Footnote:
Nach Informationen von SSRN wurde die ursprüngliche Fassung des Dokuments May 17, 2019 erstellt
Description:
The main attraction of Regression Discontinuity (RD) is local randomization around the cutoff. However, often the running variable (or just ''score'') is manipulated to ruin the local randomization. To detect manipulation, typically, the continuity at the cutoff of the score density function is tested. This paper shows via formal counter-examples that the score density continuity is neither necessary nor sufficient for RD identification. The score density continuity should not give a false sense of security when not rejected, and lack of it should not be taken as a reason to give up on the RD study hastily. Also, if the treatment cannot be implemented without manipulation of the score, then manipulation should be considered as part of the treatment effect. These findings suggest thinking more about the consequences of manipulation and the desired treatment effect, instead of mechanically applying score density continuity tests