Regression Discontinuity Design Flashcards
How does RD differ from randomized experiments?
Regression discontinuity involves regarding treatment as being as good as randomly assigned along a narrow window of a running variable. A randomized experiment involves a true randomization process
in which various subjects are actually assigned to difference treatment arms in an experiment, such as to receiving the treatment drug or a placebo.
What are the two types of RD designs and how do they differ?
Sharp and fuzzy. Sharp is a jump from E[D|below cutoff] = 0 to E[D|above cutoff] = 1 (or vice versa), while fuzzy is E[D|below cutoff] /= E[D|above cutoff].
In the context of a regression discontinuity design, what is the role of the running variable and how is it used to identify the causal effect of treatment on the outcome?
The running variable is the variable on which the cutoff is defined. So for some cutoff, usually set by a policy, and picked somewhat arbitrarily, people above the cutoff are handled differently than people
below the cutoff. In the case of a sharp discontinuity treatment shifts entirely at the cutoff, whereas with a fuzzy discontinuity the propensity changes.
