Chapter 13 — Complete Compliance: One-Sided and Two-Sided Violations When subjects don’t take the treatment they were assigned—or take treatment when assigned to control—the experiment faces a compliance problem. This chapter distinguishes one-sided from two-sided noncompliance, shows why imperfect compliance prevents recovery of the ATE, and introduces the behavioral assumptions (monotonicity and the exclusion restriction) needed to recover the local average treatment effect for compliers. It also covers intention-to-treat analysis and ATE bounds, and discusses how experimental design choices can maximize compliance from the outset.


  1. Imperfect compliance results when received treatment differs from assigned treatment.
  2. The two forms of violations are one-sided and two-sided incompliance.
  3. Without further assumptions, imperfect compliance prevents recovering the ATE, but alternative parameters of interest can be recovered by imposing new assumptions.
  4. Design considerations can attenuate compliance issues.
arrow-down-thickarrow-downarrow-leftarrow-right-thickarrow-rightcaret-downclosecodedeco-arrow-leftdeco-arrow-rightexternalfacebookfilterline-arrowlinkedinmagpaperpythonqarslider-arrow-leftslider-arrow-rightslidesxyoutubezip