Manipulation in Political Prediction Markets

My latest column in the W.S.J. assesses whether Intrade’s political prediction markets have been distorted by market manipulation:

Over recent weeks we’ve observed a pattern of large orders for Sen. McCain on Intrade … executed at times when liquidity is particularly scarce. These orders have caused markets to shift sharply, often against the broader political narrative.

The odd pattern has given market-watchers a hint that something was awry, a point recently emphasized by Nate Silver, the election analyst behind This unusual trend … has not let up.

How do we sort out manipulation from regular trading?

The biggest difference between typical market movements and manipulation is that honest traders will usually try to minimize the impact of their trades on the market price; paying higher prices for an asset only cuts into profits. But a market manipulator, intent on buoying the market’s ratings of their preferred candidate, will work to maximize the impact of their trading on the price.

And indeed, the suspicious behavior on Intrade involves large purchases of Sen. McCain’s stock, executed at times when most traders wouldn’t be active. A flurry of buying at about 8 a.m. on Tuesday drove McCain’s stock from 47 to 51; a follow-up buying spree about 12 hours later drove the price from 48 to 51. The same pattern continued on Wednesday, with an 11 p.m. buy from 43 to 46, and was seemingly repeated with another 11 p.m. buy on Thursday, which pushed McCain’s price from 42 to 47.

Consider some further evidence that these trades are due to manipulation:

These price changes are unusually large and occurred during periods of relative political calm. … Prediction markets assessing the fate of important swing states did not mirror these movements. … Alternative prediction markets, including BetFair, British sports books, or play-money markets, also failed to register similar movement for the Arizona senator.

Trying to look beyond the impacts of this attempted market manipulation, how is the 2008 race shaping up?

On Intrade, Sen. McCain is currently given a 36 percent chance to win compared to Sen. Obama‘s 64 percent, which amounts to a 20 percentage point shift to Sen. Obama in recent days. Some part of this shift toward Sen. Obama may simply reflect the fact that his price had been artificially depressed in previous weeks.

Other markets detect an even larger lead for Sen. Obama., the British prediction market, shows the Democrat to be a 74 percent favorite to win. The Iowa Electronic Markets suggest a 70 percent chance that Sen. Obama will triumph in the popular vote, with a seven-point edge in his expected vote share.

Even as BetFair and Intrade point to different chances of winning the election, they each suggest a roughly similar set of outcomes in the swing states, most of which currently favor Sen. Obama. Colorado, which is likely to be the pivotal state, is rated a 68 percent chance to go to Sen. Obama on both markets.

This isn’t the first time that traders have attempted to manipulate political prediction markets. Indeed, interesting research by Koleman Strumpf and Paul Rhode has shown that such shenanigans may be the norm, rather than the exception.

They find that manipulation may knock market forecasts off course for short periods, but they are rarely off base for long. And even despite these missteps, prediction markets have amassed a pretty impressive record, out-predicting pollsters and pundits. That is, prediction markets are clearly imperfect, but I remain convinced they are the least imperfect tool we have for tracking the political race.

All of this suggests a note of caution when reading the markets, and hopefully this posting will give you some sense of how to assess whether price changes reflect manipulation. I wouldn’t be surprised to see further mischief.

The full article, written with David Rothschild, my star graduate student, is available here.

David Zetland

Anyone with insider information (per #6 above), can comment anonymously on this topic at:


If you think someone is manipulating the price, or that the price is in any way inaccurate, then buy or sell based on that and make a profit. (many people do, which is why these fluctuations last a few hours at best usually)


Following #7 - some may bet against the one they want to win, so that they either pay for a good result or are compensated for a bad result.


I remember when this happened. I made a decent profit from selling my McCain contracts at around 49 and conversely buying cheap Obama at 51. I looked for an answer to this sudden shift and it seemed to be the Gallup poll that was released the same day that put McCain in the lead at something like 51 to 47.

If someone as you suggest is manipulating, what is their incentive?

Paul K

Remember that some people buy positions on Intrade based on what they actually think will happen (want to make a profit) and some will based on what they want to happen (partisans if you will). The latter is somewhat like donating money to the campaign, except you get it back and more if your candidate wins! I think we are seeing a lot more of the latter camp now that people realize that it can influence undecideds (who often just seem to want to vote for the eventual winner). So, influencing is probably going on constantly, just not as determined as in the case highlighted by this post.

What is interesting is why Intrade may be accurate. It may simply be that the traders in general are more in touch with what is really going on and are paying more attention to which polls to trust (although 538 is making that easier for everyone, it is still likely that the Intrade crowd knows to go to sites like 538).


Scott Supak

I wonder if the trade sites have a list of the IP addresses of the people making the odd trades?

I wonder if those trades came from the McCain campaign or some other wingnut central office, where they think the "truth" is something they can create with a press release or market manipulation?

Aren't there terms of service on these sites, and couldn't this be considered a violation?


Couldn't this person or persons just have been really short on McCain contracts and was systematically neutralizing their position?

If I was short a ton of McCain contracts, at say 55, when he was trading better, and I thought he might make a comeback by November, I'd be buying the crap out of McCain at 47-51 too.

Maybe not, but I don't see how we rule this out from the available data.


I am perplexed by what appears to be an obvious arbitrage opportunity by intrade listing four markets for the same thing:





Seeing as how these four markets will tend to echo eachother strongly, someone who owns many shares in one of these markets with many outstanding shares can manipulate the price of another market with fewer outstanding shares at less cost. The arbitrageur need only one piece of information which is not readily available to the average intrade user: the number of outstanding shares in each market.


This is going to sound radical but bear with me-- Why don't we let a nonpartisan group of economists write the bill? I know what you're thinking, there are economists working on it. But really is Paulson a economist?

If you look at the stats economists in research and academia have been against this bill for over 2 weeks. maybe we should start listening to them.

Here's what they'd say:


Polling v. prediction markets

Clearly, polling has its problems. But a more intelligent examination of polling might reveal good data in there.

What FiveThirtyEight does is a HUGE leap in interpretting polling data. I hope that in the future a couple other groups will attempt to do that kind of thing, with altertnative models and weightings.

Do you think that prediction markets are more accurate than the kind of analysis that FiveThirtyEight does?

Kevin H

I've always wondered how much better prediction markets are. First off there is a question confound. In polls, you are asked: "Who will you vote for", yet in prediction markets you are asked: "Who do you think will win." The second question allows you to draw in a lot more information, so polling with that question might make polls significantly more effective.

Also, I haven't looked too deeply into the analysis of the comparisons, but you would need to be very careful in how you time the comparisons. Because a poll is a piece of information, one would expect polls to influence the prediction markets, therefore to tell if one was significantly more accurate than another, you would have to look at the market while the poll is ongoing and secret, so that the market can't use the info from the poll. Factor that in with a large number of polls, and aggregates of polls, and the modeling because very complex.