Old Times There are Not Forgotten: Race and Partisan Realignment in the Contemporary South

American Journal of Political Science, Vol. 49, No. 3


 * Our focus is the regional political realignment that has occurred among whites over the past four decades. We hypothesize that the South’s shift to the Republican party has been driven to a significant degree by racial conservatism in addition to a harmonizing of partisanship with general ideological conservatism. General Social Survey and National Election Studies data from the 1970s to the present indicate that whites residing in the old Confederacy continue to display more racial antagonism and ideological conservatism than non-Southern whites. Racial conservatism has become linked more closely to presidential voting and party identification over time in the white South, while its impact has remained constant elsewhere. This stronger association between racial antagonism and partisanship in the South compared to other regions can not be explained by regional differences in nonracial ideology or nonracial policy preferences, or by the effects of those variables on partisanship.

The American voting public has shifted substantially toward the Republican party since the mid-1970s. The leading indicator of this shift has been the presidentialvote. The once-majority Democrats have captured only a minority of the white vote in each of the last seven presidential elections. Their only victories, in 1992 and 1996, seem to have been partially contingent on the strong third-party candidacies of Ross Perot. In 1994, the Republicans took control of the House of Representatives for the first time in nearly half a century; they have controlled the Senate for much of the past two decades; the once healthy Democratic majority of the governorships has switched to a strong Republican majority; and Republicans have come to parity in the state legislatures as well. In terms of underlying party identification, Republicans have overcome the stable majority once held by Democrats.

A number of factors are responsible for this Republican surge. Here we pursue the possibility that race and racial issues have played a more important role than ordinarily recognized. Carmines and Stimson (1989) make a convincing case that racial issues were central during the Civil Rights era, but their analyses end with data from 1980. In the decades since, policies attacking racial inequality have continued to attract the strongest opposition from Republicans and conservatives (e.g., Sears et al. 1997; Sniderman and Carmines 1997). Some have therefore seen a continuing role of racial prejudice in party divisions, particularlyin a racial ambivalence born of resentment toward blacks combined with basic commitments to fairness and egalitarianism (Kinder and Sanders 1996; Mendelberg 2001). Other scholars downplay the role of racial issues and prejudice even in contemporary racial politics: “A quarter century ago, what counted was who a policy would benefit, blacks or whites” (Sniderman and Piazza 1993, 4-5), while “the contemporary debate over racial policy is driven primarily by conflict over what the government should try to do, and only secondarily over what it should try to do for blacks” [emphasis in original], so “prejudiceis very far from a dominating factor in the contemporary politics of race” (Sniderman and Carmines 1997, 4, 73). And the conventional wisdom about partisanship today seems to point to divisions over the size of government (including taxes, social programs, and regulation), national security, and moral issues such as abortion and gay rights, with racial issues only one of numerous areas about which liberals and conservatives disagree, and far from the most important one at that (Abramowitz 1994; Abramowitz and Saunders 1998; Campbell 2002; Sniderman and Carmines 1997; Sniderman and Piazza 1993).

The White South
The major shift to the Republicans has occurred in the South. From the end of Reconstruction to the mid-twentieth century, Democratic presidential candidates in the “Solid South” invariably received far more votes than did Republicans, but their hold on the South has weakened ever since (Black and Black 1992, 2002; Sundquist 1983). In 2000, one of the most narrowly divided elections in history, the regional shift was complete, when Al Gore lost every Southern state including his own. Moreover, the main shift to the Republicans has been among white native-born Southerners: many older whites have changed parties (Beck 1977); most young, white, native-born Southerners today start out as Republicans; and while Republican migration to the South has contributed, it is not a dominant factor (Black and Black 1992; Carmines and Stanley 1990; Miller and Shanks 1996; Petrocik 1987; Stanley 1988).

Realignments generally depend on two factors. One is a change among party elites, and the other is fertile soil in the mass public’s attitudes. Beginning in the Civil Rights era and in the years since, conservative Southern whites have felt “abandoned” by the Democratic party. The reasons for this feeling are debatable. We believe they had, and continue to have, much to do with race. In the 1960s, national party elites began to stakeout conflicting positions on racial issues (Layman and Carsey 2002). The change in party positions was especially vivid to white Southerners, where the Democratic party had long protected the distinctive Jim Crow system. The Democratic party in the South also became more racially liberal, with increased African American participation and the gradual replacement of older white conservative Democrats. Also, racially relevant issues such as busing, crime, welfare, and affirmative action have continued to be quite salient in American politics in the post-civil-rights era (Kinder and Sanders 1996; Mendelberg 2001). On the other hand, party elites did not change only on racial issues. Democratic elites began to move to more liberal positions on noneconomic issues such as national defense or abortion in the 1970s, and the Reagan era heightened the distinctive economic conservatism of the Republican party.

Nevertheless, at the level of the mass public, we focus on the role of white Southerners’ racial attitudes for four reasons. First, race has been a dominant element in Southern politics from the beginning, leading to significant sectional conflicts at several of the nation’s most pivotal moments: the writing of the Declaration of Independence and the Constitution, the events triggering the Civil War, and the wrenching abandonment of Jim Crow. In each case, the white South’s formal system of racial inequality confronted substantial,though far from unanimous, opposition elsewhere in the country. Such deeply ingrained, regionally concentrated cultural differences are always difficult to change, and it seems to us implausible that they have been eliminated in the relatively brief historical time since the end of Jim Crow.

