« My favorite things Virginia | Main | Assorted links »
Why don't more businesses use prediction markets?
Last week in The New York Times (TimesSelect), Joseph Nocera quoted Robin Hanson as saying private businesses had not made a breakthrough with the use of idea futures. It seems natural to let your employees bet on future business conditions, the success of product lines, or broader questions of corporate strategy. Microsoft and Google and a few other companies have played with the idea, but it does not (yet?) seem to be taking off. Why not?
1. Prediction markets threaten the hierarchical control of top managers. It would become too obvious that most managers are idiots, unable to predict the future.
2. Prediction markets make a big chunk of the bettors into "losers." Yet within a company morale is all-important. Businesses proceed by soliciting feedback, and by reshaping their plans to pretend that everyone is on board and has an ego stake in the final outcome. Prediction markets make this coordination more difficult. Once people make bets, they start rooting for their bet to win and for the other bet to lose. They move away from maximizing the value of the firm and develop an oppositional mentality vis-a-vis other employees. Furthermore it is disruptive to have a running tally on who are the winners and losers each day.
3. No matter what they pretend, businesses are not much interested in forecasting many future variables. Successful businesses find product markets they can control for long periods of time. They do a few things really well, and let a surprisingly large number of tasks slide.
4. We already have implicit betting markets in the form of resource prices. When the information contained in those prices is sufficiently important, institutions will be organized in terms of "markets," rather than "firms." Or firms can look at resource prices in outside markets for the information they need.
5. Most employees have no rational basis on which to bet. If someone knows the truth, but is otherwise locked out from credibly signaling that knowledge to management, something is wrong with the organization of the company. The small prizes from corporate prediction markets won't be enough to elicit that knowledge from him in any case.
6. The corporate beast is far more constrained than most outsiders imagine. Interest groups must be courted, coordinated, and sometimes fought every step of the way. When it comes to choice, there are fewer degrees of freedom than one might think. The real question is not what to do, but rather having the will and effectiveness to do it. A bit like international free trade, no? Prediction markets don't help much in this regard.
7. When reward systems are created, employees view them as a means to distribute further privileges to insiders and favorites. Prediction markets would be viewed the same way and in fact this might be true. Who else is going to win all those bets? Do corporations really need more insider favoritism?
Your thoughts? Here are five open questions about prediction markets.
Posted by Tyler Cowen on March 15, 2006 at 07:27 AM in Economics | Permalink
TrackBack
TrackBack URL for this entry:
http://www.typepad.com/t/trackback/3576/4445727
Listed below are links to weblogs that reference Why don't more businesses use prediction markets?:
» Prediction Markets in the Firm from Catallarchy
Tyler Cowen asks why prediction markets are not common in business. The obvious answer is that the concept is still too new for most organizations to really adopt it. It took decades for U.S. based businesses to catch on to quality programs.
Having... [Read More]
Tracked on Mar 15, 2006 8:57:55 PM
» Do-It-Yourself Prediction Markets from Knowledge Problem
Michael Giberson You have read about prediction markets, maybe you have even participated in one, and now you want to go to the next level. You have an idea for a market, but you dont know how to make one.... [Read More]
Tracked on May 10, 2006 12:44:28 AM
» Do-It-Yourself Prediction Markets from Knowledge Problem
Michael Giberson You have read about prediction markets, maybe you have even participated in one, and now you want to go to the next level. You have an idea for a market, but you dont know how to make one.... [Read More]
Tracked on May 10, 2006 1:25:37 AM
» Do-It-Yourself Prediction Markets from Knowledge Problem
Michael Giberson You have read about prediction markets, maybe you have even participated in one, and now you want to go to the next level. You have an idea for a market, but you dont know how to make one.... [Read More]
Tracked on May 10, 2006 1:28:15 AM
Comments
I nominate:
8. Regulations, both from government regulators and from internal ethics policies. Most companies take aggressive steps to prevent anyone from profiting from either inside information, front running, or side deals with outside customers and suppliers. Letting people profit from information via prediction markets goes against the spirit of these ethics rules and could in some case provide an incentive for rule breaking. Even worse, Sarbanes-Oxley and SEC disclosure rules could provide criminal penalties for CEO's who tell the world something different from what the internal prediction markets are saying.
It is no surprise that Microsoft and Google would be interested in prediction markets, as both companies have reputations for playing fast and loose with government regulations (Microsoft for antitrust, Google for copyright and their pre-IPO quiet period interview with Playboy). [Note that I am just summarizing some common opinions of these companies; I am not judging their actions, nor am I alleging any actual violations, nor do I have any nonpublic knowledge of either.]
