Scott Alexander (Slate Star Codex, Astral Codex Ten) writes asking why none of the clever people following his blog ran for the Governor of California despite it being clear for a long time that:
The Democrats weren't running a clear backup candidate
Very small candidates have been receiving 10% chance in the prediction markets for a long time
California is the 5th largest economy in the world
As a result there is a 10% chance that complete nobody (Kevin Paffrath, yes, exactly) could be Governor of California. Perhaps for multiple terms. That could have been a more qualified candidate.
I think this kind of introspection should apply to us too. How did we miss this significant shot on expected value? How do we stop it happening again?
I'll write more below, but I think it doesn't matter that this is about Democrats. If the parties switched, the argument is the same.
Read the article. I posted the whole text (let me know if copying the whole text was bad and I'll change)
Californians love long-shot bets. Actors trying to make it big in LA, tech founders chasing unicorns in San Francisco, cult leaders trying to found religions in Pasadena. In Silicon Valley, VCs turn the long-shot bet into an art: if some new startup has a 5% chance of making a billion, that's $50 million in expectation. Just a whole state full of people looking for weird opportunities.
...which makes it extra funny that the biggest opportunity of all came by a few months ago, and they all missed it.
My claim is that basically anyone with the slightest amount of fame or money - any B-list actor, any second-tier tech CEO, any successful blogger or influencer, maybe me, maybe even you - could have maneuvered themselves into a position where they had a 5-10% chance of becoming Governor of California.
Let's start at the beginning.
Governor Gavin Newsom had a bad year. First he pissed off Republicans with his strong response to COVID. Then he pissed off the people who wanted strong responses to COVID by attending an unvaccinated unmasked dinner. Also, taxes are still high, homelessness is still high, rents are still (too damn) high, and parts of the state are literally on fire. Gavin Newsom didn't cause most of this, but he also hasn't announced any particularly inspired plans to fight it. Just a really, really bad year.
California has a long tradition of direct democracy. Citizens can circulate petitions, and if they get enough signatures, everyone has to vote on them. After several tries, Republicans finally got enough signatures on a “recall Newsom” petition to trigger an election. The way the election works is: there are two questions on the ballot. First, should Newsom be removed as governor? Second, if he is removed, who replaces him? Everyone gets to vote on both questions, so even if you want to keep Newsom you can still vote on who replaces him if he loses.
Several top California Republicans signed up as replacement candidates. But top California Democrats didn't. The state party officially recommends voters say no to recalling Newsom (obviously), but also that they leave the question about his replacement blank. In keeping with this policy, no important Democrat has signed up as a potential replacement candidate. A few randos with (D)s next to their name signed up for vanity campaigns, but that's it.
Maybe at some point this seemed like a defensible position? California is a deep blue state, so maybe Democrats thought they could just...not dignify the recall with a response? [EDIT: Commenters bring up that in the 2003 recall election, Democrats fielded a great replacement candidate, lots of Democrats who disliked the governor voted yes on recall because it was costless with such a good replacement, and then the governor got recalled and Republican Arnold Schwarzenegger won the replacement election. Now they want to try the opposite strategy of forcing Democratic voters to oppose the recall entirely.] But the mask mandates are lasting longer than expected, the wildfires are pretty bad this year, and Newsom supporters lack all conviction while his opponents are full of passionate intensity. For whatever reason, polls show the recall-Newsom question is 50-50 to pass right now.
That left Democrats under increasing pressure to unite around a Democratic replacement candidate. This wouldn't mean abandoning Newsom - just telling voters "Please vote no on the recall, but in case the recall wins anyway, please vote for so-and-so as the replacement." But since no real Democrats are on the ballot, they need one of the randos who ran a vanity campaign.
Source here. It sums to more than 100% because of usual prediction market issues like fees, transaction costs, etc.
The markets give Paffrath about a 10% chance of becoming Governor of California.
I think this makes sense. Republicans might turn out to vote in higher numbers than Democrats, which would mean the recall passes. But they might split their votes among the many interesting Republican candidates, the Democrats might all vote straight Paffrath, and Paffrath might win. This isn’t the most likely scenario, but, again, 10% chance.
