Equal turnout would have moved the mayor’s race by one point

Los Angeles voted on June 2, casting 874,715 ballots against 2,222,958 registered voters, a citywide turnout of 39.35 percent. That average covers an enormous spread. Across the 99 certified Neighborhood Councils, turnout ran from 16.03 percent in Watts (3,357 ballots on 20,938 registered) to 57.01 percent in Studio City (15,431 on 27,070), which means a registered voter in Studio City was 3.6 times as likely to cast a ballot as one in Watts.

The gap holds at council-district scale too, so it isn’t an artifact of small geography: Council District 9 turned out 22.01 percent while Council District 11 turned out 53.32 percent, and fifty-three of the 99 neighborhoods came in below the citywide rate.

The spread is the basis for a familiar argument, that a higher-turnout Los Angeles would elect different people. On the June returns, equalizing turnout across precincts would have changed the mayor’s race by about a point.

The counterfactual

The test starts with the 739 precincts that reported in the June 2026 mayoral contest. For each one we computed a scale factor, registration times the citywide rate of 0.393491 divided by ballots actually cast, then multiplied every candidate’s votes in that precinct by it and re-totaled the city.

That rescaling holds each precinct’s politics exactly where they were and changes only how much the precinct counts, so a precinct at 20 percent turnout roughly doubles while one at 55 percent loses nearly a third. Total ballots don’t change, because the citywide rate is the registration-weighted mean of precinct turnout.

The rescaling assumes the additional voters in a precinct would have split the same way as that precinct’s actual voters did. That assumption carries the whole result, and the section below sets out what happens when it fails.

Bass goes from 292,592 votes to an equalized 301,363, up 8,771, but she still doesn’t clear 50 percent, so there’s still a runoff, and it’s still against Raman. The second-place fight doesn’t tighten; it actually widens, from a 3.49-point gap to 4.04. Across all fourteen candidates, 2.09 points of vote share change hands in total. And the target rate isn’t doing the work here, because rescaling to 100 percent turnout instead produces changes identical to four decimals; a common multiplier cancels.

Why the effect is small

It isn’t small because low-turnout neighborhoods vote like everyone else. Sorting all 739 precincts by turnout and cutting them into five groups of equal registration shows just how differently they vote.

Bass runs nearly ten points better in the bottom quintile than in the top, and Pratt runs ten points worse. Ballot-weighted, the correlation between precinct turnout and Bass’s share is -0.30, while Pratt’s is +0.31.

Two things blunt that difference. The first is arithmetic: equalizing moves the bottom quintile from 12.19 percent of the electorate to 20 and the top from 28.15 down to 20, which is about eight points of electorate changing hands between groups whose preferences differ by about ten. Eight percent of ten points is under a point.

The second is that high-turnout Los Angeles isn’t one thing. The bottom quintile leans one way, but the top is close to a three-way tie, and the neighborhoods inside it are opposites: Los Feliz gave Raman 55.2 percent, Bel Air-Beverly Crest gave Pratt 45.6 percent, and both turned out above 54 percent. Downweighting high-turnout Los Angeles takes votes from Raman and Pratt at once, and the losses cancel. The two candidates who lose the most ground in this exercise are the two who agree on almost nothing.

Non-voters are assumed to vote like their neighbors

The assumption doing the work is that non-voters in a precinct would vote like the voters in that same precinct. That’s an assumption, not a fact, because nobody has interviewed the 1.35 million Angelenos who stayed home.

The error that would matter looks like this. We ran a version where nobody is un-voted, raising the 395 precincts below the citywide rate up to it and leaving the rest alone. That adds 109,605 ballots and lifts turnout to 44.28 percent. Bass gains 0.68, Pratt loses 0.70, Raman loses 0.42, and the order doesn’t change.

For that to have changed which two candidates advanced, Pratt would have needed 45.2 percent of the roughly 106,000 added mayoral votes, against the 19.2 percent his neighbors’ behavior implies, which is well over double his support among new voters in the neighborhoods where he ran weakest. For Bass to have won outright, she needed 144,542 votes beyond her pattern share, and even if every one of the 106,000 added voters had chosen her, she would have finished at 41.5 percent and there would still have been a runoff.

The returns carry no evidence at all about how non-voters would have voted. What this calculation shows is how far they would have had to depart from their voting neighbors, which is what a sleeping-giant argument has to establish.

It isn’t just the mayor

Every citywide contest on the same ballot behaves the same way.

Nothing flips. The largest movement anywhere is 1.50 points, and it comes in a two-candidate race decided by 26. Measure TT was the closest of the three citywide measures at a 5.7-point margin, and it’s the one equalization affects least, moving just two hundredths of a point.

What the result covers

A councilmember who knows Watts casts a sixth of its registered voters while Studio City casts more than half is handed an incentive about whose calls to return, and that incentive runs between elections, over budgeting and permitting and paving, where it never shows up in a vote total.

Rescaling every precinct on this ballot to equal turnout left the same two candidates in the mayoral runoff, in the same order, and changed no leader in any citywide contest. That is a result about one election, and it holds only as long as the added voters in each precinct would have voted like their neighbors.

Method

The universe is the 739 precincts reporting in the June 2, 2026 city primary, from the county Registrar-Recorder’s certified Statement of Votes Cast. Registration and ballots cast come from the county’s TOTAL rows, checked for consistency across every candidate row of each contest. Turnout is ballot-based, which means it counts a voter who returned a ballot and left the mayor’s race blank.

Neighborhood figures use an area-overlap crosswalk to certified Neighborhood Council boundaries. 494 of the 730 covered precincts sit inside one council, and the other 236 straddle a line and are split by area share, which is why neighborhood registration is fractional. The counterfactual runs at precinct level, so that step doesn’t touch it.

Nine precincts belong to no certified Neighborhood Council and are excluded from neighborhood views. All of them sit in Council District 11, covering Brentwood and Pacific Palisades, which have independent community councils instead. They cast 21,621 ballots on 36,152 registered, 2.47 percent of citywide ballots, at 59.81 percent turnout, higher than any certified neighborhood. Excluding them makes the range reported here a floor, because the true spread is wider than 3.6-fold.

Two artifacts are worth ruling out. The first is registration quality. County rolls carry voters who have moved or died and not yet been removed, and that deadwood is generally worse in high-mobility areas, which are also low-turnout areas. If so, real turnout in low-turnout neighborhoods is higher than measured and the rescaling over-weights those precincts, meaning the one-point effect is more likely overstated than understated. The second is rolloff. Blank mayoral ballots run 3.31 percent in the lowest-turnout quintile against 1.99 percent in the highest, and the method carries that difference through, which is why the equalized mayoral total of 852,723 sits just below the actual 853,893 with ballots held constant. It doesn’t affect shares.

Two smaller notes. Six citywide contests here are each short exactly one vote against the county’s certified summary published at results.lavote.gov (text results for election 4338, as of the June 26, 2026 certification), always on the leader, which is consistent with one ballot attributed to no precinct row; one vote in 853,893 moves nothing above. That comparison is against the county’s own published page, not against any file in this dataset, which contains precinct rows only. Eight precincts tied, which doesn’t matter here because nothing in this analysis assigns a precinct winner.