You have to control for the number of seats the party held going into the election. The biggest reason Dems lost so many more seats in 2010 than in 2022 is that they had so many more to lose! If you just made the Y axis seats won rather than seat change, that wouldn't matter.
That being said, Eisenhower in 1958 and Obama in 2010 "under-performed" their approval rating in the House by more than Biden "over-performed" last year, but it's pretty close (and more likely to do with the simple model)
I realize this is only looking at approval rating (and the House), but, using this very simple model, this is the largest positive deviation from expectation for the presidents party in the House since WW2.
New analysis from the data firm Catalist offers several reasons for the Democrats' surprisingly decent 2022.
One worth elevating: the generational shift in the electorate.
This is obviously more reliable in presidential elections, but recent midterms have been centralized enough such that we were able to get away it. This particular midterm raises the question whether our model would have performed better with a "tossup" fixed effect.
This decentralization raises interesting questions around election night modeling (like the needle or our fuzzy bars). At least in our case, the model depends on patterns holding across the country since what is happening in one state informs predictions in other states.
As our understanding of what happened in 22 takes shape, this is another piece of analysis that underscores
1) This was a very decentralized midterm
2) Turnout in competitive races looked more like 20/18
3) Small slippage for Dems amg Black voters. Not huge, but interesting twitter.com/jon_m_rob/stat…
Interested in what _really_ happened in the 2022 election? Look no further, our team at @Catalist_US’s take combining our voter file, demographic modeling, large volumes of survey data, and precinct/county/CD election results is live!
https://catalist.us/whathappened2022/…
With 62% of precincts reporting (not necessarily 62% of the final vote!), Donna Deegan (D) leads Daniel Davis (R) 54–46% for Jacksonville mayor. Jax is currently the largest US city with a GOP mayor. https://enr.electionsfl.org/DUV/Summary/3385/…
newsletter, I took a look at the party breakdown of mail ballots in Tuesday's primary election in Pennsylvania, based on state data:
https://datawrapper.de/_/CzW9r/
New: Compton, a city known for its Black culture, has flipped majority Latino — but its leadership remains all Black. My piece on this complicated city and the group of Latino vaqueros who are pushing for change:
Initial results have always tilted towards the AKP. This has to do with how the count is done. In past elections, we’ve seen Erdogan start at around %60-65. Who knows what will happen but let’s wait for a higher % of the count for the educated analysis and hot cakes.
As we expected: the first ballots open show a huge Erdogan lead. This is the usual; as I stated earlier this will change over the night. It is going to be a long night but we all need to be patient.
Next there is an analyze of response rates among Republicans by vote choice. Republican who voted for Biden where more likely to respond than Republicans that voted for Trump. #AAPOR
How Democratic are young voters in swing states (at least by party registration)?
NC: D+11
PA, FL: D+17
AZ: D+22
NH: D+25
NV: D+28
For @SplitTicket_, our most detailed piece yet on a *potential* generational cliff.
https://split-ticket.org/2023/05/08/a-generational-cliff/…
This is a massive issue in data professions and it seriously stymies creativity, quality, and knowledge generation. The prevailing pedagogy of data science convinces folks to start with some data and work up, rather than starting with a vision and working down. twitter.com/rlmcelreath/st…
Updating earlier analysis for @SkyNews, Labour is still seeing larger gains in areas with fewer university graduates but the Conservative vote is down more in areas with more graduates.
one of my favorite facts: since the president of France is a co-prince of Andorra, Andorra is the only country in the world with an elected monarch who is democratically elected by the population of a different country.
Beyond King Charles: Your guide to the world’s 28 other monarchs
The world of monarchies is fascinating.
https://washingtonpost.com/world/interactive/2023/monarchies-around-world/?tid=ss_tw…
The electorate overall had the highest share of non-Hispanic Whites since 2014. Midterm elections tend to see a whiter electorate, although 2018 was a bit of an aberration. I fully expect to see a sizable drop in the White share of the electorate in 2024 compared to 2022
added some more findings to the story: Turnout sagged among college-educated voters (...) though it was particularly large among Black college graduates, dropping from 73 percent in 2018 to 60 percent in 2022 (incl. 20 point drop for Black voters w/ college degree)
The 2022 midterm had the biggest racial gap in turnout in any national election since at least 2000 according to the census
w/ @sfcpollhttps://washingtonpost.com/politics/2023/05/02/black-voter-turnout-election-2022/…
The 2022 midterm had the biggest racial gap in turnout in any national election since at least 2000 according to the census
w/ @sfcpollhttps://washingtonpost.com/politics/2023/05/02/black-voter-turnout-election-2022/…
election data friends: when adjusting for CPS over-reporting turnout, why do we adjust the weight based on the turnout rate instead of the raw number of voters?
If we do the latter we get the same turnout rate (up to rounding) and we also get the raw counts for free
Our first modeling adventure was VA House of Delegate in 2019. We had a model, but mostly used it for internal testing, the only component we showed was our turnout estimate.
hmm I'm not sure I believe this. Obviously a different kind of election model, but it took us around a year and a half before we produced the core of an election night model that we were really happy with.
The communists recent surprise result in Salzburg seems to have given them national momentum. The Austrian left was hoping to put together a coalition of social democrats, greens and liberals after the next election (like Germany). It's hard to see them do the same with the KPÖ.
8. The House wants Minnesota to hand our presidential electoral votes to whoever wins the national popular vote.
No one seems to care that future presidential elections will be decided completely by large cities on the coasts like New York and Los Angeles.