Thomas J. Smith suspended his campaign for Congress on July 16, 2026. He endorsed one of his rivals on the way out. He had missed the deadline to get his name off the ballot, so the name stayed there, and on August 4 the Republican voters of Michigan's 8th District nominated him anyway, with 50.44% of the vote, over a candidate President Donald Trump had endorsed. Then he accepted. "I am humbled and honored that Republicans have chosen me to take the fight to Kristen McDonald Rivet," Smith told Michigan Public the next morning.
Kalshi, the federally regulated exchange where people buy and sell contracts on real-world outcomes, now lists a separate market on every one of Michigan's 13 congressional districts. Ten of them are priced as lopsided. Three are live arguments: MI-07, MI-10 and MI-04. On all three, the price-blind AI panel we run against these boards came back below what the market is paying for Democrats, which is not what you would guess from a state where the news has run one direction for a month. Smith's seat, the 8th, is the one place the panel and the money agree.
The Republican Who Quit His Own Race And Won It Anyway
Kristen McDonald Rivet won MI-08 in 2024 by 6.6 points, 51.25% to 44.61%, in a Saginaw Bay seat that covers Flint, Saginaw, Bay City and Midland. National Republicans spent the year treating her as their best Michigan target. What they got out of the August 4 primary was Smith, whose campaign had been over for 19 days when the votes were counted, beating Trump-backed former Navy officer Amir Hassan 50.44% to 33.18%, with Al Lemmo, the man Smith had endorsed, taking 16.38%.
Under the Michigan Secretary of State's filing rules for congressional candidates, a nominee in that position had until August 24 to formally decline and come off the November ballot. That date has passed. On August 25, the Cook Political Report moved MI-08 from Lean Democratic to Likely Democratic, and on the same day moved the open 10th District from Lean Republican into Toss Up. Those two moves are the reason people started searching this race.
Kalshi's MI-08 Democratic contract bids 90.5 cents and asks 94.1. Our panel put McDonald Rivet at 93%. Of the four Michigan races anyone is actually contesting, MI-08 is the only one where six models scored blind and landed effectively on top of the money.
Free: The Weekly PM Market Brief
The 8-model panel's graded record, the week's biggest market-vs-model gaps, and what's spiking next. One email, Sundays.
Unsubscribe anytime, one click. We never share your email.
What A Michigan House Contract Actually Pays
This matters more here than on a normal board, because the thing people assume these contracts track is not the thing they track.
Every Michigan district market settles on one sentence. MI-08's Republican contract reads: "If the House member sworn in for MI-08 for the term beginning in 2027 is a member of the Republican party, then the market resolves to Yes." The candidate names Kalshi prints beside each contract are labels. The contract pays on the PARTY of whoever is sworn in. If a nominee were replaced tomorrow, the contract would not void. Traders could of course reprice it, but the question it settles on would not change: the party of the member sworn in. Anyone reading the 5.9-cent bid on MI-08's Republican side as a bet on Thomas J. Smith personally is reading it wrong.
Prices are quoted in cents, and a cent is a percent: a contract that costs 63.5 cents pays $1.00 if it resolves Yes and nothing if it does not, so in payout terms it turns $100 into about $157 when it hits. New to this? Our explainer on how election prediction markets work covers how these contracts price and settle, and whether election trading is legal in Michigan covers the state rules.
One honesty note before the numbers. Across all 26 Michigan contracts, roughly 101,000 contracts have traded in the markets' lifetime, and that total is spread very unevenly. Kalshi does not publish open interest on these, so lifetime volume is all we can see, and lifetime volume only grows. Read it as "has this market ever been traded," not as "is this market liquid today." Seven of the 13 Democratic contracts carry a spread of four cents or tighter: MI-01, MI-03, MI-04, MI-07, MI-08, MI-10 and MI-11. Only three of those seven are races anybody is arguing about, so a tight spread here means a quoted book, not a close contest. At the other end, MI-02 bids a tenth of a cent and MI-09 bids seven tenths, which is to say neither has a real bid at all. Their 3-cent and 4.1-cent midpoints are displayed prices rather than quotes, and a midpoint on an empty book is not the market's considered view of anything.
