Lakshya Jain Profile picture
Nov 7 4 tweets 3 min read
Our final Senate forecast for @SplitTicket_. 49D-50R, with a runoff to decide it all in Georgia. If we had to pick a party to win the majority, though, it's the Republicans.

#PASen: Lean D -> Lean R
#NHSen: Likely D -> Lean D
#NVSen: Tossup -> Lean D

split-ticket.org/2022/11/07/our…
@SplitTicket_ Here's a table showing our changes from our last update. We've said everything we wanted to say and everything we think we needed to say in the article itself. We'll have one more piece out tomorrow on the overall environment. Thanks for reading, folks. Always a pleasure.
I actually had PA and NV *both* going R before @RalstonReports' update. But once we saw the data he put up and then saw his prediction for CCM, we pivoted. There's a good data argument for it, and we wouldn't bet against the guy who got Harry Reid and Dean Heller's victories down
@RalstonReports If you want to see the data argument for NV going blue, you can just read the piece; if we felt that strongly about Nevada going Republican, we would have kept it there, but we didn't. Once the Clark mail update dropped, it became clear that you could make an easy case for CCM.

• • •

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

Keep Current with Lakshya Jain

Lakshya Jain 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 @lxeagle17

Nov 8
State correlations are overlooked when analyzing turnout today IMO -- if you're seeing relatively high Democratic turnout in one state but abysmal Dem turnout in another demographically similar one, there's a chance that we're just missing data. Wait before jumping to conclusions
I feel like the Nevada rollercoaster On Here should have taught us this lesson. Democrats concluded they were hosed seven different times before the late mail resulted in Jon Ralston saying CCM was slightly favored. It's possible Dems don't turn out much today. I'd just wait.
None of this is saying Dems will win today, nor is it saying they'll get blown out (both are possible, the latter maybe slightly more likely than the former based on fundamentals). It's just saying that...well, extrapolations from incomplete data can get very, very risky.
Read 5 tweets
Nov 5
There's a real danger in data science of overfitting to a very small sample set, and it's not one any of us are immune to. What I worry about is that we're doing exactly this in political data today by bending over backwards to say that the GOP can only outperform their polling.
I've said I expect Republicans to have a pretty decent night overall, so this isn't me forecasting a blue wave or even a blue-leaning year. Just that I think perhaps our uncertainty bounds in common rhetoric seem a bit too narrow, given all that's happened newyorker.com/news/the-polit…
You see this *all* the time. Some of the best-known forecasters and several of the best pollsters in America are outright dismissing their own polls or framing it as bad for Democrats when they're clearly not, *even adjusted for 2020 bias*! Why's that? I don't know, but...
Read 5 tweets
Nov 4
The main thing that has me wondering whether we're all going to look very silly in our predictions of a R-leaning environment are the post-Dobbs primaries and special elections. Never seen a party overperform in a ton of August specials only to get blown out in November.
Polling is a direct, clear indicator that there's been a big shift towards the Republicans. Our aggregator sits at a tie right now and that's almost 4 points worse than where it was in August. So, there's been a shift, quite clearly...
But like, how much, and from what baseline? But @GalenMetzger1 and @iabvek have both shown the special election shifts were driven largely by persuasion instead of just turnout, which you can see by examining precinct results. So, maybe they went from a ~D+4 or so to an R+1?
Read 5 tweets
Nov 3
Part of the reason I say this isn't a model is because I have no clue how to model this election. If y'all see the error bounds on our tracker, they are absolutely massive because nobody can agree on anything.

CCES, Yahoo: D+2
The Economist: Even
CBS: R+2
Quinnipiac: R+4
OK, so Yahoo is RVs. Tufts has it at D+1 with LVs. So did Ipsos/538 on a poll finishing Oct. 23rd. How much has the environment deteriorated against Dems in 1 week? I don't know. I think Rs are poised for a good night, but this could easily go very badly for them too.
I have my predicted environment at R+1/R+2, which is a bit off from the tracker but well within the error bounds. But 2-pt polling errors happen *all the time*, and historically the direction is hard to predict. What if it goes the other way? May be unlikely, but could happen.
Read 4 tweets
Nov 2
Underlying this is a deeper methodological question. I've said repeatedly that this is not a model and that my forecast is an R+1 or R+2, but I'm under no illusions that people will listen. But I can't change the criteria now just because I disagree with the current result.
It would be quite easy for me to change things around a bit. I could go in and put in the Marist "definitely voting" screen, or I could relax the 538 grade criteria to take CNN/SSRS and NewsNation, and both of those things would show Rs leading by a margin that makes more sense.
But would I do that if the polls skewed it the *other* way (i.e. if CNN came out with a D+5 somehow today)? No.

So I can't do the reverse here. I'm not going to bend the criteria post-hoc simply because I don't agree with the results the aggregator shows.
Read 4 tweets
Nov 1
Deleting my Emerson tweet because we don't have proof that they polled or weighted incorrectly and I really don't want people to jump to a conclusion on the stats alone. To be honest, I have no idea what Emerson did here. I don't know how you can get a difference like that.
It is *extremely* weird for them to get a poll that very weirdly lines up with the old partisanship of the districts better than the new ones (#NM02: Trump +10 in 2020, Biden +5 in 2022, recalled vote Trump +11?). Also, why are natives 4% of the electorate in #NM03?
Same thing in #NM03, where the recalled vote is Biden +18 for a seat that was Biden +18 in 2020 and Biden +11 after a redraw.
Read 7 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

Don't want to be a Premium member but still want to support us?

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

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!

:(