So my claim is: if you (or I or whoever else) had been thinking clearly six months ago when candidacy applications were open, you could have done what Kevin did. Sign up as a candidate to replace Newsom. Put a D after your name. Campaign under a banner of "Keep Newsom, but choose me as a backup". If you didn’t have a history of elf-related misadventures, maybe the Democrats would have united around you instead of Paffrath. And then you'd have the 10% chance of becoming CA governor instead of him.
Are there any strong arguments against this claim?
Maybe it was impossible to realize things would be this close back when applications to become a candidate were open? But those didn’t close until July, and in July the prediction markets were already giving Paffrath a 10% chance. Granted, the argument was a little different (Newsom was looking stronger, and popular Republican talk show host Larry Elder hadn’t declared his candidacy yet), but 10% is 10%, right?
Maybe Paffrath is a stronger candidate than I think? His YouTube has 1.9 million subscribers, compared to (eg) only 35,000 ACX subscribers, so maybe he’s actually a really big Internet celebrity, such that minor Internet celebrities can’t compete? But I think YouTube subscribers are just really easy to get. Google Trends shows pre-campaign Paffrath gets about 5-10x times higher search volume than I do, which I guess is a lot. But I have infinity times fewer elf-related misadventures than he does, so I still think it’s pretty even. There is another Democratic candidate named Patrick Kilpatrick who doesn’t seem clearly worse than Paffrath but hasn’t gotten the same level of name recognition; maybe you would end up in the same place? I’m not sure. My guess is Kilpatrick just didn’t spend enough money.
Maybe in fact Paffrath is rich and able to outspend everyone else? But here’s the graph of campaign spending this cycle…
Source is here. I know he’s a dark horse, but I think it would be kind of fitting if ERROR got elected California governor.
…and it shows Paffrath spending $250,000 or so. There are 1.1 million millionaires in California, any of whom could presumably outspend him if they wanted.
Maybe if you joined the race after Paffrath, you and Paffrath would split the Democratic vote and a Republican would win? I think this is the strongest objection. But Paffrath isn’t getting lots of attention because he’s organically running a great campaign. He’s getting attention because the media is looking for some Democrat to tell people to vote for, and he’s the best they can find. If you had been around as an elf-related-misadventure-less Democrat for the media to unite around, maybe you could have just trounced him.
Maybe you should actually be scared of the California Democratic Party and its demand that no Democrat stand as a replacement for Governor Newsom? I think this makes sense if you want a future career in California state politics; probably they can blacklist you forever if you cross them. But I wasn’t expecting to have a future career in California state politics, neither were you, and probably neither was Kevin Paffrath. That’s why we would have been able to take this weird opportunity that all the real politicians turned down. Also, given the level of competence they’ve shown here, having the California state Democratic Party out for your head is probably the surest path to a long and healthy life.
Maybe becoming governor isn’t as cool as it sounds, because you’d only get to serve out the last year of Newsom’s term, then get replaced in 2022? But in 2022, you’d be the Democratic incumbent, and even if the rest of the party tried to remove you, you’d have a strong platform to fight back from. Also, even if you didn’t care at all about governing California, it would be a heck of a way to shape the national conversation. I would never have heard of Paffrath if not for all this. How many more landlords do you think are watching his videos now?
Maybe subjecting yourself to the misery of a campaign for a mere 10% chance of becoming governor isn’t worth it? I think this is true for a lot of people. But I’m reminded of this profile of billionaire Sam Bankman-Fried, which says:
Though it may sound like a stretch — lots of billionaires have boundless ambition — both Singh and Barbara Fried firmly believe that Bankman-Fried’s risk-tolerance stems from his effective altruism.
Barbara Fried puts it this way. “If you’re earning money for personal consumption, there is a very steep, declining marginal utility of income. After your fifth Porsche, do you really need a sixth? . . . But if you’re earning money to give it away to charity, there’s no diminishing marginal utility to money. The last life you save is worth as much as the first life you save.”
In the same way, if you don't have diminishing marginal utility to power, a 10% chance at the California governorship looks fantastic. The point of VC funds is to help people who do have declining marginal utility to money act as if they don't. The point of movements and ideologies ought to be to do something similar with power. But this time, nobody bit.
My conclusions are:
1. A lot of people in California like to think that we are Very Smart, but in this case a YouTube celebrity landlord beat all of us.
