Colorado’s June 30 primary is our first live test: state the numbers now, before the real result is certified, and grade ourselves after.
The open Democratic gubernatorial primary pits Senator Michael Bennet against Attorney General Phil Weiser for the seat Jared Polis vacates. We synthesized a Colorado Democratic primary electorate and simulated the race.
Method
We built a population of 10,000 synthetic likely Democratic primary voters, raked to the Colorado Secretary of State’s active-voter file on the margins that matter here:
| Dimension | Target source |
|---|---|
| Age band | CO active registration, Dem + unaffiliated |
| Region | Front Range / Western Slope / Eastern Plains |
| Urban density | County classification |
| 2024 turnout | Modeled likely-voter screen |
Each synthetic voter was interviewed: a vote-choice question, a confidence rating, and one open-ended “why”. We ran the full population 100 times to get a distribution rather than a point estimate.
What the simulation believes
Across 100 runs, the mean two-way share:
| Candidate | Mean share | 90% interval |
|---|---|---|
| Michael Bennet | 53.4% | 49.1–57.8% |
| Phil Weiser | 46.6% | 42.2–50.9% |
The interval crosses 50%. The simulation does not call this race a lock. It leans Bennet on the strength of statewide name recognition, but the Western Slope and Boulder-area synthetic voters split enough to keep it live.
What they said
The open-ended answers are where a simulation earns its keep. Two patterns:
- Bennet leaners cited federal experience and “someone who can stand up to Washington” far more than any policy specific.
- Weiser leaners clustered hard on local record — the AG’s consumer-protection and reproductive-rights work — and on wanting “someone who’s actually been fighting here.”
“I like Bennet in D.C. But the governor’s job is Colorado’s job, and Weiser has been doing Colorado’s job.” — synthetic voter, Boulder County, age 41
That is the kind of sentence a crosstab cannot give you.
How we will grade this
When the real result certifies, we will score three things:
- Winner call — did we name the right candidate?
- Margin error — |simulated margin − actual margin|, in points.
- Coverage — did the actual result fall inside our 90% interval?
We are publishing the number first. Check back for the report card.
Synthetic estimates describe a simulated population, not a poll of real Colorado voters. This is not a forecast of the actual outcome.