As we collectively try to refocus on winning in November, I figured it might be helpful to have a quick reminder that a primary and a general are *very different elections*. They have different electorates, different incentives, and operate in a different ideological space. Doing well in a primary does not necessarily mean you will do well in a general, and vice versa.

The two ways to think about these changing electorates are as a set of demographics, or as a set of policy preferences. The first will tell you that primary voters are older, more ideological, more educated, and tend to be more white, although this all varies slightly by geography. This is fine, but I don’t find it that useful for building a mental model of the differences. The second I find more useful.

Courtesy of the Broockman and Kalla paper (“Should Moving to the Middle Win Candidates Votes? It Depends Where Voters Are”, which laid out voter preferences across a range of issues, as well as estimating the impact of changing candidate positions on each of those issues) we have some pretty dang nice charts of these differing electorates. This gives us an ability to compare *policy* preferences in a way that’s quite hard when working across multiple surveys. All these figures can be found in their preprint, linked above, as can the full text of each position on each issue (in the appendix).

This survey was performed on voter-file-matched registered voters, weighted to the 2024 electorate. Democratic and Republican primary voters are then broken out from that population. Not an exact match for every primary, but pretty good. And the survey is huge, too (n=6k) which gives you enough power to get a good look at these subgroups. They asked voters to select from three positions on each issue, the liberal one, the elite middle one, and the conservative one. Full positions can be found in their paper, but here’s an example:

This is designed to feed into the conjoint that’s the main point of the paper (you should REALLY read it, it’s good), but it also makes a nice comparison tool because you can pin down what specific positions these voters are accepting or rejecting.

Here are your Democratic primary voters. (sorry, it’s a screenshot)

Democratic primary voters are as you’d expect, *quite* liberal. 84%1 take the liberal position on guns, 81% on minimum wage.

Here are your Republican primary voters. They’re a mirror image- very conservative on most issues.2

And HERE are the general election voters. This group contains the two groups above, plus more people who didn’t vote in a primary but would show up on election day.

Squinting at all these charts to do a comparison is a bit tricky, so I had Claude transcribe the numbers (approximately) and made this one, comparing the share selecting the liberal position from each section of the electorate. Note this obscures change from the middle to conservative, etc, but it was the only way to make the chart readable. Edit: The authors have kindly provided the underlying data, so I have updated the charts! I am still comparing the share selecting the liberal position from each section of the electorate. Note this obscures change from the middle to conservative, etc, but it was the only way to make the chart readable.

Unsurprisingly, the general electorate is far less liberal than the Democratic primary electorate on most issues. This is a little bit definitional - Democratic primaries contain *Democrats* (although some primaries are open, etc), who tend to hold Democratic positions, and therefore liberal positions. But it’s good to keep in mind when trying to make the mental pivot.

Here’s an additional chart of the gaps between the Dem primary electorate and general electorate (on % selecting the liberal position) by issue.

As we move towards November, remember: the general election is a different beast. General election voters are not the same as primary election voters. And beating Republicans is what matters.

In his final filing, July 15th, covering spending through June 30, Graham Platner’s campaign reported $20,876,528 in contributions over the cycle and $19,111,060 in expenditures, with $1,763,058 cash on hand.

This is not really a campaign that was broke at the end, unless they spent a ton between June 30th and dropping out. $1.8 million cash on hand isn’t awful, even if they were clearly losing money at the end there. It is a campaign that raised and spent a massive amount of money over a relatively short period.

So....what was the money spent on?

The Money Went To Advertising

8.4 million dollars alone is media buys through LC Media LLC1. In total, LC Media had 40% of his spending. There’s some very large buying after the primary, presumably ramp up to general election advertising, around 600k. The media buys in general are *huge*. On the advertiser side, Meta reports 3 million dollars of advertising spend from Platner on their platform. Google reports 211k, heavily focused in Maine. The Meta spending is slightly skewed towards California but otherwise nation-wide. From a quick look, the Meta ads are heavily, heavily fundraising (hence the nationwide focus), and the Google ones are mostly persuasion. I can’t be precise here, but this looks like enormous media spend mostly for the purposes of fundraising.

2.7 million dollars is “digital consulting” through Helix campaigns, with another 2.2 mill for “digital advertising” through that same vendor. This puts Helix at 24% of his spend, and a huge amount of “consulting” spend. The consulting spend is particularly notable because early in the campaign it was billed as a retainer, at $100k a week in September/October, even before Platner had started raising a ton of money. 1.5 million of this was in Q4 2025, mostly in October. The entry dates (and transaction numbers) are a little strange, it kind of looks like these payments were all logged at once, and they’re followed by a single $418k payment marked Nov 14th. The filing discrepancies in October/November make it very hard to figure out if the giant payment is an anomaly or an oopsie correction to bad payment management. $418k at once is weird. I don’t know what’s going on there.

Together, these two vendors made up ~64% of Platner’s total spending.

Share

Both Helix and LC are linked with Platner’s top strategist, Dan Moraff. His consulting firm (Dark Forest LLC, I have a side note on this name below) also got about 108k directly. I believe Workers Productions LLC (video vendor) is also Moraff-linked, bringing Moraff-linked enterprises to around 70% of Platner spend.

So....ads! It went to ads. And consulting. Something like ~12% of his total spending went to digital consulting. That’s a very online campaign, or a broad definition of “digital consulting”.

SIDENOTE

Re: Dark Forest LLC, if I am identifying my type of guy correctly, this name is a reference to a theory popularized in the Dark Forest (sequel to the Three Body Problem) by Liu Cixin. The theory posits that aliens *do* exist, but that broadcasting your location is too risky, since those other civilizations would immediately try to destroy you because of game theory. Since you can’t be sure the other civilization will be benevolent, you should immediately destroy them, and since you would expect to be immediately destroyed, you should never risk making contact, etc, etc.

This is, impolitely, dumb as hell. I have not read this book (because I didn’t really like Three Body Problem, but also to be fair I read it entirely while getting a tattoo...) so I can’t comment much more, there’s a wikipedia page, you can read for yourself. The naming is really beside the point except that “guy who is really into game theory [aliens]” is a character some readers may recognize from their own lives and I think that context is useful. Also crucially, the book came out in 2008 at which point Moraff would have been around 16, classically an age where you imprint on sci fi novels, so I think this is a reasonable enough theory. Edit: nope, the english translation was only published in 2015. I still think it’s a reasonable theory.

Also sorry Dan Moraff if you just really love forests2. I suppose that is also possible.

Sidenote 2: LC Media + Bareebe

Another quasi-side-note, the new big client for LC Media is *very odd*. They’re working with Quincy Bareebe3 to the tune of 2.7 million, on a self funded campaign that is currently getting very sued. There’s also some weirdness where the campaign has just backfilled months of FEC filings, and the top vendor (Kennedy Communications) is doing *everything*. I didn’t dig too much here, but like...hm!

Return on investment?

As with many filings, Platner’s team does not tend to mark “fundraising” spend very cleanly. Spending with fundraising-only vendors is marked (it’s around a million dollars over the whole campaign), but flipping through the ad library you can see that a lot of the general ad spend was for fundraising as well. This makes it a bit tricky to determine what the return to that spending was.

If we handwave and assume that the LC and Helix spending was significantly for fundraising purposes, we end up with returns as high as 76% in the first quarter, and as low as 2% in Q1 2026. This low number is probably an outlier- due to the massive (pre?) payment to Helix- but the latest filing still has them spending 83 cents of every dollar on fundraising, so it isn’t impossibly far off.

With slightly different assumptions, we can choose to remove the payments to Helix for “digital consulting”, treating those as more of a general campaign consulting thing. This drops spend on fundraising a bit, but the latest few filings still look QUITE bad, with Q2 2026 having 78% of money raised spent on fundraising.

Platner was paying Helix comparably with other campaigns, such as Eileen Laubacher (in CO, running against Lauren Bobert), also paying them 100k per week. Her tagging of payments is a bit different, but that is usually a bookkeeping difference. Osborne is paying much less, more like ~30k a week, which may be explained by the overall smaller scale of his campaign (~4 million vs 20 million). LC Media doesn’t have other comparable clients. The Bareebe payments look similar-ish, but that’s kind of just what media buys look like, so it’s not much of a matter.

Conclusions

I started this digging before the last filing dropped, and I wasn’t sure what I was going to see in the final (ish, it doesn’t cover July) filing. I was vaguely hoping for some exciting new spending, I guess. What’s in there is mostly “ads, but more”. It is a very large amount of money to be spending on ads in the weeks after the primary election ($600k!). I’m not sure what to make of that! There’s a branch of campaign strategy (that I think is wrong) which loves early ad spending, but early usually doesn’t mean “literally the week after the primary”. Maybe it was fundraising, trying to make up the deficit of unimpressive Q2 raising?

I do think that interested parties should keep an eye out for this kind of payment structure in other viral/insurgent/outsider campaigns. Super high retainers to firms that don’t otherwise have a ton going on indicates, if you’re skeptical, a very efficient way of sucking money out of the hot new thing. Or, the charitable version, it indicates a huge operation betting the farm on some high risk outcomes.

In any case, if you want the story of how Graham Platner spent his money, it hinges on two firms: LC Media and Helix Campaigns.

I am tired of thinking about 2016. The general election is a big ol smear across the history of my industry (when polling broke for the first time in my career, but not the last). I worked on it, it sucked, it’s caused my entire political career to be “dealing with Trump”, although I am reliably informed politics was normal-er once. I am deeply sick of Trump, but since he’s still an active threat to the country, I’m usually worried enough to not be purely exasperated.

What I am purely and entirely Done With is the 2016 primary- on which I did not work, on account of being in college at the time. My memories of the actual events of the primary are of an exciting, competitive contest. At the time, I was already a little over the endless drama of DNC insiderism/emails/who exactly was or wasn’t pushing for Clinton. I found it irritating, especially since in the end, it all came down to voters....voting.

At this point, a full decade out, my opinion has shifted: I don’t really care if a primary election is “fair”. I don’t care if the 2016 primary was fair, I don’t care if the 2020 primary was fair, the 2024 primary such as it was was definitely not “fair”, whatever. What I care about is if they selected a good candidate.

Primaries are not a democratic process. They’re just not. When I worked on the Warren campaign, I spent a year pouring over the details of the primary process (and the caucuses, and the delegate allocation rules). I know more about the 2020-era DNC primary rules than I care to recall. The upshot of all that staring was Biden sweeping South Carolina and none of the clever math really mattering, which may contribute to my cynicism here.

But those rules are just...fake1. They’re not legally binding. They are not voted on in any meaningful sense. They change every cycle. We don’t quite know what the 2028 version will be yet (Josh Putnam at FHQ is the follow if you’re interested in the decision process). The rules-wrangling starts well after candidates have begun putting out feelers. A party could decide to do some true nonsense (boxing match, Mario Kart tournament) to select a candidate and it would get them on the ballot just as well as holding primary elections. It’s a popularity contest in a silly hat.

The shadow of the 2016-era “are the primary rules fair/is the DNC biased” Sanders-Clinton fight has hung over every single primary since then and I am so tired. It has been a full decade and the battle lines in the discourse are still the same every time2, even when the actual coalitions are super different (the Chair of the Maine Democratic Party was literally a Bernie delegate)3. It’s now an easy pattern for a certain type of candidate (lefty, outsider-y, doing a populist bit) to claim that the people running the primary are biased against them in some sense. This sleight of hand lets you brush off a loss and add shine to a win, with the one teensy tiny downside of spreading conspiratorial fallout over a perfectly ordinary process. There’s no consequences for framing every setback as the fault of a nebulous evil Them, I’m sure.

