Democratic and Republican politicians agree on one thing about President Donald Trump’s tax returns: The Constitution determines who can see them. Democrats such as Representative David Cicilline insist that Congress needs access to fulfill its “constitutional responsibilities of oversight” and evaluate possible violations of the Constitution’s emoluments clauses. Republicans such as Representative Bradley Byrne insist that the House Ways and Means Committee’s request for the returns raises “questions of grave constitutional significance.” According to Republican Senator Chuck Grassley, the request goes beyond what the Framers of the Constitution “had in mind” when they “created Article I.”
Statements such as these illustrate something important about how the Constitution figures in public life. Elected officials from both parties appeal routinely to the nation’s foundational document. But, far from serving as a symbol of “unity and common purpose,” the Constitution has come to enable, or even exacerbate, partisan strife. In political debates such as the Trump tax tussle, it often feels as if the United States has two legal charters, one for Republicans and another for Democrats.
That’s not just an anecdotal impression. The tools of computational analysis shed light on how wide the chasm has grown.
The three of us recently examined the evolution of constitutional rhetoric on the floor of Congress from 1873 to 2016. We first identified the hundreds of thousands of remarks that referred to the Constitution. We then trained a machine-learning classifier to predict—based solely on the content of the remarks—whether Republicans or Democrats were speaking. If the algorithm finds this task hard to do, it implies that the parties are apt to talk in similar or overlapping ways. By contrast, if the algorithm performs this task with a high degree of accuracy, it implies that the parties are largely talking past each other.
The results are sobering. Since around 1980, it has become increasingly easy for an algorithm to predict whether any given constitutional remark was made by a Republican or a Democrat. It has likewise become increasingly easy to predict whether the speaker was a conservative or a liberal. By the time Trump took office, the machine was guessing right roughly 80 percent of the time, an all-time high by historic standards.
This result holds up across multiple machine-learning classifiers, multiple measures of algorithmic accuracy, and multiple criteria for what counts as a constitutional remark. Additional tests of the “disjointness” between the parties’ rhetoric point to the same conclusion: To an unprecedented extent, Republican and Democratic members of Congress no longer speak the same constitutional language.