Editore"s Note
Tilting at Windmills

Email Newsletter icon, E-mail Newsletter icon, Email List icon, E-mail List icon Sign up for Free News & Updates

January 26, 2008
By: Kevin Drum

BETTING ON HILLARY....Who's more electable, Hillary Clinton or Barack Obama? Beats me. However, an economics professor at UCLA emailed a couple of days ago to suggest an interesting data point: the Intrade prediction markets. Intrade allows you to bet on two separate questions: (1) Do you think a particular candidate will get the nomination? (2) Do you think a particular candidate will win the presidency?

For Hillary Clinton, Intrade currently predicts a 67.1% chance of winning the Democratic nomination and a 44.5% chance of winning the presidency. What this means is that if Hillary wins the nomination, Intrade bettors think she has a 66.3% chance of winning the election. [See technical note below for an explanation of the arithmetic.]

For Barack Obama, Intrade betters currently have him priced at levels that give him a 32.0% chance of winning the nomination and a 15.0% chance of winning the presidency. Using the same arithmetic, this means is that if Obama wins the nomination, Intrade bettors think he has a 46.9% chance of winning the election.

This is a surprisingly large spread. Bettors think Obama is likely to lose the election if he's the Democratic nominee, but they think Hillary is a strong favorite to win if she wins the nomination. The electability spread is nearly 20 points. Interesting, no?

Now, there are at least two reasons to be skeptical about this. The first is the possibility that Hillary's Intrade market is being manipulated. See here for a discussion of this from last year.

The second is the question of whether betting markets like Intrade do a good job of predicting events like elections in the first place. I'm not up on the relevant literature, but I suspect the answer is that they probably don't perform very well in an objective sense. However, what they can do is aggregate public perceptions well. So while Intrade can't say that Hillary is truly more electable than Obama, it can (in the absence of manipulation, anyway) tell us pretty reliably that people think she's more electable.

I don't believe any of this enough to have blogged about it during the week. But weekends are a good time for miscellaneous speculation, and this seems to count. Make of it what you will.

TECHNICAL NOTE: If you're wondering how the probabilities work, here's the math. First:

P(election) = P(nomination) * P(election conditional on nomination)

In other words, if you have, say, a 50% chance of winning the nomination, and a 50% chance of winning the election once you've won the nomination, then your overall chance of winning the election is 25%. Now rearrange the equation to get:

P(election conditional on nomination) = P(election) / P(nomination)

So to get the probability of winning the election if you get the nomination, just divide the election probability by the nomination probability.

Kevin Drum 1:35 PM Permalink | Trackbacks | Comments (76)

Bookmark and Share
 
Comments

I don't understand how those markets can have any predictive value when they are so volatile. Like, a few months ago, Rudy G. had a good chance to win the nomination. Now he has very little. What happened?

Intervening events, but that's the point. If a prediction can't account for what might happen in the meantime, it's not a very good prediction.

Posted by: jc on January 26, 2008 at 1:44 PM | PERMALINK

> the Intrade prediction markets.

Please look at the Intrade "markets" before and after the Iowa and NH primaries. Then ask yourself, why does anyone even bother to look at those farces much less take them seriously?

Cranky

Oh yeah, here comes the "why don't you make yourself a fortune if you are so smart" counterargument. Let's see, $500 betting limits and games so thin that 2 or 3 $500 bets move then 10 percentage points. Yeah, we're in Michael Douglas 1980s arbitrage territory there.

Posted by: Cranky Observer on January 26, 2008 at 1:48 PM | PERMALINK

Only in America, steeped in pro-market magical thinking, would anyone even imagine that these betting clubs could have any predictive ability at all. It's as if someone built a weather forecasting methodology based on polling.

Posted by: jimBOB on January 26, 2008 at 1:51 PM | PERMALINK

jc and Cranky: Yes, agreed. As I said in the post, I doubt that prediction markets work that well. However, they do aggregate perceptions pretty well, and apparently there's a strong perception that Hillary is more electable than Obama, regardless of whether that's objectively true. I thought that was interesting enough to take note of.

Posted by: Kevin Drum on January 26, 2008 at 1:51 PM | PERMALINK

What a bunch of drivel Exit polls can be wrong. betting is what people want or currenlty see. It leaves out the actual voter, voting. I can tell you that EVERYTIME that I'm approached by any pollster, online, telephonic or in person, I give out information that often has nothing to do with my actual reality. Putting one's moeny where one's mouth is, (betting) is also not a sure sign of anything. a lot of people do what the majority does. So why not manipulate elections psychologically by betting a certain way, and then creating that reality?
Our mocratic republic has drifted far afield, and his sort of stuff is jsut another example of fascise. Manipulate through propaganda. BULLSHIT! the vote is what counts, as iowa showed.

Posted by: Chris on January 26, 2008 at 1:51 PM | PERMALINK

You mean the same intrade that gave Rudy! Guliani a better than 45 percent chance of winning the nomination in early December? That intrade? Sure that's a pretty useful tool.

Posted by: Merle on January 26, 2008 at 1:52 PM | PERMALINK

And if you looked just after Iowa, you'd have seen much different results.

If you look after today, they'll change again.

