There are two ways to add votes between a June primary and a November runoff. You can persuade people who voted for somebody else, or you can bring out people who did not vote at all. In this race the two paths point in different directions, and they point differently for each candidate.
Here is the arithmetic, with no forecast attached to it.
The persuasion math
In June, 853,893 people cast a vote for mayor. Karen Bass took 292,592 of them and Nithya Raman took 247,781, a gap of 44,811 votes or 5.25 points. The remaining 313,520 votes, 36.72 percent of the total, went to the twelve candidates who are not on the November ballot. Spencer Pratt holds 217,977 of those and the other eleven split 95,543.
If every June voter comes back in November and nobody new joins them, Raman passes Bass by winning 57.1 percent of that unassigned bloc. Counting only the Pratt vote and splitting the smaller candidates evenly, the requirement is 60.3 percent.
| Raman's share of the unassigned vote | Bass | Raman | Margin |
|---|---|---|---|
| 40% | 56.30% | 43.70% | Bass +12.59 |
| 45% | 54.46% | 45.54% | Bass +8.92 |
| 50% | 52.62% | 47.38% | Bass +5.25 |
| 55% | 50.79% | 49.21% | Bass +1.58 |
| 57.1% | 50.00% | 50.00% | tie |
| 60% | 48.95% | 51.05% | Raman +2.10 |
| 65% | 47.12% | 52.88% | Raman +5.77 |
| 70% | 45.28% | 54.72% | Raman +9.44 |
An even split of the eliminated vote returns Bass's primary margin almost exactly. Everything depends on how far from even it lands.
The eliminated votes are concentrated in precincts with the city's highest homeownership rates, and precinct renter share is the variable most strongly associated with Raman's own support at +0.438 against Pratt's -0.523. These are relationships between precincts, not between individual voters. On geography alone, 57.1 percent is a demanding number for her.
The turnout math points the other way
June turnout ranged from 22.01 percent in Council District 9 to 53.32 percent in Council District 11, a 2.42-fold spread. Citywide, 874,715 of 2,222,958 registered voters cast a ballot, a rate of 39.35 percent. That leaves 1,348,243 registered Angelenos who did not vote, which is substantially more people than decided the race.
Those non-voters are not spread evenly, and the pattern runs against the primary result. Across the fifteen council districts, turnout correlates with each candidate's vote share like this:
| Candidate | Correlation between district turnout and their share |
|---|---|
| Bass | -0.575 |
| Pratt | +0.389 |
| Raman | +0.358 |
Bass ran strongest where the fewest people voted. Her best district in the city was Council District 8, where she took 61.80 percent on turnout of 29.49 percent, the second-lowest in Los Angeles. Her third-best was Council District 9 at 42.24 percent, on the lowest turnout in the city at 22.01 percent. Raman's and Pratt's support both sit in higher-participation territory.
Ranked by the raw number of registered voters who stayed home, the top of the list looks like this:
| District | Did not vote | Turnout | Bass share | Raman share |
|---|---|---|---|---|
| CD 8 | 101,884 | 29.49% | 61.80% | 18.38% |
| CD 12 | 101,784 | 42.81% | 31.72% | 19.03% |
| CD 7 | 100,495 | 32.06% | 35.26% | 20.65% |
| CD 15 | 97,651 | 31.27% | 39.16% | 18.16% |
| CD 10 | 93,818 | 36.91% | 44.49% | 31.28% |
| CD 14 | 90,869 | 37.63% | 32.67% | 37.54% |
The largest single pool of non-voters in the city sits in the district Bass won by 43 points.
Equal turnout moves Bass 1.08 points
We can put a number on it. Rescale all 739 precincts so that every one of them turned out at the citywide 39.35 percent, hold each precinct's candidate preferences exactly as cast, and recount.
Under that counterfactual Bass gains 1.08 points, Raman loses 0.63 and Pratt loses 1.18. Bass goes from 34.27 to 35.34 percent.
Equalizing turnout across precincts therefore moves the result by about a point, in the incumbent's direction. A point matters in a race decided by five, though it is a smaller effect than turnout arguments usually claim. The method and its assumption, that added voters would split like their precinct's actual voters, are set out in our piece on the turnout gap.
Putting both paths together
The two mechanisms interact. Suppose Raman wins the unassigned vote 55 to 45, which on the persuasion table alone leaves her 1.58 points short. Now vary how well each side holds its own June voters:
| Change in Bass's June base turnout | Bass | Raman |
|---|---|---|
| -10% | 49.04% | 50.96% |
| -5% | 49.93% | 50.07% |
| no change | 50.79% | 49.21% |
| +5% | 51.62% | 48.38% |
| +10% | 52.42% | 47.58% |
A five percent slip in the incumbent's own returning electorate flips a race she otherwise wins. The position is two-sided: Bass has more unrealized votes available to her than Raman does, and she is also the one more exposed if her June voters do not come back.
Precinct turnout repeats across cycles; vote share does not
Our own precinct panel shows the difference. Precinct turnout between 2022 and 2026 correlates at +0.921, so the earlier value accounts for about 85 percent of the variance in the later one. Support for Bass across the same two cycles correlates at about +0.32, accounting for about 10 percent. That ratio of explained variance is the sense in which turnout is roughly eight times more predictable than vote choice; the correlations themselves differ by a factor of about three. We can say with some confidence which neighborhoods will vote in November. We cannot say from this data how they will vote, and any model built on a relationship that weak would produce a number far more confident than the evidence underneath it.
We have not built one, and we do not intend to before November 3. What is above is a set of conditions: how much of the eliminated vote each candidate needs, how many non-voters sit in each district, and how sensitive the result is to each. The campaigns are working the same arithmetic.