Chapter 2: The Culture War Mythology: Critics of the Conventional Wisdom

Political polarization has received a great deal of attention from scholars and laypeople alike. The most recent scholarship and the ongoing debates over the nature of the Culture Wars and political polarization provide insight on the theoretical implications and empirical puzzles of political polarization. The materialization of an emergent polarization among scholars is apparent in the recent literature. With scholars on one side of the divide suggesting the Culture Wars thesis is a myth and that the trends in public opinion show the exact opposite: movement towards more moderation, centrism, and tolerance (Fiorina, Abrams, and Pope 2004). Others suggest that polarization in the electorate is real and detectable (Abramowitz and Saunders 2005; Campbell and Cannon 2006; Abramowitz and Jacobson 2006; Jacobson and Edsall 2006).

The implications of a true culture war—a truly polarized public—are apparent. With the center vacated and public opinion coalescing around two distant poles, the potential for compromise would evaporate. With the political system unable to cope with the disposition of policy opinion, other means could be sought to achieve political ends. Yet there has been no eruption of political violence in America. No end to compromise apparent in the near two decades since Hunter made his grave warning. Is it ahead? Furthermore, is it even true that that the American electorate has become more polarized? If not, it bodes well for the continued existence of a peaceful political culture, however contentious at times. Several scholars and observers of politics have challenged the conventional wisdom on political polarization making exactly that case.

Nothing to See Here! Fiorina and the Culture War Myth

Morris Fiorina, a scholar with a penchant for unconventional takes on conventional theories of politics, registered one of the first counter-strikes to the growing consensus on an emerging Culture War and increasing political polarization in his 2005 work: Culture War? The Myth of a Polarized America (Fiorina, Abrams, and Pope 2004). In this small but influential tome (later expanded upon in a second edition), Fiorina makes several arguments regarding the causes and nature of polarization among masses and elites, but his fundamental argument is that the rhetoric regarding the polarization of the American electorate over the last ten to fifteen years is apocryphal. That Hunter and those who have accepted his thesis are wrong. There is no political polarization. “The simple truth is that there is no culture war in the United States—no battle for the soul of America rages…” (Fiorina, Abrams, and Pope 2004). The trend, according to Fiorina, over the last several decades in the American public is one of growing centralization and moderation of the electorate on social issues (Fiorina, Abrams, and Pope 2004). This shot across the culture wars bow has generated much debate over what empirical evidence, if any, underlies political polarization in American politics.

Fiorina argues that the American electorate is largely centrist on the bulk of political issues (including such hot button issues as abortion and homosexual rights), that they remain relatively ambivalent on the policies related to these issues, and that the trends in public opinion in the American electorate reflect increasing tolerance rather than divergence and conflict (Fiorina, Abrams, and Pope 2004) . Fiorina points to relatively small shifts in the average and moderate opinions on abortion—the sine qua non of salient social issues—in the electorate over the last few decades as evidence that the American public’s social views are stable and centrist. The electorate over the past thirty to forty years has been relatively stable in its distribution of political opinion. “There is little evidence that American’s ideological or policy positions are more polarized today than they were two or three decades ago, although their choices often seem to be” (Fiorina, Abrams, and Pope 2004). Thus the American mass electorate is largely centrist on a range of issues (including hot-buttons such as abortion and homosexual rights) and not sharply divided as suggested by the polarization thesis.

So, why all the fuss about a Culture War? The myth of politicial polarization along cultural lines is a product of elite activists promoting the perception of a divided nation (Stephonopolus’s polarization entrepreneurs) and bad interpretation of the available data by a credulous media establishment. “The myth of a culture war rests on misinterpretation of election returns, lack of hard examination of polling data, systematic and self-serving misrepresentations by issue activists, and selective coverage by an uncritical media more concerned with news value than with getting the story right” (Fiorina, Abrams, and Pope 2004). The Fiorina posits that the only true polarization observable over this time span is among the elites in the two political parties, whom have polarized absent or despite of any signals from the voters.

One of Fiorina’s primary theoretical arguments is that partisan polarization is not the same thing as ideological or issue polarization. He argues that “sorting” is distinct from “popular” polarization. Fiorina argues that the realignment of the South has created a more conservative Republican party and a more liberal Democratic party, but the ideological disposition of the mass public has remained stable.

Partisan sorting can occur without any substantial shift in the issue opinions and policy positions of the electorate as a whole (Fiorina, Abrams, and Pope 2004). In 2004, 21 percent of voters described themselves as liberals, 34 percent as conservatives, and 45 percent were self-described moderates. These numbers are nearly the same as the average of the ideological disposition of the electorate found in the exit polls over the last 30 years (Galston and Kamarak 2005). So while the parties have become more ideologically consistent at the mass level, the distribution of conservatives and liberals and the relative position on cultural issues of the mass public has not changed.

To support his argument, Fiorina looks at two big, defining cultural issues: abortion and homosexuality. On abortion, as noted previously, he finds “remarkable” stability over the time series. Furthermore, demographic groups (e.g. gender) either do not differ in their average positions or their differences aren’t as significant as expected (religious vs. the non-religious). On homosexuality, Fiorina identifies a strong liberalizing trend in the American public. Far from a “major front” in the culture war, he posits that homosexuality is a decreasingly relevant issue dimension, since the American public is converging on a more tolerant view of homosexuals and homosexual rights. This increase in tolerance of homosexuals, from Fiorina’s perspective, is strong evidence against the Culture Wars thesis (Fiorina, Abrams, and Pope 2004).

He concludes his analysis by arguing that class remains a strong predictor / correlate of partisan and political beliefs in the American electorate, and that, while religiosity (as opposed to religious denomination) has become an increasingly relevant influence on partisan and political beliefs, income divisions are even more important than they were decades previous. “On the contrary, comprehensive research indicates that income divisions are more important now than in earlier decades” (Fiorina, Abrams, and Pope 2004). He further argues that the only polarization that has occurred is at the elite level, driven by a “hijacking” of American politics by interest groups as a consequence of the funding they provide politicians and the parties.

