- With default settings, the TWINS model gives two Republicans a 21.2% chance of advancing to the general — a one-in-five shot at the Democratic nightmare scenario.
- With updated campaign finance data and an adjusted partisan vote share, that probability drops to 1.1%. A two-Democrat runoff (16.5%) becomes far more likely than a two-Republican one.
- Tom Steyer's $53M in available primary resources — mostly from F497 contributions reported since the last filing — is the single biggest factor reshaping the simulation.
One of the biggest anxieties hanging over California's 2026 governor's race is the nightmare scenario for Democrats: two Republicans advance through the top-two primary, locking Democrats out of the general election entirely. It sounds far-fetched in a state where Democrats hold a nearly 2-to-1 registration advantage — but in a fractured field with half a dozen credible Democratic candidates splitting the vote, it's not crazy.
Paul Mitchell — Owner of Redistricting Partners and Vice President of Political Data Inc. — built a Monte Carlo simulation called TWINS to put hard probabilities on this exact question. We ran TWINS with updated inputs, and the short version is: Democrats can probably exhale. But the longer version is more interesting.
How the Model Works
TWINS runs thousands of simulated elections for California's top-two primary, each one randomized around a set of baselines built from polling averages, campaign finance data, and structural factors like party registration. It uses a Dirichlet distribution — essentially the right statistical tool for generating random vote shares that add up to 100% — and counts how often each candidate, and each type of matchup (D vs. R, R vs. R, D vs. D), appears in the top two across all those simulations.
The Default Run: 21% R/R
Running TWINS with its default settings (as of March 11, 2026) produces a 21.2% probability of two Republicans advancing — with a D vs. R matchup at 76.1% and a D vs. D outcome at just 2.7%. That's a one-in-five shot at the Democratic nightmare scenario, high enough to justify genuine concern and certainly high enough to fuel a cycle's worth of anxious fundraising emails.
The reason it's that high comes down to the asymmetry of fragmentation. The Democratic field is crowded, with vote share scattered across Swalwell, Steyer, Porter, Becerra, Thurmond, Villaraigosa, and Yee. Meanwhile, the Republican side is more consolidated around Steve Hilton and Chad Bianco — both of whom, in the default run, show higher individual probabilities of making the top two than any single Democrat. In a top-two system, it doesn't matter if Democrats collectively win 62% of the vote. What matters is whether any individual Democrat can outpoll the second-place Republican, and a crowded field makes that harder.
How the Ballot Book Updated the Inputs
Mitchell built TWINS as a framework — it's designed to be fed updated data as the race evolves. The Ballot Book ran TWINS with two adjustments to the default settings: an updated partisan vote share and updated campaign finance figures.
Partisan Vote Share. TWINS defaults to 59% Democratic and 36% Republican, with 5% reserved for minor and write-in candidates. The Ballot Book adjusted this to 62% Democratic and 36% Republican, with 2% for minor candidates. The basis for this is the last competitive gubernatorial primary: in 2018, Democrats captured 62.5% of all votes cast in the governor's race and Republicans took 36.2%, with minor parties and write-ins combining for just 1.3%. Voter registration has shifted slightly since then — Democrats have gained about 1.5 points of registration share and Republicans about 1 point — but those gains came mostly at the expense of No Party Preference voters, leaving the relative D-to-R balance essentially unchanged. That 3-point difference in the Democratic share may sound small, but in a fragmented field where individual candidates are separated by single digits, it meaningfully changes how often a Democrat clears the top two.
Campaign Finance. The Ballot Book tracks a metric called "Available for Primary": cash on hand, plus late contributions received after the last filing period, minus any funds legally reserved for the general election. Raw cash-on-hand figures can overstate what's deployable (by including general-only funds) or understate it (by missing new money reported via F497s) — Available for Primary accounts for both. Here is how our Governor election page tracks the money:
The model's default finance figures are based on the December 31 filing deadline, but candidates have since reported significant new money via F497 late contribution reports. The most dramatic example is Tom Steyer: the default shows $1.9M (his cash on hand as of the 12/31 filing), but since then Steyer has reported over $51M in new contributions via F497s — bringing his available resources to roughly $53M.
Beyond candidate committee money, independent expenditure committees have also made or announced major commitments that aren't captured in the defaults. The Deliver for California IE supporting Matt Mahan recently announced it has invested over $10M in paid communications and claims to have exceeded $10M in total IE receipts — far above the $3.2M reflected in the model's defaults. Eric Swalwell's IE support is also understated, with approximately $2M in commitments not yet reflected. The Ballot Book updated the model's finance inputs to reflect these current figures as closely as possible.
The Results
When these updated figures are fed into TWINS, the picture changes dramatically. The R/R probability drops to just 1.1%. D vs. R rises to 82.5%. And here's the twist: D vs. D jumps to 16.5% — meaning it's now fifteen times more likely that two Democrats advance than two Republicans. The question isn't really whether Democrats get locked out; it's whether Republicans do.
The biggest single driver is Tom Steyer's resource advantage. At $53M in available primary resources — almost entirely from F497 contributions reported since the last filing — versus $1.9M in the defaults, Steyer goes from middling contender to clear frontrunner in the simulation, with roughly a 75% probability of making the top two.
Money alone doesn't decide a governor's race — this is the highest-profile contest on the ballot, and voters will have no shortage of information — but the financial picture is one of several signals the model weighs, and Steyer's spending power is so far beyond the rest of the field that it shifts the overall probabilities substantially. That spending also appears to be paying off in the real world. (The Ballot Book's governor poll aggregator shows Steyer climbing from the low single digits last fall to roughly 13–16% in the most recent March polls.)
The net effect in the model is that Steyer's dominance consolidates the Democratic side enough that Steve Hilton — while still the top Republican in both scenarios — is no longer pulling a second Republican into the top two alongside him.
Caveats Worth Taking Seriously
None of this is a prediction. It's a probability estimate based on current data, and current data is incomplete by definition.
Three months is a long time in a race this fluid. Polling at this stage captures where the race is, not where it's going — and a lot can change between now and June. A major endorsement, a debate breakout, a candidate dropping out, or a shift in the national mood could reshuffle the deck in ways no simulation can anticipate.
There's also a structural limitation: California's top-two primary has only been around since 2012, so the historical track record for a system like this is short — and a field this large makes it even harder to model with confidence.
Finally, it's worth noting that spending in political campaigns has diminishing returns. Doubling a candidate's war chest doesn't double their vote share. The model accounts for this, but it means that even a massive financial advantage like Steyer's gets compressed somewhat in the simulation — the real-world impact may be larger or smaller than what the model reflects.
Bottom Line
TWINS is a genuinely useful tool for thinking about this race probabilistically rather than through vibes and anxiety. Mitchell built something smart, and it's designed to evolve as the data does. With an adjusted partisan split grounded in 2018 actuals and updated campaign finance numbers, the R/R scenario that has Democrats losing sleep looks far less likely than the headline figure suggests — not impossible, but closer to a statistical curiosity than a realistic threat. If anything, the more interesting question the model now surfaces is whether Republicans should be worried about getting locked out themselves.
The race is fluid, the data will keep changing, and the Ballot Book will keep running the numbers.