Module 4: Difference-in-Differences Estimation
Overview
This module introduces the difference-in-differences identification strategy. We review the basic setup and assumptions necessary for the diff-in-diff estimator to provide a valid estimate of causal impacts. We also survey some of the seminal early applications of diff-in-diff estimation to social policy.
This module includes one reading, four video lectures (approximately 62 minutes total), and an empirical exercise.
Readings
Impact Evaluation in Practice: Chapter 7
Review Questions
- What is the difference-in-differences estimate of the treatment effect? How is it constructed, and what must be measured in order to construct it?
- What counterfactual is used in diff-in-diff estimation, and what is used to estimate that counterfactual?
- What is the "equal trends" (or "common trends") assumption, and how might one test it?
- How did Professor Esther Duflo test the equal trends assumption in her study of school construction in Indonesia?
Video Lectures
Lecture 4.1 Intro to Difference-in-Differences Estimation (6:56)
Review Questions
- What type of data is required for difference-in-differences estimation?
- How can difference-in-differences estimation overcome the problems (i.e. biases) inherent in each of the false counterfactuals?
- What are the four cell-level means required to calculate the difference-in-differences estimator?
- Which of those four cells represents individuals/observations that have received treatment?
Lecture 4.2 Diff-in-Diff Pioneers (21:19)
Review Questions
- Who was Ignaz Semmelweis, what public health problem was he attempting to solve, and how did he propose to solve it?
- In Ignaz Semmelweis' study, what were the treatment and comparison groups?
- What did his results show?
- How did John Snow use a difference-and-differences style research design to demonstrate that cholera was likely caused by contaminated water?
- What change in minimum wage law did Marie Obenauer and Bertha von der Nienburg study?
- What data did Obenauer and Nienburg collect, and what did their results show?
Lecture 4.3 Identifying Assumptions for Diff-in-Diff Estimation (13:10)
Review Questions
- What is the common trends assumption?
- What must be true about time trends for diff-in-diff estimation to provide a valid estimate of a program's causal impact?
- What must be true about selection bias for diff-in-diff estimation to provide a valid estimate of a program's causal impact?
- Describe a setting in which the common trends assumption would likely be violated.
Lecture 4.4 Diff-in-Diff Examples (20:42)
Review Questions
- What law change did David Card and Alan Krueger study? What were the treatment and comparison groups in their research design? What were the pre-treatment and post-treatment periods?
- What were their findings?
- What policy experiment did Esther Duflo study in Indonesia?
- What were her main findings? How did school construction impact educational attainment and adult wages?
Lecture Slides
A handout version of the lecture slides is available here.
If you wish to use these slides for teaching, the underlying beamer files and supporting materials are here.
Papers Mentioned in the Lectures
Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania by David Card and Alan Krueger
Schooling and Labor Market Consequences of School Construction in Indonesia: Evidence from an Unusual Policy Experiment by Esther Duflo
Empirical Exercise
The empirical exercise is available here. In the exercise, we analyze data from Ignaz Semmelweis’ handwashing intervention in the maternity hospital in Vienna. The data come from Semmelweis’ (1861) book (some helpful person put them on Wikipedia). The empirical exercise involves an in-class activity and a homework activity.
If you wish to use this exercise for teaching, the supporting materials are here.