Is Someone Manipulating Intrade?

As Naked Dollar readers know, I love prediction markets. I love them because they have a highly successful record of predicting political outcomes with greater accuracy than polls. It's real people betting real money. This will be the third presidential election in which I have utilized the Naked Dollar Electoral Prediction Model (see here for explanation), and it has a great track record.

But I have to say, I'm confused right now. The top-line presidential contracts current place Obama's odds for re-election at 64%. I look at the polls, and the RCP average has Obama actually down by about a point nationally. The most recent Gallup poll actually has Romney up by seven points. Shouldn't Romney have a slight edge on Intrade? (Note: Intrade.com is the most popular site for prediction markets.)

Okay, I know what you're going to say next, it's about the electoral college. I know. So what does Intrade have to tell us about the swing states versus the polls? Let's take a look at Romney's numbers for each state:

                                 RCP Avg.                  Romney Intrade Odds      
Colorado                     +0.8%                               50.9%
Florida                         +2.5%                              64.5%
Iowa                            -2.3%                                41.5%
Michigan                     -4.4%                                12.1%
Nevada                        -3.3%                                34.0%
New Hampshire           -0.8%                                37.5%
North Carolina            +4.7%                                75.1%
Ohio                            -2.4%                                41.5%
Virginia                       -0.8%                                53.6%
Wisconsin                    -2.0%                                33.5%

Frankly, these numbers don't look that off to me. Maybe I would say Mitt should trade a bit higher in New Hampshire and a bit lower in Virginia, although Virginia probably reflects the fact that the most recent polls show Mitt with the momentum there and there's talk of Obama scaling back in the state.

If you award each state to the Intrade favorite, Obama wins 281-257, a helluva close race. Flip Ohio and Romney wins. Failing that, you could flip two other states such as Wisconsin and New Hampshire. Does all this add up to a 64% chance of winning overall, particularly with national polling numbers trending clearly in Romney's direction? Doesn't seem like it to me. What I'm suggesting here is that there seems to be an arbitrage opportunity between the national presidential contract and the state-by-state ones. Go long Mitt on the big contract and short him in the state contracts.

But why should such an opportunity exist? One possible explanation is that someone is supporting Obama's odds on the national contract because that's the one that everyone looks at. I had a dispirited friend call me last week bemoaning Romney's seemingly long odds. The point is, people look at this, including the media and the Twitterverse.

So, what would it take, hypothetically, if someone wanted to distort the market? There are two contracts that you'd have to manipulate, one for Obama and the other for Romney (they are essentially the inverse of each other). The open interest on these two markets is 2.2 million contracts. This works out to about $11 million in money currently at stake.

There are two ways one needs to look at this: short-term and long-term. To move the market intra-day is no big deal. Right now, for instance, you could bump Obama two percentage points for about $7000. That Intrade has been manipulated in this way is almost certain. On election day 2004, for instance, John Kerry futures shot inexplicably higher. This was, mind you, in the middle of the afternoon when no one knew anything concrete. If I recall, they rose from about 50% to 75%. The move was quickly reversed by traders, but it caused a buzz, which was likely the point. I calculated at the time it required an investment of about $10,000 to move the market that much. Ten grand for some media buzz on election day, when there's no other news? That's got to be the best deal on the planet!

Watch for this to happen again this year. Why? Because in the age of social media, particularly Twitter, it will be an even better deal. #election #obama Holy cow, Obama at 80% on Intrade! Retweet, share, repeat. By the time the contract reverses, the meme will have reverberated through the social graph. Mission accomplished.

Influencing presidential contracts for longer periods of time is a much bigger commitment. Daily dollar volume on each contract is roughly $125,000. What percentage of daily volume would one have to be to keep the price, say, 10% off the true market? For the sake of argument, let's say half the vol would do it. This would require a dollar commitment of roughly $7.5 million over the last 60 days of the campaign.

Remember, though, $7.5 million is the amount you'd have at risk. Should your candidate actually win, you make money on the scheme. For instance, assuming someone is currently manipulating in favor of Obama, there would be a profit of $4.2 million. The correct way to think of the cost is the amount by which you overpaid for the contracts. In this example, we assumed 10%, so the "true" cost is $750,000.

That's starting to sound like a better value, especially in the context of a billion dollar campaign, but the problem is the outcomes are binary, and one would probably have to budget against the worst case. Thus, I doubt is long-term manipulation going on.

What are the implications of this? If I'm right, it either means that:

1. Prediction markets are currently mispriced, or
2. Prediction market traders know something the polls don't

We will look at these possibilities in my next post.