2. There have been a lot of fights around here about whether you can really get power just by being intelligent, able to predict things accurately, etc. I think this is a point in favor - sometimes there are big opportunities just waiting for someone to notice them. But also a point against - intelligence did not, in fact, help anyone notice this opportunity. Overall kind of a wash.
3. Californians are lucky Peter Thiel isn't a Democrat, because his mind is minmaxed for spotting opportunities like this, and you can bet he would have gotten involved here if he could have. The rest of us should keep re-reading Zero To One until it penetrates our thick skulls.
4. The most likely outcome is still that Newsom manages a razor-thin victory and all of this stays hypothetical.
Some of Nathan's thoughts, written quickly
I'm not suggesting electing an EA. But I think there was a real chance at someone better than a "YouTube landlord influencer". Nor am I saying Larry Elder is good or bad, but some people are just going to vote for a democrat
I don't have a horse in this fight. I'm British. But I do want the 5th largest economy in the world to be well governed. This argument would hold if it was the Republicans who messed up
Why did we not spot this chance? This feels like exactly the sort of of low chance high impact situation an EA org should have been following. I feel like there should be a post-mortem here. How did we miss this?
Maybe some EAs did spot this and didn't run for some reason I don't understand
I propose there should be an EA org looking for races like this. Where there is positive expected value in a very different candidate
I really am asking "why did we miss this" what's your best guess?
EAs who live in California. Were you discussing this? What was the sense on the ground? Did anyone suggest that with a small push there could be a much better candidate?
Maybe you think it's good that EA stayed out of this one. I feel a bit differently. I don't see this as EA becoming political - I'm not saying a democrat should win just that a democrat probably will (either Newsom or this new guy) and that there was a real chance to change who that democrat is. If EA doesn't want to be involved with picking candidates, then is there an organisation that occasionally looks at these black swan events where a politican might be swapped for someone much more competent? I'm not seeking to disrespect Keven Paffrath, but he's probably in the middle of the distribution which isn't great in my opinon.
I'm not pointing fingers, I'm more wrong than most. I literally follow political forecasting markets and have listened to at least one entire podcast on the recall election. I guess I just figured the Democrats and Republicans would each field sensible candidates. I was wrong and I could have seen that coming
I have copied an entire blog without Scott's permission. If you liked it, subscribe. (Scott if you're reading this, I think it was the right call, but I'll remove it if you ask)
I recommend just crossposting articles by their titles rather than including "Crosspost" in the title. It makes the titles just a bit more informative/easier to read. (Also, nearly every crosspost on the Forum just uses a title.)
If you want to be extra-sure people know the content isn't yours, my recommendation is to list the author instead of the word "crosspost" (more informative, gives people more incentive to read if they like the author), in a format along these lines:
most people here want social justification via legible verbal reasoning before they do anything. pwned by the left brain. spending time reading forecasts to manage your uncertainty is anticorrelated with success at a social thing that requires selling yourself. i tried running for city council once after someone suggested it but didn't get far due to social anxiety. maybe try walking up to individuals and suggest they run for public office. the least you can do is support someone you know who is trying it. trying it requires not looking around you for foolproof evidence that it will work out.
I'm not so convinced that forecasting is anticorrelated with selling oneself. I get the sense that Galef argues that in the Scout Mindset but I've not read that yet.
Feels like we are talking about different things if you think the forecast here are disheartening. If I were a Democratic person with some connections in California I'd be kicking myself right now
Short answer, as someone who lives in California, as to why I didn't run:
45% Same situation as DonyChristie
45% Running for office is difficult when the majority of your time is spent trying to get bare minimum survival needs met. This is something many Californians struggle with, so I would let go of the 5th largest economy being a positive thing.
10% I'm not sure being Governor of California would be the best way for me personally to contribute to the world as an EA at this point in my life and this point in US history. I wouldn't be a good candidate.
I am part of the Run for Something mailing/notification list so when I get my feet back on the ground I could try giving it a go. I'm not sure how involved they are specifically with EA.
Some other factors to take into consideration are culture and politics of California. I have thought of the possibility that if someone were an EA and wanted to run for office AND win, they may have to hide the fact that they are an EA, because people don't want solutions. There are many many reasons for this. Too many for one post.
This is my first post, so I probably broke some guidelines. I wish I had a better way of explaining.
