The literature on what influences issuance of presidential directives is bifurcated between emphases on formal constraints, such as divided government, and informal constraints, notably public opinion. Here we reassess the relevance of each, while paying careful attention to the type of directive employed, the relative significance of each action, and to potential endogeneity. We analyze a comprehensive dataset of directives from 1947-2016---including Executive Orders, presidential proclamations, and presidential memoranda---using Bayesian significance scores for 1947-2016, and estimate vector autoregression models that include both formal and informal constraint measures. Our results show evidence of endogeneity between directives and opinion, that divided government if anything deters presidential actions, and that what subsets of actions are studied---EOs v. all presidential directives, all actions regardless of significance or higher significance actions---is extremely germane. Hence, our results provide a cautionary note for those trying to draw broad inferences about what influences presidential unilateralism.