« Free trade websites | Main | Consumption vs. income inequality, revisited »
How off is InTrade?
David Leonhardt weighs in. The other day John Nye and I were discussing that de facto limits on the size of effective bets are the biggest problem hindering effective price discovery. When you can become a millionaire on InTrade, that's when its prices will become much better forecasters. Nonetheless I agree with the piece's conclusion:
If you have any better ideas of where to look, let me know.
But some inefficiencies need to be taken with a grain of salt:
Mr. Ravitch has made a nice profit betting against Ron Paul, the libertarian who late last year was, amazingly, given almost a 10 percent chance of becoming the Republican nominee. “If you asked anyone in politics whether there was ever, at any point, a 10 percent chance of Ron Paul being the nominee,” Mr. Ravitch said, without finishing the sentence. “That sort of makes my case for me.”
When it comes to the long shots, remember that InTrade takes deposits in non-interest-bearing cash rather than T-Bills. In the meantime observers need to adjust their expectations accordingly and not interpret all the prices are pure percentages. I recall Paul trading at about 7 or 8 percent. Let's say you shorted Paul last December. You're locking up your cash for quite a while at zero percent interest and when Paul fails you're not going to net as much as you thought. In other words, there is a partial short sale constraint on this market and its prices need to be understood accordingly.
Of course there is a reason why InTrade insists on earning the float and that directs our attention back to the zero-sum property of the bets. That's another reason why prediction markets probably won't ever forecast as well as the stock market: their users have to be charged or taxed at a higher rate. The expected rate of return in InTrade is negative; the expected rate of return in the stock market is seven percent minus commissions. Where would you rather put your smarts to work?
Posted by Tyler Cowen on February 12, 2008 at 07:20 PM in Economics | Permalink
Comments
how about paying people for making predictions? (i.e. a dollar's bet pays $1.03 or something; or perhaps prize money for top results of bettors). if the information was actually worth anything, it would be worthwhile to make predicting a positive-sum game for the participants.
Posted by: Will Perkins at Feb 12, 2008 8:01:25 PM
Maybe the bets should be denominated not in dollars but in shares of InTrade stock...
Posted by: Grant Gould at Feb 12, 2008 8:02:18 PM
Correction: the aggregate return for Intrade is negative. That doesn't make the expected return negative. (Similarly: poker.) Ravitch is very conscious of the time-value of money, and is making short-term bets (such as shorting Obama-Michigan at 15, when he wasn't even on the Michigan ballot) that permit him to free up his bankroll. And when there are thousands of dollars of dead money in there bumping up Ron Paul to win primaries where he has no chance, that's a lot of money for the taking.
Posted by: Ted at Feb 12, 2008 8:03:01 PM
"That's another reason why prediction markets probably won't ever forecast as well as the stock market: their users have to be charged or taxed at a higher rate. The expected rate of return in InTrade is negative; the expected rate of return in the stock market is seven percent minus commissions. Where would you rather put your smarts to work?"
I disagree that prediction markets will never forecast as well as the stock market, at least in certain areas. For example, the sports prediction market is about as efficient as they come, perhaps because so many people bet on it for fun (giving others a large opportunity for profit?).
Also, an important effect in prediction markets is the aggregation effect among various markets. Even if each market has a small volume and limited liquidity, the predictive power of all markets offering bets on the same events will be aggregated due to cross-market arbitragers. So it's not an issue of InTrade's volume per se, but rather aggregate volume on a given event.
Posted by: Cliff at Feb 12, 2008 8:13:12 PM
Cliff:
Sports prediction markets are invariably bound by the large sums of money that are thrown at the bookies in Vegas. If the Vegas line is trading at a InTrade equivalent price spread of 40-50, then anytime the InTrade price floats outside of that range there is an arbitrage profit(go long with a bookie and short on InTrade) to be made that pulls the InTrade price back in range.
The Vegas bookies drive the momentum and InTrade follows. Within the range prices shift based on traders deciding which side of the spread to stay biased towards.
