NEW from me & @christinezhang:

Much was made of US exit polls showing non-white voters swinging towards Trump, but is it that simple?

We spent 10 days poring over data from thousands of precincts in battleground states to get a more robust answer

Story: ft.com/content/31a027…
1) At first glance, the precinct-level data do support the exit poll’s finding of a non-white shift towards Trump:

Majority-black, -Latino and -Asian neighbourhoods in Atlanta, Philadelphia, Arizona and California all returned higher vote shares for Trump this year vs 2016. Image
2) But there’s a problem with proportional shift analysis:

Asking e.g "did the % of Latino voters backing Trump increase?" ignores turnout, and in doing so it ignores what elections are actually decided by: numbers of votes.
3) e.g:

If black voters went 94% D vs 5% R in 2016, then 93D vs 6R in 2020, that’s a 2pt shift to Trump

But if turnout rose by 3%, the margin in *number of votes* actually goes more blue, because the ⬆️ in votes cast *among a very D demographic* offsets switching from D to R
4) That’s exactly what happened in Atlanta, except turnout actually rose by 7% in majority-black areas, so altho people focused on a small pro-Trump % shift, these neighbourhoods actually delivered a net 15,000 vote swing to Biden (who currently leads Trump in GA by 14,172 votes) Image
5) Here’s the same thing in map form:

The majority-black southern precincts of Atlanta swung slightly towards Trump, but they remained staunchly blue, and that combo of a strong pro-Dem lean with a rise in turnout meant lots of net gains in votes for Dems. Image
6) We see similar patterns in another key state; Arizona

Here, majority-Latino precincts in Phoenix shifted towards Trump by ~2.5 % pts

But turnout in these areas — which still broke 72%D to 27%R — surged by 32%, so they still added thousands more new votes for Biden than Trump Image
7) But if the proportional swing is large, or the pre-existing pro-Dem lean small, these shifts can translate into big vote swings to Trump.

That’s what happened in Orange County CA, where majority-Asian precincts swung to Trump by >30pts and delivered him a net 10,000 vote gain Image
8) However, one other thing is true of all of those maps:

Although majority-minority precincts in city centres often did shift the vote margin in Biden’s favour, the Dems made much bigger gains in majority-white suburbs both in terms of proportional swing and absolute vote swing
9) If you look at shifts in voting patterns across the US as a whole, the Dems increased their margin more in dense, large metros than in the suburbs.

But zoom in on the key battleground states that took Biden to victory and it was a suburban blue wave that made the difference. Image
10) In recent decades a huge rural-urban gap has opened up in US politics, leaving the suburbs as the key battleground. This is especially true in swing states.

This will pose challenges for both parties Image
11) The challenge for the Dems is how to keep those same suburban swing-voters on side in 2024.

Of all Biden voters, white voters were most likely to say they picked him as an anti-Trump vote. Many of these are lifelong Republicans who have said their Dem 2020 vote is a one-off Image
12) Without the anti-Trump motivation in 2024, will they revert to their Republican habits, or will they stay blue?
13) And for black, Latino & Asian voters to have shifted proportionally towards Trump in a high-turnout election suggests new non-white voters are less pro-Dem than those that have been voting for years. How do Ds combat R messaging among these groups as they join the electorate?
13) Meanwhile the Republicans are gaining ground with non-white voters (especially those without college education), but also need Trump’s white non-college base, many of whom are Trump voters more than Republican voters.
14) Some concluding points:

Demography is not destiny. If any Dems were operating on the basis that a diversifying county will naturally shift the needle in their direction, these results cast that into severe doubt
15) Terms like "black", "Latino", "Asian" etc mask huge political diversity within each of those labels.

Or as @lorellapraeli told @christinezhang, "You need to understand that [Latinos] are different in New Mexico, and we are different in Nevada, and different in Florida."
16/16 Percentage point swings are interesting for understanding the shifting sands of the electorate, but it’s critical to factor in turnout before concluding that any one group or other did or did not propel a candidate to victory.
17/16 Please also read @christinezhang’s thread, which gets into our methods & caveats

And precinct data is a jungle, so we’re indebted to US political geography heroes including @derekwillis @sixtysixwards @jdjmke @jtannen215 @Garrett_Archer @joemfox
Last but most importantly of all, a huge thanks to @AdrienneKlasa who resisted the temptation to murder me when we filed a 1500 word first draft, and then managed to extract a coherent story from our brain-dump.

