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

TL;DR

I estimate a 61.3% chance that Abdul El-Sayed wins. He is ahead by roughly one to two points after adjusting the polling, and the national climate favors Democrats (Michigan polls, generic ballot). Rogers can still win through an ordinary Michigan polling miss, older-voter strength, and a late anti-El-Sayed ad wave that no post-September 9 poll has measured (AARP, FEC). (fec.gov)

Context

As of September 15, 2026 at 10:55 UTC, the election is 49 days away. El-Sayed won the August 4 Democratic primary by 13,792 votes, 48.45% to 47.55%, while Rogers received 883,328 votes unopposed in the Republican primary (Michigan Department of State).

The public polling record is stale at a crucial moment. The latest identified survey was fielded September 7–9, and no public poll through the cutoff measured the Republican convention attacks or most of the new outside spending that followed (Trafalgar release, poll archive).

Evidence

The historical backbone says Michigan Senate races lean Democratic, but recent ones are often close. This is the full run of Michigan Senate general elections since the current Class 2 seat first became open in 2014.

Those four final averages overstated the Democratic margin by 1.7 points on average. The sample is only four elections, and methods changed, so I shrink the correction by about half and subtract 0.9 points rather than 1.7. The strongest single analogue is 2024: Rogers lost the Senate race by 0.34 points while Trump won Michigan by 1.41 points, so Rogers ran 1.75 points behind his party's presidential ticket (official certification and results).

I found 11 unique post-primary head-to-head polls. One result per underlying sample is used, preferring the likely-voter estimate:

The simple mean is D+2.6. My recency, sample-size, population and sponsor weighting gives D+2.1; the latest three average only D+1.0 (poll archive). Applying the 0.9-point historical correction gives a poll-based projection near D+1.2.

The fundamentals are better for El-Sayed than that adjusted polling number alone. National generic-ballot averages are around D+6.6 to D+7.2 in data through September 14 (RealClearPolling, Silver Bulletin). Michigan-specific surveys also show the administration's Canada tariffs deeply unpopular, while the economy and cost of living outrank Israel policy as Senate issues (CNN, Fox). Rogers also failed to carry the state in 2024 without any need to speculate about his electability.

El-Sayed gives much of that advantage back. Across six surveys that asked the same respondents about Senate and governor, Democratic gubernatorial nominee Jocelyn Benson ran an average 4.7 points ahead of El-Sayed, with gaps from zero to nine points (paired releases). AARP found Rogers ahead 52% to 43% among voters age 50 and older, while Trafalgar found roughly ten-point Rogers leads in both the 50–64 and 65-plus groups (AARP, Trafalgar). Fox found El-Sayed at 75% among Black voters, below Harris's reported 87% in 2024, and El-Sayed lost the Wayne County Democratic primary vote to Stevens, 133,824 to 151,037 (Fox, official primary results). These are noisy subgroup signals, but they point to the same risk: his coalition is younger and less automatic than Benson's.

The late money is a real downside risk. FEC notices dated September 1–15 contain about $11.42 million in rows opposing El-Sayed, including a $3.77 million television and connected-TV placement on September 10 and a $1.5 million placement on September 9 (FEC independent expenditures). The raw support and opposition categories cannot be added across both candidates because the same ad is often filed as supporting Rogers and opposing El-Sayed. Candidate cash data are also stale: reports through July 15 show El-Sayed with $2.55 million and Rogers with roughly $4.47 million across authorized committees (El-Sayed FEC, Rogers FEC). (fec.gov)

My final margin model is simple. The adjusted polling component is D+1.2 and gets 75% weight. A fundamentals and candidate-quality component is D+2.0 and gets 25% weight. That gives a central margin of D+1.4. I use a 4.9-point standard deviation, wider than a polling-only model because 49 campaign days remain, most candidates are below 50%, the primary polling failed badly, and the newest attack campaign is unmeasured. A normal approximation gives P(El-Sayed wins) = Φ(1.4 / 4.9) = 61.3%, with an approximate 80% margin interval from Rogers +4.9 to El-Sayed +7.7.

I formed that estimate before checking markets and public models. PredictIt then showed the Democratic contract at a 60% bid and 61% ask, Call the Map showed 60%, and Cook Political Report classified the race as a toss-up (PredictIt, Call the Map, Cook Political Report). Those checks are close enough that I make no market-driven adjustment. (predictit.org)

What's non-obvious

The Democratic primary miss should change the uncertainty more than the mean. Late primary polls overstated El-Sayed's margin by about 12 points, but statewide primary polling errors average about 13 points because turnout is smaller, preferences are softer, and likely-voter screens are harder; general elections are more predictable (Washington Post). Applying a 12-point correction to November would be a category error. Ignoring the miss would also be a mistake. I use a modest downward correction and a wide 4.9-point outcome distribution. (washingtonpost.com)

The new Republican spending is more useful as a variance signal than as a mechanical vote shift. The ads could persuade older moderates who already see El-Sayed as too liberal, but the unsupported religious and terrorism claims can also repel suburban voters or mobilize Muslim voters (Associated Press, FEC). With no post-attack poll, assigning a large one-way adjustment would pretend to know the response. (fec.gov)

Uncertainties

• No public survey through the cutoff has fieldwork after September 9, so the effect of the convention attacks and the largest new television buy is unknown (poll archive, FEC). (fec.gov)
• The most important crosstabs are thin. I found no representative post-primary vote-choice estimate for Michigan's Arab-American or Muslim electorate, and Black-voter, Jewish-voter and older-swing-voter estimates come with large subgroup error (CNN/SSRS methodology, AARP methodology).
• Candidate finance reports stop on July 15. They do not reveal current cash, and near-real-time independent-expenditure filings contain mirrored support and opposition records (FEC). (fec.gov)
• The historical polling correction rests on four elections and has not been stable from cycle to cycle. A high-quality new likely-voter poll could move this forecast several points in either direction (RCP history).