Prediction Markets Definitions

No Gravatar

We should try to be more careful about our distinctions and definitions. I&#8217-m less concerned about whether we use the phrase &#8220-prediction market,&#8221- &#8220-information market,&#8221- &#8220-decision market,&#8221- or whatever- I&#8217-m more concerned about what exactly these phrases are intended to mean. Here are some possible concepts to distinguish (I&#8217-ll avoid naming them).

  1. Markets that make predictions – Pretty much every speculative market ever created does this.
  2. Markets that make easily interpreted predictions – Most betting markets would be included here, while most financial markets would not. For financial markets one must understand and estimate risk premia, discount factors, and so on be able to say much what exactly the current price estimates.
  3. Markets whose prices embody information – Also pretty much every speculative market ever made.
  4. Markets whose price predictions are used as info by non-traders – Many financial markets meet this criteria.
  5. Markets whose interpretable predictions are used as info by non-traders – Many betting markets meet this criteria.
  6. Markets whose primary function is not hedging – A &#8220-primary&#8221- function would most explain the existence of the market and its volume of activity. Most financial markets might fit here, as most volume is from speculation.
  7. Markets where the main trader motivation is not material gain – Most play money markets would fit this criteria.
  8. Markets where the social value of allowing the market to exist outweighs the social cost – this would be the sort of market we want to legally allow to exist.
  9. Markets where the social value of the info gained by non-traders, but not the social value of its use for hedging, outweighs any other social costs of allowing the market to exist – These markets could justify a legal regime empowered to allow markets to exist for info reasons, not just hedging reasons.
  10. Markets whose primary function is to inform non-traders – Non-trader interest in the info would be the main explanation for the existence of the market and the volume of trade- such non-traders would somehow subsidize the existence of the market and trading activity in order to gain the info they desire.

These last two concepts are of the most interest to me – I would like to have names that clearly identify them and distinguish them from the other concepts.

30 thoughts on “Prediction Markets Definitions

  1. Crosbie Fitch said:

    ‘Decision support’ markets?

    or simply

    ‘Decision Markets’?

    But surely non-trading sponsors simply trade in slightly different predictions, e.g.

    I am offering odds of 10:1 against you picking the winner.

    If you pick the winner I pay $10 with a refund of your $1 stake. If you don’t, I keep the $1.

    Otherwise non-trading sponsors simply provide participation fees or retainers for players to go through the motions with play money.

    Wouldn’t there be a considerable temptation to provide monetary incentives as rewards to better players?

    And consequently this could be extracted out into sponsors’ being participants anyway?

  2. Robin Hanson said:

    Crosbie, by funding an automated market maker, a sponsor can pay only those who move the price in the direction of the final judgement.

  3. Crosbie Fitch said:

    Which could be equivalent to obliging traders to buy all sponsors’ predictions of their success and failure in predicting the final judgement, such that the consequences for failure are zero, and for success, positive.

    In my simple bet there’s a small negative consequence for failure, but this can be contrived to zero if preferred.

    I’m not making a big fuss here, just trying to see if, at least theoretically, sponsorship can actually be made equivalent to participation in the market.

    If the end result is identical, does it really matter I wonder?

  4. Robin Hanson said:

    Let me try to be more direct; who would object to my/our defining the most popular term “prediction market” to refer specifically to the concept I care most about, “markets where non-trader interest in the info to be gained from directly interpretable prices is the best explanation for the existence of the market and the volume of trade”?

  5. Eric Zitzewitz said:

    Robin,

    Here is the taxonomy I’ve been carrying around in my head:

    Markets whose primary purpose is capital raising or sharing/hedging risks => Financial markets

    Markets whose primary purpose is to aggregate information about the expected outcome of an event => Prediction markets

    Markets whose primary purpose is to facilitate taking financial stakes in the outcome of events for entertainment purposes => Betting markets

    I think the line between financial and prediction markets is reasonably clean. A cat bond on hurricane losses in Florida is on one side, a Tradesports contract on whether a hurricane hits Florida is on the other. This is clear from their design: one of which optimizes risk sharing, the other, a clean interpretable prediction.

    The line between prediction and betting markets is less clean. Markets designed primarily for prediction often work better when they attract entertainment-motived traders, and betting markets often have nice predictive applications, even if sometimes unintended.

    One way to draw the betting/prediction market line could be to ask whether prices’ predictions could affect decision making, which is another way of asking whether the information they contain has (non-entertainment) value. Of course, one can construct examples of markets most would consider betting that still fit this definition (think of a football coach who checks Tradesports when deciding whether it’s safe to sit his starting quarterback).

