Statistical Methods for Causal Inference

Apply now

This course introduces students to advanced statistical tools of causal inference for impact evaluation and policy analysis.

Course description

There is great interest among students and practitioners today to understand the causal mechanisms underlying major events. Identifying cause-and-effect relationships is important for impact evaluation and effective policy design. Such identification can help us answer questions like: "What causes an economic downturn?", "Does universal basic income reduce unemployment?" and "Does a carbon tax reduce greenhouse gas emissions?"

However, identifying causal relationships using data is often error prone. Differentiating causality from simple correlation requires learning and applying sophisticated quantitative tools. The golden standard of identifying causal linkages relies on designing experiments, often through randomised control trials. But designing a randomised control trial is not always feasible or ethical. Moreover, some events might have already happened in the past, such as a financial crisis or a cyclone. How can one use observational data to analyse the causal effects of such events?

This course provides a hands-on introduction to statistical methods for causal inference. Over two weeks, students are introduced to experimental and quasi-experimental methods which allow them to infer cause-and-effect relationships robustly. We teach these methods from both a theoretical and applied lens, supplementing lectures with hands-on computer tutorials in the R programming language to help students learn by doing.

Continue reading below for course topics, learning objectives and more.

Summer courses

View the full programme

Fees and scholarships

See course fees in more detail

How to apply

Application and admission details

See how to apply

Course level

Master / Advanced / PhD

Credits

3 ECTS

Contact hours

Language

English

Tuition fee

€735 - €1310

Learning objectives
By the end of this course, students will be able to:
- Understand the difference between correlation and causation.
- Apply quantitative methods of statistical data analysis to infer causal relationships.
- Identify confounding factors that threaten causal inference and hamper the internal and external validity of analytical findings.
- Critically analyse data using statistical methods like experiments, matching analysis, difference-in-differences, regression discontinuity, and instrumental variables estimation.
- Explore challenges and limitations in the use of quantitative methods of causal inference such as data availability, missing data, and measurement errors.
- Apply diagnostic knowledge to inform impact evaluations and develop evidence-based policies
About the course coordinators

Dr. Sanchayan Banerjee is an Assistant Professor of Environmental Economics at the Institute for Environmental Studies of VU Amsterdam. He holds a PhD from the London School of Economics and Political Science. Sanchayan is a visiting fellow at the London School of Economics and King's College London. He sits on the Steering and Development Committee of the International Behavioural Public Policy Association. He also holds a fellowship of the Higher Education Academy (FHEA) in the United Kingdom. He has completed a PG Certificate of Higher Education (equivalent to the STQ) and is qualified as a professional external examiner for universities. He has broad teaching experience across regular schools, summer schools, and executive education courses. Currently, he is a lecturer and course manager of the Methods for Economic Analysis and Sustainable Energy Systems courses in the MSc Environment and Resource Management program at IVM. Dr. Sanchayan taught more than 100 students in LSE Summer School annually between 2018 and 2022, where he managed a large course on Environmental Economics and Sustainable Development (average teaching score for the last two years: 4.8/5). During the last four years, Dr. Sanchayan has also taught a course on Applied Quantitative Methods (of Causal Inference) to MSc students at LSE, in addition to teaching on environmental economics, development economics, and policy. This professor is inclusive and diverse in his teaching practices. He has been awarded three consecutive LSE Excellence in Class Teaching Awards for his teaching practices between 2019 and 2022. He was also nominated by his students in the last two years for LSE Student Union Teaching Awards in the categories of Inclusive Teaching, Innovating Teaching, and Sharing Subject Knowledge.
Jack Fitzgerald is a doctoral candidate in economics at VU Amsterdam. He is based in the Ethics, Governance, and Society department at the School of Business and Economics, and he holds a teaching affiliation with the Criminology department at the Faculty of Law. His dissertation focuses on behavioural experiments analysing determinants of corporate compliance with public policies, but his research also delves into causal inference econometrics, exploring the econometrics of experiments and applying quasi-experimental causal inference techniques in practice.
Course topics and schedule

In week one, students are introduced to pitfalls of standard regression analysis by identifying multiple threats to causality, such as omitted variable bias, endogeneity concerns like simultaneity bias, and reverse causality problems. Then, they are introduced to the potential outcomes framework (Neyman, 1923; Rubin, 1977), a standard workhorse model of statistics, that forms the basis of identifying cause-and-effect relationships. Following this, they learn how to design experiments and analyse experimental data for impact evaluation and policy analysis. Finally, at the end of the first week, students are introduced to quantitative methods of causal inference for observational data, where they are taught to select observables to create treatment and control units using matching analysis.
In week two, students continue to learn four more methods to evaluate quasi-experimental phenomena (so called “natural” experiments). Here, we start with instrumental variables regression, including a guest lecture by Professor Hans Koster on the use of instrumental variables to estimate the impact of historical monument refurbishment on Dutch housing prices. Following this, students are introduced to panel data designs with difference-in-differences estimation. Finally, students are introduced to regression discontinuity designs.
All lectures are complemented with hands-on computer tutorials, where students learn how to apply these quantitative methods of causal inference using R.
Forms of tuition and assessment

Students will be taught through lectures and computer tutorials, and their final mark will be based on a final project (60%) and daily quizzes (40%).
Additional requirements

All students must bring their own laptops to the course. Said laptop should be capable of running R Studio.
Detailed course syllabus

You can download the preliminary course syllabus here.
^{*Please note that it is preliminary and that it still might be subject to change.}