As we saw in Weird Markets it can be remarkably difficult to pin down the idea of a “market”. Markets are what economists study, economists study markets; a perfectly circular relationship. On the other hand, markets do seem to be a fine way of deciding how to allocate scarce resources: economics is built on this singular, empirical observation.
Given this it’s not a huge jump to recognizing that markets may be a better way of making predictions about all sorts of things than relying on the kind of dubious expert forecasting we've met numerous times as in Forecasting a Financial Earthquake. Prediction markets have suddenly become interesting because they work; which probably means they’ll shortly stop working, because they’re now interesting. Tricky stuff, this reflexivity.
Stockmarkets are the most well known type of prediction market – they aggregate information from multiple sources and combine it all into a single number, the share price. On the whole they do this quite well, notwithstanding the fact that they frequently fail the test of perfect efficiency. Generally, though, they get a lot more things right than wrong.
Arguably one of the more powerful effects of markets is that they self-select their participants. As Robin Hanson explains in Insider Trading and Prediction Markets:
“People who realize they are not as well informed as average traders tend to stay away. People who do not realize they are not well informed lose and then go away.”
Of course, in stockmarkets this tends not to be true: their sheer popularity almost guarantees an endless pipeline of suckers who confuse and confound valuations and, at times, flood the market with arbitrary signals. But in other, less crowded markets the evidence is that the wisdom of informed crowds works rather well: they recently got the result of the 2012 US Presidential Election almost spot on, albeit the modelling methods of Nate Silver have got rather more attention.
In fact there is a very long history of prediction markets: people were betting on papal elections at the beginning of the sixteenth century and only the threat of excommunication by an annoyed Pope was sufficient to end the practice, in public at least. During the Battle of Britain London brokers ran contests on how many German planes would be shot down each night. Paul Rhode and Koleman Strumpf give a nice overview of the history in Historical Election Betting Markets: An International Perspective.
As the researchers point out the absence of betting markets in political prediction over the last fifty years has been an exception to the norm, and they suggest a number of reasons for this. Some of this was due to an expansion of other opportunities for gambling, and a subsequent lack of capital and some of it due to the rise of scientific polling methods as pioneered by George Gallup, which tended to eclipse the tried and trusted prediction markets.
However, the rise of the internet has seen a resurgence in betting markets for prediction; another facet of the long-tail, where ease of access and low-costs allow markets to be made in relatively low-volume, obscure areas. Such markets have been predictively successful in a wide range of areas as related by Robert Hahn and Philip Tetlock in Using Information Markets to Improve Public Decision Making:
“Las Vegas odds and point spreads predict the outcomes of sporting events better than sports experts. The prices in Iowa political markets are more accurate than concurrent opinion polls in forecasting elections 451 times out of 596. Hewlett‐Packard information markets beat official forecasts in predicting printer sales in two separate trials: fifteen times out of sixteen in one trial and six out of eight in the other. Even play‐money markets are a dominant source of information, outperforming four out of five columnists at forecasting Oscar winners in 2000”.
Hahn and Tetlock go on to argue that “information markets” would be a far better way of determining public policy than the standard mantra of cost-benefit analysis: at best a morally dubious way of allocating resources as we saw in Screwed: Fictional Profits, False Accounting and Financialization. They point out that even following relatively simple statistical rules usually manages to outperform the experts, a finding we discussed in Can Software Beat Penny Flippers?, and argue that markets are likely to yield better results.
A cynic might observe that a dart chucking monkey could also do that, but let's not be churlish ...
The idea is that it is possible for governments to use information markets to form policy; by auctioning off the potential rights to implement certain projects against a given baseline. The auction price achieved is essentially equated to the net benefits of the project and the project is actually implemented only if the overall benefits make it worthwhile to do so. Underlying this is the idea that, wherever and whenever possible, we should use the natural properties of markets to achieve the best economic results; after all, central planning has never been the most effective way of allocating resources.
Effective or not this leads to at least a couple of issues which aren’t easy to find an answer to. Firstly, all market led approaches require elements of financialization: as Hahn and Tetlock themselves note:
“The government must specify and monetize all verifiable benefits accruing from a possible project”.
Which is OK, but often leads to the reintroduction of subjective, non-market driven management through the back-door. The second issue, however, is potentially even more tricky because, as we often see in other markets, introducing market-led reforms opens up the possibility of manipulation – and when we’re into the creation of public policy or even voting for the next leader of the free world this may matter a great deal more than whether a few behaviorally compromised investors lose money.
The evidence of manipulation is mixed; Rhode and Stumpf find in Manipulating Political Stock Markets: A Field Experiment and a Century of Observational Data that such attacks tend to be difficult to maintain beyond the short-term but when Cary Deck, Shengle Lin and David Porter looked at the issue in Affecting Policy by Manipulating Prediction Markets: Experimental Evidence they:
“Find clear evidence that highly incentivized manipulators can destroy the predictive power of a prediction market. That is we have identified a case where manipulators do cause human forecasters to make predictions that are no better than random guessing would generate showing that prediction markets can be manipulated. In fact, the manipulators are so effective that the econometrician observing the market cannot predict the truth more accurately than a coin flip. This finding demonstrates that policy makers should not indiscriminately rely upon market predictions.”
As usual care needs to be taken in interpreting these results – the Deck, Lin and Porter research is based in laboratory conditions while Rode and Stumpf take their data from real-life situations, which may be more authentic but tends to be harder to analyse. Either way, in relatively thinly traded markets the possibility for manipulation is an ever-present danger.
Finding a way of making more objective, more effective policy decisions around (say) levels of healthcare and environmental investment is a highly worthwhile exercise. In a world dominated by the mantra and philosophy of markets, whatever they are, it’s an obvious step to attempt to apply this knowledge in areas such as these, which otherwise tend to be dominated by dogma and plagued by popularism. And, as a method, it's no less ethical than the alternative of cost-benefit analysis. But markets are morally blind: we can’t simply delegate our decision making unthinkingly to the invisible hand.
Prediction markets offer great scope for improving our world, both by reducing our dependence on the vagaries of political ideology and by improving our use of the limited resources of island Earth. However, objectivity is a chimera, don’t trust the leader who delegates responsibility to the wisdom of crowds: it’s surprisingly limited, subject to the vagaries of mass psychology and exposed to nudge-led manipulation. As usual, no free lunches, but maybe offering a taster or two to whet the appetite?