iiinova — where your digital self keeps evolving

Joel Wertheimer @wertwhile.bsky.social · 01/10/2026

From ad impact, democrats closed a 26% spending gap to just 7% in airings bc of lowest unit cost rule. Going forward might get to an advantage. Hard dollars are good!

072

Joel Wertheimer @wertwhile.bsky.social · 30/09/2026

I think 1) it's plausible MN Dems are unpopular after fraud stuff, 2) northern whites do seem to be swinging less than other voters, and 3) these polls are all largely shit

010

Joel Wertheimer @wertwhile.bsky.social · 30/09/2026

Here is IRA architect @janeaflegal.bsky.social explaining why it's so good.

0282

Joel Wertheimer @wertwhile.bsky.social · 30/09/2026

Via @purrtah.bsky.social, the ActBlue machine is kicking into gear, which will shift ad spending dynamics a lot because it's mostly hard dollars.

1211

Joel Wertheimer @wertwhile.bsky.social · 29/09/2026

From Nate Cohn on Twitter in 2024 here are actual vote probabilities by self reported likelihood of voting using the voter file.

180

Joel Wertheimer @wertwhile.bsky.social · 29/09/2026

Using these numbers and a probabilistic model of turnout would make LVs more like D+12.

23210

Joel Wertheimer @wertwhile.bsky.social · 24/09/2026

But one thing worth noting is that @gelliottmorris.com does have Maine modeled as basically about the same Trump disapproval as Texas and North Carolina. Swings have been sharpest among young and nonwhite voters and boy does Maine not have many of those.

2152

Joel Wertheimer @wertwhile.bsky.social · 23/09/2026

CNN did not post an LV model of their poll today unlike the same time period in 2018 and 2022. They did show Democrats leading by 20 among extremely motivated voters though.

191

Joel Wertheimer @wertwhile.bsky.social · 23/09/2026

The main reason I'm less concerned about polling error is that there are just smaller errors in the midterms. Using 538 data absolute error in midterms averages 1.5% rather than the 2.5% in presidentials and 3.5% the last 3 cycles.

2254

Joel Wertheimer @wertwhile.bsky.social · 22/09/2026

One thing about Maine (and New Hampshire) that I think is maybe under-appreciated is that a lot of people are about to buy like 250 gallons of diesel to heat their homes this winter.

53510

Joel Wertheimer @wertwhile.bsky.social · 22/09/2026

One issue that doesn't fully resolve things is Active vs Inactive + Active Reg. In Iowa inactive reg is just missing one election so the total is probably better. That's just -8 for Dems.

020

Joel Wertheimer @wertwhile.bsky.social · 22/09/2026

I’m as skeptical as the next guy of this particular poll but were this a likely voter poll that would be reasonably accurate I think. Democrats are going to have a turnout advantage this year and going from -9 to -6 seems reasonable.

270

Joel Wertheimer @wertwhile.bsky.social · 20/09/2026

I'm sorry if your reaction to the Hamilton campaign executing a perfect oppo hit on Roger Marshall is to complain about the campaign, you have a terminal case of Dems are bad brain. The Times article didn't just come out of nowhere!

1528

Joel Wertheimer @wertwhile.bsky.social · 19/09/2026

Ok, worked with Claude to try to build a little model of GRP advantage and money advantage. Not perfect or anything and some really rough estimates but headline is something like this when you factor in Dem candidate expected additional fundraising. docs.google.com/spreadsheets...

4133

Joel Wertheimer @wertwhile.bsky.social · 19/09/2026

Here's a chart of how things standing using that Ad Impact data and also some FEC data though it's a bit out of data. You can see 1) that the GOP advantage goes down in terms of future reservations and that's overstated bc of Talarico reservations but also bc of hard/soft distinction.

