PsyFi Search

Tuesday 22 May 2012

Parsimonious, Big Picture Behavioral Bias

107 Ways of Being Wrong

As we’ve seen in The Big List of Behavioral Biases, there are 101 (well, 107 at the time of writing) ways in which people exhibit irrational biases. The basic idea, that we’re affected by these biases in predictable ways, is now well accepted. The problem is that there are simply too many biases for this to be the be-all and end-all of the explanation of market irrationality.

So while the basic concepts of behavioral finance are understood, in the sense that irrationality is a driving force of market misbehavior, the underlying mechanisms by which they operate are not.  There's not much point in claiming that these behaviors are predictable if you can't use them to predict anything, So what's the big picture of behavioral bias: how does it all fit together?

Cottage Industry Economics

Although finding some new bias, putting a name to it and publishing has become a cottage industry for economics and psychology researchers the underlying problem remains that there’s no unifying theory of how everything fits together. This isn’t a minor problem, either, because unless we possess that theory we can’t use behavioral finance to do anything useful in the way of prediction. This is one of the criticisms levelled at the behavioral approach by economists from the more classical areas of economics, who point out that their approach does yield predictions. Unfortunately these usually turn out to be incorrect, but the idea seems to be that it’s better to be precisely wrong than roughly right.

The kinds of problems that behavioral research runs into can make your head spin. In many cases multiple biases collide, predicting different outcomes. In others, researchers are able to show that by changing the context of their experiments they can reverse results – which may turn out to be equally illogical, but are no less troubling (see: Behavioral Finance's Smoking Gun).  In general, behavioral finance describes what's gone wrong but doesn't help predict what's about to happen next.

If we're to put behavioral finance on a solid footing we need to look to the principles of science, and develop testable hypotheses.  One of the main concepts behind the scientific method is to look for unifying principles, amongst which is a rule of thumb that prefers a hypothesis which explains the most phenomena using the least assumptions.  This is the idea behind Occam's Razor, commonly referred to as parsimony

This approach has defined the trajectory of scientific knowledge, as we have increasingly found ways of integrating apparently disparate areas of knowledge while reducing the range of underlying assumptions we need to make. There’s no obvious reason why this should be different for behavioral economics – ultimately we'd expect that the vast range of biases should simplify down to a relatively small number of underlying causes. 

Genes, Revisted

A couple of recent research papers have made a gesture in the direction of an integrated approach. The first of these starts with a fairly simple hypothesis – that many biases originate in our genetic makeup. We’ve previously seen that genes can explain various types of hormone related bias in behavior (see: Are You A Born Investor? and Craving a High: Trading on Dopamine) so a wider application of this type of idea wouldn’t be at all surprising. In Why Do Individuals Exhibit Investment Biases? Henrik Cronqvist and Stephan Segel looked at the differential impact of genes and education on behavior in non-investment domains and have found a correlation with similar behavior in investment:
“We find that a long list of investment biases, e.g., the reluctance to realize losses, performance chasing, and the home bias, are "human," in the sense that we are born with them. We base this conclusion on empirical evidence that genetic factors explain up to 50% of the variation in these biases across individuals.”
The idea is that investment biases are generated out of more general, non-investment related preferences and that these are genetically moderated. So, for example, the familiarity bias (see: Home Is Where The Risk Is), the tendency to prefer stocks with which we’re personally familiar, is correlated with the person’s preference for where they live and where their spouse originates, and this, in turn, has a genetic basis   – at least to the extent that genetically identical twins are more likely to exhibit these biases than fraternal twins.

Noisy Signals, Revisted 

A second idea, as a mechanism for integrating biases together, is to look at the way the brain processes information. We’ve already seen that noise can impact information processing at a gross level – by introducing noise into the signals generated by markets it’s possible to mask the underlying information being produced sufficient to confuse investors (see: Idiot Noise Traders). However, information processing limitations should also impact people at the individual level: the brain is an information processing system, and it’s a reasonable theory that it can be impacted by noise sufficient to introduce apparent biases into its outward behavior.

Martin Hilbert has looked at this in Towards a Synthesis of Cognitive Biases: How Noisy Information Processing Can Bias Human Decision Making. In this he argues, and shows mathematically, that introducing noise into the brain’s information processing pathways can theoretically create at least eight different biases. The range of biases he looks at is interesting in that, conceptually at least, they seem to form a distinct sub-set. So, for example, he covers conservatism, aka the underestimation of high likelihood events and the overestimation of low ones, and illusory correlation where judgement is based on a relationship we expect even if it doesn’t really exist.

The paper also looks at the relationship between over-confidence and the hard-easy bias.  The former is where we are over-confident about the accuracy of our judgements, while the latter demonstrates under-confidence in comparative situations where the problem faced is very difficult, and over-confidence in the situations where the problem faced is easy; which ignores the fact that everyone else must face the same problem. The difference is context, and Hilbert shows that these biases can be explained as related consequences of noise during memory retrieval. 

