Candidate Appeals to Black and. Hispanic Voters in Democratic. Congressional Primaries

Candidate Appeals to Black and Hispanic Voters in Democratic Congressional Primaries November 3, 2016

This study examines the effects of district racial and ethnic composition on voter turnout and candidate issue appeals to voters of color. I argue that Democratic candidates in districts where the Democratic party is largely African-American or Latino are more likely to use strategic issue appeals to mobilize voters of color. To test this, I rely on Democratic House primary voter turnout data and Democratic House primary candidate websites. I find that candidates in districts with larger proportions of black Democrats are more likely to emphasize criminal justice issues, and black Democrats in these districts are more likely to turn out. Similarly, candidates in districts with larger proportions of Hispanic Democrats were more likely to discuss immigration, and Hispanic Democrats in these districts were more likely to turn out. This suggests that candidates utilize strategic appeals to voters of color in situations where these voters make a significant proportion of the Democratic primary electorate. This suggests that candidates utilize strategic appeals to voters of color in situations where these voters make a significant proportion of the Democratic primary electorate.

While African-American and Latino voters had record high levels of turnout in the 2008 and 2012 presidential elections [Frey, 2013], turnout among voters of color continues to lag in midterm elections [Krogstad, 2014]. However, a significant body of literature suggests that African-American and Latino turnout in Congressional elections is significantly boosted by the presence of co-ethnic candidates or increased co-ethnic district composition[barreto et al., 2004, Fraga, 2016, Barreto, 2010]. However, the majority of this literature focuses on general elections, rather than Congressional primaries. Since many voters of color live in overwhelmingly Democratic districts, Democratic primary elections may have a greater impact on African-American and Latino representation than do general elections, as the winner of the primary election will almost always go on to win the general. Some previous work does suggest that increases in the proportion of co-ethnic voters do increase primary turnout among African- American and Latino voters [Fraga, 2016]. Many possible explanations for the effect of district composition turnout have been proposed, including neighborhood effects [Barber and Imai, 2014], differences in political knowledge [DeSante and Perry, 2015], and empowerment [Gay, 2001, Fraga, 2016]. I advance a different argument that is rooted in the congressional candidates strategic calculus. Congressional candidates make strategic decisions to focus on specific constituencies. As candidates have limited time, resources, and credibility, they usually cannot appeal strongly to all groups within their party. Democratic candidates who run in districts where voters of color make up a significant proportion of the Democratic party may strategically appeal to these voters in order to mobilize them. Some candidates and groups have openly sought to appeal to African-American or Latino voters.leighley [2001] argues that political elites choose to mobilize African-American and Latino voters by decreasing the costs of political participation in areas with large populations of voters of color. Hayes et al. [2010] suggests that House members who were redistricted into districts with greater proportion of voters of color are more likely to sponsor civilrights legislation in the House. Furthermore, at a national level the Republican party has tried to make inroads with Latino voters, though often unsuccessfully [Baik et al., 2009, Gimpel and 3

Kaufmann, 2001]. In this paper, I argue that issue-based candidate appeals play an important role in the mobilization of African-American and Latino voters in districts where they make up a significant portion of the electorate. To test this theory, I rely on two separate sources of data. First, I use Catalist primary turnout data from 2000 to 2012 to examine the relationship between the racial composition of registered Democrats in a district and Democratic primary voter turnout in competitive primary elections. My second data source is a collection of Democratic House primary candidate website issue pages in 2014 and 2016. I then examine the relationship of specific web page issue topics to district racial and ethnic composition. I find that primary turnout is higher among African-American and Latino voters when they make up a larger proportion of a district s Democratic registered voters. I also find that candidates in districts with large proportions of African-American and Latino Democrats are likely to tailor their policy positions to appeal to these constituencies. This suggests that candidates in diverse districts strategically emphasize certain issues in order to appeal to their constituencies of color. Theory Candidates make strategic decisions to increase their voteshare and their probability of winning office. To this end, they appeal to various constituencies in order to mobilize their supporters or convince voters who may be unsure. However, Congressional candidates have limited resources and sometimes are forced to choose between conflicting constituencies. Expending significant capital to mobilize or convince a constituency that makes up only a tiny sliver of the district is not an efficient use of resources. Candidates can only prioritize a limited number of issues during their campaign, and must choose these issues carefully to appeal to the correct constituencies. 4

