will need Stata (version 15 or later) with ados psmatch2, teffects, kmatch, moremata, kdens, and cem installed. If you need Stata, please let us know the latest two weeks in advance of the course start so that we can take care of it.