Towards the Big Picture

Although this synthesis only covers a small subset of the total range of biases it does suggest that many of them may arise from similar causes, and that these causes originate from mechanisms that are already well understood in other arenas. Interestingly, Hilbert is also able to come up with a mathematical definition of the different biases under consideration – which both allows for better empirical investigation and removes some of the ambiguity associated with the often rather vague association of a name with a particular set of behaviors, something usually accepted by researchers without much debate.

These attempts to provide a partially integrated view of behavioral biases are the first, halting efforts to produce a really scientific synthesis of the topic. After all, it’s no good claiming that biases are both irrational and predictable if you can’t then use that information to useful effect. Combining these research ideas gives us an initial suggestion that our external behaviors are biased by noise in the system and that the precise effects will be moderated by our genetic makeup. Neither should be surprising, as both genetic disposition and the laws of information processing are already well understood as limitations on behavior in other sphere of operation.

Of course, these initial ideas will doubtless turn out to be wrong in some interesting and unexpected ways. Yet the basic idea, that we’re creatures sculpted by our genetic makeup and then baffled by our noisy environments is not unexpected. A nice next step would be some ideas about how to use this information to make us better investors.

Related articles:
The hard-easy bias added to The Big List of Behavioral Biases.


  1. One reason of the difficulty to find applicable lessons from bias may be that some of them are harmless.

    For example, if you reproduce synthetically the "reluctance to realise losses" bias, e.g. via stop losses that are a sizeable and fixed multiple of a fixed take profit limit, you end up with a reshuffle of the pay-off distribution that is essentially neutral: the more frequent small gains balance the larger but rarer losses; there's a paper I've lost the reference of that proves that point more formally than I can do here.

    You can observe a similar phenomenon with many bias-driven selection criteria, if it's arbitrary but broad enough it will be neutral: if you compose a well diversified index of companies whose name starts with a vowel, you should get pretty close to the matching full universe index, so a vowel-start fetish ends up harmless.

    You could even argue some biases might help information overload and analysis paralysis problems: at some point you must trim the information set you use for making investment decisions, which will necessarily be arbitrary (if you could process the info you exclude, you wouldn't have to exclude it) so you may as well use moderate biases for the purpose of making your investment universe tractable. Similarly, if people spend time and effort trying to manage their harmless biases, it's less time available for productive endeavours.

  2. Very interesting point and something I have often thought about, and have sometimes been asked about in various questions including: Where is this all leading? Are not many of these biases basically slightly different forms of the same thing? How can all this knowledge be practically applied?

    My responses tend to follow the following: Firstly I think this whole topic is still in its infancy (relatively speaking), one can even argue that economics may still be in its relative infancy. Theories are still forming, many of them can not be truly tested empirically and thus evidence emerging over time is possibly the best tool available.

    The study of human behaviour and decision making is such a huge, complex and fertile area with inputs from so many sciences, disciplines and field and outputs affecting so many things, even to the extent that it is reflexive. It is quite possible it will never be fully understood, or perhaps not for several generations at a minimum.

    With regard to practical usage. Greater knowledge and understanding surely has to be an advantage to all aspects of activity. - As an example, I can only hope that greater understanding on Behavioural Finance can start to permeate areas such as Risk management, which still functions and practices largely based on quantifying and reducing risk to numbers. At some stage the RM profession is going to have endorse behavioural practices into its functioning. Would risk managers fully understanding behavioural finance have allowed the situation at JPM to have evolved the way it did, it crossed so many boundaries on so many levels that a simple behavioural checklist would have flagged and been alert to. Behavioural practices have been used to improve the efficiency and safety standards in medicine, the military, shipping, flight, at some point they will be adopted by the rm profession and perhaps regulators. - I believe they can also improve ones ability to improve their analysis, familiarity with the concept of confirmation bias, and putting into practice some sort of check on this can improve the quality of one's analysis.

  3. I've used this post within a discussion question I have asked on my linkedin group - The question is as follows : How can Behavioural Finance and Behavioural Economics can be applied on a practical level?

    Link below.

  4. Hi Steven

    Well, I'm actually in process of writing a book on this topic, so I'd like to reserve some of my powder for that! However, I think there's a strong case for arguing that purely by publicising behavioral biases we help to eliminate them. Market forces do work, a lot of the time, and some levels of irrationality can be removed through market mechanisms. I'm not convinced that fully understanding behavioral psychology, whatever that means, will completely remove bias in financial markets, but I agree that as wider acceptance of the ideas increase then you'll get an improvement.

    I also like the idea of checklists, but as a future article discusses, they're not quite the panacea they appear ...