Behavioural Finance’s Smoking Gun

Linda

Here’s a classic demonstration of behavioural finance in action, proving the irrationality of humankind. It’s the (in)famous Linda problem:

Linda is 31 years old, single, outspoken and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in antinuclear demonstrations. Which of these two alternatives is more probable?

(a) Linda is a bank teller.
(b) Linda is a bank teller and is active in the feminist movement.

Most subjects choose (b) and are informed that they’re irrational because the conjunction of two events – Linda is a bank teller and active in the feminist movement – is less likely than her just being a bank teller, regardless of her leisure interests. This is known as the conjunction fallacy. Unfortunately there’s a teensy little problem with this finding.

It’s wrong and it’s the smoking gun of behavioral finance.

Probability Analysis isn't Rationality

Now, in terms of probability, the researchers are correct. It’s much more likely that Linda will only be a bank teller than be a bank teller and a feminist activist. Unfortunately to go from that position to arguing that people who choose the second option are irrational and are being affected by some behavioural bias or other is a huge jump than can only be explained by using too many narcotics. Allegedly. Probably.

Classically this “irrationality” is explained in terms of something called the representativeness heuristic, which is, roughly, a rule of thumb where people make decisions about something by judging how similar that something is to some readily available and similar data. This is an attractively simple idea but one that comes swathed in clouds of confusion: it superficially explains everything but is of little use when it comes to actually predicting behaviour – a problem that people trying to solve real-world problems, like when to invest in stocks, often level at behavioural finance.

Stat Homunculus on the Brain

Now, if you think about this for a moment what the research is telling us is that to "think rationally" is to behave as though inside our heads is a tiny statistical homunculus performing conjunctions and other statistical stuff. In fact, if we analyse the theories behind most of behavioural finance (and, to be fair, traditional economics) it all assumes that to behave rationally is to reason in terms of probability.

A number of researchers have begun to question whether this makes any logical sense in the real world. Take the Linda problem, for instance. Here’s a variation of the Linda problem carried out by Gigerenzer and colleagues:

Linda is 31 years old, single, outspoken and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in antinuclear demonstrations.

There are 100 people who fit the description above. How many of them are:

(a) bank tellers
(b) bank tellers and active in the feminist movement

Guess what happens to the results? Yep, the conjunction fallacy disappears – far more participants now choose (a) than (b). Something odd and deep and important is happening that the mainstream of behavioural finance research is completely failing to understand.

Frequencies, Not Environment

The second Linda question is directly asking a question about frequencies – it’s asking us to judge a probability in a direct way that is hard to misunderstand. And lo, our statistical homunculus jumps into life and correctly figures out how to handle this.

However, the first Linda question isn’t asking us about a probability frequency, it’s asking us about Linda. In the real world how likely is it that someone will present us with loads of information about Linda’s feminist tendencies and then expect us to ignore the information? It’s just not realistic – we look at the question, decide that we’ve been told stuff that must be relevant to the answer and then use it. The homunculus barely stirs from its sleep because it’s not a statistical question.

Economic Research Fallacies

The idea that the human mind is based on an internalised statistical decision making algorithm is one we’ve met before. It’s a research fallacy which each successive generation of academics falls into by firstly using their latest research tools – statistical analysis packages, digital computers, etc – as metaphors for the way the mind works and then assuming that this is the way the mind actually works. There is a huge behavioural bias behind behavioural finance but it’s not the experimental participants who are subject to it, but the experimenters.

Behavioural finance assumes that people don’t behave rationally but would if their wetware allowed them: we only behave a bit stupidly because we don’t have the processing power up top to implement the necessary statistical decision making packages efficiently. This is a step away from homo economicus, the ideal economic, rational man but one that turns out not be useful: people correctly identify, based on the situation, that Linda’s more likely to a active feminist bank-teller, it’s only economic researchers that don’t.

Basically we’re not perfect by design, and we’re bloody good at it.

Exploding Prospect Theory

Here’s Gerd Girgerenzer again:

“It is understandable that when heuristics were first proposed as the underlying cognitive processes in the early 1970s, they were only loosely characterized. Yet, 25 years and many experiments later, explanatory notions such as representativeness remain vague, undefined, and unspecified with respect both to the antecedent conditions that elicit (or suppress) them and also to the cognitive processes that underlie them. My fear is that in another 25 years researchers will still be stuck with plausible yet nebulous proposals of the same type: that judgments of probability or frequency are sometimes influenced by what is similar (representativeness), comes easily to mind (availability), and comes first (anchoring). The problem with these heuristics is that they at once explain too little and too much.”

On its own this would be provocative but addressed as it was to Kahnemann and Tversky, the founding fathers of behavioural finance, it’s fighting talk. And, indeed, a bit of a scrap was had with predictably indeterminate results. However, you can see the thrust of Gigerenzer’s argument: everything about behavioral finance is psychologically plausible yet it doesn’t actually seem to be practically useful.

Lies, Damn Lies and ...

Gigerenzer has expanded his thesis away from the Linda problem into all sorts of other areas of behavioural bias. So, for example, in an early paper he and his colleagues looked at overconfidence bias, a staple of behavioural research which is consistently reported as stable under all conditions. They promptly showed that they could manipulate conditions to make this appear, disappear and invert dependent on how their experimental situation was constructed.

Underlying all of this is subtle argument about the misuse of probability and statistics which is a bit too obtuse even for this journal. However, Gigerenzer basically believes that behavioural researchers looking at confidence statements are misusing statistical analysis to prove their points.

The Smoking Gun

This is, indeed, a smoking gun for behavioural finance. It also implies that there’s some other underlying mechanism of human judgement that’s common to all supposed biases, something that resolves the existing cacophony of competing biases which seem to randomly take turns in driving our behaviour.

What this might be will have to wait for another time, but it’s an encouraging thought that maybe we’re not so stupid as the legions of norm-defining economists who want to train our brains to be the same as the statistical analysis packages that sit on their PC’s think we are. But really, we knew that anyway, didn’t we?



Related Articles: Unpredictably Rational, Investor Decisions, Experience is Not Enough, The Special Theory of Behavioural Finance