Two newly released surveys put Republican Mike Rogers seven percentage points behind Democrat Abdul El-Sayed in Michigan’s U.S. Senate race. The similar gaps come from different samples, interview periods and response methods.
CNN’s October 29 report places Rogers at 42% and El-Sayed at 49%. SSRS interviewed 768 registered Michigan voters October 20–27 online and by telephone, then weighted the results by predicted likelihood of voting. CNN reports a margin of sampling error of ±4.7 percentage points for the Michigan results.
Mitchell’s October 28 release reports Rogers at 44% and El-Sayed at 51%, with 3% undecided and 2% choosing another candidate. Its 636 likely voters include 121 online-panel respondents and 515 text-to-web respondents. The report lists a ±3.9-point margin of error. The conflicting field-date language in the source is a reason to retain a qualification alongside these figures.
The four releases cover different portions of October.
The other recent releases in our comparison show a smaller reported gap. Glengariff’s October 27 report for WDIV and The Detroit News has Rogers at 45.1% and El-Sayed at 51.4%, a 6.3-point difference. Its 600 likely voters were interviewed October 19–22 by live telephone; the report gives a ±4-point margin of error.
Michigan State University’s October 26 report has Rogers at 45.5% and El-Sayed at 48%, a 2.5-point difference. Its online fieldwork ran October 2–19. The likely-voter findings use 836 respondents within a 1,000-adult Michigan sample, with ±4.4 points reported for the likely-voter sample.
When respondents were interviewed · October 2026
Reported uncertainty does not predict the election result.
Each survey measures a selected group of respondents under its own assumptions. A published margin of sampling error is not the probability that a candidate will win, and it should not be applied mechanically to the difference between two candidates. Likely-voter models, nonresponse, question wording and mode can also affect the findings.
The sequence of release dates alone does not establish that voters moved from one candidate to another. MSU interviewed across much more of October than the later surveys. Treating its release as a same-day reading would conceal that distinction.
Other candidates and undecided answers still matter.
This comparison displays the Republican and Democratic shares. Response options and the handling of undecided voters vary, and the two percentages do not describe the entire ballot. The atlas’s official candidate directory preserves the complete active candidate field.
Use the Michigan polling page to compare the underlying records, read source qualifications and download selected results. It presents individual ballot tests without turning them into an average or an election forecast.