Racial Composition, White Racial Attitudes, and Black Representation: Testing the Racial Threat Hypothesis in the United States Senate

The RTH has also been call the “power-threat hypothesis,” “racial backlash hypothesis,” and “power theory.”

Other research suggests that the relationship between geographic social context and whites’ support for race-targeted policies may, in some cases, reflect psychological responses of whites to African Americans, rather than a realistic threat (Oliver and Mendelberg 2000).

For a similar logic, see Rudolph and Popp (2010) who use black empowerment through business ownership as a source of white racial threat rather than simply using racial composition in the population.

The votes scored by the LCCR generally fall into one of three categories: (1) votes that are overtly tied to African-Americans, such as affirmative action, hate crimes involving racial minorities, and voting rights protections for minority citizens; (2) votes that deal with implicitly racial issues because of African-Americans’ minority or economic standing, such as welfare/food stamps, poverty programs, minimum wage increases, and the “motor-voter” registration bill aimed at poor, and minority citizens; and (3) bills without a direct tie to the African-American community, but are related to civil rights for other groups such as women, homosexuals, and the disabled. For the years in our time period, 30–45% of votes scored by the LCCR fall into the latter category. Thus, while not all of the votes scored are directly related to African-American interests, the majority of votes are either overtly or implicitly tied to the black community.

All substantive findings hold when controlling for senator ideology measured using DW-nominate scores (Poole and Rosenthal 1997) and state citizen partisanship (Wright and McIver 2006; available from authors at request), the latter of which is not statistically significant. Following the lead of most of the research, we report models that exclude these controls.

Over-reporting of voter turnout could be problematic if African-Americans overreport at higher rates than whites, which early research found to be the case (Abramson and Claggett 1986). However, recent research finds that while blacks are more likely to overreport voting, whites living in regions with larger black populations, as well as whites living in the Deep South, are also more likely to overreport (Chadha and Montjoy 2001). Thus, in states with larger black populations we would expect exaggerated reporting of turnout among both African-Americans and Whites. Given this, we have no reason to expect the estimated effect of black electoral strength on black representation will be biased.

The 2004 NAES also includes large samples but does not include a measure of white racial attitudes.

Senators with an independent affiliation are classified along with the party with which they caucus. We have run our models with these senators excluded or classified separately and the results are substantively unchanged.

Some earlier studies found a nonlinear relationship between racial diversity and black representation (e.g., Black 1978; Bullock 1981; Hibbing and Welch 1984). We explored several nonlinear specifications and found no evidence of a nonlinear relationship between either the proportion of the population that is black or black electoral strength and LCCR scores.

Percent population black and black electoral strength are highly correlated (r = 0.95), which means that the standard errors for these variables will be inflated. Since the effect of both variables are statistically significant, the inflated standard errors do not lead to false inferences because any such bias will make it more difficult to find a statistically significant effects.

We are not the first to consider partisan differences in black representation and the RTH in the Senate. Hood et al. (2001) examine the influence of black electoral strength (i.e., proportion of registered voters that are black) on black representation in the South among Democratic and Republican senators. Initially they find a positive effect of black electoral strength among Democrats and a negative effect for Republicans. However, they report that further tests show that there are no statistically significant partisan differences in the influence of black electoral strength on black representation.

It worth noting that there is no statistically significant effect found for the lagged dependent variable in the model predicting LCCR scores for southern senators. In analysis not reported, we found that this null effect is the result of controlling for senators’ partisanship in this model. The effect of lagged LCCR scores is significant when the model is estimated without senator partisanship. Consistent with this conclusion regarding these null effects, the correlations between LCCR scores and lagged LCCR scores are substantially smaller when examining southern Democrats and Republicans separately.

We find no regional differences in the interactive effect of unemployment rates and white racial conservatism, or of senators being cross-pressured between unemployment rates and constituency preferences.