190

Joel Wertheimer @wertwhile.bsky.social · 18/09/2026

The polls right now are apocalyptic for Republicans but also that is exactly what we should expect. Leave aside everything, Trump's approval rating right now is -20.4 per Nate Silver, in his first term on the same date it was -12.6. It's the single most unpopular a president has been on midterm 9/18

46611

Joel Wertheimer @wertwhile.bsky.social · 18/09/2026

Yes but if you combine shelter CPI and mortgage rates into a single mortgage payment index this is what it looks like @proptermalone.bsky.social. Obviously this isn't most people's mortgage rate but it obviously matters a lot! (Also stop citing UMich data)

4355

Joel Wertheimer @wertwhile.bsky.social · 17/09/2026

The two polls from today have Talarico improving among Hispanic voters though not to Beto levels. What they really have happening is a big gain among white voters, which is perfectly consistent with moves we've seen in Texas.

0667

Joel Wertheimer @wertwhile.bsky.social · 17/09/2026

What would it take to get Blexas this year? What if I told you, Democrats would win in Texas if they got Beto's numbers among non-white voters per exit polls and *Kamala's* numbers among white voters per exit polls?

38610

Joel Wertheimer @wertwhile.bsky.social · 17/09/2026

Underneath the hood, however, what you see in Texas is that white voters have been trending left consistently. In 2024, white voters shifted substantially left in Texas relative to the country (this is two-party share for Democrats) and Latino voters shifted substantially to the right

3648

Joel Wertheimer @wertwhile.bsky.social · 17/09/2026

Gonna do a little thread here jumping off this post because I think what's been happening underneath the hood is really interesting. First is that Texas has gotten much more Dem since 2012 relative to the country, but gave back some of that change in 2024.

111124

Joel Wertheimer @wertwhile.bsky.social · 15/09/2026

Had an AI post I needed to get off my chest. wertwhile.substack.com/p/it-doesnt-...

060

Joel Wertheimer @wertwhile.bsky.social · 15/09/2026

A quick chart with states of interest. I think Dems win Alaska because Peltola has a candidate quality difference greater than 4.5 and same Talarico v Paxton at 3.5. Ohio close. Kansas we might really have that gap also. Makes me marginally more nervous about Maine and Michigan.

3192

Joel Wertheimer @wertwhile.bsky.social · 15/09/2026

Two most important charts that also feel right to me more or less. First one is relative swing using specific demographics compared to uniform national swing and then just the state baselines.

281

Joel Wertheimer @wertwhile.bsky.social · 14/09/2026

If friends and family ask you who to donate money to this fall, really think you could do worse than this from Nate Silver, the Senate voter power index, which is basically a state's odds of being the tipping point in the election relative to its population.

190

Joel Wertheimer @wertwhile.bsky.social · 11/09/2026

Or what about when he said he would have stood up and applauded Trump at the State of the Union

020

Joel Wertheimer @wertwhile.bsky.social · 11/09/2026

Definitely an example of how moderation does not work

110

Joel Wertheimer @wertwhile.bsky.social · 11/09/2026

This guy definitely disagrees with searchlight

140

Joel Wertheimer @wertwhile.bsky.social · 11/09/2026

Comparing NYT Siena crosstabs with the (good version of the) exit poll, the big miss was with Latino voters which Allred only won by 7 rather than 18.

040

Joel Wertheimer @wertwhile.bsky.social · 11/09/2026

In Texas the last two midterms on average did not overstate Dems. In 2020, polls in Texas overstated Dems but the error was mostly in line with the national secular error. 2024 is really the only big Texas specific miss.

2234

Joel Wertheimer @wertwhile.bsky.social · 11/09/2026

Grocery inflation coming to a store near you.

33911

Joel Wertheimer @wertwhile.bsky.social · 09/09/2026

I thought Adam Hamilton was a good candidate then I saw he was going to depress turnout by being too moderate so I’ve changed my mind.

150

Joel Wertheimer @wertwhile.bsky.social · 08/09/2026

Interestingly, in NYC this has actually accelerated the other way, with shootings dropping fast over the past month. Could be as low as 240 murders in NYC this year.