One important contribution of Fiorina’s analysis is his debunking of the popular notion that close elections constitute evidence of political polarization and a “great divide” between the American populace. A close election can result just as easily from a centrist, unimodal electoral distribution of political beliefs that inform electoral choices as it does from a bifurcated and sharply divided distribution of political beliefs in the electorate. Likewise, significant polarization in society can exist despite the fact that an election is a landslide. “An election may be closely divided without being polarized, as it was in 1960, or deeply polarized without being closely divided, as it was in 1936, or neither, as seems to have been the case in the famous ‘Era of Good Feeling’ between the war of 1812 and Andrew Jackson’s arrival on the presidential stage” (Galston and Nivola 2006). The only significant evidence of political polarization that Fiorina acknowledges has occurred at the elite level. This has happened despite the stable views of the mass electorate. Indeed, Fiorina suggests that the problem facing American democracy is not a war between sharply divided groups in the electorate but rather a run-away cadre of elites pushing policy outside the mainstream of mass opinion (Fiorina, Abrams, and Pope 2004; Fiorina and Levendusky 2006). As Mayhew noted, there is clear and rational connection between elected officials and their constituents, whether or not this reelection motivation is single-minded (Mayhew 1974).

Fiorina argues very little mass polarization has occurred. In Culture War? he argues that Americans are moderate, tolerant, and relatively ambivalent towards politics and the divisive ideological battles that are so intense at the elite and activist level. “Americans are closely divided, but we are not deeply divided, and we are closely divided because many of us are ambivalent and uncertain, and consequently reluctant to make firm commitments to parties, politicians, or policies. We divide evenly in elections or sit them out entirely because we instinctively seek the center while the parties and candidates hang out on the extremes” (Fiorina, Abrams, and Pope 2004). Fiorina suggests that most of the American electorate reflects the predominance of cross-cutting ideological beliefs, with individuals holding a mixture of conservative and liberal views, depending on the issue at hand. Though sorting has occurred at the mass and elite levels, the extent to which they have sorted is as night is to day. Partisan polarization is mostly an elite phenomenon, accordingly. He suggests that geographical polarization is overstated and that social cleavages have become more 'fuzzy' rather than stark dividing lines between intractably disagreeable groups. Economics continues to drive the partisan wheel of politics and has not been displaced by religiosity. In his previous work he suggested partisan gerrymandering may explain increasing congressional polarization, though he has abandoned that argument in more recent takes on the subject (Fiorina, Abrams, and Pope 2004).

The geographic tale of a polarized public isn’t without its skeptics beyond Fiorina, even in the popular accounts of polarized politics. Robert Samuelson objects to the depiction of a deeply divided Red vs. Blue political electorate (Samuelson 2004). He argues, as others scholars have argued, that opinion polling shows an increasing consensus on cultural issues such as racial and gender equality, traditionally social hot buttons. He takes on the argument that underlies the partisan map that some observers , such as E.J. Dionne, focus on: “The divide in American politics is about more than the ideological distance between the two parties. Right now, Red Staters and Blue Staters live in two different political universes” (Dionne Jr. 2006). Samuelson counters by arguing that the technological revolution (the internet and the information revolution) has served to make distant lands in the United States more culturally homogenous (i.e. the Wal-Mart effect) rather than reflecting the growing political dichotomy between “Blue” and “Red” consistent with the Culture Wars thesis. “Texas and New York have more in common now than in 1950 or 1960,” he notes, though for some reason fails to cite the prevalence of Olive Gardens in his litany of similarities (Samuelson 2004). Samuelson asserts that the color-coded partisan map masks fairly close presidential elections in those states. In 2000, the election that birthed the Red vs. Blue political maps, nineteen states were won with 51 percent of the vote or less. Hardly the picture of geographically homogenous states bifurcated by a deep political division (Samuelson 2004).

Dude, Where’s my Polarization? The DiMaggio Study on Political Polarization

DiMaggio and his colleagues produced the most comprehensive time-series on political polarization to date in their 1996 article and subsequent update (DiMaggio, Evans, and Bryson 1996; Evans 2003). Assessing polarization across various social issue measures in the ANES and GSS, they examine trends in the variance and kurtosis of these opinion indexes over time. Consistent with Fiorina’s argument, they found little evidence of increased polarization over the period of 1972 through 1994 in both the GSS and NES, excepting abortion. The distance between various groups (age, education, sex, race, region, religion, etc.) on racial and gender issues, crime, sexual morality, and the role of the welfare state have narrowed over their time series (DiMaggio, Evans, and Bryson 1996). “We find no support for the proposition that the United States has experienced dramatic polarization in public opinion on social issues since the 1970’s” (DiMaggio, Evans, and Bryson 1996). DiMaggio and his colleagues formulate four theoretical conceptualizations of polarization: consolidation, constraint, bimodality, and dispersion. Their empirical measure of dispersion is variance:

Equation 2.1: Dispersion

This is the “No one I know voted for Nixon” phenomenon or the “separate camps” aspect of polarization as described by DiMaggio et al. The more that the distribution of opinion on an issue reflects two relatively distant poles with a vacated center, the more intractable political competition becomes. In other words, a tendency towards extremism and a resistance to compromise is inherent to the formation of separate camps, as similarly situated individuals all view the issue relatively similarly and those who differ with them on the question are that much more distant from their views on the subject. DiMaggio uses kurtosis to measure bimodality. Though it should be noted that objections regarding the extent to which kurtosis reflects bimodality have been raised in the statistics literature. Most such objections refer to relatively distinct distributions (such as two modes relatively close to one another, i.e. the gamma distribution), though there are other issues as well, leading some scholars to dispute kurtosis as a measure of bimodality (DiMaggio, Evans, and Bryson 1996; Mouw and Sobel 2001).

