Abstract
The 2004 U.S. presidential election was one of the most divisive in recent history (Pew Research Center 2004). The divisions in the electorate are popularly seen as the culmination of a process of political polarization underway since the 1970s (e.g., Frank 2004), and are epitomized by the now-ubiquitous map of the United States which shows swaths of red (i.e., majority Republican) states in the center of the country surrounded by blue (i.e., majority Democratic) states on the east and west coasts and in the north central region. In this chapter we investigate the geographic dimensions of political polarization in the United States through the lens of the 2004 election. We elucidate the principal contours of the divisions in the electorate, and characterize the manner in which the effects of the correlates of voting behavior cluster regionally. We take an ecological approach, using spatial econometrics to estimate the interregional divergence in the influences of the characteristics of populations and places on the odds of the Republican vote. To this end we employ aggregated data on 3,106 counties in the lower 48 states, which is the finest spatial scale at which both electoral returns and a variety of demographic and contextual variables are readily available. Our goal is to push the limits of ecological analysis in electoral geography. We first develop a theoretical framework in which geography plays a central role in electoral polarization. Our central hypothesis, which draws on themes in the political science literature (Johnston et al. 2004; Cho and Rudolph 2008), is that a number of social processes that operate at fine spatial scales tend to push individuals voters’ views into closer alignment with the ideological preferences of their geographically proximate majority – a phenomenon we call “localized entrenchment.” Drawing on the sociological literature on polarization (DiMaggio et al. 1996; Evans 2003), we circumvent the well-documented handicap of weak correlation between demographic attributes and ideology by employing a richer array of explanatory variables than prior spatial statistical analyses (e.g., O’Loughlin et al. 1994). We then apply spatial statistical techniques that exploit the spatial interrelationships among the electoral returns and our set of covariates, and find strong indications of entrenchment. Finally, we employ advanced methods to characterize the spatial heterogeneity in our estimated relationships – rather than re-estimate our aggregate statistical model on different regional sub-samples, we use geographically weighted regression (GWR). This technique enables us to exploit the spatial interdependencies among the entire universe of counties to estimate the fine-scale geographic variation in our covariates’ influences on the 2004 presidential vote, while simultaneously controlling for the underlying spatial distributions of the characteristics of people and places. The patterns of agglomeration in the resulting influences on voting behavior are consistent with our explanation of how local entrenchment might induce polarization of the electorate.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Glaeser et al. (2005) develop a theory of strategic extremism which illustrates the incentives political parties have to divide on issues in order to increase their chances of winning at the polls. Partisanship turns on two key elements: among voters, the existence of an intensive margin where the level of support matters (e.g., turnout or donations, as opposed to the extensive margin of voting) and which parties can activate by taking extreme positions that appeal to their respective bases, and the ability of parties to target extreme statements to their own supporters while bypassing those of the opposition, thereby avoiding a backlash. Below, we note that this sort of targeting becomes easier the more the electorate is ideologically segregated along geographical lines.
- 2.
This is an example of the modifiable areal unit problem (Openshaw 1984). Differences between the number of Democratic and Republican votes were as large between red and blue counties within some states as they were between some red and blue states. Using counties as the unit of analysis is attractive precisely because, unlike states, congressional districts or electoral precincts, their geographic boundaries are independent of electoral processes relevant to the presidential vote. The consequent absence of selection bias makes us confident in exploiting county characteristics as strictly exogenous covariates in our subsequent analyses.
- 3.
These statistics were computed for our sample of 3,106 counties in the lower 48 states. Indices of isolation measure the likelihood of Republicans’ and democrats’ exposure to the opposing group at 55% and 51%, respectively, while non-uniformity in the pattern of votes is given by the index of dissimilarity at 22%.
- 4.
Note that B-IV’s variance and excess kurtosis are larger than A-II’s These authors test whether these two moments of the distributions of survey respondents’ attitudes on a diverse array of social issues have increased over time.
- 5.
“contrasts in individual sociocultural characteristics are not direct indicators of political polarization. Hence, contrasts in such characteristics may or may not constitute evidence of polarization. Analysts must provide additional information about the strength of the links between social characteristics and relevant political variables, as well as information about the stability of such linkages.” (p. 568)
- 6.
For example, southern conservative voters switching from Democratic to Republican (Schreckhise and Shields 2003; Bullock et al. 2005; Valentino and Sears 2005), northeastern voters becoming increasingly liberal (Speel 1998), and the rise of the mountain west as a conservative voting bloc (Marchant-Shapiro and Patterson 1995).
- 7.
For example, Glaeser and Ward (2006, p. 131A): “These differences in beliefs within the United States drive home a central point about how politically relevant beliefs are formed. People in different states have been exposed to quite similar evidence through national media outlets, but they have reached radically different conclusions, and continue to hold these conclusions despite being aware that others disagree. This disagreement requires either different prior beliefs or some other deviation from Bayesian reasoning. One natural alternative model is that people base opinions mostly on the views of those around them. As such, local interactions are critical, and these provide plenty of possibility for wide geographic variation…”
- 8.
See also Johnston et al. (2004), who develop a slightly different taxonomy.
- 9.
Despite anecdotal evidence (e.g., Bishop 2008; Bishop and Cushing 2004) and statistical indications particular kinds of neighborhood environments influence their residents’ ideological leanings, irrespective of demographic composition (Williamson 2008), the political sources and consequences of self-selection have yet to be thoroughly investigated.
- 10.
For example, Mutz (2002, p. 852): “Homogeneous environments are ideal for purposes of encouraging political mobilization. Like-minded people can encourage one another in their viewpoints, promote recognition of common problems, and spur one another on to collective action. Heterogeneity makes these same activities much harder. Participation and involvement are best encouraged by social environments that offer reinforcement and encouragement, not ones that raise the social costs of political engagement.” Also, Williamson (2008, pp. 20–21): “… the spatial sorting of residents by political ideology, once it reaches a sufficiently advanced stage, may help create what Lazarsfeld, Berelson, and Gaudet (1944) termed a ‘reinforcement effect’; not only might residents of a very conservative suburb be less likely to hear a liberal viewpoint from their neighbors but such areas will likely be targeted and contacted frequently by conservative political activists while being relatively ignored by liberal political activists, further reinforcing the relationship between spatial context and individual political outlook.” Homogeneity facilitates a political campaign’s ability to mobilize voters by reducing the cost of what Lazarsfeld et al. (1944) refer to as activation (“not to form new opinions but raise old opinions over the thresholds of awareness and decision,” p. 74), and reinforcement (“to secure and stabilize and solidify […] vote intention and finally to translate it into an actual vote,” p. 88).
- 11.
- 12.
For example, Cho and Rudolph (2008, p. 277): “Casual observation exposes citizens to meaningful information through low-intensity neighbourhood cues such as the display of yard signs, bumper stickers, or simple observations and biases created by how neighbors dress and behave, what types of cars they drive, or how well their garden is groomed. Such low-intensity cues may influence behavior by subtly communicating information about the prevailing norms and sentiments within a community. In particular, they may provide signals about a local community’s political culture and ethic or the nature and distribution of political preferences within that community.”
- 13.
These data are of necessity approximate, not being adjusted for the results of recounts in Ohio and New Mexico. There were additional independent candidates on the ballot in each state, but the numbers of votes cast for them were small. Neither of these factors seems likely to significantly change our main results.
- 14.
Campbell and Monson (2008) note that this database suffers from a number of problems, principally non-response bias in survey questionnaires, omission of non-denominational churches – which account for an increasing share of religious participation, and an inability to track the number of residents of one county who attend church in another.
- 15.
These data are available online from the Alan Guttmacher Institute. Preliminary regressions indicated that the state-level incidence of abortion and teen pregnancy were not significant predictors of the odds of voting Republican, in part because of their collinearity with state fixed effects.
- 16.
Our use of the proportions of divorced persons and households headed by single females is admittedly crude. In particular, it is hard to know whether the statistical effect of these variables on electoral outcomes is driven by the voting behavior of people in these groups or by morally conservative voters’ negative reactions to the former.
- 17.
The states are: Arkansas, Georgia, Kentucky, Michigan, Mississippi, Montana, North Dakota, Ohio, Oklahoma, Oregon and Utah.
- 18.
Consistent with our discussion of the prominent role of social interactions, we defined the neighborhood of each county as a radius of 200 km, which is approximately twice the distance traveled at the highest state-mandated speed limit (75 mph) for the maximum average commute time in Fig. 2. The advantage of this scheme is that every row in W has at least one non-zero off-diagonal element, which allowed us to row-standardize the resulting matrix of distances without having to worry about divide-by-zero errors.
- 19.