I think you're right that its hard to run for positions but I don't think the cost should be borne by individuals. I think if it cost 1 million to have a 5% chance of governor of California then I would call that effective spending.
I suggest that I've not yet heard why an EA org shouldn't have been willing to spend a million here
I'll give it a go: Predictit is illiquid and has roughly a 10% margin of error because of the transaction costs. 5% withdrawal fee (on the gross amount), plus 10% of any profits, plus of course the taxes you'd have to pay.
If I think Paffrath's chance is actually 0, and he's at 11, do I buy no on Paffrath for 89 cents? At most I'd win 11. Then I lose 6.1 on the fees. Then I lose ~1 on the taxes (and if I lose money instead, I can't deduct it on my taxes). So for my 11 cents of nominal gain I only end up with 3.9. This is for keeping my money locked up for months in advance, when I might be able to make bigger profits elsewhere with the same amount of risk (this does imply predictions closer to the election would tend to be more accurate, since my opportunity cost of locking up the money is lower and it's more worth making such a bet. But it's also a lot harder for a prospective candidate to use this info). The $850 limit also limits how much money I can make, especially on a bet that is already rated as pretty likely to win, so if my time is sufficiently valuable that is also a consideration.
And, thinking about this in more detail, Predictit's fee structure actually dis-incentivizes bidding down longshots more than bidding them up. If it has him at 11 and I think his chances are really 22, what happens if I buy the yes on Paffrath contract? My expected value without fees and taxes is 11 cents. But really there's a bifurcated scenario. If I lose, I lose it all and, pay no fees. If I win, I pay 5% on my 100 cent withdrawal, plus 10% on my 89 cent winnings, for a total of 13.9. Plus taxes of ~19. So I keep roughly 56 cents out of my 89 cent winnings. So my real expected value is 63% of the gross expected value of the trade, vs. the only ~35% of gross expected value on the No bet. Further, my ability to get enough return to make it worth my time is higher. If I put $850 onto the No side at 89 cents, and I win, I get 850*100/89-850=~$105 in gross profits (of which I will then get to keep 35%). If I bet $850 on the yes side, and he actually wins, I get 850*100/11-850=~$7727. Even adjusting for my probability estimate only being 22%, I get $850 in expectation (of which I keep 63%). So my estimate of my true expected value of making the bet is $37 if I think the market has him rated 11 percentage points too high, or $536 if I think the market has him rated 11 percentage points too low.
This is very different than how e.g. the stock market works. There, the returns to being smarter than the market in either direction are symmetrical. If I think an $11 stock is really worth 22, I can buy it and make $11 (and assuming I meet the requirements, can get a margin loan from my broker and have an initial expense of only $5.5, then pay the rest when I sell the stock, so I make $11 on a $5.5 investment. Similarly, if I think an $11 stock is really worth 0, and I meet the brokerage's requirements for margin trading, I can short sell as much as I want at the $11 price, by putting up a margin amount to cover my risk of loss (usually 50% of the value of the trade), then if I am right, buy the stock at $0 later to cover my trade. So If I short sell $11 of the stock and it goes to 0, my initial investment is $5.5 and I have made $11, the same as if I correctly predicted it was underpriced.
If Predictit has him at 20%, there's at least a little something to that. But 10% has a high enough probability of being noise that I'd trust the conventional metrics over that.
Edited because some of my original example math made no sense in light of the actual mechanics of Predictit, and thinking through the mechanics added new insight.
I used AI to fix transcription errors, rerrarange the ideas, and suggest tweaks to the title and some sentences.
Three of the most exciting projects to come out of EA in recent years are, in a vague sense, CEA spinouts:
* Kairos is directly a spinout of CEA and now handles most support for university AI safety groups. Basically everyone I've found who knows them is really excited about what they do
* NEST is an opinionated ideas-fi...
This post presents the executive summary from Giving What We Can’s impact evaluation for 2025. At the end of this post we share links to more information, including the full report and...
I recommend just crossposting articles by their titles rather than including "Crosspost" in the title. It makes the titles just a bit more informative/easier to read. (Also, nearly every crosspost on the Forum just uses a title.)
If you want to be extra-sure people know the content isn't yours, my recommendation is to list the author instead of the word "crosspost" (more informative, gives people more incentive to read if they like the author), in a format along these lines: