“Wisdom of the crowds” is a pretty solid phenomenon. Ask a thousand people to guess the number of sweets in a jar, and the average (mean) of their guesses tends to be damn close to the actual number.
What’s more interesting is whether the same idea works, not just to guesses, but to forecasts. Specifically, economic forecasts. If you ask a thousand people to guess what the unemployment rate will be in two years time, how will they do?
There are certainly reasons to be hopeful. Information is widely distributed, with little advantage accruing to experts; and in fact, unlike with simply counting sweets, there’s likely to be a fair few people with “inside” information (hiring plans, perhaps, or a feel for how their sector is moving), which they may use to inform their guesses. Mix together enough guesses, and you could generate insight.
That’s what the Adam Smith Institute and Paddy Power are hoping; the two have teamed up to offer markets in key UK economic statistics. You’ll be able to bet on what the rate of inflation and unemployment will be in June 2015; the ASI’s Sam Bowman writes that:
By combining the local knowledge of thousands of people, betting markets can outpredict any panel of experts. If these markets catch on, the government should consider outsourcing all of its forecasts to prediction markets instead of expert forecasters.
But there may still be some problems, both with the idea and its implementation.
Betting markets are indeed a theoretically great way of harnessing the wisdom of the crowds. As Bowman writes, the fact that people put money on their predictions means that more confident predictions are weighted higher, and vice versa. But the necessity of teaming up with a bookmaker to launch the idea means that there is a major distortion: the odds the bookie has set. Punters can get 7/2 that inflation will be greater than 5 per cent, and just 5/2 that it will be between 4.01 per cent and 5 per cent. That means that someone who thinks that inflation is most likely to be around 4.75 per cent may take advantage of the higher odds offered if they guess slightly higher. It also means that what Paddy Power think is most likely will skew the guesses.
A better version of the same idea would be to create a prediction market. The difference between the two is that in a market, the crowd takes the role of bookmaker as well as punter. The odds themselves get set procedurally, based purely on where people are betting, and so there’s no chance of a bad guess on the bookies’ part skewing the predictions.
But even if the market was designed to perfectly get the true thoughts of everyone in the crowd, there’s still reasons to doubt that it can be that good at forecasting economic data.
There’s quite a specific set of conditions which are required for crowdsourcing to work. James Surowiecki, who coined the phrase “wisdom of crowds”, describes four: Diversity of opinion, independence of opinion, decentralisation of action, and aggregation of information. Of those, the one which is the most problematic in this case is independence. People’s guesses aren’t secret, and they affect others. That means you could end up seeing a circular mill, where everyone reinforces everyone else’s beliefs to the extent that the crowdsourcing breaks down. Think: do you hold your beliefs about what might happen to the unemployment rate based on investigation of the primary data, or based on collation of expert analysis? If it’s the latter, you’d be a net harm to the crowdsourcing, contributing largely to the flocking problem.
It would still be nice to get more financial bets. But that’s mostly so that I could join in my sportier friends in having something where I feel like my expertise could win me a bit of cash; when it comes to actually trying to work out what will happen, we might have to stick with older methods.