This is all on display in Maine. Platner and his allies have spent the last few days throwing accusations at the Maine party and preemptively complaining that the process to replace him won’t be fair. They’re also preemptively blaming the party for a general election loss, as if “losing to a highly successful Republican incumbent after your primary dissolved into chaos” is a fun situation the Maine Party chose on purpose. We don’t yet know how the replacement process will go - it seems like a snap convention, and the logistics here are a total nightmare- but I’m sure the complaining will continue.

Ten years on, what if we left this all behind for good? We could have new, different arguments. Perhaps they could be about policy, even. It might be a nice treat.

Anyways, goodbye Graham Platner, see you never.

Tom Steyer spent $216 million dollars- PERSONALLY- on his failed bid to be the Governor of California. This is around 27 times what the famous Governator1 (Schwarzenegger, obviously) spent on his successful campaign, then considered big money. Around $200 million of that was reported to go into ads alone (and congratulations to the ad buyers who got their ~20% cut of that).

Saikat Chakrabati spent almost $9 million on his (also failed) campaign for CA’s 11th district. He got (current count, NYT) 41,684 votes, putting him at around $215 per vote. That’s somehow worse than Steyer’s $106 dollars per vote.2

At a certain point, you have to conclude that they’re not simply buying votes, because it would have been cheaper to do so.

Chakrabati was reported to be spending up to $45 an hour on an army of paid canvassers, over 250 of them. That ~56k a day is a drop in the bucket for such a well funded campaign, so one must imagine the limit was ability to recruit, not funds. At the reported rate of contact of ~6 people an hour, that army can reach 7,500 people per 5 hour day, taking a bit under 6 days to reach *his total number of eventual voters*3.

Clearly the problem was not volume, but ability to persuade.

Both these campaigns had all the resources they could possibly want, as well as a huge lead time on their race. They both had enormous amounts of staff, and full ability to choose their hires in a relatively slow part of the cycle. Steyer, at least, had the benefit of name recognition holding over from the last time he burned a few hundred mill on a presidential run. They both got national write ups about their candidacy, and were talked about endlessly on twitter. For both of them, the voters said “ehhhh, nah”.

We don’t have access to a counterfactual world where these two candidates were persuaded to avoid spending their money in the interest of helping counterfactual-us learn about its impact. We can’t really calculate the marginal impact of their spending, other than as “not enough to win, I guess”. What we can conclude is that infinite money alone is not enough to assure victory, at least not when it’s spent in this manner.

I find that....kind of heartening, honestly. You’d think $200 million could buy you pretty much whatever you wanted. Evidently, in this case, it can’t buy you victory.

There’s a huge pile of firms, vendors, and individuals who work for campaigns and committees. Their business models vary, between a complete focus on politics to mostly corporate work (it pays the bills in off years!) with a little political work on the side. They are getting a lot of $ for politics, but relatively little in the scheme of corporate work. You probably haven’t heard of them, unless you work there or you’re mad at them. Let’s take a look. This is 2026 spending, from the bulk filing, to campaigns and committees. Not PACs, and nothing secondary (i.e. I can’t tell where these firms are spending their money after they get it).

A note, as usual: The amounts of reported payments to vendors can vary depending on their practices. Pass-through costs are not broken out, so things like media buyers have much higher amounts of $$ going to them than your average consultant. You can also do clever tricks with shifting money around to hide payments to a given vendor (although like, why). Also as usual, I have no particular insider knowledge- sorry if I get the purpose of your company wrong. If you DM me a reasonable edit, I will fix it.

The Top Firms

A table!1

Ads and digital

The top $-getters are ALWAYS ad buyers. They are directing spending from the campaigns, so a chunk of that $ isn’t really staying with them, which inflates their numbers. However, they do generally take a percent of spend (up to ~20%, it depends) which is really damn nice work if you can get it. I feel good assuming these folks are making some solid profits.

We start with Aisle 518 Strategies, run by Tim Tagaris and coming out of the Sanders small dollar digital fundraising operation. They look to be doing digital fundraising/advertising/consulting for many of the big raisers- Ossoff, Talarico, Kelly, Peltola, Vindman, Slotkin, the list continues. There’s one funny collision where they are getting paid by Zach Wahls and the (recently Turek-supporting) VoteVets. You can probably take this expansive list as evidence of how much the vendor ecosystem is shared across candidates who can take dramatically different (and oppositional) postures.

Thematic Campaigns looks to be a more TV buy oriented firm, with lots of Obama name-drops on their website. Adelstein and Associates is similar, more old school, with a set of PAC clients on their website who won’t be showing up in this data. That is where the truly huge $ is. It is worth noting that their big spender is Raja Krishnamoorthi, and that they didn’t make anything like this amount last cycle. Maybe they’re on their way up, or maybe they’re just diversifying.

Lower down the list, you get more of a grab bag. There’s a bunch of ads-and-fundraising: MissionWired (who I think I may be missing payments to due to name normalization issues), Middle Seat, Liftoff, Adwell. Plus, one largeish direct mail firm, RWT, with smallish amounts of spend by a ton of candidates on top of $4.6 million from the DCCC.

Boring

There’s a handful of boring entries, credit cards, payroll, etc. I am slightly amused by Gusto taking in a huge number of clients just to make up what Rippling is getting from the DNC alone (their main customer on this list), but since they both have a (presumably) far larger non-political client base, I bet this is all small-potatoes money to them.

The ActBlue amount here is *only* fees (pass through donations are counted differently), which is kind of a shocking level of income on fees alone. Of course, they’re handling donations for basically every Democratic campaign, so it adds up.

Lone Republicans

Our first Republican firm is Targeted Victory. They’re a big fundraiser and digital agency, working for a laundry list of Republican candidates. It’s all very dull (to me, at least), except for the scale comparison this presents with the multiple D firms we just talked through. The Republican ecosystem is more centralized, and (as far as I can tell) relies much more heavily on pass-throughs. On a Democratic filing, you can see a pile of firms working direct for the campaign. Republican filings are more likely to have a central consultant, presumably farming out work, or to omit this “normal” spending entirely, and have it covered by other actors in the system. Which is to say: that spending exists, but it’s in PACs.

There’s also WinRed, the ActBlue equivalent (really creative naming....), making a decent amount of fees.

They’re followed by a couple media buyers, one of whom (SRCP) has a website with an incredibly clear vibe [derogatory, screenshot below]. They both have eye-wateringly larger amounts of money earned last cycle (278 million for the bad-vibes SRCP), so I’m guessing this low-$ state is due to it being early.

More

I will not bore you to tears going through the rest of this list. You can read. If you want a larger list, here’s a dreadful chart that you can download the data from, yay! Or if you’re more willing to do work, grab the F3 bulk filing and aggregate.

It’s all ads

My interest in looking at vendors is to start getting a sense of who the players are, and watching how they change over the cycle. Right now, we’re looking at mostly primary campaigns and the start of general election work, in a midterm year. It is not yet the time of big money. I think the most interesting next steps would be to catalog what vendors a campaign was using and look for any shifts over time that might tell you someone got dropped or brought on.

The topline notable thing is just how far down you have to scroll before it stops being all media buying, fundraising, and pass through costs that scale to fundraising (like ActBlue fees). A *stunning* amount of money in politics is more like money running through politics. It comes in and goes right back out to pay for the cost of raising itself. It’s a bit of a depressing treadmill? I’ve written about my skepticism of small dollar donations before, and the picture continues to be grim.

The other thing I’m struck by how *little* money this all is in the scheme of things. This came up on twitter the other day, but like, that 20 million dollars going to Aisle 518 is a lot for a person, but also the cost to build a medium-ish hotel. Yes, you can argue that building a hotel makes a physical thing appear and is less of a waste, but surely you’ve gotta value access to the levers of power in the US pretty highly.

Anyways

If the DNC would like to stop doing things while I am trying to write a normal, dull post about some data, that would be wonderful. I don’t think I have anything novel to say, other than “oh my god, literally any of the dozens of wonderful analysts I know could have done this ten times better”. Get it together.

Share

Where the current DNC spends its cash

One stated goal of the current DNC is to distribute money more widely outside the battleground states (and thus, the richest state party organizations), in the State Partnership Program (https://democrats.org/news/dnc-and-asdc-announces-organize-everywhere-win-anywhere-strategy-largest-ever-monthly-dnc-investment-into-democratic-state-and-territory-parties/). Not all this money is federal, so only some of it is showing up in the FEC filing. In their April filing, 1.2 million dollars in federal money went to the state parties. However, around half of that was PA alone (43%, $525k), which is presumably a ramp up to the midterm elections investment. These spending commitments lock the DNC into ~1.2 million (across federal and nonfederal dollars) per month, around a tenth of its overall fundraising.

Payroll is one of the largest expenses, at 1.44 million, across ~249 employees this pay period. This appears to have been flat since the end of last year, with the total # of employees about the same as November 2025 [if you run this analysis yourself, you’ll probably initially get a larger number, because payroll is reported both as a bulk payment to Rippling, the payroll processor, and then again as payments to individuals. This is a duplication of the same spending]. If you add on the other payroll-associated expenses (benefits, taxes, etc), you land around 2.7 million for staffing costs.

The total amount of fixed, non-discretionary costs (the two above expenses, plus office rent, legal compliance, data and technology subscriptions, but not advertising in any form) lands around 6.4 million total. That is (approximately) half the DNC’s monthly fundraising amount committed off the bat to fixed costs.

The high, high cost of money

It’s tricky to accurately measure fundraising costs from an FEC filing. You can get dollars paid to vendors, for ads or texts or similar, but you can’t easily track staff time, or figure out exactly how much of an ad dollar went to fundraising versus persuasion. We are slightly lucky here in that the March 2026 DNC has no particular reason to be running a large persuasion campaign (and we can observe transfers to state parties that seem to be for that purpose), which makes me feel good about attributing all ad spending to fundraising. There’s also direct mail, which is again uncertain but probably donation requests.

Direct mail alone comes to ~1.96 million in March, and digital fundraising plus texts and other online stuff is a further ~2.7 million. The total here lands around 4.7 million- around 41% of the 11.4 million raised in March. It’s worth noting that this fundraising efficiency is the worst in the 14 month window I checked, and the average is more like ~23% efficiency, but it varies a ton. Notably, the DNC is spending more on fundraising in 2025, while raising less. This may be an early in the year aberration, but it doesn’t look great.

At the March fundraising efficiency, the DNC needs to raise 10.9 million to cover fixed costs. At the average efficiency, it needs to raise 8.3 million to cover fixed costs.

This is not great news for an organization whose fundraising looks like this

If you look closely, the thing currently propping up the DNC is a fifteen million dollar loan reflected on the October 2025 filing. The committee has around 18 million in total debt as of the most recent filing. The low September/October fundraising numbers are in spite of an increased amount of spending on fundraising activity, although it is possible this is miscategorized spending on the elections or fundraising for other groups.

Adding this breakdown to fundraising numbers, and excluding transfers from other committees to focus on individual donations, actually makes the DNCs fundraising efficiency look worse- March goes from 11.4 million total to around 7.5 million of individual donations, for a dire 60% of every dollar raised going to pay for fundraising. Ow.

A slow bleeding

You can spend a long time failing to raise money and taking out loans, and it’s likely that the midterm runup will be a fundraising bump. I don’t think the DNC is going to collapse tomorrow.