Candidates who target specific constituencies may lose support from voters not in that constituency. Hersh and Schaffner [2013] finds that mis-targeted mailers aimed at Latinos, union members, or born-again evangelicals were associated with a decrease in support among voters who were not a member of those groups. Additionally, mobilizing constituencies of color may be even more complicated than mobilizing other constituencies because there is a possibility of white backlash. Racially conservative white voters may react very negatively to appeals designed to turn out constituencies of color, and may either stay home instead of voting or turn out to vote for the opponent. There are still a number of racially conservative white voters for whom race is a salient electoral issue. For example, Schaffner [2011] finds that many racially conservative white voters believed that Obama focused a significant amount of attention on appealing to African-American voters in 2008. In 2008 and 2012, Obama received significantly fewer votes in areas with higher rates of racially-charged Google search terms [Stephens-Davidowitz, 2014]. Furthermore, Tesler [2013] suggests that a number of racially conservative Democrats migrated to the Republican party after Obama s election. Openly appealing to voters of color may only be an electorally gainful strategy when they make up a substantial portion of the electorate, so that increased turnout among this group can make up for any white backlash. Candidates who emphasize certain issue positions may be more successful at mobilizing African-American or Latino voters. While voters of color, like other voters, have diverse and individually specific sets of issue concerns, members of certain racial or ethnic groups may place greater weight on some issues. Dawson [1994] finds that due to the significant influence of race on the life chances of African-Americans, many African-American voters see their political interests as intertwined with the interests of African-Americans as a group. Therefore, it is likely that African-American voters would be mobilized by issues that disproportionately affect African-Americans. While some scholars have found that linked fate is weaker among Latino voters than among African-Americans [Sanchez and Masuoka, 2010], many Latino voters place significant importance on the issue of immigration [Barreto and Segura, 2014]. 5

Barreto et al. [2009] finds that support for and participation in the 2006 immigration rallies was not limited only to Mexican immigrants, but was widespread among Latinos. Furthermore, Pantoja and Segura [2003] finds that racially charged ballot propositions were associated with increased political awareness among Latinos. The immigration issue can serve as a mobilizing issue for Latino voters. Data and Methods Turnout To examine the effect of district composition on primary turnout, I use data from a random 1% sample of US voter registration files provided by Catalist. I limit my analysis to white, black, and Hispanic registered Democrats only. Voter turnout in primaries tends to be low, and it is difficult to tell which non-democrats would be interested in voting in a Democratic primary if eligible. Some voters with no party affiliation may lean Republican, and including them in the regression would confound the turnout effect with partisanship. Instead of using the overall percentage of black or Hispanic residents in a district to determine racial composition, I use the proportion of black or Hispanic registered Democrats out of all registered Democrats in the district. In Democratic primaries, candidates are appealing only to the Democratic primary voters, rather than the whole district. Two districts with the same proportion of African-American or Latino residents may have very different proportions of African-American or Latino Democrats. For example, a district in the deep South may have a large number of African-American Democrats and a large number of white Republicans, whereas a district in a northern city may have a large number of both African-American and white Democrats. In the first district, close to 100% of the Democrats in the district will be African-American, while in the second district the proportion will be much smaller, even 6

though both have the same absolute proportion of African-American residents. Only competitive Democratic House primaries were included in the analysis. These were primaries in which at least one of the losing candidates received PAC funding 1. As few PACs would donate to a non-viable candidate, races where at least one of the losers received PAC funding are much more likely to be competitive. Voter demographic covariates and block group income were from the Catalist voter file. Finally, 37 races were removed from the data because they had either 100% or 0% turnout among all racial groups. These percentages are likely to be an error in the Catalist voter file, so those districts were removed. In total, the analysis included 408,499 voters voting in 146 races from 2000 to 2012. A logistic regression with standard errors clustered by election was used with year and state fixed-effects. Candidate Website Issue Appeals In order to test the effect of district racial composition on candidate appeals, I examined 325 Democratic House primary candidate websites from the 2014 and 2016 2 congressional elections. I limited my analysis only to websites of candidates who had at least one primary opponent. To identify topics within the web pages, I used an unsupervised text analysis model (Latent Dirichlet Allocation) [Blei et al., 2003] with 20 topics. Table 1 lists the 15 issue relevant topics uncovered by the model. In sum, these topics present a reasonable representation of the Democratic party issue agenda. LDA allows for multiple topics per document, and returns the proportion of each topic per page. Since some sites had more than one issue relevant page, the adjusted topic proportions 1 Candidate gender, incumbency status, and funding data was from the DIME database [Bonica, 2013] 2 This paper contains data from the 2016 House elections up to June 2016. The 2014 data was collected by Karen Jusko, Allison Anoll, and Mackenzie Israel-Trummel at Stanford University; Michael Dougal and Ryan Hübert at the University of California, Berkeley; and Mike Parkin at Oberlin College [Jusko et al., 2014], whereas the 2016 data was collected by Masha Krupenkin. 7