Second, the onset of realignment was intimately entangled with race. Beginning in 1963, the national Democratic party abandoned its century-long commitment to avoid challenging the Jim Crow system. The civil rights legislation proposed by Northern Democrats immediately attracted massive resistance from Southern Democrats in Congress, and support for the Democratic party began to erode among Southern whites (Black and Black 1992; Carmines and Stimson 1989; Sundquist 1983).

Third, in the years since, race continued to generate considerable political heat. At the national level, a number of issues central to contemporary campaigns seem to have been linked implicitly to matters of race. For example, local television news seems to dramatically overrepresent blacks as perpetrators of violent crime. The result may be that opinions about crime have become tightly linked to attitudes about blacks (Hurwitz and Peffley 1997). In the short term, exposure to such stories has been shown to exacerbate negative racial attitudes and boost whitesupport for punitive crime policies (Gilliam and Iyengar 2000). Consequently, exposure to news about crime primes racial attitudes during candidate evaluation (Valentino 1999). Gilens (1999) provides similar evidence about the racialization of news coverage of poverty and welfare. Long-term patterns in the media’s framing of racialized issues have a profound influence on public attitudes about race and may determine aggregate preferences on racial policies (Kellstedt 2003). Other studies demonstrate that campaigns can capitalize on these linkages, employing subtle cues that prime racial thinking among white citizens (Mendelberg 2001; Valentino, Hutchings, and White 2002). Race, therefore, may still play a significant role in politics even when it is not discussed explicitly.

Finally, a number of quite heated and largely symbolic racial issues have arisen in the South. Several states have witnessed roiling debates about the use of Confederate battle symbols on public insignia. The NAACP organized a boycott of tourism in South Carolina in 2000 until the state legislature voted to remove the Confederate battle flag from atop the state house. The victory for opponents of the flag was limited, since the agreement provided that it be flown near a Confederate monument on the statehouse lawn. A similar flag controversy played out in Mississippi in 2001. Georgia’s flag controversy may have contributed to the victory of the state’s first Republican governor since Reconstruction, Sonny Perdue.

The Southern parties today are split quite decisively along racial lines. Republicans are almost all white, and blacks are the dominant core of the Southern Democratic party (Black and Black 2002). All this leads us to suspect that racial attitudes, in particular, might be found to structure partisan divisions today, particularly in the white South. Having said that, we see major changes in the role of race in the South along with such continuities. In the 1950s and 1960s, white Southerners strongly supported Jim Crow or “old-fashioned” racism, focused on rigid social distance between the races, legalized segregation, formal racial discrimination, and beliefs in the inherent inferiority of blacks (Sheatsley 1966). But much of that support for formal racial inequality has disappeared in the New South (Schuman et al. 1997), and is now too skimpy to be the main foundation of the party alignment. Instead we argue that its political influence has been replaced by that of a new form of racism, variously described as “symbolic racism,” “modern racism,” or “racial resentment,” blending racial animus with perceptions that blacks violate traditional American values, such as individualism (Sears and Henry2003). It is reflected in beliefs that blacks’ continuing disadvantages reflect their own lack of work ethic rather than continuing racial discrimination and that blacks make excessive demands and get too many undeserved advantages (Kinder and Sanders 1996; Sears et al. 1997). It is important to note that symbolic racism is conceptualized today as mostly expressing sincere beliefs, melding ordinary conservatism with some racial animosity, rather than hypocritical efforts to hide a deeper and pure racism.

Other analysesof contemporary Southern politics do not assign race to such a central role. For example, the most important recent work on the subject, by Black and Black (2002), arrives at a somewhat different emphasis. In their view,the distinctiveness of Southern political culture in Jim Crow days, based in a rural, largely impoverished white population that was centrally focused on racial issues, has now been replaced with a strong, white middle class and a less regionally distinctive politics. To be sure, massive resistance to change in the racial status quo drove the politics of the 1960s, epitomized by the 1968 Wallace candidacy. But with the Reagan revolution of the 1980s, white Southerners replaced that explicitly racial focus with the same broad-ranging conservatism as in the rest of the nation, centered on defense, class self-interests, smaller government, lower taxes, family values, personal responsibility, and other forms of economic and social conservatism, all personified in a president who was enormously popular in the white South. Racial questions were mainly absorbed into this broader set of views (also see Black and Black 1992, and Petrocik 1987, on the 1980s).

This impressive work by Black and Black (2002) must be taken as a starting point by anyone interested in Southern politics. We believe it leaves room for a closer look at the role of race, in three critical ways. First, the proposition of Southern political convergence to the rest of the nation, especially regarding race, is generally not tested with direct regional comparisons. Second, it argues both that race is the central demographic cleavage in contemporary Southern politics and that the nearly all-white character of the Republican party is due not to race, but to the interests of an enlarged white middle class. However, their data seem to show that the white working class differs more politically from blacks than from the white middle class, as if race continues to trump class in the South (see chapter 8). Third, the evidence comes almost exclusively from voting returns rather than from survey research,so attitudinalexplanationsof partisandifferencesaremostly inferred.

Existing Evidence
Our case that specifically racial conservatism is central to Southern white realignment requires at least three kinds of empirical evidence. One is that Southern whites continue to have more negative racial attitudes than do Northern whites. A second is that over time racial conservatism has become closely associated with Republican partisanship in the white South, as white Southerners have realigned, but similar changes have not occurred elsewhere in the nation. Finally, general ideology and/or nonracial issue preferences should not account for these regional differences. What evidence exists on these points?

