Two recent polls tell very different stories about Latino voters in the Los Angeles mayoral race. One shows Nithya Raman well ahead of Mayor Karen Bass. The other puts Bass in front.
The gap is large enough to raise a question voters deserve an answer to: How can two polls of Latino voters produce such different results, and what do we need to know before trusting either one?
A Berkeley Institute of Governmental Studies poll, co-sponsored by the Los Angeles Times, found Raman leading Bass 45% to 21% among Latino likely voters. The survey was conducted September 15–20 and included 427 Latino respondents. Another 34% were undecided.
A separate poll by BSP Research, conducted October 1–8 for Unidos Con Karen Bass, an independent expenditure group supporting the mayor, reported the opposite: Bass led Raman 40% to 28%, with 32% undecided. The survey included 602 Latino respondents, according to published reporting.
The difference is striking. In the Berkeley poll, Raman led by 24 percentage points. In the BSP poll, Bass led by 12 points. That’s a 36-point swing in the reported margin between the candidates.
The numbers alone don’t tell us why the polls differ. But they do make the methods, timing, sponsorship and underlying data worth examining.
The polls were taken at different times
The Berkeley survey finished nearly two weeks before the BSP poll ended. Voters can change their minds during a campaign, but timing alone doesn’t establish that a major shift occurred.
The Berkeley results also show why the makeup of a poll’s sample matters. Among Latino voters ages 18–49, Raman led Bass 60% to 9%. Among Latino voters 50 and older, Bass led 36% to 26%.
Those differences make age an important factor to examine. The BSP survey’s reported findings also vary by age and language preference, according to published coverage.
But we need the complete tables from both surveys to determine whether their samples differ in ways that could help explain the results. Without that comparison, we can’t tell how much of the gap reflects changing opinions, differences in the people surveyed, or the way the polls were conducted.
A bigger sample doesn’t automatically mean a better poll
The BSP survey included 602 Latino respondents, compared with 427 in the Berkeley poll. That difference matters, but sample size is only one part of the picture.
Researchers also need to explain how they found respondents, how they determined who was likely to vote, and how they adjusted the results to reflect the electorate. Question wording and the order in which questions are asked can matter, too.
The definition of a likely voter is especially important. A survey of people who identify as Latino is not necessarily the same as a survey of Latino residents who are eligible to vote and likely to cast a ballot.
Until the full BSP questionnaire and methodology are available for review, we cannot make a complete comparison of the two surveys.
Who paid for the poll matters, but doesn’t settle the question
The Berkeley survey was conducted by UC Berkeley’s Institute of Governmental Studies and co-sponsored by the Los Angeles Times. The BSP survey was commissioned by Unidos Con Karen Bass, a group supporting the incumbent mayor.
That sponsorship difference is relevant information for readers. It is not, by itself, proof that a poll is accurate or inaccurate.
The same standard should apply to both surveys: What questions did they ask? Who responded? How were the results weighted? Can independent researchers examine the methods and data?
Readers should also be able to distinguish a poll’s findings from the claims a campaign or political group makes about those findings.
A third poll offers another point of comparison
An October 9 survey from the Manhattan Institute provides another view of the race, although it measures the broader Los Angeles electorate rather than Latino voters alone.
The survey of 449 likely voters in the city found Raman at 38% and Bass at 32%, with 30% undecided on the initial ballot question. When initially undecided respondents were asked to choose, the reported result was 39% for Raman and 36% for Bass.
The survey used live telephone interviews, text-to-web interviews and online panel interviews. Its results cannot be directly compared with the Latino subgroup numbers as if all three polls surveyed the same population.
Still, the third survey highlights another reason to look beyond a headline number: the way researchers handle undecided voters can change the result they report.
What the public needs to see
To understand why the two Latino polls differ, the organizations behind them should make their survey methods and underlying documentation available for scrutiny.
That means the complete questionnaires, recruitment methods, likely-voter screening rules, weighting procedures and demographic tables, not just the headline percentages.
The cross-tabulations should include age, language preference, renter status, income, geography and party registration, along with the number of respondents in each subgroup. Those details would help researchers assess whether the polls are measuring comparable groups of voters.
Parriva has not independently verified the complete BSP questionnaire, methodology and crosstabs. Until those materials are reviewed, the reasons for the difference remain unresolved.
The question isn’t which poll to trust. It’s what the evidence shows.
Polling offers a snapshot of opinion among a particular group of people at a particular time. It is not an election result, and even a well-conducted survey has limits.
When two polls report sharply different results, voters shouldn’t have to choose between competing claims of momentum without enough information to evaluate them.
The next step is to compare the surveys’ methods, examine their underlying data and ask both polling organizations to explain the differences. Independent review can help establish which conclusions the evidence supports and which remain uncertain.
For now, the two polls provide conflicting estimates of Latino voters’ preferences in the Los Angeles mayoral race. They do not, on their own, tell us why those estimates differ.
The public deserves the information needed to judge the numbers for itself.
