Public Choice 118: 105–124, 2004. © 2004 Kluwer Academic Publishers. Printed in the Netherlands.
105
Under two flags: Symbolic voting in the State of Mississippi ∗ GÖKHAN R. KARAHAN & WILLIAM F. SHUGHART II Department of Economics, University of Mississippi, University, MS 38677-1848, U.S.A. Accepted 25 November 2002 Abstract. Participants in a special election held in the State of Mississippi on April 17, 2001, voted overwhelmingly against changing the design of the state’s flag, which incorporates a symbol of the Confederacy. The determinants of voting on the flag are analyzed and turnout rates in April 2001 are compared with those for recent gubernatorial and presidential elections. We find that the flag vote divided Mississippians sharply along lines of race, class and political ideology. A key empirical implication is that voter positions in issue space tend to be more polarized when political choices have expressive as opposed to instrumental consequences.
1. Introduction On April 17, 2001, a special election was held in the State of Mississippi for the purpose of allowing voters to choose between the state’s current flag, which incorporates the so-called Confederate (or Beauregard) battle flag (see Figure 1), and a new state flag designed by a commission chaired by former governor William Winter. The flag decision was the only issue on the ballot and voters were offered two options: Proposition A favored keeping the present-day flag, adopted as the state’s official banner in 1894; Proposition B proposed to replace the Confederate symbol in the flag’s canton corner with a double circle of 19 white stars on a field of blue, the larger central star representing Mississippi’s original admission as the Union’s 20th state in 1817 (see Figure 2). The alternative proposed in Proposition B was defeated decisively: the 1894 flag garnered about 64 percent of the 767,682 votes cast in the special election. The referendum was an unintended, if belated, consequence of a lawsuit filed on April 19, 1993, by local Mississippi branches of the National Association for the Advancement of Colored People (NAACP) and 81 individual African-Americans against then-Governor Kirk Fordice, a case ultimately decided by the Supreme Court of Mississippi on May 4, 2000.1 Plaintiffs prayed for “declaratory relief and an injunction against any future purchases, displays, maintenance or expenditures of public funds on the State Flag”,2 ∗ We benefited from the comments of Michael Reksulak, Robert Tollison and an anonymous referee. Any remaining errors are our own.
106
Figure 1. The Confederate (Beauregard) battle flag, 1862
Figure 2. The current Mississippi state flag and the proposed alternative; Proposition A: Current state flag (1894); Proposition B: Proposed state flag
contending that “the flying of the State Flag violates their constitutional rights to free speech and expression, due process, and equal protection as guaranteed by the Mississippi Constitution”.3 Affirming the trial court’s dismissal of the NAACP’s complaint, the justices were controlled by a 1998 case that raised similar issues. In that matter, Stanley J. Daniels and five co-plaintiffs sought to enjoin the Harrison County Board of Supervisors from “carrying through with their intensions to fly the Confederate battle flag on the Mississippi Beaches and other public property located within Harrison County”.4 Finding no reason to overturn the trial judge’s granting of the Board’s motion for summary judgment, the Daniels Court ruled that public display of the battle flag does not “deprive any citizens of this State of any constitutionally protected right”.5 In a separate concurring opinion, which likewise rejected the plaintiffs’ constitutional arguments, along with their narrow construction of the statutory guidelines pertaining to the proper display of the state flag,6 Justice Banks wrote that
107 “Mr. Daniels and others like him who are compelled to voice their objections to the battle flag should look to the legislature” for redress.7 In the course of reaching its decision in the NAACP matter, the Supreme Court discovered that Mississippi has not in fact had an official state flag since the early twentieth century. Although the flag now flown had been adopted by the legislature in 1894, the authorizing statute was rescinded when, inadvertently, it failed to be carried forward to the revised Mississippi code of 1906: “consequently, the State of Mississippi has no existing law establishing or promulgating a flag for the State. The current flag is Mississippi’s by custom and usage only”.8 The Court went on to say, however, that “this does not mean that the State must remove the flag currently in use”; furthermore, “it is beyond this Court’s power and intent to order the removal of the flag currently used”.9 As for remedial action, “the decision to adopt or not adopt a flag for Mississippi is wholly within the power of the Mississippi legislature and the executive branch of government, now, as it was in 1894 when the current flag was adopted and in 1906 when the flag statute was repealed”.10 Following the Supreme Court’s ruling, Governor Ronnie Musgrove, Lieutenant Governor Amy Tuck and House Speaker Tim Ford jointly appointed a special 17-member commission to study the flag issue and to recommend alternatives. Seeking citizen input, the commission held a series of unruly public meetings throughout the state. Representatives of the NAACP and other civil rights organizations, such as the Southern Christian Leadership Conference (SCLC), argued passionately that the current flag is offensive to many African-Americans.11 Black Mississippians, it was said, saw the battle flag’s inclusion in the state flag as a constant reminder of slavery and as a modern symbol of bigotry and racism.12 Proclaiming “Heritage, not Hate”, many whites, on the other hand, maintained that the battle flag, which was first carried onto the field by General P.G.T. Beauregard in the spring of 1862 at the battles of Shiloh, Tennessee, and Corinth, Mississippi, and later became the official flag of the Army of Northern Virginia (Sansing, n.d.), is an important historical emblem honoring the sacrifices and nobility of the Confederacy’s “Lost Cause”.13 Old wounds were reopened. Some groups at the very extreme ends of the ideological spectrum exploited the flag debate for their own narrow purposes, as they have before (Coski, 2000; Holmes and Cagle, 2000). Others, black and white alike, thought that Mississippi had problems to address more important than the design of the state’s flag. After the commission completed its work and recommended the adoption of a new state flag, it soon became clear that the Mississippi legislature wanted to duck the hot-button issue. The legislators decided to let the voters decide.14
108 The purpose of this paper is to model empirically the determinants of the votes cast in Mississippi’s April 2001 special election. Using a county-wide data set and a number of multivariate statistical techniques, we find that the defeat of the proposed new state flag vote was as much a matter of class as it was of race. The paper is organized as follows. Section 2 places the flag referendum in political context. Empirical estimates of a model of voting on the proposed replacement for the 1894 flag are presented in Section 3. Section 4 uses the data to shed some light on the theory of expressive voting. Section 5 concludes.
2. The politics of race and class in Mississippi More than half a century ago, V.O. Key (1949: 230) remarked that the Negro determines the policy of Mississippi in the nation; resistant to external intervention. . . . Mississippi politics may be regarded, if one keeps alert to the risks of oversimplification, as a battle between the delta planters and the rednecks. It is a battle of diminishing intensity, but the cleavage between the planters of the delta and the rednecks of the hill has persisted for half a century and even yet appears from time to time. According to Key, “the rednecks of the hill” inhabit “the northern end of the state and occupy roughly its eastern part, broadening out almost to the [Mississippi] River in the south and petering out in the pine forests of the coastal region”. The delta planters, on the other hand, occupy the fertile strip of land along the eastern banks of the Mississippi River, a region almost two counties wide meandering south from Memphis, Tennessee, to the Yazoo River at Vicksburg (Ibid.: 231). It was in the economic interests of the delta planters to care for their “Negro” sharecroppers, and the rednecks of the hill apparently did not approve of such paternalism (Ibid.: 231–235). These geographic divisions of race and class are still evident in Mississippi, trapped in a time warp where the Fourth of July is remembered as much for marking the double blow of dreadful news from Vicksburg and Gettysburg in 1863, as it is for the signing of the Declaration of Independence in 1776. Driving 25 miles due west from the hill country of north Mississippi returns one to the 1950s; drive all the way to the River and one is transported to the 1930s. Small towns predominantly populated by African-Americans loom as islands of abject poverty in an utterly flat, seemingly endless sea of cotton and soybean fields. The cotton gin, the grain elevator and the bank all are owned by one of the white planters, who frequently live not in town, but in large, isolated houses scattered across the delta, easily identified by the grove
109 of pecan trees that characteristically surround and shade them from summer’s oppressive heat. Thirty-nine percent of Mississippi’s population is black, higher than any other jurisdiction in the United States save the District of Columbia. The concentration of African-Americans is much greater in the delta, though, where the distribution of income is also highly skewed along racial lines. Largely black Tunica County, located in the northwest corner of the state, was, before the advent of casino gambling, the poorest in the nation. Academic researchers have arrived at two conflicting conclusions with respect to the political dynamics of racially mixed communities. On the one hand, it is argued that whites living in densely populated black areas may perceive blacks as threats to their wellbeing and, thus, behave in ways that promote their own narrow interests, such as voting for white candidates (Giles and Hertz, 1994; Orey, 1998). On the other hand, whites and blacks may forge alliances to defeat racially divisive candidates in favor of those who, regardless of race, are more moderate (Carsey, 1995; Sadow, 1996; Voss, 1996).15 In addition to raising issues of race and class, the April 2001 special election was seen in some quarters of the state as a referendum on economic development. One major theme the proponents of the new flag design rallied behind was that continuing to fly a Confederate emblem sends a negative signal both to businesses considering relocating to Mississippi and to prospective immigrants. George Shelton of the Mississippi Legacy Fund, an ad hoc coalition that spent about $626,000 campaigning for the new flag, listing banks, insurance companies, attorneys, food processors and actor Morgan Freeman among its contributors, was quoted as saying, “We’ve made so much progress, and the rest of the country isn’t seeing that. We need a symbol that represents the entire state, that shows that we’re moving forward.”16 Blake Wilson of the Mississippi Economic Council likewise argued that the status quo would be detrimental to the state’s economy. He continued on to assert, “States that have removed it, like Alabama, have moved forward. We can get the same kind of growth by sending that signal around the world that we have moved forward.”17 Similarly, the University of Mississippi’s head football and basketball coaches jointly declared their support for the new design, contending that the 1894 flag hampered their efforts to recruit able black high-school athletes. Opponents of the new flag countered that, if the state flag was indeed a barrier to economic development, the Japanese automaker Nissan would not have chosen Mississippi as the site of its new American production facility.18 Zogby International conducted a survey (telephone interview) of 886 Mississippi voters between January 25 and February 2, 2001. According to this
110 survey, 82 percent of whites were aware of the upcoming referendum on the flag, whereas only 65 percent of blacks responded that they knew of the April 2001 vote. Respondents who had been born out of state appeared to be more aware of the approaching special election than native Mississippians; awareness was likewise higher among college graduates than among high school graduates. Overall, 55 percent of the respondents said that they would vote for the current flag and 35 percent said they would vote for the new flag. Thirty-eight percent of those identifying themselves as Democrats and 72 percent of those identifying themselves as Republicans expressed preferences for the current flag. Respondents born in Mississippi and those born outside the state did not differ greatly in terms of support for the new flag design, while a higher percentage of college graduates than of high school graduates said they preferred the new flag. Eighteen percent of whites and 65 percent of blacks said that they would vote for the new flag; in contrast, 73 percent of whites and 20 percent of blacks favored the old flag. The poll’s margin of error was +/- 3.4%.19 3. Empirical model of the flag vote We use aggregate county data in this study. The special election results for each of Mississippi’s 82 counties were obtained from the Secretary of State’s web site and the remaining data are from the Center for Population Studies at the University of Mississippi, the Mississippi Employment Security Commission, and the Mississippi Statistical Abstract (Office of External Affairs, 1999).20 The dependent variable, OLD FLAG, is defined as votes cast for the 1894 design as a fraction of a county’s voting age population.21 The independent variables are per capita county income (INCOME); percentage of the population who have earned college degrees (COLLEGE); percentage of the population voting for George W. Bush in the 2000 presidential election (BUSH); population density per square mile (DENSITY); percentage of the population that is nonwhite (NONWHITE); percentage of the population not native to the state (BORNOUT); and the percentage change, between 1999 and 2000, in the number of businesses located in the county (BUSINESS). Two regression models are estimated by ordinary least squares. Model 1 includes all of the above independent variables. Model 2 includes everything in Model 1 as well as an interaction term, NONWHITE∗DENSITY, to test whether voting behavior is influenced by racial proximity.
111 3.1. A priori expectations Owing to more enlightened attitudes toward other races and to greater stakes in economic progress, votes cast for the old flag are expected to be negatively related to INCOME. Conventional wisdom would certainly suggest that less affluent white Mississippians – Key’s “rednecks of the hill” – were the most resolute opponents of change. To the extent that their positions in the income distribution had been achieved under the status quo or are reflective of a planter mentality, it might also be the case that high-income voters joined the opposition to the new flag. A third possibility is that many Mississippians considered the 1894 flag to be one of the few remaining emblems of their Southern heritage and, hence, regardless of income, voted to reject the proposed alternative. The latter two possibilities seem remote, however. The expected negative relationship between income and support for the 1894 flag is confounded by African-American voters who, according to the Zogby poll, strongly favored changing the flag’s design. If poor blacks voted for the new flag in numbers sufficient to offset those of poor whites for the old flag, the estimated coefficient on INCOME might turn out to be positive. Low electoral participation by the former would tend to undercut such a sign reversal, however. We nevertheless explore the income-race interaction in more detail below. As the pre-referendum poll suggests, voting on the flag is plausibly a function of a county’s level of educational attainment (COLLEGE). While one would expect support for the new flag to be stronger among more highly educated voters, an alternative description of the racial divisions in society sees the dividing line not as a matter of skin color, but of class: white animosity toward blacks may be more a function of the latter’s comparatively low education levels rather than of their spatial proximity (Oliver and Mendelberg, 2000). We enter BUSH as a proxy for political ideology. It is the percentage of votes cast for George W. Bush in the 2000 presidential election. If the prereferendum polls are any indication, one would expect support for the 1894 flag to be greater in those counties where more voters identify themselves as Republican. Although some political scientists argue that racism is correlated with political party affiliation (e.g., Giles and Hertz, 1994), we resist that interpretation, especially so for Mississippi, where, despite the recent electoral successes of Republican candidates for national office, the Democratic Party still dominates state and local politics, and, moreover, church affiliations frequently are more important than party affiliations. Indeed, the overwhelmingly Democratic Mississippi legislature overwhelmingly favored the old flag (see note 14).
112 Confusing opposition to programs favored by African-American opinion leaders with Republican racism, precedent nevertheless exists in the political science literature for entering BUSH in our model. As Black and Black (1992: 262), for example, put it, From Goldwater’s vote against the Civil Rights Act of 1964 and Nixon’s vocal opposition to “forced busing” in 1968 to Ronald Reagan’s coolness toward civil rights laws in the 1980s and George [H.W.] Bush’s veto of the Civil Rights Act of 1990, Republican presidential nominees and Republican presidents have consistently taken significant positions in opposition to the wishes of most blacks. Primarily, however, our inclusion of BUSH is guided agnostically by the results of the pre-referendum Zogby poll. Population DENSITY is used as a measure of the opportunities for social and racial interaction. Owing to greater proximity, the relations between whites and blacks may either be more discordant or more harmonious in more densely populated counties. It is also possible that, owing to heavy concentrations of poor whites and low electoral participation by poor blacks, rural areas were more resistant to change. That is pure speculation, however; the estimated coefficient on this variable is thus indeterminate a priori. NONWHITE measures the voting behavior of a county’s AfricanAmerican citizens.22 Consistent with the pre-referendum poll, we expect that counties having higher proportions of nonwhites will vote for the new flag and not for the old flag. The fraction of a county’s population born outside Mississippi is used as a proxy for immigration. If non-native voters do not identify with issues perhaps unique to the South in general and to Mississippi in particular, we would expect a negative relationship between BORNOUT and OLD FLAG.23 One argument made by the proponents of the new flag design was that the Confederate battle emblem hampers economic progress by deterring new businesses from relocating to Mississippi. We include BUSINESS to test this conjecture. A negative coefficient on this variable would supply evidence that support for the new flag was greater in counties experiencing economic growth in the recent past. Larger majorities favoring change in such counties could be a result of influxes of newcomers to the state or of local acceptance of the pro-development argument for replacing the 1894 flag. A positive coefficient would point to the opposite conclusions, namely that voters in locations with less success in attracting new businesses either tended to favor the status quo or did not think that the state flag was a significant economic development issue.