As mentioned above, Jim Crow racism has diminished sharply in the New South, at least as measured by conventional survey techniques, eroding most but not all regional differences. However there is scant research on contemporary regional differences in other forms of traditional prejudice, such as antiblack affect and stereotypes, nor on contemporary forms of prejudice such as symbolic racism. Whites living in the South consistently showed greater opposition to such race-targeted policies as busing, fair housing, antidiscrimination laws, and spending on race-targeted programs than have whites living elsewhere, at least into the 1990s (Glaser and Gilens 1997; Schuman et al. 1997; Tuch and Hughes 1996, 1997). Also, lifelong white Southerners seem to be more racially conservative than in-migrants. In the South of the 1960s, Southern rearing was a better predictor of support for George Wallace than was current Southern residence (Wright 1977). In later decades, opposition to racial policies has been greatest among those both bred and currently residing in the South (Glaser and Gilens 1997; Wilson 1986). These regional differences may even be stronger when assessed with unobtrusive measures (Kuklinski, Cobb, and Gilens 1997). Unfortunately, these studies largely antedate the Clinton era and focus especially on traditional racial attitudes, yielding an incomplete portrait of whites’ current thinking.

There is even less evidence available on the second empirical issue, the role of racial conservatism in driving partisan realignment in the white South. Rather, most available studies since the 1970s focus on putatively nonracial predictors of partisanship such as general political conservatism, religious beliefs, abortion, defense, gun control, and the role of the federal government (Black and Black 1992; Carmines and Stanley 1990; Green et al. 2003; Kellstedt 1990; Steed, Moreland, and Baker 1990). These studies do not give clear guidance on the role of specifically racial issues relative to these others in white Southerners’ partisanship. As a result there is a need for a more direct test of the role of racial conservatism in the Republican surge in the white South. And, finally, we require a comparison of racial conservatism with other attitudinal explanations of whites’ voting behavior.

Before presenting our specific hypotheses, we need to state clearly the boundaries of our goals in this study. Our intent is to assess the power of racial attitudes in shaping partisanship in the past and the present, and across regions of the country. We cannot presume to estimate precisely the amount of prejudice in any region, nor is our focus on moral judgments, however loaded racial issues are in America. Finally, our thesis is that race is central to the realignment, but not the sole force driving it, and we do not attempt to disprove previous work illuminating other mechanisms of realignment.

Hypotheses
We address four concrete hypotheses:

(1) Regional differences in racial conservatism have persisted since the Civil Rights era, despite the general decline of Jim Crow racism throughout the nation and especially in the South.

(2) These regional differences in the contemporary period are large and significant even controlling for more general political conservatism, and across several measures of racial animosity.

(3) White Southerners’ votes and partisanship have become increasingly aligned with their racial attitudes since the Civil Rights era. No similar increase exists for whites elsewhere in the country.

(4) In the contemporary era, racial attitudes have a significantly stronger impact on white Southerners’ partisanship than elsewhere. Moreover, this regional difference is not due simply to nonracial conservatism.

Data and Measurement
To compare the roles of racial and nonracial attitudes in the realignment of white Southerners over time, we need both an especially rich set of survey measures over time and large numbers of cases. Ideally, we would have been able to track the same racial attitudes back to the mid-1960s when the Civil Rights movement was in full swing, but most measures are availableonly beginning in the early 1970s, and some only inconsistently since then. Therefore, we have pooled datasets that provide consistent measures of our key variables. We will employ the cumulative General Social Surveys (GSS) and National Election Studies (NES) from the 1970s through 2000. The particular years included in each analysis vary according to the availability of measures. In every case, these decisions will be made explicit. In these analyses, we areconcerned solely with the attitudes of white respondents.

Racial Attitudes
Various items tapping Jim Crow racism have been included in the GSS, but two were asked consistently from 1976 through 1996: white people’s rights to keep blacks out of their neighborhoods [RACSEG] and laws against racial intermarriage [RACMAR]. The Cronbach’s alpha for a scale created from these two items was .57. Racial affect was measured in the NES from 1992 to 2000 using the feeling thermometer for whites minus the feeling thermometer for blacks (the difference score was used to reduce potential response-bias effects). Negative racial stereotypes were measured in the 1992 NES with a three-item scale rating blacksas “hardworking” vs. “lazy,” “intelligent” vs. “unintelligent,” and “peaceful” vs. “violent,” all on 7-point scales (Cronbach’s alpha = .67). In 1996 and 2000, the last item was replaced with “trustworthy” vs. “untrustworthy” (Cronbach’s alpha= .83, .80, respectively).

The precise and consistent measurement of symbolic racism is particularly important, given previous debate over it (Sniderman and Tetlock 1986; Tarman and Sears 2005). Its most common measures over the years have been four 5-point agree/disagree items in the NES, which provided us an additive scale for the years 1986, 1988, 1990, 1992, 1994, and 2000: (1) Irish, Italians, Jewish, and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors (agree); (2) Over the past few years blacks have gotten less than they deserve (disagree); (3) It’s really a matter of some people not trying hard enough; if blacks would only try harder they could be just as well off as whites (agree); (4) Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class (disagree). The Cronbach’s alpha for this scale was .73. In the 1972 NES, only the last two items were available (Cronbach’s alpha= .49). When we compare levels of symbolic racism across time, only the two-item scale is used for all available years (1972, 1986 through 2000; alpha= .51 for 1986 through 2000). However, when examining the relationship cross-sectionally between symbolic racism and other political attitudes, such as partisanship, we use the two-item scale for 1972 and the 4-item scale for 1986 through 2000. In all the scales described here, items were summed and rescaled from zero to one for ease of interpretation of the results. We will return at the end to the current status of the debate over measurement of symbolic racism.

