Thanks for doing this! Looking at this vs. the 2021 election results – an imperfect comparison, I know – it looks like your model predicts a total collapse in Frey support in Wards 4 and 5, two of his better wards in 2021. If that comes to pass, he's toast. I wonder if it will.
Interesting. So many variables. Worth noting?: It sounds like many have voted Frey+Vetaw+Ilhan--heterodox--in the past.
I still think this year's race is Frey's to lose, given his familiarity; ppl's continuing skepticism over progressive change in our backyard; and limits of RCV coalitions... 1/2
...as in, with Davis and Fateh, I wonder whether their coalition will have the strong ranking discipline you noted with Nezhad->Knuth. Or whether the differences that were there in 2017 among the non-Frey candidates, will reappear. 2/2
White precincts (41% of delegates) are known data, gathered from replies to @wedge.live. Grey precincts are predicted from a regression on 2024 primary votes for Ilhan Omar, median age, median income, renter status, and racial demographics.
The 95% confidence intervals were computed by dropping a ward of data, fitting the model, and then finding the error between predicted and known data. This was used to bootstrap the distribution of city-wide errors.
Tbh the model worked better than expected! Without feeding any spatial data in, just demographic, you can clearly well-defined neighborhoods that support Frey, Fateh, and other opposition candidates! Predicted delegates look like nearby known delegates, and the uncertainty is reasonably low!
Maps of Minneapolis for each of the underlying model variables. Ilhan Omar primary vote share and renter % were the most important variables in the model.
All of the model variables were highly correlated precinct-by-precinct, as you can see below, so it would be a lot of work to do any kind of inference the specific relationships between demographic variables and mayoral candidates.
All this to say, the outputs of the model are hilariously obvious. You could get my results by scaling up the 41% known precincts. The map shows that young renters, and people of color tend to prefer progressive candidates, and old, white, home-owning neighborhoods support Frey. Shocking!
Comments
I still think this year's race is Frey's to lose, given his familiarity; ppl's continuing skepticism over progressive change in our backyard; and limits of RCV coalitions... 1/2