Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
We present a test of the selection on observables assumption that neither requires instruments nor excluded covariates from the structural function. Instead, we rely on the presence of a discontinuously distributed variable among the set of controls. We develop formal testing procedures for a non-parametric additively separable model for binary and finite treatment variables. We also outline a nonparametric nonseparable extension. Our test is easy to implement and should be useful in many empirical settings. Specifically, we employ it to study selection concerns in the estimation of the impact of a nutritional aid program for pregnant women on birth weight.