How High is 44%?

Just how unprecedented is the 44% figure? Over the past four decades, the Latino vote has always supported Democratic presidential candidates, albeit by varying numbers. Contrary to some claims, there is no clear trajectory to this support—in the last two decades, there has been substantial variation.

The most concise guide to Latino preferences is the polling data brought together by DeSipio (1996) and DeSipio, de la Garza, and Setzler (1999), covering presidential elections from 1960 to 1996. One caveat is that we do not know how “Latinos” voted before 1972, as the only available data are for Mexican Americans. This is important because Cuban Americans are more likely to support Republican candidates. In this 1960 to 1972 period, the average Mexican-American vote for the Democratic candidate was 84%.

From 1976 to 1996, we see that 68% of Latinos on average voted for the Democratic candidate. This ranged from a low of 56% in 1980, when Ronald Reagan defeated Jimmy Carter, to a high of 72% in 1996, when Bill Clinton defeated George H. W. Bush and Ross Perot. 1976 and 1996 are also similar low points in the Latino Republican vote—18% and 21%, respectively. That Latino support only increased three percentage points during this period casts some doubt on the claim that a long-term realignment of Latinos is taking place (although 6% of Latinos voted for Perot in the latter contest).

Some analysts describe the relative success of Ronald Reagan in 1984 among Latinos—winning about 37% of the vote—and compare it to the present election. These analysts rarely note that Latino Republican support dropped to 32%, 28%, and 21% in the next three elections, which suggests a personal and not a partisan dynamic. While Ross Perot's candidacy in 1992 and 1996 complicates the issue, if we assume that Latino Perot voters would have split evenly in those years without Perot on the ballot, the Republican level of support in the “post-Reagan plummet” was 32%, 31%, and 24%. The same dynamic could be true for Republican candidates in the post-Bush era—or we could see the start of a Latino realignment.

We might be cautious about claims of growing Latino support for Republicans because similar claims in the past did not withstand scrutiny. While some pundits and politicos claimed that such Latino support grew from 2000 to 2002, for instance, James Gimpel showed that the Latino vote for Democratic candidates remained fairly stable. Comparing Latino voting in both elections, he concluded that: “Republicans had made no appreciable gains among Latinos. To the extent that the proportion of Latinos voting Republican had increased, it was principally a consequence of their low turnout in heavily Democratic areas coupled with the lopsidedness of some of that year's gubernatorial elections” (Gimpel 2003). Nevertheless, “Following the 2002 mid-term elections, the strong Republican showing brought legions of GOP pundits, consultants, and sympathetic journalists forward to crow about the on-going realignment of Latino voters.” Gimpel also argued that a similar dynamic took place in the Texas 1998 gubernatorial elections, whereby many thought Bush “had made inroads into the Latino community there when, in fact, legions of Latino Democrats were deactivated by an especially weak Democratic candidate. This was also the story of the 2002 midterms.”

Nevertheless, the idea that Latinos can change their partisan allegiances, or at least cross over in a particular election, is not impossible to entertain. For instance, one-third of Latino Republicans are former Democrats (de la Garza et al. 1992, Table 8.25). In addition, there are some regions where Latinos have “flip flopped” in adjacent elections. For instance, according to pollster Sergio Bendixen: “If there is a model for why the Hispanic vote is considered a swing electorate in the United States, it would be the Orlando market. The way they voted in 2000 and 2002 is classic” (Padilla 2004). This refers to the Latino (and mostly non-Cuban) vote in Florida along the I-4 corridor, which supported Gore in 2000 but then switched to support Governor Jeb Bush in 2002 (Padilla 2004).

The Exit Poll Controversy

Following the November 2, 2004 election many scholars and pundits questioned the accuracy of national exit polls that showed Kerry narrowly winning battleground states that he ultimately lost. Despite inconsistencies between the exit polls and the official state vote totals, few journalists asked questions about the “surprising” Latino vote. Instead, the 44% figure was accepted by many and often repeated.

In 2004, there were three major exit polls. The first was conducted by Edison/Mitofsky for the National Election Pool (ABC, Associated Press, CBS, CNN, FOX, and NBC) and the second was by the Los Angeles Times. The former showed Bush receiving about 44% of the Latino vote, and the second indicated a 45% level of support—both a considerable increase from 2000 when Bush won 35% of the Latino vote.¹

This was also a substantial increase in just nine days, as the final telephone survey of Latino voters conducted by the Miami Herald and Zogby International showed Bush polling at just 33% among Latino voters (Corral 2004).

