No-risk free money in the presidential betting markets

A couple of weeks ago, I wrote a couple of posts about the divergence between Intrade and fivethirtyeight.com in their estimates of the probable results of the upcoming Presidential election. In the second of those posts, I linked to Nate Silver's recognition that the betting markets themselves did not agree: Intrade consistently leans Republican relative to the Iowa Electronic Markets.

As it happens, I paid close attention to the two markets on Saturday, and I saw the prices converge, with the IEM probability of a Democratic victory steady around 71 and the Intrade probability drifting up to that level from the sixties. That evening, a hammer dropped: a huge, sudden sell order of the kind Nate had identified reinstated the Republican lean of the Intrade markets, and it has remained intact since, even growing. At this writing, the probability of a Democratic victory (which creates a slightly neater comparison than the Obama-only price) is this:

Intrade: 68.1 (ask 68.3)
IEM: 76.4 (bid 75.0)

That's spread of 8.3 points between the prices of the most current sales! (Incidentally, the Nate Silver model has the probability creeping closer to 90% now.) That spread is the kind of problem that arbitrage should be fixing, and I hope people who understand these markets better than I do can comment on why it's not. But here's how the arbitrage trade would work, and this is why I included the current ask (selling) price for Intrade and bid (buying) price for IEM as a more realistic estimate of what a trader could do right now.

The way the arbitrage trade works is that you short sell the commodity where it is priced high and buy shares where it is priced low. The principle is incredibly simple, although its application is often wickedly complicated: it's like buying a bag of peaches for eight dollars and then selling them for ten.

In this case, let's say you start by selling 100 DEM shares on IEM, where the price is higher. Buy low, sell high. (This is short selling, so you sell first and pledge to buy later; a short sell makes money if the price goes down.) At $75 each, that gives you $7500. At the same time, you buy 100 DEM shares on Intrade. at $68.30, that costs you $6830.

In a month, you're going to get the value of 100 Intrade DEM shares (the ones you bought) in exchange for the value of 100 IEM DEM shares (the ones you sold short). And this is the key: at that point, the values will necessarily be the same. Either all those shares are worth $100, or they're all worth nothing. Either the Democrats will have won, or they won't have won. The share values must converge.

So if Obama wins, you get $10000 for the 100 shares you bought on Intrade (now worth $100 each), which nets you $3170 ($10000 minus the original cost of $6830). That's balanced by your loss on the short sale of $2500 (buying back shares for $10000 that you sold for $7500). The net is $670.

And you get the same amount if McCain wins. In that case, you lose the $6830 you paid for the Obama shares, but you get to buy back your short sell for nothing, which means you keep the original $7500 you received by making the sale. The difference is again $670.

Your balance on November fifth, no matter who wins: $670.

That's a fantastic guaranteed return for a month, even if you add in the transaction costs, and even if you don't compare it with the current behavior of the equity markets.

So why on earth is this difference persisting? There seems to be a serious issue with market manipulation--as Nate suggested--or with some other kind of market inefficiency. My sense is that the problem is more with Intrade with IEM, which is important given that a lot of serious people, such as Greg Mankiw, use Intrade to represent the voice of betting markets as a whole. Something is not working as it should here.

Note: this post does not constitute investment advice. I am not a professional. If you make investments based on blog posts by English majors, well, you figure out the end of the sentence.