These are market prices and model estimates, not predictions of fact and not financial advice. Kalshi is a CFTC-regulated event-contract exchange (18+; availability varies by state).
Tom Barrett, MI-07: The Incumbent Priced As The Underdog
Tom Barrett is a sitting Republican congressman, and Kalshi has him losing.
He flipped the Lansing-area 7th in 2024 by 3.7 points, 50.28% to 46.56%, the narrowest congressional result in Michigan that year. His seat is now the deepest Michigan market on the exchange, about 21,000 contracts traded, and the Democratic side bids 62 cents and asks 65. The Republican incumbent trades at 34 to 38.
What happened in between is a fight about data centers, and it is worth the detail, because it is the reason this district stopped behaving like a normal incumbent hold.
Will Lawrence, a co-founder of the Sunrise Movement, won the August 4 Democratic primary with 43.65%, beating former U.S. Ambassador to Ukraine Bridget Brink (28.79%) and Matt Maasdam (27.56%). He ran on a 12-month federal moratorium on new data-center construction. E&E News reports that 12 data centers are operating or proposed inside MI-07 alone, and that Lawrence took the clearest position of the three, while Brink and Maasdam backed narrower transparency and environmental standards and letting communities decide locally. He won that three-way race with a plurality, not a majority.
Barrett has answered with an ad of his own:
"While politicians are cozying up with Big Tech billionaires to force data centers on our communities, I stood up to my own party and voted against giving your tax dollars to data centers." — Tom Barrett campaign ad, 2026
The American Prospect reported on August 20 that the vote the ad refers to is a 2015 state-level vote on a single Grand Rapids internet data center, and laid out the rest of the record beside it. Barrett voted for the One Big Beautiful Bill Act, which the Prospect puts at $83 billion in tax breaks to AI companies. He backed the December 2025 SPEED Act limiting federal environmental review for data centers. And he took money from a Musk-aligned super PAC reported at more than $869,000, maxed contributions from Peter Thiel and Blackstone chief executive Stephen Schwarzman, and corporate PAC checks from Google, Amazon, Microsoft and Oracle.
Lawrence carries a record of his own that the 7th District has already voted on once. On a 2024 podcast episode about foreign policy, he described older Black leaders' backing of American power this way:
"Establishment, capitalist, imperialist American power... defangs the white left and puts us in impossible positions." — William Lawrence, 2024 podcast appearance
In July the Rolling Sea Action Fund, a group affiliated with the Congressional Black Caucus, ran a radio ad against him in the district:
"The same leaders, Black leaders, that have fought and bled for this country and this party are not a big problem." — Rolling Sea Action Fund radio ad, July 2026
The same ad asserted that Lawrence had been part of a movement that helped Trump win Michigan. Lawrence said his words had been taken out of context, that he had been talking about organizing Black Americans against the war in Gaza, and that his own phrasing had been clumsy.
The same Prospect report from July 17 documented Black officials in Michigan pushing back the other way, among them Detroit city councilmember Denzel McCampbell, who called it "quite disappointing that the Black community is being used as pawns."
Here is the part that matters for the price. The ad was running by July 17. The primary was August 4. Lawrence won it by roughly 15 points. A Democratic primary electorate heard the attack and nominated him anyway. A November electorate in a district Trump carried by a point in 2024, 50% to 49%, has not been asked the question yet, and that gap between the two electorates is the single clearest reason a contract can read 63.5 cents while a model that weighs the same record lands at 52.
That is the record the two campaigns are now arguing over. Our panel would not follow the price all the way.
Six models scored MI-07 without seeing a cent of it. The median came back at 52% for the Democrat against a 63.5-cent market. Their objection was mostly about dates. Several of the "Lean Democratic flip" ratings on this seat were published in May and July, before the August primary produced Lawrence, and the only two general-election polls that exist here, showing Lawrence up 49-47 on August 13 and 48-46 on June 30, were both paid for by Democrats. A campaign releases the internal poll it likes. It does not release the other one.
Christina Hines And Michael Bouchard Jr., MI-10: The Seat John James Left
John James won MI-10 in 2024 by 6.1 points and is not defending it; he is running for governor instead, a race where Kalshi currently has him at 10 to 13 cents against Jocelyn Benson. His Macomb County seat is open, and open is the whole story.
