Why Are Prices F-CKED UP right now?
One of the things that is happening on PredictIt is that the odds that people are giving Trump to win the election and to win some select states are totally bonkers. TBH, it is hard to look at PredictIt’s map and not assume that an extremely large number of traders are taking the same crazy pills as Kanye West.
The conventional explanation for why these prices are so off is that people are retards, especially people who love to bet on Trump. In some cases, that might be true. But it is generally a good idea to understand why people are betting against trends and respect their opinion, otherwise you could get totally nuked with a surprise attack and experience an electoral Pearl Harbor come November.
So here are a few other reasons why pricing is so janky for this election:
1 – There are tons of traders on PredictIt who are not trying to predict politics at all. Instead of trying to nail the right outcome, they will just drop a few brief cases of cash into a market and hope to predict a small movement in polls or in the news cycle that will push the price in their direction by a few cents. This strategy is way more profitable when you take the cheaper side of the bet. So I would guess that there are a lot of smart traders buying Trump to win places he’s guaranteed to lose — for example, New York or New Jersey — and hoping for the right news story to push their 7c shares to 8c or 9c shares.
2 – There are a lot of novice traders who think they understand how to do what I just described. But the truth is it is actually harder than you think, because it requires you to know what your shares are actually worth — and that is its own skill. Everyone who reads this blog has had the experience of getting trapped in one of these “I’m just flipping shares” markets and getting their faces ripped-off. So be careful.
3 – The polls. There is the constant under-current that polls are wrong and Trump has more support than they show. I won’t get into this because lots of people have already written about it already. And while I do think that it’s true that the people just don’t vocalize their support for Trump due to the fact that there is at least a 99 percent chance that an Antifa cell will burn down their house and enslave their family members if they do, I don’t see how those numbers are possibly big enough to overcome Trump’s 8 point deficit in the RCP average.
4 – People are used to elections being close. The last time that we had a serious blow-out was in 1984 when Ronald Reagan basically put Democratic challenger Walter Mondale in a skirt and called him Susie. Adding to that, the Fake News Media (FNM) always portrays elections as being closer than they really are, because they need you to believe that in order to keep your eyeballs glued to the news. Jake Tapper is never going to say, “this election is over, please don’t waste your time watching my show.” That would be like Daniel Craig saying, “this James Bond movie is going to blow, so just skip this one and buy a ticket once they re-cast me.”
We know that #2 is true. And based on all observable evidence, it’s looking like #1 will be the case as well. But no one is ever going to make either of those announcements publicly. Which gets us to…
5 – People are front-running the, “This election is going to tighten-up” trend that the Fake News Media is selling them. But there’s really no guarantee that will happen, and I personally think they’re flushing a lot of money down the toilet. We can talk about that later.
So those are your answers to why pricing is so bonkers: people are stupid; some people are using a clever betting technique; a lot of people think they’re using a clever betting technique but are going to lose money; there is a kernel of truth in the “polls are wrong” conspiracy theory; and people think the election is going to tighten-up, even if there isn’t a lot of evidence to support that.
KEENDAWG.
There’s also a lot of internal mispricings on PI. Take any model–538, for instance. Trump is rather overpriced to win any state (via PI) compared the model, sure–but he’s MORE overpriced to win the election, so even if you want to say “teh modelz are wrong!11!” then he *still* sells for too much as a candidate.
Some opportunities: 538 releases its model output with conditional probabilities. Just under half (47%) of Trump’s potential wins include Colorado, so why does it sell for 13c, which is far less than half of Trump’s 44c? Similar math in Maine (52% of wins, 17c price), and for the admittedly-sparsely-polled NE-2 district, 90% of his wins include it but many of his losses do as well… so there’s no way it should sell for a paltry 75% (33c vs 44c) of Trump’s chances.
I ran a whole model on this. NV and NH are good as well. Know what all these have in common? They’re not sexy. Everyone wants to bet on Trump for the media-narrative White Working Class states (TM) that he won in 2016, making MI and WI in fact good buys for *Biden*, but the swing states Trump lost in 2016 are actually underpriced for a re-flip (exception is NE-2 as a good bet for a 2016 GOP hold, but that’s because it’s not on the front page, so the Trumpers don’t click through to it).
I think it’s more likely that Nate Silver has totally screwed up the covariance matrix in his model, and traders recognize P(Trump wins Colorado | Trump wins EC) < 47% (though the state prices are better for Biden as you say, since the main 'who will win the presidency' market attracts more casual Trump bettors)
I had to read this twice but it’s really smart. What are you watching in Maine specifically?
Hi Keendawg! Thanks for the reply. I’m not watching anything at the district or county level in Maine as of yet; my main thesis is that Biden is horribly undervalued by the markets (a couple days ago you could sell all GOP victories and ties in the Electoral-College-Margin market for 49 cents total before fees, giving you an approximate Biden price of under 54c each). Yet, I don’t want to go all in on something where I might get “unlucky” and lose everything 25% of the time. So I was looking for a hedge–a cheap state that (1)Trump can hardly win without; (2) which he can still take into the “W” column even in a national loss; and (3) which is cheaper than Trump@44c. NE-2 fits all of these, but the polling is sparse and you didn’t ask about it.
Maine appeared on my radar (meeting conditions #2+3, and half-meeting #1) because there’s a fair chance of systematic error this year: polls could be off because of undetected turnout skews, COVID-19, mail-in ballot trickery by either side, or whatever else one cares to name. If Trump under-performs with certain groups, well, he was probably going to lose anyway. If Trump OVERperforms with certain groups, that’s about the only way he wins the election. It’s far cheaper to buy White Working Class voters in Maine, where Trump brought Obama’s 2012 vote margin there from 15 points down to D+3; compare Minnesota, where he only brought a D+7.5 to a D+1.5. Much closer 2016 total for MN, much bigger 2012–>2016 movement for ME. Polls this year look very favorable for Biden, but if the polls are right, my Biden profit overwhelms anything I burn in Maine anyway, so I’ve cheaply insured myself against the most landslide-y half of the potential Trump wins. Yes, there is a chance Joe loses the nation while somehow still taking the Pine Tree State; Clinton did. But I think the odds of the same map again are low, and in 2020, bettors are still fighting the last war–so I got in cheap (17c just yesterday).
I’m using the 538 data here. People criticize me for that on the boards, but you gotta remember that Silver’s 2016 prediction, however bad, outperformed not only other models but also the betting markets themselves, and until proven otherwise, I and my bank account will assume market participants are simply bad at math; it’s worked so far.
If you’ll permit me, I’ll clean up my thoughts into a long-form guest post you can use on a rainy day. I promise a zany, bombastic writing style and more hard numbers than you can shake a stick at.
Email me through the contact page and we can discuss a guest post. Thanks!