Thread by @christinezhang

NEW: #Election2020 turnout is the highest in over a century, but who's gotten the most votes, and where? We looked at precinct & county data in key areas to find out: [THREAD]

By @jburnmurdoch & me — our second @FT collab in two weeks! ⚡️

ft.com/content/31a027…

1/ Much of the story's in the suburbs, which Biden "won back" to a certain extent. Meanwhile, a red rural wave has countered the suburban swing.

Result: an increasingly polarized US.

2/ More detail here: taking battlegrounds MI, PA, WI & GA as a whole, there's a clear & pronounced net vote increase in the suburbs for Biden in 2020 vs Clinton 2016

3/ Drilling down, let's look at Milwaukee: ~3.2m voters cast ballots in Wisconsin this year, a record high (raw & %terms). But in the state's biggest city, turnout stayed flat. It was the inner suburbs that delivered the "blue wave"

DATA CREDIT: @jdjmke johndjohnson.info/post/2020-milw…

4/ It's a similar story in Philadelphia, which (as of writing) has cast 724,898 total prez ballots in 2020, only ~2% higher than the 709,681 cast in 2016.

In both Milwaukee City & Philly, Biden underperformed Clinton in majority-Black and majority-Latino areas:

5/ Keep in mind these cities are still overwhelmingly Democratic, and majority- non-white neighborhoods are still where Biden did the best. When we say "underperformed" we mean swing *relative* to 2016

6/ MAJOR HT to @sixtysixwards, btw, for his EXTREMELY USEFUL #rstats tutorial on how to reconcile changing city boundaries over time (more on that later when I discuss our methods section!) sixtysixwards.com/home/crosswalk…

7/ In Atlanta, things are a bit different: number of ballots cast rose across the board. Majority-Black precincts swung *very* narrowly toward Trump in % terms, but still backed Biden 93 to 6

Result: majority-Black Atlanta precincts added ~15K more votes to Biden

8/ Maricopa County, AZ is another example of how % swing is different than vote swing:

Phoenix swung *slightly* toward Trump, but it still added more votes to Ds than Rs

As with a lot of election narratives, **MORE THAN ONE THING CAN BE TRUE AT THE SAME TIME**

9/ Orange County, CA, where Republicans won back 2 House seats, is a story in its own right:

The county was Dem overall (54-44) for the second prez election since FDR ('16 was the 1st: latimes.com/projects/la-po…), but majority-Asian areas swung to Trump

ocregister.com/2020/11/13/how…

10/ To conclude:

> Suburban turnout+swung Dem, while rural areas swung more Republican

> Race has played a nuanced role: different groups in different places vote & turn out differently

> Non-white areas were key to the outcomes *for both parties* in different areas

11/ All right, if you've stayed with me this long, time for some ~~ NERDY PARTS ~~ and credits:

It's impt to recognize we're not going to have a public dataset of turnout estimates by race until the Census Bureau updates its Current Population Survey

census.gov/topics/public-…

12/ This means that our analysis is a *rough* look at turnout & voting patterns in select places where we were able to obtain data. Just looking at votes in majority-minority vs majority-white areas is going to miss some key dynamics

https://twitter.com/christinezhang/status/1326955237983334400

13/ For example, the closure of colleges & universities could have depressed turnout in Philly and other cities

https://twitter.com/sixtysixwards/status/1325484375987924992

14/ Also, a D swing in majority-white suburbs is consistent w/our previous finding that Biden gained w/college-ed white voters. BUT the suburbs themselves are getting more diverse, also accounting for the shift

**2 things can be true at the same time**

https://twitter.com/christinezhang/status/1325104495156736000

15/ In the same vein, in majority-non-white areas that swung slightly Rep, it could have been white voters *in those areas* that accounted for the swing. We don't have the data/analysis atm to really disentangle these effects

16/ Speaking of data, here are some of the elements of data wrangling that went into this:
a) finding 2020 and 2016 precinct results
b) merging 2020 with 2016
c) merging demographic data to 2020 precincts

17/ HT @gerrymandr & @openelex who have done *a lot* of work putting together historical precinct results and boundary files, which helped us with (a). (b) doesn't sound so hard until you realize that precinct boundaries have changed since 2016 😬🙃 ...

https://twitter.com/christinezhang/status/1326570592821260293

18/ Major props to the Wisconsin Legislative Technology Services Bureau who has disaggregated historical data into 2020 ward boundaries!!! Amazing!! HT @jdjmke for pointing this out, and for providing #rstats code to merge the data

data-ltsb.opendata.arcgis.com/datasets/2012-…

19/ But not all places do this, so we really appreciate @jtannen215 at Philadelphia elections blog @sixtysixwards providing a roadmap (in #rstats!) on how to do this, which we applied to other places as well:

sixtysixwards.com/home/crosswalk…

20/ The Appendix in this @sixtysixwards post is also instructive for how to achieve (c), b/c only in ur dreams do precinct boundaries line up with census tracts 🙃

sixtysixwards.com/home/the-turno…

21/ A nerdy note on rural/urban: there's lots of ways to classify counties cityobservatory.org/what-is-urban; we used cdc.gov/nchs/data_acce…. That's why you'll see different numbers in different stories about "suburban" swing, though the story is largely the same

brookings.edu/research/biden…

22/ All this and more is in the STORY by @jburnmurdoch & me and edited by the amazing @AdrienneKlasa

Check it out for *even more* maps, charts & analysis
🗺️📊🤓

ft.com/content/31a027…