Does familiarity with members of out-groups breed contempt or acceptance? The relationship between the geographic distribution of ethnic and racial minorities and white attitudes has been the subject of considerable academic debate since V.O. Key’s 1949 landmark Southern Politics. In Southern Politics, Key advanced the claim that white racial conservatism was the strongest in places with large African American populations. Key noted that white turnout in support of the segregationist Democratic Party and the Jim Crow apartheid system was strongest in places where African Americans constituted a large proportion of the population and thus presented a potential political and cultural threat.
Empirical tests of Key’s claim about the relationship between out-group size and white racial conservatism have generated mixed findings—depending on the geographic unit of analysis and electoral context—leading many scholars to doubt its validity. Some of the reason for skepticism of Key’s threat hypothesis stems from the work of another social science luminary: Gordon Allport. Allport was a psychologist who was interested in understanding the roots of intergroup prejudice (among many other things). Allport found that one consistent mechanism for mitigating intergroup prejudice was contact with members of out-groups. On the aggregate level, Allport’s contact hypothesis produces the opposite empirical prediction as compared to Key’s threat hypothesis. The contact hypothesis presupposes that white racial attitudes should be the most liberal in places with lots of opportunities for interracial contact—presumably places with large minority populations. The behavioral implication of the contact hypothesis is that whites in racially diverse areas should be less likely to mobilize against black political interests, not more, as Key predicted. The threat hypothesis and the contact hypothesis generate distinctly different predictions.
Steen Thomas and I investigate this question in an article that will be appearing in Social Science Quarterly later in the year (early view here). We attempt to assess these two competing explanations by examining how white turnout is conditioned by their proximity to African American populations. We test these competing explanations in the Deep South, where the empirical referents and intellectual roots of the Key’s threat hypothesis lie. We chose to test our argument using electoral data from three presidential elections (2000, 2004 and 2008) in the state of Louisiana. We chose Louisiana because the state recorded registration and turnout by race in order to comply with Section V of the Voting Rights Act (author’s note: RIP), making turnout by race directly observable.
The question is: how is white turnout affected by the geographic distribution of African Americans? Answering this question requires us to construct a measure of a white voters racial environment. Measurement of a voter’s racial environment is complex; voters inhabit multiple overlapping political geographies (e.g. precinct, county, state). Moreover, the mean racial composition of any political geography only tells part of the story, individuals actually need to come into contact with members of other groups if Allport’s mechanism is to work. This fact implies that segregation is a critical intervening variable—racially diverse but segregated contexts might elicit dramatically different response than diverse but integrated contexts. In order to account for these possibilities, we utilize ArcGIS mapping software in conjunction with census data to construct variables for the mean racial composition of each precinct and parish in Louisiana. In addition, we also construct a measure of segregation on both the precinct and parish level and include an interaction between precinct and parish level diversity and segregation, in order to assess whether the potential for interracial contact conditions white’s threat responses and decreases political mobilization.
Because precincts are nested within parishes, we utilize a hierarchical linear model of test our hypotheses. We find that the effect of racial diversity on white behavior is conditional upon the level segregation. Precinct level racial diversity has a negative effect on white turnout when precinct level segregation is low but a positive effect on white turnout in highly segregated contexts. Figure 1 displays the interaction between the proportion of Americans in a precinct and precinct level segregation across all three elections included in the analysis.
As Figure 1 makes clear, whites in racially diverse and integrated contexts are less likely to turnout compared to whites in racially diverse but segregated settings, even when controlling for income and education. The predicted values for 2008, when holding all variables to their respective means and modes, are shown in the table below. The substantive conclusion that can be drawn from these findings is that racial diversity can have a liberalizing effect of white attitudes and behaviors, but only when diversity is accompanied by the possibility of interracial contact. If not, diversity may exacerbate perceptions of threat, as Key suspected. The degree of segregation matters a great deal. Our evidence suggests that familiarity certainly does not breed contempt, but cursory interaction may also do little to encourage harmony.
Overall, our findings help to clarify a number of the mixed findings. When studying the effect of the racial context on white attitudes and behaviors, the results one is likely to find are not only contingent on the chosen geographic unit of analysis but are also conditioned how the groups are distributed within each unit. Failing to account for overlapping political geographies and segregation introduces omitted variable bias into most analyses and this bias potentially explains why there are so many mixed findings in the literature.