However, this financial picture is dire. The combination of high fixed expenses and slow fundraising, despite significant investments in that fundraising, paints a picture of a lifeless entity barely holding on. This is all particularly ironic in the context of Ken Martin’s goals of giving money to the state parties, since you can’t give away what you don’t have. Much has been made about the need for a stronger (or less hollow) party, but the DNC is in no shape to be a strong anything.

Photo by Mulyadi on Unsplash. Retrieved by searching for “sad robot”. This is the saddest I could find and it is not very sad.

Once again I got distracted while working on a more productive long term project. This is partially inspired by this tweet about AI uptake among political consultants. That’s using a survey of consultants, which is a reasonable (if tiny) way to study it. Since I had my hands in FEC data *anyways*, I took a look.

Tl;dr: There is evidence of AI uptake among candidates, PACs, and committees, but either it’s relatively small, or spending on AI tools is mostly going through passthroughs that I can’t measure.

Couple of assumptions here:

1. To be counted as “using” an AI tool, a candidate or committee must be paying for it through their org- I would guess use of the free version of ChatGPT is way higher (if this is you and you’re a Democrat, *stop that*), and this can’t capture staff upgrading AI tools on their own

2. That payment would have to flow directly from the campaign/committee to the AI company- this doesn’t capture passthrough orgs (from credit card payments, or IT vendors)

3. This doesn’t capture AI use by consultants at all, since they’re doing it on their own dime initially, even if they’re reimbursed.

4. AI tools that are bundled aren’t included initially- Google Gemini just shows up as normal Google payments

AI Tools and Their Users

Democratic candidates seem to be more into tools that use AI for extremely specific purposes- see Otter.ai, a transcription tool- than they are into generalized AI. ChatGPT is the most popular general-AI tool, but 9 candidates out of a pool of ~hundreds isn’t huge. That’s on the scale of a popular but not dominant fundraising firm. It’s hard to figure out the precise size of the pool of possible candidates, since the size of campaign and detail level of spending filings drops off rapidly once you get out of really large campaigns, so I can’t give you a “X% of candidates” number. Claude is popular with PACs, but not so much with candidates.

It basically drops off a cliff after that. There’s a handful of other identifiable-AI tools, with one or two users, but not much else.

Share

Frankly, this is way lower usage than I was expecting. I figured that ~most campaigns would have shelled out for some kind of AI subscription at this point, if only defensively to get their staff off less-secure free versions.

I have a couple possible explanations:

1. Republican campaigns run everything through vendors, and those vendors don’t have to report in the same level of detail. This may result in them being functionally excluded from this analysis, and suggestive data (survey results as discussed at the top of this post) say that Republicans are more likely to use AI.

2. AI tools aren’t in formal-enough use to merit enterprise contracts, and they’re being paid for by a random credit card that’s getting reimbursed. This would obscure the spending, since it would show up as Bob Doe getting a credit card reimbursement for a slightly larger amount than normal.

3. Most campaigns just aren’t using these tools.

It could be a combination!

Gemini and Copilot

You may have noticed these were not included in the table above. This is because they are all billed through enterprise software subscriptions that include other stuff. For example, a candidate might be giving Google a ton of money per month for Google Workspace already, and adding Gemini use to that wouldn’t change the nature of the line on the report.

One thing that is distinctive about AI use is that it (often) requires purchasing credits for tokens. This should show up as something different than a flat monthly payment for enterprise software. Many many caveats here, this only applies if you’re using AI enough to exceed the minimum plan, it depends on the program, etc. Can you use this to detect campaigns or committees using AI through these orgs?

Let’s go with “maybe”. The only one I could find was the DCCC, which has payments to Google and Microsoft that vary in amount and have increased steadily over time. These look like payments for something other than the basic plan. My experience here is with Google, not Microsoft (this post composed in a Google Doc), but I know that varying-amount monthly payments can also happen if you’re making use of Google Cloud and API features. So, either it’s AI use, or it’s working with largish datasets through Google tools.

HUGE SHRUG. I dunno.

Grok

Grok is included with a X Premium account, or an X organization account. I don’t really know what these do but I’m told they’re useful. This puts it in the same category as Gemini/Copilot in terms of AI use I can’t easily detect. The main difference is people don’t spend a ton of money on Twitter. Republicans spend some, but not lots, and none of it matches my theorized AI spending characteristics.

Anyways, no one is spending all that much money on X. Rest in pieces, etc.

Methods

I was working with a dataset of spending data from the top-100 fundraising candidates. Via Claude, I went through expenditures in that dataset for anything plausible AI related, including the major products and stuff that just happened to have AI in the name. There isn’t a comprehensive dataset of AI companies, but I think we got the big ones.

I then took that and used the FEC API (via the very useful agent-fecfile) to look up all payments to those vendors, as well as anything with AI in the name, across the 2026 cycle. I did a bunch of spot checking and removed records that were clearly wrong, including a reimbursement of a donation to an AI company. I also did some googling to see if I was missing any other major AI companies, and to exclude things with AI in the name that I didn’t think qualified.

More important than any concern about AI hallucinations (which I did check for at length) is the fact that FEC data is fundamentally messy. Assume these numbers are all slightly off. Assume there’s double counting and undercounting and misfiling.

The text of this post is, as always, human written.

I read a couple of nonfiction books recently (see the post about Blood In The Machine), which of course meant I needed to immediately retreat into some genre. Sci-fi is my usual jam, and frankly, I am willing to kind of read whatever within the genre because I find the use of tropes and references interesting, even if the book is bad. This is what I reach for when I want to read something, but don’t want to spend too much brain power on it.

A friend of mine categorizes a certain type of book as a “chuggable”. Like the cocktail (or beer?) variety in question, this sort of book needs to meet a couple criteria:

1. Good enough- somewhere between good and okay, nothing that would immediately put you off, but also not so good that you need to spend extra time on it.

2. A decently fast read, maybe a few days. Not a project read.

3. Ideally, part of a series, such that you can get another without worrying about it

4. Engaging, not surprising, fun, but not a shock in any way

This is a book you might grab for an hour after a long work day, with no concern that you’ll miss important details. You shouldn’t need to remember all that much from a previous book in the series, maybe a few key plot points. These books should resolve about how you expect, with every main character either fine or dying gloriously, any big crises resolved, maybe a wedding or a funeral. Notable examples include the Honor Harrington series, most things by Seanan McGuire, the Laundry Files, and the Strange Practice series. If you can read 5 of it in a month, it’s a chuggable.

Share

I’ve started developing another mental category of book in direct comparison to the chuggable- the “shot”. This is a very short book, ideally experimental or otherwise odd, that you can inhale over the course of an evening and spend several days considering. Something you feel the need to immediately reread, and can’t find an easy equivalent to. Ideal reactions include “what the FUCK what that” and “I…liked that? maybe???”

Here are a couple recent “shots” that I have read. If the book description when you google it appeals, go ahead. If not, don’t. Also, shoutout to my local library, I was able to borrow the first two of these—

Walking Practice by Dolki Min (translated)

The Employees by Olga Ravn

Hermetica by Alan Lea

The first two are SO baffling that I am not going to attempt to describe or respond to them here, maybe later. The third is at least in the realm of things I can generate immediate comment on.

Hermetica is nominally in a genre favorite of mine, the generation ship book. I love a long haul spaceship story, and I love a reason for people to be in forced proximity and stressful situations. What is the human condition if not navigating stressful things with groups of people you didn’t get to pick and can’t escape? This was recommended to me by an IRL friend who I hope is not reading this but if you are, hello, I liked it, we have already discussed this.

[from here, spoilers abound for Hermetica. It is 8 dollars and fairly short, if you want to read it go do that, there’s an ebook version]

.

.

.

.

.

Okay neat. Anyways, Hermetica is not actually a generation ship book. It’s a book about the trope of generation ships being used to confine a bunch of people on earth and prompt them to make advances in science. It’s also a pandemic novel- the idea of an airborne disease and reduced human contact play a huge role in how the fake-ship of the book functions. It is fundamentally a book about the human affinity for stories and struggle and how you might use this to control people. It’s published by a small avowedly anarchist publisher, and you can see this ideological bent throughout. The best parts of the book are the protagonist using scientific practices which they were brought up with to unravel their reality. Maybe this is nerd shit, but it was utterly delightful to watch the characters go “hang on, we CAN check if we’re in a rotating ship, with a fairly simple experiment”. If you’re going to do a book where the setting as presented is basically false, this is the kind of ingenuity I want from your characters.

The worst are the bits where the protagonist (or a exposition character) faces the camera reader and explains their politics, or how the world of the book functions. Given a little more space to breathe, I think the author could have handled the exposition gracefully. In a worse book, this exposition would have been boring. Here, it isn’t even boring! It’s cool as hell! It’s just something I wish I would have been able to discover via clues like I did the main setting’s entire deal. The concept is that the government figured it would be easier to slice the population up into chunks and give each the narrative that fit their biases, rather than trying to govern under a shared reality. There’s a group that thinks they’re succeeding from the US (or, presumably the US) and fighting some kind of border war, there’s the generation ship folks who think they’re in space, there’s another group that thinks they’re in Mad Max. The exact purposes of these groups are lightly defined- the secessionist folks do farming (but like, also capitalism), the gen. ship folks do science, everyone else does…..stuff. I would read a vignette from each of these (except perhaps the secessionist folks, I got quite enough of that with the character from that culture that we briefly meet). The Ship people feel much better thought out, with lots of attention paid to education systems and jobs and living arrangements. There’s a lovely balance between extreme detail and lightly sketched concepts that lets you fill in the rest.

As far as I can tell, this is the only work by this author. This is, imo, a shame. There’s a lot in here that speaks to an interesting authorial voice, and novel ideas. As with many first-or-only novels, it could probably benefit from an editor, but so could my blog posts, so I can’t judge. The biggest thing I can say for this book is that it is a “the premise of this was fake” book that didn’t feel like a cop-out. I loathe fiction that tries to meta-lie to the reader (“it was all a dream!”, shut up, you’re a book of fiction, don’t give me this) because it usually feels unfair or deceptive. Hermetica doesn’t feel like that. It feels more like a good detective story, letting you discover things along with the protagonist.

If you too need a short little weird book, consider this one. If you know of short little weird books, I am interested (admittedly, far more interested if I can get them from the library).

So who are these people? I am talking about Democratic donors, because I don’t care to help Republicans win elections, and also because ActBlue is a far better source of data for this than WinRed.

They are, on the whole, of average income, and their contribution sizes reflect that (this is obvi somewhat tautological). They’re very liberal , they’re a bit more female than average, they’re heavily educated, and they tend to donate in the aftermath of big events. The NBER paper on this (here) has the best bibliography, but a lot of the work on this is a bit dated. I don’t have a strong reason to believe these characteristics have changed, though.

[A note to my future self, I should really try using the Cooperative Election Study data to look at donor characteristics more. That seems to be the option for people who don’t have a voter file on hand. ]

The other thing of note is that small dollar donors tend to give more than once. Campaigns heavily push recurring donations (which has raised some valid concerns about scam-like behavior), but in addition, donors tend to give to multiple candidates. In 2020, there was enormous overlap between the competing primary campaigns. Some of this can probably be attributed to how lists of contact information are handled (i.e. frequently passed around and sold between campaigns and orgs), but not all of it.

Donors are a pretty small group. OpenSecrets thinks that around ~1% of the population gives over 200 dollars to a campaign. The ANES reports that around 15% of people gave money to help a campaign in 2024, including 21% of Democrats. We know the number of people donating to campaigns in midterm years is lower, especially before the general election approaches.

