Michigan Election 2026 Live Desk: Turnout, Polls and Forecasts

Michigan Election 2026 live desk. This is BadPD’s maintained record for early voting, absentee participation, statewide polls and published third-party forecasts ahead of the November 3 general election. It is not an early-results page. Candidate vote totals are not available from participation data. Every number needs a source, a date and a clear description of what it measures.

Michigan’s required statewide early in-person voting period began October 24 and runs through November 1. Check your own location and hours at Michigan’s official voter information center. Communities can have different arrangements. Do not use a news headline as a substitute for your clerk’s current listing.

Participation update, October 30: the official dashboard’s 7:13:26 a.m. snapshot records 1,858,624 ballots statewide and 286,478 in Wayne County. The board uses that daily snapshot, not a real-time count. Older figures remain below as dated history. These are participation records, not candidate totals or a county executive race forecast.

Original-source poll check, October 30: our maintained poll report now includes MRG’s October 22 ballot-measure release, covering October 9-14 interviews. It reports Proposal 1 at 31% yes/49% no/20% undecided and Proposal 2 at 63% yes/16% no/22% undecided. The second total is 101%; the release expressly warns about rounding. This is one separate ballot-only sample, not a new October 30 survey, another candidate-board row or a reliable trend against a different pollster. The original release’s methods, undisclosed details and inconsistent footer remain attached.

October 30 evening poll update: the Mitchell companion check now supplies governor figures of Benson 49%/James 44% for the existing October 26-27, 636-likely-voter sample. CBS Detroit’s new question report adds Gilchrist 51%/Forlini 42% for secretary of state and Savit 44.6%/Lloyd 42.2% for attorney general. These are additional questions, not additional samples or official votes. Thirteen candidate samples remain thirteen; ten now have both Senate and governor findings. The original Senate date conflict, governor chart-label error, attorney-general rounding and unverified detailed question wording remain disclosed.

Voter service, after 5 p.m. October 30: the Friday online/mail-received absentee-application cutoff has passed, not the completed-ballot return deadline. Our updated guide puts the remaining clerk-office, early-voting and ballot-return routes near the top. For voters already registered at their current address, the state gives November 2 at 4 p.m. for an in-person absentee application. Eligible late registrants have a separate Election Day clerk-office route. Check official local hours and your own eligibility; Monday early voting is not universal.

Voting-access update, October 27: our voting guide now separates absentee application cutoffs, ballot-return deadlines and the clerk-office registration route. The state’s October 27 notice also says some communities offer additional early voting on Monday, November 2. That is optional, not a statewide extension; check your assigned site’s published hours.

Ballot-problem update, October 28: the guide now covers tracking limitations and lost-ballot replacement dates, plus the state’s ballot-envelope signature-cure process. The official 2026 calendar sets November 6 at 5 p.m. for that cure; it is not extra time to return an ordinary domestic ballot. Ask your clerk about your own notice and do not vote twice. This service update does not refresh the daily turnout snapshot or change poll/forecast figures.

Accessibility update, October 28: the voting guide now separates ballot-marking equipment, curbside help, permitted assistance and accessible absentee voting. The electronic absentee option described by state instructions still requires printing and physically returning the ballot; it is not online vote submission. The guide links the October 19 civil-rights notice, the accessible-ballot request cutoff and the responsible accessibility/complaint offices. Local arrangements need clerk confirmation. This service addition does not create new turnout, poll or forecast figures.

Polling-place service update, October 29: the voting guide now explains the photo-ID affidavit route, limits on watchers questioning voters and police body-camera privacy guidance. Michigan’s official in-person instructions say a voter without photo ID can use the prescribed form. This does not replace registration, residency requirements or a lawful eligibility review. The guide also separates campaign-activity boundaries from a blanket ban on observers or responding police. These are newly checked service details, not new laws, a finding against a local site or additional turnout observations.

Go to the data board | Polls | Wayne County executive | Update policy

October 30: Michigan records 1,858,624 ballots; Wayne County 286,478

BadPD reloaded the official participation dashboard, confirmed the November 3 general election and selected Wayne County. The source card reads October 30, 2026, 7:13:26 a.m. It does not print a time zone; the ledger uses Michigan’s local Eastern time while preserving the exact wording.

Statewide, 1,422,463 absentee returns plus 436,161 early in-person ballots total 1,858,624. Wayne County has 220,525 absentee returns plus 65,953 early in-person ballots, totaling 286,478. The source still shows Election Day data as unavailable, which is not a reported zero.

