Response #2 on Michele Bachmann
Daniel S. made some very good comments today about last night’s post which I’m going to attempt discuss intelligently despite my brain being addled by circa 10 hours of math today.
So here is Daniel’s first comment:
“I’m not totally convinced of Intrade’s track record on these issues? There are a lot of takedowns of intrade by other economists and researchers that point noting the low trading volume on a lot of these contracts, particularly with closing dates this far in the future, leading to plenty of opportunities for arbitrage.
This doesn’t mean I think Michele Bachmann is likely to be the Republicans’ actual presidential nominee, but Nate Silver has some posts the last couple days where he develops some regression models demonstrating that the Iowa Straw Poll is at least a decent predictor of who is likely to win the Iowa caucus, the first caucus/primary of the season. At the same time, it seems that the Iowa caucus itself has more predictive power for who will be the Democratic nominee than for Republicans, at least in the last few election cycles.”
Daniel raises two very fair points about prediction markets and so I should preface this comment by saying that both the factors he discusses could attenuate the reliability of Intrade. However, for a few reasons I think that as far as the Michele Bachmann issue is concerned, Intrade is about the best predictor out there. First, as to the issue of trading volume, there is no magic number that a market needs to achieve to operate efficiently. Economics suggests, however, that observing the bid-asks spreads quoted in a market will provide an indication of how efficiently it is working. A bid-ask spread is a little discussed but very important aspect of financial markets. When you buy a stock or any other financial product, if you were to turn around and sell it immediately, you would find that you could only sell it at a slight discount to where you bought it. For heavily traded stocks like GE, bid-ask spreads are extremely small and you can effectively ignore them unless you are day-trading and you shouldn’t be day-trading anyway. However, in many bond markets and markets like Intrade where there are fewer traders bid-ask spreads may show significant deviation. Consider this stylized example from the financial crisis. Imagine that I was the last bank to buy a mortgage backed bond before the market crash. I may have paid a pretty penny for that bond, but I would then turn around and find no buyers for my bond…effectively its value has dropped to zero because no one will buy it from me. What is the value of the bond? Somewhere between zero and what I paid for it, but that doesn’t really tell you very much, does it? In such situations markets are not particularly reliable indicators of value.
However, if you go to intrade right now:
you’ll notice that the bid-ask spread on a Bachmann contract is 6.6-6.5 or 10 cents in other words. This is a very small margin and indicates to me that the market for Bachmann futures is working quite well. Note for instance, if you go to Intrade’s scientific page and look at the contracts regarding when someone will detect the Higg’s Boson:
The bid-ask spreads vary by as much as 40. Thus, Intrade is probably quite unreliable on that measure. As for the regression estimates, a quick glance at the data indicates that the Ames Straw Poll has been a pretty poor predictor of outcomes for Republicans, so whatever is driving the results, its not the ability of the Ames Straw Poll to predict Republican outcomes
In fact, the only nomination it has correctly predicted was Bob Dole and even then, he tied with Phil Gramm that year. Also note that Pat Robertson managed to win this thing in 1987. Enough said.
Finally, to the extent that Intrade is not accurate at the moment, given the hype surrounding Bachmann, if anything I would expect it to be exaggerating her probability of getting elected. But actually today saw gains for Perry and Romney and further (small) losses for Bachmann.
As for the second article Daniel posted, I agree that it is simplistic to assert that the absolute probability of the event occurring is the very last Intrade price. It is probably better to think of Intrade as predicting some range of outcomes. But I still think you can draw the conclusion that based on the Intrade data it is highly unlikely Bachmann will be nominated and that the most likely outcome is that either Rick Perry or Mitt Romney will be nominated.