But *how small* is this group? What does the Venn diagram of candidate donor lists look like? How much are we recycling the same people from campaign to campaign, and how much are we bringing in new folks?

The answer to this is nominally knowable, but requires a lot of data processing. I have an okay system, but it’s not that good just yet. So instead, I’ve started with a smaller universe: Senate candidates.

Test Case: Current Senate Candidates

There are so many candidates, and ActBlue data is very large. As a starting point, I pulled together a database of donations to 13 current Senate candidates across a few key states. This involved merging ActBlue (for small donations) and candidate filings (for larger donations). By merging those datasets, you get okay coverage across the spectrum of donation size. Anyone who gave less than 200 dollars directly to the candidate is still excluded, but given how popular ActBlue is, that population is small. I would describe this database as “good enough”.

To distinguish individual donors, I am using name + zip code. This is again not perfect! Some people have folks in their zip code with the same name. Unfortunately the way campaigns report full addresses can be very messy, so I think this is the best compromise. Since this data is from a single quarter, hopefully only a small number of donors have moved and thus gotten two records with two zip codes.

A couple of the candidates in this set are running in the same primary. They still have shared donors! Shoutout to the 7k ish Talarico/Crockett donors, are you guys just really not online? Do you just want everyone to have a good time?

How Many Donors Give To Multiple Candidates?

Keep in mind this is limited to candidates *in this pool*, *in this reporting period* (Q4 2025). This is certainly less overlap than you’d see taking more time/candidates into account.

The answer is “tons”! By far the most common candidate to share donors with is Ossoff, because he has a very large number of donors. I would also guess his lists have made their way back into the Democratic ecosystem at this point, and that other candidates are fundraising from lists of Ossoff donors. He is serving as a proxy for “the Democratic donor base”, in a way. If you ran this with Harris donors, you’d have a similarly massive overlap, I bet. [Yes, on the to do list, but that donation file is *huge*.]

Eight candidates in this set got the *majority* of their donors from this shared pool. I don’t currently have a way to track which candidate was the earliest contribution from an individual (i.e. who should get credit for recruiting this person to donating behavior), so it’s hard to know if these folks are recruiting donors who then make many donations elsewhere, or benefiting from that recruitment behavior. I don’t even have a particularly good picture of lifetime donation behavior, although I think it’s a bit like voting, where doing it once strongly predicts doing it again. This sort of super high overlap was surprising to me. And you’ve gotta keep in mind, this is one quarter’s filings, so almost certainly an underestimate.

Number of donors doesn’t always match up to amount of money raised. It seems like small dollar donors are more likely to be shared than large dollar ones, so the amount of money coming from shared donors tends to be smaller than their percent of the donor population. In this set, there’s just over three thousand large dollar (1k+) donors who gave to at least 2 candidates, compared to 179k small dollar donors with the same behavior. There’s 11 people who gave $1k+ to every single candidate in this set, who must be *extremely committed* to Democratic donations*[Footnote].

That being said, most of these candidates are getting significant shares of their money from shared donors. Pappas has the highest (and I’m not sure why! Joint max-out fundraiser with someone? idk!), and Talarico sticks out with having low numbers of shared donors overall, and resulting low shared $$.

El-Sayed also sticks out, because he has a very large share of shared donors (mostly with Platner), but they aren’t responsible for the bulk of his money. If I squint, this looks like using small dollar optimized lists from similar orgs/candidates and deliberately making smaller asks? That’s pure speculation, I’m not sure.

The Full Matrix

The really cool thing we can do here is look at who has donors in common with who!

You should read row-wise, where the number in each cell is the percent of [row candidate]’s donors shared with [column candidate]. Very sorry if you’re on your phone. You may want to click through for the full interact-able one, this is pretty wide, but it does get far less interesting as you scroll to the right side of the table.

Like I said above, everyone has donors in common with Ossoff. Talarico seems to be fundraising off a different set of people than everyone else (OR, has done something weird with his reporting that is breaking my script, both are possible). There’s a Platner-El-Sayed overlap, where El-Sayed is pulling a lot of his very small $ donors (14k of them), but not that much of his money. El-Sayed seems to mostly *only* have heavy overlap with Platner, so maybe this was a list passed over? That is backed up by the fact that Platner doesn’t seem to be heavily drawing from the Ossoff pool (34k of his 168k donors). You can also see Platner intersections with McMorrow, Flanagan, and Stevens, that I am less sure how to explain.

Note: I started going down the logic puzzle route of “Platner shares X donors with Ossoff and Y with El-Sayed, but El-Sayed only shared Z with Ossoff, so the intersection must be....” and then realized I can *just check*. It’s 9,506 donors giving to Platner + El-Sayed but not Ossoff, for the record, a sizeable chunk of El-Sayed’s 34k.

I also looked into what state these overlapping donors were from. It is California for every pairing except candidates in the same state (i.e. Crockett/Talarico) and Craig/Flanagan (it’s MN, idk). There’s so many Democrats in California and they love to donate.

Share

So What Does It Mean?

Great question. It’s interesting? It means the donor polls for all these Democratic candidates are far more shared than I expected. It could maybe let you look for heavy donor overlap as an indicator of shared fundraising practices or agencies, I guess, unclear if that works.

This also gestures towards donation behavior as an expressive act rather than something particularly tied to a candidate. There’s been some good writing on “rage donating” and you always see giant spikes during dramatic news events. It seems like giving small-$ donations is less reasoned ideological support and more...cheering?

Also, the super high donor overlaps gives me even more heartburn about the churn and burn strategy of small dollar fundraising. If the donors you’re hitting with texts are part of this shared pool, your bad behavior risks ruining it for everyone. Then again, the fact that these donor populations still exist means they haven’t become burned out yet, so perhaps they’re self selecting for liking that sort of messaging? I am skeptical. If you work for a fundraising firm and you have existing analysis on lifetime donor behavior and burnout rates that I can see, pretty please let me write about it, I will leave you anonymous.

I previewed this for a couple folks, and immediately got interesting follow up questions (mostly about expanding the pool of included candidates). If you also have interesting questions, lmk, and I’ll see if I can address them. I make no promises on timing.

Notes and Caveats

As usually happens in FEC data, there’s some weirdness with donations around the max. I have a couple records that look to be over max- I think this is probably something like joint spousal contributions, or maybe oddly reported bundling. It’s not that much money in the scheme of things, so I’m not chasing it down for now.

I have in several places given precise numbers for counts of donors or amount of money. Please understand those all to be APPROXIMATE. I guarantee you I have matched incorrectly or failed to match some donors, deduped wrong, or missed refunds. It is in no way going to match internal candidate records exactly. I’m very sorry if I misstated your number of donors.

This analysis is limited to a set of candidates and a particular quarter’s filing. I have no outside information other than what you can find on twitter.

Addendum: more charts by request!

Here’s the high dollar only version. Higher overlap overall, these folks love giving to multiple candidates. Talarico is still noticably low here, not sure why.

Low dollar- less overlap overall, but some distinct spikes. Big El-Sayed/Platner spike again, this might be “lefty” small donors?

Blood in the Machine and Textile Tech

As such, I’ve gotten really stuck on the way the Luddites and the mechanization of weaving/cloth production live in the modern mind in complete isolation from the actual technology involved. I came to Merchant’s book really late (it came out in 2023), so I have to take him at his word that the popular conception of Luddites at the time was mindless anti-technology smashers. His story paints them as principled labor agitators reacting to being undercut by new machines that could do their jobs somewhat worse but hugely faster. Right now, I mostly see Luddites referenced as being an inspiration for anti-AI or anti-tech agitating. And like, the metaphor works!

To me, what the fixation on this one (famous, well documented) case of pushback against mechanization misses is that 1. This is a thing that happened many times as many different types of labor (relating to textiles, specifically) were replaced and 2. Clothing produced without mechanization was expensive to a degree that’s hard to contemplate these days.

Just offhand (again, not a historian of this), the introduction of the spinning jenny (mechanizing spinning of yarn/thread) caused riots, the framework knitting machine was refused a patent and drove the inventor to go to France about it, and the power loom caused riots. Arguably, the Luddites were starting from a place of high autonomy and high wages only *because* of previous automation (they were framework knitters, on machines that had displaced hand knitting, using yarn created by more efficient spinning jennys). Merchant’s book argues that the output of the new machines displacing the Luddites was lower-quality, which seems to have been true, but people at the time seem to have preferred “worse stockings, but I can afford them” to “nicer stockings, but I can’t afford them”. Personally, I own a number of socks that would have been unfathomable to people of the era (more than 10, even), so I don’t feel comfortable judging them about that choice.

If you’re going to lean on the Luddites as an inspirational example, even if you are arguing more that they should have been compensated/protected rather than arguing the tech shouldn’t exist, I think you have to grapple with what that would look like. There’s some good literature on household inventories over time, frequently including clothes, and it seems like even the moderately well off just didn’t own that many items. It was *incredibly expensive* to buy fabric, because of all the labor involved in producing it. I’m generalizing, but the amount of clothes in a normal wardrobe today would be shocking to most people in history. Things were mended, passed down, repurposed, and taken care of to an enormous degree. There are items in museum collections that were remade into new styles some 50 years after the original construction of the item.

Share

Fix your clothes

Anyways, this returns me to my original point: if you’re going to be using the Luddites as an example, and thus clothing manufacture as an AI parallel, then you should probably learn to mend your clothes. Pretending for a moment that it’s hideously expensive to get a new coat, you should know how to reattach a button or replace a cuff. I suppose for true authenticity to my point, I should be talking about darning socks, which you should also try (although you’ll have a hell of a time with the modern super tight knit elastic-y variety). Re-sewing rips and torn hems is probably the easiest thing to learn, although making it look nice is a challenge. This is especially handy if you’re the sort of person who catches your clothes on things all the time (not that I know anyone like that). It’s a useful set of skills, I think it’s fun, and it’ll let you hang on to your stuff for longer.

It is possible, even likely, that you’ll try this and think it sucks and is boring. That is maybe a little the point- this was a thing that people DID until not all that long ago, even when it sucked and was boring. (Nitpickers may point out that there used to be an industry of people who would do these things for you, for relatively low prices, due to the abundance of cheap labor. I think the point stands.) I think it is healthy to understand what would have been necessary to exist in an environment before ultra cheap clothing manufacture when we’re using it as a casual point of comparison to our modern developments.

Addendum

Do I have a point here? This may be an oops all vibes post? I started writing a very detailed point by point account of why I didn’t like Blood in the Machine, but half of that was “it doesn’t actually care about the technologies in question”, which is only a problem if you happen to think those technologies are neat. The other half was “the author is wrong about some things I noticed”, including the irritating lazy thing of confusing life expectancy with how long an adult could expect to live, and a point about where trousers came from. While those things DO make me wonder if he’s wrong about things I can’t easily verify, I don’t actually know, and I didn’t really care to dig and find out. I’m fairly sure my objection is mostly, as described above, that he doesn’t grapple with the knock-on effects on the entire population of clothes-wearers. It’s an interesting story, don’t get me wrong, but I kept coming back to “okay, and the alternative is…..??”

Instead, here’s a sneak peek of what I’m working on! I’m (as always) super interested in FEC data, and trying to chase down a particular theory about the asymmetry between donors and voters. We know that donors cluster hard in Democratic strongholds, and that they tend to donate to multiple campaigns. We also know there’s huge incentives to cultivate that national donor base. In theory, we can get a ton of detail on this from the FEC, but the details are tricky.

