Thread by @ArbitragePi

I think it’s conventional wisdom that PI traders skew Republican and contracts for events that are favorable for Republicans have inflated prices. Here’s an attempt to see if that was true leading up to the 2018 election.

Here’s a chart from 538 that shows their expectations for the 2018 Senate results. The shaded area is a range that their formula said had an 80% chance of containing the actual result.

That means they estimated a 10% chance of the result being above the shaded area (and more favorable for Dems) and a 10% chance of the result being below the shaded area (and more favorable for Republicans).

I looked at what kind of returns were offered in the GOP senate seats market for betting against the top 10% of outcomes for both parties on each of the 30 days before the election.

For democrats, this meant buying no on 49- in the GOP senate market. For Republicans, this meant buying no on 55 - 60+, and 54 on days where the 538 expectation had shifted lower.

If both parties’ irrational partisans were equally represented on PI (and if they cared an equal amount about polling) you’d expect the returns for betting against the top 10% of outcomes for each party to be similar.

Going against the top 10% of dem outcomes consistently offered a return between 15 and 25 percent. The return for going against the top 10% of GOP outcomes is way higher and all over the place.

I guess some difference is to be expected because we’re comparing one bracket on the dem side to six or seven on the GOP side. But still, the returns shouldn’t be 2-3x higher when the probabilities of winning, according to a pretty good polling average, are similar.

So, at least in 2018, there was more dumb money on wagered on improbably good outcomes for Republicans that on improbably good outcomes for Democrats. And it looks like the best time to go against those wagers is 2 - 3 weeks out.