FOR ALL THE florid journalistic commentary about voter polarization, extensive empirical studies have shown that the American electorate is no more polarized today than it was in the 1970s. What's changed is that the parties have sorted: Democrats have become more homogeneously liberal, Republicans more homogeneously conservative.
But even taking sorting into account, partisan hostility appears greater than any objective differences on the issues would seem to justify. To explain this divergence, a number of scholars have argued for a view generally known as "affective partisanship." Ordinary Democrats and Republicans may not differ that much on public policy, the argument goes, but they dislike the other side more than they used to, ascribing negative traits to them and even claiming they would be upset if their offspring married someone from the other party.
The latest and arguably strongest contribution to this line of work is Uncivil Agreement by Lilliana Mason, an assistant professor of government and politics at the University of Maryland. But before looking at her findings, I want to note some problems with studies of affective partisanship.
FIRST IS THE simple fact that the proportion of Americans who admit to having any partisan identity is at an all-time low. When the American National Election Studies began in the 1950s, three quarters of those queried claimed to be either a Democrat or a Republican. In the 2016 study, only about 60 percent did. Commercial polls similarly report that 40 percent or more of Americans today claim to be independents.
Some scholars contend that most independents are closet partisans, a contention that far exceeds the empirical support for it. You do not ordinarily think of an identity as something one denies or hides. It's usually something one affirms, as when Red Sox fans wear their hats and shirts in the hostile confines of Yankee Stadium. Indeed, Samara Klar and Yanna Krupnikov make a strong case in their 2016 book Independent Politics (Cambridge University Press) that a large swath of the electorate today finds the available partisan identities unattractive.
Measurement issues are a second problem. The workhorse variable in affective partisanship studies is the "thermometer" measure: Survey respondents rate individuals and groups according to how warm or cold they feel toward them. To my knowledge, no one has empirically demonstrated that such scores do in fact measure affect. Do I feel warm...