This level of detail shows incredible respect for your subscribers and professional colleagues, Elliott. Thank you. Investing in you has been worth every dime.
Focus group of one: The young people in my family either were unaware of the election or weren't passionate enough about the candidates to show up. Ironically, the last couple of primaries have had Republican "poison pill" amendments, and that drive them in to vote. This election did not. I did make them aware of the primary, BTW, and tried to get them to turn up. They will in November, so pollsters had better adjust.
My non-quantifiable impression, based on the young people in my family/social group, is that young people in Madison and Milwaukee have some awareness that showing up can move the needle in the main election, but they largely dismiss the primary as not worth the effort. Tbf, I had a preference, but I would have been fine with any of the candidates left standing. After 15 years of Walker and the gerrymander that outlived his tenure, your bar is lowered to "Not A Raging A," (a bar that Tom Tiffany defiantly does not clear).
In a chat on another Substack, someone claimed that 75% of 18-25-year-olds vote. I questioned the accuracy of that number. As it was a chat, I have no idea where the poster got it from, but my reading these past few years suggests it was an inaccurate statement.
Very cool that you have recontact data. Might be cooler to recontact the recontacts and ask how they actually voted. (Unless you already have switching data from somewhere)
My takeaway is that until polling becomes less expensive, this ‘error’ will be how it is.
And, you cannot discount the fact that people are flaky. They may actually vote or think they will and don’t for many reasons. Or they may change their mind when in the voting both.
I don't get it. State Navigate, L2A, and L2B all weighted on age, among other things, and ignored ideology and income. So why did State Navigate end up with way too many young voters after weighting? If more young voters voted in primaries in 2025 and 2026, that could increase State Navigate's weighting from 2% to 9%, but how did they get to 17%?
Well, I don't have access to the microdata, so no imputation with joint variables. In the analysis above I allocated age pro-rata. I did a sensitivity analysis too and I'll add that to the footnotes.
It is one of those frustrating oddities in trying to come up with weighting targets in primaries for Wisconsin in that you can't simply run counts by age on the file by vote history.
Considering that all of the state schools were out for the summer, seemed quite foolish to weigh so heavily on younger voters. In Madison alone, there are probably 60-70k students during the school year, but only about a third in the summer months.
For a non-nerd, a tough read. But the result is remarkable. Could the diagnostic be run in the future on other sample polls to show whether or not the weighting corrections that brought the State Navigate poll results close to the actual results? Can you later comment on responses to your analysis from other pollsters (or even other nerds)? Thanks to you and State Navigate for this discussion.
Wow, what an article! Great reporting, Elliot, and even a non-pollster like me understood what I was reading. Thank you for this, it helps explain yesterday's piece much better.
You beat me to it. To expand on this, my questions are, is there a chance that the Republican and undeclared demographics weren't sampled at all for the Democratic Primary? Do these voters vote in numbers sufficient to sway the results significantly?
Wisconsin (and Michigan) have open primaries. A voter doesn't have to go to the effort to register as a Democrat to vote in the Democratic Party primary.
Neither Republican primary was competitive. Democrats are favored in both races. So if you're a Republican or undeclared voter and you want your vote to count the most, you should vote in the Democratic Primary.
Almost 500k people voted in the GOP gubernatorial primary in Wisconsin, even though Tiffany was going to win no matter what, which is about 150k more votes than in 2018, compared to a 300k increase on the Democratic side, from about 500k to nearly 800k. Considering massive increases in Democratic turnout broadly, I don't think crossover voting was much of a factor.
This level of detail shows incredible respect for your subscribers and professional colleagues, Elliott. Thank you. Investing in you has been worth every dime.
Focus group of one: The young people in my family either were unaware of the election or weren't passionate enough about the candidates to show up. Ironically, the last couple of primaries have had Republican "poison pill" amendments, and that drive them in to vote. This election did not. I did make them aware of the primary, BTW, and tried to get them to turn up. They will in November, so pollsters had better adjust.
Good analysis.
But I just can't get past the fact that 2% of the voters were 18-29. Two Percent.
Honestly, kids, until you start getting off the couch and vote, I don't care at all what you think about politics. You're irrelevant.
My non-quantifiable impression, based on the young people in my family/social group, is that young people in Madison and Milwaukee have some awareness that showing up can move the needle in the main election, but they largely dismiss the primary as not worth the effort. Tbf, I had a preference, but I would have been fine with any of the candidates left standing. After 15 years of Walker and the gerrymander that outlived his tenure, your bar is lowered to "Not A Raging A," (a bar that Tom Tiffany defiantly does not clear).
In a chat on another Substack, someone claimed that 75% of 18-25-year-olds vote. I questioned the accuracy of that number. As it was a chat, I have no idea where the poster got it from, but my reading these past few years suggests it was an inaccurate statement.
Very cool that you have recontact data. Might be cooler to recontact the recontacts and ask how they actually voted. (Unless you already have switching data from somewhere)
My takeaway is that until polling becomes less expensive, this ‘error’ will be how it is.
And, you cannot discount the fact that people are flaky. They may actually vote or think they will and don’t for many reasons. Or they may change their mind when in the voting both.
Great analysis.
Ooooh…. Might do this…
I don't get it. State Navigate, L2A, and L2B all weighted on age, among other things, and ignored ideology and income. So why did State Navigate end up with way too many young voters after weighting? If more young voters voted in primaries in 2025 and 2026, that could increase State Navigate's weighting from 2% to 9%, but how did they get to 17%?
How did you handle the missing age data on the Wisconsin file? It is substantial and not randomly distributed. I assume some imputation or age model.
Well, I don't have access to the microdata, so no imputation with joint variables. In the analysis above I allocated age pro-rata. I did a sensitivity analysis too and I'll add that to the footnotes.
It is one of those frustrating oddities in trying to come up with weighting targets in primaries for Wisconsin in that you can't simply run counts by age on the file by vote history.
Considering that all of the state schools were out for the summer, seemed quite foolish to weigh so heavily on younger voters. In Madison alone, there are probably 60-70k students during the school year, but only about a third in the summer months.
For a non-nerd, a tough read. But the result is remarkable. Could the diagnostic be run in the future on other sample polls to show whether or not the weighting corrections that brought the State Navigate poll results close to the actual results? Can you later comment on responses to your analysis from other pollsters (or even other nerds)? Thanks to you and State Navigate for this discussion.
Absolutely!
Wow, what an article! Great reporting, Elliot, and even a non-pollster like me understood what I was reading. Thank you for this, it helps explain yesterday's piece much better.
Was any of this manipulated by Republicans voting as democrats?
You beat me to it. To expand on this, my questions are, is there a chance that the Republican and undeclared demographics weren't sampled at all for the Democratic Primary? Do these voters vote in numbers sufficient to sway the results significantly?
Wisconsin (and Michigan) have open primaries. A voter doesn't have to go to the effort to register as a Democrat to vote in the Democratic Party primary.
Neither Republican primary was competitive. Democrats are favored in both races. So if you're a Republican or undeclared voter and you want your vote to count the most, you should vote in the Democratic Primary.
Almost 500k people voted in the GOP gubernatorial primary in Wisconsin, even though Tiffany was going to win no matter what, which is about 150k more votes than in 2018, compared to a 300k increase on the Democratic side, from about 500k to nearly 800k. Considering massive increases in Democratic turnout broadly, I don't think crossover voting was much of a factor.
Joel, that is what I was thinking too, including that Elon Musk could have been greasing their palms too.