Course Title: Selection on Observables: Unconfoundedness and Regression Discontinuity

Dates: Monday, June 3rd to Friday, June 7th

Instructor: Dr. Scott Cunningham

Preferred Email: [email protected]

Website: Scott Cunningham's Website

Note: This syllabus may change, but I am making an effort to keep it fixed.

Selection on observables is an approach to causal inference that takes advantage of a known and quantified variable that assigns units to treatment or control. We will cover two methods in that category: the unconfoundedness methods and the regression discontinuity design methods. Unconfoundedness methods involve estimating aggregate causal parameters using the known and quantified confounders directly through the construction of weights, matching techniques or regression adjustment. Regression discontinuity design is a method where the selection into treatment is based on an observable variable called the "running variable". We will explore both methods, focusing on heterogeneous treatment effects, regression, and matching using both coding examples and lectures.

Econometrics or equivalent.

• Develop comprehension of causal inference as a theoretical field.
• Build confidence in understanding and applying the methods to data.
• Gain competency in using modern statistical software to implement the methods practically.

Required:

• Cunningham, Scott. Causal Inference: the Mixtape (Yale University Press). Available for free on Scott Cunningham's website.

• Crits: Three assignments where you will critique podcast interviews with winners of the 2021 Nobel Prize in Economics.
• Class Participation: Come to class, take notes, read the lecture slides ahead of time, read the chapters in the Mixtape

• Monday, June 3rd: 14:00 to 17:00 (3 hours, no breaks)
• Tuesday, June 4th:
  • 10:00 to 12:00 (2 hours)
  • 12:00 to 13:00 (lunch)
  • 13:00 to 16:15 (3 hours plus a 15-minute break)
• Wednesday, June 5th:
  • 10:00 to 12:00 (2 hours)
  • 12:00 to 13:00 (lunch)
  • 13:00 to 16:15 (3 hours plus a 15-minute break)
• Thursday, June 6th:
  • 10:00 to 12:00 (2 hours)
  • 12:00 to 13:00 (lunch)
  • 13:00 to 16:15 (3 hours plus a 15-minute break)
• Friday, June 7th: 14:00 to 16:00 (2 hours, no breaks)

• Potential outcomes and DAGs
• Nonparametric methods (matching and weighting)
• Semiparametric methods (propensity scores)
• Regression, regression adjustment, and regression weighting

• Local linear regression with polynomials
• Optimal bandwidths
• Density tests
• Visualization

Feel free to copy and paste this into your GitHub repository README file. Let me know if there are any other details or modifications you need!