Having grown up just south of Saratoga, NY, I spent many teenage summer days sitting in the picnic area of the majestic Saratoga Race Course and many August nights bar-hopping the small-town, mardi-gras-like atmosphere on the streets of downtown Saratoga. You get some funny looks in most of the rest of the country when it’s revealed that you know how to calculate the payout on a $2 exacta ((total exacta pool – takeout)/(total winning dollars bet * 2)) or can name the horse that beat Secretariat in the ’73 Whitney (Onion), but it’s neither trashy nor rare to know horse racing where I come from.

One thing you become quite familiar with at Saratoga, virtually by intuition, is the economic concept of the favorite-longshot bias: horse bettors seem to vastly prefer betting longshots rather than favorites, independent of the true odds on the horses. And since the odds are set by the betting market at the track, this creates an attenuation: horses that should be massive longshots by their true (but unknown) odds of winning (like 150-1) end up selling as only big longshots (perhaps 75-1), while horses that should be ridiculous favorites by the true odds (say, 1-9) end up going off for much less (perhaps 1-2).

As a result, most longshots are absolutely horrible bets (since their pay odds are vastly less than their true odds), while there’s usually a fair amount of value in just betting the favorites (because they pay much better than their true odds would imply). And despite having a century-old reputation as the “graveyard of  favorites” — just ask Man ‘O War, whose only career loss came at Saratoga to, yes, a horse named Upset — savvy horseplayers at Saratoga know that longshots are still for suckers.

Favorite-longshot bias has a long-known and well-documented empirical history in the economic study of the gambling world: using past performances, it has been shown over and over again that bets on longshots lose much more money, on average, than bets on favorites. For example, in the linked study of 6 million horse races between 1992 and 2001, bets on all 100-1 or bigger longshots returned -61%, while bets on all favorites returned -5.5% and utterly random bets returned -23%. Uninformed betting is expected to have a negative return because of the state’s take from the prize pool, which in NY is 16% on win bets. But that only highlights how good blindly betting the favorites can be: if the state wasn’t removing 16% of the prize pool, blindly betting the favorites would generate a significant positive return!

There are two theories about why gamblers tend to behave this way. The first is known as “risk-love.” In this view, gamblers are rational utility-maximizers, and get utility from the excitement of the risk of the longshots, whereas betting the favorite, well, sucks. Anyone who has ever cashed a winning $2 bet at Saratoga for a return of $2.40 (and thus a profit of 40 cents) knows how that feels. The other theory is behavioral: people are not good at discerning between two extreme sizes. And thus gamblers don’t distinguish between 200-1 and 100-1, they are cognitively the same odds. And therefore bettors systematically take 100-1 prices on 200-1 true odds.  And the betting market responds.

I say all this because longshot bias is at least a partial explanation of what you see on Intrade right now, for both the Democratic and GOP nomination markets. Let’s start with the Democratic market:

Intrade Democratic Nomination Market (1/5/2012)

Obama: 94.5
Clinton: 3.0
Biden: 0.5

These prices reflect only the functioning market — where there are actually sellers and buyers — and does not include candidates who have no current offers to buy at any price. As a reflection of the true odds, this seems certifiably insane. I would guess the odds on Obama are much closer to 99.5 (leaving a tiny amount of space for death, massive scandal, or incredible political turn in the party), which means the sum of the true odds of all other plausible candidates can’t be more than 0.5. I would suspect that neither Clinton’s or Biden’s true odds are greater than 0.1.

But the observed odds are easily explained by favorite-longshot bias. Right now, in order to bet Obama, you are effectively laying approximately 17-1 (i.e. bet $17 now and return $1 profit this summer). That might be a nice investment strategy — it looks like a pretty darn stable 5.5% return to me — but it’s absolutely no fun as a gamble. Conversely, Biden at 0.5 looks like a horrible financial investment (a 1 in 200 chance Joe B is the nominee? No way.), but one heck of an exciting gamble — think of the psychological utility you could get from $10 on Biden, which could get you dreaming about two grand every time President Obama has to go the doctor. The Democratic market seems very much to be reflecting a risk-love bias.

Now, let’s turn to the GOP market:

Intrade Republican Nomination Market (1/5/2012)

Romney: 79.9
Gingrich: 5.1
Santorum: 4.8
Huntsman: 4.8
Paul: 2.4
Perry: 2.0
Jeb Bush: 0.3
Bachmann: 0.1
Chris Christie: 0.2
Sarah Palin: 0.1
Paul Ryan: 0.1
Mitch Daniels: 0.1
Buddy Roemer: 0.1

This market strikes me as equally biased, but somewhat different. The existence of the true mega-longshots at 500 or 1000 to 1 seemingly presents a greater possibility that cognitive misperception, rather than risk-love, is at work here, at least for the most extreme candidates. Does Chris Christie have a 1 in 500 chance of being the nominee? I doubt it, but he might very well have a 1 in 5000 chance (0.002%) and definitely a 1 in 50,000 chance (0.0002). Ditto with Bush, Ryan, and Daniels. Those all look like terrible, terrible bets.

But, of course, that’s not the interesting question. What we’d love to know is how the favorite-longshot bias is affecting Romney’s price and the price of the other top tier and second tier candidates. This is, of course, seemingly an imponderable, since we can never know the true odds and we don’t have millions of past performances to empirically examine like we do in horse racing. Still, anyone who’s ever stared up at a 1-4 favorite on the Saratoga board and even passingly thought to themselves “this just isn’t worth it for that return” can probably relate to someone who’s irrationally passing on Romney right now on Intrade.

And the instant you have that feeling, you have proof there’s at least some marginal value in Romney due to the favorite-longshot bias. Now, marginal value is not actual value. Romney’s true odds may be lower than 79.9 for fundamental reasons that are not being captured by the market. But so long as any bettors are passing on Romney in favor of another candidate because it’s just no fun to bet the big favorites, there’s a market distortion that can theoretically be capitalized on.