Posted by: john on January 26, 2008 at 1:59 PM | PERMALINK

Kevin wrote: "Who's more electable, Hillary Clinton or Barack Obama? Beats me."

Thank you, Kevin. If only everyone would be so honest, all discussions about "electability" could start and end right there.

Voters do themselves a grave disservice if they choose who to vote for based on speculation about who other voters will choose to vote for.

Obsession with "electability" gave us John Kerry as the 2004 Democratic nominee, and a second term for Cheney and Bush, instead of a first term for President Dean.

Posted by: SecularAnimist on January 26, 2008 at 2:02 PM | PERMALINK

Kevin,

Your analysis of the probabilities is a little too simple. While it is certainly true that anyone who bets that Hillary will win the election must believe (and likely bet) that she'll win the nomination, going the other way may not work. That is, someone who bets that she'll win the nomination but does not bet that she'll win the election is not necessarily saying that she'll lose the election. It's just that for those people the uncertainty is sufficiently high in their minds to preclude their betting, since it's real money at stake. They may, for example, hedge their bets by betting simultaneously for Hillary and John McCain in the general election, which presumably raises the share price on Hillary less than it would if it was a straight bet on her.

Posted by: Greg in FL on January 26, 2008 at 2:06 PM | PERMALINK

All of you who think that the Intrade markets are so bad can surely make tons of money by proving them wrong. Put your money where your keyboard is! I sold Giuliani short a long time ago and it turns out I was right, but I don't have strong views about most of the things in that market.

There is a large academic literature judging the accuracy of prediction markets. The claim is not that they are perfect predictors, just that they outperform by a significant margin the predictions of any individual "expert." The same is true of the odds given in Vegas for the outcome of sporting events. That is also a market and if you can beat it, you can make tons of money. Good luck.

The problem with predicting things -- that predictions are often wrong -- is not limited to prediction markets. When the best weather prediction model says there is an 80% chance of rain, it means that 20% of the time such a prediction is made, there will be no rain. One can't conclude that weather predictions are useless by one observation. So one can't just say that the markets got New Hampshire wrong and therefore they are crap.

Still, I admit I haven't been all that impresssed by Intrade's performance in these primaries. Not like I was on the evening of the midterm elections. While all the pundits on TV were saying that the Democrats would have to win all three of the last three Senate races to be called to win the Senate, the Intrade value of the Democrats winning control went over 50%. NOT ONE expert I saw -- and I watched many channels -- said it was likely the Democrats would win all three, but the market said so and it was right.

I know my defense of this stuff will cause me to be flamed. My only response is: if the market is so bad, then it would be easy to outpredict it and make tons of money. Go forth!

--RiMac

Posted by: RiMac on January 26, 2008 at 2:13 PM | PERMALINK

Hey, I made this point almost 4 months ago...

http://www.dailykos.com/storyonly/2007/10/4/14052/1675/771/393819

Posted by: DaveOinSF on January 26, 2008 at 2:13 PM | PERMALINK

Intrade markets provide an excellent window into the minds of the sorts of persons who are willing to stake money on Intrade markets.

Outside of that function, they are completely worthless.

Posted by: lampwick on January 26, 2008 at 2:17 PM | PERMALINK

However, they do aggregate perceptions pretty well, and apparently there's a strong perception that Hillary is more electable than Obama, regardless of whether that's objectively true.

Maybe I'm missing something, but I really don't understand how betting markets are supposed to reveal anything that can't be captured in a reliable poll of likely voters. Presumably the idea is that you start getting away from peoples preferences ("I like Candidate X") and into their expectations ("... but Candidate Y is really going to win"), making this somehow a more realistic appraisal of the situation. But I think it's questionable, not only because of the reactive nature that jc points out, but because I question the presumption that the people placing these bets are in any way representative of the voting population. Whatever they may be telling us about the election might be interesting, insofar as it relates to the expectations of a demographic that gambles on line (mostly white guys between 20 & 50, I would guess), but I'm not sure that it tells us much about what the rest of the country expects will happen in the election.

Posted by: junebug on January 26, 2008 at 2:18 PM | PERMALINK

InTrade accurately predicts the past, not the future. Check out the wisdom of inTrade prior to New Hampshire to see how accurate you think they are.

Posted by: BombIranForJesus on January 26, 2008 at 2:26 PM | PERMALINK

These markets have already proved their worthlessness as predictors. All they do is provide a way of averaging together the conventional wisdom, with more weight attached to strongly held beliefs (beliefs so strong that people are willing to bet on them).

These markets "predicted" that Hillary would win Iowa, and that Barack would win New Hampshire.

Posted by: Joe Buck on January 26, 2008 at 2:32 PM | PERMALINK

Kevin,

I think that performing math like this to tease out whether people think Hillary or Obama is more electable is actually not empirically sound.

People aren't nearly as rational as you think when it comes to probabilities. One example is the so-called conjunction fallacy.

The conjunction fallacy is a logical fallacy that occurs when it is assumed that specific conditions are more probable than a single general one.
The most oft-cited example of this fallacy originated with Amos Tversky and Daniel Kahneman:

Linda is 31 years old, single, outspoken, and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations.

Which is more likely?