Policy Attitudes
To assess the effect of racially driven policy preferences, we employ racial and nonracial policy attitude items. In the NES beginning in the 1980s a 7-point item has been used to measure racial policy preferences, involving special aid to blacks (“...the government in Washington should make every effort to improve the social and economic position of blacks” or “...the government should not make any special effort to help blacks because they should help themselves”). We compare this with items on two non-racial policies often thought central to Southern realignment, a 4-point item on abortion and a 7-point item on defense spending.

Partisanship and Ideology
In both the NES and GSS, party identification runs from 1 (“strong Democrat”) to 7 (“strong Republican”) with pure Independents at 4. Political ideology is also measured from 1 (“strongly liberal”) to 7 (“strongly conservative”). For presidential vote, we dichotomize votes for the Republican candidate versus for all other candidates.

Region
Our main comparisons arebetween the South as a whole, defined as the 11 states of the former Confederacy, and all other states, described as “North and West.” This is a conservative test of our since it excludes hypotheses from the South border states with some Southern tinges. A few studies have also found more pronounced regional differences when the “Deep South” (Louisiana, Mississippi, Alabama, Georgia, and South Carolina) is distinguished from the “Outer South” (Texas, Arkansas, Tennessee, Florida, North Carolina, and Virginia; see Black and Black 2002; Glaser and Gilens 1997; Tuch and Martin 1997). We make that further distinction when sample size permits it, but usually the number of cases in the Deep South is insufficient, so we pool these two Southern regions for most analyses.

Party Realignment of Southern Whites
The phenomenon we begin with is the shift of white Southerners to the Republican party since the 1950s, both in absolute terms and relative to those living in the North and West. Although this is a well-established fact, the magnitude of the change is impressive. Two findings stand out. First, in NES data, fewer than 10% of the whites in the Deep South were Republicanin 1956, and fewer than 30% were in the Outer South. By 2000, these had risen to over 60% and nearly 50%, respectively. Second, this is a case of realignment, not dealignment. The Southern increase in Republicanism has been mirrored by an equally substantial decline in Democratic identification, from 87% of the whites in 1956 to 24% in 2000. In contrast, both Republican and Democratic identifications were quite stable over that period in the North and West, hovering around 40% and 45%, respectively. We tested the differential trends across regions with an OLS model on the standard7-point party identification scale, incorporating dummies for region and interactions between those dummies and time. As shown in the second column of Table 1, the region * time interactions are both highly statistically significant, suggesting that both regions of the South were indeed realigning relative to the North and West.

Trends over Time
Our general proposition is that the white South is primarily responsible for the national shift to the Republican Party and that this shift was driven substantially by the politics of race. Our first hypothesis is that Southern whites have continued to hold more negative racial attitudes than do whites living elsewhere, net of other factors. There is much evidence that Jim Crow or “old-fashioned” racism has declined greatly, but we doubt that the more contemporary symbolic racism has. Moreover, if the historically greater levels of racial animus in the white South have persisted, symbolic racism should have been consistently higher in the South than elsewhere throughout the last several decades, despite the decline of Jim Crow racism.

TABLE I

Jim Crow racism declined sharply throughout the country between the 1970s and the 1990s, as shown in the second panel of Figure 1. Multivariate analyses yield a large and highly significant drop for the entire sample from 1976to 2000 (see Table1, column 3), from .33 to .13 on a 0-1 scale (p <.001). But both regions of the South house higher levels of Jim Crow racism than the North and West over this period, as reflected by the large coefficients for the two regional dummies (3 = .23 and .14, both p < .001). When we interact region with time, we see that the size of the regional difference diminishes significantly over the period. This is due mostly to the fact that the South began further from the minimum on this scale than the North and West.

The story for symbolic racism is much different. First, symbolic racism has remained stable or even increased slightly over time (see the first panel of Figure 1), as reflected in a significant effect for year in column 5 of Table 1. Second, the regional difference in symbolic racism is statistically significant and of moderate size, with more symbolic racism in the South than the North and West at all time points. This is shown in Figure 1 and in significant region effects in Table 1. Third, the South has not converged to the rest of the country in symbolic racism; neither region * year interaction on symbolic racism in column 6 is statistically significant.

FIGURE 1

Though Southern realignment is not commonly thought to be driven by increasing ideological conservatism among whites over time, it bears checking. In fact political ideology among whites has been quite stable over the available period (1972-2000), shifting only from 4.23 to 4.42 on 7-point liberal-conservative scale. As shown in the third panel of Figure 1, the Deep and Outer South were slightly more conservativethan the rest of the country throughout this time period. There appear to be no large trends over time in the size of this regional difference. The slightly increased regional divergence in 1998 and 2000 comes too late to explain the long realigning trends across three decades. In multivariate analyses (Table1,column 8), the Deep South * yearinteractionis nonsignificant while the Outer South * year interaction shows a slightly increasing convergence to the rest of the country. In other words, the massive shifts in Southern whites’ party identifications do not result from conservatizing shifts in political ideology in the South.