Equation 2.2 Kurtosis

Kurtosis is, in theory, positive when a distribution is peaked (large percentage of responses located at one point) irrespective of whether the peak is located in the center or elsewhere in the distribution. As a distribution flattens, kurtosis becomes more negative. As responses vacate the center in favor of the extremes, kurtosis becomes further negative. The constant “3” is subtracted from the kurtosis equation so that the normal distribution is roughly equivalent to zero. Kurtosis is distinct from skewness in that skewness provides a measure of the direction a distribution is lopsided (one way or the other) while kurtosis distinguishes between a dimension heavily weighted towards one side of the distribution and where a distribution is characterized by significant proportions of the responses on both sides of the center (Mouw and Sobel 2001; DiMaggio, Evans, and Bryson 1996)

In their analysis of polarization the singular exception noted to their ‘absence of polarization’ finding is that of abortion. “If attitude polarization entails increased variance, increased bimodality, and increased opinion constraint, then only attitudes toward abortion have become more polarized in the past 20 years, both in the public at large and within most subgroups” (DiMaggio, Evans, and Bryson 1996). Yet even that finding is suspect according to some scholars (Mouw and Sobel 2001; Shaw 2003). Mouw and Sobel argue even on abortion the American public has failed to polarize. Though Dimaggio and his colleagues maintain that polarization on abortion is a real phenomenon (Evans, Bryson, and DiMaggio 2001).

Wagging the Dog: Elites hold the Median Voter’s Leash and do all the Barking

In his most recent take on the subject, Fiorina argues that sorting has markedly increased at the elite level, while very little sorting has occurred among the masses (see Table 2.1 for illustration of sorting). Abramowitz counters arguing that the masses, in some respects, have outpaced elites in partisan sorting. He finds that polarization in Congress has increased only half as much as the partisan polarization evident among voters (Abramowitz and Jacobson 2006). Fiorina explains the elite sorting disconnection from mass sorting as a function of the primary system (rather than gerrymandering), where candidates preemptively move to the relative extremes in anticipation of challenges from the ideological poles. “Because sorting produces a more homogenous and a more extreme primary electorate, the pressure increases for candidates to take consistently liberal or conservative positions on

Table 2.1: Partisan “Sorting” with No Ideological “Polarization”

IDEOLOGY	TIME PERIOD 1	TIME PERIOD 2
Liberals	65 Democrats, 35 Republicans	95 Democrats, 5 Republicans
Moderates	100 Independents	100 Independents
Conservatives	65 Republicans, 35 Democrats	5 Democrats, 95 Republicans

Source: Adapted from Fiorina & Levendusky, 2006
most issues, even when moderation would be more helpful in the general election. Thus sorted partisans move candidates toward non-centrist positions“ (Fiorina and Levendusky 2006). All well and good for the primaries, but that doesn’t really explain elite polarization outside of the primary context. Running to the center in a general election is an ancient political axiom, and primary challenges to sitting partisans are as rare as a blue moon in congressional races. So why are elites diverging if the mass public is increasingly moderate and centrist? The classic Downsian model illustrates the centripetal logic of party competition in an electoral setting with rational voters: the median voter theorem (MVT). This simple yet powerful insight provided a logic for the classic observation of candidates moving towards their parties extremes in primaries (or moving towards the median primary voter) only then to run to the center in general elections (move back towards the median voter in the electorate. DiMaggio et al. in their study of polarization argue explicitly that political polarization is implicated in median voter models. “Once the median is no longer the mode, majorities may form around either mode. If bimodality is great and the issue salient, rational candidates may embrace extreme positions in order to prevent a sit-out by purists or a third-party challenge” (DiMaggio, Evans, and Bryson 1996).

Yet in order for distributional changes to be relevant in median voter models, a step away from the stylized model developed by Downs is necessary. Downs argued that the location of equilibria in an election would be dependent on the shape of the distribution of citizen preferences (Downs 1957; Hotelling 1929). Downs was wrong (see Figure 2.1). The pure MVT with complete turnout and sincere voting predicts that the median voter is, in fact, a Condorcet winner: no position defeats the ideal point of the median voter in a pair-wise vote irrespective of distributional qualities (Black 1958). Thus, parties (or candidates) must move towards the center to maximize votes. However, there are countless elections where parties and candidates have failed to converge to a single policy point or even a vector of policy points in a continuum of policies. While 1950’s political competition tended to conform to moderate and centrist expectations, current rhetoric and serious analysis alike decry the stark choice (as opposed to an echo) that the parties present today and the ratcheted conflict that accompanies it. Many analysts believe that diverging citizen preferences must be related to diverging candidate and party positions, as the realignment literature implicitly assumes. Either this is the consequence of a more polarized population, a more polarized electorate, or a nomination process that generates polarized candidates, i.e. the primaries effect (Fiorina 1999).

Fiorina provides theoretical justification for his argument on polarization through the use of a formal model of voter response to elite polarization (Fiorina, Abrams et al. 2005). It employs a Downsian proximity model of the vote decision with which he attempts to explain how the cultural dimension could become increasingly important while the voter positions remain stable (Platt, Poole, and Rosenthal 1992). As can be seen in Figure 2.1, Fiorina argues that elite polarization yields divergent candidates and parties. This thus elevates the importance of a cultural or social dimension despite the fact that voter positions on this dimension remain the same. This is a key point in Fiorina’s arguments on the behavior of the elites relative to the masses. He argues that political and partisan elites, unanchored by the pragmatic and materially-oriented party machines of the past, have become beholden to extremist advocacy groups. As such they have polarized irrespective of the growing ideological centrism of the American electorate. The bi-modal distribution of polarized voters (and thus responsive political
Figure 2.1: Polarized Elites? Are Elites Located Distant from Median Voter and the Mass of Voters?


