>
"What you’re doing is collecting bits and pieces of information and
aggregating it so we can watch it and understand what people know. People picked this up and
called it the ‘wisdom of crowds’ and other things, but a lot of that is
just hype."–California Institute of Technology economist Charles Plott
>
Interesting piece in the WSJ on prediction markets and politics, Traders’ Calls Just as Bad On Elections. (I may have mentioned something about this in the past).
"John McCain’s presidential campaign is doomed — at least, if you still believe what political futures markets indicate. At the Irish electronic exchange Intrade, on which people bet on election outcomes and other events, the futures market suggests Mr. McCain has a 38% chance of becoming the 44th president. In the Iowa Electronic Markets, set up at the University of Iowa, Mr. McCain’s Republican Party gets a 41% chance of winning the popular vote for the White House.
Then again, six months ago, the Iowa markets gave Barack Obama less than a 30% chance of winning the Democratic nomination. Academic studies suggest these markets are more reliable than opinion polls, but that might be giving the markets too much credit.
Intrade futures had John Kerry beating President Bush well into the evening of Election Day 2004. They also said there was a good chance Mr. Obama would top Hillary Clinton in January’s New Hampshire primary, which she won."
The article details many of my favorite quibbles: thinly traded, plagued by bad information, skewed participation, bubbles, head-fakes and manipulation.
What did it reflect when all those people bought all those Hillary Clinton and Rudy Giuliani presidential futures when each was a front runner? Somehow, the phrase "Wisdom of Crowds" just doesn’t seem to capture the full essence of that . . .
Previously:
Iowa and Prediction Markets, January 24, 2004 http://bigpicture.typepad.com/comments/2004/01/iowa_and_predec.html
Why Prediction Markets Fail January 11, 2008 http://bigpicture.typepad.com/comments/2008/01/prediction-mark.html
Misunderstanding Prediction Market Failures February 14, 2007 http://bigpicture.typepad.com/comments/2007/02/misunderstandin.html
Source:
Traders’ Calls Just as Bad On Elections
MARK GONGLOFF
WSJ, May 13, 2008; Page C1
http://online.wsj.com/article/SB121063385437486555.html
The fix is in. Forget about the elections.
I am all for some voter ID. But proof of citizenship? That is hard to prove for an average person without putting in a lot of effort. And yes, I would like to know which illegal immigrant is going to invite trouble by trying to vote. Who are these people kidding? These measures smack of other motives.
http://www.nytimes.com/2008/05/13/opinion/13tue1.html?ref=opinion
Anybody believing in the “Wisdom of Crowds” has never watched a British soccer game or observed a full nightclub when some idiot starts waving a gun around. Wisdom my ass.
Yes, there are problems with prediction markets like Intrade. For example, a lot of the contracts have very low volume, are not very liquid, and the bid/ask spreads are quite large. This means that they can be not very accurate and can be manipulated.
But it is well known that prediction markets become more accurate as more people participate. There are several other conditions that must be met for them to work well. When these conditions are met, the markets are very good at aggregating information, and even if they’re wrong, they represent the crowd’s best guess of the outcome *using the information currently available*. With new information and developments, the possibility of a given outcome changes–and with it the price.
As for New Hampshire, there might have been some funny stuff going on there. I think Kucinich tried to get an investigation going and some of Ron Paul’s people participated in the recounts, but I don’t know what came of it. Unlike some of the other contracts, the presidential election contracts have high volume and should be pretty accurate.
One minor disagreement here. These markets shouldn’t really be called predictions markets. They’re really odds-setting markets. Just because a favored outcome didn’t come to pass doesn’t mean that the markets somehow “failed.”
Google’s IPO was at $85. That’s about $500 off today’s price. Did the markets fail? No, the markets adapted to new information.
Though I do agree with your quibbles about these markets.
Was Google $85 six months ago ?
Look at how recently Rudy and Hillary were the frontrunners on Intrade.
Kerry won in 2004. The fact that he isn’t in the White House has to do with OH. OH, Oh-hio.
The thinly traded argument is, however, the best one against believing in these betting booths.
Sorry BR. I agree with Eddy Elfenbein. Just because an expected outcome did not come to pass does not mean that the odds-makers were imprecise to begin with.
Circumstances change with time and unexpected outcomes do occur.
