v0.2.0Evidence preliminaryMichigan Governor · Research Preview · Nov 3, 2026
Jocelyn Benson is clearly favored, though the forecast still allows for a meaningful upset.
This research preview is centered on reviewed polling and calibrated gubernatorial polling error. Certified past results are shown for context and add zero forecast points.
Conditional modeled chance to win · two modeled major candidates only · not projected vote share
Jocelyn Benson +7.480% range: Jocelyn Benson +3.1 to Jocelyn Benson +11.6
Conditional modeled chance to win · two modeled major candidates only · not projected vote share
Jocelyn Benson +7.380% range: Even race to Jocelyn Benson +14.7
Evidence desk · Snapshot note
Jocelyn Benson is favored by a polling-only research preview; certified historical results are shown for comparison, add zero forecast points, and published uncertainty leaves both candidates a credible path.
Published fallback
The published polling-only research preview favors Jocelyn Benson, with current polling pointing to Jocelyn Benson. Certified historical results produce a Democratic comparison margin, but that reference belongs to earlier candidates and adds zero forecast points. The uncertainty range leaves room for either candidate, so this is not treated as a certain outcome.
Jocelyn Benson's evidence-based path is for the polling signal to hold up as more independent surveys arrive. The other candidate needs new verified polling to move that signal; these are conditional paths, not campaign recommendations or separate probabilities.
Watch whether future independent polling confirms or changes the present estimate. That evidence would update the polling center, while the historical comparison would remain descriptive rather than becoming a hidden model input.
Prepared from this forecast snapshot and the sources below.
Jocelyn Benson's clearest path is for the current polling signal to be confirmed by additional independent surveys.
John James's clearest path requires new verified polling to move away from the present signal; the published uncertainty still leaves room for that reversal.
Official forecast unchanged · Commentary only · Forecast effect 0
Deterministic race brief
Built directly from the published forecast and reviewed evidence. Model-assisted answers may explain this record, but they cannot revise it.
Jocelyn Benson is the model favorite. John James remains inside the model's central range of outcomes.
4 reviewed polls from 4 pollsters. Latest fieldwork ended Aug 11, 2026; sponsor: AARP Michigan.
Latest release · AARP ResearchThe central range reaches an even race after rounding, so a close outcome remains within its boundary.
Open the published forecast artifactThese evidence-cited readings explain plausible pressure on the race. Dependency groups separate mechanisms; shared-sample labels identify reused respondents across mechanisms. Neither is independent or additive.
No candidate-directed context has passed review yet.
No candidate-directed context has passed review yet.
The path is strongest if independent polls confirm the current lead and the supportive conditions below continue to hold.
No reviewed candidate-directed context yet.
No opposing directional context has passed review.
If these conditions change, the path changes. This is a scenario, not a second forecast.
The path requires reviewed polling to move toward John James while the supportive conditions below hold.
No reviewed candidate-directed context yet.
No opposing directional context has passed review.
If these conditions change, the path changes. This is a scenario, not a second forecast.
The race can preserve a Democratic executive check or install a Republican governor during a period of divided legislative control.
Observed: Michigan currently has a Democratic governor and Senate alongside a Republican House, so state government is divided.
What would change this: Certified chamber results would establish the post-election control configuration.
Dependency group · Mi State Control 2026
MultiStateThe model does not transfer Gretchen Whitmer's 2022 result to Benson; that result is displayed only as history.
Observed: The sitting governor is term-limited, Jocelyn Benson and John James won their party nominations, and Mike Duggan ended his independent campaign before the primary.
What would change this: A certified final candidate list and repeated full-field surveys would refine the field treatment.
Dependency group · Mi Candidate Field 2026
Associated Press / NPR · Michigan Bureau of Elections · Michigan PublicEvidence desk · Published record
Ask what drives the forecast, what would need to change, or how a conditional scenario could play out in Michigan Governor.
Questions worth asking
Evidence map
The certified-result stations are descriptive references, not sequential model updates. The final two stations show the modeled polling center and forecast uncertainty.
Gretchen Whitmer versus Tudor Dixon, shown only as a statewide historical comparison.
Michigan's certified presidential vote, shown only as a recent statewide comparison.
The midpoint of the certified governor and presidential two-party margins. It is a descriptive comparison with exactly zero separate forecast effect.
Quality-, recency-, sponsor-, and correlation-adjusted estimate from reviewed polls.
The current polling center with uncertainty set by the nearest out-of-time gubernatorial polling-error calibration; historical reference results add no points.
Since the last run
Future changes will be tied to a poll, data release, or documented model update.
Calibration before confidence
Verified context stays outside the probability until it demonstrates predictive value in an out-of-sample election test.
Jocelyn Benson and John James won their major-party primaries. Mike Duggan ended his independent campaign in May; the complete November ballot remains provisional.
The prior governor and presidential results are displayed for context and add zero forecast points.
Reviewed polls use the same recency, quality, sponsor, and pollster-correlation controls as the archived backtest.
2 reviewed assessments explain direction and caveats with exactly zero forecast effect.
The error layer is historically calibrated; the full race-specific publication remains a research preview.
Historical check
Out-of-time production-aggregator polling error by days before Election Day. This calibrates the polls-only baseline and does not validate the full fundamentals model.
Read the complete backtest limits →Race replay
Move through the record one dated input at a time. The tape reveals a probability only when an archived model run exists; news never receives an automatic sentiment score.
The election-day median margin was +7.3 points with an 80% range from -0.0 to 14.7.
Reviewed releases
Swipe the table for other share and model use →
| Pollster / dates | Sample | Jocelyn Benson | John James | Other / undecided | Review / use |
|---|---|---|---|---|---|
| Fabrizio Ward / Impact ResearchSponsor: AARP Michigan | 877 LV | 49% | 44% | — / 7% | AcceptedIn current average |
| TIPP InsightsSponsor: League of American Workers | 1,215 LV | 53% | 37% | — / 10% | Accepted With PenaltyIn current average |
| Beacon Research / Shaw & Company ResearchSponsor: Fox News | 1,006 RV | 52% | 47% | — / 1% | AcceptedIn current average |
| Mitchell ResearchSponsor: Mitchell Research | 672 LV | 46% | 44% | — / 10% | Accepted With PenaltyIn current average |
News with a paper trail
The Democratic and Republican nominees are set for the open-seat general election.
Model treatment: Set the two modeled major candidates; do not transfer the term-limited incumbent's prior result.Duggan's withdrawal removes the best-known independent candidacy from the current field.
Model treatment: Do not model Duggan as a candidate; retain a generic aggregate remainder for any final minor candidates.Three August surveys report Benson leads of varying size, with sponsor, population, and mode differences handled by the aggregator.
Model treatment: Include through the documented poll-quality, sponsor, recency, and correlation controls.Built to be challenged
Every published run points back to its inputs, rules, simulation seed, and model version. Corrections create a new record instead of rewriting history.
Bind the candidate field and certified historical references to dated sources.
Adjust current polling for recency, sponsor, quality, and repeated pollsters.
Apply the nearest out-of-time gubernatorial polling-error calibration across the simulated outcome.
Publish the forecast artifact, source manifest, warnings, and validation state together.