As a regular reader of The Monkey Cage, I am delighted to help launch what I hope will be a vibrant series of discussions of new and old books by political scientists. And as the author of _Unequal Democracy_, I am very grateful to John Sides and Henry Farrell for recruiting such a smart and stimulating set of commentators to share some of their thoughts about my book. I will attempt here to respond to just a few of their points, trusting that they and others will continue the conversation.
Lane Kenworthy seems to be working his way chapter-by-chapter through Unequal Democracy, and by the time he gets to the end the book will be much improved. Having produced the “best critique”:http://www.u.arizona.edu/~lkenwor/challenge2010-presidentsinequality.pdf I have seen of the analysis of partisan income effects in chapter 2, Kenworthy turns here to class politics (chapter 3) and myopic retrospection (chapter 4). He asks why support for Democratic presidential candidates among low-income white voters was “never particularly high,” and answers “that voters are myopic, and Democratic presidents have been less likely than Republican ones to produce healthy income growth in election years.”
Kenworthy “extends this argument”:https://themonkeycage.local/2010/12/why_dont_low-income_whites_lov.html by suggesting (1) that shifts in party identification may also reflect differential income growth, but with less myopia than we observe in voting behavior, and (2) that the distinctive partisan politics of the post-war South may reflect distinctive regional patterns of differential income growth. I am sympathetic to both suggestions, and intrigued by the preliminary evidence he has adduced in support of them.
In much the same spirit, Chris Achen and I have attempted to reinterpret the New Deal realignment as largely a matter of myopic retrospection rather than ideological conversion (see “here”:http://www.princeton.edu/~bartels/depress3.pdf). FDR’s electoral gains in 1936 came mostly in states where income growth in 1936 (but not in 1934 or 1935) was especially robust–not in states that otherwise displayed any particular affinity for progressive economic policies. Voters in a variety of other democracies hard hit by the Great Depression demonstrated an impressive willingness to replace “ins” of every ideological complexion with an equally varied cast of “outs,” including conservatives (in Britain and Australia), nationalists (in Ireland), Nazis (in Germany), social democrats (in Sweden), socialists (in the Canadian prairie province of Saskatchewan), and followers of a funny-money radio preacher (in the adjacent prairie province of Alberta).
In every one of these cases, the party that happened to be in power when the economy rebounded went on to dominate politics for a decade or more. Although we lack survey data on party identification from the New Deal era, our analysis of congressional voting provides some evidence that they did so, at least in part, because myopic retrospections got solidified into partisan loyalties. Pooling data from non-southern congressional districts in the elections of 1936, 1938, and 1940, Achen and I estimated that election-year income growth (measured at the state level) had a fairly strong positive effect on Democratic vote shares (.23). Income growth in the previous election year, two years earlier, also had a significant effect (.17). Was that because voters employed a relatively long time horizon in assessing the Democratic Party’s performance? We think not, since income growth in the intervening (odd-numbered) years had a noticeably smaller effect (.11). Our interpretation is that voters made their assessments of party performance at election time, with a good deal of myopia, then updated their party identification accordingly and carried it forward to the next election (see “here”:http://www.princeton.edu/~bartels/myopicretrospectionrealignmentdepression.pdf).
The idea that party identification reflects an accumulation of short-term retrospections is consistent with the Bayesian model of partisanship developed by Achen and others. However, the “saw-tooth” pattern discernible in congressional election results in the New Deal era suggests that the “running tally” of party identification (in Morris Fiorina’s felicitous phrase) may be something more like a limping tally, incorporating economic experience primarily in even-numbered years, when American voters happen to go to the polls.
Of course, aggregated congressional voting data shed only very indirect light on the evolution of partisanship, and our statistical results from the New Deal era are no more than suggestive. Why not examine party identification directly? One way to do this would be to look for evidence of election-related retrospection in aggregated time-series of “macropartisanship.” Another would be to harness more than half a century of high-quality individual-level survey data gathered by the American National Election Studies. I once attempted something like the latter in “a paper”:http://www.princeton.edu/~bartels/learning.pdf focusing on the development of party identification over the life cycle; the results were not pretty, but they may help inspire Kenworthy or someone else to produce a more tractable and comprehensive analysis of partisan updating.
Kenworthy’s interest in class politics adds some additional complexities. Most notably, in the case of the postwar South, his focus on low-income whites leaves unaddressed the partisanship and voting behavior of middle- and high-income whites. Unlike low-income white southerners, more affluent white southerners have abandoned the Democratic Party in droves over the past half-century, producing class polarization in voting behavior well beyond the levels observed in other parts of the country. For their behavior to fit Kenworthy’s story would seem to require that their experience of income growth (or their response to income growth) under Democratic and Republican presidents was quite distinct from that of low-income white southerners, and of middle- and high-income whites in the rest of the country. If that turns out to be the case, then Kenworthy’s account (which he modestly refers to as “the Bartels theory”) will look even more promising.