From Correlation to Causation: Introduction to Causal Inference Methods

Key topics covered

The course moves from the foundations of causal inference to the tools used to adjust for observed confounders. Days 1 and 2 define causal effects through potential outcomes and show why randomised experiments are the benchmark for causal claims. Days 3 and 4 turn to observational data: how causal graphs help decide what to control for, and how regression, matching and weighting estimate causal effects when treatment is as good as random given observed covariates. Day 5 asks how much we should believe such estimates, and gives a first overview of research designs for settings in which confounders are unobserved.

Each day follows the same structure: about 1 hour 45 minutes of lecture, a short break, and one hour of hands-on work. On Days 1–4 the hands-on work takes the form of an R lab replicating a published study; on Day 5 it is a research design workshop. Required readings are one or two book chapters per day. The course also prepares participants for Causal Inference for Observational Data, which covers difference-in-differences, instrumental variables and regression discontinuity in depth.

Day 1 – Causal questions and potential outcomes

What makes a question causal, and how does causal inference differ from description and prediction? We introduce potential outcomes, average treatment effects and the fundamental problem of causal inference, and show why naive comparisons mix the causal effect with selection bias. In a short group exercise, participants turn research topics into precise causal questions. In the R lab, participants refresh their R skills, use simulation to see selection bias at work, and analyse data from a published study.

Day 2 – Randomised experiments

Why does randomisation solve the selection problem? We cover estimation with differences in means and regression, t-tests and randomisation inference, the use of covariates for balance checks, precision and blocking, and power calculations for designing an experiment. In the R lab, participants analyse data from published field experiments in political science and economics, run a randomisation test, and calculate the sample size needed to detect an effect.

Day 3 – Regression, confounding and causal graphs

What does a regression coefficient estimate when we 'control for' other variables? We cover regression anatomy and omitted variable bias, introduce causal graphs (DAGs) and the backdoor criterion, and discuss bad controls: colliders, mediators and post-treatment variables. In a group exercise, participants decide whether a variable should be controlled for in a series of DAGs, then draw the DAG for their own research question. In the R lab, they simulate confounders, mediators and colliders to see how each affects regression estimates, and compare regression results with an experimental benchmark.

Day 4 – Selection on observables: matching and weighting

When treatment is as good as random given observed covariates, we can compare like with like. We cover subclassification, nearest-neighbour matching, propensity scores and inverse probability weighting, with diagnostics for balance and overlap, and ask whether these methods can recover experimental benchmarks. In the R lab, participants apply matching and weighting to observational data from published research, check covariate balance, and compare their estimates with experimental results.

Day 5 – Sensitivity analysis and an overview of research designs

How much should we believe a selection-on-observables estimate? We cover placebo tests and sensitivity analysis: how strong would an omitted confounder need to be to overturn a result? We then give an overview of difference-in-differences, instrumental variables and regression discontinuity. In the research design workshop, participants work in small groups to develop a causal question, an ideal experiment, a DAG and a realistic design for their own research, and pitch it to the class for feedback.

Post-course assignment

Participants seeking the additional ECTS credit submit a written assignment of approximately 2,500 words (excluding references, tables and figures). The assignment builds on the research design developed and presented in the Day 5 workshop, and can be completed individually or in the same groups as the workshop. Participants choose one of two options:

1. Research design
Develop the workshop presentation into a full causal research design. The paper should state the causal question precisely (units, treatment, outcome, estimand); describe the ideal experiment you would run if you could; present a DAG distinguishing observed from unobserved confounders; propose a realistic design and justify its identifying assumptions; and discuss the main threats to identification and how you would probe them (e.g. balance checks, placebo tests, sensitivity analysis). The paper should take into account the feedback received in the workshop.

2. Methods review
Critically review how causal claims are made in the literature on the question you presented in the workshop (about 4–6 empirical studies). For each study, identify the causal question, the estimand, the design and its identifying assumptions, and assess their credibility using the concepts from the course. Conclude with what a more credible design for this question could look like, drawing on your workshop presentation.

Papers are assessed on the precision of the causal question, the appropriateness and justification of the design, the critical discussion of assumptions and threats, and clarity of presentation.

Prerequisite knowledge

Participants are expected to have a working knowledge of linear regression (OLS), statistical inference (standard errors, hypothesis tests, confidence intervals), and basic R, meaning they can load data, run a regression and read the output. They don’t need any prior knowledge of causal inference.

Software: R and RStudio.

Before the course

Before the course begins, participants are asked to install R and RStudio, and complete the reading for the first day. This should take no more than two hours.

Learning commitment

The course includes three hours of live teaching each day for five days, alongside readings and other learning activities.

Once you have registered and paid, you will receive access to our Learning Management System (LMS) approximately two weeks before the course start date. Course materials, including pre-course readings, will be available here. Allow approximately 20 hours per week before the course to review the materials and complete readings and any pre-course tasks.

During the course week, you should allow approximately two to three hours per day for preparation and assignments, in addition to the live teaching sessions.

The course offers the opportunity to earn three ECTS credits. To earn a fourth credit, you will need to complete a post-course assignment involving approximately 25 hours of work.

Disclaimer

This course description may be subject to subsequent adaptations (e.g. taking into account new developments in the field, participant demands, group size, etc.). Registered participants will be informed at the time of change.

By registering for this course, you confirm that you possess the knowledge required to follow it. The instructor will not teach prerequisite items. If in doubt, please contact us before registering.