Stephan Litschig- GRIPS- Balanced Subgroup Analysis in Regression Discontinuity Designs

Title: Balanced Subgroup Analysis in Regression Discontinuity Designs

Abstract: Subgroup analyses to explore mechanisms and conduct falsification tests are increasingly common in regression discontinuity designs. However, existing methods fail to account for other differences across subgroups beyond the characteristic of interest. As a result, evidence of subgroup effect heterogeneity - or lack thereof - is often difficult to interpret. This paper introduces a new approach to conduct regression discontinuity subgroup analysis based on inverse probability weighting. Observations in a given subgroup are weighted by the inverse of their conditional probability to belong to that subgroup, given a set of observable characteristics. Successful balancing of observables across subgroups helps to isolate effect heterogeneity due to the subgroup characteristic of interest from effect heterogeneity driven by observed confounders. The approach is illustrated with two studies that explore a particular mechanism (Fujiwara 2015) or run a falsification exercise (Solis 2017).

Presenter: Stephan Litschig, The National Graduate Institute for Policy Studies (GRIPS)

If you would like to subscribe to the Melbourne Institute Seminar Series email list, please contact us.