113 Interacting NONWHITE and DENSITY in Model 2 allows us to test for racial conciliation or racial conflict. In other words, given the evidence from the Zogby poll that black Mississippians in general said they preferred the new flag, we will attempt to determine if that support tended to be greater or lesser in more densely populated counties. If the sign of the estimated coefficient on the interaction term is positively significant, meaning that larger majorities tended to favor the status quo in such counties, and if African-Americans continue to vote heavily for change, then we may interpret this as evidence of a white backlash. A positive coefficient on the interaction term would be consistent with racial conciliation only if black opposition to the current flag is muted in more urbanized areas of the state where they presumably interact more frequently with whites. A conclusion of harmony between the races would also follow if the estimated coefficient on NONWHITE∗DENSITY is negative in sign, suggesting that fewer votes tended to be cast for the 1894 flag in densely populated black counties. 3.2. Descriptive statistics and regression results The results of Mississippi’s April 2001 special election are shown in Figure 3. Counties where a majority of the votes were cast for the proposed new state are indicated by square dots. A total of 18 counties voted for the new flag and 63 counties voted for the current flag. The vote in Yazoo County ended in a tie: 4,370 voters supported the status quo (Proposition A); exactly the same number expressed a preference for change (Proposition B). Sharp divisions along ideological and geographical lines are evident in the results. Of the 25 counties that returned majorities for Al Gore, the Democratic Party’s presidential candidate in 2000, 16 voted for the new flag and nine for the present-day flag. George W. Bush won 57 of the state’s counties in 2000; nine of these voted for the new flag and 47 voted for the current flag. Thirteen of the 14 counties located along the Mississippi River’s delta that voted for the new flag also returned majorities for Al Gore. In short, the gulf identified by Key more than half a century ago between the delta and the rest of Mississippi apparently persists. The means of the independent variables are reported in Table 1.24 These data suggest that voters in the counties that returned majorities for the 1894 flag tend to have lower levels of education, to have voted for George Bush, to be less likely to be African-American, and to be more likely to have been born outside the state. These same counties also appear to have experienced greater-than-average reductions in business formation, to have higher per capita incomes and lower population densities. Whether these relationships hold up in a multivariate regression framework is a question to which we now turn.
114
Figure 3. April 2001 special election results for Mississippi’s 82 counties. Counties returning majorities for the proposed new state flag are indicated by square dots. The vote in Yazoo County was split evenly between the two flag options.
115 Table 1. Descriptive statistics: Means of the independent variables Variable
COLLEGE (%) BUSH (%) NONWHITE (%) BORNOUT (%) BUSINESS (%) INCOME DENSITY
State of Mississippi as a whole
Counties voting for New Flag
Counties voting for Old Flag
20 28 39 17 –0.0005 $15,856 61
23 21 62 16 0.00006 $15,635 66
15 30 30 27 –0.0007 $15,984 60
Table 2 presents our results. It appears that counties with lower incomes and lower education levels were more likely to vote for the old flag. Majorities in favor of the status quo were also more likely in counties voting for George Bush in the 2000 presidential election. Less densely populated counties and those with smaller population percentages of nonwhites and non-native voters were more likely to vote for the old flag as well. Finally, counties experiencing relatively low rates of business formation tended to vote against changing the state’s flag. Of all of the variables included in Model 1, only DENSITY and BORNOUT are not different from zero at standard levels of statistical significance. No more than modest changes in the precision of the coefficient estimates materialize when the interaction between DENSITY and NONWHITE is added (Model 2). The algebraic signs of all variables remain the same. Compared with Model 1, BUSINESS declines in significance, while BORNOUT becomes significant at the 7 percent level. The interaction term itself enters with a positive sign and is marginally different from zero at the 10 percent level. When placed in the context of the coefficient estimate on NONWHITE, the positive sign on its interaction with DENSITY supplies evidence of racial backlash on the flag issue. Other things being the same, there was more support for the old flag in densely populated black counties. On the other hand, the marginal impacts of NONWHITE in the two models are indistinguishable.25 This suggests that the additional votes for the old flag in densely populated black counties were cast by white voters. In other words, white support for the 1894 flag (and black opposition to it) appears to have been stronger in more urbanized areas of the state where African-Americans comprise larger fractions of the population and the two races presumably interact more frequently.26
0.29023 –0.00001 –0.00297 0.73315 –0.00007 –0.21742 –0.00076 –0.14835
Intercept INCOME COLLEGE BUSH DENSITY NONWHITE BORNOUT BUSINESS NONWHITE∗DENSITY Adjusted R2 F-statistic p-value for overall significance
Coefficient estimate
Variable
<0.0001
8.56 –2.55 –5.12 12.21 –2.03 –8.82 –1.82 –1.41 1.70
Student’s t
Model 2
<0.0001
0.29539 –0.00001 –0.00317 0.72923 –0.00023 –0.23464 –0.00059 –0.13970 0.00041
Coefficient estimate
0.95 180.58
<0.0001 0.0109 <0.0001 <0.0001 0.2775 <0.0001 0.1443 0.0173
p-value (2-tailed)
0.95 200.84
8.34 –2.61 –4.82 12.13 –1.09 –8.72 –1.48 –2.43
Student’s t
Model 1
Table 2. Regression results for the April 2001 flag vote
<0.0001 0.0130 <0.0001 <0.0001 0.0462 <0.0001 0.0729 0.1641 0.0936
p-value (2-tailed)
116
117 Overall, the two regression models explain 95 percent of the variation in votes cast for the 1894 flag across Mississippi’s 82 counties. The evidence suggests that Mississippians were divided sharply along lines of race, class and political ideology when voting on whether to change the state’s flag: more educated, higher income, most likely liberal white voters, joined large numbers of African-Americans in rejecting the 1894 flag, which drew its support primarily from rural, low-income, native Mississippians in counties that voted Republican in 2000.27 That supporting coalition was decisive: when the votes were tabulated, the old flag had defeated the proposed new flag by a margin of nearly two-to-one.
4. Expressive versus instrumental voting Mississippi’s April 2001 special election supplies a unique opportunity for exploring the theory of expressive voting (Brennan and Buchanan, 1984; Brennan and Lomasky, 1993; Brennan and Hamlin, 1998). That theory, proposed as an alternative to orthodox explanations for democratic participation, starts, as do all rational-choice models of the vote motive, by recognizing that any one person’s vote is highly unlikely to be decisive: in a two-candidate race, a single vote determines the winner only if all other votes are evenly divided, the probability of which is vanishingly small. Rather than concluding that the logic of collection action will deter voters from participating in democratic processes, however, the theory of expressive voting suggests that elections offer citizens a low-cost way of registering their personal preferences. Precisely because his vote will not sway the outcome, the voter can afford to take positions that are inconsistent with his own narrow interests. In other words, expressive voters support policies or candidates they would not support if they expected their votes to have instrumental consequences. Empirical evidence on the existence of expressive voting motives is mixed. Copeland and Laband (2002), for example, report evidence that citizens who voluntarily wear their favored candidates’ campaign buttons or display political signage in their yards are more likely to go to the polls on Election Day. Faith and Tollison (1990), by contrast, find more explanatory power in instrumental voting than in expressive voting.28 The April 2001 referendum asked Mississippians simply to register their opinions on the state flag. There was no other issue on the ballot. Other than the unsubstantiated claims of some proponents of change that the 1894 flag raised barriers to Mississippi’s economic development, voters were confronted with a choice having no obvious instrumental consequences. Approximately 40 percent of Mississippi’s registered voters participated in the April 2001 special election (Mitchell, 2001). What motivated these voters
118 to express their preferences? In particular, were the determinants of voter turnout in the flag referendum similar to or different from those in comparable statewide elections where outcomes were more consequential? Answers to these questions are provided in Table 3, which reports the results of estimating Mississippi voter turnout models for three elections: the April 2001 flag vote, the November 1999 gubernatorial election, and the November 2000 presidential election. The dependent variable in each case is the ratio of total votes cast to total voting age population in each of Mississippi’s 82 counties. The explanatory variables are the same as those included in our empirical model of voting on the flag issue, except that, to facilitate cross-equation comparisons, BUSH (George W. Bush’s vote share in 2000) is omitted. Voter turnout is positively related to county per-capita income in all three models, significantly so in the flag referendum and the 2000 presidential election. With the exception of NONWHITE, which enters with a positive, but insignificant sign in the two general elections, it is negatively correlated with all other explanatory variables. Other things being the same, voter turnout is significantly lower in more densely populated counties and, in the two statewide elections, it is significantly lower in counties where more of the voters are not native Mississippians. College-educated voters are less likely to vote, but that relationship is only marginally significant in one model. Economic growth in a county, as measured by the change in the number of businesses located there, does not appear to be a key determinant of voter turnout in Mississippi. The flag referendum differs from the other elections in two important respects. First, race mattered. Other things equal, nonwhite voters were less likely to participate in the special election on the state flag than they were to vote in either the presidential or the gubernatorial election. Indeed, the estimated coefficient on NONWHITE indicates that, at the margin, race was by far the numerically largest determinant of turnout on April 17, 2001. Second, in comparison with the two general elections, the same set of independent variables explains significantly more of the variation in voter turnout across counties for the flag referendum: adjusted for degrees of freedom, the turnout model accounts for 44 percent of the variation in the dependent variable in voting on the flag, whereas it explains 27 percent of the variation in turnout in the gubernatorial election and only 21 percent of the variation in turnout in the presidential election. One explanation for this difference in explanatory power follows from differences in the number of issues on the three ballots. Voters faced only one decision in April 2001 – to replace the 1894 flag or not. (The flag vote took place in unidimensional issue space and thus might be called an “ear-
0.44743 0.00001 –0.00262 –0.00044 –0.15715 –0.00323 –0.12609
Variable
Intercept INCOME COLLEGE DENSITY NONWHITE BORNOUT BUSINESS
Adjusted R2 F-statistic p-value for overall significance
Coefficient estimate
<0.0001
3.11 1.56 –1.15 –3.25 0.93 –2.52 0.53
<0.0001
0.34319 0.00001 –0.00252 –0.00070 0.06278 –0.00257 0.18954 0.28 6.14
<0.0001 0.0722 0.0929 0.0052 0.0014 <0.0001 0.6195
p-value (2-tailed) 0.0027 0.1226 0.2540 0.0018 0.3575 0.0137 0.6010
1999 Gubernatorial Election Coefficient Student’s p-value estimate t (2-tailed)
0.44 11.60
5.78 1.82 –1.70 –2.88 –3.31 –4.54 –0.50
Student’s t
2001 Flag Vote
Table 3. Models of Mississippi voter turnout in three elections
0.35923 0.00002 –0.00199 –0.00062 0.03666 –0.00106 –0.22448
0.0005
0.21 4.60
4.41 2.67 –1.23 –3.88 0.73 –1.41 –0.84
<0.0001 0.0093 0.2217 0.0002 0.4658 0.1631 0.4017
2000 Presidential Election Coefficient Student’s p-value estimate t (2-tailed)
119
120 marked” one.) The other elections were multidimensional, presenting voters with choices on numerous ballot propositions and political contests. Indeed, more than two candidates vied for the political offices in the races at the tops of the ballots in November 1999 and November 2000. Democratic participation in single-issue elections evidently is easier to explain than it is in general elections. An alternative interpretation of these results is that voters tend to be more polarized – their positions in issue space are both farther apart and more rigid – when voting expressively than when voting instrumentally. Strong positional separation in expressive voting situations suggests that voters’ preferences are bimodal rather than unimodal. The explanatory power of turnout models predictably increases in the presence of such bimodality. Finally, it is worth emphasizing that expressive voting seems to be amenable to empirical explanation by the same variables routinely employed in modeling voter turnout in elections with instrumental consequences. That observation, which points to the difficulty of disentangling the two vote motives empirically, has fundamental implications for the theory of voting in general. It also points the way to future empirical testing on larger datasets, with observations collected across states or nations, which include both single-issue and multi-issue elections.
5. Concluding remarks This paper has studied the special election held in the State of Mississippi on April 17, 2001, asking voters to express their preferences on the design of the state’s flag. Should the flag continue to incorporate a symbol of the Confederacy in its canton corner, or should that symbol – offensive to many African-Americans but exemplifying fading southern heritage to others – be replaced by a field of stars representing Mississippi’s accession as the Union’s 20th state? Despite heavy lobbying by the proponents of change, the results were unambiguous: 64 percent of the Mississippians who turned out on Election Day voted in favor of retaining the 1894 flag. Using the election returns for the state’s 82 counties, we find that support for the proposed new flag was concentrated in the Mississippi River’s delta, an area that is predominately black, rural, poor, and Democratic in its politics. College-educated Mississippians and those born out of state were also more likely to vote for the new flag. Evidence that support for the 1894 flag tended to be higher in densely populated black counties contributes additional evidence that voters were divided sharply along lines of race, class, and political ideology. A comparison of voter turnout rates in the April 2001 special election with those for the state’s 1999 gubernatorial election and the 2000 pres-
121 idential election pointed to two important conclusions. One is that voter positions in issue space are more polarized when political choices have expressive as opposed to instrumental consequences, the other that expressive voting seems to be amenable to empirical explanation by the same variables routinely employed in modeling voter turnout in elections with instrumental consequences.
Notes 1. Mississippi Division of the United Sons of Confederate Veterans v. Mississippi State Conference of NAACP Branches, et al., 774 So.2d 388 (Miss. 2000). A subsequent motion for rehearing, filed on May 16, 2000, was denied by the Court on December 29, 2000, which issued its decision and order on January 10, 2001. The appellant in the matter before Mississippi’s Supreme Court was permitted by the trial judge (the Honorable W. O. Dillard of the Hinds County Chancery Court) to intervene as a defendant (Ibid.: 389). 2. Ibid. 3. Ibid., p. 390. 4. Daniels v. Harrison County Bd. of Sup’rs, 722 So.2d 136, 137 (Miss. 1998). 5. Ibid., p. 139. 6. The plaintiffs argued “that the trial court had erred in failing to apply the doctrine of inclusio unius est exclusio alterius in its interpretation of Miss. Code Ann. § 3-3-15 (1972)”. In particular, they contended that, because the relevant statute does not expressly provide for the display of any flag other than the state flag, it should be construed as prohibiting the flying of the Confederate battle flag. The Supreme Court held that “this argument is without merit” (Ibid.: 138). 7. Ibid., p. 141; emphasis added. Daniels, 774 So.2d 388, 389 (Miss. 2000), stands for the proposition that “the decision to fly or adopt a state flag rests entirely with the political branches”. 8. 774 So.2d 388, 391–92 (Miss. 2000). 9. Ibid., p. 392. 10. Ibid. This is not the first time Mississippi has been without an official state flag: “Mississippi did not officially adopt a state flag until 1861, when it seceded from the United States and joined the Confederate States of America”. See 774 So.2d 388, 390 (Miss. 2000) and Sansing (n.d.). 11. In NAACP v. Hunt, 891 F.2d 1555 (11 Cir. 1990), the court rejected an earlier attempt “to bar the display of the confederate flag by the State of Alabama”, concluding that “the federal judiciary is not empowered to make decisions based on social sensitivity”. Like the Mississippi Supreme Court, the Court of Appeals held in Hunt that disputes about the battle flag are political matters, “the remedy for which lies within the democratic process”, quoted in 722 So.2d 136, 138 (Miss. 1998). The NAACP has actively lobbied to redesign Mississippi’s banner at least since 1988, when “the late Aaron Henry, a member of the Mississippi Legislature and president of the Mississippi Conference of the NAACP, introduced a bill to remove the battle flag from the state flag . . . . This bill was never brought to the floor for a vote, nor were any of the others he introduced in 1990, 1992, and 1993” (Sansing, n.d.). 12. Justice Banks framed the issue in the strongest possible terms when he wrote in Daniels, 722 So.2d 136, 140 (Miss. 1998), that “the Confederate battle flag conjures up images of
122
13.
14.
15.
16.
17. 18. 19. 20.
21.
22.
and is popularly used to symbolize slavery, white supremacy, racism and oppression. It takes no back seat to the Nazi Swastika in this regard”. Following the adoption of Mississippi’s post-Reconstruction constitution in 1890, the joint legislative committee charged with designing the present-day state flag avoided all references to Confederate symbols. The committee’s report in fact described the 13 stars overlaying the St. Andrew’s cross depicted in the flag’s canton corner as “corresponding with the number of the original states of the Union”. “How many Mississippians”, Justice Banks asked, “are aware of the fact that the thirteen stars on the Mississippi flag purportedly represent, not the states of the Confederacy, but the original thirteen states of the Union?” 722 So.2d 136, 139–40 (Miss. 1998); emphasis in original. Although there were only 11 official states of the Confederacy, stars representing Kentucky and Missouri, which “sent delegates to the Confederate Congress and supplied troops to the Rebel army” are included in the battle flag (Sansing, n.d.). While the Joint Legislative Committee for a State Flag failed to acknowledge the obvious historical symbolism, Fayssoux Scudder Corneil, the daughter of the Mississippi State Senator credited with designing the 1894 flag, said in a 1924 address that her father “wanted to perpetuate in a legal and lasting way that dear battle flag under which so many of our people had so gloriously fought” (Ibid.). In a story appearing in the Jackson, Miss., Clarion-Ledger on July 6, 2001, “Tunica lawyer Greg Stewart, who’s pushing to put the 1894 flag in the state constitution, predicted lawmakers will never touch the flag question”. Perhaps that is because there was no doubt about the outcome: according to House Speaker Pro tem Robert Clark, “if lawmakers had voted on the flag, 90 percent would have backed the old flag” (Mitchell, 2001). The career of Louisiana’s David Duke is the poster child of much of the literature addressing the “racial threat” hypothesis. See Kuzenski, Bullock and Gaddie (1995) for detailed analyses of Duke’s political rise and fall. By contrast, groups opposing the new flag, led by the Sons of Confederate Veterans, spent about $125,000. See Mississippi votes Tuesday on state flag, Inside Politics, 16 April 2001, http://www.cnn.com/2001/ALLPOLITICS/04/16/Mississippi.flag/?s=2 and Mississippi votes 2–1 to keep existing flag, Inside Politics, 17 April 2001, http://www.cnn.com/2001/ALLPOLITICS/04/17/mississippi.flag/?s=2. Mississippi votes 2–1 to keep existing flag. Ibid. Statewide survey of Mississippi voters on flag issue, Clarion-Ledger (Jackson, Miss.), 8 February 2001, http://www.clarionledger.com/news/0102/08/08fullpoll.html. The URLs for our data sources are http://www.sos.state.ms.us/elections/FlagVote/ CountyScans/CertStateResults.pdf and http://www.olemiss.edu/depts/sdc. The latter site contains the 2000 U.S. Census data for Mississippi. A county’s voting age population is entered in the denominator of the dependent variable because information on voter registrations is not maintained in any systematic fashion by Mississippi’s Secretary of State. The available county voter registration lists appear to be highly unreliable in any case. Coahoma County, for instance, ostensibly has 31,765 registered voters; the total population of the county is 30,622. Hence, we do not have access to the numbers of the white registered voters voting for the old flag or for the new flag, nor do we have any data on party affiliation by race. As reported below, however, our empirical results are not influenced materially by the manner in which the dependent variable is defined. Although NONWHITE includes races other than African-Americans, the population percentages of these other races in Mississippi’s counties are negligible. We note that African-Americans tend to be underrepresented as registered voters. Defining NON-
123
23. 24. 25. 26.
27.
28.
WHITE as a percentage of a county’s voting age population thus tends to overestimate the fraction of blacks (and other racial minorities) eligible to vote in the special election. This bias is unavoidable given the unreliability of voter registration lists in Mississippi (see note 21). It would have been helpful to identify the demographic characteristics of this group. Such detailed information is not available, however. The observations on all variables are from 2000, except for the dependent variable (2001) and INCOME (1999). At the variable mean, the marginal impact of NONWHITE in Model 2 is –0.23464 + 0.00041(39) = –0.21865. Diagnostic tests on both regression models produced no evidence of heteroscedasticity. Moreover, the variance inflation factors were all low, indicating that multicollinearity is not a problem, even when the interaction term in included. Hence, the possibility that a high partial correlation between REPUBLICAN and INCOME (r = .49) masks a pure income effect in multivariate models of voting on the flag is not evident in our data. The same conclusion holds for any other variable pair. In order to explore further the influence of race and income on votes cast for the 1894 flag, we re-estimated the first model in Table 2 by including a new explanatory variable, INCOME∗NONWHITE, the interaction between per capita income in a county and the county’s population percentage of voting-age minority races. The estimated coefficient on that variable was statistically indistinguishable from zero, while the algebraic signs and statistical significance of the other independent variables were largely unchanged. What is most important about these results is that INCOME and NONWHITE both remained negative and significant. This empirical finding suggests that black Mississippians tended to vote against the old flag regardless of income or, to put it differently, that support for the 1894 flag originated primarily from low-income whites. To address the criticism that the use of a county’s voting age population in the denominator of the dependent variable confounds a citizen’s decision to vote (electoral participation) with his expression of preference on the flag’s design, we re-estimated our regression models using two alternative definitions that use information only about the votes actually cast in the special election: OLD FLAG/NEW FLAG and OLD FLAG/(OLD FLAG + NEW FLAG), where OLD FLAG and NEW FLAG are the votes cast in a country for Propositions A and B, respectively. The results from both specifications were quantitatively similar to those reported in Table 2. See Carter and Guerette (1992), Fischer (1996) and Sobel and Wagner (2002) for additional empirical tests of the expressive voting model.
References Black, E. and Black, M. (1992). The vital South. Cambridge: Harvard University Press. Brennan, H.G. and Buchanan, J.M. (1984). Voter choice: Evaluating political alternatives. American Behavioral Scientist 28: 185–201. Brennan, H.G. and Hamlin, A. (1998). Expressive voting and electoral equilibrium. Public Choice 95: 149–175. Brennan, H.G. and Lomasky, L.E. (1993). Democracy and decision: The pure theory of electoral preference. Cambridge: Cambridge University Press.
124 Carsey, T. (1995). The contextual effects of race on white voter behavior. Journal of Politics 57: 221–228. Carter, J.R. and Guerette, S.D. (1992). An experimental study of expressive voting. Public Choice 73: 251–260. Copeland, C. and Laband, D.N. (2002). Expressiveness in voting. Public Choice 110: 351– 363. Coski, J.M. (2000). The Confederate battle flag in historical perspective. In Martinez, J.M., Richardson, W.D. and McNinch-Su, R. (Eds.), Confederate symbols in the contemporary South, 89–129. Gainesville: University Press of Florida. Faith, R.L. and Tollison, R.D. (1990). Expressive versus economic voting. In Crain, W.M. and Tollison, R.D. (Eds.), Predicting politics: Essays in empirical public choice, 231–244. Ann Arbor: University of Michigan Press. Fischer, A.J. (1996). A further experimental study of expressive voting. Public Choice 88: 171–184. Giles, M. and Hertz, K. (1994). Racial threat and partisan identification. American Political Science Review 88: 317–326. Holmes, R. and Cagle, M.C. (2000). The great debate: White support for and black opposition to the Confederate battle flag. In Martinez, J.M., Richardson, W.D. and McNinch-Su, R. (Eds.), Confederate symbols in the contemporary South, 281–302. Gainesville: University Press of Florida. Key, V.O. (1949). Southern politics in state and nation. New York: Alfred A. Knopf. Kuzenski, J.C., Bullock, C.S. III and Gaddie, R.K. (Eds.). (1995). David Duke and the politics of race in the South. Nashville and London: Vanderbilt University Press. Mitchell, J. (2001). Mfume leaves open option to boycott state over flag. Clarion-Ledger (Jackson, Miss.), 6 July, http://www.clarionledger.com/news/0107/06/m01.html. Office of External Affairs, Division of Research. (1999). Mississippi statistical abstract. Starkville, MS: Mississippi State University. Oliver, J.E. and Mendelberg, T. (2000). Reconsidering the environmental determinants of racial attitudes. American Journal of Political Science 44: 574–589. Orey, B.D’A. (1998). The race race in black and white: The 1995 Louisiana gubernatorial election. Southeastern Political Review 26: 908–920. Sadow, J.D. (1996). David Duke and black threat: Laying to rest an old hypothesis, revisited. American Review of Politics 17: 59–68. Sansing, D.G. (n.d.). Flags over Mississippi. Mississippi history now, http://mshistory. k12.ms.us/features/feature2/flagsoverms.html. Sobel, R.S. and Wagner, G.A. (2002). Expressive voting and government redistribution. Unpublished manuscript. Voss, S. (1996). Beyond racial threat: Failure of an old hypothesis in the new South. Journal of Politics 58: 1156–1170.