In economics, many questions revolve around causality. For example, one question that interests labor economists is: What are the returns to education? More specifically, our students might wonder about the labor market returns of a college degree from ISEG. Simply observing a positive relationship between education and lifetime earnings in observational data is not sufficient to answer this question, as individuals with higher earnings potential may be more likely to pursue education. However, conducting controlled experiments to determine causality is often infeasible in this context. While society may accept randomizing the COVID-19 vaccine to test its efficacy, randomizing access to education would be considered unethical. As a result, economists frequently turn to Instrumental Variable (IV) methods, a powerful statistical tool that helps establish causal relationships when experiments are not feasible.
My recent publication in the Review of Economics and Statistics advances this methodology by addressing a common challenge in IV analysis: the issue of "partial compliers". Partial compliers are individuals who receive only part of the treatment for which economists want to estimate the causal effect. In the context of college degrees, partial compliers are students who drop out after initial enrollment or who enroll late and graduate after transferring from a different program. My paper introduces a new IV framework that accounts for these partial compliers and identifies the (local) average treatment effect of receiving the full treatment compared to no treatment at all. In contrast, standard IV methods are often biased in the presence of partial compliers or can only estimate the effect of one specific part of the treatment. We apply our framework to policy evaluation in the education market, healthcare market, labor market, and developing world, demonstrating that accounting for partial compliers provides more reliable insights into the effectiveness of policy interventions. By refining IV methods to handle partial compliers, this research contributes to more informative policy analysis, helping economists and policymakers better understand the effects of their interventions.
