when the models are hitting

individual level generic ballot support is heavily influenced by which party you view more favorably and some nuance is lost not breaking out strength of fav.

these are from over 10k responses in NY 2022 and 2024. when i ran it i was shocked at the consistency

a bit more breakdown tracking age by borough and ev day (still only first 7 days)

interesting age chart for early voting in nyc among dems (first 7 days only though) in raw numbers 29 and 30 had the highest numbers

i was testing out the census's field of bachelor's degree for first major on a small scale (upstate ny senate district) at the election district level and got some similar results (science and engineering + related fields are science)

i think your defection rate on this is too low, based on polling i have from ny last year (outside of nyc) defection for dems and reps who didn't vote in the 2022 gov primary was 15%, i would expect it even higher among non voters