Here's what my simulator would have said Intrade's state markets say Obama's chances were, had I run it on each of the last 60 days (gold line). I also plot the overall Obama wins contract (red line). Oddly, while the simulator results are volatile, they are highly-sensitive to who is actually leading in electoral votes at the moment in intrade - because it perceives the flipability of each state as low unless the prices in that state are both close to 50. Here is a chart over the same time period of how many EVs Obama was in control of by intrade's reckoning. That's the red line, and the blue line is the mean of my simulation on that day.
One possible resolution is that the state market traders expect volatility to be higher in the future than it has been in the past (or possibly they have some expectation about how correlations will be different in the future than they were in the past.) I will solve for what value of forward volatility and correlation would make the simulated results line up with the national contract. There is also an options market on intrade now, which I'l look into for this as well.
Also, everything to date had been based on a 90-day window of the past. I tried using a 45-day window, and that made no difference. That was another possible resolution to the always tricky issue of how to observe volatility, so I at least had to try it out.
1 comment:
This suggests that intrade may be uncalibrated. [1] Specifically, underconfident. I seem to recall that in 2006, every congressional seat in which intrade quoted a >50% chance of democrats winning, democrats actually won. Which I would take as more evidence that intrade prices are too squashed towards 50%, as you're claiming. I think there are formal studies on the calibration of intrade. Anyone have a citation?
[1] One's predictions are said to be calibrated if, historically, of all the times you said "such-and-such will happen with 90% probability" the such-and-such actually happened 90% of the time.
Post a Comment