Constitutional Arguments in Presidential Veto Messages

We analyze presidential vetoes throughout the U.S. history and use algorithmic natural language processing methods to identify, encode, and classify constitutional arguments raised by veto messages. Going beyond the familiar hypothesis that modern presidents employ constitutional arguments instrumentally, we will classify those arguments as institutional, i.e., designed to promote the institutional interests of the executive, and policy-oriented, i.e., designed to further the president’s policy preferences. We hypothesize that institutional arguments in veto messages are more similar to mainstream constitutional arguments, more likely to find acceptance, and more consistent across parties, while policy-oriented arguments are more likely to be addressed to the public (and thus less professionalized), and less consistent across parties. We will test those hypotheses by comparing the style and semantic structure of presidential constitutional arguments with other sources of constitutional arguments, such as Supreme Court opinions and briefs or OLC opinions, again using NLP and machine learning algorithms. The results will be validated against a qualitatively coded subset of vetoes, including all modern ones.