杏吧论坛

Dr Daniel De Kadt

This course is available on the MSc in Applied Social Data Science, MSc in Behavioural Science, MSc in European and International Politics and Policy, MSc in European and International Politics and Policy (杏吧论坛 and Bocconi), MSc in European and International Politics and Policy (杏吧论坛 and Sciences Po), MSc in Human Geography and Urban Studies (Research), MSc in International Social and Public Policy (Research), MSc in Political Science (Political Science and Political Economy), MSc in Social Research Methods, MSc in Statistics, MSc in Statistics (Financial Statistics), MSc in Statistics (Financial Statistics) (Research), MSc in Statistics (Research), MSc in Statistics (Social Statistics) and MSc in Statistics (Social Statistics) (Research). This course is available as an outside option to students on other programmes where regulations permit.

This course is not controlled access. If you register for a place and meet the prerequisites, if any, you are likely to be given a place.

Knowledge of multiple linear regression and some familiarity with generalised linear models, to the level of MY452 or equivalent. Familiarity with notions of research design in the social sciences, to the level of MY400 or equivalent. Familiarity with R.

This course provides an advanced introduction to modern quantitative causal inference in the social sciences. The class covers the canonical approaches to causal inference and includes excursions to the leading edge of the field. We begin with a foundational introduction to both the potential outcomes and graphical frameworks for causality, before considering a range of applied research designs for causal inference. We first discuss identification and estimation for classical randomized experiments, with brief forays into more complex designs. We then turn to a range of observational designs, which will be the primary focus of the class. The first of these is selection on observables (SOO), and we cover regression, matching, and weighting as estimations strategies, before discussing sensitivity analyses and interval estimation (bounds). We then consider instrumental variables (IV) from both the modern potential outcomes perspective and, briefly, the classical structural approach, before delving into new IV settings like examiner designs, shift-share designs, and recentered instruments. From IV we move to regression discontinuity designs (RDD); we approach identification from the continuity perspective and introduce local polynomial approximation for estimation. Finally, we pivot to causal inference with time-varying data, focusing first on the canonical two-period difference-in-differences (DiD) design. We then consider generalised DiD with many time periods, treatment effect heterogeneity, staggered assignment, and non-absorbing treatments. Throughout the class examples are drawn from different social sciences. The course includes seminars for each of the major methods, which combine the close reading and discussion of an applied paper with a brief session on implementation in R.

Combined hours across lectures and classes will be equivalent to a minimum of 30 hours of face-to-face teaching across the WT.

This course has a reading week in Week 6 of WT.

Problem sets from the computer classes can be submitted for feedback.

• Imbens, G. W. and Rubin, D. B. (2015). Causal Inference for Statistics, Social, and Biomedical Sciences: An Introduction. Cambridge University Press.
• Angrist, J. D. and Pischke, J.-S. (2009). Mostly Harmless Econometrics. Princeton University Press.
• Rosenbaum, P.R. (2010). Design of Observational Studies. Springer.

Take-home assessment (100%) in the ST.