1. Linda is a bank teller.
2. Linda is a bank teller and is active in the feminist movement.

85% of those asked chose option 2. However, mathematically, the probability of two events occurring together (in "conjunction") will always be less than or equal to the probability of either one occurring alone.
Posted by: frankly0 on January 26, 2008 at 2:37 PM | PERMALINK

Another prediction market that has been fun to watch is the one that is run by the business school of the University of Iowa:

http://iemweb.biz.uiowa.edu/quotes/Nomination08_quotes.html

It was running 55 to 41 in favor of Hillary on the morning of the Nevada Caucus. By the end of the day it was running 61 to 33 in favor of Hillary.

Posted by: emmarose on January 26, 2008 at 2:49 PM | PERMALINK

I think John Edwards is the most electable. He has consistent, deeply held beliefs, acts on them, and doesn't get rattled easily.

He's also talented and smart, but those characteristics are not differentiators.

Posted by: Deggjr on January 26, 2008 at 3:08 PM | PERMALINK

Multiplication of independent sequential probabilities doesn't work here. Bettors are wagering on (2) seperate questions. These are independent wagers rather than sequential probabilities.

The bet that counts the most is that 55% wager she will not be elected, 45% bet she will. To a certain extent, the yes/no bet on nomination is 'encapsulated' in that wager.

Posted by: Buford on January 26, 2008 at 3:09 PM | PERMALINK

I don't even think one can use this in the limited fashion KD is here in terms of "aggregate perceptions" in the wider public. Within a niche of that public maybe but that is as far as I can see it going. This is far too open to manipulation by partisans within each camp to in my mind be useful as anything more than as a way to gamble your money with on politics, sorry.

Essentially meaningless in my view, and most certainly not a proven tool to predict anything other than some are going to make money in the trading and some are going to lose money in the trading.

Posted by: Scotian on January 26, 2008 at 3:13 PM | PERMALINK

The intrade congressional markets were abysmal in 2006. You could have made a bundle betting that the house would go Democratic. (On the Senate, even more so, but that really was filling an inside straight.)

I was watching the bettors on election day. They were mostly a bunch of Walter Mittyish Republican dittoheads, and insanely overconfident. When the Senate was called for the Dems, their heads exploded.

Posted by: mac on January 26, 2008 at 3:16 PM | PERMALINK

I think John Edwards is the most electable. He has consistent, deeply held beliefs, acts on them, and doesn't get rattled easily.

Funny, I thought he was the candidate who apologized the most for the votes he made in the Senate.

Posted by: Foo on January 26, 2008 at 3:24 PM | PERMALINK

Can we deduct a few points from Hillary as a penalty for having to listen to the Repubs talk about the Rose Law firm, the blue dress, Vince Foster, the "W" key, etc., etc. for 8 months?

Posted by: Nannyberry on January 26, 2008 at 3:26 PM | PERMALINK

I came in here to say what lampwick said.

"tell us pretty reliably that people think she's more electable." That depends on your definition of "people" Kevin. More accurately, it's people that would spend money at intrade, which I suspect is not really representative of "people" as "people" commonly think of it.

Posted by: jerry on January 26, 2008 at 3:30 PM | PERMALINK

Hillary will win it all because McCain will either
become too old/ill to continue, or his true bitter and angry self will become universally evident prior to November 2008.

Posted by: Rula Lenska on January 26, 2008 at 3:33 PM | PERMALINK

I'm surprised this hasn't come up before. CNN itself is running a prediction market; seems like there have been tens of thousands of trades in the various markets they are running - probably a little more representative of the electorate than INTRADE which is likely a bunch of 30/40-something men. :) That's not the diverse crowd you really need says Wisdom of Crowds theory.

Posted by: Sandy Rodgers on January 26, 2008 at 3:50 PM | PERMALINK

Unless America has changed overnight into a color-blind society, the answer is very simple.

Posted by: Realist on January 26, 2008 at 3:51 PM | PERMALINK

Please, don't tell me anyone needed to have those probability formulas explained to them.

Posted by: Jim W on January 26, 2008 at 3:57 PM | PERMALINK

Either you don't want free money or you don't actually believe what you are saying. If you really believe that Intrade's predictions of the elections (or anything else) are "worthless," then you can make tons of money at Intrade by putting some money behind your oh-so-much-better predictions. Is there a single person trashing Intrade who is willing to do this? Talk is cheap.

--RiMac

Posted by: RiMac on January 26, 2008 at 4:02 PM | PERMALINK

Either you don't want free money or you don't actually believe what you are saying. If you really believe that Intrade's predictions of the elections (or anything else) are "worthless," then you can make tons of money at Intrade by putting some money behind your oh-so-much-better predictions.

If this represents the logical ability of the average Intrader, it must truly be crap.

Posted by: Boronx on January 26, 2008 at 4:20 PM | PERMALINK

The math is wrong. The nomination and the presidency are not independent probabilities. Those betting on the presidency by definition already are also factoring the probability of the nomination.

Posted by: Benjamin on January 26, 2008 at 4:23 PM | PERMALINK

"Using the same arithmetic, this means is that if Obama wins the nomination, Intrade bettors think he has a 46.9% chance of winning the election."

