structures also include mechanisms for ensuring that traders can and do make good on their offers, and
they include choices about who can see what offers and trades.
Whatever the details of the exchange mechanisms, the important thing would be the appearance
of two market prices, or trading rates of exchange. In the first decision market we need, people would
exchange one asset “Pays $L if S” for some chosen fraction of the asset “Pays $100 if S.” In the second
market people would exchange one “Pays $L if not S” for some fraction of “Pays $100 if not S.” These
fractions could be taken as market estimates of the conditional expected values E[L | S] and E[L | not S],
respectively. If the first fraction were higher than the second, market speculators would be saying that
they expect U.S. lifespan to be higher if the SPS bill is approved than if not. Such estimates are what we
would want to advice SPS policy.
Requirements for Decision Markets
We have good reasons (including formal theory, lab experiments and field data) to think that decision
markets can give accurate estimates. But decision markets are not cure-alls; they require:
Important-enough Claims
There are fixed costs of using this decision market process. You have to set up markets, get
traders to pay attention, and so on. Yet, if other forecasting institutions are working well, market estimates
might only be slightly more accurate than estimates from other institutions. The policy question asked
thus needs to be important enough for this added accuracy to be worth the added fixed costs.
Careful thought should be given to what question one wants to pose to market speculators. If you
ask the wrong question, the answer you get may not be very useful. For example, if reducing lifespan
inequality is as important a benefit of SPS as increasing average lifespan, one might want to ask markets
about inequality as well. If one cannot afford to ask about both, one might ask about a weighted average
of the two.
Enough Influence
9
Convention