Christina Hines, a former assistant prosecutor, won the Democratic primary with 47.01% over Eric Chung (29.67%) and former Pontiac mayor Tim Greimel (23.32%). Michael Bouchard Jr., an Army veteran and the son of the Oakland County sheriff, took the Republican primary with 72.21%. Kalshi's Democratic contract bids 64 and asks 68. Our panel said 57.5%.
The models' case for the price sitting above their own read is the same one they made on MI-07, plus one structural point they raised themselves: the 6.1-point Republican margin in 2024 included John James himself, and an incumbent's personal vote does not transfer to a first-time nominee who has never appeared on a ballot here. That argument cuts toward the Democrat. Our models priced it, and still landed eight and a half points under the money.
Bill Huizenga, MI-04: An 11-Point Win, Priced At Even Money
Ask which Michigan seats are in play and most people will say the 7th, the 8th and the 10th. Cook's own Michigan analysis names a different three: the 4th, the 7th and the 10th, with the 8th as the Republican target rather than a Democratic one. The market agrees with Cook.
MI-04 is Bill Huizenga's Kalamazoo and Holland seat. He won it by 11.7 points in 2024, 55.07% to 43.36%, which is not a swing-seat number. It is also the only Michigan contract priced anywhere near even money: the Democratic side, state Senator Sean McCann, bids 43 and asks 47. The handicappers are openly split, with Cook, Inside Elections, Sabato and Decision Desk HQ all at Lean Republican, and Silver Bulletin on August 15 and Race to the WH back on May 22 calling it a toss-up.
Our panel took the Lean Republican side hard: a median of 33.5% for the Democrat against a 45-cent market, the joint-widest gap on the board at 11.5 points. Their reasoning was arithmetic rather than atmospheric, and it is worth doing on the page.
Republicans won the national House popular vote in 2024 by 2.6 points, 49.8% to 47.2%, per the Clerk of the House's certified election statistics. Six public aggregators now put Democrats ahead on the generic ballot by about 6.2 points as of August 27, 2026. Going from R+2.6 to D+6.2 is a swing of roughly 8.8 points of margin. Take 8.8 off Huizenga's certified 11.7-point win and he is still ahead by about three. To get McCann to even money you need a swing meaningfully larger than the national numbers describe, or a district-specific collapse that no public evidence documents.
The three general-election polls that exist in MI-04 do not settle it, and none of them is neutral. Every one was commissioned by a campaign or a party-aligned group, and a campaign releases the internal poll it likes.
| Poll | Field dates | Sample | Huizenga (R) | McCann (D) |
|---|---|---|---|---|
| GBAO, For McCann's Campaign | May 28 - June 1, 2026 | 500 LV, ±4.4 | 45% | 48% |
| Ragnar Research, For Conservatives For America | March 12-14, 2026 | 400 LV, ±5.0 | 48% | 42% |
| Public Policy Polling (D) | Nov 20-21, 2025 | 559 RV, ±4.1 | 44% | 42% |
The newest of the three is nearly three months old, and the two most recent disagree about who is ahead. That is the entire public polling record for a district priced near even money.
The Full Michigan Board
The AI blend is the median of the six models that returned a board, after a revision round in which each one read the others' reasoning without ever seeing a price. Where a median falls between two models it is shown to one decimal. The Kalshi column is the midpoint of the live bid and ask for the DEMOCRATIC contract in each district, fetched August 28, 2026.