For the basic model, LMρ = 2695. 41 (p < 0. 01) and LMλ = 1044. 49 (p < 0. 01), while for the fixed effects model, LMρ = 523. 08 (p < 0. 01) and LMλ = 278. 27 (p < 0. 01).
- 20.
For the basic model LMρ ∗ = 172. 99 (p < 0. 01) and LMλ ∗ = 1823. 91 (p < 0. 01), while for the fixed-effects model LMρ ∗ = 73. 96 (p < 0. 01) and LMλ ∗ = 318. 77 (p < 0. 01).
- 21.
Specifically, Zc = diag[Z 1c, …, Z Nc ] is an N ×N diagonal matrix of c’s distance-based weights expressed as a local kernel, \({Z}_{jc} =\exp \left (-0.5{\left ({d}_{jc}/h\right )}^{2}\right )\), in which d jc is the distance between c and other counties j, and the spatial interaction radius is given by a fixed bandwidth parameter, h, that we estimate using a crossvalidation procedure.
- 22.
- 23.
For example, as in racially polarized voting, where whites and non-whites have divergent ideological preferences which push their vote distributions in opposite directions away from the mean, like B-II and B-III in Fig 1 b.
- 24.
Our results suggest that the “social multiplier” associated with voting in the 2004 U.S. presidential election is around 1.4. This is substantially smaller than the values found by Glaeser et al. (2003) for the peer effects of college roommates, criminal behavior in cities, or the human capital spillovers in urban labor markets. This outcome is not surprising given that ballots are secret, and that even with early voting, individuals are only exposed to the influence of neighbors’ self-announced behavior for at most three weeks. (Although more prolonged exposure might result from proximity to intensely partisan voters.) It therefore seems more plausible that ρ is picking up the influence of correlated effects associated with counties’ common exposure to political campaigns, and the reflection of that stimulus in their residents’ everyday social interactions.
- 25.
Kruskal-Wallis rank-sum tests indicated significant differences between suburban and core metro counties’ distributions of the vote and the proportion of the population self-identifying as white only.
- 26.
Kruskal-Wallis tests indicated significant differences between suburban and rural counties’ distributions of the proportion of jobs in agriculture, but not their voting patterns.
- 27.
Given our coding of GMB as a state dummy variable, the significance of the indirect spatial lag (as opposed to the direct) coefficient is to be expected, as it is by definition a wide-area effect.
- 28.
Globally, the signs and relative magnitudes of the estimates are similar. However, the magnitudes of almost half of the estimates shrink while the rest increase. The median values of the parameter distributions are in closer agreement with the signs of our spatial Durbin estimates, though slight differences in their magnitudes persist.
- 29.
In conducting these analyses we employ our original 200-km bandwidth kernel.
- 30.
These authors find that a simple model with two independent variables, the coefficients associated with each covariate may exhibit collinearity even if the underlying exogenous variables in the data generating process are uncorrelated, and a high degree of spatial correlation between two covariates increases the potential for the two sets of coefficients to exhibit interdependent, spatially opposing patterns of effects. In both cases the upshot is spurious spatial trends in the GWR estimates.
References
Abramowitz A, Saunders K (2008) Is polarization a myth? J Polit 70:542–555
Allard, SW (2007) The changing face of welfare during the Bush administration. Publius 37: 304–332
Ansolabehere S, Rodden J, Snyder JM (2006) Purple America. J Econ Perspect 20:97–118
Anselin L (2002) Under the hood: issues in the specification and interpretation of spatial regression models. Agric Econ 27:247–267
Anselin L, Bera AK, Florax RJGM, Yoon MJ (1996) Simple diagnostic tests for spatial dependence. Reg Sci Urban Econ 26:77–104
Baldassarri D, Bearman P (2007) Dynamics of political polarization. Am Sociol Rev 72:784–811
Baldassarri D, Gelman A (2008) Partisans without constraint: political polarization and trends in American public opinion. Am J Sociol 114:408–446
Bartels LM (2000) Partisanship and voting behavior, 1952–1996. Am J Polit Sci 44:35–50
Bivand R (2006) Implementing spatial data analysis software tools in R. Geogr Anal 38:23–40
Bivand R, Brunstad R (2006) Regional growth in Western Europe: detecting spatial misspecification using the R environment. Pap Reg Sci 85:277–297
Bishop W (2008) The big sort: why the clustering of like-minded America is tearing us apart. Houghton-Mifflin, New York
Bishop W, Cushing RG (2004) Response to Philip A. Klinkner’s “red and blue scare: the continuing diversity of the American electoral landscape.” Forum 2(2):Art 8
Brunsdon C, Fotheringham AS, Charlton ME (1996) Geographically weighted regression: a method for exploring spatial non-stationarity. Geogr Anal 28:281–298
Bullock CS, Hoffman DR, Gaddie RK (2005) The consolidation of the white southern congressional vote. Polit Res Q 58:231–243
Burridge P (1981) Testing for a common factor in a spatial autoregression model. Environ Plann A 13:795–800
Campbell DE, Monson JQ (2008) The religion card: gay marriage and the 2004 presidential election. Public Opin Q 72:399–419
Cho WKT, Rudolph TJ (2008) Emanating political participation: untangling the spatial structure behind participation. Br J Polit Sci 38:273–289
DiMaggio P, Evans J, Bryson B (1996) Have Americans’ social attitudes become more polarized? Am J Sociol 102:690–755
Dixit AK, Weibull JW (2007) Political polarization. Proc Natl Acad Sci 104:7351–7356
Evans JH (2003) Have Americans’ attitudes become more polarized? – an update. Soc Sci Q 84: 71–90
Evans JH, Nunn LM (2005) The deeper “culture wars” questions. Forum 3(2):Art 3
Fiorina MP, Abrams SJ, Pope JC (2006) Culture war? the myth of a polarized America, 2nd edn. Longman, New York
Fiorina MP, Abrams SJ (2008) Political polarisation in the American Public. Ann Rev Polit Sci 11:563–588
Fleisher R, Bond J (2004) The shrinking middle in the U.S. congress. Br J Polit Sci 34:429–451
Fotheringham AS, Brunsdon C, Charlton M (2002) Geographically weighted regression: the analysis of spatially varying relationships. Wiley, Chichester
Fotheringham AS, Charlton ME Brunsdon C (1997) Measuring spatial variations in relationships with geographically weighted regression. In: Fischer MM, Getis A (eds) Recent developments in spatial analysis. Springer, London, pp 60–82
Frank T (2004) What’s the matter with Kansas? How conservatives won the heart of America. Metropolitan Books, New York
Frey WH (2000) Regional shifts in America’s voting-aged population: what do they mean for national politics? Population Studies Center Report No. 00–459, Institute of Social Research, University of Michigan
Gensler H (ed) (1996) The American welfare system: origins, structure, and effects. Praeger, Westport
Gimpel JG, Schuknecht JE (2001) Interstate migration and electoral politics. J Polit 63:207–231
Glaeser EL, Ward BA (2006) Myths and realities of American political geography. J Econ Perspect 20:119–144
Glaeser EL, Sacerdote BI, Scheinkman JA (2003) The social multiplier. J Eur Econ Assoc 1: 345–353
Glaeser EL, Sunstein CR (2009) Extremism and Social Learning. J Legal Anal 1:263–324
Glenmary Research Center (2004) Religious congregations and membership in the United States: 2000. Nashville
Goodman L (1953) Ecological regression and the behavior of individuals. Am Sociol Rev 18: 663–664
Hanushek EA, Jackson JE, Kain JF (1974) Model specification, use of aggregate data, and the ecological correlation fallacy. Polit Method 1:87–107
Hetherington MJ (2001) Resurgent mass partisanship: the role of elite polarization. Am Polit Sci Rev 95:619–631
Hillygus DS, Shields TG (2005) Moral issues and voter decision making in the 2004 presidential election. PS: Polit Sci Pol 38:201–210
Huckfeldt R, Sprague J (1995) Citizens, politics, and social communication: information and influence in an election campaign. Cambridge University Press, New York
Hunter JD (1992) Culture wars: the struggle to define America. Basic Books, New York
Johnston R, Jones K, Sarker R, Propper C, Burgess C, Bolster A (2004) Party support and the neighbourhood effect: spatial polarisation of the British electorate, 1991–2001. Polit Geogr 23:367–402
Kim J, Elliott E, Wang DM (2003) A spatial analysis of county-level outcomes in U.S. Presidential elections: 1988–2000. Elect Stud 22:741–761
Klinkner PA (2004) Red and blue scare: the continuing diversity of the American electoral landscape. Forum 2(2):Art 2
Klinkner PA, Hapanowicz A (2005) Red and blue Déjà Vu: measuring political polarization in the 2004 election. Forum 3(2):Art 2
