Trump Support? Conventional Measures in the Era of an Unconventional President

• In Event: White House and Congressional Politics

Abstract

Conventional indicators reported in CQ’s 2017 Presidential Support Study show that President Trump racked up a “Record Success Rate”, winning 100 percent of House votes on which he expressed a clear position. Trump’s unconventional style, however, poses a fundamental challenge to measuring presidential support in Congress. If the president’s position taking behavior is neither honest (valid) nor consistent (reliable), interpreting the outcomes of roll call votes as record high presidential success is like giving an “A+” to a student who turned in a plagiarized paper. Fortunately, the data journalism website FiveThirtyEight started identifying roll call votes on which Trump expressed a position in 2017. While this analysis provides additional conventional indicators that do not address the unprecedented threat to empirical research posed by Trump’s behavior, analysts trained in statistics at 538 developed an innovative indicator—“Trump plus-minus”—that estimates how much more or less members support Trump than should be expected given how well he ran in their constituency. Expected support is estimated with regression analysis, and a member’s “plus-minus” is the difference between actual and expected support. This indicator is similar in concept and construction to past research on presidential support in Congress. The underlying theory—that members have an electoral incentive to give higher support to a president who ran well in their district—is consistent with political science research. Yet, past research consistently finds that party is a stronger determinant of presidential support in Congress, and there is evidence that the effects of popular support are conditioned by party—i.e., the effects of popular approval of the president are different for co-partisans and opposition partisans. Scores estimated by models without the interaction of vote margin and party may be misleading. Furthermore, we typically estimate models with data from more than one president to make findings more generalizable. This paper estimates “plus-minus” scores with a model that interacts vote margin with party for presidents from Carter to Trump.

Author

• Jon R. Bond, Texas A&M University, College Station