Posted by: Jay at Feb 12, 2008 8:34:38 PM
I'd rather put my smarts to work where my likelihood of winning is greatest while the prize is large enough to be worthwhile.
Given that InTrade is in its infancy and (to me) resembles the Texas Hold'em World pre-ESPN, there's more likelihood of winning if you're really good (vs. the stock market or present day Texas Hold'em where in an expanded field luck or other strategies not based on pure smarts play a much greater role).
Posted by: meter at Feb 12, 2008 9:15:24 PM
If a 6% return in 3 months isn't good enough for you, how the heck do you explain all the contracts for sale at 0.1%? After commission, you'd be losing money by selling these, and Intrade claims they are not market making...
Posted by: Paul N at Feb 12, 2008 9:41:21 PM
For this reason, Mr. Ravitch has recently been betting that the odds of Mr. Obama’s getting the Democratic nomination will keep going up this month, even though they were already around 60 percent when we spoke late last week. As Mr. Obama wins more primaries, Mr. Ravitch figures the contracts will continue to gain value, and he can then sell them before the March primaries, which look more favorable to Mrs. Clinton.
I'd like to thank Mr. Ravitch for exposing THE EXACT THING I've been doing.
Posted by: Neema at Feb 12, 2008 10:04:44 PM
Jay,
Perhaps you don't realize that bookies are de facto prediction markets? The lines are set by the bettors, not the books. You basically repeated what I just said. Although, Vegas is increasingly irrelevant to sports lines.
Posted by: Cliff at Feb 12, 2008 11:29:36 PM
It would be relatively easy for a prediction market to pay out interest, say from investments in a money market fund. If there was demand for it, people could bet shares of stock indexes. Intrade could, presumably, charge the same sort of transaction fees as stock trading services. It would have to do way more volume before it could really push the cost down, though.
I also think Will is correct; in many instances people are willing to pay for predictions, so charging the traders may not be necessary, and may be counter-productive to producing good predictions for corporate customers.
As soon as prediction markets offer open APIs for all users, arbitrage could be semi-automated.
Posted by: Grant at Feb 12, 2008 11:56:04 PM
There wasn't a 10% for Ron Paul to win the Republican primary in retrospect. Just as it is obvious now that McCain will win the primary, but that wasn't so when the bets were made. In fact, McCain had less of a chance than Ron Paul according to Intrade.com several months ago. Easy to belittle the odds after the fact. That being said, Intrade still suffers from lack of liquidity. It's hard to find contracts (other than at ridiculous prices) for lesser known candidates. Intrade has to reduce the transaction costs to be a better predictor.
Posted by: BiJian Feng at Feb 13, 2008 1:51:00 AM
How important is the lack of liquidity vs. the lack of valid private information in effective price discovery? Can anyone point me to some literature that compares these two components?
Posted by: Allan Friedman at Feb 13, 2008 1:56:16 AM
And some inefficiencies are pretty tasty, even unsalted:
"BTW, another pick I was gonna make that's no good anymore due to the length of time this is taking: For about two weeks, there were ~25 Hillary shares for sale between 80-90 in MI. She's the only one on the ballot there and Obama/Edwards write-in votes don't count, but apparently it took two weeks for somebody else to figure that out and buy them up. Inefficient market ftw."
http://forumserver.twoplustwo.com/showthread.php?t=88375&page=7
Posted by: burger flipper at Feb 13, 2008 3:16:14 AM
I bet a couple hundred bucks that would win, because I thought he was the worst choice as a president and an incredibly strong personal distaste for the man. If Kerry won, I'd be out several hundred dollars but I wouldn't be depressed about the political leadership so the cost wouldnt bother me. If Bush won, I'd be depressed, but at least I'd have several hundred dollars in consolation.
Posted by: Stiven at Feb 13, 2008 8:14:18 AM
The biggest problem with people assessing the "success" of most prediction markets is that most people have problems understanding probabilities.
If every favourite won, the market would actually be pretty inaccurate! Sports bettors understand this, but when it's applied to other fields journalists and others tend to forget about it. I've written more about the phenomenon here:
http://blog.mercury-rac.com/2008/01/25/how-to-interpret-prediction-market-results-on-elections/
I agree that InTrade would likely be even better if the market had more people participating, but that doesn't make it inaccurate.
Best regards,
Jed
Posted by: Jed Christiansen at Feb 13, 2008 8:22:53 AM
bookies are de facto prediction markets... lines are set by the bettors, not the books
I know this is off-topic, but there is a school of thought which suggests that the Federal Reserve behaves the same way. That is, most people are under the impression that the Fed can influence the federal funds rate, just as most people are under the impression that bookies can influence the lines. But in fact, the Fed is merely reflecting (its view of) market predictions of overnight rates over the next couple of months. I can't remember the name of this theory, but I thought the parallel was interesting (since no one disputes that bookies do not, in fact, "set" the lines in anything other than a mechanical sense).
Posted by: Rich at Feb 13, 2008 8:38:23 AM
shouldn't the probabilities from intrade be assumed to be risk neutral probabilities and not actual expected probabilities?
To the extent the contracts represent events with real economic value, we should expect people to use them to hedge rather than to place bets. e.g. buy Hillary contracts if you work in banking and stand to lose if she starts ripping up mortgage contracts.
Posted by: DK at Feb 13, 2008 8:40:48 AM
I would think there would be more turnover of long-dated contracts if something close to a prevailing risk-free rate of interest were paid on tied-up capital; perhaps it would be more straightforward to pay the shorts interest on 100 and allow the contracts to drift upward toward expiration. I don't know whether this would result in enough new commissions for intrade to be happy having done it, though I suspect as the market grew it would.
The game is zero-sum ex post, though need not be ex ante; the same is true of futures markets. If there were more hedgers (like Stiven) then there would be more potential rewards for speculators; whether there are ever likely to be enough hedgers to make the market efficient is yet another question to which I don't have the answer.
Posted by: dWj at Feb 13, 2008 10:41:12 AM
The second problem is that the market seems to react to new information too slowly.
I've commented about this in other places. I expect financial markets to snap near-instantly to news, but it took InTrade a good hour or so to react to the news of Clinton in New Hampshire.
Posted by: Dan Weber at Feb 13, 2008 12:05:51 PM
Cliff:
I was commenting on that the difference between sports markets on InTrade and other markets on InTrade is that there is a much more liquid auxillary market for sports betting than for say betting on political campaigns.
Posted by: Jay at Feb 13, 2008 12:33:00 PM
"The expected rate of return in InTrade is negative; the expected rate of return in the stock market is seven percent minus commissions. Where would you rather put your smarts to work?"
As prediction markets grow more popular and are used to find valuable information (when to fire your ceo, where to locate your company, ect) business will subsidize the markets, $1 bet brings a $1.03 reward as given in an above example.
Posted by: Mason at Feb 14, 2008 10:40:48 AM
I looked into selling contracts at Intrade, but was put off by the time-value of money and high deposit requirements.
Let's say you wanted to sell a contract on Paul at 10 (Predicting he won't be the nominee). You must deposit your maximum loss with Intrade (about 90) for the entire term of the bet, which might be a year.
I can get 24%+ annually doing Hard Money loans that are backed by real property worth 2x what I'm loaning. Heck, I can spread my risk and get 14%+ from a Hard Money fund.
Why would I "risk" money on even a sure thing (not even discounted for risk of losing) in exchange for waiting a year for at most a 10% return with no possible upside?
Their model just doesn't facilitate real investing in large amounts, and it won't until them come up with a mechanism for investing on margin where you can pledge collateral of some sort (including intrade contracts that are worth something) instead of tying up large amounts of cash.
Why was Ron Paul at 10? Lack of supply in the market. Imagine if you sold a contract at 5 instead. Then you make 5%/year on your money. You might as well make a virtually no-risk investment in a CD or Bond. I think 10 reflected the length of the contract and the time value of money more than it reflected the actual odds of getting the nomination.
Posted by: Sharper at Feb 15, 2008 10:11:55 PM