Barry you’re wrong on this one…
The unexpected SHOULD happen once in a while…
~~~
BR: Ok, I’ll bite:
1) When? How often should the outlier occur?
2) What does that do to the value of the prediction markets?
Intrade had Clinton winning Indiana.
Circumstances change with time and unexpected outcomes do occur.
Adam, wouldn’t the whole point of predicting something involve anticipating changes with time over some amount of time?
And wouldn’t an “unexpected outcome” then represent a failed prediction?
The wisdom of crowds? Hmmm. Let me think of some examples of wise crowds in history:
-1929 crowd of stock speculators
-1933-45 crowd of German Nazi facilitators
-1970’s crowd of “nifty-fifty” stock investors
-1980’s crowd of oil and gold speculators
-1960’s crowd of Malthusians claiming there would be mass starvation by the 80’s, then by the 90’s, then by, etc.,
-1999 crowd of dot com speculators (“this time it’s different”)
-2003-06 crowd of housing price speculators (“no really, this time it’s different”)
-2008 crowd of oil and grain speculators (“no, no, no…it really is different, this time.”)
With crowds as wise as all that, I’d rather just go it alone.
I am not suggesting that they need to be perfect — what I have been saying is that they are flawed, and that many people put way too much stock into them.
By the way, if you couch EVERYTHING in percventage terms, well than, you can never be wrong — the fallback position is that the lower probablity event occurred.
Google’s IPO was probably at $85 because it was low balled so execs could get cheap stock options…not a good example of free markets.
ottnott and BR, perhaps you have a problem with the phrase “predictions market”? The Intrade crowd is not predicting that McCain is going to lose; rather, it predicts he has a 38% chance of winning given the information the crowd knows at this time. I don’t see how McCain winning in a landslide in November would invalidate that estimation on May 13.
It reminds me of the experiment in which people are asked to guess how many jelly beans are in a big clear bag. Very few correctly guess the right number, but the more people that are asked, the closer their average comes to the correct number.
You may be correct about “too much stock” but I argue that these markets are valuable. I use them to hedge when I have an emotional interest in an outcome (e.g. I want Michigan to beat Ohio State; Kerry to beat Bush; or the Cubs to win the pennant). I ask myself how much would I pay to guarantee my preferred outcome, then I bet that amount of money against my preferred outcome. I call it Cubs insurance, because I would have made more than 10 to 1 on my money had I bet against the Cubs just before Steve Bartman resurrected the curse.
I thought the wisdom of crowds was the same as mobs: take the average IQ and divide by the number of participants (as per old Ted Sturgeon). ~m
Adam:
Sounds to me like we aren’t in direct disagreement. You are defending the ability of the prediction markets to generate a representative percentage. I’m defending the notion that the representative percentage generated is not a valuable predictor just because a large number of people were involved in generating it.
I still have to quibble, of course. I’m not convinced that the percentage generated is representative for several reasons:
1) Using the McCain example, the 38% reflects only those trading McCain futures, which might not be a representative sample in spite of the number of people involved
2) The 38% is a derived number – it is not a direct answer to the question “what do you believe is the probability of a McCain victory?”
3) related to #2, putting money on the “prediction” changes the predictions people are willing to make. As an example, let’s say that I believe there is a 5% probability that Al Gore will be the Democratic nominee (I don’t). I can buy the Al Gore futures for 2.9. As an intellectual exercise, I would say that the Intrade futures price greatly understates the probability. As an investor, I’m not interested in something that I believe has a 95% chance of returning 0.
4) The combined Intrade futures prices for the outcomes of an event can exceed 100.
Well, at the risk of sounding like the obvious, a 20% longshot should probably win about 20% of the time… there isn’t enough data to really judge how often these 20% chances happen, (and we won’t use the Iowa markets due to their massive illiquidity and horribly skewed sampling of who actually trades them).
But I think probably, in your defense there are three points.
1) you modified your stance. you should in fact couch things in percentage terms.
2) The Time value of money. Once McCain cleared 95, even though he was all but the nominee, it still wasnt worth it. The 5.1% or so you’d gain on your investment is offset by the fact that you’d be locked up for upto 6 months, plus the chance McCain had a stroke or something. This skews the odds, since McCain’s true odds were closer to 98%, but his trading price was a few percent lower.