    As to the other terms (many of which you coined …):

    Decision markets could be defined as any market that helps make a decision, but this seems too broad to be useful. I’d prefer a narrower definition: markets desigend to predict outcomes conditional on a decision being taken. They are clearly a subset of prediction markets.

    Information markets I’ve never found to be a useful term(apologies to its advocates). Largely this is for the reason you alude to, that essentially every market price provides information.

    Event futures and prediction markets seem to be pretty close synonyms. Prediction markets is slightly broader, so maybe it’s OK that we seem to have standardized on that term.

    Idea futures is a subset of prediction markets.

  6. Eric Zitzewitz said:

    “Who would object to my/our defining the most popular term “prediction market” to refer specifically to the concept I care most about, “markets where non-trader interest in the info to be gained from directly interpretable prices is the best explanation for the existence of the market and the volume of trade”?”

    I’d drop the last 5 words, and then I think the correlation between us is 0.99.

    Many markets that we’d all probably call prediction markets work fine without subsidization. It’s hard to say why people trade in them, but I suspect that much trading has betting or (perhaps overly confident) trading profit motives.

  7. Chris Masse said:

    QUOTE

    Markets whose primary purpose is to aggregate information about the expected outcome of an event => Prediction markets

    Markets whose primary purpose is to facilitate taking financial stakes in the outcome of events for entertainment purposes => Betting markets

    UNQUOTE

    With all due respect to Professor Eric Zitzewitz:

    – In your taxonomy, I suppose, the “betting markets” are BAD, and the “prediction markets” are the GOOD incarnated.

    – Now what about this: a “prediction market” on topic A, with few transactions (because the non-profit exchange is run by unskilled marketers), and a “betting market” on that same topic A, with, this time, high volumes (because the commercial exchange is run by skilled marketers). The so-called “betting market” will have (and correct me if I’m wrong) a BETTER predictive power than the so-called “prediction market”.

    – I’m not impressed by the GOOD INTENTION of the designer of any prediction market. I’m more interested in the end game: Is that market-generated prediction accurate, yes or no? I don’t care how that market was classified by tenured economists in the first place.

    – As for the definition of the term “prediction market”, it is well understood by people as “a market that floats a prediction“. Tenured economists can try to change that definition, but will their special definition make it outside Academia?

    ========

    QUOTE

    Markets whose primary purpose is to facilitate taking financial stakes in the outcome of events for entertainment purposes => Betting markets

    UNQUOTE

    A last remark (if I may). ENTERTAINMENT PURPOSES??? I would like to make a survey among TradeSports and BetFair traders and see what they think of their trading being cataloged along side of Hollywood movies and late night TV talk shows.

    ====

    All that said with deep respect for these two professors of economics.

  8. Robin Hanson said:

    Eric Z.,

    Your preferred definition of “decision market” is what I proposed when I introduced the term, so I of course have no objection. :)

    Regarding “prediction market,” I won’t push very hard on including the volume of trade aspect in the definition. But I want us to be clear on the implications of the definition I proposed.

    Looking at the early publications on the Iowa Electronic Markets, they seem to have been created primarily to allow researchers to study the details of trader behavior and the information aggregation process, and not because they cared who would win in the elections. Similarly, I’ve heard that the Hollywood Stock Exchange exists more because it lets studios see which kinds of people are interested in which kinds of movies, and which are attracting the most trading activity, and not so much because studios want the directly interpretable estimates of movie success found in the prices. Finally, many corporate prediction markets seem to exist more so management can keep track of employee sentiment, and help employees feel involved and listened to, rather than because managers want to use prices to inform their decisions.

    According to our proposed definition, these would not count as prediction markets. Are we OK with that? Or should we leave them to be called prediction markets, and identify another term with the more restrictive definition we are considering?

  9. Robin Hanson said:

    Chris, if you are right that “the term `prediction market’, it is well understood by people as `a market that floats a prediction’,“ then we should of course use another name for any different concept. A survey to help us learn about such perceptions would be useful here.

  10. Yiling Chen said:

    In my mind, I have categorized both prediction markets and betting markets in Eric’s taxonomy under the term “prediction market”. I seem to take the concept of prediction market as:

    – A market whose primary purpose is to aggregate information about an event;

    OR

    – A market that makes easily interpreted predictions.

    As long as the market-generated predictions may be useful for some people, the market can be considered as a prediction market.

