Cutoff

The cutoff is the threshold value of the Running Variable that switches treatment on or off in a Regression Discontinuity design. Above the cutoff units are treated (or treatment probability jumps); below they are not. The design exploits the fact that units just on either side of the cutoff are essentially comparable, while their treatment status differs.

When to use

Identify the cutoff first — it is the institutional rule that creates the design. PS_5's cutoff is 0.50 (union forms iff >50% vote in favour). Other examples: Medicare eligibility at age 65, legal driving age at 16, scholarship eligibility at 1,200 SAT points. With the cutoff fixed, RDD estimation centres the running variable ($r - c$) so the coefficient on the treatment dummy reads off the jump at the threshold.