per site were calculated according to the following equation. T ik is the proportion of topic i for candidate k, p represents all of the candidate s issue-relevant pages, and n represents all of the 15 issue relevant topics. Essentially, T ik represents the proportion of space that the candidate devotes to the issue out of all of the space they devote to discussing policy issues. p topic i T ik = 100 p n j=1 topic j Additionally, candidates who spoke about a topic less than 1% were assigned an adjusted topic proportion of zero. To analyze the data, I used a Tobit model [McDonald and Moffitt, 1980] with a lower threshold of zero. This model assumes that the observed data is a censored version of the true data. Any value of the true data which falls below a certain threshold (usually zero) is replaced with that threshold value. A Tobit model is appropriate for modeling candidate issue agendas, because if a candidate does not discuss an issue, we do not observe the actual importance they put on that particular issue relative to the other issues they do not talk about. There were two issue topics of particular theoretical interest. The first was the criminal justice topic, which dealt with crime and policing. Given the significant racial disparities in Americans perceptions of policing [Gabbidon and Higgins, 2008] and the current debate over policing practices, it is likely that these disparities would be reflected in candidate appeals to African-American voters. The second topic of interest was immigration. Scholars have found that the immigration issue has acted as a catalyst for political participation among Latinos[Barreto and Segura, 2014]. Thus, it is also likely that candidates making appeals to Latino voters would discuss immigration. Finally, I used data from the 2012 and 2014 CCES to determine the proportion of black and Hispanic Democrats out of all Democrats per district. The Party ID question was used to determine the subpopulation of Democrats per district - any respondent who answered that 8

they were a strong Democrat, weak Democrat, or a Democrat leaner was considered to be a Democrat. Data on district ideology was from Tausanovitch and Warshaw [2013]. Results Both turnout and candidate issue appeals were positively and significantly correlated with the proportion of black or Hispanic Democrats in a district. As the proportion of black Democrats out of all Democrats in a district increased, candidates were more likely to appeal on criminal justice, and turnout among black voters increased. Furthermore, increases in the proportion of Hispanic Democrats is associated with increased candidate appeals on immigration, and increased turnout among Hispanic voters. Turnout Black voters are more likely to turn out when a larger proportion of the Democrats in their district are black. Similarly, Hispanic voters are more likely to turn out when a larger proportion of the Democrats in their district are Hispanic. Table 2 shows the results of these regressions. Figure 1 shows that increasing the proportion of black Democrats from 0 to 80% is associated with a decrease in the black-white turnout gap among Democratic primary voters by about 15%. As the proportion of black Democrats increases, the turnout gap between white and black Democrats virtually disappears. Similarly, Figure 2 shows that as the proportion of Hispanic Democrats increases, the turnout gap between Hispanic and white non-hispanic voters decreases. Increasing the proportion of Hispanic Democrats from 0 to 60% is associated with about a 15% decrease in the turnout gap. 9

Figure 1: % Black District Democrats and Racial Turnout Gap 10

Figure 2: % Hispanic District Democrats and Racial Turnout Gap 11

Candidate Website Issue Appeals For both the immigration and criminal justice topics, there is a relationship between the ethnic/racial composition and candidate appeals. Table 3 shows that an increase of 10% in the proportion of Hispanic Democrats out of all Democrats in a district is associated with a 1.2% increase in candidate appeals on immigration. Similarly, Table 4 shows that an increase of 10% in the proportion of black Democrats is associated with 1.1% increase in candidate appeals on criminal justice. None of the other issue topics had a significant correlation with the district racial or ethnic composition 3. These effects were robust to the inclusion of candidate race, ethnicity, and district ideology and income. Conclusion Voters of color are more likely to participate in Democratic primaries when they comprise a greater proportion of the Democrats in their district. For both black and Hispanic voters, increasing the proportion of co-ethnic Democrats in a district nearly eliminated the turnout gap between white voters and voters of color. This result is consistent with previous work, which has found associations between district co-ethnic composition and turnout [Fraga, 2016]. Furthermore, candidates in districts with large proportions of black or Hispanic Democrats were likely to emphasize issue positions that could mobilize voters of color. Candidates in districts with many black Democrats included issue appeals on criminal justice, whereas candidates in districts with more Hispanic Democrats appealed on immigration. These effects were not the result of African-American or Latino candidates. While Latino candidates were more likely to discuss immigration than non-latino candidates, the effect of an increase in the proportion of Hispanic Democrats was much larger than the effect of the candidate s ethnicity. 3 With the exception of the energy/environment topic, which was negatively associated with the proportion of black Democrats in a district 12

