This course provides an overview of a variety of modern techniques for studying the ability of changes to produce effectiveness outcomes, whether these are policy changes, or observed natural or human events whose causal effects we wish to measure.
The methods you will study have immediate practical implications for evaluating policy outcomes and their effectiveness, but have broad application to social science problems across many fields. As approaches for making causal inferences from observational data, these techniques have moved to the centre of social science fields such as economics, political science, development, and health policy, where understanding causes and effects is of paramount importance but also where controlled experiments are seldom practically feasible or ethically possible. As such, knowledge of these techniques is increasingly important for consumers and practitioners alike.
Combining basic statistical theory, principles of smart research design, and hands-on experience with real data, this course will give you a basis for being good consumers and practitioners of modern quantitative social science. Through a variety of applied examples drawn from multiple fields of social science and policy, we will see how these methods are applied to evaluate outcomes. In class exercises, we will also see how to implement such techniques using example data, and lectures are complemented with computing exercises using real-world data and the statistical software R.
The course begins with a brief introduction to various philosophical approaches to causality. We focus the remainder of the course on the potential outcomes framework, which is increasingly the lingua franca of causal analysis in the quantitative social sciences and related fields. Using this framework, we begin with experimental logic, presenting the randomized experiment as the gold standard for causal inference. Following this, we use the potential outcomes framework to apply experimental logic to a variety of common non-experimental settings, a situation often referred to as a quasi-experiment. Quasi-experimental designs covered include matching, instrumental variables, difference in differences, synthetic control, and regression discontinuity.
This course is aimed at third-year undergraduates, postgraduates, and interested professionals.
Dates: 3 – 21 August 2020
Lecturer: Dr David Hendry