Main Workshop - Monday, August 12 - Friday, August 16
The Main Workshop will be hosted by Duke and Northwestern at the Duke Law School, Room 3041. The workshop will run daily from 9:00am - 5:00pm. Breakfast will be served at 8:30 each morning and lunch will be provided.
Registration is limited to 125 attendees, please register as soon as you are able. For questions about the workshop logistics or registration, please email Isabel Fox (firstname.lastname@example.org). Please email Bernie Black (email@example.com) or Mat McCubbins (firstname.lastname@example.org) for substantive questions or fee waiver requests.
Faculty | Registration | Schedule | Questions | Hotels
Research design for causal inference is at the heart of a “credibility revolution” in empirical research. We will cover the design of true randomized experiments and contrast them to natural or quasi experiments and to pure observational studies, where part of the sample is treated in some way, the remainder is a control group, but the researcher controls neither the assignment of cases to treatment and control groups nor administration of the treatment. We will assess the causal inferences one can draw from a research design, threats to valid inference, and research designs that can mitigate those threats.
Most empirical methods courses survey a variety of methods. We will begin instead with the goal of causal inference, and emphasize how to design research to come closer to that goal. The methods are often adapted to a particular study. Some of the methods are covered in PhD programs, but rarely with a focus on credible causal inference and on which methods to use with messy, real-world datasets and limited sample sizes.
Quantitative empirical researchers (faculty and graduate students) in social science, including law, political science, economics, many business-school areas (finance, accounting, management, marketing, etc.), medicine, sociology, education, psychology, etc. –anywhere that causal inference is important.
We will assume knowledge, at the level of an upper-level college econometrics or similar course,of multivariate regression, including OLS, logit, and probit; basic probability and statistics including conditional and compound probabilities, confidence intervals, t-statistics, and standard errors; and some understanding of instrumental variables. Despite limited prerequisites, this course should be suitable for researchers with recent PhD-level training and for empirical legal scholars with reasonable but more limited training.
Donald Rubin is John L. Loeb Professor of Statistics Emeritus, at Harvard. His work on the “Rubin Causal Model” is central to modern understanding of causal inference with observational data. Principal research interests: statistical methods for causal inference; Bayesian statistics; analysis of incomplete data.
Brigham Frandsen is Associate Professor of Economics and Visiting Associate Professor at MIT. Research interests include developing methods for causal inference on how treatment effects vary across the treated population, and applying those methods to labor, education, and health economics questions.
Joshua Angrist is Ford Professor of Economics at MIT. His work on “causal IV” is at the central to the modern revival of use of instrumental variables methods. Principal research interests: labor economics; econometrics. Author of Joshua Angrist and Jorn-Steffen Pischke, Mostly Harmless Econometrics: An Empiricist’s Companion (2009) and Mastering ‘Metrics: The Path from Cause to Effect (2014), and the Mostly Harmless Econometrics blog.
Jens Hainmueller is Professor in the Stanford Political Science Department, and co-Director of the Stanford Immigration Policy Lab. He also holds a courtesy appointment in the Stanford Graduate School of Business. His research interests include statistical methods, political economy, and political behavior. Papers on SSRN.
Main workshop tuition is $900; $600 for graduate students (PhD, SJD, or law) and post-docs; $400 for Duke and Northwestern affiliated attendees who register with a valid duke.edu or northwestern.edu email. The workshop fee includes all materials, temprary Stata 15 license, breakfast, lunch, snacks and an evening reception on the first workshop day.
You can cancel by July 1, 2019 for a 75% refund, and by July 22, 2019 for a 50% refund, but there are no refunds after that date.
We know the workshop is not cheap, we use the funds to pay our speakers, for meals, and other expenses; we don't pay ourselves.
Monday, August 12 (Donald Rubin)
Introduction to Modern Methods for Causal Inference
Overview of causal inference and the Rubin “potential outcomes” causal model. The “gold standard” of a randomized experiment. Treatment and control groups, and the core role of the assignment (to treatment) mechanism. Causal inference as a missing data problem, and imputation of missing potential outcomes. Rerandomization. One-sided and two-sided noncompliance.
Tuesday, August 13 (Brigham Frandsen)
Matching and Reweighting Designs for “Pure” Observational Studies
The core, untestable requirement of selection [only] on observables. Ensuring covariate balance and common support. Subclassification, matching, reweighting, and regression estimators of average treatment effects. Propensity score methods.
Wednesday, August 14: Morning (Joshua Angrist)
Instrumental Variable Methods
Causal inference with instrumental variables (IV), including (i) the core, untestable need to satisfy the “only through” exclusion restriction; (ii) heterogeneous treatment effects; and (iii) intent-to-treat designs for randomized trials (or quasi-experiments) with noncompliance.
Wednesday, August 14: Afternoon (Brigham Frandsen)
Going beyond estimating “average treatment effects,” and estimating quantile and marginal causal effects.
Thursday, August 15 (Jens Hainmueller)
Panel Data and Difference-in-Differences
Panel data methods: pooled OLS, random effects, correlated random effects, and fixed effects. Simple two-period DiD. The core “parallel changes” assumption. Testing this assumption. Leads and lags and distributed lag models. When does a design with unit fixed effects become DiD? Accommodating covariates. Triple differences. Robust and clustered standard errors. Introduction to synthetic controls.
Friday, August 16: Morning (Jens Hainmueller)
(Regression) discontinuity (RD) research designs: sharp and fuzzy designs; bandwidth choice; testing for covariate balance and manipulation of the threshold; discontinuities as substitutes for true randomization and sources of convincing instruments.
Friday, August 16: Afternoon
Feedback on Your Own Research
Attendees will present their own research design questions from current work in breakout sessions and receive feedback on research design. Session leaders: Bernie Black, Mat McCubbins, Jens Hainmueller, Vladimir Atanasov (William and Mary). There will be additional parallel sessions if needed to meet demand.
Stata and R sessions
On selected days (tentatively, Tuesday, Wednesday, and Thursday), we will run parallel Stata and R sessions to illustrate code for the research designs discussed in the lectures, or the speakers will build Stata code into their lecture slides. Presenters: Bernard Black (Stata) and Joshua Lerner (R).
Please email Bernie Black (email@example.com) or Mat McCubbins (firstname.lastname@example.org) for substantive questions or fee waiver requests, and Isabel Fox (email@example.com) for logistics and registration.