As a proof of concept, I grabbed donations from the October 2025 reports filed by Cory Booker and John Ossoff (since they are the top 2 fundraisers are reported by the FEC). I’ve also mapped them geographically, since the FEC helpfully provides donor details.

Here’s Ossoff:

and here’s Booker:

You may notice that these are basically “home state, California, New York”. This seems to be true for ~every candidate. From initial results, that also seems to be the case for Actblue data (i.e. donations under $200 total), but mostly because a lot of large dollar contributions come from candidate home states. The distribution of number of contributions is basically the same in itemized vs unitemized. I have zip code data as well- we will see if that looks different! Maybe there’s hotspots in CA that cluster around different candidates? Idk.

The next thing I’m working on is figuring out how many of these donors are literally the same people between candidates. The universe of donors seems under-studied. I don’t have an amazing sense of the actual size of the group we’re talking about, and candidates have every incentive to exaggerate their numbers or only discuss number of donations rather than unique donors. Deduplication and matching are hard, but I think I can get a good-enough answer.

More later!

Share

At best, I find this kind of messaging very trite. At worst, it makes candidates say things that are obviously false if you actually think about them. Jasmine Crockett had one of the latter that annoyed me enough that I went and dug it back up.

https://x.com/grace_panetta/status/1998175192657404043

I want to be clear: Jasmine Crockett isn’t the only one to take this line of reasoning, but she’s the most recent, and highly public.

Is Texas a red state?

Yes. Trump won it by 13.7% in 2024, which was *up* from his 9% victory in 2016. Ted Cruz won by 8.5% in 2024. A Democrat hasn’t won statewide office in Texas since 1994. This is, by all metrics, a red state. That’s not like, a dig! I think some people understand calling a state “red” to be derogatory, or a way to write it off. “Red state” is a description of the voting patterns of the state.

Crockett is not the first or only one to say something like this about states not “really” being red. It’s usually deployed as a sort of “we can win, rah rah” thing, and probably the literal truth of it doesn’t matter to most people hearing it. But it’s still wrong.

The logic behind her statement goes like this- there’s enough people in the state (either registered as Democrats, or who might be Democrats if they registered) that if they all voted, they’d outnumber the Republican votes cast last cycle, thus the state is “actually” not a red state.

Friends, this is every state. *Of course* if Democrats voted and Republicans didn’t, Democrats would win! That’s how numbers! Most elections have turnout under 70% of the registered population, and a far lower proportion of the eligible population. That’s a lot of people who didn’t vote. You would have to be losing *so* badly for them to not be numerous enough to technically make up the margin. This does not work in practice.

Subscribe now

You can go down a fascinating rabbit hole in academic lit about the thing where election results by geography are relatively stable, even as the voting population fluctuates (I’ve seen estimates of something like 20% churn of the voter rolls each cycle). This is an interesting problem. What *isn’t* interesting is wishcasting about how you’d totally win if loads more people voted for you and not for the other guy.

Dreaming of never changing anything

My general position is that turnout vs persuasion is a false choice. Not only is it always both, but they’re correlated. Do really good persuasion, people like you more, they turn out more. The idea that you can “just” do one or the other is fake. Crockett’s comments are symptomatic of a turnout-only strategy (or at least, the impulse to tell primary voters that you’ll run a turnout-only strategy).

Also, the idea that you can run a campaign in a state like Texas and not even try to win over Republican voters is ridiculous. I know this is primaryslop, but come on. I know you won’t, you know you won’t, why are we playing this game. I’d take a perfectly good non answer about how they’ll see the strength of your vision or whatever over this.

I see the rhetoric of “this isn’t really a red state” as part of a broader mindset that treats Democratic failure to win elections as something imposed by outside forces. It’s not that Democrats need to do anything or change anything- if things were fair/perfect we’d already be winning, and we should fight about that fairness instead. To me, this feels like a very conspiratorial and complicated explanation for election results. The cleaner, more accurate answer is that some people just do not agree with Democrats. Many (not all) of these people can be persuaded! Some of them disagree on vague sketches of ideas they half-heard on tiktok, and can be talked out of that! But they’re real people, with real opinions, and not just pawns of some mystery machine.

It would be convenient if you could win by just shifting around the population of voters like they were checkers. That would be more comfortable than trying to appeal to them- and it plays really well with primary voting Democrats to pretend that absolutely zero compromises have to be made in order to win. It’s also a hideously undemocratic impulse that we should resist.

I don’t know what the perfect way to do persuasion in Texas is. Honestly, I mostly think of it as a state where money goes to be lit on fire. The dream of a blue Texas has been “a couple years away” for many years. This is going to be a good midterm for Democrats (for reasons of thermostatic shifts and also Trump being bad at governing), so we’ve probably got a shot, and there’s some interesting infrastructure work going on that may help. I would love to see it happen. And I am going to be watching as whoever wins the primary pivots to a general election message- I can just about guarantee it will be different from what they started the primary with.

Being a huge nerd, I have an interest in history, mostly historical sewing and lace making. A while back, I picked up the book Punch from a local bar supply store on a whim, and really enjoyed it. I’d recommend it if you like cocktails, history, or both.

The main thing I learned from the book is that historical punch recipes are *wild*. Massive portions, alcoholic as hell, often alarmingly sweet. One of the stories involves an entire fountain full of punch. Googling about “fish house punch” led me to the Mount Vernon website with its extensive collection of recipes, and that led me to make something called “Tavern Rum Punch” for Thanksgiving a couple years ago.

It quickly became obvious that 1. this punch is extremely good and 2. this punch is EXTREMELY ALCOHOLIC.

Here’s the original recipe:

Tavern Rum Punch, from the Mount Vernon Tavern

• 24 oz boiling water

• 1 bottle George Washington’s Rum

• 1 bottle George Washington’s Apple Brandy

• 8 black tea bags

• 4 limes +2 limes

• 2 lemons + 2 lemons

• 3/4 cup of cane sugar

• 6 whole cloves

• 2 cinnamon sticks

• 1 tbsp whole allspice

Directions

1. In a large pitcher (something you can store in the fridge) combine the sugar and boiling water and the teabags. Steep the tea for 10 minutes and make sure the sugar has dissolved.

2. Remove the tea bags and discard them. Add the spices straight in followed by the rum and the brandy. Stir this mixture gently.

3. Next, add the juice of 4 limes and 2 lemons. (It’s okay to get seeds in it, we will strain them out later.) Stir the juices in and refrigerate overnight or 8 hours.

4. After the time has passed, remove the mixture from the fridge and strain the mixture into a secondary pitcher or punch bowl. Garnish with the remaining citrus fruit cut into slices. Enjoy!

They claim this serves 5-6. This is only true if those 5-6 wish to crawl home.

Share

My Errors

If you are me, and not paying that much attention, you read that recipe and purchase a bottle of nice rum (whatever was dark-ish and a medium price) and a bottle of apple brandy (I think I got Lairds) at your local liquor store. These bottles would be a fifth, or 750ml.

This is not what the recipe called for.

After investigating why we were all quite so intoxicated, I learned that the bottles specified in the recipe and sold for many dollars on the Mount Vernon website are 375ml each. This makes the punch more like 1-1 alcohol to mixer, rather than the 2-1 concoction I made.

The good (?) news is this ratio is still delicious, but like, maybe don’t.

The Recipe So You Don’t Make My Mistakes

Tavern Rum Punch, larger and simpler

• 42 oz boiling water, or about one and a quarter liters

• 1 bottle nice dark rum (750ml)

• 1 bottle apple brandy (750ml)

• 12 black tea bags (not Lipton’s, something nice)

• 8 limes (or cup of lime juice) + 4 limes if you want to decorate

• 4 lemons (or half cup of lemon juice) + 4 lemons if you want to decorate

• 1 1/2 cup of sugar

• 6 whole cloves

• 4 cinnamon sticks

• 2 tbsp whole allspice

• 1 tbsp nutmeg, for that “we just learned about spices” taste

1. In a heat proof container, combine boiling water, sugar, and tea bags. Steep 10 minutes and stir until sugar dissolves. Remove tea bags.

2. Add spices, using a teabag to contain them if you only had ground spices. Let sit 5 minutes while mixture cools down a bit.

3. Transfer to larger contain that you can fit in your fridge. Add rum and brandy. Add lemon juice and lime juice, stir to combine.

4. Refrigerate overnight or for however many hours you have time to, while leaving the spices in. If refrigerating more than a day, remove the spices at roughly 12 hours and it’ll stay fine.

5. Serve, probably with a warning about it being caffeinated if you hang out with people who mind that sort of thing.

You can in theory just halve this, but I find it really annoying to use half a bottle of something I bought specifically for a recipe. You do you. If it’s still too strong for your guests, make more tea and use that to water it down.

Happy Thanksgiving, I hope you have only funny accidents during your dinner and festivities, in whatever form you’re having them.

Personally, I adore state legislatures. It’s a whole pile of elections, with extremely different dynamics, and many of them are so tiny that it’s all kind of adorably local. As an election forecaster, I am damn proud of the fact that for a bit there I was running a model predicting all five thousand ish state leg elections, and the results were pretty good. It was a huge logistical nightmare and I now know so many things about multimember districts and the hated floterial district. I take seriously the idea that they are laboratories of democracy, and a place where early stage campaign tactics and strategy can get tested out.

Here’s my pitch: we, as Democrats and people interested in Democratic elections, are not getting nearly weird enough with it. Local races are (relatively) cheap to campaign in, there’s a ton of them, most of them are safe districts or longshots. I will be on here hassling you to donate to swing district candidates and defending their need for professional, competent campaigns and adequate funding, but for all those +30 / -30 seats? Try some shit! Take whatever strategy you have cooked up online or in a group chat, find a plausible candidate, run a campaign where you communicate entirely through puppet theater, I don’t know. See what works.

State legislative candidates are already pushing the bounds of weirdo scandals. This is part of why I love it: we can theorize plenty about what exactly WOULD happen if a candidate got caught pouring water into someone else’s backpack but now it’s happened. We will know if the water pouring thing mattered once she faces reelection in 2026, btw. Candidates are already going to make odd choices. We should pair that with the longest shot campaign strategy we can muster.

To some extent this is tricky because the sort of people who want to run for office are nice, serious, and want to improve their communities. They don’t tend to be in it for the zany antics. Part of the job of campaign strategists (and comms people, of which I am not one) is to get our candidates out of their shells enough to show the world how cool they are. In the scheme of things, I’d guess it’s harder to get some candidates to do direct to camera videos than it would be to have them do Zoom calls from a canoe while fishing. Maybe what your district needs is an entirely row-boat-based campaign. There’s essentially no downside here. If you’re running in an R +20 district, that marginal 10 votes from a careful, ordinary campaign is just not going to matter. It should be hail-mary all the way.

On the RCT nerd side, we should be using these races to *test stuff*. One of the most difficult parts about running RCTs on campaigns is getting people to agree to have a control group. No campaign or IE I’ve ever worked with has been chill with the idea that we have to leave people out of our treatment group and potentially lose those votes. I have a whole speech about how that’s mathematically fine and that 2 votes you might be leaving on the table won’t matter. But in a long shot state legislative election, the odds you have funding to get your treatment to every single possible target are slim to none. This scarcity is the original condition for an RCT- if you can’t target everyone, randomize your targets so you can learn. I’m not inventing an original idea here- there’s already structures in place to help campaigns and groups run RCTs- but I want us to do it more, and with higher uncertainty ideas. I love the 100th RCT on mailers, but I would even *more* love the first RCT on leaving paper cranes at voters’ doorsteps, or on putting canvassers in bear costumes.