Compared with the dated October 29 snapshot, statewide participation is higher by 153,909 and Wayne’s by 24,410. Those are net differences between daily observations. The state explains that clerks’ voter-history entries can be corrected or added later. Do not call the differences a precise count of people who voted during Thursday, candidate votes or evidence that a party is ahead.

The state now reports 7,415,593 active registered voters and 25.1% active-voter turnout; that denominator also changed from the prior snapshot’s 7,412,507. Wayne’s displayed active-voter turnout is 23.1%. These rounded percentages do not measure all residents or predict final turnout. Wayne’s 354,076 absentee requests and 335,201 permanent-mail voters are different categories from its 220,525 returned absentee ballots. None reveals a voter’s choices, knowledge or current location.

October 29: Michigan records 1,704,715 ballots; Wayne County 262,068

BadPD checked the official Department of State dashboard with the November 3 general election selected, then selected Wayne County. The source card reads October 29, 2026, 7:14:55 a.m. Its card does not print a time zone; the data ledger uses Michigan’s local Eastern time and retains the exact card wording.

Statewide, 1,346,372 absentee returns plus 358,343 early in-person ballots make 1,704,715 recorded ballots. The dashboard reports 23.0% turnout against 7,412,507 active registered voters. That is a specific voter-file denominator, not a share of all residents or a prediction of final turnout.

Wayne County’s 207,969 absentee returns and 54,099 early in-person ballots sum to 262,068, with active-voter turnout of 21.1%. It separately shows 353,451 absentee requests and 334,877 permanent vote-by-mail voters. Those categories are not interchangeable: a request does not prove a ballot was returned, and a permanent-mail designation does not show where someone currently lives or whom they supported.

The recorded totals exceed the October 28 snapshot by 152,243 statewide and 23,167 in Wayne County. Within those differences, absentee returns rose 81,293 statewide and 12,013 in Wayne; early-voting records rose 70,950 and 11,154. These are changes between source snapshots, not an independently counted number of votes cast on October 28. Clerks can add or correct records retroactively. Election Day entries remain unavailable, not zero.

There is still no candidate tally here. These figures cannot establish support for Michael Evans, Warren Evans, John Olumba, or a Senate or governor candidate. They cannot measure voters’ local knowledge or establish nonresidence or fraud. Candidate performance belongs in the linked records comparison; participation belongs in this separately dated count.

October 29 tracker check: zero displayed is not zero early voters

Our evening check of On Point Politics’ Michigan tracker found 1,376,007 ballots, all assigned to mail, with zero in person. That wording persisted in the normally rendered page, not just a search snippet. Its Wayne row likewise showed zero in person. The page dates its data October 29 and attributes counts to civicAPI; it does not establish an exact matching cutoff for the official snapshot.

The official dashboard, rechecked with the November 3 election and Wayne County selected, still gives the October 29, 7:14:55 a.m. snapshot: 358,343 early in-person ballots statewide and 54,099 in Wayne. The two displays cannot both be complete measures of that category. Different mail-count cutoffs are a separate issue; do not subtract their headline totals and call the difference missing or fraudulent votes. The cause of the tracker’s zero is not established here.

On Point explicitly labels Michigan’s party mix as modeled because the state does not publish early voters’ party. Its party advantage is therefore an estimate, not a count of ballots marked for a candidate. That disclosure matters. A model can be useful without becoming an official return, and it cannot establish a county executive leader or how informed voters are.

The state explains that its daily figures come from voting-history records entered in the Qualified Voter File, can change after corrections and do not equal tabulator-based election results. We retain the official observation, source time and category definitions on this board. This source check adds no survey, turnout observation or BadPD winning odds. No provider clarification, inquiry or refusal is claimed.

Previous snapshot, October 28: Michigan 1,552,472; Wayne County 238,901

BadPD checked the Department of State’s official dashboard with the November 3, 2026 general election selected, then selected Wayne in the county filter. The source card reads October 28, 2026, 7:05:07 a.m. It is a daily snapshot; refreshing this page does not create another source observation.

Statewide, the dashboard lists 1,265,079 absentee returns and 287,393 early in-person ballots. Those sum to 1,552,472. Its active-voter turnout measure is 20.9%, against 7,411,144 active registered voters. That denominator is not Michigan’s total population.

Wayne County has 195,956 absentee returns and 42,945 early in-person ballots, totaling 238,901. The county’s active-voter turnout measure is 19.2%. The dashboard separately lists 352,816 absentee requests and 334,534 permanent vote-by-mail voters. A request is not a returned ballot, and a permanent-mail designation is not a candidate preference or evidence that someone lives elsewhere.

