With two weeks before the election, most political followers are glued to national and swing-state polls. But some junkies and journalists swear by Intrade, an online betting market where investors place money on the next president (among other things).
But this morning, something very weird happened on Intrade. Mitt Romney began the day trailing the president 60 to 40 (i.e.: his chance of winning was priced at 40%). Suddenly, Romney surged to 49%, and the president's stock collapsed, despite no game-changing news in the press. The consensus on Twitter seemed to be that somebody tried to manipulate the market.
But the more interesting question might be: Could a campaign re-direct their tens of millions of marketing dollars to bid up the candidate's stock price for a final month to bolster his Comeback Narrative in the press? We hashed it out with Justin Wolfers, an economics professor at the University of Michigan.
DT: What just happened?
JW: At around 9:57am this morning, I noticed something funny happening on InTrade: Obama's stock was tanking, and this was happening in the absence of any concrete political news. Barnard College's Rajiv Sethi alerted me over Twitter that this was really due to some unusual trades in the Romney stock (which then ultimately affect Obama).
In the chart below, you can see this very clearly: Romney's stock shot up from 41 to 48 in a matter of minutes (suggesting that his chances of winning the election had risen from 41% to 48%).
To be clear: We don't know what caused this. It may have been an attempt at manipulating the market. It may also have been a trader with fat fingers making a mistake. But the fact that the uptick took several minutes, rather than occurring instantly, suggests that perhaps it wasn't fat fingers. It still may have been some other form of naive trading. Remember: When you're buying a stock, you want to minimize the extent to which you bid the price up, because that only makes it more expensive for you. These trade did the opposite. It's hard to think of a way of having a more immediate impact on the price.
How much might this sort of manipulation have cost?
The total quantity of Romney stock traded between 9:57 and 10:03 was around $17,800. But that's not the "cost" of this manipulation (if that's what it was), because the buyer got stock in return. If we value that stock at 41 (rather than the higher price he paid), the net cost of this manipulation/error was about $1,250.
What did the trader get in return?
About six minutes where Romney's stock rose sharply. Notice though that the effect disappeared very quickly. The Obama Flash Crash disappeared nearly as quickly as it appeared.
Two conclusions follow. First, you can manipulate prediction markets fairly easily. But second, you won't get much bang for your buck.
Considering the attention Intrade prices get in media, is is possible that sustained Intrade manipulation would be a relatively smart use of campaign money?
My guess is that a campaign would have been better off with another $1250 spent on get-out-the-vote efforts than on the six minutes of a higher stock price that Romney got this morning.
But that doesn't say much. There are surely more subtle, and hence less expensive ways of manipulating political prediction markets. And there are times when it will likely also yield some useful publicity -- something that a short-term blip at 10am on a Tuesday won't do.
There are several scholars right now looking at manipulation of political prediction markets. They tell me that they have some circumstantial evidence that InTrade has been the subject of several manipulation attempts, but I'm yet to see their work, so I'm not sure how seriously to take it.
Could we see more serious attempts at manipulating prediction markets?
The money involved in this morning's shifts was small, relative to the millions that someone like Sheldon Adelson has put into the race. But just because $1,250 bought 6 minutes for Romney doesn't mean that Adelson could easily or cheaply scale this up. One reason that these unusual trades had any effect at all is that they occurred during a fairly quiet time for political prediction markets -- at 10am, the day after a debate.
Also, this morning's attack succeeded in moving the Romney stock up sharply, but had a more muted effect on the Obama stock, and little effect on stocks tied to whether Dems or Republicans would win the White House. Also, there were no similar effects in other prediction markets like BetFair or the Iowa Electronic Markets. So a "successful" attempt at manipulation--or one that was less transparently not driven by fundamentals -- would be much more expensive.
Moreover, if shifting the odds for 6 minutes costs $1,250, I don't think you can conclude that shifting the odds for 60 minutes would cost $12,500. My guess is that creating a sustained Romney boom would require a lot more money than this. Now I can't be sure of this, as there are two forces in opposing directions here:
This flash crash occurred so quickly that many people who would have been willing to sell Romney stock at 44, 45, 46, 47 or 48 hadn't even had a chance to log on and see that someone was willing to buy at that price. This morning's rogue trader was able to soak up the liquidity that existed in the market only in a very brief and unusual window.
But countering that: The price movements this time were so clearly not driven by fundamentals, that they likely gave traders more confidence that they should lean against whoever was so willing to buy Romney.
I should also add: Shifting the odds at a more important point in the political cycle--say, during a debate, or on election-eve--would be enormously more expensive, because there's a lot more liquidity.
How can we trust prediction markets if they are susceptible to this sort of manipulation?
Yes, prediction markets are imperfect. But that's not the point. The question is whether they are less imperfect at predicting elections than the alternative approaches. The evidence so far says that yes they are, and this evidence holds despite previous attempts at manipulating markets.
But there are still smarter and dumber ways to read prediction markets. For instance, in this case, the Romney stock moved sharply, even in the absence of news. That should make you suspicious. Also, the Romney stock moved up, even though the Obama stock wasn't moving down (at first). A more telling issue is that the Intrade prices moved sharply, but those on other prediction markets: the Iowa Electronic Markets, BetFair, and with the British bookies didn't. That's a big hint that there may be mischief afoot in the InTrade markets.
My conclusion is that there's useful information in prediction markets, but you need to be careful in interpreting it.