Rationality of early release data is typically tested using linear regressions. Thus, failure to reject the null does not rule out the possibility of nonlinear dependence. This paper proposes two tests that have power against generic nonlinear alternatives. A Monte Carlo study shows that the...