2) The odds fluctuate. mccain at one point was truly seen as dead in the water, and his resurrection was miraculous. Yet if you asked me circa 2004 what I thought of McCain, i woulda taken him (back when he was a fiscal conservative) over Bush or Kerry in a heartbeat, and I think many others thought that way too. The fact is our society tends to overvalue the NOW. Someone stumbles on a speech? They have no chance. The Washington Nationals start out 3-0? They are winning the world series! There is profit in those sorts of things.
ottnott:
You state your case eloquently, but for whatever reason, I think I disagree with most points.
2. Not sure I understand #2.
3. I disagree. As an investor, if I correctly estimate that something has a 95% chance of returning 0 and a 5% chance of returning 100, I could make money by regularly betting 2.9 for such contracts.
4. This is extremely rare. But rare arbitrage opportunities exist in all markets.
Here’s what I don’t get about these predictions, anyway:
Let’s say that Indicator XYZ has determined that there is 38% chance that Obama will win.
Let’s then say that Obama wins.
Was the indicator right or wrong?
Do you get what I mean? Since there was still a 34% chance he might win, then the indicator was right, right? Because there was still a chance. But if this is true, whether Obama has a 34% chance of winning or a 56% or a 42.9% chance, isn’t it basically impossible, with only one election, to determine if the indicator was correct? Even if Obama has a 99% chance of losing, there is still a 1% chance of winning, so either a win or loss would still verify the indicator, right?
To put it another way, since we can’t run elections 100 times to see if Obama wins 34 of them and loses 66 of them, how can we ever verify whether these things are accurate or not?
I know next to nothing about statistics, and so maybe I am just exposing myself as an idiot here, but this is a part I have never understood.
No problem, Adam.
Regarding #2: I’m just stating my belief that the answer to “How much will you pay for McCain to win?” is not the same as the answer to “What do you think is the probability that McCain will win?”. When answering the question, I can give a specific number. When dealing with the futures, response is blunted to “higher”, or “lower”.
Regarding your disagreement with #3: As an investor, if I correctly estimate that something has a 95% chance of returning 0 and a 5% chance of returning 100, I could make money by regularly betting 2.9 for such contracts.
You are right…
If you correctly estimate, and
if you can regularly make such bets, and
if the opportunity cost (mental, time, money) of regularly making such correct estimates is not too high.
I’d argue that Intrade and similar prediction markets
The proper benchmark for telling whether prediction markets are worthwhile or not is to compare them to what “experts” think. In the Wisdom of Crowds, I thought it was pretty evident that he was arguing that collective wisdom is simply better than the average expert and for many fields beats all but the top couple of experts.
Also the argument is that over time they are quicker to react to changes than even the top experts who for the most part are at the top because they are highly specialized and have gotten fortuitous circumstances that play into their specialization.
In all honesty, there are very few properly constructed benchmarks to tell whether they are working or not. I certainly don’t know of any political one. But if the market gives a 20% chance to the actual outcome while the experts gave it a 10% chance, it’s still a win.
Steve Barry> Google’s IPO was probably at $85 because it was low balled so execs could get cheap stock options…not a good example of free markets.
What? Google’s IPO was priced using a “Dutch Auction”. Can you explain how “execs” could have possibly “low-balled” the offering price?
Setting the record straight, the only time John Kerry led at Intrade in 2004 was on Election Day and that was because of the strong exit polling for Kerry. GW Bush was leading at Intrade consistently by wide margins prior to election day.
RFK Jr. wrote an excellent well-footnoted article entitled “Was The 2004 Election Stolen?” in Rolling Stone in 2005 detailing the unlikely outcome of the 2004 election given the exit polling, and all sorts of election shenanigans in Ohio. The Intrade traders were duped by those shenanigans, too.
It’s kind of surprising that RFK Jr’s article didn’t get paid more attention because close reading of it provides some scandalous details. Here’s the link for some interesting reading if you have a good 30 minutes:
http://www.rollingstone.com/news/story/10432334/was_the_2004_election_stolen/print
What? Google’s IPO was priced using a “Dutch Auction”. Can you explain how “execs” could have possibly “low-balled” the offering price?
__________________________________________
First off, a Dutch Auction, by its very nature, finds the lowest price at which a company can sell all its shares. from CNN Money 4/04:
“The IPO price will equal the lowest price bid on any of those 150 million shares. The price all bidders pay will be the IPO price — even if they had bid higher.
