Pundits, data nerds, and fellow pollsters have noted the rapid changes in the partisan attachments of Latino voters. Story after story suggests that a fundamental transformation has occurred in Latino political attitudes and partisan preferences. A counter theory is that many “mainstream” polls have small and unrepresentative samples of Latinos and should not be read as gospel.  Allow me to explain.

It’s true that the Latino vote is fluid at this uncertain moment in U.S. politics. Why wouldn’t it be? Latinos are hardly a monolith, and demography is not destiny.

We know, for example, that across racial and ethnic groups, highly religious voters have for decades been moving toward the Republican Party; Latinos are no exception. Despite conventional rhetoric aligning Latino voters with the Democrats, roughly 1 in 3 Latinos have long identified with the Republican party. Observers should not be shocked that somewhere around 30 percent of Latinos are indeed Republicans.  Still, Latino voters have proven consequential to Democratic victories given their rapid growth rates and continued majority support for Democrats.

In 2004, Latinos propelled then-Governor George Bush to win the Presidency, a fact that seems all-but forgotten in today’s partisan rhetoric. Bush won Nevada, Colorado, New Mexico, Arizona and Florida with about 38-40 percent Latino support. The truth is, the idea that Latino attitudes are universal or fixed is absurd. But Bush’s victory did not signal a “realignment,” not at all.  In 2008 and 2012 Barack Obama made strong in-roads with Latinos and pushed the Democratic support levels over 70 percent in his 2012 re-election, while carrying Florida. In 2020, Arizona flipped Blue precisely because of the rapid growth and two-thirds Democratic vote among Latinos. 

But equally wrong is the notion that Latino partisan attachments have shifted dramatically in the Trump era. Unfortunately—and often as the result of either too-small or unrepresentative sample sizes—too many polls overstate support among Latinos for either Trump or the Republican Party. Let’s examine these sampling problems and the confusion they can cause.

–   Collecting a Latino sample significant for statistical analysis: The first, and what should be the most obvious factor, is having a significant Latino sample in a study. Many researchers conduct national surveys with standard sourcing of participants. Largely because of population sizes, but also the saturation of White individuals in the panels used to source participation, it is often the case that Latinos are under-sampled. A sample of n=100 Latinos has a +/- of 10 percent. And that assumes the underlying 100 Latinos are actually representative, they usually are not. This sampling error overgeneralizes Latino demographics. While not large in the overall results, the margin of error for subsamples are smaller, allowing for greater interpretative discrepancies. To ameliorate this problem, some seek to include oversamples of Latinos, but what they often fail to capture is the heterogeneity of the sampled population. In one survey, NYT/Sienna reported that 98 percent of their Latino interviews were in English. Um? No. Inaccurate interpretations of the general Latino population can lead to two polls garnering vastly different results.

–  Over-and-under sampling of Latino subgroups: Within the Latino sample, the oversampling and undersampling of Latino subgroups is a second problem that may influence the accuracy and significance of results. Many national samples of Latinos do not properly consider the subgroups and the role they play within the Latino community. When not accurately weighted, shifts in aspects like ideology can appear bigger and more significant than otherwise would be. Small moves in political ideology by one subgroup in a given region can be exaggerated and cue to larger shifts than present.

–  Setting demographic quotas relative to Latino populations: The second error leads directly to a third: The need for proper weights and quotas in samples (and subsamples) that accurately represent the Latino electorate. Demographics such as education and age are different from those of the general White population. The use of inaccurately weighted samples leads to misinterpretation and an eventual mischaracterization of the electorate.

–   Understanding what a “first time voter” is in the context of the Latino electorate: When it comes to polling Latinos in advance of an approaching election, many polls fail to categorize and understand the uniqueness of the “first-time voter” in Latino populations. In the 2024 presidential, one in five Latinos will be voting for the first time. This surge in new voters is driven not only by Latinos entering the citizen voting age population, but also the naturalization of hundreds of thousands of Latinos since 2020. Most registered voter screens, and especially likely voter screens, will miss these voters because they simply do not exist in the datasets from which pollsters draw their samples.

This last point reaffirms an important aspect about both Latino voters and the ability to predict their behavior: As the fastest growing demographic group, Latinos become a more pivotal voting block with each new election cycle, and often the hardest to predict. The Latino population is not just growing the fastest, but arguably is changing most rapidly, too.

And it’s not just the rapid growth of age-eligible Latinos that makes them difficult to poll accurately. Since the dawn of the new century, Latinos have come to populate more states and regions, attend and graduate from college at higher rates, and have diversified in their occupations, religious attitudes, family structures, wealth profiles, and even their expressions of sexuality. Against that backdrop of rapid and profound change, it would be shocking for their political attitudes or partisan attachments to remain fixed.

Nevertheless, Latinos remain a reliably Democratic constituency. Perhaps not as reliable as they were in Barack Obama’s 2008 presidential victory 16 years ago, nor even for Hillary Clinton’s defeat eight years later. It also remains unclear how the dramatic events of this year’s presidential campaign will affect Latino voter splits in the Trump/Harris presidential contest, or relevant down-ballot races this November.

But expect Democrats to win Latinos by comfortable margins this November, even if those margins are not as pronounced as they have been in recent cycles.

Anaís Xitlali López, MPP is a senior analyst at BSP Research