Old economy companies are more careful. I work at a bank, and I'm just glad they haven't banned visiting or commenting on blogs yet.
Posted by: DK at Mar 15, 2006 8:25:38 AM
9. Public prediction exchanges have not *yet* openened prediction markets aiming at divining future corporate aggregats.
You said: "It seems natural to let your employees bet on future business conditions, the success of product lines, or broader questions of corporate strategy."
These examples could well be settled by public observables.
[Psstt... In the link under my name, you'll find an excerpt of the NYT news article.]
Ciao,
Posted by: Chris. F. Masse .COM at Mar 15, 2006 9:01:55 AM
To some extent this is present in (2), but perhaps it could be spelled out a bit more. To what extent would such markets introduce bad incentives? For example, suppose I bet (sorry, "predict") that a product I'm working on will have a late release date? or will fail commercially? Someone has to bet against success for these markets to be interesting, and such bets do not encourage whole-hearted commitment to success.
Posted by: Tom Slee at Mar 15, 2006 9:05:24 AM
Item number 6 on your list seems the most plausible to me.
Many people seem unaware of the extent to which corporations are politicized.
Maybe not in the same way that politics is politicized, but navigating corprate bureaucracy still requires a degree of diplomacy.
Posted by: Dave at Mar 15, 2006 9:12:24 AM
1 (a): Many managers assume they know they know at least as much about the state of the world as any of their employees (after all they are 'in charge') so why ask?
1 (b): Managers think they provide the same function by asking their suboridinates (and recursing through the organization) their beliefs and aggregating the response. The incentives however may be quite different when talking directly to your boss versus logging on to an internal prediction website.
Other reasons:
Size. It takes a 'large enough' business unit for the aggregation benefits to show up. Obviously there are a number of companies large enough but it's usually smaller companies that are willing to attempt innovations like this first.
Cultural norms. If you thought the project was going poorly why didn't you tell management instead of betting against your company's success? Aren't you a team player?
In any non-annonymous market reason 6 is going to be a big factor. Who wants to bet against the boss? If you do is it better to be right or wrong?
Posted by: Trevor at Mar 15, 2006 9:27:26 AM
I'd say that there are already de facto prediction markets in most companies. Managers are nearly always asked to predict their unit's performance in the coming year, and they in turn will coordinate with their line employees on what that prediction should be.
A person (manager or employee) who is wrong risks their bonus and a person who is seriously wrong or consistently wrong over time loses their promotability and perhaps their job. All those things have real economic value. Additionally, there are emotional benefits to being right and costs to being wrong that we'd be foolhardy to ignore.
Every time a CEO is asked (and they get asked this a lot) "What's the bet-your-job profit number for next quarter?" there is effectively a derivative being created which takes on a very high value (CEO compensation) if the result is at or over the prediction and a very low value (a job hunt) if the result is under the prediction.
That said, I'm very interested in the idea of letting line employees who disagree with their manager's prediction explicity bet against it. Normally the more information you get from people with specific line-level knowledge, and the less you get from management, the better.
Posted by: Andrew Edwards at Mar 15, 2006 9:30:04 AM
Tom Slee's point above about incentives is a good one.
You'd need to cap employee's bets below their bonuses, so that it was always better to increase performance, and the betting market could be used only to "hedge" against a management predicition that is too high or "double down" on a prediction that is too low.
The other danger is that since other aspects of compensation are tied to "meeting targets" employees may systematically climb on the short side of the market as hedgers. You'd need to find a way to account for the risk aversion of everyday people.
Posted by: Andrew Edwards at Mar 15, 2006 9:35:43 AM
Andrew Edwards wrote: "I'd say that there are already de facto prediction markets in most companies."
Surely, you're joking Mister Edwards.
In a prediction market, there's public *trading*. When there're enough uninformed traders, the marginal traders (a.k.a. market makers) take over and direct the market price close to an objective probability.
[Psstt... Under my name, there's a link to a news story hinting at some General Motors prediction markets, soon to be floated.]
Posted by: Chris. F. Masse .COM at Mar 15, 2006 9:35:48 AM
Here's an example of a set of *public*, real-money business prediction markets aiming at the release date of the PS3 machine.
So maybe the answer to Tyler Cowen's question is:
10. Business prediction markets already exist, but Tyler Cowen hasn't noticed yet.
[Under my name, link to the InTrade/TradeSports PS3 prediction markets.]
