On the Validity of Geographically Local Average Treatment Effect Estimates From Spatial Regression Discontinuity Designs

• In Event: Roundtable Session 97
  In Roundtable Session: Accommodating Data From Quasi-Experimental Designs

Abstract

Regression discontinuity design (RDD) theory and analysis can be used to estimate geographically local average treatment effects (GLATEs) of programs and policies precisely at borders separating treatment and comparison areas. The method, referred to as spatial regression discontinuity design (SRDD), shares some of the elements of RDD that strengthen internal and external validity, but GLATEs estimated at preexisting geographic borders are subject to a number of validity threats. This study identifies the validity threats and proposes design and analytic enhancements to help rule them out. An evaluation of a teacher performance pay program demonstrates SRDD and the proposed enhancements.

Author

• Christopher T. Moore, University of Minnesota