Furthermore, African-American candidates were actually less likely to discuss criminal justice issues after controlling for district composition. This suggests that candidates are adapting their issue positions to the characteristics of their voters. Additional research would be necessary to further illuminate the link between voter turnout and candidate issue appeals. While the association between the proportion of Democrats of color and specific candidate issue positions is suggestive of a link between issue appeals and turnout, more research is needed to fully clarify this link. Even though issue appeals may be only one part of the candidate s mobilization strategy, they nevertheless show that candidates are committed to mobilizing constituencies of color. References Ellen R Baik, Jessica Lavariega-Monforti, and Adam J McGlynn. Latino cabinet appointments and young latino voters: A preliminary look at gop efforts to attract latino voters. The Social Science Journal, 46(3):601 608, 2009. Michael Barber and Kosuke Imai. Estimating neighborhood effects on turnout from geocoded voter registration records. Technical report, Citeseer, 2014. Matt Barreto and Gary Segura. Latino America: How America s Most Dynamic Population is Poised to Transform the Politics of the Nation. PublicAffairs, 2014. Matt A Barreto. Ethnic cues: The role of shared ethnicity in Latino political participation. University of Michigan Press, 2010. Matt A Barreto, Gary M Segura, and Nathan D Woods. The mobilizing effect of majority minority districts on latino turnout. American Political Science Review, 98(01):65 75, 2004. 13

Matt A Barreto, Sylvia Manzano, Ricardo Ramirez, and Kathy Rim. Mobilization, participation, and solidaridad latino participation in the 2006 immigration protest rallies. Urban Affairs Review, 44(5):736 764, 2009. David M Blei, Andrew Y Ng, and Michael I Jordan. Latent dirichlet allocation. the Journal of machine Learning research, 3:993 1022, 2003. Adam Bonica. Database on ideology, money in politics, and elections: Public version 1.0. Computer File, 2013. URL http://data.stanford.edu/dime. Michael C Dawson. Behind the mule: Race and class in African-American politics. Princeton University Press, 1994. Christopher D DeSante and Brittany N Perry. Bridging the gap how geographic context affects political knowledge among citizen and non-citizen latinos. American Politics Research, page 1532673X15606967, 2015. Bernard L Fraga. Candidates or districts? reevaluating the role of race in voter turnout. American Journal of Political Science, 60(1):97 122, 2016. William H Frey. Minority turnout determined the 2012 election. Washington, DC: Brookings Institution. Retrieved January, 3:2014, 2013. Shaun L Gabbidon and George E Higgins. The role of race/ethnicity and race relations on public opinion related to the treatment of blacks by the police. Police Quarterly, 2008. Claudine Gay. The effect of black congressional representation on political participation. In American Political Science Association, volume 95, pages 589 602. Cambridge Univ Press, 2001. James G Gimpel and Karen M Kaufmann. Impossible Dream or Distant Reality?: Republican Efforts to Attract Latino Voters. Center for Immigration Studies, 2001. 14

Matthew Hayes, Matthew V Hibbing, and Tracy Sulkin. Redistricting, responsiveness, and issue attention. Legislative Studies Quarterly, 35(1):91 115, 2010. Eitan D Hersh and Brian F Schaffner. Targeted campaign appeals and the value of ambiguity. The Journal of Politics, 75(02):520 534, 2013. Karen Jusko, Allison Anoll, Mackenzie Israel-Trummel, Michael Dougal, Ryan Hübert, and Mike Parkin. 2014 congressional election cycle campaign website archive. Archive-it.org, 2014. Jens Manuel Krogstad. Hispanics punch below their weight in midterm elections. Washington, DC: Pew Research Center. April. http://www. pewresearch. org/facttank/2014/04/02/hispanicspunch-below-their-weight-in-midterm-elections, 2014. Jan E Leighley. Strength in numbers?: The political mobilization of racial and ethnic minorities. Princeton University Press, 2001. John F McDonald and Robert A Moffitt. The uses of tobit analysis. The review of economics and statistics, pages 318 321, 1980. Adrian D Pantoja and Gary M Segura. Fear and loathing in california: Contextual threat and political sophistication among latino voters. Political Behavior, 25(3):265 286, 2003. Gabriel R Sanchez and Natalie Masuoka. Brown-utility heuristic? the presence and contributing factors of latino linked fate. Hispanic Journal of Behavioral Sciences, 32(4):519 531, 2010. Brian F Schaffner. Racial salience and the obama vote. Political Psychology, 32(6):963 988, 2011. Seth Stephens-Davidowitz. The cost of racial animus on a black candidate: Evidence using google search data. Journal of Public Economics, 118:26 40, 2014. 15

