Who Wins

In Table 2, I provide the first analysis of the fate of different groups across American elections. The table shows the proportion of voters who end up supporting losing candidates in presidential, Senate, gubernatorial, House, and mayoral elections broken down by a range of core demographic and political characteristics. The last column also shows the proportion of voters from each group that voted for the loser in presidential, senatorial, and gubernatorial elections in a single year—a group that I refer to as “superlosers.”

The analysis offers a range of interesting insights into group outcomes. The first and probably the most obvious conclusion is that no one group is totally shut out from the winning side of electoral democracy. Across all of the demographic and political groups and all of the types of elections, the worst outcome for any single group is to end up on the losing side of the vote 65% of the time. In most cases, the majority of voters from each group end up on the winning side of the vote. America's democracy may be divided, but it is not the case that an overwhelming majority consistently wins out against united minority voting blocs.

This does not mean that all groups win equally often. The top of the table reveals a clear racial hierarchy with whites on the top and African Americans on the bottom. During this period, blacks are the least successful racial, demographic, or political group in American elections. Blacks are the only group that loses more than half the time in most contests. A majority of all black voters ends up on the losing side in presidential (59%), Senate (55%), gubernatorial (52%), and mayoral elections (53%). Aside from Latinos, no other group (racial, political, or demographic) loses more than half the time in more than one type of contest. The one important exception is House elections where some 74% of black voters end up getting their favored candidate into office. I discuss the distinctiveness of House elections in more depth below.

The limited success of African Americans in mayoral elections is especially striking. These elections occur in cities where blacks make up a much larger share of the electorate (27%) than at the national level and where the electorate tends to be significantly more Democratic than the national population (64% of voters in these cities chose John Kerry in 2004). If black voters were going to be on the winning side anywhere, one would expect it to be in these disproportionately black and liberal cities. The fact that the vast majority of these mayoral elections are nonpartisan also suggests that the black defeat in mayoral elections should not be entirely ascribed to a pro-Republican tilt in American politics during this time period.

At the other end of the racial spectrum, white voters stand out for their consistent success. In particular, for four of the five different kinds of contests, white voters are substantially more likely than black and Hispanic voters to end up winners. Whites win half or more of the time in every contest, and in most cases close to 60% of white voters end up winners. This leads to a fairly substantial racial gap. The difference between white and black success is 13 points for the Senate, 13 points for the mayoralty, 12 points in gubernatorial contests, and 9 points in presidential contests.

Latino voters fare almost as poorly as blacks. In three of the five types of contests, a slim majority of Latino voters ends up on the losing side and in two sets of these elections Latinos lose just as often as blacks. Latinos fare substantially better than blacks in only one type of contest. In mayoral elections a slim majority of Latinos ends up winners whereas a slim majority of African American voters ends up losers in these contests. Asian Americans are the most difficult racial group to characterize since they have the most mixed outcomes—sometimes surpassing whites and sometimes losing more than any other racial group.

In some ways, these figures understate the depth of America's racial hierarchy. By looking at each type of election in isolation, they cannot tell us how often individual voters lose across multiple elections. If, however, one looks simultaneously at an individual's vote in three elections (presidential, senatatorial, and gubernatorial) as is done in the last column of Table 2, the consistency with which African Americans lose is even more apparent.Footnote ¹⁶ Overall, 41% of all black voters can be characterized as superlosers, meaning that they choose the looser in all three contests. By contrast, only 9% of whites, 4% of Latinos, and 14% of Asian Americans end as superlosers. By this measure, blacks stand out from every other demographic or political group in America. The next least successful voters are liberals and Democrats who are, respectively, superlosers only 20% and 24% of the time. Blacks also stand out in terms of how often they win. The average black voter is a winner only 0.76 times across the three contests. That is far below the figure for whites (1.67 wins on average), the figure for Latinos (1.87 wins), and the figure for Asian Americans (1.14 wins).

The fact that, across this range of different contests, blacks are consistently more likely to end up losers raises concerns about equity in American democracy. This pattern seems to fit classic accounts of minority vote dilution, which is defined as “elections that permit a bloc voting majority, over a substantial period of time, consistently to defeat candidates publicly identified with the interests of and supported by a politically cohesive, geographically insular racial or ethnic minority group” (Davidson and Grofman Reference Davidson and Grofman1994, 231). For the many individual black voters who lose time and time again across the spectrum of American elections, one could reasonably argue that American democracy is not serving their interests.

Importantly, none of the other demographic factors seems to have a consistent impact on winning and losing in American elections. Every demographic group in Table 2 wins more than half of the time in at least three of the four contests. Class, in particular, is not tied to losing. Members of the lower class—as measured by income and education—are no more likely to end up on the losing side of democracy. Similarly, there is no consistent pattern in outcomes for age, gender, religion, urbanicity, or sexual orientation. City residents, for example, may have fared worse than any other group in these presidential contests but they ended up as regular winners in the other three types of contests. Further tests also reveal no connection between region or employment status and electoral victory.

There are, however, some signs of a rightward tilt in outcomes during this time period. In these presidential elections, most liberals (60%) and most Democrats (64%) end up on the losing side of the vote. The partisan or ideological skew is not, however, nearly as consistent as the racial skew. In the other three types of elections for which data are available, most Democrats and liberals end up winners. Moreover, in these Senate, gubernatorial, and House elections, those on the left and those on the right win at about the same rate.

Is it Race?

Findings about the rightward tilt of American democracy raise an important question. Are black voters simply losing because they happen to be liberal and Democratic in a time when conservatives and Republicans are more frequently coming out on top? Or is there something about blacks and their place in American democracy that sets them apart and leads to consistent losing? In other words, does race have anything to do with it?

To isolate the independent effect of race, in Table 3, I examined the probability that each respondent ended up on the winning side of these elections, controlling for a range of individual characteristics. The dependent variable is whether the respondent voted for the losing side (in other words, whether she voted against the majority of all voters in that particular election). For this analysis, I pool the responses of every respondent for all elections of the same type (e.g., all Senate contests in all VNS years). As independent variables, I include dummy variables for each racial and ethnic minority group (black, Latino, Asian American), with non-Hispanic whites forming the residual category. To see if it is really race, I also include controls for political ideology, partisanship, and a range of other measures of socioeconomic status, and region in the logistic regressions.Footnote ¹⁷

The results are consistent and clear. Even after controlling for political views and party identification, African American voters are significantly more likely than whites to end up on the losing side of democracy for every type of election.Footnote ¹⁸ Moreover, the gaps are reasonably large. Since it is difficult to interpret the magnitude of these effects from the logit coefficients, I converted the coefficients into probabilities that members of each group will wind up on the losing side of the vote (King, Tomz, and Wittenberg Reference King, Tomz and Wittenberg2000). In each case I hold all of the other variables constant at values for a hypothetical median voter. These calculations indicate that blacks are 11% more likely than whites to end up losers in Senate contests, 6% more likely to do so in gubernatorial elections, 5% more likely to lose in House contests, and 4% more likely to lose in presidential elections.Footnote ¹⁹ This doesn't mean that all blacks lose. But many do.Footnote ²⁰

No other political or demographic group ends up consistently as losers across the four types of elections. In terms of race and ethnicity, the results for Asian Americans and Latinos are quite mixed. In two cases (Senate and House elections), Latinos are substantially more likely than whites to end up losers (2% and 13% more likely, respectively) but in one other case (gubernatorial elections) they are less likely to end up losers (4%), and in one other case they are about as likely as whites to be on the winning side. Asian American outcomes are similarly mixed.

Importantly, during a time period when Republican officeholders did well, party and ideology do not play a clear role in this analysis. Democrats are more likely than Republicans to end up on the losing side of presidential democracy. But the effects are not particularly large (3% more likely to lose). And in one case (House elections) Democrats fare marginally better than Republicans (6% less likely to lose), all else equal. In addition, in the other two types of elections, Democrats are no more or less likely to lose than Republicans. Similarly, liberals are disproportionate losers in one case (presidential elections), disproportionate winners in another case (House elections), and on par with conservatives in the other two types of contests. The effects for income and other socioeconomic characteristics are also decidedly mixed. It is African Americans and only African Americans who always end up losing disproportionately.

Another way to assess whether race is the central factor driving outcomes is to look at different subtypes of voters. Are outcomes for lower-class whites, for example, different from outcomes for middle class whites? And more generally, is there a “type” of white voter that loses as consistently as black voters? To answer this question, I examined outcomes for a wide range of groups of white voters that combined different characteristics of education, income, religion, and political orientation. Although there were marginal differences and some groups such as white liberals were somewhat less successful than the median white voter, no combination of characteristics led to losing on par with black voters.Footnote ²¹

These results raise even more questions about the representation of African Americans in American democracy. Does a group that consistently loses more regularly than all other groups have enough of a say in the political arena to be consider well represented? Ultimately, these kinds of questions are normative ones that require subjective answers. But the fact that blacks are substantially more likely than whites and other racial ethnic groups to be outvoted in American democracy across a wide spectrum of elections suggests that there may be a problem.

Validating a Count of Winners and Losers

Selective Turnout. One concern with counting votes is that selective turnout could skew an analysis of winners and losers.Footnote ²² The regressions in Table 3 do control for overall voter turnout in each election but they do not control for selective turnout by different groups. It might be that members of a group who feel alienated from the menu of candidates or who feel that their favored candidate has little chance of winning might choose not to vote at all.Footnote ²³ If true, a count of winners and losers could overstate the representation of a group.

Additional analysis indicates, however, that selective turnout has at best a marginal effect in these elections. First, supplementary tests show that each racial and ethnic group's turnout is unrelated to its success rate. Specifically, after acquiring the turnout rate by race for each general election from the CPS, I found that each group is not significantly more (or less) successful when its turnout is low than when its turnout is high. I arrive at similarly insignificant results if I look at relative turnout rates (e.g., black turnout minus white turnout in a given election).

Another way to demonstrate that strategic nonvoting has little effect on the relative success of different groups is to directly compare changes in turnout across groups and elections. For turnout to affect the relative success of a group, members of that group must be more strategic than another group in their decision to turn out. The data, however, reveal no such pattern. Black, white, and Latino turnout does vary from year to year but all three racial groups follow an almost identical pattern. When one group turns out more in one year, the other groups also turn out more.Footnote ²⁴ Across the years in the data set, black, white, and Latino turnout rates are correlated at a minimum of 0.95. There is, in short, little evidence of selective turnout by race in these elections. Importantly, the structure of American elections limits the likely impact of strategic nonvoting. Since the electorate votes on multiple offices simultaneously, there is less room for strategic nonvoting in one contest. If you turn out for one contest (e.g., the president), you generally vote in the other contests (e.g., Congress)—whether the outcome in the second contest is likely to be in your favor or not.

A third way to show that strategic turnout is not widespread is to compare the candidate preferences of voters and nonvoters across a variety of different contexts. As I noted earlier, there is ample evidence from a range of different elections and cases indicating that the preferences of nonvoters are similar to the preferences of voters (Bennett and Resnick Reference Bennett and Resnick1990; Ellcessor and Leighley Reference Ellcessor, Leighley and Menifield2001; Highton and Wolfinger Reference Highton and Wolfinger2001; Norrander Reference Norrander1989). This is even true if we compare voters and nonvoters from the same racial or ethnic group (Hajnal and Trounstine Reference Hajnal and Trounstine2005). To supplement this literature, I turned to two alternative data sets, the NAES, which has data on the choices of voters and the preferences of nonvoters in general and primary elections for president in 2000 and 2004, and the General Social Survey, which has data on presidential preferences among voters and nonvoters from 1968 to 2000. In this limited set of elections, there are few substantial differences between the preferences of voters and nonvoters in either data set. Nonvoters would have chosen the same winners in roughly the same proportions. Moreover, differences between voters and nonvoters within the same racial group (or demographic or political group) tended to be even smaller and were generally insignificant.Footnote ²⁵ Including the preferences of nonvoters would likely alter a count of winners and losers on the margins, but the overall conclusions would be nearly identical.

Primary Elections. Another concern is that almost all of the results to this point focus on general elections while ignoring primary election outcomes. Although there is a logic to focusing on general elections—winning a primary matters little if you ultimately lose in the general election—primary elections can also affect the kind of representation minorities have. Groups that are consistent losers in primary elections are less likely to have a meaningful choice in general elections and may be forced to choose the lesser of two evils.

Analysis of primaries at the mayoral and presidential levels reveals a mixed picture. The same racial hierarchy that we saw in general elections also emerges in mayoral primaries. African American voters in the typical Democratic primary in the mayoral data set lose 48% of the time compared to 37%, 41%, and 38% of the time for whites, Latinos, and Asian Americans, respectively.Footnote ²⁶ At least at the mayoral level, African Americans are failing disproportionately at both stages of the electoral process. They are, in essence, doubly disadvantaged. The pattern in presidential elections is, however, very different. Here, African Americans—and other minorities—are particularly apt to end up winners in the primary.Footnote ²⁷ The limited available data suggest that blacks are marginally more successful than whites—winning 43% of the time versus 40% of the time for whites.Footnote ²⁸ Latinos (48% winners) and Asian Americans (42% winners) also fare better than white voters at this stage. This pattern suggests that African Americans (and other minorities) will often have a favored candidate in the general election. The problem for African Americans and to a lesser extent for Latinos is that even when they have a meaningful choice in the general election, at the end of the day they still regularly wind up losers. Much more work needs to be done across a broader array of primary contests before we can offer anything close to an overarching assessment of primaries in American democracy, but none of these limited primary results would lead one to dismiss the concerns that emerge out of the general election analysis.Footnote ²⁹

Comparing Different Measures of Representation. Another very different way to assess this new measure of representation is to compare its results with assessments of minority representation based on existing measures. If we look first at descriptive representation and compare outcomes for several key demographic groups, it is apparent that a count of winners and losers is correlated with but distinct from a count of descriptive representation.

For African Americans, the parallels between the two measures are striking. Blacks fare poorly on both measures for all offices except the House of Representatives. In terms of descriptive representation, black elected officials account for only 2% of the nation's governors, 1% of its senators, and 2% of its mayors, but almost 10% of House seats (Hajnal Reference Hajnal2007; JCPS 2003; MacManus and Bullock Reference MacManus and Bullock1993). Similarly, most black voters lose in contests for governor, senator, and mayor, but the vast majority (71%) end up winners in House elections. The one difference between the two measures is that a count of black elected officials suggests that when blacks are underrepresented, they are grossly underrepresented whereas a count of winners and losers suggests that blacks are not that severely undercut by democratic elections. The reason for the difference in this case is simple. A count of winners views blacks who vote for successful white candidates as winners. By contrast, descriptive representation views these same black voters as being unrepresented. Since the bulk of the black population lives in majority-white states and districts and votes in elections without black candidates, descriptive representation is dismissing most of the voting done by African Americans as meaningless. As such, it may be ignoring most of the influence that African American voters have in elections across the country.

The case of gender provides a sharper contrast between the same two measures. In a count of winners and losers, women fare only marginally worse than men. By contrast, a count of descriptive representation indicates that women are greatly underrepresented in all of these offices.Footnote ³⁰ Again, it is not clear which measure provides the more telling account, but the contrast does highlight the fact that the two measures are likely to diverge when the group in question does not vote cohesively. Female voters don't lose in overwhelming numbers in a count of winners and losers because they seldom vote in overwhelming numbers for one candidate over another.

A comparison of a count of winners and substantive representation is more difficult because we have few overall measures of substantive representation for a group in any given democracy. We know that black constituents do have influence on the voting patterns of their legislators (Griffin and Newman Reference Griffin and Newman2007; Whitby Reference Whitby1997), there is evidence that expanded black voting has altered policy outcomes in a given state (Keech Reference Keech1968; Parker Reference Parker1990), and there are studies that link the election of black elected officials with a host of symbolic and concrete effects (Browning, Marshall, and Tabb Reference Browning, Marshall and Tabb1984; Karnig and Welch Reference Karnig and Welch1980). But it is extremely difficult to say whether African Americans are, as a whole, being well represented in American democracy from studies that focus on the actions of a few leaders across a small number of policy arenas.Footnote ³¹ Scholars do offer broader assessments of black substantive representation but because of the imprecise way in which both black opinions and policy outcomes are measured, the accounts differ wildly. Thus, while one set of scholars can claim that both of the nation's major political parties ignore the interests of black voters when deciding policy (Frymer Reference Frymer1999; Walters Reference Walters1988), others can claim not only that blacks are often well represented but also that electoral rules give African Americans special advantages (Swain Reference Swain1995; Thernstrom Reference Thernstrom1987; Thernstrom and Thernstrom Reference Thernstrom and Thernstrom1997).Footnote ³²

A sharper distinction between substantive representation and a count of winners and losers might be made in the case of gay and lesbian representation. By a count of winners, gays and lesbians fare reasonably well, winning between 52% and 57% of the time in the elections in this study. But it would be hard to maintain that policy outcomes in America favor gay and lesbian interests on the most important issues facing the two groups. Fundamental rights such as the right to marry are often not recognized in American democracy. The insight here might be that it is important to assess winners and losers in elections that are of particular importance to the group in question. For gay and lesbian voters, we might want to focus on statewide propositions concerning gay rights.

It is also informative to compare all three measures in other historical periods and across different geographic contexts. A focus on the representation of blacks in the South after the Great Depression would likely conclude that blacks fared extremely poorly in terms of substantive representation (blacks were barred from white schools and many jobs and had little access to the vote) and descriptive representation (black officeholding was negligible), but well in a count of winners and losers since the few black voters that did make it to the polls often won (blacks were starting to vote Democratic in an era of Democratic domination). Roughly the same contrast might emerge if we shifted the focus to the Apartheid era in South Africa. These historical and geographic comparisons suggest that a count of winners and losers is only appropriate if the group in question has broad access to the vote.

These few paragraphs offer a far from systematic and exhaustive comparison of how these measures fare in different circumstances, but they do help to recognize some of the qualities and limitations of counting winners. In terms of limitations, these comparisons suggest that a count of winners should be limited to cases where there is equal access to political participation and should focus on elections that are important to the group in question. At the same time these comparisons also help to establish the validity of counting winners by demonstrating some level of correlation with existing measures. They also suggest that a count of winners can incorporate different kinds of minority influence that other measures fail to capture. Based on these insights, it is clear that we should not perceive a count of winners and losers as anything close to a perfect measure of representation. Instead, we should view a count of winners as a valuable new tool that compliments rather than replaces existing measures of representation. What that tool is telling us is that even today, African Americans appear to be less well represented in American democracy than most other groups.

Why Do Blacks Lose?

The disproportionate failure of African American voters in American democracy leads to an important question. Why are black voters losing more regularly than other voters? One part of the answer is surely institutional. By all accounts, the fact that almost all American elections are winner-take-all likely diminishes black influence (Davidson and Grofman Reference Davidson and Grofman1994; Guinier Reference Guinier1994). Comparative studies suggest that countries such as the United States that have majoritarian systems tend to have fewer minorities in office and more limited spending on minority-friendly policies than countries with proportional representation (Lijphart 1994; Taagepara and Shugart Reference Taagepara and Shugart1989). Within the American context, analysis of the few elections that are held with alternative voting methods suggest that minority descriptive representation is greatly enhanced under many of these alternative electoral systems (Bowler, Donovan, and Brockington Reference Bowler, Donovan and Brockington2003). There are also a number of more specific local electoral institutions that have been linked in recent decades to reduced black representation. In particular, black representation continues to be impaired by at-large elections (Welch Reference Welch1990), off-cycle local elections (Hajnal and Trounstine Reference Hajnal and Trounstine2005), reduced council size (Bullock and MacManus Reference Bullock and MacManus1987), and other electoral barriers. If we are concerned about the underrepresentation of black voters, reform to each of these institutional arrangements should be considered.

Another factor that, at least intuitively, could do even more to explain the defeat of black voters is a pattern of racial bloc voting.Footnote ³³ The potential relevance of bloc voting for any minority group is obvious. No minority can overcome the will of a united and active majority in a democracy. But do blacks really lose out because of bloc voting? If bloc voting is to blame, we would see three underlying patterns. First, the minority group must, to a certain extent, vote as a unified block. If members of a minority group do not agree on what policies should be passed or who should be elected, then large numbers of that group will necessarily end up on both the winning and the losing side. Second, the majority group—or white voters in this case—must be cohesive. If the white majority is fairly evenly divided, then nonwhite minorities will be able to influence the outcome of many elections. There can be no tyranny of the majority without a clear majority preference. Finally, the preferences of the minority and majority groups must diverge. The interests of white and nonwhite voters must be opposed. If most voters agree on which candidates should be elected, regardless of their race or ethnicity, then clearly there is no tyranny of the white majority.

In the next two tables, I examine the vote by race and ethnicity across elections in American democracy to see how well these conditions are met. The results suggest that the American electorate is divided by race and that blacks are uniquely disadvantaged by these bloc voting patterns.

Cohesion. The figures in Table 4 indicate the degree to which each racial and ethnic community votes as a bloc. To measure the cohesiveness of a group, I measured the proportion of the voters from each racial group that supported their group's most favored candidate. If a group was wholly united and all voted for one candidate, the measure would equal 100. A totally divided community that evenly split their vote between two candidates would score 50. All of the elections for a given office are averaged together.

The first conclusion to emerge from this analysis is that African Americans do tend to vote as a bloc—very much raising the possibility of tyranny of the majority. In a typical electoral contest, the vast majority of black voters end up on the same side. Despite arguments about the declining significance of race in American politics and life and growing class divisions within the black community, electoral politics still appears to bring blacks together. In presidential contests, for example, 90% of all black voters support the same candidate in the typical contest. Black cohesion is as high or almost as high in senate, gubernatorial, and House elections. Black unity in mayoral contests suggests that this black cohesion is not just a function of partisan labels or partisanship. In the mayoral contests, the vast majority of which are nonpartisan, blacks are still exceptionally cohesive. On average, 82% of black voters support the group's favored mayoral candidate. In short, it is generally clear who the “black” candidate is, and the vast majority of the community supports that candidate.

Unfortunately for African American voters, white voters are also somewhat cohesive. Although white cohesion does not come close to matching black cohesion, in all but one set of elections at least 60% of whites voted together. In mayoral elections, the figure rises to 71% cohesion. Only in presidential contests where 54% of white voters favor their group's preferred candidate are whites close to evenly divided. These patterns suggest that if blacks favor candidates that the white majority opposes, they are likely to lose heavily in most contests.

Another interesting conclusion that emerges out of Table 4 is that it is possible to talk about Asian American and Latino voting blocs. Even Asian Americans, generally the least cohesive group, vote together much more than chance would suggest. In the typical senatorial contest, for example, 68% of Asian Americans voted together. In other contests, Asian American cohesion ranges from 59% to 79%. Latinos are slightly more apt to vote together. Across the five types of elections, between 64% and 79% of Latino voters supported the same side on average. This suggests that the Asian American and Latino communities are often able to at least partially overcome differences of national origin group, immigration status, and socioeconomic status. We still cannot think of Asian Americans and Latinos as monolithic voting blocs, but we should probably consider them as more of a voting bloc than some recent accounts suggest.

Division. It is also clear that American elections generate considerable racial and ethnic division. Table 5 presents figures for the average divide between each set of racial and ethnic groups across the different types of contests. The table shows the average difference in the percentage favoring the winning candidate. As Table 5 illustrates, American democracy often pits racial and ethnic groups against each other. Across the entire set of elections, the smallest average divide is 10 percentage points between whites and Asian Americans in presidential contests. The typical racial divide is over a 20-percentage-point gap. Racial and ethnic communities are not only voting cohesively, they are also voting against each other.

The more important story revealed by Table 5—at least in terms of understanding the poor outcomes for black voters in American democracy—is the especially sharp divide between black and white voters. On average, the black vote for the winner differs from the white vote by between 34 and 50 percentage points. With few exceptions, the black community and the white community are not supporting the same candidates. This clearly explains why African Americans are ending up more regularly than other groups on the losing side of democracy. Voting cohesively and voting against the white majority seal the fate of the black community in American democracy.

A more detailed look at Table 5 suggests that African American voters do not just disagree with white voters over whom to elect, they also often differ from Latino and Asian American voters. The racial gaps here are much smaller, but the black-Latino divide is still 20 points or higher for every case but House elections. In the mayoralty where partisanship is much less of a factor, the black-Latino divide grows to 43 points. Similarly, the black-Asian American divide averages between 18 and 33 percentage points. If the black community was hoping to form a coalition with other racial and ethnic minorities, their hopes have clearly not been fulfilled. In short, African American voters are finding themselves somewhat isolated in the American political arena. We can now more fully understand the electoral losses that black voters are incurring.Footnote ³⁴

The flip side of the distinctiveness of the black vote is the possibility for coalition building among the other three racial and ethnic groups. Judging by the vote across this wide range of elections, whites, Latinos, and Asian Americans appear to be the three groups most likely to form a viable rainbow coalition. The divides between the three groups are, relatively speaking, small. Across most of the different types of contests, the Latino-Asian American divide was either the smallest intergroup divide or close to it. But whites and Asian Americans appeared to coalesce almost as often. In mayoral elections, for example, the white-Asian American divide averaged only 13 points. To a certain extent, American democracy is pitting African American voters against voters from the three other racial or ethnic groups.

When Do Blacks Lose?

The results so far are somewhat discouraging for those interested in the representation of African Americans. These results to this point do, however, only provide a general overview of black success rates. These aggregate patterns hide considerable variation across time and context. If we can garner a better understanding of where and when black voters are successful, we may be able to identify reforms that could lead to greater balance in American democracy.

One of the starkest differences in outcomes for blacks is one that we have already seen—greater success in House elections. These House elections are, of course, the only elections in which electoral boundaries are regularly manipulated. The fact that blacks (and Latinos) appear to be doing much better in the one area of democracy where the drawing of district lines affects outcomes reinforces the view that gerrymandering can strongly benefit minority voters. Importantly, the success of black voters in House elections points not only to black success in majority-black districts but also to black success in majority-white places. Since around 70% of black voters in House elections live in majority-white districts, the present results imply that even in predominantly white districts outcomes regularly favor black voters. One likely explanation is that black voters are being grouped with liberal or Democratic white voters who help elect minority-favored candidates. The larger implication is that minority interests may be more fully incorporated in democracy if district lines can be drawn so that minority voters are paired with other minority voters or like-minded members of the majority group.

