The Open Question About Prediction Markets
As Robin Hanson succinctly put it, "the whole idea of prediction markets is that someone who wants to know the answer to a question might be willing to pay to create a market to entice traders to help answer their question." Identifying these sponsors along with the best means of subsidization and the best way to aggregate diverse, knowledgeable traders is THE open question of prediction markets. Real-money markets whose primary purpose is to reveal information and that provide no significant risk-sharing, capitalization or entertainment value must be subsidized, otherwise no-trade conditions will snuff-out activity. Why trade against someone with better information unless you are consciously paying them for it? (This could be what Chris Masse means when he says that aggregating information is not the primary purpose of prediction markets — that a pure prediction market is an imaginary locus.) And needless to say, the market participants can't be in a position where they would be able to otherwise comparably profit from their information without the market structure.
There is also a trade-off to the sponsor in terms of making such markets public in order to attract as diverse a set of traders as possible. Insofar as the sponsors consider the information they are seeking to be valuable, maybe they don't want to make it immediately public through an open market?
If the markets are subsidized via market-making as with a market scoring rule, this presents another reason why binary options might not be the best fit for aggregating information: the sponsor is exposed to much higher percentage losses.
There is no general overarching answer to "the" question, but rather many specific answers that depend on identifying individual niches and inefficiencies. Once these are identified, issues such as interpreting prices are secondary. Promoting corporate prediction markets does approach a general answer however, and here the subsidization happens by way of the company paying for the infrastructure and letting employees' time be diverted, as opposed to direct real-money market-making. Not that corporate prediction markets must operate with play-money. It's not too hard to imagine employee-traders being rewarded by varying their end-of-year bonuses by market winnings or losses, and perhaps the markets could be shuttered during normal working hours to mitigate any time diversion.
Here insider-trading laws may intrude, which brings-up the legality issue that many, including myself at one time, have considered to be the most important open question on prediction markets proper. But perhaps the legality issue is a little overrated and US-centric? Real-money markets in much of the Anglosphere and play-money corporate markets are relatively untouched by it, and there may be some viable strategies for US-based firms seeking to operate hybrid markets. More on this last point in the coming weeks...