There is nothing in finance that gets people's blood going like thinking about bubbles and crashes. We don't really have a precise definition, but we know them when we see them, and we are definitely in a crash right now of at least once every 50 year proportions. Agent-based financial market simulations seem to easily generate these market instabilities across a wide variety of model specifications. Market instability in the world of adaptive agents seems to be almost generic. I think this ease of generating bubbles in the machine has important implications for the real world.
Financial instability is highlighted by price crashes because they stand out in everyone's mind as the most visible part of instability. However, the large, and very persistent swings in deviations from fundamental valuation are probably the part of picture which is most costly in economic terms. As valuations get out of line and stay there, bad decisions made by consumers and firms are made. This is pretty obvious for large overvaluations in assets, but it can be equally problematic for undervaluation as well.
The dynamics of trading strategies that generate overreaction and bubbles is very simple. Positive feedback strategies which increase their position in a risky asset when the price rises will clearly generate a bubble. These strategies alone are not stable, and they can be tamed when combined with strategies based on a perceived return to some fundamental value. The dynamics of various representations of these strategies has actually been studied extensively. The interesting, and critical question is whether destabilizing combinations of strategies can survive in an evolutionary environment where they are competing with other strategies. The answer to this question, from computational models, appears to be yes. Markets which converge to strict rational expectations equilibria are usually the exception rather than the rule.
If bubbles are caused by misperceptions, are all mistakes created equal, and should they all be treated in the same way? Bubbles can be due to mistakes in estimating expected returns, and also in estimating risk. Either mistake can have a long run impact on returns and policy makers. I worry that too much emphasis is often placed on mistakes made on returns, and not those made on perceived risk. Also, I think in most cases irrational return expectations are often endogenous to the traders and would therefore be difficult to mitigate. However, misperceptions about risk can often be part of a financial environment which is very much in the realm of improved policy making. The current financial situation is a perfect example of this where various securities were sold as being much lower risk than they actually were. Investors were fooled into thinking that complex AAA rated debt instruments really shared the risk characteristics that would go with a AAA rating. When proved wrong, their entire picture of the world changed. In a learning sense they were devastated at seeing a random event that was outside what they were expecting, and forced them into a very cautious position since no internal model of the world could be trusted. Compare this to an investor holding a dot com stock in the late 1990's. Obviously, the investment came with over optimistic return forecasts, but most investors still maintained some reasonable perspective on risk. When one of these stocks crashed, they lost lots of money, but the low return draw might still be in their perspective of the world, and they would not completely reject all models that were being entertained for how the world worked. I like to think about these two types of bubbles as good (return) and bad (risk) bubbles. The former are not particularly good, but they are generally less disruptive, and few policies can be implemented to prevent them. The latter are more costly in terms of economic upheavals, and also may have bigger policy implications. Policy makers need to concentrate on preventing bad bubbles.
If bubble like behavior is difficult to remove from most financial markets, then what should policies be designed to do? The objective for eradicating bubbles should be off the table, since this will not succeed and might even be counter productive. Policy makers should look for potential securities and institutions that confuse investors about risk. It seems likely that this will often be concentrated in debt instruments which are infrequently traded, and where there is no transparent market mechanism which allows them to reveal their true risk. Beyond this simple recommendation some progress might also be made to try to change the distribution of bubbles and crashes. Policies which try to shift these from a few very large crashes to smaller more manageable crashes might be a possibility. Complex systems analysis has an example which has some curious connections to financial markets, forest fires. It turns out that attempting to put out all forest fires can be a problem in that it allows for the build up of low brush which is often destroyed by small fires. This brush serves as a form of gasoline, and eventually causes small fires to quickly turn into out of control large fires. In financial markets attempting to eliminate all instability may push investors assessments of risk to unreasonably low levels. When a market is full of investors no longer cognizant of the risk they are bearing, it is probably poised for a large and persistent crash. The tricky policy is to make sure that all security prices do a good job of revealing not just valuations, but also the risk behind the various instruments. Only in such a situation will effective learning allow markets to function well.