Compared with the dated October 27 snapshot retained here, the recorded totals are higher by 160,302 statewide and 23,108 in Wayne County. Those are differences between two published snapshots, not an independently measured number of people who voted yesterday. The state says clerks can upload or correct records retroactively. Election Day data is currently unavailable, not zero. None of these figures identifies support for Michael Evans, Warren Evans, John Olumba or any statewide candidate.

Previous snapshot: the October 27 turnout receipt

The state’s participation dashboard was updated October 27 at 7:05:36 a.m. It showed 1,392,170 ballots statewide: 1,176,375 absentee returns and 215,795 early in-person ballots. Active registered voter turnout was 18.8%. The active registration total was 7,409,996. The percentage uses that active-voter denominator; it is not the share of all Michigan residents.

For Wayne County, the same dashboard showed 183,703 absentee returns, 32,090 early in-person ballots and 215,793 total ballots recorded. Active registered voter turnout was 17.4%. Absentee ballots requested totaled 352,067. Requests and returns are different stages. They must not be added together as votes cast.

The figures are a dated snapshot, not a second-by-second counter. They can change when clerks add records or correct entries. The dashboard explicitly distinguishes participation from official or unofficial results. It cannot reveal a candidate’s support, a voter’s knowledge or a motive for using the mail.

BadPD keeps the snapshot’s date attached even if you are reading this later. A fresh page load does not make old source data new. Use the official dashboard for the latest available aggregate figures and local filters. If its refresh is delayed, a blank or unchanged value is not evidence of misconduct.

What the 2022 mail-ballot comparison actually says

Added October 27: the state’s one-week-out comparison lists 1,176,375 returned mail ballots on October 27, 2026, versus 1,126,246 on November 1, 2022. Subtracting those counts gives 50,129 additional returns, or about 4.5% above the earlier count. That percentage describes returned mail ballots at comparable points before Election Day, not a 4.5-point increase in final turnout.

The state also says this is the first statewide midterm general election with early in-person voting available to every Michigan voter. Adding this year’s early-site ballots to its mail returns and comparing that sum only with 2022 mail returns would mix unlike measures. A voting-method shift can change the composition of the early count without identifying who benefits.

This comparison cannot tell us how absentee voters voted, whether they follow local issues, or which Wayne County executive candidate leads. It supplies historical participation context, not candidate results or a turnout forecast. The live board still uses the dated daily snapshot; this explanation does not refresh its source timestamp.

Michigan polls: thirteen samples, different methods

The data board now lists thirteen documented samples, not thirteen new polls released today. Ten provide both Senate and governor figures; three report Senate only after the late-October Mitchell governor companion was checked. Today’s CNN/SSRS source check adds October 20-27 fieldwork: El-Sayed 49% / Rogers 42% and Benson 53% / James 41%, including follow-ups. The original report’s governor question for already-voted respondents mislabels the state as Kansas while naming Michigan candidates. That unresolved wording defect stays visible; it is not silently corrected.

Secretary-of-state question added: WDIV’s October 28 release reports Gilchrist 50.6% / Forlini 39.2%, with 10.2% don’t know/refused, from the existing October 19-22 Glengariff sample. The linked question check preserves its named-candidate framing and method. It does not add another survey or establish a Wayne County result.

Attorney-general question added, October 29 evening: WDIV’s October 28 release puts Savit at 45.0%, Lloyd at 40.1% and don’t know/refused at 14.9%. Our original-question check keeps the October 19-22 interviews, named-candidate wording and four-point reported margin attached. The 4.9-point difference is not a win probability. This adds a question to the existing Glengariff sample, not a fourteenth survey or a Wayne County executive result.

Mitchell source check, October 28 evening: the latest PDF reports a seven-point Senate gap, not a win probability. Its cover/footer range conflicts with a date inside the methods paragraph. The table preserves that warning. The older companion releases count as one sample, not two. The historical comparison does not establish that the same people switched sides, a statistically significant change or a Wayne County executive result.

Marist interviewed registered voters October 15-18 and separately reports likely-voter estimates. MSU/YouGov’s October 2-19 toplines distinguish initial answers from leaner-inclusive results. Neither overall sample should be mislabeled the likely-voter sample.

RedBox/AFWH’s October 20-23 survey is super-PAC-sponsored and retains its unresolved invitation/interview timing caveat. The Glengariff governor release and Senate companion describe one 600-person telephone survey, October 19-22, not two samples. On Point’s October 18 interviews supply initial and model-allocated answers; the board uses initial likely-voter estimates, with its modeling limits attached.

