Two strategies reduce the likelihood of RD mistakes, though neither provides perfect insurance. The first models nonlinearities directly, while the second focuses solely on observations near the cutoff. We start with the nonlinear modeling strategy, briefly taking up the second approach at the end of this section.
