A pitfall in behavioural economics
Danny Finkelstein describes how the Tories are embracing behavioural economics. This chart shows one of the pitfalls in this. It shows the performance of a portfolio of loser stocks - an equal-weighted basket of the 20 biggest fallers in the FTSE 350 in the previous three years, re-formed at the start of every quarter. 
This portfolio has hugely under-performed the market not just during the credit crunch but in the years before then.
What’s this to do with behavioural economics? Simple. It contradicts the message of one of the founding papers in behavioural finance. Back in 1985 Werner de Bondt and Richard Thaler published a paper, Does the stock market over-react? (pdf) in which they showed that loser stocks - the biggest fallers in the past three years - subsequently out-performed the market. This, they hypothesised, was because investors over-reacted to long runs of bad news, causing shares to eventually become under-priced.
Any UK investor who followed this message recently would not have thrived*.
This is an example of the first law of policy-making, Goodhart’s law: any statistical regularity breaks down when relied upon for policy purposes.
There’s a reason for this - people learn. For example, stock-pickers who read de Bondt and Thaler a few years ago would have bought past losers, believing them to be under-priced. But this buying would have stopped prices falling sufficiently. The upshot is that loser stocks did not become genuinely under-priced.
And if the institutionally imbecilic fund management industry can learn, so can Joe Public. Take an example. Nudge thinking says we can rely upon peer pressure to get people to do what we want, so if we say: “give to charity, as most people do”, more people will give. But what if people learn about Nudge? Then they’ll see through this, and think: “Hey, we’re being manipulated here.”
What’s this to do with behavioural economics? Simple. It contradicts the message of one of the founding papers in behavioural finance. Back in 1985 Werner de Bondt and Richard Thaler published a paper, Does the stock market over-react? (pdf) in which they showed that loser stocks - the biggest fallers in the past three years - subsequently out-performed the market. This, they hypothesised, was because investors over-reacted to long runs of bad news, causing shares to eventually become under-priced.
Any UK investor who followed this message recently would not have thrived*.
This is an example of the first law of policy-making, Goodhart’s law: any statistical regularity breaks down when relied upon for policy purposes.
There’s a reason for this - people learn. For example, stock-pickers who read de Bondt and Thaler a few years ago would have bought past losers, believing them to be under-priced. But this buying would have stopped prices falling sufficiently. The upshot is that loser stocks did not become genuinely under-priced.
And if the institutionally imbecilic fund management industry can learn, so can Joe Public. Take an example. Nudge thinking says we can rely upon peer pressure to get people to do what we want, so if we say: “give to charity, as most people do”, more people will give. But what if people learn about Nudge? Then they’ll see through this, and think: “Hey, we’re being manipulated here.”
So, perhaps behavioural economics might work for policy purposes in the short-term. But, in some cases at least, it could fail once people wise up.
* Strictly speaking, I’m not comparing like with like. De Bondt and Thaler say that much of the good performance of past losers comes 2-3 years after portfolio formation. However, it’s not clear this is true recently; the past losers in June 2005 have since only performed in line with the market, as good performers (Cookson, Invensys) have been offset by stinkers like JJB or Logica. And recent UK data does not support de Bondt and Thaler’s view that past loser portfolios earn great returns in January.
Nudge has its place, but spotting the Mavens or "information specialists", (tipping point) and getting them to lead the group is much more effective both for good and bad (its up t us to make it for the greater good). But then again you are not a fan of leadership are you Chris?
seems to work here.
http://www.eurekalert.org/pub_releases/2008-05/l-tis050708.php
Posted by: passer by | July 17, 2008 at 01:32 PM
Isn't just another application of Heisenberg's Uncertainty Principle?
Posted by: Wolfie | July 18, 2008 at 08:54 AM