If the distribution of voters is normal and centrist, then the party elites have moved away from voters



If the distribution of voters is bimodal, then the party elites are located with their respective constituents



Republican Party Elites

Democratic Party Elites

Fiorina’s Normal Distribution of voter ideal points

Bi-Modal Distribution of voter ideal points

Median Voter Position
_XMED

POLICY

Recall Thomas Frank’s arguments that the lower class has been duped into voting for Republicans on the cultural issues. But is Frank right? Has class-based voting declined? Has it been replaced by a now dominating social/cultural cleavage? Morton says no, arguing the stylized fact of an electorate, and particularly a lower class electorate, more concerned with social and cultural issues than economic issues is wrong. Citing recent election data from the 2004 election, Morton points out that lower-class values voters are much less likely to support Republicans (30%) than upper-class values voters (60%), contrary to Frank and the Culture Wars thesis (Morton 2006). Stonecash’s Class and Party in American Politics presents a substantial time series examining determinants of partisanship since the 1950’s and concludes that Republican gains have come among middle and upper-class voters, thus widening the traditional gap between the c lasses in party support (Stonecash 2000). This trend is also evident in Nadeau and Stanley’s research on partisan trends among Southern whites (Nadeau and Stanley 1993). McCarty, Poole and Rosenthal present a similar finding in an analysis of class, partisanship, and presidential voting, showing class as an increasingly predictive favor—subject to fundamental economic and social changes—in electoral choice (McCarty, Poole, and Rosenthal 2006). Stonecash and his coauthors conclude Democratic competitiveness in the post-New Deal realignment was rooted in their increasing appeal to working class voters outside of the South (Stonecash et al. 2000). In a later work focusing on congressional districts and partisan polarization, Stonecash shows that Democratic candidates increasingly win in districts with lower median incomes, higher urbanization, and a relatively high percentage of non-whites (Stonecash, Brewer, and Mariani 2003; Stonecash 2005).

Larry Bartels, in his vivisection-like review of Frank’s work, argues that, despite bad analysis of the 2004 exit polls and Frank’s “stew of memoir, journalism and essay,” the theory that Republican success is a consequence of a growing cohort of working class conservatives is a myth (Bartels 2006). According to Bartels, the bottom third of the income distribution among working whites have become more reliably Democratic in presidential elections. He finds surprisingly little class differences in Democratic support from 1952 to 1974. However, from 1974 to 2004 he finds an exacerbation of the differences between the upper and lower thirds of the income distribution. The gap between lower-class whites and upper-class whites increases from 4% to 14% between the two periods (Bartels 2006). While he does find a decline in white working class party identification this is offset by an even greater decline among upper income voters. His analysis of issue trends in the National Election Studies suggests that working class whites have not become conservative over the last 30 years and that, indeed, the working class trends towards moderation in comparison to the more extreme social views of the upper class (Bartels 2005).

In an update of the review, Bartels addresses Frank’s rejoinder to his review. Frank suggests that the original thesis of What’s the Matter with Kansas was that the uneducated, not the lower income voters that Bartels examined, make up the working class. Bartels’ rebuttal identifies a particularly puzzling trend that tends to puncture a well-established conclusion regarding the partisan divisions of the 1930’s and the 1940’s. The traditional story of the New Deal realignment was that it was strongly grounded in class conflict, with the poor solidly in the Democratic camp aligned against greedy and rich Republicans. Yet Bartels finds that educational differences were more strongly associated with the partisan divisions in the 1950’s and the 1960’s. Contrary to Frank’s revision of the “What’s the Matter” thesis that suggested the uneducated have trended Republican, Bartels’ analysis suggests that education has vanished as a relevant partisan cleavage since the 1970’s. Income disparities in party affiliation and partisan political choice have escalated considerably at the expense of educational differences (Bartels 2006). He again examines issue trends, but assesses here the extent to which education differences explain these trends. Once again, Bartels argues Frank’s take on the macro-trends in partisanship cannot stand up to scrutiny. The trend in educational effects on partisanship has actually declined over the last few decades.

One exception to Bartels’ generally damning case against Frank is on the abortion issue. Bartels finds a strong and significant positive coefficient for the electoral significance of abortion among white voters without college degrees. Pro-life voters have been pulled towards the Republican Party over the last twenty years. Before Frank dances a jig, however, Bartels also finds that this effect has been more significant among whites with college degrees, suggesting that the educational attenuation of vote choice, to the extent one exists, is not operative in the direction Franks argues. Furthermore, since 1996, among whites without college degrees abortion has declined in importance while it has increased in importance for whites with college degrees (Bartels 2006). What’s more, the newer analysis conforms with Bartels’ earlier review in that economic issues continue to be relevant in structuring the electorate’s political choices and their impact is slightly larger among the white working class voters (Bartels 2006). However, all this must be set against the backdrop of a significant trend upward in the number of college educated and bachelorette-earning adults in the United States over the past 50 years. High-school graduation rates have increased from around 30% to nearly 90% of the 25-and-older population. The college degree-earning population has gone from well under 10% to just-under 30% of the population. Hence, it may be the case that the apparent trends Bartels identifies among the college and non-college educated sub-populations may be a consequence of the significant changes in composition those groups have undergone in the last 50 years. The effect of education on voting itself may not have changed while the composition of the group of ‘educated’ voters has, possibly explaining the puzzle.

Bartels’ lower income = support-for-Democrats finding presents a vexing puzzle regarding the observed empirical regularities in politics. On the one hand, it appears that lower class voters remain solidly in the Democratic camp and perhaps have increased their support for the party of Jefferson in the last 20 or 30 years. However, at the state level of analysis, states with lower per capita income and gross state product have trended on to the Republican side of the ledger. The reliably Blue states, on the other hand, are also the wealthiest states in the Union. As can be seen in Table 2.2, the top states in per capita gross state product (GSP) and median household income (MHI) are reliable Blue states while the bottom states are just as reliably Red. The bottom ten states in MHI contain only one blue state, New Mexico. Likewise, only one red state (Virginia) cracks the top ten in MHI. Blue states overall have an average MHI of $55,000 while Red states trail well behind with an MHI of just above $47,000. The per capita GSP differential isn’t as large, but the blue states still have a sizable advantage. For the red states the average GSP is $34,000, but the blue states have an average per capita GSP just under $40,000. So, evidence at the individual level suggests that lower-income voters continue to throw in their lot with the Democratic Party, that they have become increasingly Democratic over the last few decades, and that economic issues are a significant factor in their party identification and electoral choice. On the other hand, at the state-level, rich states overwhelmingly go Democratic in presidential elections while poor states are solidly Republican. One answer is that this is a simple manifestation of the ecological fallacy—generalizing from the aggregate to the individual. Hence political observers such as David Brooks can project “the liberal stereotype onto the map” and “admit on behalf of everyone who lives in a blue zone that they are all snobs, toffs, wusses…and utterly out of touch with the authentic life of the people” (Frank 2004). As Brooks puts colorfully, “We’re [Blue Staters] more sophisticated and cosmopolitan…But don’t ask us, please, what life in Red America is like. We don’t know…We don’t know what James Dobson says on his radio program…We don’t know about Reba and Travis…Very few of us know what goes on in Branson, Missouri, even though it has seven million visitors a year, or could name even five NASCAR drivers…We don’t know how to shoot or clean a rifle” (Brooks 2001). Polarization is thus a mirage: a feverish, imagined divergence that is the product of mapping our own political biases on to the rest of America.

