Deceit, diversity, or mobilization?

Abstract

What explains President Bush's increased vote in the State of Florida in the 2004 election? A common perception is that implementation of electronic voting machines and a surge in GOP registration increased Bush's vote margins relative to the 2000 election. In this paper we offer an alternative explanation: massive Puerto Rican immigration combined with successful Republican mobilization of this group explains about 14% of the increase in Bush's margin of victory—approximately 50,000 votes. Scholars’ failure to account for intra-ethnic diversity, by employing a “panethnic” approach that treats Latinos as having identical political preferences, leads scholars to overlook the important role of Hispanic subconstituencies in the 2004 election.

Notes

2 One columnist, for example, wrote “George W. Bush's vote tallies, especially in the key state of Florida, are so stunning that they border on the unbelievable” (Perry, 2004).

3 We acknowledge the debate over which term is most appropriate: “Latino” or “Hispanic” (Hero, 1992). However, following the lead of de la Garza (2004) we use these terms interchangeably.

4 While the importance of intra-ethnic differences are widely recognized by race and ethnicity scholars, non-specialist scholars commonly rely on catch-all ethnic and racial classifications (e.g., de la Garza, 2004). For example, we used the JSTOR search engine to count the number of articles published in the American Political Science Review, the American Journal of Political Science, and the Journal of Politics between the years 2000 and 2002 that mention the words “Black”, “African-American”, “Hispanic”, or “Latino”. Of the 461 articles published, 33% mention a racial or ethnic group. Of these articles that mention a racial or ethnic group, 87% are not specifically about race or ethnicity.

5 U.S. Census American Community Study (2004). Voting Supplement November 2004.

6 For example, the 1982 amendment to the 1965 Voting Rights Act led states to create majority–minority congressional districts based on the premise that the members of a minority group have similar political views.

7 Racial and ethnic minorities in the United States tend to support the Democratic Party over the Republican Party (de la Garza, 2004; Smith and Seltzer, 1992; Tam, 1995). This is largely the case because in the aftermath of the 1960s partisan dealignment, the Democratic Party has been seen as the party of civil rights. Moreover, minorities tend to be less prosperous and more reliant on the state for support. Cuban Americans and Nicaraguan Americans are the most prominent exceptions.

8 For example, Cuban immigrants are far more likely to say that they are “Cuban American” than to identify themselves as a “Latino” or “Hispanic.”

9 While we successfully replicated all of their results, this paper builds on the full model of the Bush vote as reported in Table 2 model 2 in Hout et al. (2004a). We build on this result because it serves as the basis for their substantive conclusions about electronic voting fraud in Florida.

10 However, Puerto Ricans in Florida tend to be more conservative than those in other states (March 2008).

11 According to the 2004 US Census American Community Survey, November Voting Supplement as calculated by the proportion of Hispanics who report that they were registered to vote.

12 According to the 2003 US Census American Community Survey, the average Mexican American immigrated to Florida in 1991. In contrast, the average Cuban American immigrated in 1978. The average non-Cuban or Puerto Rican immigrated in 1988.

13 Much of this spending, particularly in Tampa and Orlando was offset by Democratic 527 groups, like the New Democratic Network, but such expenditures are less likely to be effective because they both spent on general messages about the parties and they are not allowed to directly endorse Senator Kerry.

14 Support from Cuban Americans was seen as essential to Clinton's 1996 victory in Florida in which he is reported to have received about 40% of the Cuban vote. That showing was contrasted by Gore's poor showing among Cubans in 2000 in the wake of the Elian Gonzalez affair where Gore received about 17% of the Cuban vote.

15 Nonetheless, the differences are robust across a range of alternative specifications.

16 The electronic voting variable misses significance at conventional standards dropping from p ≤ .007 in model 1 to p ≤ .059 in model 2, an 857% decrease in our confidence that this coefficient is not 0. Substantively, the size of the coefficient also drops substantially, about 25%.

17 Moreover, the accuracy of existing polling data examining Latino behavior has been called into question (e.g., Leal et al., 2005).

18 One limitation of these analyses is that data allowing for estimation of change in Puerto Rican and Cuban American's between 2000 and 2004 are unavailable as Census data via the American Community Study are available only for the 17 most populous of Florida's 67 counties. Importantly, however, survey data validates our results. The Latino National Study shows that Bush obtained 56% of the Puerto Rican vote in Florida in 2004. While figures from 2000 are not available for Puerto Ricans, this figure is much higher than commonly expected, as the national Latino Bush vote in 2000 is estimated at 40%. Similarly, Bush obtained 73.8% of the vote among Cuban Americans in 2004, down from 83% in 2000. Otherwise stated, Bush's vote total was slightly depressed among Cubans in 2004, as our results in Table 2 predict.

19 We ran a full suite of regression diagnostics which suggest that our results are quite robust. While we find no evidence of collinearity, heteroskedasticity is indicated by a significant coefficient on the Breusch Pagan (Cook-Weisberg) test. Hence, we re-estimated the results using White's corrected (robust) standard errors. All variables remain highly significant across estimations. Finally, the results are quite robust to the influence of particular observations as dropping the largest outlier (Orange County 2004), increases the strength of our findings as both the Puerto Rican and Cuban American variables remain highly significant and correctly signed.

20 Of course, these analyses assume that the campaigns were effective enough in their targeting and strategy for us to detect the effects.

21 Specifically, the coefficient on Puerto Rican is significantly smaller than the coefficient on the interaction between Puerto Rican and Year (p < .01, F = 41.71). The coefficients on Cuban and Year are also significantly different (p < .01, F = 8.45).

22 One limitation of these analyses is that allowing for estimation of change in Puerto Rican and Cuban American's by county between 2000 and 2004 are unavailable as Census data via the American Community Survey are available only for the 17 most populous of Florida's 67 counties are available. Importantly, survey data validates these shifts as the 2006 Latino National Survey shows that 59.2% of Florida's Puerto Ricans supported Bush in 2006 consistent with the estimates presented above.

23 At no other time in the 2000 election was this antipathy towards Gore clearer than when the Democratic Cuban American Mayor of Miami-Dade County, Alex Penelas, chose to go on vacation rather than campaign with Al Gore during the last two weeks of the election (DeFede, 2000). Gore would later call Penelas “the single most treacherous and dishonest person” he dealt with during the 2000 campaign (Associated Press 2004).

24 For example, the 2004 American National Election Study contains responses from only 54 Mexican Americans, 16 Puerto Ricans and 4 Cuban Americans.

25 Leal et al. (2005) argue that this problem affects attempts to identify the preferences of the Latino community as whole in 2004.

26 Similar problems are encountered in every Florida survey conducted by a national organization in 2000 and 2004. We examined the following polls: Voter News Service 2000, CBS/NY Times Florida Poll 2000, National Annenberg Election Study 2000, The Pew Hispanic Poll 2004 and the National Election Pool Exit Poll 2004.