Posted by: Chris. F. Masse .COM at Mar 15, 2006 9:43:27 AM
(1) This is surely a factor. I never worked at a company where I didn't observer the managers making obviously dumb decisions.
Posted by: Half Sigma at Mar 15, 2006 9:43:29 AM
Chris,
on General Motors prediction markets...this has been around for a long time ,it's called cds [credit default swaps]
Posted by: jck at Mar 15, 2006 9:50:27 AM
Chris F. Masse is on the right track. The fundamental problem with prediction markets is that you need uninformed traders, or "dumb money" -- people who are willing to LOSE money -- to work. Because, without people who are willing to lose money on bad trading, there's no incentive for the "smart money" to do the research to drive the prices to the correct level.
Since the dumb money expects to lose money on average, they have to get something in return. Effectively, they are buying "something" with their expected losses.
In the case of a site like tradesports, the dumb money gets the positive benefit of the "fun" from gambling on their team or whatever in return. In the case of stock markets, it is possible for the dumb money to still make money by buy-and-hold. In the case of futures markets, the dumb money can reduce their financial risk. At Microsoft and Google, you have all these smart math/computer type people who probably view it as an interesting intellectual challenge. But what does the dumb money get out of guessing whether sales are going to be up 10% or 15% at some random company?
Posted by: Foolish Jordan at Mar 15, 2006 9:55:55 AM
JC Kommer,
Fair remark (on the 'credit default swaps'). However, with prediction markets, there's openness and the interpretation is easy. Prices (0--100) can be understood as probabilities. My 4 year-old nephew can understand that. Can you say the same with swaps or with traditional futures markets?????
Posted by: Chris. F. Masse .COM at Mar 15, 2006 9:56:42 AM
#3 is close, but not quite it. Companies are too busy doing work, making and selling things, and trying to create the future they want (where the public buys their product) to predict stuff. What sort of predictions do you think companies would want to make? "Is our product/system or our competitor's going to catch on?" Hell no, they want employees to go out and, whether they're true believers or not, sell/improve the hell out the product. If there are external factors (price of raw materials, etc.) which affect a business model, there are already markets in those things--in fact, a counterargument to your assertion that companies don't provide markets in the future is the fact that industries aren't shooting for vertical monopolies like they did a century ago; companies are in fact choosing markets to predict external factors in their production processes.
Posted by: Tim at Mar 15, 2006 10:06:31 AM
I go with a variation of (3) and (5). Prediction markets would only work if a large number of people are participating in them, but opening up prediction markets and encouraging an internal workplace culture of using them to the extent that they can be effective, is going to divert employees from more productive tasks. (8) is huge, too -- see Ribstein and Butler on Sarbanes-Oxley at the March 13 AEI event.
Posted by: Ted at Mar 15, 2006 10:20:27 AM
A few thoughts:
1) There are many more companies interested in the possibilities of these markets -- or similar mechanisms for aggregating the collective knowledge of their employees -- than there were even two years ago. But actually using these markets in a meaningful way requires a fairly dramatic shift in the way corporations have traditionally thought about knowledge and decision-making. That's not something that we should expect to happen overnight.
2) The "dumb money"/"smart money" distinction is, I think, generally not a useful one to rely on in explaining why prediction markets work, and especially not in explaining why people participate. Study after study shows that "dumb money" doesn't know it's dumb, and that knowledgeable investors are no more sure that they're smart than dumb ones are. More important, there is no reason to think that people who are smart in one market are also going to be smart in another market.
3) Within corporations, people participate in these markets not just for the financial/ego rewards, but also because they would like to feel that their opinions were having some impact (however minor) on what the company does. In large corporations in particular, many employees feel as if they are fundamentally irrelevant. Prediction markets get around that.
4) Finally, and most importantly, I would question the implicit assumption in Tyler's post (and in some of the comments) that the "knowledge" these markets are trying to tap is something that one employee (or a very employees) have and that most employees are completely without. No one person knows what a company's sales are going to be in forthcoming quarters, or whether a given product is likely to be a hit or not, or whether a software project will be done in time. The information that's relevant to answering any of these questions well is not concentrated in the hands of one person or a few people. Instead, it's distributed among lots of people, each of whom likely has a small bit of knowledge that's germane. The genius of a prediction market is that it's able to aggregate all of that distributed information and produce a collective forecast. And I think in the long run, the benefits of this approach are going to outweigh the pitfalls (many of which, I think, Tyler has accurately enumerated).
