Regression Discontinuity Models and the Variance Inflation Factor

Abstract

The regression-discontinuity design (RD) is a powerful methodological alternative to the quasi-experimental design when conducting evaluations. The RD design involves testing post-test differences between the experimental and comparison group regression lines at the cutoff point for statistical significance. Regression discontinuity models can involve linear, curvilinear, and interaction terms in the
model specification, which are not orthogonally specified. Consequently, a variance inflation problem may exist when using regression discontinuity models in evaluation designs. This study investigated the impact of variance inflation on parameters specified in full and restricted regression discontinuity models. It is recommended that VIF be considered when including interaction effects in RD designs.