I don't think this is right, because the set of people buying shares in the nominee differs from the set of people buying shares in the general election. So, while the market as a whole may believe this, its bettors might not.

Posted by: Matt on January 26, 2008 at 4:31 PM | PERMALINK

AFAIK, the equation is incorrect.

It doesn't take into account how loyal voters are to Party (E.g., if Clinton wins nomination, does she get all Obama's voters?)

Plus, there are 203% of voters involved in the "party nomination" totals, and only 100% of voters involved in "president elected."

Maybe adjusting for party affiliation in the first groups will help. This would make the adjust the 203% figure back to 100%. Assuming complete party loyalty, you'd force the Dem nominee total to add up to 62%, since that's how many think any Dem will win the Presidency. And force the Rep numbers to add up to 36%. 2% goes to the Paultards and Bloombergs.

So, Adjusted chances of winning party nomination:
Clinton [67.5% * 62%] = 42%
Obama [29.4% * 62%] = 18%
Other D [ 3.1% * 62%] = 2%
McCain [50.8% * 36%] = 18%
Romney [34.6% * 36%] = 12%
Other R [14.6% * 36%] = 5%

Then,

Ratio of [Winning Election]/[Winning Nomination]:
Clinton 1.05
Obama .86
Other D .82
McCain 1.09
Romney .99
Other R .41

The front runners have a much higher ratio than the 2nd tier candidates, which I'd conclude shows again that people are risk averse. They will go with the sure bet (E.g., the front runner of each party wins the elections) when there are many complicating factors.

Nothing to see here.

Posted by: absent observer on January 26, 2008 at 4:32 PM | PERMALINK

Your math assumes independent variables. Investors may believe that it would be very difficult for a candidate to win a primary but, if they did, they would have an easy general election victory. This would not work with the equations you provide.

Posted by: Joe on January 26, 2008 at 4:35 PM | PERMALINK

The conditional probability formula does not require independent variables. In fact, it is a test of independence. These two are the very definition of conditional.

Posted by: Matt on January 26, 2008 at 4:38 PM | PERMALINK

I want to say something in defense of the prediction markets. You can't really tell if they work or not by looking at one individual event, since the defenders of the markets can always respond by pointing to other events the markets got "right". An anecdote simply is not data. And it's far from clear they got New Hampshire "wrong": they were predicting an Obama win with about 85% probability, which means that given the evidence available you'd expect Obama to lose 3 times out of 20, which isn't that unlikely—it's slightly more likely than rolling 8 on a pair of dice.

The way to test the Intrade markets is to look at a large number of historical data and see how close their probabilities are to the number of times things happen. For example, if they were perfectly accurate, 40% of bets that traded at 40 cents on the dollar should have paid out, 75% of bets that traded at 75 cents on the dollar should have paid out, etc. I'm sure there are studies like that, and I have no idea what they show. But the arguments being leveled against the markets here are just not very good.

Posted by: Elliot Reed on January 26, 2008 at 4:39 PM | PERMALINK

Lol @ the idiot who uses a futures market to try and predict an election.

Posted by: soullite on January 26, 2008 at 4:46 PM | PERMALINK

If you really believe that Intrade's predictions of the elections (or anything else) are "worthless," then you can make tons of money at Intrade by putting some money behind your oh-so-much-better predictions.

Thinking that InTrade's predictive capacity is nil doesn't necessarily mean we think our own predictive capacity is better. If some delusional guy has a "system" for figuring out the results of a coin flip, it doesn't follow that we can beat him by betting against him. It just means that he doesn't comprehend the fact that coin flips are unpredictable.

Posted by: jimBOB on January 26, 2008 at 4:57 PM | PERMALINK

For my money they're both no good.

What it comes down to is that Hillary Clinton is the can-do executive and Barack Obama is the can-inspire leader and that their policy views are almost indistinguishable. But what are their politics? What do they actually believe in?

Both of them have so hedged their bets or are so maddingly vague or are so small-bore policy-wise as to be meaningless: maybe we’ll get out of Iraq and maybe we’ll leave bases and trainers/Blackwater and the air force stay, maybe we’ll have some sort of health care for America’s uninsured and underinsured and maybe we won't, perhaps we need to raise taxes or we need to cut taxes, maybe so-called illegal immigrants must return or perhaps they can stay, and their views on how to help stem the foreclosure mess are unknown.

Compare this to the Republican side of the aisle where the full throated, red meat ideologies of God, war on terror, starve the government, free markets and tax cuts for the rich thrive and rally its supporters: it is this that gives the party its vehemence, its strength. Republicans can put most of their message on a bumper sticker: ”Markets work, governments don’t.”

The democrats have no such strong message. It helps explain the tepid campaigns of both Kerry and Gore. Indeed it continues to this day and explains why both Clinton and Obama lack a strong political ideology and just spin their wheels on race, gender, change vs. experience, she’s a liar, he’s a liar issues.

Clinton might have the executive ability to get things done and Obama might have the inspirational leadership to get things done but they both beg the question: what is their forceful, put it on a bumper sticker message of what they want to get done?

You know what the problem is with modern American politics? It's corporate money driven. As ABSCAM Congressman Ozzie Myers famously remarked, “Money talks, bullshit walks.”; it ought to replace “E pluribus unum” on our currency.

