CROSSTABS: CLICK HERE
The generational split is the whole race, and an older August electorate is the only thing keeping it close
The Public Sentiment Institute · Fielded July 25 to 29, 2026 · Election August 4
Abdul El-Sayed enters the final weekend of the Michigan Democratic Senate primary leading Rep. Haley Stevens 54.8% to 45.2% in our projected result. Before undecided voters are allocated, he leads 47.6% to 39.5%, with 12.9% still undecided.
Our survey gives El-Sayed a 96% probability of leading the field. That is a confident finding about who is ahead. It is deliberately not a confident finding about by how much. More on that at the end.
Below is every question we asked, reported across both the registered voter universe and the screened likely voter universe. We publish all of it, including the questions that didn't produce a headline.
THE BALLOT TEST
"If you walked into the voting booth and you had to choose, who are you voting for on the August 4 Democratic primary ballot?"
Abdul El-Sayed: 49.7% registered, 47.6% likely, 54.8% projected
Haley Stevens: 36.0% registered, 39.5% likely, 45.2% projected
Undecided: 14.3% registered, 12.9% likely
Read those three numbers left to right and you can watch the race tighten. Among all registered Democrats, El-Sayed leads by nearly fourteen. Screen for who actually shows up in an August primary and the lead falls to eight. Allocate the undecideds and it settles near ten.
That middle step is the story. The turnout screen costs El-Sayed 5.6 points of margin, and it costs him that for one reason: Michigan's August electorate is old. Voters 65 and over are 39.8% of the registered Democratic universe and 48.9% of the people we project will actually vote. They are Stevens's strongest group by a wide margin.
Mallory McMorrow withdrew but remains printed on the ballot. We did not read her name, and we estimate her residual support near 2%. Treat the two-way numbers above with that in mind.
WHAT VOTERS THINK THEIR FRIENDS ARE DOING
"Thinking about the 5 to 10 people you are closest to, which candidate do you believe most of them are voting for?"
Abdul El-Sayed: 39.2% registered, 36.6% likely
Haley Stevens: 22.7% registered, 25.5% likely
Not sure: 38.1% registered, 37.8% likely
El-Sayed leads here too, by eleven points among likely voters, but the striking number is the 37.8% who have no idea. Nearly four in ten Michigan Democrats can't name a favorite among the people closest to them.
We did not allocate those undecideds, and we want to be explicit about why: this question measures a respondent's perception of other people. Reallocating someone's uncertainty about their friends would invent information we never collected. The number stays as measured.
LOOKING AHEAD TO 2028
"If the 2028 Democratic Primary was held today, who would you vote for?"
Among likely voters:
Pete Buttigieg: 24.2%
Undecided: 21.7%
Jon Ossoff: 13.9%
Alexandria Ocasio-Cortez: 10.7%
Gavin Newsom: 10.0%
Kamala Harris: 5.1%
JB Pritzker: 3.8%
Josh Shapiro: 3.6%
Buttigieg leads a genuinely unsettled field among Michigan Democrats, with Ossoff a clear third. Kamala Harris draws just 5.1%.
The second line is the real finding. Undecided finishes second. More than a fifth of this electorate has no 2028 preference at all, and we left every one of them there. A contest more than two years out is exactly where an undecided answer is a result rather than a gap to be filled. Pushing those voters onto a candidate list would be a fabrication dressed up as precision.
PRO-ISRAEL PAC MONEY
"If a candidate for U.S. Senate accepted campaign contributions from a political action committee supporting Israel, would that make you more or less likely to support them?"
Among likely voters:
Much less likely: 41.0%
Somewhat less likely: 14.8%
No difference or no opinion: 36.7%
Somewhat more likely: 2.7%
Much more likely: 4.8%
A 55.8% to 7.5% split against. Among Michigan Democratic primary voters, accepting pro-Israel PAC money is a net liability by roughly seven to one, and the intensity is lopsided. "Much less likely" alone outweighs the entire favorable side five times over.
The age gradient is severe. Among voters under 30, 65.2% say much less likely. Among voters 65 and over, that falls to 26.1% and the modal answer becomes "no difference."
We did not allocate the 36.7% who say it makes no difference. That is not indecision. It is a measurement of how many voters the issue simply does not move, and it belongs in the result.
WHO THIS ELECTORATE IS
"Please select the option that best describes your general political outlook."
Among likely voters:
Progressive or Socialist Democrat: 37.4%
Lean Democratic Independent: 16.2%
Mainline or Institutional Democrat: 12.1%
Moderate Independent or Centrist: 6.3%
Working-Class or Union Democrat: 5.3%
Coalition or Civil Rights Democrat: 4.5%
More than a third of this electorate self-identifies as progressive, three times the share calling themselves mainline institutional Democrats. That composition is the structural foundation of El-Sayed's lead, and it is why a race that looks close on paper has been stable for two weeks.
TURNOUT: THIS ELECTORATE IS LOCKED IN
93.2% of our likely voters say they are certain to vote and highly motivated. 61.0% expect all or nearly all of their close contacts to turn out.
And a large share have already voted. 29.2% say they have already returned a mail ballot, with another 13.4% holding one. Add early in-person voters and roughly 56% of this electorate has either already cast a ballot or has one in hand. For a race being litigated in its final weekend, that is a meaningful constraint on how much anything can still move.
HOW WE DID THIS, AND WHAT WE WON'T CLAIM
All reported figures rest on 762 SMS text-message and online interviews conducted against the Michigan voter file between July 25 and July 29. Every one of those respondents is matched to a real voter record, which means our turnout model runs on validated prior primary participation rather than asking people whether they usually vote. Those responses were then modeled across a 19,999-record voter universe.
Margin of error: plus or minus 4.0 points at 95% confidence, adjusted for design effect. Subgroup figures carry wider error than that, and the Hispanic and Asian rows should be read as directional only.
One last thing, stated plainly: this survey supports the finding that El-Sayed leads. It does not support defending the size of that lead to a decimal point. Depending on how old Tuesday's electorate actually turns out to be, our own model puts the margin anywhere from roughly five to fifteen points. Among the hardest-core primary voters, those who have taken a Democratic ballot in all four primaries since 2018, the race is effectively tied.
If August turnout collapses to that core, this is close. If turnout is merely typical, El-Sayed wins.
Full cross-tabulations and technical documentation are available on request: contact@tpsielections.com