What the Heck Happened on Predictit?

Published: 12/21/2020

There are few facts more apparent or which evidence of surprises me more than 'other people have different beliefs than me'. So I was greatly surprised and confused by what happened on Predictit after the election.

First a disclaimer: I started brainstorming this post around the 10th of November and am writing it on the 17th of November but don't expect to publish it until the facts of the matter, both with regards to the election and the settlements on Predictit are long settled (Predictit paid out Wisconsin and the electoral college markets on the 21st of December). Frankly I'm a coward. I did make a tweet about it and got an unsettling response from someone I don't know.

And maybe he's right [1]. I don't think he is but I'm too much of a coward to call my shots. I put 2k on my beliefs but frankly feel I'm something of a coward for not maxing my credit cards on this belief. Dollars are for sale for 90c. If someone bet you that the sun wouldn't rise tomorrow with 9:1 odds you wouldn't say "Sure, I've got a $20 bill on me". You'd say "hold on, let me visit an ATM". [2]

A second disclaimer: I've always sort of thought 'fake news' was sort of 'fake news'. Republicans complain about voter fraud and Democrats complain about Russians posting memes on twitter. But there really are people who believe false things. Demonstrably false things. And what is really crazy is that they'll place bets on those falsehoods. It's crazy (in the purest sense of the word).

I've recently come around to this perspective that political beliefs are sort of like another religious belief. Unlike religious beliefs I do think with time and energy we could really get to the bottom of what an optimal tax policy or minimum wage was. But none of that matters to *me*. It does matter to me what tax policies and minimum wages our glorious leaders will create. But I am not creating them. My thoughts and desires on those policies are causally unrelated to the realities that are being designed. I should no more have an opinion on the largest marginal tax rate than I should have an opinion on how many angels can dance on the head of a pin or how the holy trinity makes logical sense. Maybe one class of such propositions can be 'settled' but neither class is going to guide me in deciding what to eat for dinner or what career to pursue. Of course the beliefs I choose may effect who I want to be friends with and who wants to be friends with me. But that's not the same thing as using a belief.

I think that captures what's so strange about the Predictit nonsense to me. It's so common to see people with crazy political beliefs. In fact I think it's not merely a possibility or just the default but a truly inescapable tar pit that afflicts anyone who spends enough time thinking about politics. But it's rare to see people paying any sort of social cost for their beliefs. They may pick up some enemies. But really the enemies they make are people they weren't interested in being friends with anyway. It's strange to see people pay any sort of cost for having false political beliefs. Let alone droves and droves of people literally throwing money away.

I guess I shouldn't be so surprised though. I missed the totally obvious. There are people who believe the things Trump says. They read a tweet from him and think 'Huh' (I think 'huh' too but with a different tone). If you believe what Trump says and not what the vote tallies say than of course you're getting quite the bargain on the prediction markets.

I've often been an apologist for prediction markets. I'll say things like 'well of course the way Predictit is set up it can't accurately price things that are four months away and have probabilities not between 20 and 80'. It's a platform for gambling. Not enough people are logging in and figuring out how to get 25% apy. Too many people are thinking 'well I can imagine a scenario where Clinton get's the nomination again so clearly it has probability bounded away from 0'. But this is another level of delusion. The markets no longer show our disagreement about the future. They show our disagreement about the past. Speaking of apologists for prediction markets, here's a tweet from Robin Hanson:

I really love this because there are so many examples of Hanson asking people if they put money on their beliefs when they point at a market as an example of prediction market irrationality. And in the past month when the largest, most obvious, and most sustained prediction market irrationality occurred [3] Hanson didn't even know about it? I guess when your position is that prediction markets could help human decision makers you don't look for examples where they're beyond stupid, when they stop estimating probabilities of events and start estimating the proportion of people who believe conspiracy theories.