    For “markets where non-trader interest in the info to be gained from directly interpretable prices is the best explanation for the existence of the market”, the information-seeking intention of the non-trader is more likely to affect the mechanism adopted, such as using an automated market maker, and maybe the accuracy of the predictions. But to me the situation is more like non-trader uses prediction market as a tool to obtain information. The tool should be named according to what it can do, not where it is used.

  11. George Tziralis said:

    To me, all markets incorporate by their very nature and facilitate in one way or another the fundamental functions of investment, hedging, speculation and information aggregation.

    But, while there exist well-established markets created and operating with the primary purpose of capital allocation (e.g. stock markets), hedging risks (e.g. futures markets) and wagering (e.g. betting markets), the fourth fundamental function was not a raison d’etre of a market till the comparative recent emergence and development of prediction markets.

    Adopting this view, I propose the following definition:

    Prediction markets are markets which are designed and run for the primary purpose of mining and aggregating information scattered among traders and which -by subsequently transforming this information into market values- produce directly interpretable predictions about specific future events.

  12. Robin Hanson said:

    George, would your definition include or exclude the examples of IEM, HSX, and many internal corporate markets, created to study trading behavior or correlations between volume and demographics, or to help employees feel listened to?

  13. George Tziralis said:

    I think yes. As long as a market’s scope is more closely related to the amount of information that is aggregated during its operation, rather that the other three fundamental poles of market’s functionality I referred to in my previous comment, namely investment, hedging and speculation, then this market could be denoted as a prediction market.

  14. Robin Hanson said:

    Would it make sense to call any market whose existence is due more to the information it provides than than to these other common causes “information markets,” and reserve the name “prediction markets” for markets that exist more specifically for the information in their directly interpretable prices?

  15. Chris Masse said:

    QUOTE

    Prediction markets are markets which are designed and run for the primary purpose of mining and aggregating information scattered among traders and which -by subsequently transforming this information into market values- produce directly interpretable predictions about specific future events.

    UNQUOTE

    TradeSports and BetFair’s prediction markets are “designed and run” primarily to satisfy their traders’s needs for profits. Nevertheless, and correct me if I’m wrong, they display as much if not more predictive power than IEM (the exchange that seems to fit George Tziralis’s definition).

    Again, I repeat that, in my view, I don’t give the first fig that the predictive power comes as primary, secondary or even a tertiary motive, as long as it is there… for real —and you know that I mean relative accuracy, not absolute.

    I favor classification, but maybe what counts is the end result not the intention.

    And I keep my idea that a “prediction market” is “a market where predictions are traded” —just like a “fresh market” is a market where you can buy fresh fruits and vegetables, or a “futures exchange” is a marketplace where futures are traded.

  16. George Tziralis said:

    The terms of information and prediction in the notion of the markets that we all have in minds are highly correlated, I could characterize them as the two faces of the same coin.

    In other words, in every such market the information -of the traders- is aggregated and transformed into directly interpretable -by traders or not- prices, which always concern future events.

    Whether these prices are used finally to educate, decide or help somebody feel listened to, is of minor importance to me and such diversification seems to rather disorientate than help.

    But, regarding the comment of Chris, what I ultimately believe is the outcome of this commentary, is that we ultimately need not a strict classification of markets, but a more flexible topology of them.

    If for example we try to locate all markets in a ‘unit square’ with the features of investment, hedging, speculation and information aggregation on its angles, we could be more reliable in our ‘characterization task’. In such a topology, for example, Betfair would be farther than IEM from the information aggregation angle.

  17. Chris Masse said:

    QUOTE

    Whether these prices are used finally to educate, decide or help somebody feel listened to, is of minor importance to me and such diversification seems to rather disorientate than help.

    UNQUOTE

    I respectfully disagree. I enjoy categorizing the blog posts according to goals (TO FORECAST, TO PROFIT, TO HEDGE, TO ENTERTAIN, etc.). Makes me think.

  18. George Tziralis said:

    Chris, I’m referring to the need of a unique definition of the concept of prediction markets in a strictly academic framework. In a blog’s framework, I find your categorization very useful and more than welcome.

  19. Eric Zitzewitz said:

    Yiling,

    “As long as the market-generated predictions may be useful for some people, the market can be considered as a prediction market.”

    I think this is a fine place to draw the prediction/betting market line, if one must be drawn. It’s what I was after in the second half of my post above when I suggested drawing the line based on whether decisions would be conditioned on the information in the markets’ prices (one way of defining whether or not the information is useful).