The salty part of this proposal is that it would create a prove-it zone for commentators. If you’ve talked a big game about how your one weird trick will make Democrats win, you should be shoved towards a R +20 district and told to go to town. If it works, amazing. If it doesn’t work, no real harm done. We should not be trying our newest hottest shit in competitive races. There are an absolute PILE of elections where you can safely experiment (and, ideally, build the party brand) without giving me heartburn about a competitive House or Senate seat.

Democratic campaigns tend to look pretty similar. Nationalization and the big national Democratic brand come for us all, but we don’t HAVE to run traditional campaigns. We can just do...whatever. Data-wise, it would be a huge help to expand the standard deviations so we aren’t all just arguing about the marginal advantage of an essentially-ordinary Democrat who said tariffs were good as if that was the maximum imaginable deviating posture. Do it for the modeling. What happens if you run zero TV ads and go all in on TikTok? If I had to guess, it’s not great, but I *do not know*. Find out! This playing it safe stuff is a path to barely eking out a majority occasionally and then having most of our policy thrown out the next cycle. It’s a path to increasingly perfect mailers that go immediately in the trash. We’re stuck in a local maximum and unwilling to hunt widely enough to get to the real high point.

So, please, try some shit. Give people money to try some shit. I bet there’s someone who would run a hell of a state legislative campaign if you dropped 50k on their heads and said “try something funky, let’s measure how it goes”. And if you’re talking online about how everyone would win elections with your cool new idea, in all seriousness, local elections need you. Find out if it works. Go, be free, write us a fun report afterwards.

TLDR: Send me a receipt for a $50 donation to your local food bank or to Share our Strength, and I’ll send you a book recommendation. You can add a note about what genre or type of thing you’re interested in, if you care to, or take the random option. Book quality not guaranteed, book lack-of-trashiness not guaranteed, read at your own risk, etc, etc. Replying to the newsletter email works, DMs work. Okay, now for the post.

From this blog’s occasional dips into book reviews when I don’t have a better idea, you can probably work out that I’m a pretty constant reader of fiction. Often not very *good* fiction, more of a broad meandering swath of whatever I can get at the library. What may or may not be obvious is that I’m extremely, extremely committed to sci-fi as a genre. I really like it, I grew up on it, I enjoy knowing things about the genre and discussing various forms of it. (at one point, my fun ice breaker presentation for a work thing was “Male Sci-Fi Authors and Where They Should Have Stopped” which I stand by entirely, looking directly at you Larry Niven)

Sometimes people who are incredibly into sci-fi and also into tech can talk a big game about how “science fiction predicts the future”. This is mostly garbage- for every oddly dead-on book there’s a pile of wildly off takes that we’ve simply forgotten1. It’s more likely to tell you about the present anxieties of whoever wrote it than anything else.

One thing that older sci-fi CAN do is let you pinpoint exactly what book someone has pulled their clever new idea or vision of the world from. Nerds of a certain age are extremely likely to have read many of the same books, usually as teenagers (the golden age of sci-fi is about 13, as the joke goes). Some of those nerds are going to try and build the torment nexus. These books vary on if they portray their torment nexus as bad or not. Even with books that are HEAVILY criticizing something, there’s a tendency to half remember it years later and go “yeah well, what if we did that but good?”.

Personally, I think understanding these references is useful as a psychological insight. My area of interest is written sci-fi, but I’m sure there’s a lot of “one episode of Battlestar Galactica” or whatever getting referenced that I’m missing.

Assume these are not recommendations unless specifically listed. Many of these books aren’t good. Without further ado, here are some topics you may wish to understand the references for, and the books I think inform them.

Libertarians and Mars Colonization Weirdos

The Moon Is a Harsh Mistress, by Robert Heinlein2. I am not recommending you read this, it’s not very good, it’s uncomfortably racist, weird about women, and also at one point fully stops the plot to talk about the optimal size of terrorist cells. However, it’s a super important reference across a variety of ideologies.

People often describe Elon Musk’s mars colonization stuff as coming from Asimov’s Foundation (he himself has said this), but I think this is a much more likely source. The entire point of the book is a small collection of people on the Moon entirely outclassing Earth by being smarter, tougher, harder working, and above all that boring earth shit like “social safety nets” and “non-mob-based justice”3. It’s also one of the foundational texts of libertarianism, and spends a great deal of time on it4.

I’m biased because I really do not like Foundation (despite loving other Asimov works), but I think it appeals to a certain type of guy who thinks he’s smarter than everyone and 10 steps ahead. It doesn’t really provide models for actually doing anything other than 1. Go to space 2. Be really smart 3. Create a nuclear-reactor cargo cult religion. It’s a model if your goal is for everyone to venerate you specifically after your death. Moon spends a lot more time extolling the virtues of the specific governing and lifestyle decisions that made people “fit” to colonize the moon. Ick, but useful to understand.

Subscribe now

Reproductive tech and screening ventures

Ethan of Athos, by Louis McMaster Bujold. Bujold is clearly really interested in future possibilities for reproductive tech, specifically artificial wombs and how they’d impact childrearing. The Vorkosigan saga spins what is very clearly Star Trek Federation x Klingons fanfic5 into a very fun combination of palace intrigue plot and mercenary company nonsense. Ethan of Athos comes out of left field with an oddly touching story of a planet that literally bans women (religious reasons, because they’re corrupting and demonic, which #goals) and relies on artificial womb tech for its population. I know this sounds bad, it’s actually kind of sweet. For the full picture, you’d have to pull from the other books which touch on the possibilities for things like embryo selection and genetic tweaking, but the use of “uterine replicators” is a constant in the books.

I find her work interesting particularly because it’s somewhat about the way that the available technology changes the relationships between men and women, and alters the cultural understanding of who gets to be a family. Later books deal with the way replicators are changing the uptight imperial society, and their impacts on the lives of women. IMO, you can see clear pulls from Athos in the way Sam Altman especially talks about reproductive tech.

Election Forecasters and AI

Stay with me here. “Big smart AI that can predict everything” is a common enough theme, with less tech-y roots in Laplace’s Demon (which is also the name of software for running Bayesian analysis as I literally just learned). There’s enough references that it’s hard to pin down. My favorite (and I think a likely enough one) is the Multivac stories from Asimov. These stories are specifically cool because they’re an early enough treatment of AI that they conceive of it as a huge calculator, rather than a being with a personality. Multivac explicitly requires data on all things happening on earth to make its decrees and predictions. Most of them are kind of ~fine, but that era of proto-AI is imo influential in how people think about forecasting and understanding human behavior.

Specifically worth your attention is a single Asimov story called “Franchise”, in which the single most representative voter in the US is called in to do his civic duty of answering baffling questions so a supercomputer can calculate the outcome of the election. No actual election is held, because the computer is so accurate that what would be the point? It’s set in 2008 and the only specific question we are given is about the price of eggs. If you wrote this now, people would be calling it a bit on the nose. It’s dated, obviously, but it hits on the pull to perfectly predict behavior so well that there’s no need for the behavior to like…occur.

I don’t actually have evidence that our election forecasting nerds of the moment have read this, but I am one, and I certainly read this, so I’m counting that. You should go read this one, it’s short and available online.

Share

The Goddamn Phones

The Naked Sun, Asimov, second book in the R. Daneel series. Matt Darling made this point on Twitter in response to some truly awful “AI friend” advertising and I think he’s right and want to add on. The first R. Daneel book, Caves of Steel6, is more about super dense future cities and their resultant agoraphobic (or, claustrophilic, a word I just learned you could use) population. I adored this book as a child (I was very fun and normal) for its synthesized yeast-based foods7 and cities with tunnels full of moving walkways. Naked Sun takes you off Earth onto a Spacer world where everyone stays in their extremely far apart houses, hangs out via video call, and can’t stomach the thought of being in a room with another person IRL.

Because of when it was written, Naked Sun sees the alternative to human contact as “calling your friends on video calls”, which is far LESS cynical than current possibilities. I think we’d be pretty happy if the Youth were on facetime constantly and not on tiktok. The other part of this vision of the world I think folks are pulling from is the idea that the great robotic replacement of all human labor will free us up to...scroll our phones hologram video screens. I find this hideously depressing, but the vision of work-free personalized entertainment in our own homes seems hugely attractive to people. I’ll keep my knitting, thank you.

My favorite eerily-accurate take on The Phones is “The Machine Stops”, a 1909 short story about a society controlled by The Machine. It perfectly captures what it’s like to be so busy with your phone that you barely have time for people who you actually know.

Niche beef: Dune

One thing that is NOT a useful way to understand the present is Dune’s Butlerian Jihad. I know this is a thing like 5 people care about, but the Butlerian Jihad is a setting shorthand to let Herbert write a futuristic novel in which they have sick knife fights8 and aren’t using robot armies. The backstory of it is mostly made up in later novels while fishing for material, and it’s dumb as hell. It is also notably not a reaction against normal-ass computers or robots taking your jobs. I am begging you to stop referencing it.

Fundraising reminder: Send me a receipt for a $50 donation to your local food bank or to Share our Strength9, and I’ll send you a book recommendation.

Photo by Giorgio Trovato on Unsplash

Metapoints on Money

The Democratic party, especially the more progressive side, is pretty leery of the general concept of “money in politics”. Groups like End Citizens United are pushing for limits on outside spending (generally IE spending), and Democratic candidates periodically pledge to reject certain types of donations, or tout their small dollar fundraising credentials. This is all at a time when Democrats have raised staggering, record breaking quantities of money, especially for national races.

Cards on the table, I think much of the anxiety about money in politics is pretty silly. Yes, it’s a bit weird that we light this much money on fire every cycle, but are other likely uses for that money any more virtuous? Would we rather it sat in rich people’s pockets? I haven’t seen much evidence of “corrupting effects” of taking donations, other than a strong, strong bias towards “anything that keeps getting us donations”. This doesn’t usually mean switching policy views, just doubling and tripling down. Which is, to be clear, bad and annoying, but would still happen if you halved the overall $$.

A somewhat-formative piece of literature for me was the Ansolabehere paper on “Why is there so little money in US politics?”, which (in addition to having a very decent overview of campaign finance rules) walks through the logic of: if spending money in politics produced any sort of clear financial benefit, wouldn’t there be *way more* money in campaigns? I.e. if you really could buy certain outcomes, and it was an efficient investment, wouldn’t it happen way more? This paper frames political spending as an expressive good, which matches what I see happening in practice. Donations are mostly on vibes and a desire to look like a good person, not because specific outcomes are expected.

Share

Virginia

Virginia’s campaign finance laws are approximately “go for it”. There’s no limit to contributions to campaigns, from either individuals, PACs, parties, anything. This includes in-kind contributions, which can be things like running mailers for a candidate, handing over polling data, or other varieties of goods and services. Normally, you’d need to strictly avoid coordinating with a candidate and keep your spending classified as “independent expenditure”. In a state with no limits, you don’t have to worry about that.

This absence of spending limits means that you can meddle in the VA elections extremely freely. You could go drop a million dollars into a candidate’s account tomorrow, if you cared to (and had a million dollars to spare). VA is an outlier in this. Eleven states (Alaska is in court about this) allow unlimited individual contributions, nineteen do not restrict party spending, and seven allow unlimited PAC spending (see the NCSL for more). Those states mostly don’t overlap, although I haven’t been able to pull together a nice map of this. Point is, VA is *weird*.

Anyways, it is possible I’m hideously wrong here and political spending is secretly corrupting us all. I raise the Virginia example, because *if* political spending has negative effects, you’d expect to be able to see them in the most go-hog-wild off-cycle state.

So, what do we actually see?

What do we see in VA?

If spending in VA politics is meaningfully skewing outcomes and governance in the state, via something like corruption, bribery, or just weird incentives, I would expect:

1. Outsized spending, particularly in less competitive races

2. Altered election outcomes, reflecting different patterns from the rest of the country

3. Tax policy favoring the very wealthy, and corporations

4. Labor laws doing the same

5. A limited social spending regime, or low state spending overall

6. Looser regulations, particularly on safety issues

I’d also expect that VA would be notably different from comparable states in a way we could attribute to that spending.

Let’s go through those points.

Outsized Spending

This is going to be a pain to measure, particularly because VA runs Gov/State Legislative races in off years when everyone has money to burn and not much to do. Also, state legislative spending data is a disaster. NJ is the easiest comparison, since it runs on the same cycle.

It looks like, in 2021, about 59 million dollars was spent on the races for the NJ Assembly. [The best source I could find here was a whitepaper by the NJ Election Law Enforcement Commission, link here] The Assembly is the lower house chamber, with 80 seats. For VA’s House of Delegates (100 seats), in 2021, it looks like about 84 million was spent [thank you VPAP]. That is more! It’s like 730k vs 840k per seat. This isn’t a precise comparison at all, but I would treat that as notably more spending.

Is that spending going anywhere unusual that might indicate non-competitive purposes? Hard to tell, but doesn’t look like it. VA has really great disclosure laws, and at a glance it doesn’t seem like sure-thing candidates and races are getting unexpected piles of money. You can also see if business donations are a large part of campaign spending, and it looks like not really.

So, outsized spending: ehhhhh, maybe a little

Altered Election Outcomes

Do VA’s election trends look really different from the rest of the country?

Not really. VA has gotten increasingly blue as education polarization has set in, and college educated white voters have swung increasingly towards Democrats. The state has gone back and forth between Democratic and Republican governors, and control of the state legislature has been tightly contested (tbt when they literally drew a name from a hat). It also doesn’t seem to have any huge gaps between Presidential results and local results that you might expect if local votes are super skewed.

There’s two possibilities here. One, the money influence is roughly split evenly by partisanship, so it doesn’t drag election results in any given direction. Two, it just doesn’t matter that much.

Altered election outcomes: nah

Tax Policy

Absolutely NOT my area of expertise, but I found this website from the “Tax Foundation” that ranks states by tax policy (which is to say, they highly rank states with low taxes). I disagree with their fundamental conceits, but as far as I can tell they are reporting numbers accurately. The wild disagreement also makes their website a very funny read: high taxes, the horror!

This site ranks VA as 28th in their “state tax competitiveness index”, which is right in the middle. The index rates states with lower taxes higher (Wyoming is 1st) and high taxes, lower (NY is 50th). It seems like their sales taxes are on the low end, income taxes are a bit higher, and corporate taxes are basically in the middle. There’s no estate or inheritance tax, which is a win for the very rich, but those things don’t seem to be particularly common, so I can’t ding the state too much.

[PS, I am lightly haunted by this section in their writeup:

“Tax competition is a little like WAR—not conflict, but Wins Above Replacement. The term comes from baseball, where it is intended as a sabermetric statistic to measure how many more wins a team can claim due to a specific player above the amount that would be generated by a replacement-level player. It’s much the same way in public finance: a well-structured tax code won’t make the Wyoming Basin a metropolis, nor will poor tax structure make Manhattan a ghost town.”

It really is the same everywhere, huh?]

Tax policy: basically normal, I think, probably

Labor Laws

VA has a right-to-work law, a $12/hr minimum wage, and otherwise pretty normal labor rules, per Oxfam. It looks pretty dang good when compared to its southeast neighbor states. It gets a B from the Commonwealth Foundation on labor laws overall.

You could argue that VA is somewhat behind very blue states in terms of labor laws, but it also looks wildly better than NJ. If you compare to the also very competitive PA, VA looks better. I’m not too surprised by the right-to-work law, given how southern the state is.

Labor laws: no effect

Social Spending

According to the Urban Institute, completely normal. Per capita public welfare spending looks on the high-middle of the range, especially using the per-low-income-individual metric. The state expanded Medicaid in 2019, with a trigger law that will cut coverage if Trump cuts Medicaid funding.

Social spending: no effect

Regulation

Absolutely all the sources on this are right-leaning, and some of them seem to be literally counting the number of times the phrase “shall not” appears in state law. Is this really representative of burden? I have no idea. Presumably the regulatory state also varies a lot by topic, but I’m not clear what topics I should try to use as a model, or really how to do that.

As a very, very initial cut, the Mercatus report on regulatory burden puts VA in the middle on “number of restrictions”. Again this seems like a really odd way to measure anything, but it’s what I’ve got. There’s also a rating of states by theoretical friendliness to small business, with VA again in the middle. I never realized just how many websites there were handing out “bad grades” to liberal states, lol.

Regulation: no effect

Share

Is that an impact?

I mean, this is not a greatly detailed investigation. I haven’t really touched the specifics of donation behavior, or who the major donors are. There’s also the entire blob of whatever is going on with Dominion Energy, which is a major VA donor and also utility, which may have specific political asks I am simply unaware of.

But from the metrics I set: not really!

Along the handful of measures that I expected to be skewed or altered by unlimited campaign spending, VA looks pretty dang normal. It looks roughly like what you’d expect from a blue-but-competitive state. There’s a bit more spending than in other places, but not a huge amount. It’s all kind of.....meh.

*If* there is an effect of the bonus money, it’s awfully muted. You can argue that the competitive nature of VA mutes the effect, or maybe the strict disclosure requirements do. But if those things are enough to stop the negative effects of money, that money isn’t very powerful in the first place.

You could do an expanded version of this analysis where you try and correlate spending limits and dark-money-volumes with state policy, or trace particular VA spending to outcomes/actions by legislators. There’s a lot more that could happen here. My meta-argument, however, is that such a corrupting effect should be *obvious*, which it sure isn’t.

Maybe money in politics doesn’t actually do all that much.

Did you know there were elections coming up? Living under a rock sounds so peaceful sometimes.

First, Some Graphs

I went and grabbed recent statewide races in both places. Everything is converted to 2way Democratic vote share for convenience.

Here are two graphs on which you could hypothetically draw trend lines. Probably don’t, because they contain extremely few data points.

This is what I pull up in my brain when asked about these two upcoming elections, and so I have helpfully pulled it up in graph form for you.

So, looking at that: Are the Democratic candidates in these two elections going to win?

Maybe, but I wouldn’t put money on it.

The Logic

Biden won both states by a good margin, and then the subsequent Governor elections went far worse for Dems. However, that was during COVID, and a Democratic administration trying to handle the absolute chaos of 2021. However however, both states saw worse results for Harris than for Biden. But Harris didn’t win, which means that a Republican administration is running the current federal chaos, and that Republican administration is doing a terrible job. Neither state has an incumbent governor running.

Also there’s a government shutdown happening (which may or may not still be happening at time of posting), VA has a lot of exposure to the federal workforce, NJ has exposure to the now-threatened NY transit system, both of them have exposure to things like “planned federal grants”. I’d be a little surprised if the shutdown lasts until the actual election, but it could. It’s also extremely possible that more DOGE-type measures or impoundments come down as a result of the shutdown. We just have no idea right now.

Oh, and also Virginia’s GOP lieutenant governor candidate seems to be having a scandal involving his Tumblr use. Will that matter for the Governor election? Unclear!

In retrospect, you could read the 2021 results in NJ/VA as indicative of the unpopularity of the Biden admin, or even as a portent of the 2024 Harris loss. At the time, my vague memory is of people going “damn, folks are really mad about COVID, huh”. There’s a lot of posting (poasting?) trying to get narratives in place to explain a win or a loss. Given how the explanations of 2021 shifted and changed over the next cycle, that seems premature to me. Currently, I’d be inclined to read worse-than-expected results here as a bad sign for Democrats, but I’d find it difficult to pin down exactly what that sign meant. Ideally we’ll learn more by specific demographic breakdowns of the results! Once we, you know, have those.

An Informed Guess

If I had to guess, I would probably say that Sherrill is going to lose and Spanberger is going to win narrowly. Let’s say like Spanberger 52% in the 2way and Sherrill 49%.

NJ was one of the states where we saw markedly worse Democratic performance in the recent presidential, and seems to be one of the historically deep blue places that Democrats are starting to struggle. I’m guessing that trend holds up in this election. Combined with the state’s apparent distaste for electing Governors of the same party as the previous Governor, it seems like a rough path for Sherrill.

The logic for a narrow Spanberger win is similar. VA held up more in the presidential, and it’s a relatively college educated state (a population Dems have continued to do well with). It’s also a state that is positioned to care a lot about Trump’s interference with the federal workforce, and the shenanigans of folks like Hegseth due to the high military population.

I should note that this is *not* what the current polling says. The little polling that exists seems to have Sherrill up, and Spanberger roughly tied. That polling is super sparse, and given the miss in 2021, I’m not inclined to lean on it. I’m maybe being contrarian here and betting against the direction of the polls in NJ.

I could be wrong if: I’m overestimating the NJ trend. Murphy did win in 2021, albeit with difficulty, and both the presidential and senate elections in 2024 outperformed that race. It could be that NJ has leveled off after the steep 2020-2024 drop, and is going to stick around as a state where Democrats pull in 53ish%.

I could also be wrong if: It ends up being a really strong Democratic night overall. Maybe the lower turnout of an off-year election combined with anger at Trump drives Dems to the polls and both governor candidates win by a mile.

Why Write This Post?

Part of what I want to do with this blog is use it as a place to put things on the record. One of my biggest frustrations with forecasters (and pundits) is the rapid retconning of their opinions when they’re wrong. I’m wrong a lot! That’s fine! I think it’s healthy for me to write down even approximate guesses, as a measure of accountability and to promote the practice.

Also, I have found that people are interested in the logic of election forecasting. While complex models can help, sometimes they obscure what’s going on under the hood. You *could* stick all of the above data plus polling data into a model. I’m not sure you actually need to. In this data-first age,

Addendum

I was recently at a conference where folks suggested that people with high levels of political knowledge should be betting on that knowledge in prediction markets. I will not be doing that, because I don’t really like gambling. However, if you use the info from this post to bet and you win, you owe me 20% of your winnings, I take venmo, thank you for your time.

If You're Reading This It's Meant For You

Gothic novel set in a legally-distinct-short-form-video-platforn hype house. It's better than it sounds. If you're a very strong fan of the genre it might be a little cliche, but fun.

Vita Nostra, A Novel

This is “books I am currently in the middle of”. How is this book so tiring to read? I'd strongly recommend it if you want to have a weird time, or if you've ever been like “hey, what if the currently everpresent magicschool book was a Russian (Ukrainian) novel?” I'll probably have more feelings once I finish the damn thing.

The Helios Syndrome

What if a necromancer worked for the NTSB investigating plane crashes? What if that? Extremely what it says on the tin.

The entire ACOUP series on pre modern peasant farmers

Link here. I love some pre modern social history, and ACOUP is really good at giving it texture.

Before They Are Hanged

Sometimes I read grimdark when I'm stressed

Exciting personal news, I have a new job and am firmly back in the world of polling. I’m the Director of Analytics at the newly launched Searchlight Institute, and I am absolutely delighted to have a mandate to run a bunch of thoughtful and experimental polls for the purpose of finding things out about the American people. Plus, I’m going to be releasing the entire questionnaire for every poll we publish, in the interest of getting more high-quality interesting polls out into the public. I’ve got a whole set of polling principles written up if you want to hear more about how I’m planning to run things there.

Highly related to the new job thing, I am having a ton of conversations about polling with folks, and spending time explaining why I care about it. Kinda in my feelings about it, tbh. So here are some notes on that.

I really, really care what the American people think. I mean, I care what people across the world think, I just specialize heavily in American politics so that’s where my focus is. The reason I went into survey research in the first place is because I was fascinated by human behavior and especially the way people formed opinions and beliefs. There are so many people walking around with completely different things going on in their heads, and yet a large chunk of them manage to form opinions on the issues of the day. And you can ask them about it! They’ll just tell you! They’ll get things wrong, they’ll misunderstand what you’re asking, but they will just tell you stuff. That’s really damn cool.

I also think it’s moral to stay interested in what average people think. Not because you’re looking for approval (turning a big dial, etc), but because fundamentally the broader public is who you’re doing politics for. There’s tension between the delegate and trustee models of politics, but I really think that assuming you know what’s best for people runs you into trouble every time. People have wants and needs and feelings about the direction of the country. Those things won’t necessarily be coherent (especially on policy!). They are still real.

Sometimes maintaining this interest is hard. I’ve spent a lot of time reading survey open end question responses and man, some of those are *not great*. It’s depressing when you run questions and get back results that seem to show cruelty, disregard, or outright hatred. It sucks even more to report out those results. But I don’t think it serves anyone to pretend those things aren’t real, or that they aren’t present. At its best, opinion research can help you figure out where those opinions are coming from, what’s driving them, and what points of connection and compassion can help you reach people who hold them. At minimum, it gives you a better idea of what you’re dealing with.

Polling helps you reach people you normally wouldn’t talk to or hear from. Doing that reaching is *really hard*, as the evolving state of polling methods would tell you. I think it’s absolutely worth it for the chance to figure out what is going on in people’s minds. Are they like, doing okay? (sort of, but not as okay as they’d like.) Are they content with the direction of things? (noooooope.) I firmly believe that this can make you a more connected, more empathetic actor in the political space, if you do it right.

That’s enough mushy moral feelings for now. If I’ve timed this correctly, my new housing polling should be live on the Searchlight site. You should go check it out.

Photo by Jacob Padilla on Unsplash

I’m slightly late to this on account of I just moved and I am very tired, but the WIRED article on Chorus and Democratic influencer outreach extremely caught my attention. If you (somehow???) missed this, it’s a very dramatically framed write up of a Dem-leaning nonprofit (Chorus) paying influencers.

Meta-note: I am playing fast and loose with the distinctions between “Democrats” and “liberal nonprofits” here. Legally, Chorus is nonpartisan, as is 1630. I am saying “Democrats” a lot because I am interested in the implications of this sort of work for Democrats, not because the orgs themselves would say they are Democratic. Not a lawyer, etc.

So Spooky

I do think it’s very funny that “a Democratic dark money group” is used as part of the lead here. When you’re talking about stuff done by the larger liberal universe, your options for “who funds this” are 1. The DNC/party org directly (in which case people get mad) 2. A PAC directly (in which case people get mad) 3. One of the blob of nonprofits (likely mad, tbd) 4. An overarching group like 1630 (SO mad) 5. A rich person directly (quantum-mad. Probably mad)

I do not have access to Chorus’ internals, at all. They appear to be a 501(c)(4), and a project of the 1630 fund, which means they won’t have their own FEC filing. This is not unusual, I have worked at multiple orgs like this. Housing a program in a larger nonprofit, as a fiscal sponsorship or something else, can be great for small organizations who don’t want to spend time and energy running their own HR, payroll, benefits, etc.

One of my favorite sections from the WIRED piece is:

“Steven Buckley, a digital media sociologist at City St. George’s, University of London, says that these sorts of programs have been “happening in the right wing for ages.” But Heider said that the structure of The Sixteen Thirty Fund deal raises the question, “Is it ethical to match the tactics of your opponents?”

To which I would say: Yes. Obviously. This isn’t even something that’s illegal! I would bet you 100 bucks this whole situation has been approved by more lawyers than you could fit into my kitchen. I am of mixed feelings about if it’s effective, or a bad look, but that’s separate.

Anyways, I think the freakout about the organizational structure is silly, and I am much more interested in the substance of what they’re actually trying to do.

What’s Actually Happening Here?

The WIRED piece notes that they reviewed a contract, but doesn’t actually post it, so I can’t use that. We know they’re reaching out to creators, and that on signing a contract with them, you get paid to do something. The exact nature of that thing is left uncertain by the article.

Chorus’ website FAQ says the following:

Will I be paid to create content?

Nope. Chorus is a creator-led nonpartisan nonprofit with a really different model in the creator space. We don’t pay creators to produce content—we just give you space to learn more about how to communicate effectively and grow your platforms. In other words, creators are free to post what they want, when they want. This includes their affiliation with Chorus. Please note that we do ask Incubator participants to submit some pieces of content for growth analysis and retain the right to share our analysis of overall performance.

I did check the Wayback machine, and this page was present when the article went live (although as a subset of the “join” page, not its own thing). You can’t expand the dropdown, but if you check Page Source, the above text is clearly there on the page.

Here’s what that text looked like:

“Nope. Chorus is a creator-led nonpartisan nonprofit with a really different model in the creator space. We don’t pay creators to produce content – we just give you space to learn more about how to communicate effectively and grow your platforms. Please note that we do ask Incubator participants to submit some pieces of content for growth analysis and retain the right to share our analysis of overall performance.”

There’s a difference between that text and the current text, namely that they’ve added “In other words, creators are free to post what they want, when they want. This includes their affiliation with Chorus.”

As far as I can tell, Chorus is hosting events where creators learn things about their issues, work on creating content, or otherwise listen to trainings. It’s a little bit like the free vacation that sells you a time share- you get something of value in exchange for listening to a pitch. Or I guess the more apt comparison here is the influencer brand trip! Free stuff, no explicit requirement to post, but a strong assumption that you will in fact post. Presumably the hope is that whatever is talked about/worked on at these events then gets incorporated into the content these creators put out. It also sounds like Chorus is bringing these creators to events with politicians, likely with the same goals.

Per the article:

“The creators who joined the incubator are expected to attend regular advocacy trainings and daily messaging check-ins. Those messaging check-ins are led by Cohen on “rapid response days.” The creators also have to attend at least two Chorus “newsroom” events per month, which are events Chorus plans, often with lawmakers.”

It sounds like being paid a decent chunk of change to go to liberal messaging school and read memos. I read messaging memos for free, but it makes sense you’d have to pay a normal person to do this because oh my god boring much?

Restrictions on creators

Much has been made about pieces of the contract (which again, I have not seen) that restrict creator’s interactions with electeds and officials. There’s two bits that address this:

“according to copies of the contract viewed by WIRED, creators in the program must funnel all bookings with lawmakers and political leaders through Chorus. Creators also have to loop Chorus in on any independently organized engagements with government officials or political leaders.”

“Creators in the program are not allowed to use any funds or resources that they receive as part of the program to make content that supports or opposes any political candidate or campaign without express authorization from Chorus in advance and in writing, per the contract.”

The second piece here seems extremely normal from a campaign finance perspective. You can‘t use 501(c)(4) primary purpose dollars to do direct ads for or against a candidate. Paying creators money that they then use to make ads for campaigns would be illegal under this finance status. Sorry! I know this is boring! Unfortunately not knowing these things will make you see weird conspiracies everywhere when what’s actually happening is tax compliance.

The first is maaaaybe also a finance thing, but certainly weirder. It’s hard to tell from the outside, and this could either be a legal CYA thing of trying to avoid passing material support from a 501(c)(4) to an elected, or a legit attempt to restrict who creators talk to. I am inclined to think the former, but you may be more paranoid than me.

What’s The Goal?

There are three theories of how influencer marketing might work for Democrats that I have heard. The first treats influencers as another media channel, like digital or TV, that can spread specific messages and talking points. The second treats influencers more like endorsers, figures who can swing votes by talking up a candidate or issue, but who shouldn’t produce notably political content most of the time. The third focuses on influencers as drivers of culture. This is the “Rogan of the left” idea, that popular figures can bend culture to the left without even talking about politics.

Of these options, (3) is clearly the most powerful, but also the most squishy and difficult. How do you make people more liberal? You make famous people they listen to more liberal. How do you make *those* people more liberal? Uhhhhhhhhhhh. If we knew that, we wouldn’t be having this problem. I am not really on board with the great man theory of culture, and so I’m just not sure this actually works. Of course I would like big cultural touchstones to be huge Democrats, but I’m not clear that produces votes.

Achieving (1) is right in line with a standard ad-based program trying to drive Democratic ideals and issues. My general feeling is that this works just about as well as you might expect for a standard message based program, which is to say, “fine”. I don’t hate expanding messaging into short form video if that’s where the eyeballs are. Honestly, I kind of wish more Dem-adjacent orgs were leaning in to just straight doing sponsored content and paying people to do an ad read. We all love authenticity, but if it works for mascara, it probably works for this.

Where I do think (1) can work well is in cases where it matches an influencer to a message that actually resonates. A mommy blogger talking about the price of groceries, a young person talking about finding health insurance for the first time, etc. Influencers have the opportunity to make these kinds of core Dem issues less boring than the 1000th tweet going “anyways prices are going up”. This is (theoretically) what separates influencer marketing from your bog standard ad: human connection. If you manage to pull this off, I expect the returns to be higher than your average ad.

It isn’t clear to me which model Chorus is pursuing, and they could be doing all 3 at once. Plenty of vague money spending efforts have been labeled “cultivating talent”. They do seem to be walking creators through specific talking points, and they don’t seem to be doing direct endorsements. So, (1) with aspirational (3)? Maybe?

Ethics and Authenticity

A lot of the discourse has circled on the idea that it’s somehow unfair to incentivize people to be liberal. Between posts accusing the DNC of paying creators to say nice things about Democrats (not the DNC, also not what seems to be happening) and posts saying it’s sad that Dems have to resort to this, there’s been a lot. I know I spend more time thinking about tactics than average, and probably have an overly cynical view of how politics works, but if I could in fact pay people to be liberal (legally), of course I would do that? That seems like a slam dunk! I also don’t think it’s a bad indicator that Dems are trying this. Expanding your range of tactics is good.

There’s meta-concerns about having content that seems inauthentic, or being perceived as shady or corrupt, obviously. There’s a reason I spent the whole first chunk of this piece trying to explain the financials here. Being seen as shady is bad. However, people are constantly perceiving politicians and Democrats as shady. This isn’t gonna be the last straw, especially while we are 100% certain Republicans are doing this and worse. You also cannot reasonably be down on Democrats for not having a strong engagement with influencers and the online ecosystem and then get mad when people try to have that engagement. I know this is probably just the usual online bullshit but like, come on.

Good Enough

My boring take is that this is like, fine. I don’t think this is going to magically fix Democratic problems with the online ecosystem, and I also don’t think it’s some kind of evil attempt to control people. It’s fine. It seems like an okay attempt to figure out how to do the right wing thing of letting folks get paid for years while they hone their brand and technique. If it leads to more influencers talking about prices and healthcare, I’m for it.

I just really, really hope they’re running some RCTs.