I suppose I should briefly and explicitly say what happened. On Predictit. As of 11/17/2020. One could buy for 90c, assets that are worth a dollar if Biden won Georgia, Nevada, AZ, Wisconsin. And there are even cheaper assets worth a dollar if Biden won the presidency. He did in fact win all those things. He won those things. They are worth a dollar. But they were on sale for 90c. Sometimes as low as 87c when Trump makes a confident tweet.

I usually try to avoid anthropomorphizing my counterparty on Predictit. I guess that's the wrong word. Of course they are human. At least the terms of service says they had better not be bots. What I mean is I don't want to have a developed theory of mind for them. I don't want to be buying BidenY and think to myself 'my counterparty is a brainwashed MAGA idiot who will buy TrumpY no matter what'. Because of course I don't know why they're buying TrumpY. Maybe they have live poll data and they think he's more than 45% to win [4]. Maybe they actually think Biden is likely to win but they're hedging their exposure in another market. I don't know that my counterparty wants Trump to win. I only know they think this is a good transaction for them. I tend to think people overestimate the likelihood of the political outcomes they want so they tend to buy the assets they want to happen [5]. So I think it's likely I'm betting against people who want Trump to win. But who knows.

But my counterparties now must be true Trump believers. I don't believe there exist democrats who think to themselves 'Well I definitely don't want Trump to win but I think with probability >10% a giant voter fraud will be revealed overturning the result of the election'. People aren't just betting that Trump won the election. They are betting that he won by over 280 electoral votes. It's delusional! I suppose there might be people who are thinking 'obviously this asset is worth 0 but people will buy it for 20 before this shit show is over'. But are there that many of those?

It's also possible that there are people afraid of counterparty risk. They bought BidenY but now they are worried Predictit is a scam and they just want to get their money out. Looking back on it this is sort of reasonable. Predictit did do some vaguely scammy things, keeping markets up past when they'd normally be closed, opening new markets which were for things that had essentially already happened etc. Everything is all paid out now but I can see how some people might have been getting a little nervous for a while.

But despite those possible other archetypes of traders with motivations and beliefs I can understand I can't escape the conclusion that there were far more people who believe in election fraud conspiracies than I realized. And furthermore they believe in those things far more strongly than I realized. I didn't write this blog post to bash the 'other' for being wrong (though I guess it mostly turned out that way) but more to acknowledge that I myself was wrong. I was wrong about how bad fake news is. I was wrong about how strongly people with crazy beliefs actually believe their crazy beliefs (in the sense that they're more willing to act on their beliefs than I realized). I'm not sure what I should do with this new perspective. Maybe I'll look for more opportunities to bet on my beliefs.

[1] Of course now we know there wasn't a successful coup. But maybe I'm the fish for thinking the probability of a successful coup was less than 10% in November when it was really 30%. I don't think so. But we can't know.

[2] After writing this sentence I did put down another 5k. But you could get a 10% return in a month for up to a 40k investment on Predictit for weeks and weeks.

[3] The effect of fees is greatly exaggerated on Predictit. If a sure thing asset is priced at 94c or less then you're making money net of fees. If you have a 2% credit card (and if you don't, stop reading this blog post and get one right now. I recommend Paypal's but I'm sure they're all good) then there's profit opportunity up to 96c. A bigger problem is the $850 per trader cap. Without a cap I'm sure some Predictit whales would have swept up 100s of 1000s of dollars of shares in some markets. But there are no caps on betfair and the prices were about the same. So what do I know?

[4] If this transaction was happening the day before the election this is plausible. I myself had difficulty squaring my belief that markets are efficient and Nate Silver is a domain expert in election forecasting. Based on what happened after the election I guess the answer is the markets are not efficient.

[5] Not to say I'm not guilty of this. During super Tuesday I had many a Bernie share and few Biden shares. I'd see a market where Bernie was priced ahead and think 'this is a sure thing' and I'd see a Biden ahead market and think 'seems about right'. Of course I got absolutely destroyed by Klobuchar and Pete dropping out. Heckin' Obama ruining my day.