Heuristics! In academia this is the study of natural patterns in any given knowledge domain. For example: say you want to organize your pictures but you have too many pictures and you can't really decide how to organize them to begin with. What do you do?
Simple, create a program that can look at how the different pixels in a photograph are arranged and look for patterns. The way you do that is to have a pre-existing database of patterns with human understandable names such as: person, flower, building etc. You can even organize information within each category so that the program first looks for a pattern that fits a category then looks for a particular pattern within the category that would represent, say, the Eiffel Tower or the Chrysler Building. Thus, auto-organizing your photo album for you. Say you wanted to look up "when you were in europe?" in your virtual album then you could type in things from your europe like the Eiffel Tower or Big Ben and find pictures accordingly. You can even search for particular people within your photographs and within your web albums. That way all those pictures of Nana can easily be recalled or accessed from your various albums.
Obviously there are AI implications as well as model building implications within this idea. The coolest of which are its applications to economics or AI(singularity). In economics this is causing a massive struggle between classical economic theorists and new behavioral model theorists. According to the classical theory of economics the simple idea of supply and demand creates the invisible hand of the market which then can be used to predict the outcome of various economic scenarios. Of course this model is good during the middle periods but towards the end of each boom and bust cycle, the model falls apart except to say that capitalism is a process of creative destruction. Of course there are many casualties along the way which have real world consequences. Behavioral economics wants to turn the traditional wisdom upside down by using heuristic or pattern searching to create complicated models of the market. Then use these models to create databases of common patterns that can be used to detect a coming crisis and help prevent it. Sort of like what your antivirus software tries to do.
One idea that has emerged from the recent crisis is that regulations for banks in regards to their leverage and capital requirements should be counter-cyclical and that maybe the Fed should go beyond just setting interest rates to actually looking for problems down the road so that it can take specific measures to thwart bubbles in different areas of the economy:(stocks, bonds, securities, commodities). Of course this would represent a huge challenge for the Fed and is a step away from dire theories of anti-capitalism such as socialism or even worse communism. But proponents argue that this is the only way of preventing the large booms and busts that have occurred in the past decades as well as the current one. The Fed has been skeptical about the entire idea because it thinks that classical economic theory forbids the government from intervening in the markets actively to burst bubbles. The end result as predicted by classical economic theory is that the problem would be made much worse. Further research has now indicated that this is usually true throughout history precisely because government agencies are bad at predicting what the market is doing or even going to do in the years to come.
Behavioral Economics turns this idea upside down by challenging the classical understanding of economic theory. They think that classical economics only models human behavior in the most general terms such as supply/demand or monetary theory. This, they argue, ,is not enough. They think that new modeling techniques using heuristics can help create models that are so detailed that they can then be compared to actual economic data and be used then to make specific predictions of future market events. The idea here is that while most market cycles have a predictable pattern of booms and busts this idea is too simple and doesn't really do anything because it cannot be used to accurately predict future behavior. For example take the current crisis: While a lot of people where warning about various issues with the market since 2002-03 up until 2007-08 when the sub-prime crisis finally struck, nobody seemed to know exactly what would happen and when. Moreover, nobody predicted the depth and breadth of the crisis despite the number of people issuing warnings and now stepping in to do retrospective analysis. This is the pattern that behavioral economists have noticed for each crisis, a clear lack of specific predictions. One way to look at it would be to follow the money, most of those predicting the crisis could happen also do not take mitigating actions to protect their investments, something that would be a rational decision if anyone really predicted that something would happen.
Their idea is to look for patterns that would explain how individual market actors behave and then create models that would predict different scenarios. Then use these scenarios to look for patterns in real economic data to make specific sets of predictions each taking into account different events or series of events and thus determine places where the Government could step in and take effective action in order to avert market turmoil. For example imagine that the Fed had used this technique and had special powers at its disposal to take action. It could have used heruistics to determine in 2005-6 that bad lending was going on because the prices of securitised debt instruments did not accurately reflect the risk being taken as their models would predict. That would have led them to look for an underlying cause of the irrational behavior and come up with a set of prescriptions to cure that behavior and fix the system before the crash. In this case stepping in during the 05-06 years to intervene with rating agencies that were not doing a good job of rating the debt instruments or looking for a way to reduce the inflow of excess cash by raising capital requirements for banks and thus reduce the amount of risk the banks could take.
Classical economists counter-argue that instead of using complex models that depend upon untested science and in the end the competency of government agencies, the better idea would be to create a set of rules that would be set into motion according to preset limits in the markets. This would, they argue, encourage predictability and thus ensure that the Fed wouldn't seem to act arbitrarily, increasing market volatility instead of decreasing it. An example of this would be to force the Fed to have counter cyclical capital requirements for banks tied to GDP growth or tied to spikes in the volume of lending. Behavioral economists think that this would be a blunt instrument that would step in to lower economic growth and thus economic potential. They prefer to give the Fed agility so that it would not be forced to do anything. They say that it would be hard to figure out when to set the pre-set limits and that each economic growth cycle might require different pre-set limits which would complicate the predictability that is so highly prized by the Classical economists.
One interesting way to grasp the different arguments of Classical vs. Behavioral economists is to look at the betting that goes on in futures markets, such as the one for volatility that Classical Economists love but which the Behavioral economists view with disdain because they suggest that these index's only tell you contemporary information about what the market thinks the future will be like. They suggest that this is simply a filter of information rather than an actual prediction engine. Something that they say can only be created using behavioral economics and its model market simulations.
In the end the winner of this debate will most likely lay down the rules for the rest of us. You will hear about this again, I guarantee it!
Comments are closed on this story.