In sum, party identification has swung sharply to the Republicans among white Southerners since the Civil Rights era, but not elsewhere in the country. Jim Crow or “old-fashioned” racism has diminished drastically, but the South has retained slightly more of it than the rest of the country. In contrast, symbolic racism has remained stable over time, but again the South has been consistently higher than the rest of the country. White Southerners have consistently been more ideologically conservative than other whites over this period, but there has been no change in the size of that difference. Regional differences in these three attitudes have persisted, then. Changes in these regional differences are therefore unlikely to account for the white Southern realignment over the past three decades.

Contemporary Regional Differences
Our second hypothesis is that Southern whites continue to be more racially conservative than whites living elsewhere in the contemporary period above and beyond their more general political conservatism. Earlier we saw that white Southerners have been higher in Jim Crow racism and symbolic racism over the past three decades, at least at the bivariate level. But would we find such regional differences in the contemporary era on all dimensions of racial animosity, or are they specific to these two dimensions? If so, are such differences merely a spurious effect of white Southerners’ conservatism or demographic distinctiveness? To find out, we examine the period beginning in the 1990s and broaden the range of racial attitudes by pooling available surveys across years (all NES surveys since 1992 with relevant measures, and all the GSS surveys since 1990). We again examine Jim Crow and symbolic racism, but add two conventional measures of traditional prejudice, stereotyping of blacks, and the feeling thermometers. To isolate race from other factors, we control for ideology and demographics.

TABLE 2

The South, and especially the Deep South, includes a disproportionate share of the highly racially conservative whites on each of these four measures, as displayed in Table 2. For example, 55% of the whites in the Deep South, but only 32% of those in the North and West, fall in the top third of the national distribution of symbolic racism. The impact of region is highly significant for each measure in analyses of variance that include controls for education, age, gender, and ideology. White Southerners are more racially conservative than whites living elsewhere on all conventional dimensions of racial attitudes. These regional differences in racial conservatism are not explained by differences in general political conservatism or demographics.

Trends over Time
Our third hypothesis is that the association between racial conservatism and Republican partisanship has strengthened over time in the South, both in absolute terms and relative to the rest of the country. Before the civil rights era, Jim Crow racism was not a defining component of party differences in the nation as a whole. Today it continues not to be a central force on partisanship, but for a different reason: because it has dwindled nearly to the vanishing point even in the South. But the more contemporary symbolic racism fits the language of today’s racial politics more closely. So we expect that it has increasingly drawn Southern whites to the Republican party. If so, the relationship between symbolic racism and both Republican presidential vote and party identification should have increased over the last 30 years, particularly in the South.

The NES series provides measures of symbolic racism in the presidential years 1972, 1988, 1992, and 2000. Table3 contains results for models of vote choice over that span, separated by region. Time (“year”) in these analyses is modeled continuously, with the first year in the series (1972) set to “0” and the last year (2000) set to “28.” The first row displays the impact of symbolic racism in the first year of the time series, 1972. At that point, the association between symbolic racism and the vote was tiny in the South, but was quite large and statistically significant in the North and West (the negative coefficient for the “year” variable in both models is substantively uninteresting, primarily reflecting the 1972 Nixon landslide that outperformed later Republican candidacies).

TABLE 3

Most important, since then the impact of symbolic racism on presidential vote has increased in the South but not in the North. The symbolic racism * year interaction is positive and statistically significant in the South, but indistinguishable from zero elsewhere. Important also is that ideology has not had an increasing effect in the South over the past 30 years once racial attitudes have been taken into account: the ideology * year interaction is nearly zero in the South, while it has grown slightly larger elsewhere.To wrap these findings together, an omnibus model was run to determine whether the differential growth curves for the impact of symbolic racism were significantly distinguishable across region. The three-way symbolic racism *time*region interaction was significant (p<.01) even when controlling for the three-way interaction between ideology, time, and region, which was not significant in that model (p = .31).

To illustrate the sizeof these differences in the association between symbolic racism and the vote, the logistic regression coefficients were transformed into probabilities. These are presented in Figure 2. Bars in the figure represent the change in the likelihood of voting for the Republican candidate associated with moving from two standard deviations below the mean to two standard deviations above the mean on the symbolic racism scale, holding ideology and all other variables constant at their means. The top panel shows that the impact of symbolic racism on vote preference in the South rounds to zero in 1972, so no bar appears for that year.The association between symbolic racism and the vote grows in each subsequent year. By 2000, moving from low to high on the symbolic racism scale led to an increase of 52 points in the likelihood of voting for George W. Bush in the South. The bottom panel shows that the association was more substantial in the North and West in 1972 than it was in the South, but also that it is quite stable overtime. Though not plotted here, the association between ideology and the vote is constant over time in the South and increasing slightly over time in the North. In sum, the Southern white presidential vote has become more tightly aligned overtime with racial attitudes but not with ideology. This trend has not occurred elsewhere in the country.

FIGURE 2

Next we explore the over-time trends in the association between racial attitudes and party identification. Party identification tends to be acquired relatively early and to be quite stable through the adult years. Presidential votes are likely to be less consistent, not least because they involve a variety of candidates over time. As a result party identification should be a lagging indicator of realigning changes in partisanship. Table 4 displays the results of regression analyses predicting to party identification similar to those presented for presidential vote. Party identification was regressed on variables for year (1972= 0,2000= 28),symbolic racism, and interactions between the two. Again, to provide a conservative test of the effects of symbolic racism, we included controls for ideology, the interaction between ideology and year, education, age, and gender.

