The Theory and Practice of Point-Optimal Testing

It is argued that the general formulation of point-optimal tests, as summarized in King (1988), (i) makes use of a benchmark which is excessively stringent, (ii) fails to integrate the method with traditional methods of testing, (iii) involves using data to form a specific null hypothesis, (iv) offers no guidance as to proper procedures in the event the null is rejected, (v) eschews the use of Bayesian methods while demanding the use of prior information, and (vi) yields tests with unknown power in the presence of specification error. For these and other reasons, point-optimal testing cannot yet be wholeheartedly recommended as part of "the econometrican's repertoire"