The Wisdom Of Crowds

I like probably every other person on the planet have a disposable Gmail account that I use to collect rubbish. One of the intriguing things about this account is the specificity of ads that Google generates. All of them relate to trading/investing which means in some way shape or form Google is reading my mail – which is a reason why I only use it for rubbish. Whilst perusing my mail I came across one of these ads that stated I could make 400% pa by listening to others in chat rooms. So being unable to resist I fired up the old compound interest calculator to what 400%pa compounded looked like. Apparently if I started with 10k after 10 years I would have $97mil – not bad going.

Leaving aside the clearly insane suggestion of 400%pa the central question to me becomes is the wisdom of the crowd greater than the wisdom of the individual. Much has been made of using social media for the generation of decisions, all of which are based upon the assumption that groups make better decisions than individuals. Based upon this premise it should be possible to harness this group think into something profitable hence the gibberish that is Stocktwits and this other idiot site promising me 400%pa.

However, it should be noted that this is an assumption and assumptions are dangerous things. There has been some work done on the crowds sourcing of decision making which has found that crowds can make more accurate decisions than an individual. But this is a statistical phenomena based upon the aggregation of values. It is not as if the crowd contains a collective wisdom that is being accessed in some touchy feely, the universe loves me sort of way. Because of this the notion of social influence is not taken into consideration when looking at crowd wisdom. To date questions surrounding crowd wisdom have looked at questions where there is no feedback in the decision making process nor is there any monetary gain or penalty locked within the decision.

To test the notion that crowds make better decisions in a setting where social feedback is available a very clever experiment was performed by Jan Lorenz and a team from the Swiss Technology University (How social influence can undermine the wisdom of crowd effect Jan Lorenz, Heiko Rauhutb, Frank Schweitzer, and Dirk Helbing. Proceedings of the National Academy of Sciences). A series of questions to where given to an  experimental group in order to gauge crowds decision making ability. The questions were as follows –

1. What is the population density in Switzerland in inhabitants per square kilometre?

2. What is the length of the border between Switzerland and Italy in kilometres?

3. How many more inhabitants did Zurich gain in 2006?

4. How many murders were officially registered in Switzerland in 2006?

5. How many rapes were officially registered in Switzerland in 2006?

6. How many assaults were officially registered in Switzerland in 2006?

It turns out that both the individual and the group are not very wise as shown by the table below.

However, the clever thing about this paper is not the demonstration that the individual and crowd are not that bright but rather than when given access to other peoples estimates the participants did even worse. It is thought that this decrease in performance occurred for three reasons.

1. Social influence Effect – the diversity of decisions within the group declined. That is people tended to revise their own estimates to be more in line with what the group thought. But if the groups consensus was wrong this revision only made the groups average or wise decision worse.

2. Range Reduction Effect – if you have a narrowly clustered range of values then this conveys a false confidence as to the veracity of the data. It provides confidence where none should exist. My take on this is to corrupt a saying by Socrates – just because every bugger thinks it’s a good idea it doesn’t mean that it is. As the experimenters say –

….would gain confidence in advice that is actually misleading. In fact, the close clustering around a wrong value makes the group less “wise” in the sense that the group delivers a wrong hint regarding the location of the truth. This is the case because the truth would not be located centrally but at outer regions of the range of estimates.

3. Confidence Effect – the experimenters define this as an undermining of the wisdom of the crowds based upon the previous two effects. Confidence grows as a function of the convergence of opinion but this convergence does not in any way boost the accuracy of opinions.

The implication of this for crowd sourced trading decisions is fairly obvious. Instead of improving the accuracy of decisions it actually diminishes their veracity where feedback in available and it is the very nature of instantaneous feedback that is at the heart of social media.

Therefore it cannot only be me that sees the irony in the name StockTwits.