| District | 2024 certified margin | Kalshi (D) | AI blend (D) |
|---|---|---|---|
| MI-07 Lawrence Vs Barrett | R+3.7 | 63.5¢ | 52% |
| MI-10 Hines Vs Bouchard | R+6.1 (open) | 66¢ | 57.5% |
| MI-04 McCann Vs Huizenga | R+11.7 | 45¢ | 33.5% |
| MI-08 McDonald Rivet Vs Smith | D+6.6 | 92.3¢ | 93% |
| MI-03 Scholten Vs DeBoer | D+9.9 | 93¢ | 95% |
| MI-11 Moss Vs Baker | D+18.6 (open) | 94.8¢ | 95.5% |
| MI-12 Tlaib Vs Hooper | D+44.3 | 97¢ | 99% |
| MI-13 McKinney Vs Nykoriak | D+44.1 | 97.2¢ | 98% |
| MI-06 Dingell Vs Smiley | D+27.0 | 96.8¢ | 98% |
| MI-01 Barr Vs Bergman | R+21.2 | 8.6¢ | 2.5% |
| MI-05 Vukasovich Vs Walberg | R+32.9 | 7.1¢ | 1% |
| MI-09 Pooley Vs McClain | R+37.3 | 4.1¢ | 1% |
| MI-02 Ambrose Vs Moolenaar | R+33.5 | 3¢ | 1% |
Read the bottom three rows with the spread in mind rather than the midpoint. MI-05's Democratic contract bids 3.2 and asks 11, MI-09 bids 0.7 and asks 7.5, MI-02 bids 0.1 and asks 5.8. On a book that wide the midpoint is an artifact of where the offers happen to sit, and the panel's 1% is closer to what the certified 2024 margins describe than MI-05's printed 7.1 cents is. MI-01 is the exception in that group: it is quoted 7.1 bid, 10 ask, which is a real two-sided book on a seat Jack Bergman won by 21.
Two rows in the safe column carry a story anyway. MI-13's Democratic contract now says Donavan McKinney, because the incumbent, Shri Thanedar, lost his own primary on August 4, 51.70% to 47.91%. MI-11 is open because Haley Stevens is running for the Senate. Neither changes which party gets sworn in, which is the only thing these contracts ask, and the market and the panel agree on both.
| Model | MI-07 | MI-10 | MI-04 | MI-08 |
|---|---|---|---|---|
| Fable | 55% | 57% | 33% | 94% |
| Opus | 50% | 58% | 34% | 93% |
| Sonnet | 51% | 57% | 33% | 91% |
| Gemini | 48% | 57% | 33% | 93% |
| Grok | 53% | 59% | 34% | 92% |
| Kimi | 53% | 58% | 35% | 93% |
| Panel Median | 52% | 57.5% | 33.5% | 93% |
| Kalshi Midpoint | 63.5¢ | 66¢ | 45¢ | 92.3¢ |
Every seat on this panel is graded against real market settlements — records to date: Gemini 86% on 5,151 graded calls · Grok 84% on 4,427 graded calls · Kimi 82% on 2,576 graded calls. Recomputed daily; the full scoreboard is public.
These are model estimates, not predictions of fact and not financial advice. Every number in this piece gets graded in public once the market settles, on the full graded scoreboard.
Six of the eight models on the panel scored this board. Of the two that did not, one declined to price an election at all.
More live boards from the same panel: Michigan Senate, where Abdul El-Sayed's contract bids 61 cents against Mike Rogers at 39 · Michigan governor, where Jocelyn Benson bids 88 and John James bids 10 · House control, the chamber-level board these 13 seats feed. Prices fetched August 28, 2026.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $197M traded
- 2027 Pro Football Champion · $73M traded
- 2028 U.S. Presidential Election winner? · $63M traded
- Pro Baseball Champion · $62M traded
- 2028 Republican presidential nominee · $60M traded
- TOUR Championship Winner · $43M traded
Every market above links to our full AI model verdict; browse them all on the OddsShopper prediction markets hub, and see how every settled call actually scored on the full graded scoreboard.
Where The Panel Changed Its Mind
The second round is where this board earned its keep, because two models read the same primary result and moved in opposite directions.
Gemini cut its Democratic number in MI-07 by eight points:
"I moved MI-07 down from 56% to 48%, persuaded by Forecasters C and E that the pre-primary handicapper ratings failed to price in the Democratic nominee's 43% fractured primary win and Sunrise Movement baggage against a first-term incumbent." — Gemini
Opus went the other way on the same district, and said exactly why it would not go further:
"Forecaster A and B made me re-read the Sunrise/data-center fact as two-sided rather than one-sided. A 12-month moratorium is a local-populist message in a district with 12 data centers operating or proposed, and the Aug 13 Democratic internal, discounted for sponsorship, reads as a tie rather than as Lawrence collapsing after a 43.65% plurality win. I still won't go to 0.53-0.58 with the peers, because the only post-primary rating on the card is Race to the WH's Aug 20 Tossup and every Lean-D-flip rating predates the primary that produced this nominee." — Opus
[Editor's note: Opus is describing the round-one peer range it was shown, when one model had MI-07 at 58%. After the revision round the highest number in the table below is 55%.]
