Regression Discontinuity design

2 min readDec 25, 2021


In my previous articles, I wrote about Quasi-experiment designs to reduce the effect of interference on the analysis. I have also written about switch-back experiment as a type of Quasi-experiment designs. In this article, I will be discussing regression continuity designs.

Regression-discontinuity (RD) designs is a kind of experiment design which helps us measure the effect of treatment in cases where it’s not possible to randomize the assignment of treatment or intervention. It is a quasi-experimental impact evaluation method used to evaluate programs based on a cut-off point. RD design is used to estimate treatment effect around the cut-off point, as at this point the treatment and control experiment units are most similar.

An example of Regression-Discontinuity experiment designs

The unique difference in characteristics of a RD designs as compared to those of pre-post experiment designs is that the assignment of treatment or control is based solely on a cut-off score in RD designs. This feature help us target a program or intervention to the users who most need it or deserve it.

Some other design considerations and variations:

  1. The cut-off point could be deterministic or probabilistic. In case of deterministic designs, our cut-off point is sharp (singular) value. This is also called a sharp design. The design with probabilistic cut-off point is called a fuzzy design.
Variations of Regression-Discontinuity experiment design example

2. We can also have multiple cut-off point if we want to decide on the effectiveness of program for multiple categories in a sample. We can also evaluate the effect of multiple treatments simultaneously on different cut-off points.