Tomorrow I'm going to go the American Economic Association's annual meeting, to interview for economics jobs. (For the uninitiated, most jobs that are specifically for new econ PhDs conduct interviews at this meeting, which is what most economists mean when they talk about "the job market.") One of the key elements of getting a job as an economist is to have a "job market paper," which is generally the first chapter of your dissertation, and which is what shows potential employers that you can do quality research. My job market paper is here in case you want to read it. I'm not going to spend a heck of a lot of time discussing the paper (that can wait until it's ready to send off to journals), but it gives me an opportunity to write a post I've been wanting to write for a while anyway, about the research that motivated my paper in the first place.
That research is the Vernon Smith "bubble experiments." "Bubbles, Crashes, and Endogenous Expectations in Experimental Spot Asset Markets" (Smith, Suchanek and Williams, Econometrica 1988) is definitely a Paper You Should Know.
What is a "bubble"? Well, it's something that looks like this:
Prices go way up, then they crash back down. Look at any long-term plot of any asset price index (stocks, housing, etc.) and you're likely to see some big peaks like this. That's what I call a "bubble." It's also the definition used by Charles Kindleberger in his book Manias, Panics, and Crashes.
But the real question is why we care about bubbles. Some people believe that bubbles are merely responses to changes in expected fundamental value of an asset (the "fundamental value" is the expected present value of the income you get from owning an asset). According to this view, the NASDAQ bubble happened because people thought that internet companies were going to make lots and lots of profit, and when those expectations didn't materialize, prices went down again. This view is held by many eminent financial economists, including Eugene Fama, the most cited financial economist in the world.
If bubbles represent the best available estimate of fundamental values, then they aren't something we should try to stop. But many other people think that bubbles are something more sinister - large-scale departures of prices from the best available estimate of fundamentals. If bubbles really represent market inefficiencies on a vast scale, then there's a chance we could prevent or halt them, either through better design of financial markets, or by direct government intervention. For a discussion on theoretical reasons to believe or disbelieve that bubbles = departures from market efficiency, see this email exchange between Eugene Fama and Ivo Welch. Very good stuff.
But in the end, empirical reality has to have the last word. And here's the problem we run into when we try to answer the question by looking at the data: It's nearly impossible to know what fundamentals really are. So in 1988, Vernon Smith (now a Nobel laureate) had the idea of using lab experiments to study bubbles. In a lab experiment, the experimenter knows what the fundamental value is, so whether the price exceeds the fundamental can be known with a high degree of confidence.
So Vernon Smith and his co-authors put groups of subjects in a lab, gave them some cash, and let them buy and sell a computer-generated financial asset. The asset payed dividends to whoever held it, so its fundamental value was equal to the expected value of the dividends it paid (or less, since people might be risk-averse). So not only did the experimenters know the fundamental value of the asset, but the subjects themselves were told everything they needed to know to calculate the fundamental!
And guess what happened? Prices rose way above the fundamental value, and then crashed at the end of the market. Check out this graph of the prices from one of their experimental markets:
The black stair-step line is the fundamental value (it goes down, since the market has a fixed finite lifetime). The dots are prices at which subjects exchanged the asset. As you can see, prices went way above the expected value of the asset in periods 3 and 4, then crashed back down somewhere around periods 9 and 10.
You are looking at a bubble that is also a market inefficiency. It is real. It exists.
Now if we knew that the market in this lab experiment was the same as real-world financial markets, this graph would spell death for the Efficient Markets Hypothesis. Done. Kablooie.
But of course, we do not know that this market is the same as a real-world market. This market is 8 inexperienced college kids in a lab, trading about $200 worth of an asset that is very different from real-world assets, over a period of maybe an hour. Real-world markets are made up of thousands or millions of experienced traders trading trillions of dollars worth of assets, with plenty of time to make their decisions. In other words, the Smith, Suchanek, and Williams experiment does not instantly explode all of efficient-market financial economics, because it may lack external validity.
But if it does have external validity, it's the most important empirical result in all of financial economic history.
A lot of work has gone into figuring out whether the Smith-Suchanek-Williams result has external validity. On one hand, it's reliably true that when the same group of traders - or even part of the group - repeats the market 3 times in a row, there's no bubble. On the other hand, if you take those "experienced" traders and run the experiment with a different fundamental value process, the bubbles come back.
The Smith-Suchanek-Williams type experiment has been run with professional traders. It has been run with large numbers of traders. It has been run under various different market institutions - allowing short selling, using different kinds of markets, under various rules, etc. Always, the bubbles come back. As the list of "things that don't prevent lab bubbles" grows, so does the nagging feeling that this type of inefficient bubble may be a universal phenomenon.
But there is one critique of the Smith experiment that has never really been put to rest. What if subjects just fail to understand the fundamentals? Maybe subjects simply learn by doing. Maybe you can tell them the fundamentals, but they don't really understand and believe in the fundamentals until they've had some experience with seeing the dividends roll in. In fact, there is research to support this critique. For example, when you let the subjects watch the dividend process before they trade, or when you explain the fundamental value in a different way, the bubbles disappear.
Nobody is really interested in a bubble that only happens because subjects are confused. It's easy to confuse a bunch of college kids. So this critique has limited the degree to which the "bubble experiment" literature has turned the financial econ world upside down.
That's where my job market paper comes in, actually. My experiment was about determining demand for assets in experimental markets - i.e., figuring out why people buy assets, not just whether markets lead to bubbles. Instead of the normal market setup, where all the traders trade with each other, I had traders who simply watched the prices that were produced by such a market, and then let them trade at those prices without affecting the price (think of this as one guy sitting at home buying and selling on E*trade; he's too small to move the market, but we can learn a lot from watching how he makes his decisions).
And one of the things I did was to ask my subjects how much dividend income they thought they would get from holding a share of the asset (i.e., the fundamental value). So I was able to tell who understood the fundamental value and who didn't. So I could tell if people who understood the fundamentals would knowingly overpay for the asset.
And guess what? They did. Even people who understood exactly how much the asset was worth were overwhelmingly likely to pay, say, 60 yen for a share of the asset (the experiment was run in Japan), even when the most dividend income they could possibly get for it was only 40 yen. They gave up certain gains. (Why on Earth did they do this? Well, that's what the rest of the paper is about, and is a subject for at least two other long blog posts.) So this result really answers one critique of the Smith bubble experiments: understanding what's going on is not sufficient to make subjects act like the "rational arbitrageurs" in financial economics models.
Now, that's only one critique of bubble experiments. It's impossible to answer them all. But science is a process of excluding causes one by one. The more uninteresting causes of bubbles we exclude, the closer we get to finding whatever interesting causes may exist.
To sum up: If we believed that Vernon Smith type bubble experiments could reliably show us how real financial markets work, the finance industry would be falling over itself to do a billion lab experiments, and both business and government would have to seriously rethink policies that depend on markets being efficient (e.g., stock options for executives). But the jury is still out. Still, the bubble result of Smith, Suchanek, and Williams (1988) is one of the most important experimental findings in the history of economics, and a research effort to which I am happy to (attempt to) contribute.
No comments:
Post a Comment