The third exit poll, conducted by the Willie C. Velazquez Research Institute (WVI), conflicted with the first two. It found that 31.4% of Latinos support Bush while 67.7% supported Kerry.

Such large differences between the polls sparked some debate. According to the research director of the WVI: “We were kind of shocked ourselves at the [exit polls] and started re-checking our figures. We reverified everything and came to the conclusion that the weighting system used by the networks was inaccurate” (Olivera 2004). Andy Hernandez at Trinity University in San Antonio said that only 30% of the respondents in the Edison/Mitofsky survey were from urban areas, while 46% were from the suburbs, and 25% were from rural areas. “In other words, 70 percent of the people surveyed are not from urban areas. Suburban areas tend to have higher incomes, and the precincts are Anglo-dominant, so they tend to vote more Republican” (Gomez 2004). The problem, according to Antonio Gonzalez, president of WVI, is that: “We're the most urban electorate in America. There are not lots of rural or suburban Latinos anywhere. What you get when you have a general market survey is one that shows more Latinos who are Republicans” (Tolson 2004).

On the other hand, Warren Mitofsky responded that, “We stand behind the [44%] number” (Fears 2004).²

Pollster Sergio Bendixen suggested that the problem lies not in the exit polls but in the pre-election surveys:

A recent report by the National Council of La Raza (NCLR) discusses the possibility that the 44% figure could be credible if there was a late shift of undecided Latinos toward Bush and an even split of new (since 2000) Latino voters (NCLR 2004). Using this scenario, which NCLR considers “unlikely” (4), one arrives at a Bush Latino support figure of 43.75%. The NCLR report also cites other points made by supporters of the exit poll results. For instance:

According to Louis DeSipio of UC-Irvine: “The national exit poll [Edison-Mitofsky] is trying to be representative of the population of the 50 states. It's not paying attention to the composition of the Latino community.” DeSipio also noted that the WVI survey likely overrepresented Latinos in urban areas, while the Edison/Mitofsky poll likely underrepresented urban Latinos (Puzzanghera 2004). He concluded that Latino support for Bush is likely in-between the figures generated by WVI and Edison/Mitofsky. “I think it's more realistic to say that 37, 38 or 39 percent of Hispanics voted Republican. And that still shows that more Hispanics vote Republican, so it's still significant” (Gomez 2004).

Comparing the accuracy of these polls will not be possible unless they all provide information on their sampling strategies and fully release their data.³

In addition, we might keep in mind what occurred in the previous election: “Immediately after the 2000 election, exit polls put Mr. Bush's support at 40 percent. That figure was quoted all over the country by major news media and GOP pollsters. It ended up being closer to 35 percent, a figure now widely accepted as more accurate” (Olivera 2004).

Data from the Pre-Election Surveys

As noted above, one reason why the 44% statistic is suspect is that it conflicts with almost all of the major preelection surveys of Latino registered voters. An assessment of 10 surveys conducted between June 2003 and October 2004 found strong and consistent support for John Kerry and the Democratic Party. On average, they suggest that Kerry might have expected a 2-to-1 advantage in the 2004 contests. Table 1 lists the name of the polls, their respective sample sizes, and the Latino anticipated vote between Kerry and Bush. Figure 1 graphically illustrates the trend over time.

In none of the 10 surveys does President Bush reach 40% among Latinos, and in nine of the 10 he was found to receive less than 35% of the Latino anticipated vote. Support for Kerry hovers around the 60% mark. Overall, the surveys conducted in 2004 combine for a weighted average of 32% for Bush and 60% for Kerry, with less than a 1% margin of error (given a combined 9,923 sample size).

Furthermore, two of the surveys were carried out by the same research/media partnership—the Tomás Rivera Policy Institute (TRPI) in conjunction with the Washington Post and Univision Television—which provides an ideal comparison of Latino preferences at two points in time. The two national telephone surveys of Latino registered voters, using the same sampling methodology,⁴

found Latino vote preference was quite stable during the 2004 campaign. The July TRPI survey found Kerry with a 60–30 advantage among Latinos, and three months later, the October TRPI survey found Kerry with the same 2-to-1 advantage. It is therefore difficult to imagine that Latino vote preferences changed to a 53–44 split in just two weeks.

