Will Abdul El-Sayed win the 2026 US Senate election in Michigan?

TL;DR

Abdul El-Sayed has a 62.3% chance of winning Michigan's 2026 U.S. Senate election. He leads the post-primary poll set and benefits from a Democratic national environment, while recent Michigan Senate polling errors and his weaker performance than the Democratic governor nominee keep the race close (270toWin, Silver Bulletin, Inside Elections). My median forecast is El-Sayed +1.4 percentage points; a normal-sized polling miss elects Rogers.

Context

At the September 1, 2026, 13:20 UTC cutoff, the matchup is settled. Michigan's certified August 4 primary gives El-Sayed 744,952 votes, or 48.45%, to Haley Stevens's 731,160, or 47.55%, a 13,792-vote margin; Rogers was unopposed and received 883,328 votes (Michigan Department of State).

Six public polls were fully fielded after the primary. Five show El-Sayed ahead; the Fox News poll shows Rogers ahead (poll archive). The simple average is El-Sayed +2.8 points, while my recency-, sample-, screen-, and sponsor-adjusted average is +2.4 points. That is a lead, not control of the race.

Evidence

The historical backbone favors Democrats, but less than the streak suggests. The table is the full Michigan U.S. Senate election history in the state's online archive; there was no Senate election in 1998.

Democrats won all nine, but the last three margins fell from 6.51 to 1.68 to 0.34 points. The closest analogue is 2024: Rogers lost by 19,006 votes while Trump carried Michigan by 80,103 votes, or 1.41 points. Rogers therefore ran 1.75 margin points behind Trump in the same electorate (Michigan Senate and presidential results).

The post-primary polling is broad enough to matter and thin enough to fail together.

The polls disagree by 11 points from best to worst. Fox is the strongest Rogers result, but its registered-voter topline left only 2% undecided and El-Sayed led by 4 points among respondents certain to vote. EPIC-MRA is the freshest result. It moved from Rogers +3 before the primary to El-Sayed +4 afterward (same-pollster series), and found El-Sayed at 39% favorable and 38% unfavorable versus Rogers at 29% favorable and 40% unfavorable (Fox methodology and crosstabs, EPIC-MRA report).

Michigan's recent Senate polling error is the main reason not to convert D+2.4 into a 70%-plus forecast. The final RCP average overstated the Democratic margin by 1.8 points in 2018 and 3.7 points in 2020; the final 538 average overstated it by 3.3 points in 2024 (2018 RCP, 2020 RCP, 2024 aggregation). The sample is only three elections, and error is not stable: the 2022 governor RCP average had Whitmer ahead by 1.0 point, while she won by 10.53 points (2022 RCP, official result). I therefore subtract 1.2 points, not the full 2.9-point recent Senate average.

The fundamentals still favor El-Sayed. Silver Bulletin's August 31 voter series put Trump at 38.3% approve and 57.7% disapprove, while its generic congressional ballot was D+6.7 (Silver Bulletin archive). Inside Elections rates the race a toss-up but gives Michigan a D+4.5 baseline (Inside Elections). The three major polls that also tested governor had Jocelyn Benson ahead by 5, 5, and 4 points, while El-Sayed ran 9, 4, and 0 points behind her in the same samples (Fox, AARP, EPIC-MRA). I read that as a real candidate penalty, but one that has narrowed since the primary.

El-Sayed's coalition has a turnout risk. AARP found Rogers leading 52% to 43% among voters 50 and older, while El-Sayed led voters under 50 by 14 points; 86% of the older group said they were extremely motivated, versus 71% of younger voters (AARP research). Yet this was already a likely-voter poll and still had El-Sayed ahead 48% to 47%. The age split should widen uncertainty, not be charged twice as another full margin penalty.

Money favors Rogers, but not by enough to override the polling. FEC reports covering January 1, 2025, through July 15, 2026, show Rogers with $4.47 million cash on hand against El-Sayed's $2.55 million (Rogers FEC, El-Sayed FEC). From August 5 through September 1, reported independent expenditures totaled about $10.0 million helping Rogers and $3.8 million helping El-Sayed, counting support for one candidate and opposition to the other (FEC spending against El-Sayed, FEC spending against Rogers). These reports include estimates and can later be amended. El-Sayed also won the primary after his side was outspent nearly 10 to 1 on broadcast, cable, and radio, so another advertising imbalance is not automatically decisive (Axios).

The Temple Israel and Hasan Piker disputes are real liabilities. El-Sayed apologized to Jewish Democrats on August 29, while declining to make a clean break with Piker; Republicans kept the issue alive through August 31 (Associated Press, Axios). The newest poll ended on August 28, so there is no clean measurement after the apology and the latest attacks. Rogers's request that AIPAC avoid visible general-election advertising cuts both ways: it removes a potential pro-Rogers spender, but also removes the foil that helped El-Sayed turn outside spending into an anti-establishment message (Axios).

My model uses percentage-point Democratic margins:

$$\mu = 0.70(2.4 - 1.2) + 0.30(1.9) = 1.41$$

The 2.4 is the adjusted poll average. The 1.2 is a shrunken Michigan Senate polling-error correction. The 1.9 is the fundamentals estimate after applying El-Sayed's candidate penalty to Michigan's Democratic baseline. I use a 4.4-point standard deviation for correlated polling error, turnout error, undecided allocation, 63 days of campaign movement, and shocks. That gives $\Phi(1.41/4.4)=62.6\%$. A 0.3-point candidate-continuity adjustment for withdrawal, death, disqualification, or replacement yields the final 62.3%.

I checked external prices only after making that estimate. PredictIt's midpoint prices de-vig to about 60.0% Democratic, while Polymarket was near 61.0% with roughly $588,000 traded; two small Manifold markets were around 64.0% (PredictIt, Polymarket). Public models are scattered: 535 reported 69.0%, while Call the Map reported 51.0%; Cook and Inside Elections classify the race as a toss-up (535, Call the Map, Cook, Inside Elections). The market check sits slightly below my estimate and does not supply a new fact that warrants a larger move.

What's non-obvious

A naive forecast above 70% is too high. It treats the current poll lead as unbiased even though the last three Michigan Senate polling averages all leaned too Democratic (2018, 2020, 2024), and it treats the D+6.7 national environment as an extra boost even though much of that environment is already inside the state polls (Silver Bulletin). The opposite move, subtracting the full 2.9-point historical error, is also too aggressive. Three elections do not establish a fixed house effect, and Michigan's 2022 governor miss ran sharply the other way (RCP).

The demographic and advertising warnings are also easy to double count. AARP's older-voter gap was inside a likely-voter topline (AARP), and EPIC-MRA fielded after weeks of Republican attacks and still found El-Sayed ahead by four (EPIC-MRA, FEC notices). Those signals justify a wide error bar. They do not justify subtracting the same risk again from the mean.

Uncertainties

• No high-quality poll covers the full period after El-Sayed's August 29 apology and the August 31 Republican escalation. Two or three new likely-voter polls would close the largest gap (AP, Axios).
• The surveys leave 5% to 15% outside the two major candidates. Their final allocation, and whether El-Sayed can reach 49% before October, matters more than small movements within the current average (270toWin).
• Candidate finance reports stop on July 15. A new full filing would show whether El-Sayed's primary win produced enough fundraising to narrow Rogers's cash and advertising edge (FEC).
• The historical polling correction rests on three recent Senate elections. Better point-in-time state-race error data, separated by office, poll screen, and days to election, would tighten the 4.4-point uncertainty estimate.