Added October 27, Eastern time: InsiderAdvantage’s October 23 original release reports Senate headline figures of El-Sayed 47% and Rogers 46%. It lists October 21-22 fieldwork, 1,200 likely voters, mixed-mode text/panel, demographic and political-affiliation weighting, and a 2.83-point margin of error. Detailed tables, question wording, response options, leaner treatment and an outside sponsor remain unverified or undisclosed. No governor figure is given; the board labels that cell “Not reported in this release.” These are older interviews newly checked here, not a new October 27 survey.

Source resolution, October 28: the recovered Guidant memo dated October 6 reports 800 likely voters interviewed September 20-27. Its full-field Senate result is El-Sayed 48% / Rogers 43%; governor is Benson 50% / James 45%. The separate head-to-head Senate question is not another poll. The date-check section separates the memo from October 7 public reporting, not a proved first-release date. It identifies the attributed Public First Action sponsorship and missing methods. This is not new October 27 fieldwork, despite conflicting aggregator labels.

Earlier-record batch, October 28: added YouGov’s October 7 publication and Rasmussen’s October 16 Senate summary. The source comparison also checks Tyson’s October 15 sponsor memo: its older Rogers lead is not today’s race call, and its claimed advertising effect is not independent evidence. Missing uncertainty stays labeled, never invented.

Field periods overlap and methods differ. The table is selective, not a polling average or a measured surge. A poll’s margin of error is not a win probability. No statewide poll here measures the Wayne County executive race.

Proposal 1 release, October 28: WDIV published another question from Glengariff’s October 19-22 interviews: 34.4% yes, 48.5% no, 15.9% don’t know and 1.2% refused on convening a constitutional convention. The poll report explains the question and official process. This is the same sample, not an additional survey. The largest answer is not a majority, and a yes vote would not itself replace the constitution.

Getting there: the voting guide now checks free-fare offers from SMART, CATA-area providers and TheRide. Dates, coverage and reservations differ. Some offers cover early-voting or registration trips; others concern November 3 only. Free fare is not a guaranteed ride or an eligibility decision. Verify your destination and return trip with the responsible provider.

Proposal 2 release, October 28: WDIV’s campaign-finance question reports 73.8% yes, 17.0% no and 9.2% don’t know/refused after a description of the measure. The source check covers the official ballot’s utility/contractor restrictions and disclosure provisions. Same Glengariff interviews, not another survey or enacted law.

Forecasts: attributed, dated and fallible

The data board shows published third-party probabilities in a section apart from polls. Read the source-dated cards for the last verified figures from The Hill/DDHQ Senate page and DDHQ governor page, rather than treating a fixed article paragraph as a live quote. A successful source check does not mean the provider issued a new forecast. A failed check preserves the dated last verified snapshot with a warning, not a claim that those figures are current.

Those are model probabilities of winning, not predicted vote shares. A favored candidate can lose. These are not BadPD’s own odds or official results. DDHQ describes a blend of polls, structural factors and market information. Its published governor methodology explains the approach and its limits. We do not reproduce a betting recommendation.

Automated source checks are scheduled hourly. That does not mean the provider updates hourly. The board preserves the provider’s timestamp and flags stale data or a failed check. If a page changes format or refuses a fetch, the last verified figure stays dated; it is never replaced by zero or relabeled fresh. Forecast display ends when the election reaches the configured polls-close boundary.

Wayne County executive: three candidates, no invented leader

The official county candidate list, dated September 10, lists Warren C. Evans, Democrat; Michael A. Evans, Republican; and John Olumba, with no party affiliation. Michael Evans and Warren Evans are different candidates. The office is county executive.

BadPD has not located a verified public head-to-head poll of that contest in this review. We are not assigning a win probability or declaring Michael Evans ahead or behind. A statewide Senate or governor survey cannot answer that county question. August primary totals also cannot be treated as the November head-to-head vote.

Records comparison published October 27: our expanded Wayne County report examines Warren Evans’s financial-agreement exit, housing and service oversight, John Olumba’s enacted legislation and recorded health-coverage vote, and the evidence gaps in Michael Evans’s campaign service claims. It separates documented actions from proposals and incomplete corrective work. It is not an endorsement or a county race forecast.

Three reports to read next

Wayne County executive comparison: Michael Evans, Warren Evans and John Olumba, checked against records. Read the dated laws, roll call, court opinion and oversight findings, including what they do not prove.

Michigan Senate and governor polls: thirteen documented samples and the date/leaner checks. Read the comparison before treating a published survey as movement or a forecast.

Michigan early voting and absentee return guide. Check the statewide window, official lookup, return rules and Detroit-specific Election Day absentee distinction.

How this live desk is maintained

Live desk means a maintained reporting page, not a promise that every source is real time. The board checks its public data endpoint while open. Forecast source checks run on an hourly server schedule. Turnout snapshots and polling entries require a source review before being changed. The official participation service follows its own daily update cycle.

We keep confirmed facts, attributed claims, estimates and pending records separate. A social post is a lead until checked. No voter-level file, ballot image, private identifier or individual voting history is needed for this desk. Reporting on election access should preserve the exact notice, location and date without exposing voters.

When official returns begin, they belong in a separate results section with the election, jurisdiction, reporting time and certification status. Missing returns are not zero votes. A media projection is not certification. A changed total needs an explanation from the reporting source, not an automatic fraud accusation.

Coverage also needs a way to hear from readers without turning rumors into facts. A useful tip identifies an agency, a public decision, a dated notice or a specific access problem. Tell us what record would settle the question. Do not send a private voter’s identification, signature or ballot selection. A concern about a candidate’s record should point to a decision or service request that can be checked. We will not publish a claim merely because it helps one party. We also will not ignore a supported claim because it hurts a favored candidate. The same source test applies throughout this desk.

Update ledger, October 27, 2026: launched with the state/Wayne participation snapshot, two original statewide poll releases, dated DDHQ forecasts and the candidate roster; expanded to five poll releases and a linked Wayne County candidate-record comparison. Pending: newer corrective-action evidence, further primary poll releases and official general-election results. No county executive race call has been made. Corrections identify the affected claim and source date.

October 27 follow-up: added a third poll release, RedBox/AFWH, with sponsor and method caveats. Refreshed the Senate forecast to its October 27 provider update. Participation remains the October 27, 7:05:36 a.m. source snapshot; it has not been relabeled real time.

October 27 afternoon: the polling table now covers five surveys after original Glengariff and On Point releases were checked. The voting guide also adds official intimidation-reporting contacts. Forecast checks continue separately; no turnout observation or candidate result has been invented.

October 27 late update: checked and added the original October 23 InsiderAdvantage Senate release, with disclosure gaps labeled. Six survey entries now appear; the daily participation timestamp and third-party forecast values were not manually changed.

October 28 participation update: replaced the board’s October 27 participation values with the source’s October 28, 7:05:07 a.m. snapshot after checking statewide and Wayne County views. Added the count differences and retained the older receipt as history. Poll entries, candidate-record findings and provider forecast values were not manually changed.

October 28 polling-source update: added the recovered Guidant memo as an older seventh survey, not a fresh late-October poll. Preserved the original field dates, question versions, attributed sponsor and missing disclosures. The daily participation snapshot and server-owned forecast values were not manually changed.

October 28 source-expansion update: the board now has ten surveys after original earlier YouGov, Rasmussen and Tyson records were checked. The missing-MOE display does not invent uncertainty. No turnout count or server-owned forecast probability was manually changed.

October 28 afternoon service update: added the newly released Proposal 1 question to the existing Glengariff record and linked the transit-service checks. The poll count remains ten; turnout and provider forecast ownership are unchanged. Original source dates and service restrictions remain attached.

October 29 update: advanced the board to the official 7:14:55 a.m. participation snapshot, preserving the older receipts and source limits. Added CNN/SSRS as one paired sample and the secretary-of-state question to the existing Glengariff record. The board now has thirteen samples; server-owned forecast values and hourly checks remain separate. No new post, image or county race call was created.

October 29 evening: linked the newly checked attorney-general question from the same Glengariff sample. The daily participation source, thirteen-sample count and server-owned forecast cards are unchanged by this editorial update.

October 29 night source check: compared the rendered On Point participation tracker with the still-dated official dashboard. Preserved the zero-in-person discrepancy, the provider’s modeled-party disclosure and the unknown cause. No candidate votes, fraud finding or extra poll row inferred.

October 30 source update: advanced the board to the official 7:13:26 a.m. daily participation snapshot. Added the original October 22 MRG ballot-only survey to the maintained report and a separate data-ledger category, not the thirteen-row candidate table. The record preserves field dates, rounding, methods, missing sponsor details and inconsistent source labels. Original article dates, candidate identities and server-owned forecast fields remain intact.

October 30 evening update: completed the existing Mitchell governor row and linked its new statewide-office question check, keeping thirteen candidate samples and the separate ballot-only MRG record. Added the remaining voting routes after the Friday application cutoff. The official daily participation source and server-owned forecast fields were not manually changed.