Public polling for Abdul El-Sayed in the 2026 Michigan Senate Democratic primary and Zohran Mamdani in the 2025 New York City mayoral race. Blue is Michigan. Oranges are New York, in three shades for three different measures. Every mark is one discrete survey. The curves are fitted trends rather than connectors — they pass through no particular poll and are drawn only from independent surveys. Two open diamonds are the only aggregates, and they are labeled as such.
Click a row to find it on the chart, or a point to find its row. Source tier: pollster the pollster's own release or tabs; report a news outlet reporting the figure; compilation aggregator table only. Rows marked date unconfirmed have no published field dates and are positioned approximately.
| # | Fielded | Pollster | Sponsor | Sample | Margin | Share | Und. | Measure | Tier | Notes & link |
|---|
What's counted. Forty-seven surveys — 20 Michigan, 27 New York — each verified to a source as of July 30, 2026 and tiered by what that source is. An earlier version of this chart asserted that no one polled El-Sayed between his April 2025 launch and that November; that was false, and the polls that filled the gap — a Detroit Regional Chamber survey in May 2025 and an NRSC survey that July — both showed him running second. They are on the chart now.
About the curves. Each is a locally weighted regression fitted to the independent polls only — campaign, party, and PAC surveys are plotted but excluded from the fit — with each poll weighted by the square root of its sample size. The curve passes through no particular poll and asserts no value for any particular day; it is a model of the shape, and a different smoothing width would give a slightly different one. No curve is drawn for the ranked-choice series, which is a simulation rather than a measured ballot share. Both curves are least reliable at their ends, and the Michigan curve is unstable in the final week, where a 41, a 46 and a 54 sit within eight days of each other.
What the Michigan rise is made of. The field went from four candidates to effectively two when Mallory McMorrow suspended on July 5, 2026, and the undecided share fell from 38 percent in January to about 6 percent in late July. Turn on the undecided series to see how much of the climb is the denominator closing rather than opinion moving.
Why the New York oranges are separated. First-choice share, simulated ranked-choice final round, and general-election share are three different quantities. The ranked-choice numbers sit higher by construction, because they redistribute eliminated candidates' ballots. They are given different marker shapes so they cannot be read as one trajectory, and the three certified results at the end of each are three different figures, not one.
On margins. Whiskers appear only where a margin was published. Emerson reports a Bayesian credibility interval rather than a classical margin of error, marked as such. Where a primary number is a subsample of a larger survey, the subsample margin is used and is materially wider — Mitchell's November 2025 Michigan read carries ±6.1 on 261 Democratic primary voters, not the ±3.7 of its full sample. Points with no published margin are drawn hollow-cored rather than solid, so that absent uncertainty does not render as precision.
On the partisan share of the record. Nine of the twenty Michigan surveys were sponsored by a campaign, a party committee, or an allied group, and they cluster in the final two months. The independent reads in that window are consistently tighter than the sponsored ones. Filter by eye: hollow markers are the sponsored ones.
disclaimer Polls are snapshots of sampled opinion with stated margins of error, not forecasts. Partisan and internal surveys are marked but not excluded. Certified results come from the New York City Board of Elections. The 2026 Michigan primary had not been held at the data edge, so the blue series simply stops eight days short.