Second, with little exception, support for John Kerry was strong across a wide variety of Latino respondents. As seen in Table 2, when broken down by education, income, age, and immigration status, the Latino vote consistently favored the Democratic candidate in the TRPI survey taken two weeks prior to the election. For example, with respect to education level, Kerry held a 40-point advantage over Bush among Latinos with less than a high school diploma (63.5 to 23.5), but he also maintained a 25-point edge among the college-educated (58.6 to 33.9). While Kerry lost and Bush gained support among more educated Latino voters, the margin was still close to 2-to-1 in favor of the Democrat. Similarly, the Kerry lead over Bush was 25 to 30 points among Latino households that earn less than $25,000 and also among those that earn more than $50,000 per year. Further, immigrants and U.S.-born Latinos displayed very similar vote preferences for Kerry over Bush. In sum, with regard to education, income, age, and immigrant status, every subsection of the Latino electorate stated a vote preference lower than 35% for President Bush.

The one exception appears to have been religion, whereby a Catholic/Protestant split is evident. Among Catholics, who compose about 70% of all Latino voters, Kerry held over a 40-point lead. Among nonCatholics, Bush was favored by 13 points (51.1 to 38.3). Given the large lead for Kerry among Catholics, and their numerical majority among the Latino electorate, non-Catholic Latino turnout would need to be almost 100% to elevate Bush's Latino support to 44%—a feat even Karl Rove would have trouble achieving. However, the data in Figure 3 from the October TRPI poll suggest that non-Catholics were only slightly more likely to report being mobilized than Catholics—28% versus 24%—which seems to suggest that reports of hypermobilization among Latino Protestants are overstated.

Building on the vote preference numbers, Table 3 reports party identification, including leaners, for Latino registered voters by demographic subgroups. In short, the percentages reported in Table 3 are quite consistent with the vote percentages reported in Table 2. Overall, 66% of Latinos ally themselves with the Democratic Party, 24% with the Republican Party, and 10% identify as independents. While some of the percentages differ across various socioeconomic indicators, the patterns are very similar: Latinos continue to identify with the Democratic Party by well over a 2-to-1 margin. While pundits and scholars alike often point out that polls do not account for turnout differences, the implication here is that higher turnout among the high-SES groups would not considerably skew the Latino electorate or their preferences. High-SES and low-SES Latino voters reported strong attachments to the Democratic Party and support for its nominee, John Kerry.

Along these same lines, when asked which party has more concern for Latinos, the Democratic Party is favored 57 to 15 over the Republican Party, with 24% citing no difference. Again, within each subsection of the Latino electorate Democrats are seen as having more concern for Latinos. Even non-Catholics rate Democrats more favorably than Republicans by nearly 25 points.

While the Democrats enjoy a registration advantage, it is possible that in 2004 they were disadvantaged when it came to the issues. Following the election, some journalists suggested that Latinos' conservative religious values led them to the Republican's “moral values” campaign. However, polls conducted on issue salience prior to the election do not bear this out. Among every slice of the Latino electorate—educated and uneducated, poor and rich, young and old, foreign- and native-born, Catholic and Protestant—the number one election issue was the economy. The war on terrorism consistently rated second, followed by the situation in Iraq, education, health care, and immigration-related concerns. These findings are consistent with previous research that suggested economic/education/health issues are high priorities for Latinos and that socio-religious values are less critical on Election Day (Barreto et al. 2002).

Religion and the Latino Vote, Part II

Given the possible role of religion in this election, we should further explore this dynamic. According to Adam Segal, director of the Hispanic Voter Project at Johns Hopkins University, “The Bush campaign used moral values, and especially the national discussion over gay marriage and abortion rights, as wedge issues within the Hispanic community to try to break off a conservative religious segment” (Johnson 2004). Commentary in the New York Times said that Latino support for Bush “happened, in part, because the Republicans went to church” (Curiel 2004). An article in the San Jose Mercury News reported that: “The conservative group Focus on the Family had a major Latino outreach program revolving around opposition to abortion rights and gay marriage. The group aired Spanish-language advertisements on 200 radio stations nationwide and sent information to about 13,000 churches” (Puzzanghera 2004).