Posted by: James Surowiecki at Mar 15, 2006 10:51:18 AM
In a corporate setting, if everyone above you has a vested interest in a certain outcome, and you correctly predict that the outcome will not happen, you will be punished and possibly fired. In many large organizations, being wrong when everyone above you is wrong too is fine, being right when everyone above you is wrong is dangerous.
Posted by: cactus at Mar 15, 2006 10:53:46 AM
James Surowiecki wrote: "The 'dumb money'/'smart money' distinction is, I think, generally not a useful one to rely on in explaining why prediction markets work, and especially not in explaining why people participate."
Study after study made by the Iowa Electronic Markets refer to a population of traders that they call "marginal traders" or "market making traders" (a.k.a. human "market makers").
They appear to "be responsible for the accuracy of the IEM prices."
They are "more male and more highly educated" than the Joe-Six-Pack traders (who are already more male and more highly educated than the rest of the population).
This is page 26 of the IEM paper, "The Iowa Electronic Markets: Lessons learned and answers yearned" by Berg & Rietz.
And that's hardly the first paper where they raise this issue.
[If you follow the link under my name, you'll find the link to the PDF.]
Posted by: Chris. F. Masse .COM at Mar 15, 2006 11:33:37 AM
On the market makers/uninformed traders issue:
Below are excerpts from "Five open questions about prediction markets" - by Justin Wolfers and Eric Zitzewitz - (2005-01-21) - 2006-02-00
1. How to attract uninformed traders? Counterintuitively, the problem for most prediction markets is attracting sufficient uninformed order flow.
An important implication of the model sketched above is that the success of the prediction market in generating trade depends critically on attracting uninformed traders. --- Risk love or the “thrill of a gamble” provide obvious motives for uninformed traders, and both Tradesports and Betfair have successfully attracted many sports bettors to their markets. --- As the wonkishness of the contract rises, however, volume and liquidity falls rapidly. --- To summarize, three routes to attracting order flow have been successful thus far: offering sports betting, subsidization, and, possibly, exploiting career concerns. Each has their drawbacks, however.
Posted by: Chris. F. Masse .COM at Mar 15, 2006 11:54:38 AM
On the so-called marginal trader point:
1) Collectively, "marginal traders" control a small fraction of the amount of capital invested in the market (the IEM and other prediction markets) as a whole. In other words, they do not have enough capital to set prices. By definition, if the "Joe Six-Pack traders" collectively think the price should be dramatically different from what the marginal traders think it should be, the price would move to where the Joe Six-Pack traders want it.
2) Marginal traders are defined, in part, by the fact that they tend to make bid/ask offers, while other traders accept it. This is why they're called "market makers." But this does not make them price setters. In an open market, a price is set, obviously, by both buyer and seller. The buyer's decision to accept the seller's price (or vice versa) makes the market as much as the offer to sell does.
3) In any group of traders, there are going to be some people who know more than others, and who are going to be more accurate in their forecasts than others. This is just tautological. That does not mean that those people are going to be the same from market to market, nor does it mean that without them the market's collective judgment is going to be radically off-base -- which is what the "marginal trader" hypothesis implies. In a sense, the "existence" of marginal traders is simply an artifact of any collective decision-making mechanism.
4) The price of a security in a prediction market represents the aggregated collective judgment of all the traders in that market. That's why it's accurate.
Posted by: James Surowiecki at Mar 15, 2006 12:25:09 PM
As a supplier of prediction market software, I think more work needs to be done educating the marketplace about this methodology. Articles like this one and books such as "The Wisdom of Crowds" have done *much* to advance the concept but you'd be surprised at how many executives I talk to know nothing of prediction markets and the value that they can bring to the organization.
The existing software options have also impeded wider adoption. While they work well in academic settings, many prediction market platforms have clunky interfaces which are hard for corporate clients to use. From our experience, the markets with the simplest users interfaces have also been the most successful.
Posted by: David Perry at Mar 15, 2006 12:45:35 PM
"The price of a security in a prediction market represents the aggregated collective judgment of all the traders in that market. That's why it's accurate."
This is a very foolish claim,marginal opinion determine market price,not colllective intelligence or the majority.see "theory of investment value" john burr williams [chapter 3]
Posted by: jck at Mar 15, 2006 1:02:43 PM
James Surowiecki wrote: "In other words, [marginal traders] do not have enough capital to set prices." & "But this does not make them price setters."
From what I read from the IEM papers, I disagree on the ground that the Joe-Six-Pack traders (a.k.a. the typical traders) are more likely to *accept* prices proposed by the market makers (a.k.a. marginal traders). And even though it's true that the market makers "control a small fraction of the amount of capital invested in the market" (as James Surowiecki wrote), since they trade like a machine gun, they very often *propose* prices.
Note the final line, in the paragraph below: "Marginal traders, not average traders, drive market prices and, therefore, predictions."
The excerpt below is from "Results from a Dozen Years of Election Futures Markets Research" - page 6 - by Joyce Berg, Robert Forsythe, Forrest Nelson and Thomas Rietz - 2000-11-00 - [Link to the PDF file under my name.]
"Specifically, the core group of traders that tend to set market prices appears less biased and error prone than typical traders. [T]ypical traders have unrealistically hopeful beliefs for their preferred candidates []. However, 'marginal traders (those who regularly trade or place bids and asks near the top of the queues) are much less prone to this bias. [T]ypical traders often trade at a price that is not the most advantageous price for the trader or that violates arbitrage restrictions. Such 'mistakes' are 'irrational' because they decrease the trader's payoff regardless of expectations or outcomes. However, 'market makers' (those who actually set market prices by placing the best bids and asks) make mistakes much less often. [...] Marginal traders, not average traders, drive market prices and, therefore, predictions."
Posted by: Chris. F. Masse .COM at Mar 15, 2006 1:10:35 PM
Out here on the front lines, we encounter our share of challenges, but I can't say that I recognize too many of them in the list provided earlier.
As James points out, the surge in interest from corporations in the last two years has been phenomenal. That is mostly thanks to his book which has many fans in the CEO and VP ranks. But businesses are fundamentally conservative, and before they change their current way of doing things, they ask for proof of the bottom-line benefits. As Robin Hanson points out in Saturday's NYT article, no company has yet gone on the record to say that they saved serious money with better forecasts generated by prediction markets. However, its very clear to us, from where we stand, that as soon as this happens, a lot of fence-stting companies will rush to implement teir prediction markets. Companies are hurting out there like you would not believe, and bad forecasting is a big part of it. (Some companies even tell us that their forecasts are so bad that they've learned to operate without any, and wouldn't know anymore how to do things differently if they suddenly had good forecasts! -- I'll nominate that for #10.)
Ok, so why don't we have that bottom-line proof yet?
1) Pilot experiments take time. To be convincing, a prediction market must be applied to an actual strategic business problem within the company. The Corning LCD TV Futures market is a good (public) example of such (see http://lcdtv.newsfutures.com). But we won't be able to assess results and business impact before several more quarters have passed.
2) Most of the better-publicized pilots, from the early HP printer-sales market to the recent Google internal market, have implemented binary markets that generate event probabilities. But event probabilities are just not actionable. People in a business environment don't know what to do with them. They need specific numbers. Not "What is the probability that I'll hit my deadline or sales target?" but "What will my sales be?" and "When will I be able to ship this?", or "What will my sales be if I include this feature or that one?" Besides the our own relatively recent work with companies such as siemens, Eli Lilly, Corning, Arcelor and others, I know of no other corporate prediction markets that seek to produce actual actionable forecasts rather than probabilities. And, again, coming back to my point (1), these take time to run, and then need time to be assessed properly for a chance to impress the bottom-line focused decision makers.
That said, the work within corporations is progressing steadily and the early results are good, despite the challenging corporate environments. We're definitely rounding the bend.
Posted by: Emile Servan-Schreiber at Mar 15, 2006 1:22:32 PM
Surowiecki is right about "marginal traders." Sorry, market makers do not set or control prices, except in conspiracy theories. You can't set prices unless you have a real demand or supply curve. The classic market maker (e.g. on the floor of the NYSE) has no demand for stock, relatively little capital, and usually tries to have zero net stock holdings at the end of each trading day. Instead, the market maker supplies liquidity, by quoting bid and ask prices around the aggregate of other traders' supply and demand curves, which the market maker learns by observing order flows. The price setters are Fidelity and other big stakeholders with their own real views on the market, and supply and demand curves reflecting their views.
Prediction markets work the same way, although since they are smaller than US equity markets, the players are less specialized, and you are more likely to find people who blend market making with committing real capital to specific ideas. As prediction markets get larger, IMHO, you will see increasing specialization into passive market makers and people with views.
Posted by: DK at Mar 15, 2006 2:25:49 PM
