Why we can’t figure out the Hispanic vote

Hispanic voters might be the most closely watched demographic in 2022. But we still don’t have much reliable electoral data for them.

by Natalie Jackson, NJ

Hispanic voters could easily be the x-factor that determines whether Democrats or Republicans hold Congress in November. In 2020, a rightward shift among some Hispanics—most notably in South Texas and Florida—resulted in a few surprise congressional wins for Republicans and stronger-than-expected support for then-President Trump. In general, the 2020 trend doesn’t seem to be reversing. Trouble is, no one seems to agree on whether it’s accelerating or leveling off.

Why don’t we know more? Two reasons: First, the entire political field tends to treat Hispanics as if they are a monolithic ethnic group that is attitudinally the same across the country—and they are not; secondly, public political surveys in the media often don’t have large enough samples of Hispanics to be able to definitively say where they fall in elections.

When we talk about Hispanics in the U.S. (the terms “Hispanic” and “Latino” are often used interchangeably), we’re lumping together people from dozens of different countries, with different histories, unique experiences, and distinct reasons and timelines for emigrating to the U.S. Hispanic populations weren’t even counted on the Census until 1970, when they comprised less than 5 percent of the U.S. population, so we’re talking about a population that has mostly emigrated in the last 50 years.

As of the 2020 Census, Hispanics comprise about 19 percent of the country (although they were likely undercounted). According to analyses of voter files and exit polls, Hispanics comprised 9-13 percent of the electorate in 2020. (Race and ethnicity are not recorded on most voter data, so these are estimates based on models and samples.) Most of the election surveys in the media are national population surveys—meaning they are designed to represent the national population of all adults, registered voters or likely voters. That means that in a survey of adults, about 19 percent of respondents should be Hispanic, and in a survey of voters, somewhere around 9-13 percent should be Hispanic.

Most election surveys have total sample sizes of around 1,000 Americans (generally considered a good sample size for overall representation), which means there are at most fewer than 200 Hispanics included. Voter surveys would have maximums around 120. Surveys in states with large Hispanic populations would have more, as their proportion dictates, but state surveys are often conducted with smaller sample sizes.

These sample sizes of 120 or so are generally sufficient to report results, but it’s critical for consumers to realize that those numbers are small and the margins of error are large. Whereas a survey of 1,000 Americans has an overall margin of error of about 3.5 percentage points, a sample of 120 will have a margin of error of 9 percentage points. When we translate that to election questions, it becomes impossible to tell whether a shift from 65 percent Democrat support in 2020 to 60 percent in the most recent poll is a real change—yet that would be more than enough of a shift to alter outcomes in a close election.

A recent New York Times/Siena College survey “oversampled” Hispanic voters (meaning, they polled more Hispanic voters than they would need to in a nationally representative survey) and got to a sample size of 522 with a margin of error of 5.9 percentage points. That’s better, but still imprecise for determining whether Hispanic voters are moving a few percentage points to the right or not.

Additionally, these sample sizes are unable to differentiate by nationality and lived experience. Mexican Americans in Texas and Cuban Americans in Florida might have drastically different reasons for moving rightward in their voting patterns in 2020 that could reverse or continue independently of the other group. Nor can analysts look at other political and demographic dividing lines within such small samples. There is clear evidence that Hispanics are as divided by religion as anything—but again, most election polls in the media are not large enough to capture that dynamic.

The Hispanic population is not the only group election polls struggle to understand. Black Americans are treated similarly to Hispanics in surveys. Asian American and Pacific Islanders—a similarly broad, diverse category—are rarely even in the conversation. Native Americans, multiracial Americans, Middle Eastern Americans, and anyone who does not identify with a broader category are often combined with AAPI into an “other” category, if they are reported at all.

There are survey groups that are increasing sample sizes and designing polls to understand Hispanic groups better, and there are plenty of private firms that focus on Hispanic voters for campaigns and candidates. We’re not seeing that much in public election polls, though. What we do see is a lot of headlines out of major media polls based on sample sizes that are far too small to offer any sort of precision. That means we’re not going to learn anything about small but critical changes to Hispanic voting patterns until they actually vote.


Natalie Jackson writes the “Leading Indicators” column for National Journal. She is currently director of research at the nonprofit, nonpartisan Public Religion Research Institute, where she develops surveys and research on religion, culture, politics, and policy.

Be the first to comment

Leave a Reply

Your email address will not be published.