Bandwidth selection for parametric RDD

[Duccio Milani]

Greetings to all,

I am using an RDD method for a research project, and to add robustness to my results, I am using both a parametric and non-parametric approach.

Given the scarce literature on this, I wondered if the MSE bandwidth selection (Calonico, Cattaneo, Farrell and Titiunik, 2017) is also the best option for the parametric approach or I should resort to other methods (CV, IK, CCT).

Thanks in advance.

[Fei Wang]

It's difficult to say which one is the best, but you may use the default bandwidth for all major analyses and check the robustness of the estimates with alternative bandwidths.