Klofstad C, McClurg SD, Rolfe M (2006) Family members, friends, and neighbors: differences in personal and political networks. Paper presented at the 2006 annual meeting of the The Midwest Political Science Association, Palmer House Hilton, Chicago, IL
Lazarsfeld P, Berelson B, Gaudet H (1944) The people’s choice: how the voter makes up his mind in a political campaign. Columbia University Press, New York
Manski CF (1993) Identification of endogenous social effects: the reflection problem. Rev Econ Stud 60:531–542
Manski CF (2002) Economic analysis of social interactions. J Econ Perspect 14:115–136
Manza J, Brooks C (1997) The religious factor in US presidential elections, 1960–1992. Am J Sociol 103:38–81
Marchant-Shapiro T, Patterson KD (1995) Partisan change in the mountain west. Polit Behav 17:359–378
McClurg SD (2003) Social networks and political participation: the role of social interaction in explaining political participation. Polit Res Q 56:449–464
McKee SC (2007) Rural voters in presidential elections, 1992–2004. Forum 5(2):Art 2
McKee SC (2008) Rural voters and the polarization of American presidential elections. Polit Sci Pol 41: 101–108
McMillen DP (2003) Spatial autocorrelation or model misspecification? Int Reg Sci Rev 26: 208–217
McPherson M, Smith-Lovin L, Cook JM (2001) Birds of a feather: homophily in social networks. Annu Rev Sociol 27:415–444
Mellow N, Trubowitz P (2005) Red versus blue: American electoral geography and congressional bipartisanship: 1898–2002. Polit Geogr 24:659–677
Miller AS, Hoffmann JP (1999) The growing divisiveness: culture war or a war of words? Soc Forces 78:721–745
Mutz DC (2002) The consequences of cross-cutting networks for political participation. Am J Polit Sci 46:838–855
Mutz DC, Mondak JJ (2006) The workplace as a context for cross-cutting political discourse. J Polit 68:140–155
O’Loughlin J, Flint D, Anselin L (1994) The geography of the nazi vote: context, confession and class in the Reichstag election of 1930. Ann Assoc Am Geogr 84:351–380
Openshaw S (1984) The modifiable areal unit problem. Geo Books, Norwich
Pew Research Center (2004) Moral values: how important? voters liked campaign 2004, but too much “mud-slinging.” Pew Research Center for the People and the Press Report, 11 November
Poole KT, Rosenthal H (2001) D-NOMINATE after 10 years: a comparative update to congress: a political-economic history of roll-call voting. Legis Stud Q 26:5–29
Schreckhise WD, Shields TG (2003) Ideological realignment in the contemporary U.S. electorate revisited. Soc Sci Q 84:596–612
Speel RW (1998) Changing patterns of voting in the northern United States: electoral realignment, 1952–1996. Pennsylvania State University Press, Philadelphia, PA
Stonecash JM, Brewer MD, Mariani MD (2002) Diverging parties: social change, realignment, and party polarization. Westview Press, Boulder
Sue Wing I, Walker JL (2005) The 2004 presidential election from a spatial perspective. Mimeo, Boston University
Thornton A, Young-DeMarco L (2001) Four decades of trends in attitudes toward family issues in the United States: the 1960s through the 1990s. J Marriage Fam 63:1009–1037
Valentino NA, Sears DO (2005) Old times there are not forgotten: race and partisan realignment in the contemporary south. Am J Polit Sci 49:672–688
Wheeler DC (2007) Diagnostic tools and a remedial method for collinearity in geographically weighted regression. Environ Plann A 39:2464–2481
Wheeler DC, Tiefelsdorf M (2005) Multicollinearity and correlation among local regression coefficients in geographically weighted regression. J Geogr Syst 7:161–187
Williamson T (2008) Sprawl, spatial location, and politics: how ideological identification tracks the built environment. Am Polit Res 36:903–933
Woodberry RD, Smith CS (1998) Fundamentalism et al.: conservative protestants in America. Annu Rev Sociol 24:25–56
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2010 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Wing, I.S., Walker, J.L. (2010). The Geographic Dimensions of Electoral Polarization in the 2004 U.S. Presidential Vote. In: Páez, A., Gallo, J., Buliung, R., Dall'erba, S. (eds) Progress in Spatial Analysis. Advances in Spatial Science. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-03326-1_13
Download citation
DOI: https://doi.org/10.1007/978-3-642-03326-1_13
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-03324-7
Online ISBN: 978-3-642-03326-1
eBook Packages: Earth and Environmental ScienceEarth and Environmental Science (R0)