Causal Estimation and Inference
How can we learn what a treatment, policy, or decision causes? Begin with the overview, then work through the resources in order—from potential outcomes and research design to matching, panel methods, instruments, and heterogeneous effects.
Overview
Start with Causality Overview, then watch What Is Causality?
Causality Overview
The starting point for the course sequence.
Potential Outcomes
Define causal effects by comparing potential outcomes and identify the missing counterfactual.
Directed Acyclic Graphs (DAGs)
Represent causal assumptions with graphs and reason about which variables to adjust for.
Random Experiments
Use random assignment to build credible comparisons between treatment and control groups.
Matching
Build comparisons using observed pretreatment characteristics, then examine overlap and balance.
The Torpedo Bat Study
Explore a reproducible simulated study. Inspect the original data, try matching or weighting, check balance, and reveal the estimate and simulation truth separately.
Opens in a separate tab. Saved comparisons last for the current page session.
Regression Adjustment
Use regression to adjust comparisons for observed characteristics.
Omitted Variable Bias (OVB)
Examine how omitted confounders can distort a regression estimate.
Panel Fixed Effects
Use repeated observations to account for unobserved differences that remain fixed over time.
Difference-in-Differences (DiD)
Compare changes across groups and examine the assumptions behind the comparison.
Instrumental Variables (IV)
Study how an instrument can identify causal variation when treatment is endogenous.
Regression Discontinuity (RD)
Use a treatment-assignment cutoff to compare observations near a threshold.
Synthetic Control
Construct a comparison unit from a weighted combination of untreated units.
Synthetic Difference-in-Differences
Combine weighting and difference-in-differences to build a counterfactual comparison.
Heterogeneous Treatment Effects (HTE)
Investigate how treatment effects differ across people, groups, or characteristics.
Simultaneous Systems
Study identification when outcomes and explanatory variables are jointly determined.