080

Joel Wertheimer @wertwhile.bsky.social · 04/09/2026

There's so much changing in the underlying data constantly but if you zoom out the Fed has done a very good job of keeping aggregate nominal income growth at 4.5% for 15 years absent the pandemic and it's probably the best thing for them to target in a lot of ways.

041

Joel Wertheimer @wertwhile.bsky.social · 29/08/2026

Did in fact get to 8. We need a poll of just decoy Dan voters to figure out what this means but this is like 60/65% of general election turnout I don’t see how Peltola isn’t a big favorite at this point.

1272

Joel Wertheimer @wertwhile.bsky.social · 28/08/2026

Here's through July 2026, direct from BEA. It's still holding true and has actually plateaued. People should just say they don't like the ads! apps.bea.gov/iTable/?reqi...

330

Joel Wertheimer @wertwhile.bsky.social · 28/08/2026

It’s literally just Joey’s chart divided by PCE.

110

Joel Wertheimer @wertwhile.bsky.social · 27/08/2026

Using the DFP poll of Alaska allocating decoy Dan's voters based on his crosstabs you get something like this with Dems having about 53% of the 2-party vote. I think I feel pretty good there given I don't think low propensity voters are coming to save the GOP this election.

082

Joel Wertheimer @wertwhile.bsky.social · 27/08/2026

These are the two elections I think that you can do the most comparison. It's pretty close but obviously these were two elections where the general electorate was expected to be more GOP than the primary electorate. Not sure that's true this year.

130

Joel Wertheimer @wertwhile.bsky.social · 20/08/2026

If Nate Cohn published numbers like these from Emerson (which all seem reasonable) I just don’t think the democratic primary electorate will risk it.

4141

Joel Wertheimer @wertwhile.bsky.social · 19/08/2026

That salary is not 5-6 market rate for those responsibilities come on. These are big complicated events and marketing people make a lot of money especially ones living in San Francisco.

0130

Joel Wertheimer @wertwhile.bsky.social · 19/08/2026

On X @winwithjmc.bsky.social posted this total for turnout earlier today. So even though Democrats in Florida had a huge turnout drop it looks like relative to registration a slight improvement from 2022. Cursed state.

062

Joel Wertheimer @wertwhile.bsky.social · 19/08/2026

You can download and read them here, it tracks. Here's DW-Nominate in congress also here are social trends from DW nominate. dataverse.harvard.edu/dataset.xhtm...

010

Joel Wertheimer @wertwhile.bsky.social · 18/08/2026

Bonica himself (as cited by Morris) published a paper that said ideology matters particularly in competitive races and high salience ones like a US Senate seat. They don't publish it, but you can infer something like 4 points of margin in a competitive Senate race from ideology alone.

041

Joel Wertheimer @wertwhile.bsky.social · 18/08/2026

I do not get this claim about the Bailey Reese paper. How the candidates talk is about policy and how would the effects be mediated through voters other than through voter perceptions of policy and ideology as understood through the candidate's communications? Candidates are what they say they are.

3152

Joel Wertheimer @wertwhile.bsky.social · 17/08/2026

Fun with LLMs. Using the same method as @weisenthal.bsky.social did for FedLock I made a pairwise tournament (ELO basically) asking Deepseek to read state party platforms going back to the 1950s and on two dimensions, social and economic, evaluate which platform was more liberal/conservative.

4325

Joel Wertheimer @wertwhile.bsky.social · 17/08/2026

But also your own measure of WAR has Baldwin underperforming. Nate's point, to me, was that she's viewed as left wing rather than extremely so, and I think the gap to Warren is demonstrative.

261

Joel Wertheimer @wertwhile.bsky.social · 16/08/2026

People are mad at Nate Cohn over this statement but it seems pretty consistent with Bailey Reese (2025) which looks at how candidates talk about themselves in evaluating the impact of moderation on campaign outcomes particularly in competitive races. papers.ssrn.com/sol3/papers....

2153