Wednesday, October 29, 2008

On Economic Thought - 6

From behavioural finance, we move on to behavioural economics. The original and most pervasive idea in behavioural economics is that of bounded rationality. This simply means that contrary to the assumptions of neoclassical economics, human beings are not perfectly rational and often make errors in judgment, though full rationality could be a plausible first cut approximation. We do not optimize economic well-being functions like utility, because we lack the information processing capabilities required to optimize. Instead, we satisfice, i.e. we have certain targets in our mind that we set through a combination of experience, incomplete information and intuition, and then we work towards achieving those targets. We are happy with results that are sub-optimal but still good enough to meet the thresholds that we set for ourselves.

Computer engineers or others with some exposure to the field of artificial intelligence will recognise the difference between optimizing and satisficing to be exactly analogous to the difference between a best-result algorithm and a quickest/computationally easiest good-result heuristic. Thus, the central idea used to move from standard programming to AI is exactly the basis of the move from neoclassical to behavioural economics. It is no surprise then that Herbert Simon, the economist who introduced the idea of bounded rationality, was also a computer scientist. In fact, he is the only person ever to win both the Nobel in economics as well the ACM Turing Medal (possibly the highest recognition in the field of Computer Science). Herb Simon is probably the economist most under-explored by the mainstream, though he finds some pride of place in management literature.

Apart from bounded rationality, behavioural economics also makes some very interesting changes to neoclassical utility theory. Research by Kahneman and Tversky showed that people are not risk-averse, rather they are loss averse, i.e. they place higher weight on the possibility of losing some money as opposed to gaining the same amount. People also measure outcomes relative to some benchmarks that they set for themselves, and a typical example of this is an investor refusing to sell a stock at a loss in a market that is going downwards. Thus, if I have bought a stock at Rs 1000, I will refuse to sell it at 900, even if I believe that tomorrow the price is going to be 800 and rationally, I am better off selling today and holding cash. People value the same thing more once they posses it (status quo bias and endowment effect) and discount future cash flows by inordinately large discount rates (hyperbolic discounting). Importantly, people give different answers to questions that are logically equivalent depending upon the way they have been framed and give logically incorrect answers to seemingly simple rational choice questions if they are posed differently (framing). All this work led to what is known as cumulative prospect theory, which Avataram has sometimes written about. It is essentially an alternative to neoclassical utility theory.

Many of these assertions explain a lot of phenomena that we observe in practice to be near-ubiquitous. The endowment effect results in people demanding higher prices for houses they own than they would be wiling to pay themselves if they were customers, resulting in many unsold houses and reduced liquidity in the real estate market. Hyperbolic discounting means that people systematically err by saving too little for their retirement, something that we see rationalized away in the name of 'living for the moment'. Framing effects imply that people will respond to the exact same information and decision problem differently if the information is unstructured.

Behavioural economics has not only introduced new ideas, its central technique of conducting experiments with people to reveal utility preferences and cognitive biases is also very different from the largely arm-chair approach of mainstream microeconomics. It seems only intuitive that a science that is founded upon assertions of human behaviour follow this approach, but for some strange reason economists of the past have had a disdain for experimentation. The Austrian school was the epitome of such disdain. But more on the Austrian school later.

Back to experiments in economics. Enthused by what I read about behavioural economics, I tried conducting an experiment myself on the relevance of how decision problems are framed.

Consider the following question

There's a red-haired woman who is single. She plays tennis, is a feminist and a member of Greenpeace. What is she more likely to be ?
1) A newsreader
2) A lesbian newsreader

If you answered 2, you'd be in the majority but you'd be wrong. While this may seem to be an effort to illustrate the stereotypes and biases prevalent in society, it is actually nothing of the sort. Bias or no bias, the set of lesbian newsreaders is a subset of the set of all newsreaders. The probability that somebody is a newsreader is thus always greater than or equal to the probability of her being a lesbian newsreader. If you are a rational economic agent who makes the correct choices in decision problems, you would have chosen 1, irrespective of the mass of information presented to you and irrespective of any biases that you may have. However, the vast majority of people across levels of education, political beliefs and geographies pick 2.

I got this simplest of decision problems off something I was reading on the internet, and have posed it to many friends of mine, all MBAs from top institutes in the country. Barring one, every single one of them has given me the wrong answer. One of them (Frust), when offered the solution, suggested that the context matters - had the question been asked in a test measuring his quantitative abilities and knowledge, he might have given the correct answer. Over a dinner table conversation, he faltered. He may very well be right, and that only goes to illustrate the overarching importance of framing - the way and context in which information is presented to you may radically alter your decision. And we all know that real world information and decision choices resemble a dinner-table conversation far more than a probability test.

