Few topics have generated as much scholarly and popular attention over the past two decades as the debate over polarization in American politics. While scholars have concluded that political elites have become more divided over time, there is much more disagreement about whether the same is true of ordinary voters, and whether such polarization has important consequences for politics and government more broadly. James Campbell’s new book, Polarized: Making Sense of a Divided America, takes up the question of mass polarization and, in so doing, makes an important contribution to this literature.

This book comes down squarely on the side of those claiming that the electorate is sharply divided and polarized, declaring that “America is a politically divided nation, it has been for some time, and has become more so in recent decades” (p. 2). Not only are Americans more polarized today, but Campbell also argues that it was changes in the mass public during the 1960s and 1970s that caused elites to polarize, inverting the typical top-down logic of this process. The book lays out evidence for this argument, and explores some of the consequences of this shift. Even if one disagrees with his ultimate conclusions, anyone seeking to better understand the dynamics of polarization in American needs to engage with Campbell’s thoughtful analyses.

The core of Campbell’s argument is that Americans are polarized today, and have been polarized for some time. He supports this argument through an analysis of a variety of over-time surveys, most notably the American National Election Study. He provides several key pieces of evidence for the argument. First, Campbell shows that Americans are increasingly willing to identify as liberals or conservatives, rather than identifying as moderates or being unable to place themselves on the liberal-conservative self-identification scale (the core of this evidence comes in Chapters 3 and 4). Moreover, such ideological self-identifications are correlated in meaningful ways to a variety of issue-positions, suggesting that these aggregate self-identifications actually tap latent political preferences.

Few would question the evidence Campbell brings to bear on this question, but one could reasonably challenge his interpretation of it. Take, for example, table 4.2 (p. 100), which examines how answers to the 7-point guaranteed jobs and a standard of living scale, and the 7-point government health insurance scale, have changed between 1972–1974 and 2004–2008. The percentage of respondents who take a position at the extremes (the 1 and 7 positions) actually shrinks over time; what increases is the fraction of respondents who are at the somewhat moderate positions (2–3 or 5–6), though the absolute center (the scale midpoint) and “don’t know” responses decline. Further, the two items also become more strongly correlated over time (refer to the discussion on p. 101). Campbell interprets this as polarization, but I would claim this is more properly seen as ideological consistency. Polarization is, at its core, about movement to the ideological extremes, but the extremes actually shrink in his data (a pattern seen more generally, as Morris Fiorina, Samuel Abrams, and Jeremy Pope document in their Journal of Politics piece, “Polarization in the American Public: Misconceptions and Misreadings,” 70 [2]: 556–60, 2008).

Showing that issue positions become more strongly related to one another is more about consistency, which is important, and related to polarization, albeit conceptually distinct. At some level, one could claim that this is just a semantic difference in one’s preferred terminology, and at some level that is true. I prefer to reserve the term polarization for movement toward the ideological extremes, as this maintains needed conceptual clarity. Part of the reason why there is so much disagreement in this literature is that authors often speak past one another without clear references to what core concepts mean. Quantities like consistency or constraint may well play a role in making politics more combative and disagreeable, but they are not fundamentally about attitudes tending toward the extremes of the ideological spectrum—they are about those attitudes becoming more interrelated. To be clear, both polarization and increased consistency are important, consequential, and have changed our politics in fundamental ways. My claim is simply that they are distinct enough to make it worthwhile to differentiate them conceptually and empirically.

Campbell also argues that this mass polarization occurred first, and that shift in turn generated elite polarization: the political upheaval of the 1960s divided the mass parties, which in turn drove elites apart. By his own admission, this evidence is more circumstantial, and I appreciate his care in noting the limits of this analysis. That said, I am not ultimately convinced that the evidence demonstrates that polarization was driven by voters, rather than by elites. The evidence for a shift in voter preferences comes from an analysis of electoral returns (and to a lesser extent, changes in partisanship). But as Morris Fiorina, Samuel Abrams, and Jeremy Pope argue in their 2005 book Culture War? The Myth of a Polarized America, one cannot draw conclusions about changes in voters’ preferences from electoral data, because it conflates choices and preferences. As frustrating as it may be, the data we have simply cannot give us a clear answer on this point.

But further, to the extent that we do have data from this period, they suggest that the elites led and the mass public followed. Take, for example, attitudes toward the Vietnam War. As John Zaller demonstrates in his 1992 classic, The Nature and Origins of Mass Opinion, elites in Congress divided in the mid-1960s, then the voters followed in subsequent years. Further, as Greg Adams shows, if the masses in the 1970s would have led on abortion, the Democrats would have become the pro-life party and Republicans the pro-choice one (see his “Abortion: Evidence of Issue Evolution,” American Journal of Political Science 41 [3]: 718–37, 1997; James Stimson makes a similar point in his work as well). While there are undoubtedly important exceptions to this general pattern, absent some additional evidence it is hard to sustain the claim that voters changed before elites. To be clear, my claim is not that voters follow elites blindly wherever they lead. Rather, it is that voters and elites dynamically respond to one another, and demonstrating this empirically is extremely difficult, as Nolan McCarty, Keith Poole, and Howard Rosenthal discuss in their 2006 book Polarized America: The Dance of Ideology and Unequal Riches.

My differences of interpretation aside, there is much to like and recommend in this book. In particular, Campbell’s care in parsing a wide variety of public opinion data, as well as a range of historical evidence on polarization over the past 50 years, makes the book well worth reading. Anyone working on polarization, public opinion, or political parties will want to read this book, and it will no doubt inspire important work well into the future.