The impact of symbolic racism on party identification, net of controls, was negligible throughout the country in 1972, as is indicated by the coefficients for symbolic racism in both regional models. Its impact increased significantly over time in the South, as represented by the symbolic racism * year interaction in the first column. That shift was not significant outside the South (second column). By 2000 its impact on white Southerners’ partisanship had grown to more than twice its influence in the North and West, even after controlling for general conservatism and demographic variables. Beyond that, Table 4 also shows a growing relationship between ideology and party identification in the South, as reflected in the significant ideology * year interaction in the first column. This result is consistent with the conventional wisdom about the increasing regularization of ideology and party in the South. What we add is that the impact of specifically racial conservatism on party identification also has been growing in the white South.

Our hypothesis is that racial conservatism, as reflected in symbolic racism, increasingly drew white Southerners to the Republican party over this period. But what if Jim Crow racism had already been associated with Republican partisanship in the 1970s? Perhaps symbolic racism has simply replaced those earlier political effects of old-fashioned racism as the latter gradually disappeared over the period of our analysis. Then our data would not have shown that racial conservatism was driving party realignment in the South. Rather,they would be consistent with a contrary view, that the essential link of racial conservatism, writ large, to party identification had not changed in any important way: only the language in which it was expressed had changed.

TABLE 4

TABLE 5

The empirical implication for the white South would be that Jim Crow racism had a significant link to Republican party identification in the 1970s, but that link would have declined significantly in the years since then as Jim Crow gradually disappeared. In fact, though, as might be expected from the historic centrality of the Democratic party to the traditions of the old South, Jim Crow racism was actually significantly correlated (r = .12, p < .05) with Democratic party identification in the South in 1976 (the first year both variables are available in the GSS). Even that modest relationship disappears with demographic controls in a multivariate regression analysis, however. Jim Crow racism was not significantly related to party identification in the South in 1976, as indicated by the upper left-hand entry in Table 5. Furthermore, Jim Crow racism, unlike symbolic racism, did not become increasingly linked to Republican partisanship over the following 20 years.The correlation of Jim Crow racism and party identification was only r= -.03 in 1996, and there is no strong trend over time, as indicated by the small and statistically nonsignificant coefficients for the Jim Crow * year interaction. In other words, the finding of an increasing association of symbolic racism and Republican partisanship over time is not an artifact of the evolution of racial attitudes from old-fashioned racism to more contemporary forms of racial animus.

Contemporary Regional Differences
Our fourth hypothesis is that in the contemporary period, racial attitudes are tied more strongly to partisanship in the white South than elsewhere. This difference should hold above and beyond the impact of ideology, and it should be stronger for racial than for nonracial attitudes. We begin by regressing partisanship on symbolic racism, region, and ideology. The entries in the first column of Table 6 are unstandardized logistic regression coefficients for the likelihood of voting for the Republican presidential candidate (a dummy variable), while the second column contains unstandardized OLS coefficients for party identification (the standard 7-point scale). The first row shows that symbolic racism is linked significantly to both political indicators in the contemporary North/West.The third row shows the key finding, that symbolic racism is tied even more closely to both the vote and party identification in the former Confederacy, as reflected in the significant symbolic racism * region interaction.

TABLE 6

Alternative explanations for these linkages with symbolic racism might focus on general political ideology, both because some believe it explains Southern realignment and because others believe symbolic racism taps nonracial conservatism as well or better than it does racism (e.g., Sniderman and Tetlock 1986). However, the greater impact of symbolic racism in the South is not merely an artifact of stronger general political conservatism. The fourth row of Table 5 shows that ideology is, by itself, a strong and significant predictor of both the vote and party identification outside the South. But the two nonsignificant ideology * region interaction terms in the fifth row show that ideology has no greater association with partisanship in the South than elsewhere.

As before, we have converted the logistic regression coefficients in the vote preference model into probabilities. These results are presented in Figure 3. The height of each column represents the change in the probability of voting for the Republican presidential candidate associated with moving from two standard deviations below the mean to two standard deviations above the mean on the symbolic racism or the liberal-conservative ideology scales. The first two bars show that such a shift in symbolic racism amounts to a 16-point increase in the probability of voting for the Republican candidate outside the South, but a 45-point increase in the South. The second set of bars show that the impact of ideology is larger than that for symbolic racism, but there is no regional difference in the size of that effect. The regional difference in the effects of symbolic racism, but not in the effects of ideology, suggest that specifically racial attitudes have structured the Southern-based partisan realignment of the past four decades.

FIGURE 3

We have relied primarily on symbolic racism to make the case that racial attitudes are central to Southern realignment. To isolate its specifically racial component we have controlled on standard political ideology. Still, it is worth making further effort on this front, since the non-racial hypothesis is a prominent alternative among both academics and the lay public. One approach is to replace symbolic racism with stereotypes as a measure of racial attitudes in models predictingvote choice and party identification in the contemporary period. We have performed these analyses and the results, not presented here, support our black stereotypes are associated hypothesis. Negative significantly with Republican party identification and Republican vote choice in the South but not in the North in the 1990s.

