Scott Sumner argues that Nominal GDP expectations are critical to macroeconomic health, and I find his arguments pretty convincing. Low NGDP expectations (relative to trend) imply that the money supply is low relative to the demand for money. Unfortunately it is non-trivial to estimate NGDP expectations because there are no NGDP futures markets. The absence of a good measure of NGDP expectations makes assessing the performance of the central bank more difficult.

Fortunately, as I recently learned, InTrade will list basically any contract you like on their exchange for a fee of $100. Since this is pretty cheap, I will create a set of NGDP prediction markets for several future dates. Hopefully these markets will be active and produce meaningful estimates of NGDP expectations. If these markets are active, I think it should be easy to convince InTrade to continue to create new contracts as old ones expire.

I think it’s important to design these contracts well so their prices are both accurate and useful. The InTrade Real GDP markets (under Financial->US Eco Numbers) are a good anti-example; the various Real GDP markets are either not very informative or not very useful. The contracts tied to the Advance Real GDP figures are totally inactive; all of the markets I saw had 0 trades. The markets that rely on Final Figures are not terribly illiquid, but since they are binary markets that indicate only the probability that Real GDP will be positive/negative they are not terribly useful.

I think it is clear that a single market with a continuous payout is preferable to a set of binary markets as a single market will be much simpler for traders to trade in.

As I see it, the critical contract design questions are:

  1. How far apart should the contracts be spaced? 3 months, 6 months, 12 months?
  2. Should the contract payout be proportional the level of NGDP or the rate of change of NGDP in some period?
  3. How far out should the contracts start? What are the most important expectations? Expectations about NGDP 6 months out? 12? 24?
  4. What BEA data should the contracts be based on? Advance, Primary, Secondary or Final NGDP estimates?

Finally, there is the more difficult question of how to generate interest in the markets if regular InTrade traders are not inherently interested in the contracts. The most direct method that I know of is to subsidize trading by employing some type of automated market maker that is willing to lose some money. Peter McCluskey used this method to subsidize conditional prediction markets during the last election (see here). Unfortunately, he doesn’t seem to think his experiment was very successful. However since the last election there has been some research done on improved automated market makers.