Evaluation of Social Policies & Programs

This PhD-level course offers an introduction to the practical application of microeconomic principles and cutting-edge statistical techniques to the evaluation of social programs. The substantive focus will be on programs in health, education, welfare, and workforce training, focusing especially on cases drawn from Latin America. Students will be invited to work through a series of concrete program evaluations conducted for a number of international and other organizations ranging from the World Bank to NGOs to the governments of Mexico, Brazil, and elsewhere. The primary focus is on the design and execution of program evaluations (the assessment of on-going programs or of programs after the fact). This will call for rigorous treatment of:

•The formal logic of experimental and quasi-experimental research design

•Multivariate statistical techniques

•The management of large databases

•The use of the Stata and R statistical analysis packages

•Statistical tools and techniques for dealing with selection bias and other commonly encountered threats to internal validity

•And much more…

Readings draw from the literature on program evaluation, quasi-experimental design, and microeconometrics. The econometrics of evaluation has become a particularly exciting area of research in recent years. While technically challenging, it has important implications on public policy analysis. Students will write occasional commentaries on real-world studies and analyze large real-world datasets using the methods and techniques developed in the course. Students will be evaluated in terms of the quality of their participation in class discussion, two take-home exams, and on their performance on take-home exercises that will be assigned approximately every two weeks.

Prerequisites: Priority is given to PhD students campus-wide; qualified LBJ masters students are also welcome (instructor consent required).  Assumes that students have a good command of graduate-level econometrics (eg, familiarity with maximum likelihood estimation and logit/probit models) and be comfortable working with summation and matrix notation.