Posted by: Dr WU-the last of the big time thinkers on January 26, 2008 at 5:01 PM | PERMALINK

jimBob, your general point is valid, but your delusional guy would probably be willing to take odds greater than 50/50, and, over the long haul, you'd profit from him.

Posted by: matt on January 26, 2008 at 5:03 PM | PERMALINK

Elliot Reed

So when are we supposed to "count" the predictions of these markets? Before or after their predictions switch massively in response to some event (say, the New Hampshire primary)? Or just before the election? If so, then why are we paying any attention to the predictions months out, before they "count?"

Posted by: jimBOB on January 26, 2008 at 5:03 PM | PERMALINK

These markets are completely reactive, thus, not in the least predictive. After Iowa Obama futures were hot after N.H. Clinton went back up to where she was. Clinton futures have actually dropped a bit the last few days.

Posted by: Chris Brown on January 26, 2008 at 5:22 PM | PERMALINK

Of the available candidates, HRC is the only one with a chance of winning the election.

I'm in two ways about this Dem process.
Given the absolute mismanagement by the GOP and the WH of just about everything they've had anything to do with, we're at a point where the electorate seems ready to hand the Dems a landslide victory in the races for the WH and Congress.

At the same time, the only party you'd think would ever field a female or black candidate for the presidency is the Dems, and one could argue that the only chance they have of seeing that be successful is when the GOP has painted itself into a corner.

It's a chancey game, though, as Rove must be somewhere on his knees, praying and hoping that Obama gets the nomination. Rove then knows he has the WH in GOP hands.
With Clinton as a candidate Rove has a fight on his hands.
And if Edwards hadn't faded against the two "watershed candidates" he would have been the candidate and Rove would have been really worried.

And this has me in two ways - I wish the Dems had fielded a shoe-in candidate, thus ensuring the landslide victory. On the other hand, it's about time you got a woman or Afro-American as president.

Good luck!

Posted by: SteinL on January 26, 2008 at 5:47 PM | PERMALINK

I was always a lousy math student. But In-Trade futures seems a risky proposition upon which to base any assessment of electability. I might semi-agree with the Clinton assessment, but Obama's seems somewhat undervalued.

Posted by: Donald from Hawaii on January 26, 2008 at 5:50 PM | PERMALINK

I don't know how long Intrade has been in business, but it seems that there would be a good scientific paper in showing how the results of Intrade betting compare overall to standard polling techniques.
There are several interesting questions that might (or might not) be answered:
- Comparing the date of collection of data for poll against odds at that time. Any correlation?
- Comparing date/s for any candidate being sold "short" with standard polling. This would require correlating the release of damaging information about the candidate/s. Guiliani's Shag fund for example.
I don't know if it would "prove" anything, but the results could be very interesting. Of course. if it turns out that the bettors are more accurate than the pollsters, what do the latter do?

Posted by: Doug on January 26, 2008 at 5:57 PM | PERMALINK

So Kevin?

Why weren't you talking about the betting markets after Iowa? Just a question. We all know who you support.

Also -- If you really think it is legitimate to assumption independence for those probabilities you are crazy.

For most Hillary supporters:

P(Hillary elected|Hillary nominated) = 1

which kind of throws things off.

Posted by: Adam on January 26, 2008 at 6:05 PM | PERMALINK

Unless the people active on Intertrade are a representative random sample of all people who are going to vote than it isn't worth shit.

Economist assume that markets are statistically efficient or BLUE (Best Least Unbiased Estimator), however there is no actual scientifically credible evidence that they are.

So this fetish about markets ability to "predict" Presidential election outcomes, or for that matter deriving all the relevant information to allow for the proper allocation of economic resources, is just cargo cult science.

Posted by: Dr. Morpheus on January 26, 2008 at 6:06 PM | PERMALINK

Rasmussen has a free version of Intrade. Lots of fun. And informative. Based on this post, I went long Obama as President and short Obama as nominee at a ratio of 16/-30. If Obama goes down I make about 30-16=14 points. If he is winning the nomination, his probs to win the presidency will rise faster and I shd be able to cover my trade. I'm also short Paul, Rudy and Huck. I think Edwards will be gone by Feb 29th. Even though I have made about 8% in about a week, I'm only ranked 400. It's a big crowd.

BTW Mitt was surging until today....he's back down close to even with Mac.

Posted by: wren on January 26, 2008 at 6:07 PM | PERMALINK

JimBob--I don't see why you'd have to pick any one point. I'd think if you wanted to study this rigorously you'd want to check accuracy at various points in time and see if and how much the predictions get better (in the sense of matching the true probability) as you get closer to the event being predicted.

Posted by: Elliot Reed on January 26, 2008 at 6:09 PM | PERMALINK

The markets are probably even less reliable on derived, secondary statistics like Obama's or Clinton's spread, than they are on the outcomes people are directly betting upon. If a futures product were crafted whereby traders could bet directly on the derived statistics you point to, I imagine Obama's electability would be closer to Clinton's. That the anomaly exists only means the market hasn't thought through the implications of its supposed collective wisdom.