    Chris,

    I’m certainly not one of the “betting markets = BAD” people. In fact, I’m on record as saying that prediction markets are the first form of “gambling” that should be legal.

    I do think that calling a market on the Super Bowl a prediction market might not be helping us with the “betting = BAD” crowd.

    Robin/George,

    People may have ulterior motives for running prediction markets, like the IEM, HSX, and corporate examples you cite. The problem though is that these motives often depend on traders ignorance of them (just as lab experiments often depend on subjects not knowing what the experimenter is trying to test).

    Suppose a company was really running a market for morale purposes. Would calling it a “morale market” really help the cause? :-)

    For that reason, along with just a general preference for big tents, I’d keep the examples you mentioned under the prediction markets umbrella.

  20. Chris Masse said:

    Thanks George Tziralis.

    Eric Zitzewitz, very interesting comment.

  21. Chris Masse said:

    QUOTE

    I do think that calling a market on the Super Bowl a prediction market might not be helping us with the “betting = BAD” crowd.

    UNQUOTE

    With all due respect to professor Eric Zitzewitz, and providing that I’m right about the prevalence in the general public of *my* definition of a “prediction market” (”a market that floats a prediction”), then THE TENURED ACADEMICS WILL HAVE TO ADJUST, PERIOD.

    All that said in deep respect to Robin Hanson and his quest.

  22. Crosbie Fitch said:

    Floated predictions are merely the subjects of the goods that are traded.

    Predictions are not traded, but probabilities of any given prediction being true are traded. Knowledge/information/expertise/wisdom/confidence is traded.

    A fish market does not trade species, but fish (instances of species). There may be many species of fish at a fish market, but it is not the species that are traded.

    You may say that a fish market is a market where at least one species of fish is traded.

    So a prediction market is a market where at least one prediction is the subject of probability trades relating to that prediction.

    But, given that all fish belong to a species, you could just as easily say that a fish market was a market in which at least one fish was traded.

    Similarly, a prediction market was a market in which at least one probability estimate was traded (with respect to a prediction).

    A prediction market is a market in which probability estimates are traded with respect to one or more floated predictions.

    A prediction market must therefore necessarily involve the floatation of a prediction.

    QED I concur with Chris Masse.

    FAME IS MINE!!! :)

  23. Jason Ruspini said:

    Masse writes: “I enjoy categorizing”

    Don’t we all.

    I think that is actually a telling part of the conversation.. people simply enjoy categorizing things. It reduces tension and gives at least the appearance of control.

    Despite the sexual joke here, I don’t mean to invoke the academic/real-world distinction. Many of the prediction market “academics” have more experience in commercial settings than those who label them as such.

  24. Chris Masse said:

    QUOTE

    Many of the prediction market “academics” have more experience in commercial settings than those who label them as such.

    UNQUOTE

    I can take the slam, no problem.

    Seriously: I meant that I would welcome professor Robin Hanson’s proposed survey on terminology.

  25. Robin Hanson said:

    Eric Z., I’m not sure that IEM, HSX, and the corporate markets could not achieve the hypothesized objectives if the traders knew of these objectives. But if so, I don’t feel a strong urge to adjust our vocabulary to help them with such deception. However, these functions would all come under the heading of “information” functions, so it would not be deceptive to call them “information” markets. But it seems more deceptive to use the word “prediction” to imply that they exist primarily because of the predictions they make. On the other hand, if “prediction market” just means that the markets make interpretable predictions about events, the phrase would be appropriate then.

  26. Niall O'Connor said:

    “Prediction markets are markets which are designed and run for the primary purpose of mining and aggregating information scattered among traders and which -by subsequently transforming this information into market values- produce directly interpretable predictions about specific future events.”

    Ah I see; so the fact that Tradesports, an online betting exchange, that amongst its activities, matches persons wishing to bet on sporting events, seeks to hide under the cloak of convenience provided by the term “prediction market,” will be enough to save it from the attention of the US Department of Justice??

  27. Chris Hibbert said:

    The DoJ will make its own definitions, and the fact that they overlap with other definitions will not slow them down, nor reduce the usefulness of those other definitions.

  28. Chris. F. Masse said:

    Prediction markets produce dynamic, objective probabilistic predictions on the outcomes of future events by aggregating information that traders bring when they agree on prices during their transactions.

    http://www.midasoracle.org/200…..ransactio/

  29. Chris. F. Masse said:

    Niall, then maybe you can sense the reason of the TradeSports-InTrade splitting.

Leave a Reply

Your email address will not be published. Required fields are marked *