Editors, they are good.

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with John Burn-Murdoch

John Burn-Murdoch Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @jburnmurdoch

15 Sep
NEW: lots of news recently on waning immunity against infection, but a study has now landed from Public Health England on how vaccines are faring against *severe disease & death*

This chart summarises key findings, but the paper is a real goldmine, so let’s dig into more detail:
First up, fresh data on protection against symptomatic infection. Key results:
• We knew protection started out lower among older groups. Now we know they also see the most waning
• Waning much more muted (if happening at all) among under-65s
• Moderna > Pfizer > AstraZeneca
Sticking with symptomatic infection, they also looked at a hot topic: the impact of the interval between first and second dose

Very short intervals (3 weeks, e.g in US) produce lower protection than longer intervals (e.g UK & Canada), though note overlapping confidence intervals
Read 23 tweets
23 Aug
NEW: in the last couple of weeks there have a *lot* of new studies out assessing vaccine efficacy, many of which have touched on the question of waning immunity.

Unsurprisingly, these have prompted a *lot* of questions.

Time for a thread to summarise what we do and don’t know:
Let’s start with last week’s Oxford paper, the most significant study to date on waning immunity to Covid.

The researchers found signs that vaccine efficacy against symptomatic infection erodes over time, and that waning may occur faster in some vaccines than others.
Unsurprisingly, this prompted a lot of questions.

Could it just be that the first people to be vaccinated were older, and perhaps more vulnerable to waning?

Nope. The study controlled for age.
Read 33 tweets
30 Jul
NEW: lots of attention on ONS Infection Survey today, but some confusion over how it should (and should not) be used to asses whether England’s fall in cases is "real"

Quick thread:

Most attention has gone on ONS “% of people testing positive” metric showing a continued rise
But "testing positive" is a lagging indicator of cases. It estimates how many *have* Covid today, not how many are *catching it* today.

Fortunately, ONS has re-introduced its incidence data (blue line), which is a much better yardstick for cases, though always 2 weeks old 😩.
So how to resolve issue of one lagging indicator, and one that’s 2 wks old?

Look to Scotland, where cases peaked 2 wks before England, so ONS indicators have had time to catch up

Turns out ONS incidence fell at exactly same time as cases 🙂. ONS positivity likewise, just lagged
Read 10 tweets
27 Jul
There’s a wild story about the women’s gymnastics at the Sydney Olympics in 2000, which I think is very relevant to what we’re hearing about Simone Biles, and the wider point of how the top level of elite sport is just as much mental as it is physical.
In the women’s all-round final in 2000, the organisers set the vault at the wrong height. Two inches too low. This was a pretty huge deal.

For competitors who have done thousands, maybe tens of thousands of vaults at a specific height, a two inch difference is night and day.
In the first round, 17 of 36 finalists fumbled the vault

One landed on her back. Clear gold-medal favourite, Russia’s Svetlana Khorkina (comfortably won qualifying) landed on her knees.

Total chaos, and nobody knew why. Athletes second-guessing themselves.
Read 13 tweets
27 Jul
Correcting an important misconception (this is my chart, but misleading commentary):
•There were thousands more cases among young men than women after ⚽️ matches, showing impact of Euros on transmission
•But not due to attending matches. It was indoor gatherings to watch games
Of course, that still means the transmission bump was driven by the football, but match attendance is only a small part of the cause. The bulk is mixing in pubs, bars, homes etc, plus some from crowded transport to and from those indoor gatherings (and matches).
Thoughts on implications:
• I would guess these watch parties happen at a much larger scale for England games at major tournaments than they do for typical club games, but we've not had pubs fully open during the season yet so that will be worth monitoring.
Read 5 tweets
25 Jul
I feel like I've seen this before somewhere 🧐🤔
Lol that they couldn't even be bothered to change a single one of the numbers.
Read 4 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!

:(