Posted by: kth on January 26, 2008 at 6:11 PM | PERMALINK

Wow. I guess a lot of people do need the equations explained. Some people even think they are wrong!

Posted by: Jim W on January 26, 2008 at 6:14 PM | PERMALINK

Prediction markets are no more reliable than the internet polls they put up on CNN.com, for the same reason: participants are self-selected.

Posted by: nota bene on January 26, 2008 at 6:22 PM | PERMALINK

This, I think, is rubbish.

People are emotional beings and not computers

Posted by: Jet on January 26, 2008 at 7:08 PM | PERMALINK

Total crap! cleve

Posted by: cleve on January 26, 2008 at 7:11 PM | PERMALINK

Obama winner South Carolina. Congratulations.

Posted by: Mike on January 26, 2008 at 7:26 PM | PERMALINK

JimBob:

Thanks for making an argument rather than just calling something crap. You say that the predictive capacity of Intrade may be very bad and yet your own predictive capacity may be no better. So you could make no money betting on it, yet consider it worthless.

Good point. I have two responses.

First, note that unlike you many commenters argue that Intrade is obviously wrong about this or that. Its participants are biased or have a narrow perspective. My comment is logically true only as applied to these commenters. If a person feels strongly that Intrade is systematically wrong about valuing something, he or she should be able to make money on it.

Second, as an empirical matter, prediction markets are simply not that bad at predicting. Perhaps it would offend people less if we focused on Vegas odds making for sporting events or the odds in track betting on literal horse races. The bet taker "adjusts the odds" based on how people are betting just as Intrade's prices adjust based on how people bet. In both cases, we have a "market" where the price of (or return to) betting on a favorite is higher than betting on a long shot. And the price or odds make a prediction. And it turns out that sports betting markets are extremely good predictors of sports outcomes. Not compared to perfection of course; just compared to any other human actor. If you accept this for sports, there is no intrinsic reason why it can't be true for politics.

Indeed, the academic literature finds that predictions markets can be extremely accurate. Here is a link for one paper (from an economist who also studies sports betting markets): http://bpp.wharton.upenn.edu/jwolfers/Papers/PredictionMarkets(Palgrave).pdf

Here is one paragraph from the paper: "Figure 3 compares the forecasting performance of the Iowa Electronic Markets [which predates Intrade by many years but now share very similar predictions] and the Gallup Poll in predicting the outcomes of Presidential elections in the United States. Over the 13 candidacies from 1988-2004, the average absolute error of the market-based forecasts was 1.6 percentage points, while the corresponding number for the Gallup Poll was 1.9 percentage points. . . . The initial success of these forecasting methods in the United States has led to similar analysis of the forecasting power of prediction markets in Austria, Australia, Canada, Germany, the Netherlands and Taiwan."

--RiMac

Posted by: RiMac on January 26, 2008 at 7:48 PM | PERMALINK

My mother is a computer, you insensitive clod.

Posted by: absent observer on January 26, 2008 at 7:48 PM | PERMALINK

Brokers in the stock market don't care whether it's going up or down, just as long as it's moving. Are Intraders buying Edwards votes today with the expectation he'll look slightly better tomorrow and hence earn them some money? If it's anything like other futures markets I would think that's all that matters, ie. is this particular candidate going to look better tomorrow long enough for me to profit from the move. Who cares about the results 10 months from now. Between now and then there are all those moves to profit from.

Posted by: dennisS on January 26, 2008 at 8:35 PM | PERMALINK

Bill Clinton is a piece of shit.

Posted by: Econobuzz on January 26, 2008 at 8:41 PM | PERMALINK

Yes, South Carolinians don't agree with the In-trade crowd. Good for them.

Posted by: Neil B. on January 26, 2008 at 8:54 PM | PERMALINK

@RiMac

can you repost the link? that one doesn't work.

My comment -- not having seen the paper:

1) If the predictions are a day or two before the election then I'm not suprised that they are accurate.

2) That's because those prices should factor in all the information contained in the polls immediately before election. Which are going to be the dominant input in the prices.

3) If you wanted to see if intertrade works as a prediction market then you need to seperate out the effect of the polls. Perhaps set up a market with access to all non-poll information and see how well it works. In the absence of such a market you're not proving anything.

Posted by: Adam on January 26, 2008 at 9:04 PM | PERMALINK

I'm making money on my arbitrage....and I'm happy Obama won big tonight! It's 2008. Time for a change.

Posted by: wren on January 26, 2008 at 9:14 PM | PERMALINK

We've really come down to betting on nominations, and now Drum's blogging on it with all the salivating intensity of a 15 year old on Speed playing Rock Band.

What a waste of time and bandwidth.

Posted by: Quinn on January 26, 2008 at 11:09 PM | PERMALINK

So far in this campaign, the candidates have spent the overwhelming majority of their time in Iowa, New Hampshire, and South Carolina. Obama destroyed Clinton in two of these (one with only white voters), and came close to a tie in the other. Don't the results in these three states already say a lot about which candidate would wear the best in a general election?

