Since writing on Frydman and Goldberg, I have had some email correspondence with one of the authors. I think I should do more to say where I agree with the authors.
1. I think that the rational expectations hypothesis in finance, much like the representative agent model in macroeconomics, incorrectly assumes that everyone has the same information set. It is tempting to presume, as do Cowen and Hanson, that rational individuals will converge to holding the same beliefs. However, I think that a more Hayekian view would allow for people to start out with and retain different models of the world. On this, I agree with Frydman and Goldberg.
2. I also agree that we observe asset price swings. There may be profit opportunities in these swings, although they are not arbitrage opportunities. One example seems to be the oil cycle. At the time that I wrote that post, it seemed highly predictable that the price of oil would rise. Eventually, I found an effective way to bet on that rise, via the fund PCRIX. Over the past month, I have sold out of that position, because I assume that at some point there will be another period where the price of oil overshoots on the down side. Of course, I could turn out to be wrong about that. Perhaps the price of oil will rise going forward, and I will never again have the opportunity to purchase shares in PCRIX at its current level. We will see.
3. Note, however, that there can be price swings when everyone has the same information. That is what Vernon Smith’s “bubble experiments” show (the link takes you to the web site of Mike Covel. As it happens, I interviewed him as a case study for my first book, Under the Radar, on entrepreneurship. Odd coincidence).
I conclude that people easily form models for asset price behavior that diverge from one another. An economist who has a view of what the fundamentals imply for an asset price is just one more individual with a unique model.
READER COMMENTS
fundamentalist
Mar 16 2011 at 4:51pm
De Soto “Money, Bank Credit and Economic Cycles” pages 461-462
fundamentalist
Mar 16 2011 at 5:03pm
But what are the “fundamentals”? In modern finance they are the NPV of expected future earnings. But as Benjamin Graham pointed out for decades, people who think they can forecast future earnings as well as future discount rates (both of which are required to calculate NPV) are only fooling themselves.
Pedro P Romero
Mar 16 2011 at 5:21pm
Arnold,
I really agree on your fist point about agents forming different models of asset pricing or the economy in general. I would like to add that I have seen many models of heterogenous agents, and by that is meant two or at most three different types of agents. within many of these models heterogeneity just implies different parameters within the same behavioral schemes. A more interesting approach that this post implies would be to make a model (or a theory) with agents having different, and even opposing, ways of interpreting the market wherein they interact. That is to say different rules of behavior not just parameters about the same phenomenon.
P
kebko
Mar 16 2011 at 5:32pm
I think these are all good points, including the other comments. In addition, I wish the finance literature would get away from the use of “rational”. It is much more useful to me to have a more post-modern perspective that every decision has some “rational” basis, and we need to put effort into figuring out what bases there are other than pure percentage cash returns. For most investors, and especially professional investing agents, there is real value in managing where gains are losses are taken relative to other investors, and at times these motives can make it perfectly reasonable for an investor to allocate capital in a way that the traditional models would call irrational.
An investor who doesn’t have these other motives can make predictable trading gains by trading against these other motives. This is no different than making higher returns as a result of having a longer time horizon or some other context which is widely accepted as producing higher returns.
Dave
Mar 16 2011 at 9:26pm
Along the lines of what Pedro P Romero wrote, agent-based modeling allows for the modeler to distribute various types of asset models/trading strategies to its agents.
Comments are closed.