Morton suggests that the ecological fallacy is the answer, pointing out that Fisher’s Paradox can disguise radically different individual support for Republicans among individual low and high income

Table 2.2: Income Polarization at the State Level by Presidential Voting 1996-2004*

voters that wouldn’t show up at the aggregate level (Morton 2006). However, noting that there can be differences between individual characteristics and aggregate characteristics doesn’t rule out the possibility that the aggregate results are a real political phenomena rather than a statistical artifact. Gelman and his coauthors attempt to tackle this apparently intractable problem through a multi-level modeling of the aggregate and individual effects on partisanship in yet another “What’s the Matter” article, this time with Connecticut as the foil (Gelman et al. 2005). Gelman et al. examine four separate levels of aggregation: the individual voter, the county, the state, and the national level. As Gelman and his colleagues note, given institutional aspects of the system such as electoral rules, parties, and geographic-based representation, elections are not “simple cumulations” of voter decisions and thus aggregate analysis can reveal real phenomena missed by a study of just individuals (Gelman et al. 2005). They find that the income and voting differences between Red and Blue states are real, despite the fact that richer voters within states tend to support Republicans. Their resolution of the apparent paradox is that within each state, income is positively correlated with Republican vote choice, but average income varies by each state. Thus individual income (positive) and state average income (negative) are independent predictors of presidential vote support.

The puzzle of a rich state such as Connecticut consistently supporting Democratic candidates is revealed to be not much of a puzzle at all. Rich voters in Connecticut are not all limousine liberals. Richer voters in Blue states tend to support Republicans just as they do in the Red states. This is attenuated by the effect of the higher median income of the state such that the net effect of income is more or less a wash. Gelman argues this debunks Frank and Brook’s characterization of “latte” Democrats and “NASCAR” Republicans while, contra Bartels and Fiorina, suggests the Blue/Red dichotomy is a real political and cultural phenomenon (Gelman et al. 2005). Gelman provides little by way of explanation, theoretical or otherwise, for this pattern but contends the pattern is real.

All this said, that income is still a relevant partisan cleavage and Republicans remain the party that appeals to upper-income voters with Democrats favored by the working class does not mean that social issues have failed to emerge as an important cleavage in partisan politics or that it is irrelevant to the secular realignment of the political system. Employing a soft conceptualization of realignment can allow for the fact that income still matters—still structures the political cleavage of the parties—while at the same time changes along other issue dimensions have resulted in significant shifts in the political coalitions of the parties that have had a lasting impact on the electoral fortunes of both parties. As noted earlier, the new and the old can persist through a secular realignment.

The argument over polarization wasn’t ended by Fiorina’s publication of The Culture Wars Myth? Each of his main points remain a source of division among scholars. Abramowitz and Saunders, in a direct response to Fiorina’s myth thesis, find deep divisions in America between Democrats and Republicans, red state voters and blue state voters, as well as religious voters and secular voters within the mass electorate. Polarization is not confined to a small cadre of governmental officials and activists but rather is a phenomena of the general public and one reflective of an increasing trend in American politics (Abramowitz and Saunders 2005). Abramowitz and Saunders take issue with Fiorina’s argument that there has been no increase in opinion polarization in the American public and that it is exclusively an elite phenomenon. Their analysis examines issue opinions from the 2004 election (ANES) in arguing that while most of the American public is moderate, “there are sharp divisions between supporters of the two major parties that extend far beyond a narrow sliver of elected officials and activists” (Abramowitz and Saunders 2005). They further suggest that Red and Blue state voters differ “fairly dramatically” in their social characteristics and political beliefs. This is consistent with Gelman et al.’s finding that there are apparently distinct cultures that inform the effect that aggregate and individual income (or class) levels have on partisan behavior and political beliefs (Gelman et al. 2005; Abramowitz and Saunders 2005).

Abramowitz addresses the Fiorina argument suggesting weak sorting of the mass electorate by pointing out that, in fact, the American electorate is more educated and the educational process has resulted in an electorate more capable of understanding and using ideological concepts. This combined with elite polarization, providing a clearer set of ideological cues for the public to react to, has culminated in a more engaged, partisan, and polarized public. This segment of the public is much larger than Fiorina and his colleagues define it. “In 2004, active citizens made up 46 percent of all Democratic identifiers and 49 percent of all Republican identifiers” (Abramowitz and Jacobson 2006). Only nonvoters, according to Abramowitz, are consistent with the characteristics Fiorina attributes to the mass electorate as a whole. Fiorina and Levendusky criticize Abramowitz’s polarization finding in suggesting it is an artifact of how he has coded the NES ideological scale (Abramowitz uses a recoded trichotomous ideology variable). Greater political activism is a consequence of greater party mobilization, ideological thinking is not on the rise in the American public, and, of course, “sorting” is not political polarization (Nivola and Brady 2006). Abramowitz suggests that changes in trends overtime cannot be a consequence of recoding, since the same code is used in all years (Nivola and Brady 2006). These points of contention remain unresolved between these researchers, suggesting further analysis is necessary to bridge their scholarly divide.