Indeed, the 2004 election reveals a “religion gap” in the electorate. According to the NEP exit poll, George W. Bush did particularly well among Protestants, born-again Christians, and regular church-goers. For example, Bush beat Kerry by 19% (59 vs. 40) among Protestants, by 57% (78 vs. 21) among born-again Christians, and by 22% (61 vs. 39) among weekly church-goers. It is also true that Bush fared better than he did in the 2000 election among less religious voters. Few would doubt, however, that religious conservatives have become a pillar of the president's coalition.

One might therefore wonder how religious conservatives in the Latino electorate cast their ballots in the 2004 elections. If Bush indeed garnered 44% of the Latino vote, Latino religious conservatives should have supported the president in disproportionately large numbers. No other particular segment of Latino voters—with the exception of Cuban Americans and Republicans—seemed to rally behind the president. Cuban Americans—a traditional, staunch voting bloc for the Republicans—were solidly behind the president, and so were Latino Republicans. As reported in Table 6, the October Washington Post/Univision/TRPI poll shows that 80% of Cuban-American voters, who made up about 6% of the Latino electorate, indicated they would support Bush in the presidential election. Twenty-four percent of Latino voters in the same poll identified with or leaned toward the Republican Party, and 86% of Republican identifiers and leaners reported their intention to support Bush. While such levels of support for the president are impressive, these two groups of voters alone are not enough to make up 44% of the Latino vote.

As shown in Table 7, Bush indeed did very well with Latino evangelical or born-again Christians. Latino evangelicals favored Bush over Kerry by 58% to 33%, and Bush also garnered 49% of the vote among other Christians (including mainline Christians). However, Kerry enjoyed strong support from Latino Catholics (68% vs. 24%) and non-Christians or seculars (62% vs. 29%). This indicates there was a religion gap within the Latino electorate; Latino non-Catholic Christians were part of Bush's coalition of religious voters and helped with his bid for re-election. One should note, however, that nonCatholic Christians constitute only 18% of the Latino electorate, as the great majority of Latino voters are Catholic. Despite their relatively high level of support for Bush, the number of nonCatholic Christians is too small to help Bush reach 44% of the Latino vote.

With that said, we have observed that there are two segments of voters in the Latino electorate that are considered to be part of the Republicans' winning coalition—Cuban-American and nonCatholic Christian (mostly evangelical) voters. Table 8 reveals that non-Catholic Christians were also more likely than Latino Catholics or seculars to identify with the Republican Party. For the Republicans, appealing to Latino evangelicals and other non-Catholic Christians may be the key to making (small) inroads to the Latino electorate, while Latino Catholics would be an important part of a future winning coalition for the Democrats.

Evidence from Texas

In addition to the pre-election surveys, there is evidence to suggest that Latinos may not have supported Bush in the commonly-cited mid-40% range. Consider the state of Texas, where one out of four Latino voters lives. The national exit poll numbers suggest that Bush won the “Tejano” vote by about 59–41 over Kerry,⁵

but there is some preliminary evidence to suggest that Kerry did much better among Latinos in Texas than the exit poll data suggest.

In Table 9, official vote totals from heavily Latino counties along the Mexico border show Kerry winning comfortably. For example, in Starr County, which is 98% Latino, Kerry won 74% of the vote. In Zavala he won 75%, in Duval he won 71%, in Brooks he won 68%, and in Dimmit he won 66%—all counties with over 85% Latino populations. Given that he did so poorly in the South Valley region, for Bush to have captured almost 60% of the statewide Latino vote, he would have had to win 80% or more in the urban centers.

However, the electoral data in Table 10 for heavily Latino precincts in Dallas suggest otherwise. In precinct 3100, which is 92% Latino, Kerry won 87% of the vote. Across town in precinct 1120, Kerry won 84% in a neighborhood that is 85% Latino. According to ecological inference analysis for all precincts in Dallas, Kerry won an estimated 71% of the Latino vote. This is consistent with the Washington Post/Univision/TRPI survey, conducted two weeks before the election, which showed Kerry with a 2-to-1 margin over Bush among Latinos in Texas. While the border county and Dallas analyses do not include Latinos in many parts of Texas, there are not enough probable Latino Republican votes in the rest of the state to get to the 59% level (Teixera 2004).