This story was originally published by CalMatters. Sign up for their newsletters.
California’s Latino voters are showing signs of moving away from Donald Trump after helping drive a rightward shift in 2024. CalMatters analyzed precinct-level results from the 2025 election in 57 of 58 California counties, representing over 99% of the statewide vote. Our analysis shows that Proposition 50, Gov. Gavin Newsom’s plan to gerrymander congressional districts in Democrats’ favor in response to a Republican gerrymander in Texas, outperformed the 2024 Kamala Harris campaign among voters of color.
We observed the largest change in majority-Latino precincts. In precincts where Latinos are the majority of the citizen voting-age population, there was a 25 percentage-point net shift to the Prop. 50 redistricting measure. In precincts where the majority of votes cast were by Latino voters, there was a net shift of 29 percentage points.
To determine the net shift, we subtracted the percentage point voting gap in the 2024 presidential contest between Kamala Harris and Donald Trump from the percentage point voting gap between the “Yes” and “No” votes on Prop. 50.
This change in voter behavior could foreshadow a backlash against President Trump and Republicans in the coming midterm election after Latino voters helped put Trump back in the White House in 2024.
These two highly partisan and closely timed elections represent a rare opportunity to analyze concrete voter behavior after the inauguration of the returning president, especially given the Trump administration’s targeting of California for immigration enforcement and federal spending cuts.
Methodology
Data collection and standardization
Statewide Database: We collected 2024 and 2025 election results for 55 counties from the Statewide Database, the official redistricting database for the state of California. The dataset includes information on precinct results, boundaries, and voter racial and ethnic demographics.
Shasta and Tulare counties’ election offices: At the time of publication, the Statewide Database did not have data for Mono, Shasta or Tulare counties, so we collected 2025 results and boundaries from Shasta and Tulare counties’ election offices.
Redistricting Data Hub: For the demographic analysis, we downloaded 2024 Citizen Voting Age Population (CVAP) Census data disaggregated to 2020 census blocks by the Redistricting Data Hub. This Census tabulation was originally produced for voting rights analyses and to assist the redistricting process.
Estimating precinct voter population demographics
We estimate racial and ethnic population demographics at the precinct geography level using the CVAP dataset. The Redistricting Data Hub disaggregates CVAP demographic data from 2020 block groups to 2020 blocks by proportional allocation, based on the ratio of the contribution of a block’s population to a block group.
We used areal interpolation to allocate CVAP population counts from census blocks to precincts. This method allocates population counts proportionally based on how much of a precinct’s land area overlaps with a block. For example, if a block had 100 eligible voters and half its land area was within Precinct A, then 50 eligible voters from that block unit are allocated to Precinct A.
Interpolate 2024 vote counts to 2025 precincts
We used areal interpolation to estimate the 2024 election vote counts for 2025 precincts. The estimate is only used to calculate the vote shift at the precinct level.
We did not rely on this interpolation count for county or state findings; rather, we used it to evaluate changes within individual 2025 precincts. We exercised caution with this measure because it is vulnerable to statistical bias (specifically, the modifiable areal unit problem). We cannot determine with certainty whether the observed vote shift is due to changes in voting behavior or to changes in precinct geography.
Analysis
For our majority-group analysis, we aggregate the vote tabulations for all precincts where a racial group is the majority of adult citizens. This does not mean a majority of voters from a racial demographic group voted a specific way in a precinct; it’s a proxy for estimating voting behavior in geographic areas where those eligible voters are a majority.
We calculate the net shift between the 2024 and 2025 elections. Net shift measures the overall voter swing between elections. For example:
Net Shift = (“Yes” on Prop. 50 vote share – “No” on Prop. 50 vote share) – ( Kamala Harris’ vote share – Donald Trump’s vote share)
Findings
Latino-majority precincts
The largest difference was observed in majority-Latino precincts, where the measure won 73% of the vote statewide, compared to Harris’s 59% in 2024.
Majority-Latino precincts in Southern California reported the largest shift in vote percentage: Los Angeles, Orange, Riverside, and San Bernardino counties shifted a net 27.5 percentage points, voting 60.6% for Harris and 75.9% for Prop. 50.
This finding was not isolated to counties with large urban centers. Kern and Stanislaus counties, both in the Central Valley, reported a 24.5 percentage-point net shift in majority-Latino precincts.
Black-majority precincts
Black-majority precincts, which are located largely in Los Angeles and Alameda counties, had the strongest support for both Harris in 2024 and Prop. 50 in 2025. Harris won with 85% of the vote; Prop. 50 won over 92%.
Asian-majority precincts
Kamala Harris won 60% of the vote statewide in Asian-majority precincts in 2024. The following year, Prop. 50 received 67% of the vote in Asian-majority precincts, a net shift of 10.3 percentage points.
Los Angeles and Orange County have a combined population of over 1.5 million voting-age Asian-American citizens. Analyzing results from the majority-Asian precincts in these two Southern California counties shows a 13.1 percentage-point net shift.
White-majority precincts
Majority-white precincts were the only grouping we observed where the statewide net shift between the 2024 Presidential contest and Prop. 50 was below 10 percentage points. The net shift was 1.6 percentage points: Harris won 56.3% in 2024; Prop. 50 won 58.6%.
Precincts with no racial majority
Our analysis of precincts where no racial group comprises a majority found that the difference in support between Harris in 2024 and Prop. 50 roughly matched the shift in majority-Asian and Black precincts statewide: Harris won 60.7% of the vote in 2024 while Prop. 50 won 68.3%.
Limitations
Comparing a presidential election to a special election
One drawback to our approach is that we are comparing a presidential election to a special election held in an off-year. Presidential general elections often have higher turnout than special elections: 15.9 million votes in 2024, compared to 11.6 million in 2025. Any shifts in voter behavior from the 2024 election to the 2025 election represents a smaller pool of voters as a result. Turnout for the 2025 special election was on par with recent midterm and special elections in the state at 50%.
Population distribution and areal interpolation
Areal land interpolation estimates voter population demographics under the assumption of a uniform population distribution. The population is not distributed uniformly. We attempt to minimize this concern by using CVAP data disaggregated to census blocks.
This differs from the Statewide Database’s approach, which uses restricted voter roll data and registration addresses to assign voters to their Census block and precinct. Using this methodology, the Statewide Database has determined that 80% of census blocks place all their respective registered voters in the same precinct rather than splitting them across multiple precincts.
Racially Polarized Voting Analyses
Racially Polarized Voting Analyses (RPV) are used to evaluate compliance with the Voting Rights Act (VRA). One method used is the homogeneous precinct method, which compares precincts in which 100% of voters belong to a single racial demographic group. We conducted a geographic analysis of the election results and did not attempt to make a conclusive finding regarding the VRA. We reran our analysis with increasingly high thresholds up to 100% and observed that the net shift towards Prop. 50 remained above +5 percentage points in voter-of-color precincts and reached as high as +36.6 percentage points for precincts where Latino eligible voters are over 90% of the eligible voters. It decreased in majority-white precincts, reaching as low as -7.4 percentage points in precincts where white eligible voters accounted for over 90% of eligible voters.
Redactions
Election administrators redact results from precincts with a small number of registered voters to protect secret ballots.
Census Survey Data vs Voter Rolls
Our analysis relies on Census Survey data that estimates the number of voting-age citizens in a census geography. Similar analyses use information pulled directly from election administrators’ voter rolls. Voter rolls include information such as name, address, race, etc. Voter roll data can be used to run an ecological inference to draw conclusions about the voting behavior of different groups.
We reran a version of our analysis incorporating voter file data from the Statewide Database. We categorized precincts according to whether the majority of the voters are Latino and calculated the net shift.
Acknowledgements
We thank Redistricting Data Hub’s Spencer Nelson and Peter Horton for reviewing an early draft of this methodology.
We would also like to thank Matt Barreto (UCLA) for methodological feedback.
Download Data
We used AI coding tools to assist with the analysis. All results were manually reviewed by members of our team. You can access and download our analysis here.
Mountain View moves forward with $4.6M plan to fix eroded banks at Stevens CreekMay 5, 2026 11:19 am