Kimi refused to move at all, and its reason is the sharpest line on the board about MI-04:
"I held MI-04 at 35% because the observation that no non-partisan poll exists to adjudicate the structural-vs-tossup conflict cuts both directions, not toward a lower number." — Kimi
Five of the six models revised something. The one district where nobody moved more than a point in either direction was MI-08.
What Would Change The Panel's Mind
Four things, each of them checkable, each with the direction it pushes.
A non-partisan poll in MI-04 or MI-10. Five district general-election polls exist in Michigan this cycle, two in MI-07 and three in MI-04, and every one was commissioned or released by an interested party: a campaign, a party-aligned group, or a firm that works for one side. One independent survey in MI-04 would be worth more than the three partisan ones already there, and MI-10 currently has no public district poll at all, so the first one published is the whole evidentiary base. Either would move our models in whichever direction it lands, because right now they are working from the certified history and the national numbers alone.
A confirmed change to the MI-08 Republican line on the ballot. The August 24 decline deadline has passed, and the panel treated the resulting uncertainty as a floor under the Democrat rather than a source of variance. A documented replacement nominee with a funded campaign would push the Republican contract up from its current single digits; confirmation that the ballot is set as it stands pushes it further down.
The next Cook or Sabato move on MI-07. Cook has had this seat at Toss Up since May 13, which is before the primary that produced Lawrence. The first rating published after the nomination is the first genuinely post-primary read anyone gets, and it moves the panel in whichever direction it goes.
Third-quarter fundraising, due in mid-October. The newest campaign-finance numbers in these races are the July 15 FEC reports. A first-time nominee in an open Macomb County seat either raises like a battleground candidate in the third quarter or does not, and that is a fact, not a forecast.
When This Settles
| Election Day | Tuesday, November 3, 2026 |
| What Settles It | The party of the House member sworn in for the term beginning in 2027 |
| Expected Contract Expiration | January 4, 2027 (Kalshi's published expected expiration) |
| Early Settlement | Kalshi may settle early once media organizations project a winner |
| Next Known Catalysts | Q3 FEC filings mid-October; any post-primary Cook or Sabato rating change |
| Refresh Policy | This page is re-scored when the story moves, and the as-of stamp changes with it. Prices and panel numbers here are as of August 28, 2026. |
- NL Manager of the Year Odds: Mattingly Surges As Kalshi Reshuffles
- Why Abdul El-Sayed (D) Will Win the Michigan Senate Race
- Why David Brock Smith (R) Will Win the Oregon Senate Race
- Why Ed Markey (D) Will Win the MA Senate Primary
- Why James Talarico (D) Will Win the Texas Senate Race
The Bottom Line
Michigan has 13 House seats and the market is only arguing about three of them, which is roughly the right number to be arguing about. Where the market and the panel part company is how far the national environment has already carried those three.
Our models are not calling the market wrong on direction. They have the Democrat ahead in MI-07 and MI-10 too, and they agree the 8th is close to over. Their argument is that the market has priced a six-point national lead into three individual Michigan districts more aggressively than the certified results underneath them support, and that the district-level evidence underneath that pricing is thin to the point of absence: two partisan internals in MI-07, three in MI-04, and in MI-10 not one public poll at all.
MI-04 and MI-07 are tied for the widest split between our models and the market, at 11.5 points each. The difference between them is the history underneath: MI-07 was decided by 3.7 points in 2024 and MI-04 by 11.7. Whichever number turns out to be the better description of Michigan's 4th District in November, the market and our models are not currently describing the same race.
Hero illustration: OddsShopper, in the house collage style. Tom Barrett photo by Brian Thorpe, licensed Public domain; photos cropped, toned, and composited.
