We are unable to display your institutional affiliation without JavaScript turned on.
Browse Book and Journal Content on Project MUSE

Find using OpenURL

Comments and Discussion

From: Brookings Papers on Economic Activity
Fall 2012
pp. 61-81 | 10.1353/eca.2012.0020

In lieu of an abstract, here is a brief excerpt of the content:

Comment by David Gergen and Michael Zuckerman

With this paper, Jacob Jensen, Ethan Kaplan, Suresh Naidu, and Laurence Wilse-Samson offer a methodologically innovative contribution toward answering a longstanding question: what, exactly, drives the national political conversation? In this comment we summarize their paper and, in a constructive spirit, recommend that future research into this topic be broadened to answer three basic questions:

  • —   What role do the media play in driving the political conversation?

  • —   How do historical changes in the nature of Congress affect its rhetoric (as well as its results), and to what extent is the historical Congressional Record both a consistent source of data and a reliable predictor of polarization and results going forward?

  • —   To what extent are elites (such as members of Congress and published authors) responding to rather than driving the ideology espoused by more popular voices?

Jensen and coauthors probe the words behind the American political conversation, building on the ascendant "text-as-data" approach to create intriguing new measures of polarization and partisanship. They then pair these newly constructed longitudinal measures with preexisting data to seek out "important national political phenomena that correlate in time with the polarization in political discourse" and may be driving its direction. The data powering these ambitious efforts are a set of quantifiably partisan three-word phrases ("trigrams") stretching back to 1873, sifted out by the authors (using, presumably, a lot of computing power) from both the newly digitized Congressional Record archives and the Google Books corpus of the English language.

The paper's first contribution, then, is this new measurement of polarization and partisanship in both congressional and broad "political discourse" stretching back over the past 140 years (and limited only by digital availability of the Congressional Record—the Google corpus extends more than 350 years earlier). The series of data they construct is no small undertaking and a welcome addition to an area of inquiry that has previously been dominated by the DW-NOMINATE score, a measure of polarization drawn from congressional voting patterns and curated most prominently by political scientists Nolan McCarty, Keith Poole, and Howard Rosenthal. Run historically, the authors' new model correlates closely with DW-NOMINATE in the post-1930 era but diverges from it before 1930, finding high polarization toward the end of the 19th century and less in the early 20th century. What it shows—that "recent political polarization may be high relative to the 1970s, but it is a far cry from the open violence of the late 19th century"—offers some useful historical scope and context. It also builds an empirical scaffolding for some anecdotal bits from history that challenge the contention that polarization is worse now than ever before: recall that in 1856 Congressman Preston Brooks of South Carolina caned Massachusetts Senator Charles Sumner to within an inch of his life on the floor of the Senate, or that one debate in the House in the 1800s grew so rancorous that some 30 members drew their guns. Politics, to paraphrase Mr. Dooley, was never beanbag.

From this initial measurement, the authors extend their exploration in an effort to tease out correlations between polarization over time and other, broader factors. Among their more targeted observations are a robust correlation between polarization and political violence; a switch in relative polarization in the direction of the minority party when House control switches (the authors reasonably hypothesize that minority parties may "talk more and use more partisan language in order to slow the enactment of policies they oppose"); and, perhaps most interesting (but "statistically weak," in the authors' words), a lag between the Google Books database and Congress for polarized economic phrases only, suggesting that "although Congress may take its economic language from public intellectuals, it does not adopt its language on social issues from the same sources."

A discussion of the statistical techniques that the authors employ lies beyond our training, but the range of their search—through Gallup polling, military casualty figures, economic data, multiple measures of legislative efficiency, major historical events, and other data—is extensive, and promising for future inquiry. Although the paper deliberately stops short of addressing causality, the authors do report that they see their "evidence as...

You must be logged in through an institution that subscribes to this journal or book to access the full text.


Shibboleth authentication is only available to registered institutions.

Project MUSE

For subscribing associations only.