HS 315F — Evaluating Evaluation Methods

Meets for one-half semester and yields half-course credit.

The ideal goal when evaluating an intervention is to show causal effects, but this is often impossible to achieve with non-experimental data in real world settings. The purpose of the course is to open a tool box with potential methods so students learn the advantages and disadvantages of different methods and when to use them. Randomized-controlled trials (RCTs) have become the gold-standard to judge the causal impact of interventions in development but face many limitations and a growing chorus of critics who advocate a more eclectic approach. We start the course with RCTs but quickly move on to examine near neighbors that merit attention because they can also show causal impacts. In fact, near neighbors can sometimes do a better job than RCTs in impact evaluations. The neighbors considered in this course include propensity score matching, instrumental-variable estimations, intent-to-treat, and natural experiments. One important gap in all the evaluation literature is how to deal with puzzles after an evaluation. In the course we break new ground by examining the use of ethnographic and other methods to return to the field to answer puzzles left by the initial evaluation. Usually offered every year.
Ricardo Godoy