Course Instructor: Nan Zhang (Office hours by appointment)
Course Website: https://nanzhangresearch.github.io/QMPS_HWS24
Date and Time: Fridays, 10:15-11:45 in 358 Pool-Raum (L 7, 3-5)
This course will teach students how to analyze questions from the field of political sociology through the application of causal inference methods. We will begin by asking what it means for X to cause Y using the framework of potential outcomes. We will then look at the most popular research designs in causal analysis including experiments, regression discontinuity designs, difference-in-differences and instrumental variables.
Students will learn to apply these methods to real data in Stata.
The overarching goal is to provide students with the foundation to perform their own analyses (e.g. for their BA theses) by transferring the acquired skills to their research interests.
Class-time will consist of a mix of lecture, “live-coding” exercises, reading discussion, and student presentations of data analysis assignments.
Weekly problem sets will provide students the opportunity to practice the methods they learn in class by analyzing real research data. Students are encouraged to work in groups.
There are 6 problem sets in total. Students must complete all 6 to pass the course. Problem sets must be submitted to me via email by noon (12:00) on the following Wednesdays:
I will randomly select students to present their problem set answers during our class sessions. Problem set answers will be posted in the following week.
Students are also required to read applied research articles that implement the methods we learn about, and submit discussion questions in advance of class. It is not always necessary to understand every detail of each article; focus on how and why they apply the methods we covered, and whether or not they do a good job.
There are 3 required readings in total. Students must therefore submit 3 sets of discussion questions. To be clear: your submissions should consist of questions about the reading. Please focus your questions on how the methods we have learned in class are applied (e.g. data analysis, the assumptions involved, etc.). There is no need to send me a summary of the readings. I will assume that you have read them.
Your discussion questions should be submitted to me via email by noon (12:00) on the following Wednesdays:
Diligent preparation and active participation are essential for the success of this class! Of course, situations may arise when you cannot attend class. As a courtesy, please let me and your fellow students know beforehand.
Towards the end of the semester, you will have to complete a take-home exam consisting of a data analysis project.
The take-home exam (and accompanying dataset) will be handed out on 21 December, 2024
Your answers consisting of (1) an analysis .do file and (2) a substantive write-up should be submitted no later than 17 January, 2025.
Please email me your exam directly (no need to upload to Ilias). Failure to submit your take-home exam on time will result in failing the course.
I expect you to complete the take-home exam individually. You must make a written declaration that the work is wholly your own when submitting your answers. Simply put: don’t cheat.
Note: the current version of the syllabus is a work in progress.
Session 1 (6 Sept): Introduction.
Meet and greet. Discuss descriptive vs. causal research. Form groups for
the remainder of the semester.
Session 2 (13 Sept): Statistics Review.
T-tests and regression with dummy variables. Discuss problems of omitted
variable bias.
Session 3 (20 Sept): Potential Outcomes
Potential outcomes and selection bias. Randomized experiments and the
fundamental problem of causal inference. Average Treatment Effects.
Session 4 (27 Sept): Uncertainty, Covariates, Blocking and
Natural Experiments.
Uncertainty and balance tests. Blocking and precision. “As-if” random
assignment in observational data. Differential treatment probabilities
and fixed effects.
Session 5 (4 Oct): Treatment Heterogeneity and Interaction
Effects.
Interpreting interaction models with dichotomous and continuous
moderators. Data visualization.
Session 6 (11 Oct): Discussion, Review and Taking Stock.
Session 7 (18 Oct): Difference-in-Differences
Visual logic and estimation. Discuss the parallel trends assumption.
Prior trends and placebo effects.
Session 8 (25 Oct): Instrumental Variables I.
Experimental 1-sided non-compliance. Introducing compliers and
never-takers.
Intent-to-treat effects, the compliance rate, and the Complier Average
Treatment Effect. Two-stage-least-squares estimation.
Allerheiligen (1 Nov): No Class.
Session 9 (8 November): Instrumental Variables
II.
Review from last time. 2-sided non-compliance. Defining compliers,
always-takers, never-takers, and defiers. Continuous instruments and
treatments. IV Assumptions.
Session 10 (15 Nov): Reading Discussion
Session 11 (22 Nov): Regression Discontinuity.
Data visualization. Parametric estimation, functional form and bandwidth
selection. Checking continuity assumptions and sorting tests. Robustness
and placebo checks.
Session 12 (29 Nov): Student Presentations of the Problem Set.
Session 13 (6 Dec): Discussion, Wrap-up, and
Q+A.
Discuss the reading and fuzzy-RDD. Answer student questions. Explain the
final exam.