Yet another approach to isolating the impact of specifically racial attitudes is to turn to policy preferences as predictors of partisanship. To accomplish this, we compared the effects of racialpolicy issues (government aid to minorities and affirmativeaction) with those of abortion and national defense, two issues often invoked to explain Southern conservatism but deliberately chosen to have little spillover from racial issues. The dependent variable is dichotomous Republican vote choice with all other candidates in the excluded category, as before. We test the effect of each policy attitude and of region on the vote, along with demographics. The key term is the interaction of policy attitudes and region, which we hypothesize will be significantly positive in the 1990s for racial issues but close to zero for nonracial issue preferences. We also expect that the regional difference in the impact of racial policy attitudes has increased over time.

The model for the impact of racial and nonracial issues is shown in Table 7. Here we compare the regional difference in the1980s with that in the1990s.The first row shows the effects of opposition to assisting blacks outside the South, since that is the excluded group in the dummy variable for region. It strongly boosts the probability of Republican presidential voting in both decades. The second and third rows indicate that opinions about abortion and defense also have a large impact on Republican vote (again, outside the South) in both decades. Then, most important, the fifth row shows that the racial policy attitudes * region interaction switches from negative and statistically nonsignificant in the 1980s to positive and statistically significant in the 1990s. In other words, racial policy attitudes are much more strongly associated with voting in the South than elsewhere, but only in the most recent decade.

In contrast, the linkage of nonracial policy attitudes to vote preference shows no significant regional differences in either the 1980s or the 1990s. Opposition to abortion and support for defense spending increase support Republican presidential candidates in the North and West. However, the nonsignificant abortion * region and defense * region interactions indicates no regional difference in the impact of these nonracial policy issues on the vote, in either decade. In short, racial issue opinions have become more strongly linked to vote choice in the South than in the North and West in the 1990s, which was not true in the 1980s. Nonracial issues operated almost identically in both regions and in both decades.

Conclusions
The phenomenon we began with is the greatlyenhanced standing of the contemporary Republicanparty since the Civil Rights era. The change is due largely to the sharp movement of Southern whites out of the Democratic party into the Republican party. At the elite level, the parties had long been divided over the economic issues central to the New Deal. They began to diverge on racial issues in the 1960s, and then on other issues in the 1980s, most prominently taxes, abortion, national defense, and faith-based issues. Why the sharp sectional difference in whites’ responses to essentiallythe same events?

The conventional wisdom appearsto be that Southern realignment began when the racial agendas of the national parties changed in the 1960s, with the national Democratic party moving to stronger support for civil rights. However race later became much less important, both because of the of the old Jim Crow belief system and because nonracial issues such as abortion, defense, gun rights, and so on became more important to conservatives.In this view, then, Southern realignment is mainly due to a long-overdue harmonization of white Southerners’ party identifications with their basic conservatism on the other issues.

We are more skeptical, however. Plainly, the anomalous mixture of conservatism and Democratic partisanship among white Southerners has been much reduced, regularizing the relationship between ideology and party. But beyond that we argue that realignment has resulted from yet another playing out of white Southerners’ historic conservatism about race in particular. A cultural way of life ingrained for so long is unlikely to have been eradicated thoroughly enough to have been shunted to the political sidelines so quickly. Its persistence has been facilitated by the polarization of party elites over racial issues, which has provided clear alternatives for voters who disagree about race. As a result, over the past four decades Southern whites abandoned their previous allegiance to a racially conservative Southern Democratic party in favor of a Republican party newly conservative on racial issues.

We present three general sets of findings. One is that at the end of the Civil Rights era Southern whites were more racially conservative than whites living elsewhere. More important for our purposes, the regional gap in racial conservatism has not closed since then, despite the sharp drop in Jim Crow racism. Southern whites remain more racially conservative than whites elsewhere on ev- ery measure of racialattitudes ordinarilyused in national surveys.

Second, we looked at the linkage of racialattitudes to partisanship. Over time, racial conservatism has become more tightly linked to both Republican presidential voting and party identification in the South. Those linkages have generally been weaker outside the South and have not increased over time. In the South, the linkage of symbolic racism to party identification, a lagging indicator, has developed more slowly than on presidential voting, a leading indicator of partisanship.

The most prominent alternative viewpoint points to the increasing influence of nonracial forces, such as general conservative ideology and/or nonracial policy preferences, rather than to specifically racial conservatism. Three findings shed doubt on this alternative. First, since the Civil Rights era, Southern whites have not become more conservative relative to other whites, in terms of general political ideology. Second, we present evidence of an increasing regional difference in the linkage of racial conservatism to partisanship, above and beyond the effects of ideology. The stronger link between racial attitudes on partisanship among white Southerners than among whites in the North and West held up after ideology was controlled. Given the considerable evidence that racial attitudes have spilled over into some other domestic policy issues such as welfare (Gilens 1999), crime (Kinder and Sanders 1996; Mendelberg 2001; Valentino 1999), and taxes and spending (Sears and Citrin 1985; Valentino, Hutchings, and White 2002), seemingly race-neutral conservatism may itself have become partially racialized. If so, our tests may in fact “over-control” for nonracial conservatism, and so underestimate the effects of racial conservatism. Third, we found that racial attitudes have increasingly influenced partisanship in the white South when we explicitly compared the effects of racial and non-racial policy attitudes on partisanship in separate analyses. Racial policy attitudes were more closely linked to the vote in the South than elsewhere in the 1990s, but nonracial policy attitudes yielded no such regional differences. In the South, ideology itself did have an increasing impact overtime on party identification (though not on presidential vote), so we assume some regularization of ideology and partisanship has occurred independent of race. But racial conservatism in particular seems to have played a potent role in the realignment of Southern whites in the late twentieth century, above and beyond the effects of putatively race-free ideology or nonracial issues.

