Well, it’s “silly season” again–that time in each electoral cycle when pollsters release results that produce panic on one side, jubilation on another, based on non-credible results. On October 5, Republican pollster Rasmussen released an unweighted, English-only, small-sample Latino vote result that implausibly claimed Trump winning 62% of the Latino vote. This is not true.  

The Rasmussen Latino sample was only a small portion of a non-Latino national poll they conducted and their Latino sample has long been inaccurate. Indeed, other independent pollsters have regularly noted Rasmussen has a strong GOP-bias. Rasmussen’s release is not transparent on the methodology of their Latino sample. While Trump did make some gains in 2020 relative to 2016, there is no credible evidence that Trump continues to gain with Latino voters in 2024.

In fact, it is Kamala Harris who has seen consistent gains in Latino support over the soft numbers Joe Biden faced back in June 2024. The problem with Rasmussen and many other White polling firms is that they do not understand how, nor even try, to accurately poll Latinos. These polls often have very  small samples which are not independently weighted to match Hispanic demographics, and nearly all non-Latino polls have an inaccurate sample of Latinos – too acculturated, too English-dominant, too few immigrants and second generation Latinos. The end result is misleading noise at best and more likely biased incompetence. 

So where does the Latino vote stand today? I’ve reviewed polling for the last two months and focused on pollsters who have accurate Latino samples, large-sample sizes, accurate Spanish-language availability and polls that accurately reflect the true demographics of Latino voters. Overall, there is no question that Democrat Kamala Harris enjoys a large and stable lead with Latino voters, and as Latinos learn more about Harris, her numbers appear to grow stronger.

The most recent evidence comes from the weekly tracking poll by Entravision Media, which owns and operates more than 100 Spanish-language TV and radio stations that are part of the Univision media network. The Entravision poll releases a rolling n=500 national Latino voter sample every week which is weighted to match the demographics of the Latino electorate.  As of Monday, October 7, Harris leads Latino voters 59% to 36%, a margin of +23.  Entravision has tracked this stable lead for Harris, along with steady increases in her net favorability rating for Harris among Latinos, from +25 on September 18, up three points to +28 on September 30.

However it is not just the Entravision weekly tracking poll that shows this advantage: Multiple other national polls with accurate Latino samples confirm this pattern. Based on where Harris stood in early August, she appears to have grown her advantage with Latinos by 6 to 9 points, on average across Latino polling. In fact, there is not a single high-quality Latino sample poll that has Trump over 40% with Latinos, and most have Trump somewhere around 33-35% support, with Harris around 58–60% support. Even if undecided voters break evenly between the two nominees, Harris should finish support in the low- to mid-60% range.

Latino polling methodology and some final thoughts on 2024

In general, for a poll to be reliable, we need a sample size sufficient to reduce the margin of error to something reasonable. For example, 400 cases yield a nominal margin of error of +/- 5%. Extremely small samples should always be suspect. And sub-samples of larger samples are often particularly suspect since they tend to be small and not gathered or weighted with an eye toward representativeness of the smaller group. 

For a sample of Latinos to be reliable, however, the sample must be drawn in a manner that does not produce systematic bias. At a minimum, this requires interviewing in both English and Spanish, as  roughly 16% of the Latino electorate are Spanish-dominant and not comfortable answering sometimes complicated questions in their second language. Additionally, sample weights–calculated to make certain the sample reflects the population–need to be created independently so that national general population weights do not produce weird distortions in the sub-group.

As election day approaches, pollsters of various stripes will release many polls, most ok, some excellent, some less reliable. We encourage readers to check the fine print. See how many respondents were interviewed, whether the results were weighted, whether Latinos had the choice of English or Spanish, and whether smaller populations like Latinos, African Americans, or other groups were over-sampled to have enough for a proper estimate. 

If a result is so shocking to you as to appear wrong–like a shift of one racial group of over 20 points above their previous highest vote mark for a party as we saw here–it probably *is* wrong.

Gary M. Segura, Ph.D. is president of BSP Research and former Dean of the UCLA Luskin School of Public Affairs, and a member of the National Academy of Sciences and the former director of the National Science Foundation marquee project, the American National Election Study (ANES).  Segura has taught courses on polling methodology, statistical sampling and Latino politics for more than 25 years at UCLA, Stanford, University of Washington, University of Iowa and more.