(You could also use this fact to infer that MBAs are stupid. That conclusion, however, will be dependent on your biases and has nothing to do with such googly-type solutions to decision problems.)

While Herb Simon would count as the big daddy of all behavioural economists, Richard Thaler is the one most prominent now. However, whenever I have read about Keynes and his theories, I have had a sneaking feeling that he may be the original behaviourist. The man who called the stock markets a beauty contest seemed unlikely to be anything else. Indeed, recent research into interpreting his works has followed that path, and the foremost researcher on that front happens to be George Akerlof.

Akerlof won the nobel for his seminal work on information asymmetry, which resulted in the formulation of the adverse selection problem. We don't need to go into the details but suffice it to say that his work introduces the last of the most significant tenets of behavioural economics. Not only are people incapable of processing large amounts of information and prone to multiple cognitive biases, they often do not even have the information rquired to make the requisite optimization or satisficing, and this leads to multiple complications. When this aspect is introduced, even Kenneth Arrow could be counted as a seminal behaviourist. Indeed, his 'learning by doing' model was probably the first endogenous model of growth.

The most interesting thing is that Akerlof has been looking at a theory of 'behavioural macroeconomics' by re-interpreting he works of John Maynard Keynes. He has been among the foremost New Keynesians with his efficiency wages theory that explains how people who get jobs during good times lock in higher-than-equilibrium salaries, and people who look for work during busts have to continue with lower-than-equilibrium salaries even when the times are good. His formal exposition of this intuitive idea has led to considerable new developments on the Phillips curve, the emprically known trade-off between reducing unemployment and reducing inflation. He has also been collaborating with Robert Shiller towards this goal of a theory of behavioural macroeconomics, which can be loosely described as the current New Keynesian macromodel with microfoundations explicitly in behavioural economics. Currently, the microfoundations of the New Keynesian macromodel are largely neoclassical, with a few market imperfections.

From whatever little sense I have been able to make of economics, George Akerlof happens to be my favourite economist. Robert Shiller is another one to watch out for, and let me predict here that he may get the Nobel soon enough, in 2009 or 2010. Shiller correctly called the dot com and the real estate bubbles, but more on him later.


Robert Frust said...

Nice post. You're putting all the reading and wiki-ing to great use :).
You provided that missing spark that led me to a lot of interesting game theory and related stuff on wiki today. Experienced the simple and sublime pleasures of wikipedia after a long time :).

PS: I'm surprised only one out of 11 respondents opted for what must surely strike everyone as the more PC answer when posed the question.
When I talked about the framing being misleading, I meant that one tends to assume a shared context in conversation. We fill in gaps, ignore incorrect usage or grammar, and read between lines in conversation. In this case, the respondent will almost always assume that the choices are, like choices usually ares, mutually exclusive and will fill in what the mind assumes is left unsaid in the first. I don't believe it illustrates much by way of the rationality of agents, though it says something about the assumptions we make.

Ritwik said...


Wiki plus some other stuff. And not 11, that word was 'all'. Georgia is a confusing font, though I like the way it looks.

People aren't so PC in personal conversations!

While I ask his question in conversations, people have asked he same question in written forms and gotten the same result. The assumption that choices are necessarily mutually exclusive may be the cause of tihs mistake - it only serves to illustrate that we have a lot of assumptions and tat we mentally frame similar information in certain ways which are not in accordance to the way we should have framed that information had we been rational. (Mental accounting and the sunk costs fallacy in logic are similar concepts).

Just as we infer thing that may not exist in conversations, we do it with any unstructured information. We look at small portions of a highly non-linear graph and try to infer the rest of it. We tend to think that recent data is robust data. I agree with all your susbtantiave points, and yet think that they only illustrate the systemic irrationality that we have.

If you will allow me to add one more of my pet favourite fallacies common on this campus of ours, what do you think of the way people judge their salaries? Students exposed to recency effect, business cycle booms and busts, and the idea of smoothing out curves by using moving averages will still compare everything only to the latest year's data. It's very natural, and very irrational. :-)

Robert Frust said...

Let's talk about this when we meet. I'm not clear about your last point. It seems to me you believe people should look at macro-eco carnage and scale down their expectations. They will do that for certain. However, the trouble is that comparisons are made with one's peers and the top percentile of jobs on campus will still pay just the same as last year. The candidate will therefore be further away from the magic number. That said, perceptions of jobs and the hierarchy that most people seem to share are also interesting in themselves.
Also, in conversation, if I was not reading between lines, communication would be compromised. Mental accounting is, I agree, truly irrational.