Test Design Under Falsification

We study the optimal design of tests with manipulable inputs. Tests take a unidimensional state world as input and output, an informative signal to guide receiver's approve or reject decision. The receiver wishes only states that comply her baseline standard. An agent preference for approval can covertly falsify at cost. characterize receiver‐optimal show they rely on productive falsification by compliant states. They work setting more stringent operational standard, granting noncompliant positive probability deter them from falsifying also how falsification‐detection technologies improve tests. allow designer build implicit cost into test, in form devaluations. Exploiting this channel requires enriching space.