Chris Tausanovitch and Christopher Warshaw. Measuring constituent policy preferences in congress, state legislatures, and cities. The Journal of Politics, 75(02):330 342, 2013. Michael Tesler. The return of old-fashioned racism to white americans partisan preferences in the early obama era. The Journal of Politics, 75(01):110 123, 2013. 16

Table 1: Words Associated with LDA Topics Label Budget Campaign Finance/Wall Street Criminal Justice Economy Education Entitlements Environment/Energy Foreign Policy Healthcare Immigration Infrastructure Labor Technology Veterans Women/LGBT Top 10 Words tax, debt, budget, billion, years, income, year, defi, federal, president war, money, campaign, finance, reform, wall, citizens, street, people, profits gun, violence, guns, justice, law, criminal, federal, background, drug, police jobs, economy, businesses, small, business, create, tax, work, economic, infrastructure education, students, college, school, schools, student, public, programs, children, teachers security, social, medicare, seniors, programs, benefits, retirement, protect, protecting, congress energy, clean, environment, water, climate, environmental, oil, change, renewable, natural security, national, foreign, policy, military, states, nuclear, world, united, peace care, health, affordable, insurance, healthcare, act, coverage, access, quality, costs immigration, reform, immigrants, country, undocumented, comprehensive, izenship, status, states, border transportation, district, federal, housing, public, local, funding, congressional, development, communities wage, minimum, working, workers, families, economic, middle, work, pay, economy science, technology, national, research, government, stem, internet, innovation, access, economy veterans, military, health, families, care, support, women, service, veteran, va women, rights, equal, equality, discrimination, pay, work, congress, civil, woman 17

Table 2: Voter Turnout Dependent variable: Voter Turnout (1) (2) Black 0.273 (0.061) 0.105 (0.044) Dist Dems % Black 0.070 (0.182) Black x Dist Dems % Black 0.412 (0.167) Hisp 0.390 (0.072) 0.761 (0.103) Dist Dems % Hisp 0.891 (0.380) Hisp x Dist Dems % Hisp 1.908 (0.464) Dem Incumbent 0.129 (0.080) 0.151 (0.080) Age 0.026 (0.001) 0.026 (0.001) Block Group Income 0.003 (0.001) 0.003 (0.001) Male 0.042 (0.018) 0.042 (0.018) log loser PAC contribs 0.049 (0.015) 0.046 (0.015) State Fixed Effects X X Year Fixed Effects X X Constant 3.144 (0.167) 3.053 (0.179) Note: Straightlining Districts (all voters either 0 or 1) removed; SE clustered by primary election N = 408,499 p<0.1; p<0.05; p<0.01 Table 3: Candidate Immigration Issue Appeals Dependent variable: Immigration % District Dems Hispanic 12.620 (5.371) Hispanic Candidate 3.944 (2.271) % District Dems Black 5.539 (3.630) Black Candidate -2.092 (1.885) Female Candidate 2.109 (1.288) District Income 0.077 (0.037) District Ideology 1.909 (2.281) (Intercept):1 7.463 (2.950) (Intercept):2 2.163 (0.065) N = 325 p<0.1; p<0.05; p<0.01 18

Table 4: Candidate Criminal Justice Issue Appeals Dependent variable: Criminal Justice % District Dems Black 11.331 (4.088) Black Candidate 4.440 (2.152) % District Dems Hispanic 1.315 (6.419) Hispanic Candidate 0.043 (2.768) Female Candidate 0.103 (1.435) District Income 0.080 (0.042) District Ideology 5.301 (2.591) (Intercept):1 8.503 (3.344) (Intercept):2 2.291 (0.064) N = 325 p<0.1; p<0.05; p<0.01 19