Policy Markets

by Robin Hanson

What They Are

A policy market is a market created to directly inform policy decisions with its price. That is, while the market may also serve other functions, such as hedging or entertainment, its primary function is to create prices which embodies information which is directly relevant to people considering some choice between policy alternatives. The perceived function of most financial markets, in contrast, is to allow people to hedge and rebalance their portfolios. (See, for example.) And the perceived function of most gambling markets is to entertain.

For example, consider the following policy questions:

These are questions about verifiable and directly relevant consequences of various policy choices. But while there is no shortage of pundits with opinions on these sort of questions, the possibility of self-interested bias makes most people reasonably skeptical about such opinions. And so people lack reliable sources of information to guide their policy choices.

The right market, however, could give us neutral expert estimates on such question! For example, an estimate of the probability that the U.S. will go to war given that Clinton is elected would be given by the market price (or odds) of bets on war, with the bets being called off if Clinton is not elected. And the price of bets on war which are called off if Clinton is elected would estimate the probability of war if Clinton is not elected. The difference between these prices estimates whether war is more or less likely given we elect Clinton, and is directly relevant to the question of whether we should re-elect Clinton.

This same called-off-bet approach can be used with any policy question. For example, prices in a market which traded a stock market basket (such as S&P500 futures) for cash, but which called-off these trades depending on who became the next president, would estimate which current candidate would be best for the stock market (and should be insensitive to who actually wins).

Why We Need Them

This market consensus estimate has many advantages over alternative ways of generating consensus estimates on policy questions. It would be precise and continuously updated, relatively cheap to create, and would be open to contributions from anyone, regardless of how articulate they are, what degrees they have, or whether they look good on TV. And anyone who claimed that the market price was a poor estimate could be subject to the challenge: "Put up or shut up"; they should expect to make money by trading and helping to correct this estimate.

Since it would be expensive to bet wrong in such a market, the market price would be an expensive signal, aggregating and communicating information not easily found in the "cheap-talk" of media commentary, nor easily disentangled from prices in other markets (if it can be found there at all). In fact, I argue that you should have a free-speech right trade in political policy markets.

Note that while the prices in most markets do contain information, some of it at least indirectly policy-relevant, most markets are constructed for other purposes. And in those few markets that are contructed primarily for the information they create, such as the Iowa Electronic Markets, the questions are at best indirectly relevant to specific policy choices. (They bet on who will be elected, not on what would happen then.)

For more information on other uses of such markets, on prototype trials, on current legal barriers, and on publications on the topic, see my Idea Futures page.

How to Make Them Easier

We can avoid the paper-work involved in calling off bets, if we instead trade certain "contingent assets" with each other. For example, consider the following four possible states of the world, depending on whether Clinton is re-elected as U.S. president, and on whether the U.S. is involved in a war during the next presidential term of office.

ClintonNot Clinton
WarC&WNC&W
Not WarC&NW NC&NW

Imagine that we created four types of contingent financial assets, each worth money in only one of these states. For example, C&W = "Worth $1 if Clinton and War", NC&NW = "Worth $1 if Not Clinton and Not War", etc. And imagine that we created markets where people could trade these assets.

Then the market price (i.e. asset amount ratio) of trades between assets C&W and the package C = C&W + C&NW would be an estimate of the conditional probability of war, given Clinton being elected. Similarly, the market price of trades between NC&W and NC = NC&W + NC&NW would estimate the probability of war, given that Clinton is not elected.


Robin Hanson April 25, 1996
known by AltaVista