Schedule

A schedule of topics and readings is provided below. Each week will cover a single topic or group of topics. Monday lectures will typically be an introduction to the topic while Wednesday lectures will go into greater detail and involve some applications of the method. You should make sure to review the readings prior to that week’s lectures with an aim towards completing the reading assignments prior to Wednesday’s lecture.

Applications are optional but useful examples of a given methodological approach implemented in a published political science (or adjacent discipline) paper. I have tried to curate some examples for each week that provide a contrast between an “older” vs. a more “modern” paper using a particular methodology, highlight a debate over an interesting empirical question that hinges on a methodological problem, and/or just generally provide a cool empirical application using the framework being taught that week.

All hyperlinks to papers are either to the published version of the paper (if published) or to the working paper. You should use your university library access to obtain the full version if the article is not open access. Likewise, textbook readings will be to the free/open version of the book (if available) or to the UW-Madison university library eBook.

Week 1: Statistical Review

Wednesday, January 21

Topics:

  • Estimands and estimators
  • Review of statistical properties of estimators (bias/variance)
  • Large sample theory, big-O/little-o notation

Readings:


Week 2: Potential Outcomes + Randomized Experiments

Monday, January 26 & Wednesday, January 28

Topics:

  • Counterfactual reasoning and the potential outcomes model
  • What assumptions are needed to identify average treatment effects
  • Why randomized experiments satisfy these assumptions
  • Estimation and randomization inference in standard experimental designs

Readings:

Applications:


Week 3: Experiments: Attrition and Generalizability

Monday, February 2 & Wednesday, February 4

Topics:

  • What do we do when things happen after the experiment (e.g. attrition)?
  • Bounding treatment effects and partial identification.
  • Generalizing experimental results beyond the sample ATE

Readings:

Applications:

Week 4: Experiments: Incorporating Covariates

Monday, February 9 & Wednesday, February 11

Topics:

  • Stratification/blocking and using covariates in experiments
  • Analysis of cluster-randomized experiments
  • Introduction to regression estimators

Readings:

Applications:


Week 5: Selection-on-Observables

Monday, February 16 & Wednesday, February 18

Topics:

  • What to do when random assignment of treatment is not possible - common challenges of observational designs
  • Assumptions behind “no unobserved confounding” designs
  • Representing assumptions using graphical models
  • Covariate adjustment using stratification

Readings:

Applications:


Week 6: Selection-on-Observables - Regression and Weighting

Monday, February 23 & Wednesday, February 25

Topics:

  • Covariate adjustment using regression estimators
  • Propensity scores and covariate adjustment via weighting
  • Combining treatment and outcome models for a “doubly-robust” estimator

Readings:

Applications:


Week 7: Selection-on-Observables - Matching

Monday, March 2 & Wednesday, March 4

Topics:

  • Covariate adjustment using matching estimators
  • Midterm exam on Wednesday March 4th

Readings:

Applications:


Week 8: Instrumental Variables

Monday, March 9 & Wednesday, March 11

Topics:

  • Identifying causal effects under unobserved confounding using exogenous variation in treatment induced by an instrument
  • Estimation via the Wald estimator and two stage least squares (TSLS)
  • Interpreting the IV estimand - the LATE theorem

Readings:

Applications:


Week 9: Differences-in-Differences

Monday, March 16 & Wednesday, March 18

Topics:

  • Leveraging repeated outcomes over time to address unobserved confounding
  • Assumptions behind the “differences-in-differences” strategy – parallel trends
  • Estimation and diagnostics for the identification assumptions
  • Connection to regression estimators with “two-way fixed effects”

Readings:

Applications:


Week 10: Modern Differences-in-Differences

Monday, March 23 & Wednesday, March 25

Topics:

  • Differences-in-differences under staggered adoption
  • “New” DiD estimators - TWFE regression imputation vs. AIPW
  • Diagnosing pre-trends violations and assessing robustness
  • Relaxing parallel trends assumptions

Readings:

Applications:


Spring Break

No Class: Monday, March 23 & Wednesday, March 25


Week 11: Panel Data Causal Inference

Monday, April 6 & Wednesday, April 8

Topics:

  • Estimating effects when past outcomes affect future treatments
  • Estimators for effects of treatment histories
  • Pitfalls and cautions with lagged outcome regressions

Readings:

Applications:


Week 12: Regression Discontinuity Designs

Monday, April 13 & Wednesday, April 15

Topics:

  • Identification under unobserved confounding using quasi-random assignment at a cutpoint
  • Local randomization vs. continuity-based approaches
  • Estimation using local polynomial regression
  • Diagnostics for design assumptions - “bunching” tests and placebo outcomes

Readings:

Applications:


Week 13: Mediation and Sensitivity Analysis

Monday, April 20 & Wednesday, April 22

Topics:

  • Assessing the robustness of selection-on-observables results to hypothetical unobserved confounding.
  • Decomposing causal effects into “direct” and “indirect” components
  • Estimating mediation effects under ignorability assumptions
  • Challenges and pitfalls of mediation analysis

Readings:

Applications:


Week 14: Shift-share designs and experiments with interference

Monday, April 27 & Wednesday, April 29

Topics:

  • Identification with instruments that are inner products of unit-specific “shares” and independent global “shifts”
  • “Share-exogeneity” and connections to differences-in-differences/“multiple instruments”
  • “Shift-exogeneity” and connections to experiments with interference

Readings:

Applications: