The Politics of Losing

Home > Other > The Politics of Losing > Page 19
The Politics of Losing Page 19

by Rory McVeigh


  The theory that has guided our analysis explains when and where white nationalist movements are likely to emerge. Conceptualizing power—economic, political, and status—as an exchange relationship subject to supply and demand lets us see precisely where power might be waning. Power losses, by themselves, don’t automatically produce collective response. But they can make us receptive to social movements and vulnerable to politicians who promise to restore our place in America. When racial and cultural identities overlap with our place on the economic ladder, these kinds of promises to restore power are even more potent, because they draw on cultural solidarity of the aggrieved group and place blame on cultural outsiders.

  In the 1920s, William Joseph Simmons dreamed of resurrecting the Ku Klux Klan as a special kind of fraternal lodge. But he attracted only modest interest until 1920, when more savvy organizers like Edward Young Clarke, Elizabeth Tyler, D. C. Stephenson, and Hiram Evans developed a strategy that linked the Klan to the shake-ups of the day. Americans outside of the industrialized Northeast were flailing from the economic transitions—transitions enabled by immigrant labor. Because of the immigrants’ cultural differences, they were a convenient foil for the Klan, a soft target for the prejudices held by so many Americans.

  A century later, lost power helps us understand that the strong loyalty to Trump reflects something more than simple prejudice. He had a special appeal to Americans who were floundering in the global economy, an attraction all the more potent because he linked their losses to cultural identities. This framing, combined with geographic segregation, made it difficult for his supporters to see that many nonwhite Americans were also struggling in the new economy—and many white Americans were thriving.

  We are acutely aware of the stakes now. The growing interdependence of global markets is disruptive—it generates prosperity for some and hardship for others. Although these hardships cut across social cleavages, segregation and a strong overlap between social position and racial or cultural identity can divide us into clear groups that can be pitted against each other.

  Trump revealed and exacerbated these deep divisions among the American people—but he did not create them. Our political system forces a choice for voters between two competing visions of what America should be. When James Madison made his case for the Constitution in the Federalist Papers, he described a government designed to make constructive use of “factions,” which he knew were inevitable in any free society.1 In a purely democratic system, he argued, a numerical majority, such as the unlanded, could consistently dominate the minority, such as property holders. By forcing citizens to reconcile interests with a vote for a single candidate, Madison envisioned an America where competition could lead to cooperation, as groups with different interests would form alliances with other groups, diminishing the extent to which any single issue shaped our democracy. Centuries later, scholars still analyze American political institutions the same way. They focus on how cross pressures—where voters may side with one party on some issues while siding with another on other issues—can reduce the hostility citizens have against the other party.2

  This pluralist view, however, overlooks how the American political system can, at times, fall into divide-and-conquer strategies and disgruntle large portions of the population when they feel that the system neglects them. We have argued that factions organize not just around differences of opinion but around privilege hierarchies. White Americans, regardless of their social class, tend to enjoy certain benefits that come from white identity. Men hold society-granted advantages over women in the labor market and at home. Some religious groups can assert their values and enforce conformity in opposition to other religious groups or the nonreligious.

  By forming coalitions around privilege hierarchies, a privileged numerical minority can maintain advantages over others even when outnumbered by those without such privileges. But these alliances are unstable. Power devaluation can not only spawn right-wing movements but also fracture alliances within parties.3 Before Trump’s election, many voters who sided with the Republican Party became dissatisfied with the way the party ignored their economic struggles. Trump was a candidate who seemed to recognize their plight, and who offered to align their interests with the agenda of a potential U.S. president.

  * * *

  In August 2017, neo-Nazis, Klansmen, and other white supremacists gathered in Charlottesville, Virginia, at the Unite the Right rally to protest the removal of a Robert E. Lee statue from a public park. On the second day, the ralliers clashed with counterprotesters. James Fields Jr., a member of the supremacist organization Vanguard America, drove his car into a crowd of counterprotesters, killing a young paralegal named Heather Heyer.

  “I’ve condemned neo-Nazis,” said Trump in the aftermath of Charlottesville. “I’ve condemned many different groups. But not all of those people were neo-Nazis, believe me.” He added, “The press has treated them absolutely unfairly.” And, “You also had some very fine people on both sides.”4

  On August 12, 2017, James Fields Jr. slammed his car into a group of counterprotesters after a rally by white nationalists in Charlottesville, Virginia. Photo by Ryan M. Kelly, Associated Press.

  Since 1990, the Southern Poverty Law Center (SPLC) has published annual reports on hate groups in the United States. These organizations are growing. In 1999, the SPLC identified 457 hate groups. By 2017, that number had risen to 954.5 Lost power has radicalized some Americans. But here we have been more interested in explaining how ordinary people, who do not think of themselves as extremists, were attracted to Trump’s message and were willing to at least overlook, if not embrace, his appeals to prejudice. White nationalism is most consequential when it enters the mainstream—so mainstream, in this case, that it captured the White House.

  When considering the rise of the Klan, it’s easy to assume that it represented an intensification of racial and ethnic animosities. But its bigotry only looks unusual when we forget what the 1920s were like. Overt racism and religious prejudice were everywhere. This was an era when even top scientists promoted the idea that racial inequality is rooted in genetic differences, and that Anglo-Saxons are naturally superior.6 Many Americans, including many who would join the Klan, did not think of immigrants from Italy or Poland as white.7 These kinds of broadly accepted and openly expressed prejudices did not trigger the growth of the Ku Klux Klan, but they fueled it as it spread. All across America, native-born white Protestants were losing power and looking for a way—any way—to reclaim it.

  The nature of race relations is changing. Americans are much less likely to express overt prejudices than they were in the past. In the 1940s, a majority of white Americans supported “segregated neighborhoods, schools, transportation, jobs, and public accommodations.” By the 1970s, that support had dropped to about 25 percent.8 By the 1990s, more than 90 percent of white Americans supported equal treatment, by race, in schools and employment.9 Despite this, those who study race relations are quick to point out that change in our willingness to express bald-faced bigotry does not mean that racism has been driven out of American society. It has just taken subtler forms. Although contemporary racial resentments may be buried a few inches deeper beneath the surface than they were in the 1920s, they are still there, and still combustible.

  We began this book by describing a massive Klan rally on the Fourth of July, 1923, in Kokomo, Indiana. In towns like Kokomo, white Americans viewed the Klan as a civic asset and a source of empowerment. As popular as the Klan was in the 1920s in many parts of the country, the movement’s rise, like the Trump candidacy, exacerbated deep divisions among Americans. We see this in the raucous and contentious Democratic Convention of 1924, and in the way the Klan was often met with angry and, at times, violent counterprotesters. It was at once massively popular and massively unpopular.

  In 1926, the African American sociologist and civil rights activist W.E.B. Du Bois wrote an article about the Klan in the North American Review. “Until last year I was of th
ose mildly amused at the KKK. It seemed to me incredible that in 1925 such a movement could attract any number of people or become really serious. And then at first hand and at second I saw the Klan and its workings in widely different places.”10

  To Du Bois, everyday Americans were complicit in promoting the Klan and its goals. He wrote:

  Thus the Ku Klux Klan is doing a job which the American people, or certainly a considerable portion of them, want done; and they want it done because as a nation they have fear of the Jew, the immigrant, the Negro. They realize that the American of English descent is not holding his own physically or spiritually in this country; that America survives and flourishes because of the alien immigrant with his strong arm, his simple life, his faith and hope, his song, his art, his religion. They realize that no group in the United States is working harder to push themselves forward and upward than the Negroes; and over all this rises the Shape of Fear.11

  Appendix

  METHODS OF STATISTICAL ANALYSIS

  The quantitative analysis presented in chapter 5 examines relationships between attributes of U.S. counties and the percent of the vote that went to Trump in the Republican primary and caucus campaigns and in the general election. This focus on counties provides a substantial amount of statistical leverage, as we are able to examine variation across 2,876 cases. As we mention in chapter 5, we excluded some counties because of data limitations. Counties in the state of Alaska were not included because Alaska does not report electoral results at the county level. Similarly, we exclude counties from Colorado (64 counties), North Dakota (53 counties), and Minnesota (87 counties) because these states report results only for legislative/congressional districts. Kansas reports results of some caucus events at the county level, while others cross county lines. This forces us to exclude 20 counties in Kansas. We exclude these cases not only from our analyses of Republican primary and caucus voting outcomes but also from the analyses of general election results, in order to ensure that all models are estimated using the same set of cases. This facilitates comparison of findings across models.

  The timing and processes involved in primaries and caucuses vary across states. Some states select candidates based on preference polling while others select based on caucus events. Some states hold “open” primaries in which members of any party can vote in the Republican contest, while others are “closed” to voters who are not registered with the Republican Party. Ballots for states that held primaries early in the campaign season naturally featured more options for Republican voters, as less-successful candidates dropped out after early losses diminished their chances of earning the nomination. We use a fixed-effects design to increase confidence that our findings capture the effect of county-level variables, and not these differences in state-level processes for selecting delegates. The process is the equivalent of including a dichotomous variable in the analysis for each state. We also use the robust cluster option in the statistical package Stata to adjust standard errors to account for the clustering of counties within states.

  We use two dependent variables in the statistical models: the percentage of Republican primary voters in the county who voted for Trump and the percentage of general election voters in the county who voted for Trump. Data for these key variables were obtained from the Atlas of U.S. Presidential Elections, which compiles data on voting outcomes from Secretary of State offices, or their equivalents.1 We estimate all models with ordinary least squares (OLS) regression. We also use data from this source to create a measure of the percentage of voters who voted for the Republican presidential candidate, Mitt Romney, in the 2012 general election. We use this variable in figure 5.2 to illustrate the high degree of correlation between the Trump vote in the 2016 general election and the vote for Romney in 2012, and we also include it as an independent variable in our regression analyses of the 2016 Republican primary and general election voting outcomes.

  We derive data for most of our independent variables from the U.S. Census Bureau’s American Community Survey (ACS) five-year summary files for the period 2010–14, the most recent five-year estimates available for U.S. counties at the time we conducted this analysis.2 Importantly, because of the differences between urban and rural counties, we control for population density, measured as the number of residents (in thousands) per square mile.

  Throughout the book, we emphasize that Trump’s campaign message was likely to resonate with residents of communities that were on the losing end of a globalizing economy. In particular, we expect that the benefits of the economic recovery following the Great Recession overwhelmingly flowed toward communities where the economy had a critical mass of individuals who were prepared to thrive in the new economy. In particular, we argue that residents of communities with larger proportions of individuals having a college education were well positioned to benefit from recovery efforts, while communities with fewer college-educated residents were unlikely to thrive in that environment. As a measure of this crucial feature of counties, we use ACS data to create a variable, percent with a college degree, that represents the proportion of the population age twenty-five years or older that has earned a bachelor’s or higher degree. We also include a measure of median household income (measured in thousands of dollars). In addition to education and income, rates of unemployment could influence the size of the pool of voters who were responsive to Trump’s economic nationalism. We measure percent unemployed as the percentage of the population age sixteen years or older in the civilian labor force that reported their status as unemployed. We include a measure of median age in the county as well as measures of percent retail occupations (the proportion of the total employed civilian workforce age sixteen and older that is employed in retail occupations) and percent manufacturing (the proportion of the total employed civilian workforce age sixteen and older that is employed in manufacturing).

  In our analysis, we are interested in examining the role of other forms of privilege besides economic privilege. We include three variables intended to capture the degree to which residents of local communities tend to adhere to traditional gender and family relations and norms, versus more progressive or egalitarian gender relations and family arrangements. These measures include percent women in the labor force, which reflects the percent of women in the county who are age sixteen years or older who are in the labor force. Our measure of percent married is simply the percentage of county residents age fifteen years or older who are married (not including married but separated). We also use a measure of male educational advantage, calculated as the difference between the percentage of men age twenty-five years or older with a bachelor’s degree or more education and the percentage of women age twenty-five years or older with a bachelor’s degree or more.

  We also include a variable for percent nonwhite—the percent of total population that the ACS categorized as either Hispanic/Latino or a member of a minority racial group. For religion variables, we rely on data from the 2010 Religious Congregations and Membership Study, conducted by representatives of the Association of Statisticians of American Religious Bodies.3 The study is unique and comprehensive in that it represents 236 religious bodies in the United States—reporting 344,894 congregations with a total of 159,686,156 adherents, and making up 48.8 percent of the total U.S. population.4 These data are particularly useful for our analysis because estimates of “adherents”—that is, individuals having any affiliation with a congregation—are reported at the level of U.S. counties. Study organizers categorized congregations into established religious classification schemes. We use their coding classifications to calculate the percentage of total adherents in the county who are affiliated with evangelical congregations (percent evangelical) and the percentage of total adherents affiliated with Catholic congregations (percent Catholic).

  We report results of OLS models predicting the vote for Trump in the primary and caucus campaigns and the vote for Trump in the general election in tables A.1 and A.3, respectively. The key findings from these statistica
l analyses are discussed in chapter 5. We also present graphs for six interaction effects (figures 5.4–5.9). Here we provide additional documentation related to those interactions. The first five interactions involve our variable for percent with a college degree: (1) percent college x percent unemployed, (2) percent college x median income, (3) percent college x percent women in labor force, (4) percent college x percent married, and (5) percent college x percent evangelical. Last, we interact (6) percent nonwhite population x percent manufacturing. All interactions are statistically significant. We display the full results of these analyses in table A.2.

  To create figures 5.4–5.9, which help visualize the strength of these interactions, we used the margins command in Stata to generate predicted values for the dependent variable (percent vote for Trump in the primaries and caucuses) at various values of the two terms included in the interaction, while holding all other covariates at their mean values. To ensure that we are not extrapolating beyond our data, for each graph we limit the range of values shown on the x-axis to only those that are observed with some frequency in the actual dataset. Our variable for percent college educated has a mean value of 20 and a standard deviation of 9. We choose to plot predicted values for counties at the mean (20%) and those with slightly more than one standard deviation above (30%) and below (10%) the mean. This arrangement allows us to display how the effect of various independent variables on the Trump vote differs for counties with low, average, and high proportions of college-educated residents (figs. 5.4–5.8). We use a similar approach for figure 5.9. The variable for percent manufacturing has a mean of 12 and a standard deviation of 7. We plot predicted values for counties at the mean (12%), for counties slightly less than one standard deviation above (18%) and below (12%) the mean. The graphs, then, show the effect of percent nonwhite population on the Trump vote in the primaries and caucuses for counties with low, average, and high proportions of the population employed in manufacturing.

 

‹ Prev