18/18 Media: When you run your own polls, insist they ask about all three ranks, and report with a proper ranked choice analysis. Anything else is misleading.

Voters: You get to rank the candidates you like the best, in order. You don't have to pick just one, and this is not a two person race.

17/18 So how is Jazz going to gain 9+ points of support? The same way he gained 8 points between August and September. He'll keep talking and listening to people. He'll spread his vision online, knock doors, and meet at houses, apartments, condos, and neighborhood coffee shops.

16/18 Methodology matters! Our poll uses rigorous sampling and factors in Minneapolis' ranked-choice system for transparent, accurate data.

Don't blindly trust us. Also: don't trust any polls that only use first choices, assume 100% turnout, skip elimination analysis, or hide voter models.

15/18 This looks way different from a first-choice-only poll assuming 100% turnout.

Reporting on first-choice only polls is misleading. Don't do it.

A real ranked-choice analysis shows:

Frey leads, Fateh's challenger spot is shaky, and Hampton is 5-10 points from round 2.

14/18 So what's the right model? The honest answer is nobody knows, and it's an educated guess at best. We discuss this at length in the whitepaper.

13/18 No, Jazz isn't winning yet. But it's a tight 5-10 point race to round 2. Fateh's 2nd place is fragile: In the 2017 model, he is eliminated (total support < Frey's first-round), but rules mandate at least two advance.

If Jazz gains 9 points in total support, he advances, and Fateh is out.

12/18 Let's look at the 2021 model, too. Again, the step-by-step simulation is in the whitepaper. So are the rest of the likely voter models.

Jazz needs +5 points of total support to get to round 2. If Fateh loses 7%, he does not progress.

11/18 We've got the vote-counting mechanics, voter preference data, and tons of likely voter models. Let's simulate!

For full methods and step-by-step details, check the whitepaper in our bio.

But here are the interesting results using the 2017 model.

10/18 For example, look at the first round votes again. But instead of looking at all voters, let's see what it would look like if voters turned out like they did in 2017. Pretty good for that Frey fellow!

But remember: this is only first choice votes and we use ranked choice voting in Minneapolis.

9/18 'Reflect turnout' means weighting responses to match actual voter distributions from past elections, by age and ZIP code. E.g., if 1.2% of 2021 votes came from 18-29 in 55401, we won't over-represent them.

Standard polling technique, but we analyze multiple likely voter models, not just one.

8/18 Before we get to the good stuff, let's talk voter turnout. We know it won't be 100%, but how do we pick a likely voter model?

We're not good at picking over here on the data team, so let's not bother. We made voter models to reflect turnout in the 2017, 2021, 2023, and 2024 elections.

7/18 1. Eliminate the single candidate with the fewest votes.

2. If the candidate received all the votes below them, could they surpass the candidate above them?

3. If the candidate received votes from all the times they are ranked, could they beat the current highest total votes?

6/18 When counting votes, they eliminate the lowest ranked candidate, then redistribute their 2nd and 3rd choice votes, rinse and repeat, right?

It turns out candidates are eliminated in three ways every round. The city provides a nice overview of how this works: www.facebook.com/watch/?v=942...

5/18 And look who is trending upwards! #GoJazz

Yes, we polled in August, too.

4/18 In a ranked-choice election, it's not just about first picks. We asked about voters' 2nd and 3rd choices too. And the full picture?

Way more promising: Jazz trails the incumbent by just 5 points in total support across all ranks.

3/18 Minnesota is known for high voter turnout, but the highest turnout for a Mayor's race in recent history is just over 50%. We'll talk about likely voter models in a bit.

2/18 Our numbers for first round preference match the other poll pretty closely. This might mean:

1. We ran an okay poll! Well done.

2. Our numbers reflect the demographics of all registered voters in the city. They are also likely reporting those numbers. That is: this assumes 100% voter turnout.

1/18 The August polls seemed discouraging on the surface.

So we did our own. A deeper dive that accounts for ranked choice tells a different tale: Jazz gained 8 points from August to September, putting him just 5 points of total support behind the incumbent.

Thread ahead... let's break it down. 📈