We have emphasized symbolic racism as an indicator of the racial attitudes we see as important in white Southern realignment. This concept has been criticized in the past (e.g., Sniderman and Tetlock 1986), leading to much relevant research (see Sears and Henry 2005). We should briefly address that literature. One criticism was that symbolic racism is not a coherent belief system or measured consistently. In recent years, however, it has consistently been conceptualized and measured in terms of four themes: the denial of discrimination, criticism of blacks’ work ethic, and resentment of blacks’ demands and treatment by the broader society, which together form a logically, psychologically, and statistically coherent belief system (Tarman and Sears 2005). Its origins were said to be obscure, but now have been shown to lie, at least partially,in the theorized mixture of antiblackaffect and individualism (Searsand Henry 2003). Its distinctiveness from Jim Crow racism was questioned, but whites’ support for the latter has been sharply diminished while support for symbolic racism remains quite widespread, and the political effects of symbolic racism dwarf those of Jim Crow racism (Searset al. 1997).

Finally, symbolic racism was said to reflect nonracial political conservatism rather than racial prejudice.To be sure, they are correlated, but in factor analyses including all three sets of variables, symbolic racism loads about equally on otherwise distinctive racial prejudice and political conservatism factors (Sears and Henry 2003); the best-fitting structural equation models require that the symbolic racism items be collected as a separate factor rather than allocated to other constructs like ideology (Tarman and Sears 2005); and controlling on ideology does not materially reduce the effects of symbolic racism on racially-relevant dependent variables (Sears et al. 1997). In other words, once tested empirically, those earlier critiques have almost all turned out to be inaccurate (for reviews, see Hutchings and Valentino 2004; Krysan 2000; Sears and Henry 2005).

Going one step further, might measurement artifacts explain our findings of regional differences? One possibility is that the symbolic racism items might carry different meanings for respondents in the South than elsewhere, which would complicate inferences about regional differences in the linkage of racism and partisanship. But respondents seem to have understood the symbolic racism items similarly in both regions: the reliability of the measure is practically identical in both regions (Cronbach’s alpha of .73 in the South and .74 elsewhere), and symbolic racism is correlated identically with the racial policy scale in both regions (r = .53 in both cases). Finally, in results not presented here, we find that the relationship between ideology and symbolic racism is nearly identical across regions. These findings suggest symbolic racism has the same meaning in both regions.

Another possibility is that the more negative racial attitudes in the South might be an artifact of stronger social desirability pressures against expressing racial animosity in the North, because of the longer history of socially acceptable overt racism in the South. However, Kuklinski, Cobb, and Gilens (1997) show that white Southerners actually exhibit a larger gap than do Northerners between standard survey measures of racial attitudes and unobtrusive measures of them. This suggests both that, if anything, we are underestimating true regional differences in racial conservatism, because of white Southerners’ greater tendency to hide true prejudices, and underestimating true regional differences in the linkage of racial attitudes to partisanship, because such correlations should contain more error in the South.

We have argued that racial conservatism has been a significant contributor to party realignment in the white South. But what have been the mechanisms by which this has happened? At the level of the individual voter, the primary cause of the persistence of these regional differences is presumably the transmission of a broad culture of racial conservatism in the South across generations. For example, lifelong white Southerners seem to be more racially conservative than in-migrants (Glaser and Gilens 1997), and even young white Southern adults were consistently more racially conservative than their counterparts in other regions in the late 1980s (Steeh and Schuman 1992). Beyond that, our reasoning suggests that the linkage between racial attitudes and political preferences should be strongest for the youngest white Southerners, who were socialized as the parties were realigning. In-migrants to the South in the latter half of the twentieth century may also have adopted partisan identities consistent with their racial attitudes prior to, or following, their migration. It is also possible that race continues to be more salient in Southern culture than elsewhere, which might explain the added potency of racial attitudes there. These questions go beyond the scope of this article, but they are important for understanding the persistence of regional cultures and the dynamic processes underlying partisan realignments.

Turning to the elite level, explicitly racial issues have not been prominent in recent Southern presidential campaigns (Black and Black 2002; Mendelberg 2001). But race remains a salient political issue, in two ways. First, the sharp racial differences in the composition of the two parties have often been salient, as in Jesse Jackson’s 1984 campaign for the Democratic presidential nomination (Sears, Citrin, and Kosterman 1987); the prominence of black and Latino appointees in the Clinton administration and his highly publicized links to the congressional black caucus; and the redistricting in the 1990s that substantially changed the colors of Southern congressional delegations, replacing many white Democratic congressmen with either black Democrats or white Republicans. Second, as we noted above, some central issues in recent campaigns have been implicitly racialized and have been shown to evoke racial attitudes.

We would conclude that racial conservatism seems to continue to be central to the realignment of Southern whites’ partisanship since the Civil Rights era. But the scope of any single article must alwaysbe limited in some ways, especially in attempting to explain as broad a phenomenon as party realignment. So, for example, we could not test social class (Black and Black 2002; Petrocik 1987) or religion-based (Green et al. 2003) explanations for Southern realignment. Nor could we address other recent changes in the party system, such as a Republican shift in the Mountain states or a Democratic shift on the coasts	(see Marchant-Shapiro and Patterson 1995). Clearly it would go beyond the data presented above to assume that racial issues were as central to those shifts as we have suggested is the case in the white South. But a full understanding of this most recent realignment will require considering those elements explicitly.

= Resources =