Learning in Investment Decisions: Evidence from Prediction Markets and Polls – (PDF file) – David S. Lee and Enrico Moretti – 2008-12-XX
In this paper, we explore how polls and prediction markets interact in the context of the 2008 U.S. Presidential election. We begin by presenting some evidence on the relative predictive power of polls and prediction markers. If almost all of the information that is relevant for predicting electoral outcomes is not captured in polling, then there is little reason to believe that prediction market prices should co-move with contemporaneous polling. If, at the other extreme, there is no useful information beyond what is already summarized by the current polls, then market prices should react to new polling information in a particular way. Using both a random walk and a simple autoregressive model, we find that the latter view appears more consistent with the data. Rather than anticipating significant changes in voter sentiment, the market price appears to be reacting to the release of the polling information.
We then outline and test a more formal model of investor learning. In the model, investors have a prior on the probability of victory of each candidate, and in each period they update this probability after receiving a noisy signal in the form of a poll. This Bayesian model indicates that the market price should be a function of the prior and each of the available signals, with weights reflecting their relative precision. It also indicates that more precise polls (i.e. polls with larger sample size) and earlier polls should have more effect on market prices, everything else constant. The empirical evidence is generally, although not completely, supportive of the predictions of the Bayesian model.
Has anyone looked at the performance of prediction markets with and in the absence of relevant polling information? For example, states which had no or very little polling activity still had markets. That comparison may not be very good, as presumably most states with little polling data weren’t close races. Surely there is a nice natural experiment to be analyzed.
You are embarrassed like Hell by this new study. (You’ll have a chocolate binge tonite, to relief your anxiety.) Wait for my next blog post about it. I’ll offer my vision about what is the real social utility of the prediction markets, and what the Open Institute Of Prediction Markets should be doing about all that.
I’m embarrassed by some things, but this study is not one of them.
It’s unsurprising that prediction markets are informed by polls. I suspect if that was not the case, some would criticize PMs for failing to aggregate available information.
I also suspect that the comparison I suggest above would provide some insight on the delta provided by polls. Am I crazy?
I believe Nate (fivethirtyeight.com) was driving the prediction markets based on his analysis of the polls.
Polls, poll aggregators and poll analysts were driving the political prediction markets.
David, so did the markets move when Nate posted? Should be testable.
Chris, what was driving political prediction markets lacking relevant polls?
Mike, small states with few polls had prediction markets with thin liquidity whose probabilities reflected those polls.
I would say that.
I didn’t investigat that closely, to be honest.
And go fugging yourself with the comparison between the polls and the prediction markets. Enough already with that. You should compare PMs with the mass media.
From what I saw in the final two weeks, non-longshot state-level divergences between Intrade and 538 mainly came down to latency in 538 updates. It seemed like the markets interpreted new polls earlier, creating divergences that largely disappeared when 538 was updated hours later.
About the paper, the markets may have immediately integrated new poll data as they should, but smoothed/averaged the news with traders’ previous outlooks. If you superimpose the charts as above, it will look like the markets are integrating polls with a delay.. or are they instead reflecting polls immediately but discounting them? Replacing the market prices with 538’s output on the percentage scale above probably looks the same, as 538 smoothed new polls. New polls are noisy and it’s desirable to wait for confirmation before jumping to the conclusion suggested by the latest poll. As the prediction is made to respond more quickly, the risk of head-fakes increases, like we see in the first three poll spikes above. If you have two series with the same mean absolute error, the more volatile series will score worse with squared or log errors.
[…] F. Masse January 5th, 2009 Midas Oracle is the only media to have published about the Lee/Moretti paper. Is there censorship going on? Here is the listing of the prediction […]
[…] F. Masse January 8th, 2009 Bo Cowgill and Midas Oracle are the only media to have published about the Lee/Moretti paper. We are awaiting insightful takes […]
The truth about prediction markets
Inkling Markets CEO Adam Siegel speaks out on the current state of enterprise prediction markets.