This is a visualization comparing the prices listed on the horserace gambling website PredictIt versus candidate polling in the 2020 Democratic primary. The gist of what I’m interested in is how closely PredictIt’s market offers follow candidate polling averages. Keep in mind that the PredictIt market is trying to measure probability of a candidate’s ultimate nomination, not their current polling average. The difference in the two measures will probably get extreme as the campaign rolls on. For example, if there were two candidates left with a week before the DNC, and one was polling at 70% and the other at 30%, you’d probably put the former’s probability of winning the nomination at a lot higher than 70%. But this far out, and with this many candidates, polling average and the market prediction seem to track fairly closely to one another.
I’m writing this on August 29th, 2019. I’m not putting a date up anywhere in the top of this post because the visualization is dynamic and should be current for a good long while yet. But I’m obviously writing about a particular snapshot in time, so don’t hold it against me if my observations don’t hold in your temporal reality.
The first thing that jumps out to me is that PredictIt really likes Andrew Yang and Kamala Harris. And Elizabeth Warren. And Pete Buttigieg. And Cory Booker. And Tulsi Gabbard…. In fact, if you add up yesterday’s (8/28/19) closing share price of every candidate with a market value of $0.02 or higher, you get $1.18. And since the cent value is supposed to equal probability of winning the nomination, that means that the PredictIt community thinks that there’s a 118% probability that the Democrats will nominate somebody. Which leads to our first lesson in gambling on PredictIt - In general, bet that candidates are doing worse than PredictIt says they are.
Note that I say “in general.” While PredictIt is bullish on, for example, Andrew Yang, it’s also consistently bullish on Andrew Yang. Contrary to the beliefs of the Yang Gang, Andrew Yang doesn’t really have a great chance of winning. He has an enthusiastic group of supporters, but they just don’t make up that big a proportion of the Democratic electorate. I assume that said enthusiastic supporters are responsible for Yang performing so much better in PredictIt than in the polls, even relative to other over-performing candidates. And eventually, he’ll drop out. But he probably won’t drop out for a while because he has an enthusiastic base, plenty of money, and the party doesn’t really have any leverage over him. So while a heavy bet on Andrew Yang to lose the nomination would almost certainly be a safe bet, it would also be one that you couldn’t cash out for a good long while, and that would be boring. If you’re on PredictIt to be bored and make money, then stop and go invest in a more stable marketplace with higher returns. If you’re on PredictIt because it’s fun, maybe throw some cash at Andrew Yang to lose, but don’t get too much of it locked up with a bet that’s not going to pay out until the Yang Gang comes home.
So far I haven’t seen anything to suggest that PredictIt is more predictive than polling. That’s why at the moment I think PredictIt is way too bullish on Elizabeth Warren. This could be a good place to gamble a bit. Warren has been doing better in the polls of late (it’s a bit hard to tell from this snapshot below because this is an unweighted polling average that only goes back 90 days. Flip over to the 2020 Presidential Polling aggregator to get a better idea of her weighted trend), and there’s a narrative emerging that the race is narrowing to her and Biden. But I think that’s premature. She still isn’t polling ahead of Bernie in my polling average, and she’s only barely ahead of Bernie in the slighly-less-good version put out by Real Clear Politics. Meanwhile, her PredictIt market value has been skyrocketing at a much higher rate than her polls have. I’d guess that she’s going to come back to earth a little before the September debates.
There’s one other thing worth pointing out. While trade volume tends to be high the night that a candidate is on the debate stage, it’s even higher the day after a debate. Maybe the folks on PredictIt have absorbed the age-old political lesson that perception of a political event is more important than the political event itself, so they’re waiting to place bets until a post-debate narrative has started to solidify. From my own obsessive reading and podcast listening, I’d say that in the last two debates, a solid narrative had emerged by 12pm EST the day after the second debate.