So it seems that you may be able to buy shares of Google, but should you?
Ideally, the auction process enables sellers to price the issue “right.” That is, its price should reflect the reasoning of thousands of investors who will determine for themselves how much they are willing to pay for a share.
This type of auction should cut down on the huge run-up in share price experienced during the first days of trading experienced by other tech IPOs during the 1990s.
So bid on shares if you believe in the company — remembering, of course, that few companies make it from start-up to titan without experiencing big setbacks along the way. But don’t count on being able to flip your shares for a large gain the day after the IPO.”
Note that the auction was designed to cut down the runup after the IPO. Google was very secretive of the goldmine they were sitting on…this from an SEC action against them:
“5. Rule 701 promulgated under the Securities Act provides an exemption from registration for certain issuers offering and selling stock options (or other securities) to employees and consultants under compensatory benefit plans. However, Rule 701 requires (among other things) that any company issuing more than $5 million in stock options over a 12-month period provide detailed financial statements and other disclosures to the option recipients. The Rule allows privately-held companies to compensate their employees with securities without incurring the obligations of public registration and reporting, while ensuring that essential information is provided to employees.
Google’s Failure To Comply With Rule 701
6. Since its inception, Google has granted stock options to its employees and consultants as a form of compensation. Under Google’s stock option plans, Google’s Board of Directors granted the company’s employees and consultants options to buy a certain number of Google unregistered shares at an exercise price set by Google’s Board. Although the stock options were not registered, Google relied on Rule 701 of the Securities Act to exempt those securities from the registration requirements of the federal securities laws.
7. In September 2002, Google became aware that its continued issuance of stock option grants might reach levels requiring financial disclosures under Rule 701. Google temporarily stopped issuing stock grants. In contrast to its chief competitors, Google was a private company, and did not have to report its financial results and other significant business information to the public in filings with the Commission. Google viewed the public disclosure of its detailed financial information as strategically disadvantageous, as Drummond recognized, and the company was concerned that providing option recipients with the financial disclosures required by Rule 701 could result in the disclosure of this information to the public at large and, significantly, to Google’s competitors.”
Google let the “small investor” get a shot at their IPO, also letting the execs get nice strikes on their options, which they now exercise quite liberally.
Couldn’t agree more about prediction markets merely being a distillation of the media-pushed CW. Up until the NC/Indiana primaries you could basically get 1/3 odds of Barack Obama winning the Democratic nomination, when the actual delegate numbers showed it would take a run of massive (20 point plus) wins for Hilary to get close. So in the end you were getting 1/3 odds on Barack Obama being revealed to be the anti-christ or being assassinated.
These so called political FUTURES market are not really futures markets at all but MOST RECENT INFO markets. They just jump around reacting to the latest bit of political info to hit the airwaves. In short they are a joke.
It’s so en vogue to be market-sceptical these days isn’t it?
Whoever claims or expects markets to “predict” should get a head check.
Markets don’t predict. If one person can’t, neither can 1,000,000.
Markets are just the least imperfect allocators of resources.
In terms of future outcomes, they’re just the best way to know what everyone is thinking – not what’s most likely to happen.
Knowing what the aggregate expectation of the future is however, also has its uses.
Marketplaces, like everything else, also have varying degrees of quality. It may be that the traded instrument has too much uncertainty associated with it, it may be that it’s a small and illiquid market trading it (like Barry said), and it may as well be both.
In short, markets are just fine as long as you don’t expect perfection from them. As to politics….
I saw a fascinating presentation by Justin Wolfers at the CFA conference stating otherwise. His research argues otherwise. According to Wolfers’ work, prediction markets have been more accurate at picking winners of elections than polling.
As to why prediction markets are sometimes wrong, i.e. that Obama had a 70% chance of winning New Hampshire but lost, the answer is because the market was assigning a 70% chance that he would win, not a 100% chance. There was a 30% chance he would not win. And when you have a 30% chance of losing, one expects to lose 3 times out of 10, not 0 times out of 10. The fact that prediction markets are assigning a 40% chance of McCain winning does not mean McCain has a 0% chance. It means he has a 40% chance.
Not my work, but fascinating nonetheless.
T.