There are also important variations in black success across time and context. In Table 6, I attempt to show how different contexts affect the success of black voters in senatorial and gubernatorial elections—two of the cases where blacks fare the worst.Footnote ³⁵ Each of the two regressions examines the role of a series of state-level factors that have in the past been linked to black representation (Huckfeldt and Kohfeld Reference Huckfeldt and Kohfeld1989; Key Reference Key1949; Manza Reference Manza2006).Footnote ³⁶ In particular, I assess the role of state politics and partisanship (percent Democrats, percent Republicans, and mean citizen ideology), key state demographics (percent black, percent black squared, and percent with only a high school degree), and measures of institutional barriers to voting (whether a state is under the jurisdiction of Section 5 of the Voting Rights Act, level of felon disenfranchisement, how long before the election voters must register, and eligible voter turnout).Footnote ³⁷ To gauge how electoral outcomes vary for blacks over time, I also include the year of the election.Footnote ³⁸ Finally, to determine if black failure is restricted to certain types of black voters, I incorporate typical measures of individual socioeconomic status and individual political orientation. The dependent variable is the same—whether the respondent ended up on the losing side of the election. All predicted probabilities below vary the variables of interest from the 10th to the 90th percentile.

TABLE 6.

Where and When Do Black Voters Lose?

Logistic Regression. ** p < .01. * p < .05.

Source: VNS Exit Polls, 1994–2006.

The results tell a largely familiar story. Black voters are more successful when they live in states that are more Democratic, more liberal, and less Republican. The effects are dramatic. All else equal, black voters in more liberal, Democratic states are 74% less likely to lose in Senate elections (38% more likely to lose in gubernatorial contests) than black voters in more conservative, Republican states. Black success in American democracy rests substantially on the willingness of their neighbors to support a liberal, Democratic agenda.

Also as expected, the size of the black population matters. The relationship is, however, a complex one as the significant and oppositely signed coefficients on percent black and percent black squared. Mirroring other studies of black-white relations, there is an inverse U-shaped relationship between the proportion of the state's population that is black and electoral outcomes (Blalock Reference Blalock1967). Black voters in Senate elections are quite successful when the black population is small (only 44% lose in states that are 7% black) but as the black population grows, white resistance to black interests grows. Black failure peaks in about the median state (15% black) where the likelihood of losing grows to 65%. Then as blacks become a more numerous and decisive voting bloc, the likelihood of ending up on the losing side of the vote declines to 58% (in states with a 29% black population). In gubernatorial elections, the benefits of living in a state with a large black population are even more pronounced. The likelihood of losing drops from 58% in the median state to 32% in states with a large black population. It is not just the racial makeup of the population that matters. In line with research showing that less-educated Americans are more intolerant and less prone to support pro-black policy, the analysis here indicates that black voters in Senate elections are 8% more likely to end up winners in more educated states (Schuman et al. Reference Schuman, Steeh, Bobo and Krysan1997).

The last critical factor for black voters is whether they reside in a state that is covered under Section 5 of the Voting Rights Act (Black and Black Reference Black and Black1987; Menifield Reference Menifield2001).Footnote ³⁹ States with a history of disenfranchising African Americans continue to have reduced black representation today. Even after controlling for the partisan leanings, the racial makeup, and the educational attainment of a state's population, blacks in these Voting Rights Act states are 15% more likely to be defeated in Senate elections. In gubernatorial elections, the VRA gap grows to 26%.

Unfortunately, for activists interested in reform, identifiable institutional levers appear to be much less consequential for African American representation.Footnote ⁴⁰ Contrary to concerns about the anti-black implications of harsh felon disenfranchisement laws (Manza Reference Manza2006), these laws have no consistent effects on black outcomes. The effect of registration deadlines on outcomes for blacks is also mixed. Alternative tests further indicate that a range of reforms that attempt to make voting easier and more open (e.g., allowing early voting, mailing sample ballot, and voting by mail) have no clear effect on black representation. This parallels results that demonstrate that these reforms have at best a marginal impact on turnout (Blais Reference Blais2006; Gronke, Galanes-Rosenbaum, and Miller Reference Gronke, Galanes-Rosenbaum and Miller2007). It may be, as some have argued, that reforms in recent decades have so reduced the ability of states to set their own registration procedures and alter their electoral institutions that state laws today are of little consequence to democratic outcomes (Highton Reference Highton2004a).