The source or linked survey record has changed since this report was reviewed. An editorial update is pending.
El-Sayed 50%, Rogers 48%; Benson holds 51% to James’s 45% in the governor race.
| Response | Share |
|---|---|
| Abdul El-Sayed, the Democrat | 50% |
| Mike Rogers, the Republican | 48% |
| Another party's candidate | 1% |
| Undecided | 2% |
| Vol: No response | 1% |
Michigan likely voters; Marist turnout-probability weighting of registered voters · Fieldwork 2026-10-15–2026-10-18. Bars use a 0–100% scale. Less-than results retain the source threshold and have no exact-length bar.
What the result tells us
The two-point Senate difference does not establish a clear leader. The same respondents give Benson a larger numerical advantage for governor. That difference between races is useful context, but it does not predict either result.
The governor contest
| Response | Share |
|---|---|
| Jocelyn Benson, the Democrat | 51% |
| John James, the Republican | 45% |
| Another party's candidate | 2% |
| Undecided | 1% |
| Vol: No response | 1% |
Michigan likely voters; Marist turnout-probability weighting of registered voters · Fieldwork 2026-10-15–2026-10-18. Bars use a 0–100% scale. Less-than results retain the source threshold and have no exact-length bar.
Comparable historical context
Registered-voter context: El-Sayed’s margin was seven points in both the September 24–27 and October 15–18 surveys. The likely-voter result measures a different population; its smaller gap is not proof of a late swing.
Earlier registered-voter record · Later registered-voter record
Survey details and sources
Marist interviewed 1,185 registered voters 2026-10-15 through 2026-10-18, using live telephone, text and online methods. The charts use likely-voter weighting; a separate likely-voter question sample is not reported. The published margin of error for likely-voter shares is ±4.3 percentage points. Released 2026-10-21.
Source: Marist Poll. Caldwell Research analysis and presentation. Rounded percentages and less-than bounds are retained.
Original release · Questionnaire and methodology · Survey record