Posted by: Matty G. on January 26, 2008 at 11:53 PM | PERMALINK

Adam:

If you go to this page, http://bpp.wharton.upenn.edu/jwolfers/research.shtml#PM, which lists Wolfers' papers, the fourth paper under the section on "prediction markets" is the one I referenced. Basically I just googled "prediction markets research" and found this.

In response to your three points:

Sure, it is easier to be accurate just before the election than earlier. But that is also true for polls. The sentence I quoted compared the prediction market to polls for a great many elections. They picked the last poll done before the election, which you would think would be the poll that best predicts the election. Yet the prediction market at the time of the last poll consistently outperformed the poll. I agree it would be nicer to have even more evidence, such as a comparison of the market to a poll one month before an election, but this evidence is a start; no one else has offered any.

There is no reason to prove that the markets work in the absence of polls. The idea behind prediction markets is that they are good at aggregating informaton. If there is no information to aggregate, then of course the market will fail. But if the market incorporates poll data and produces something better than the poll, then what's silly about that?

--RiMac

Not by a huge quantity, but that's precisely because the comparison is made right before the election. But if the market beats the poll right before the election by a little, it probably beats the poll by a larger margin a longer time before
If we compared one month before the election and the market at the same time, but I haven't seen that data, though I really haven't read much about this).

2) That's because those prices should factor in all the information contained in the polls immediately before election. Which are going to be the dominant input in the prices.

3) If you wanted to see if intertrade works as a prediction market then you need to seperate out the effect of the polls. Perhaps set up a market with access to all non-poll information and see how well it works. In the absence of such a market you're not proving anything.


Posted by: RiMac on January 27, 2008 at 12:15 AM | PERMALINK

Given that polls can gyrate wildly in the months preceding an election, if the markets beat them for accuracy it still may not be terribly meaningful. And to be truthful, this isn't really an important knock against polling or markets - both are highly dependent on minutia and chance events that occur day to day. Inevitably they'll contain a lot of noise.

As long as we're talking about predicting election results, I wonder if you have seen this guy's stuff. His predictive system:

• Works as far out as a year before the election

• Includes NO polling data

• Has prospectively predicted the popular vote
winner in all presidential elections since it was created, from 1984 onwards

• Has retrospectively predicted the popular voter winner in all presidential elections as far back as 1860

It's an interesting system, with, to my layman's eyes, a far more accurate record than either polls or markets, at least when considered from the the vantage point of months before the election occurs.

Posted by: jimBOB on January 27, 2008 at 1:11 AM | PERMALINK

Some interesting trades seem possible in this market, though apparently you cannot bet serious money because trading volumes are low.

Looking at the history of the SC primary contracts, you could have bet on Obama at 80% of expiry price only a week or two ago. That seems like it would have been a good trade (25% profit if you were sure Obama would win). No similar opportunities are available for the upcoming primaries yet, as far as I can tell.

It's also possible to make good bets without predicting the winner exactly:

Right now, you can bet that the Republican nominee will be one of the two front-runners (McCain or Romney) for about 85% of expiry price (for a few contracts) or 90% (if you want more contracts). This assumes you buy both the McCain (55%) and the Romney (29%) contract.

Bid/ask spread for the second bet above seems reasonable for about 100 contracts ($1,000) each. Not clear what the SC Dem primary bid/ask spread was, but judging from volume it must have been tight.

Posted by: JS on January 27, 2008 at 1:34 AM | PERMALINK

Multiplication of independent sequential probabilities doesn't work here... The bet that counts the most is that 55% wager she will not be elected, 45% bet she will. To a certain extent, the yes/no bet on nomination is 'encapsulated' in that wager.

This seems to be based on a misreading of what Kevin did: He did take the 45% as the combined probability of nomination and final election (which recognizes the "encapsulation" mentioned). From this, and from the independent probability of nomination, he derived the conditional probability of election if nominated. A correct computation -- though, as Kevin made clear (and many in the comments seem to have ignored) these numbers show current perceptions, not probabilities of actual outcomes.

Posted by: JS on January 27, 2008 at 2:05 AM | PERMALINK

As long as we're talking about predicting election results, I wonder if you have seen this guy's stuff.

Impressive indeed. Over 50% profit can be made on intrade right now if his prediction (the the Dems will win in November) is correct.

Posted by: JS on January 27, 2008 at 2:19 AM | PERMALINK

I am deeply sadden by Caroline's endorsement of Senator Obama. I understand that she feels he is inspiring, but what we need now is more than just inspiration.

I have to wonder if Caroline would have endorsed Senator Obama, if Jackie Kennedy Onassis was still alive. Jackie was very fond of the Clinton's stating how Bill and his visions reminded her of her husband. Jackie was a friend to Hillary often sharing stories of raising children in such a public eye.

Upon her death, John F Kennedy,Jr. sent the Clinton's a hand written letter that said " I wanted you both to understand how much your burgeoning friendship with my mother meant to her. Since she left Washington I believe she resisted ever connecting with it emotionally-or the institutional demands of being a former First Lady. It had much to do with the memories stirred and her desires to resist being cast in a lifelong role that didn't quite fit. However, she seemed pro fondly happy and relived to allow herself to reconnect with it through you. It helped her in a profound way- whether it was discussing the perils of raising children in those circumstances or perhaps it was the many similarities between your presidency and my fathers".

It appears that the Clinton's not only inspired a nation, but also a former First Lady. Though I respect Caroline's decision to talk of Obama's inspiration and his great work in Chicago. It still does not take away from the facts, that his "resume" comes up short!!! Americans need more than inspiration after the Bush administration.

Americans need a proven leader who has traveled the "World" building relationships with foreign leaders. We need someone who has dedicated their entire adult life to helping the under privileged in our Country. Someone who has fought for civil, human and yes, woman's rights. Not only here in America but across the globe. We need a proven voice that can reach out across party lines. We need someone who is battle tested. We need change for our failing economy. We need universal heathcare. We need jobs. We need to focus on our energy crisis. We need "our" men and woman home. Americans need a leader. Americans need Hillary Clinton!

Though Obama is a beautifully spoken and inspiring gentleman. I am inspired by the years of hard work and dedication that Hillary brings to the table. Kitchen table at that! I am inspired by her ability to stand strong during the media's continuous bias. I am inspired by a woman who has been fighting for American since before I was born. I am inspired by her continued wiliness to make America better to help make Martin Luther King Jr. dreams a reality. The speeches she heard LIVE all those years ago when segregation was thought to be impossible there stood in Chicago a very young white girl, on a mission to learn and help. Hillary has dedicate her life to a cause so great it has empowered her during some of the toughest times. When most would have given in, she stands to fight for us again and again. That to me, is inspiration. Hillary Rodham Clinton is an inspiration to many Americans and Many Woman all over the WORLD!!!

Today I am endorsing Hillary Clinton. My voice is only one and I am not famous. I will not be on front pages of the dailies across the Country. I am not a Kennedy who swam along side of Chelsea, with her Mom and Bill and Hillary at Menemsha Harbor. I am not
"a perceptive friend and role model to Chelsea", as Caroline Kennedy was regarded.

I am an American endorsing the only other American running who is qualified to lead and passionate enough to never forget us.

I am an Inspired American from Kentucky and I am proud to endorse, support and volunteer for the 44th President of the United States, Hillary Rodham Clinton.

My name is Robin Davidson and I not only approve this message, but I wrote it.

Posted by: Robin Kentucky on January 27, 2008 at 2:41 AM | PERMALINK

"Bettors think Obama is likely to lose the election..." An unfortunate choice of words. If one went to an Indian Casino and said - I like the way you treat your bettors - one would likely go home with less hair.

Posted by: nonheroicvet on January 27, 2008 at 8:00 AM | PERMALINK

Watch Clinton's value plummet and Obama's skyrocket tomorrow.

Posted by: Chris Brown on January 27, 2008 at 12:37 PM | PERMALINK

InTrade is not a reliable predictor.

I've won some good dough betting against the spread. My best take was betting that John Bolton would never be confirmed.

I tend to agree that there's some manipulation there, but I suspect that the herd mentality prevails.

I've got Obama across the board for a significant part of my prior winnings - and I expect to do quite well.

Posted by: Jaime Frontero on January 27, 2008 at 1:00 PM | PERMALINK

Look, Kevin, these trading sites are no better than an internet poll. In fact, that's exactly what they are. You might as well poll your commenters here to see what they think as look at the trading figures. Traders are idiots, too (cf. recent stock market).

Posted by: David in NY on January 27, 2008 at 1:04 PM | PERMALINK

David in NY:

Do you really think that the odds set in Vegas for sporting events are "no better than an internet poll" for predicting what team will win? The key difference in an internet poll and a prediction market, which is what the Vegas odds-makers create, is that people put up money in the market but not the poll. Talk is cheap; betting is not. That is why prediction markets beat even the most scientifically conducted polls, not just internet polls. I cited some evidence for this point in a prior comment.

--RiMac

Posted by: RiMac on January 27, 2008 at 11:57 PM | PERMALINK
The second is the question of whether betting markets like Intrade do a good job of predicting events like elections in the first place. I'm not up on the relevant literature, but I suspect the answer is that they probably don't perform very well in an objective sense.

They do (or did, when the first studies were done): if you look at the market state immediately before the event predicted. Any early than that, and they don't work very well. Of course, looking at polls right before the event works pretty well.

However, what they can do is aggregate public perceptions well.

No, they can't. First, the participants are self-selected, so the population being aggregated over wouldn't be "the public" but "political gamblers". Second, for several years there has been major media coverage of them as predictors, so there is a considerable motive to manipulate them to make your preferred candidate appear to be more likely to win, given that a media perception of that likelihood translates into a real increase in the likelihood of victory. Its probably best to see them as a wealth-weighted proxy for political preferences.

(Of course, because perception of prospects drives media coverage, campaign donations, etc., and thus has a positive feedback on actual prospects, if such gambling systems are perceived as good predictors, they will become good predictors, and those who have the money and inclination to manipulate them will have great power to dictate political outcomes.)

Posted by: cmdicely on January 28, 2008 at 5:37 PM | PERMALINK




 

 

Read Jonathan Rowe remembrance and articles
Email Newsletter icon, E-mail Newsletter icon, Email List icon, E-mail List icon Sign up for Free News & Updates

Advertise in WM



buy from Amazon and
support the Monthly