Campbell and Cannon argue that not only are the masses polarized today, they were polarized in the 1950’s when heterogenous parties and centrist policies were the norm (Campbell and Cannon 2006). But it is not on culture, as suggested by the culture warriors. Campbell and Cannon argue that “…it should be remembered that for much of the twentieth century some of the most intense polarization in nations around the world centered on the economic and political philosophy of Marxism” (Campbell and Cannon 2006). Campbell argues there is direct evidence of mass polarization as evidenced in the increase in polarization on the ideological measures in the NES. He finds a decline in the “nonideologicals” in the mass electorate (Campbell and Cannon 2006).

The Culture War skeptics present a strong case against the emergent polarization conventional wisdom. Fiorina points out that close elections do not connote a polarized electorate. Distinguishing political polarization from partisan sorting, he argues that the culture war is a myth. The American electorate has become increasingly centrist and moderate on social issues. DiMaggio et. al. fail to identify a polarization trend across a variety of cultural indicators, with the sole exception of abortion. And even there, Mouw & Sobel suggest that the American electorate has not polarized on abortion. Bartels and Morton punch a hole in the hull of Frank’s culture war argument. Frank’s contention that economic and class divisions have receded in favor of social concerns takes on water in Bartels analysis, suggesting class remains and is in fact an increasingly predominant factor in partisan affiliation. Whether these arguments have sunk the culture wars thesis remains, however, an open and hotly contested question.

“America is a divided nation. This is an indisputable fact. We are split on the war, split on abortion, split on the unions, split on what the Constitution really means and split on the role of religion in this nation.”

- Cherry 2007

Having examined the evidence and arguments of the conventional wisdom and its critics, the task here is to assess where the polarization literature has gone wrong. There are problems with both the conventional consensus on the culture wars as well as the criticisms of that thesis. The research to date on polarization has poorly developed the concept of polarization, employed invalid and inaccurate measures of polarization, and failed to draw appropriate conclusions regarding polarization from the available data. A big part of this problem is the lack of consensus on what polarization is. Thus, the second task is to develop a rigorous conceptualization of political polarization: to determine what polarization is and what it is not. Third, I will develop empirical measures of polarization that will serve as the basis for the analysis in the subsequent chapters.

We have seen that there is evidence suggesting geographic distinctions between states and regions are indicative of territorial-based political cultures that have a significant part to play in the tale of political polarization. However, looking at political polarization solely from a “Red vs. Blue” perspective misses a great deal of the empirical story. Indeed, the Red vs. Blue dichotomy maps poorly on to the true political changes that have wrought a more polarized political environment, requiring scholars to move past this false choice and examine the true cultural, social, and political fissures that determine partisan politics as well as mass and elite political behavior.

The culture war proponents give us the argument that America has become more geographically polarized along cultural lines. On the other hand, there are those that suggest American culture has moved away from distinctiveness towards a “standardization” that encourages similar perceptions of cultural issues throughout the country (Bowles and Pagano 2003). Homogeneity, in this instance, operates globally. We’ve all become more alike, period. This ‘Americanization” of America is typified by chains such as Wal-Mart and Olive Garden where “the discreet hint of Tuscan decor and the passable wine list disguise the fact that there are 476 other Olive Gardens across North America, all with precisely the same menu”(Engel 2002). But is veal what we have in mind when we talk about cultural issues? The extent to which “Chain-America” and the influence of national popular culture produces cultural homogeneity ,or if it does, remains a controversial topic (Jia 2007). Less esoterically, there is the argument that most states are “purple” rather than red or blue. Furthermore, there is a great deal of partisan diversity even in the deeply red and blue states in the Union. Oklahoma, Kansas, and North Carolina have Democratic governors, while California, and until very recently New York and Massachusetts , has a Republican governor (Galston and Nivola 2006).

Then again, New England just defeated its last Republican hold out, Chris Shays, in the November, 2008 election. Georgia (as is noted later once a bastion of Democratic power) has become one of the more consistently Republican states in America. The potentially confounding geographic impact on cultural and political views presents a problem to the study of political polarization. Important political polarization of the electorate may be masked in purely national-oriented studies, such as Fiorina’s examination of trends in views on abortion and gay rights in the American electorate (Fiorina, Abrams, and Pope 2004). The contrasting trends of the homogenization of national culture versus and increasing heterogeneity between geographic regions and the cultures that manifest in those regions necessitates accounting for increasing or decreasing regional differences on social and other political issues. For example, it is possible for both trends (a trend towards homogeneity on the national level coupled with increasingly disparate views when the electorate is broken down in regional subcategories or reference groups) to be true. Any analysis of political polarization must account for trends in the body politic as well as changes in the grouping of that electorate that could yield polarizing

Figure 3.1: Red, Blue & Purple – State Coded Composition, by Party, of the 110th United States Senate

http://en.wikipedia.org/wiki/File:110th_US_Congress_Senate3.PNG

factors independent of an overall centralizing trend. The American electorate as a whole could become more centrist on gay rights and abortion, while at the same time the reference groups and geographical ties that influence political opinion produce polarization in views between these groups and extreme views within them. So, while it is true that “Red vs. Blue” is an overly simplified and inaccurate measure of geographic polarization, this fact does not mean that geographic polarization hasn’t happened. The key is to sync the measures of geographic boundaries with the political cultures they serve as a proxy for.

The 50:50 Nation & Polarizing “Close” Elections: Being Right for the Wrong Reasons.

Journalistic treatments of the Culture Wars thesis have mostly expressed agreement with it, contributing to a perception of polarized politics irrespective of its empirical reality. Their concurrence is based primarily on the ‘evidence’ of closely contested elections and in examining the culture war from an ‘up-close-and-personal’ standpoint. These stories are characterized by personal interviews of ideological characters and biopics on cities that illustrate the stark differences between conservatives and liberals and Republicans and Democrats—culturally, geographically, and on policy. “The polarization in US politics has both spawned and is being fed by a huge cultural rift that extends from the airwaves to the altar and from the bookshelves to the bedroom. Alongside concrete issues like the economy, the war and healthcare, a bitter battle has emerged over individual lifestyle and moral values, oscillating between the personal, racial, social, sexual and religious, which is shaping the battleground for the forthcoming election” (Young 2004). As David Broder— that paragon of Beltway journalistic wisdom—warned, “This nation has rarely appeared more divided than it does right now” (Broder 2000).

However as some scholars have noted, close elections, in and of themselves, are not evidence of polarization. A close election can be the product of a moderate, centrist and unimodal electorate just as much as it can be the result of a highly polarized and divided electorate. This does, of course, represent a divergence from traditional party positioning expectations given the central location of the median voter. Subjectively chosen communities that reflect stark partisan and cultural differences may make for good copy, but may not be representative of political, geographic, and social trends in the larger society (Wildermuth 2004). Jonsson notes Georgia’s remarkable adoption of the Republican party in a traditional Democratic stronghold where John F. Kennedy secured his second largest margin of victory, in any state, over Nixon (Jonsson 2004). But is Georgia’s shift in to the Republican side of the ledger emblematic of a shift in the nature of national politics and the issues relevant to party identification? A trend of Republican victories is suggestive, but not dispositive. Partisan trends may be a consequence of polarization trends, however they are not necessary for nor sufficient evidence of polarization.

So while these accounts suggest to the reading public that there is a culture war going on and political polarization is a real phenomenon, the primary evidence they marshal to demonstrate this fact may be misleading. Polarization may have indeed increased over the past several decades, but the data points commonly employed in popular accounts of polarization fail to make the case.

Fiorina’s primary empirical evidence against polarization is based on time-series data of the average positions on a variety of cultural issues. He notes that abortion attitudes over the ‘polarization’ time period of the last thirty years has been relatively stable (Fiorina, Abrams, and Pope 2004). Yet, as can be seen in Table 3.1, average policy positions in the electorate can be stable while the distributional characteristics shift drastically. One of Fiorina’s most important pieces of evidence against the Culture Wars thesis, just like the ‘close elections’ meme of the Culture War proponents, doesn’t actually speak to political polarization. If we imagine Table 3.1 represents the distribution of opinion on a cultural issue such as abortion over time, then it is apparent that the distribution of opinion on abortion can radically change while the location of the mean and median of the distribution is stable over the three time periods. Clearly a measure of polarization needs to account for the nature of the distribution of

opinion on an issue and not merely the central tendency in an opinion dimension. Fiorina’s analysis that tracks the movement of the average issue positions of the electorate over time completely fails to account for the distribution of that opinion over that time period. Further we cannot assess whether the direction of a change in the average issue position is evidence of depolarization without knowing what change in the distribution of opinion on an issue that move reflects. A move to the center may be evidence of depolarization…it may also be evidence of polarization. It isn’t about where you are. It is about where you’ve come from.

Fiorina argues that elite polarization is explained by the fact that activists polarize due to their attentiveness to politics, their high levels of information, and their sophisticated ideological outlook in comparison to an uninformed, uninterested, and unconstrained mass public, consistent with Converse’s evidence from a half-century ago. Rather than shifts towards the extremes in the mass electorate’s policy preferences, Fiorina and his colleagues argue that their choices have become increasingly polarized due to the polarization at the elite and activist level. Thus ideological conservatives have increasingly moved into the Republican camp and ideological liberals have moved into the Democratic camp as the parties at the elite levels have presented increasingly stark ideological choices in elections. This results in what Fiorina calls “sorting” which, he argues, is distinct from polarization. “Some analysts prefer to refer to [sorting] as ‘partisan polarization.’ We prefer the term ‘party sorting.’ Reserving the term polarization for bimodal distributions of opinion: voters are polarized on an issue if more of them cluster at the extremes than locate themselves in the center, or if they are moving from centrist positions toward the extremes” (Fiorina and Levendusky 2006). The sorting process that occurs absent preference polarization is illustrated in Table 3.1. Note that the number of moderates in the electorate never changes nor does their choice to align as an “independent” rather than with one of the parties. What sorting entails is conservatives dealigning from the Democratic Party and realigning with the Republican Party combined with the opposite shift in party allegiance among liberals. Thus the parties become more ideologically coherent while the distribution of opinion in the electorate remains unchanged.

Is sorting distinct from polarization? Fiorina’s distinction between sorting and polarization seems to hinge on what is being polarized rather than the distributional characteristics themselves. But polarization is generally applicable to any distribution, opinion or otherwise. Income, the attribute which economists have been primarily concerned with in defining and conceptualizing polarization, is not an opinion or issue dimension at all (Esteban and Ray 1994; Duclos, Esteban, and Ray 2004). Rather than defining sorting as a different species from partisan polarization, opinion polarization and partisan polarization (i.e. sorting) should be conceived as two types or classes of polarization. Just like a square and a parallelogram are both rectangles, so too is sorting and opinion polarization two kind of political polarization. The unifying symmetry between the variety of polarizations (issue, ideological, partisan, etc.), and thus the justification for this classification scheme, is that we can assess changes in their distributions, from polarized to depolarized, similarly. The particular attribute, be it partisanship, abortion opinion, or ideology, is irrelevant.

It follows that while the “Culture Wars” thesis is one form of political polarization, it is not synonymous with political polarization. As should be evident from the realignment discussion in chapter 1 and the conceptualization of polarization outlined in chapter 2, political conflict can occur across a plethora of dimensions of which the social issue dimension is just one. The New Deal involved economic polarization that produced the New Deal coalition and the partisan divide that would characterize American politics for three generations. The Iraq war may have proved more polarizing in the ‘Oughts’ than abortion and gay rights. Perhaps a new issue of potential political conflict and distributional polarization will arise as a result of the financial crisis and recession. Certainly the November, 2008 election results do nothing to dissuade us from crediting such a possibility.

As noted in Chapter 2, Fiorina argues that the elites have polarized (or sorted) independent of the mass electorate, which has largely moved to the center on social issues. But Fiorina’s argument regarding elite behavior presents a significant theoretical hurdle that he ultimately fails to clear. Political elites in a democratic republic must endure the judgment of the mass electorate periodically, and the United States is no exception. Hence Fiorina’s depiction of political elites as polarizing while the masses have centralized necessarily means that political elites have increasingly become unresponsive to the signals the mass electorate is sending. Fiorina speculates that primaries and ‘polarization entrepreneurs’ in Washington with big dollars behind them may explain this elite polarization independent of the masses. However, both of those explanations are unsatisfying. Primary fights might explain some polarization during the primaries, but what about the general election? Given the unlikelihood of a sitting legislator losing a primary fight, why would these elites continue to respond to the subset of primary voters rather than the larger class of general election voters? Party is one answer. Given increasingly uncompetitive districts, the intra-district median would shift to the poles of the ideological distribution. Centrists from a national perspective would find competing for that extreme median voter difficult. Once in the general, the ‘extreme’ candidates rely upon the partisanship in the electorate to yield victory. As for the money in politics angle, how does it make sense to raise a lot of money for re-election, only to lose that re-election because the candidate has wandered too far from his general election constituency?

One possible alternative explanation would be a polarized electorate. Where multiple modes exist, candidate divergence may be rational. The possibility of entry by a third party may cause parties to diverge from the median in order to discourage a third party challenge on their extremes in a polarized electorate(Fiorina 1999; Palfrey 1984). Indeed, entry at the candidate level might explain polarization at the congressional level where moderates may vote with hardliners in a move anticipatory of a third party challenge from hard-core ideologues in their districts such as challenge endured by Joe Lieberman for his pro-war stance (Fiorina 1999). Hinich and Munger develop a theory of ideology which permits party divergence (Hinich and Munger 1994). Incorporating previous studies that have pointed to incomplete information and uncertainty in voter policy locations and candidate locations, they argue that the creation and maintenance of an ideology by parties is a necessary component of political competition, given the traditional MVT model’s problem of representing choice in elections as having the same fundamental structure of economic decisions (Alverez and Franklin 1994; Bartels 1986; Calvert 1985; Enelow and Hinich 1981; Ferejohn and Noll 1978; Shepsle 1972; Wright and Goldberg 1985; Alverez 1997; Calvert 1986; Dalton, Beck, and Huckfeldt 1998; Feddersen and Pesendorfer 1999, 1997; Huckfeldt and Sprague 1988; Huckfeldt and Sprague 1987; Hinich and Munger 1994). In a political environment where Republicans have become much more consistently and strongly conservative (likewise for Democrats and liberals) vote-seeking parties rationally diverge in order to create a credible ideology which they can ‘sell’ to their constituents. Establishing an ideological flag at one of the ‘poles’ in a bimodal distribution would account for the divergence Fiorina fails to adequately account for.

A ‘compassionate conservative’ with a centrist agenda may have to tack right in the political environment of 2004, whereas he could comfortably govern from the center in the 1970’s. This raises an interesting theoretical question: how polarized must the electorate be in order for candidates to diverge? The two possibilities that Fiorina mentions do not exhaust the possible distribution of voters. Other distributions exist which may have a significant numbers of voters located in the center of the distribution and yet still permit divergence, as Downs himself noted. But clearly Fiorina has a serious theoretical problem. He has adopted a Downsian proximity model to explain the increase in the ‘weight’ associated with the cultural dimension; however his candidates behave in a decidedly un-Downsian fashion.

What mechanism produces candidate divergence (clearly contrary to classic Downsian logic) is an ambiguity in Fiorina’s initial argument, though he has attempted to account for it in later discussions (it remains “puzzling” to him despite his speculative explanations). He does not adopt Rabinowitz’s directional theory which calls for candidate divergence, yet his interest-group responsive model of party divergence on the issues irrespective of centripetal voter preferences may require it. Fiorina argues that electoral pressures on candidates have weakened while candidates have become more beholden to advocacy groups (with views on issues decidedly distant from the mass electorate’s mean). “The result is political activists and candidates whose ideological commitments run deeper than a generation ago, whose fear of losing elections is less than a generation ago, or both” (Fiorina, Abrams, and Pope 2004). It is a somewhat controversial position to take, and it has some apparent problems from a theoretical perspective and in terms of the empirical evidence (McKelvey and Ordeshook 1985, 1986; Morton 1993; Palfrey and Poole 1987; Hinich and Munger 1994).

It is difficult to assess the relative strength of the electoral connection over the past 50 years, however even if the ‘risk’ of loosing for incumbents has been lessened through a variety of factors, even Fiorina does not suggest it is not a or even the ‘prime mover’ of candidates for office. Fiorina accounts for the ‘irrational candidates’ implication of his argument by suggesting that as long as parties diverge a relatively equal distance from the mean, then voters cannot punish the candidates of those parties for adopting stances away from the center of the electorate distribution. The problem with this argument should be apparent: a party need merely move some upsilon towards the mean in order to secure electoral success. It is difficult to see how even if parties have become more beholden to advocacy groups that this would be powerful enough to overcome the obvious strategy: move towards the center and win. However, with an increasingly polarized electorate, parties, candidates, and political elites could diverge in response without having to postulate irrational elites beholden to the special interests.

“As the struggle proceeds, ‘the whole society breaks up more and more into two hostile camps, two great, directly antagonistic classes: bourgeoisie and proletariat.’ The classes, polarize, so that they become internally more homogenous and more and more sharply distinguished from one another in wealth and power” (Deutsch 1971)

Justice Potter Stewart, tasked with imposing a definition of obscenity in order to rule on the constitutionality of a fine imposed on a filmmaker for showing the French film The Lovers, said of obscenity that he could not intelligibly define it but that “I know it when I see it” (Jacobellis v. Ohio 1964). It is tempting to do the same with polarization. Polarization is relatively easy to visualize but much more difficult to define. Much of the empirical discussion of political polarization has glossed over what polarization is. There is a great deal of references to polarization, but few efforts to rigorously define it. Some conceive it as increasing extremism in the electorate or among social groups on issues, ideology, partisan and electoral choices. Others look to election results.

Polarization Nuts & Bolts: Necessary and Sufficient Components