Saturday, 28 March 2009

A behavioural theory of money

If you have clicked onto this article just from reading the title, then I may disappoint you slightly. I am not (yet) going to propound a behavioural theory of money, though I think there's one coming in the future.

But I will point to a couple of results which may indicate where to get one. The first is a quote from John Moore and Nobu Kiyotaki of LSE, in the beautifully titled lecture Evil Is The Root Of All Money.
"Money is the medium of exchange.

Notice that for this argument to hold together, there has to be a set of mutually-sustaining beliefs, stretching off to infinity. I was willing to hold money yesterday because I believed the dentist would accept it today. She is willing to hold money today because she believes someone else will accept it tomorrow. And so on. If there were a known end-point to history, the entire structure of beliefs would collapse back from the end."
This, as they point out, is the conventional view among microeconomists about the existence of money. However, experimental results from Vernon Smith (see references 11 and 15 on this paper, though I'll try and update later with a proper reference that's available on the web) show that it might not be true.

A laboratory experiment was set up with a number of participants. A bond-like instrument was invented, with predefined returns over a known finite period. For example, a dividend to be paid every 5 minutes over the course of an hour, with a final payment at the end of the hour. Participants were then issued with a number of shares in this instrument, and asked to trade them with each other over the course of the hour.

The participants used computers to trade with each other, and (importantly) the computers showed them the precise discounted return of the instrument at all times. Thus, this bond had a known, rational valuation at all times and all traders knew what it was. In theory, a perfectly efficient market with no external events, and thus the price should never vary from the theoretical return: it should drop by a predictable amount each time a dividend is paid, and at the end of the hour it should trade at face value and then mature. There should be no way to make money from trading it.

Despite this, the bond developed a price dynamic which was independent of this rational level. Specifically, the traders bid up the price of the instrument during the middle of the hour, and it declined back towards its rational value at the end. Thus, the optimal strategy was to buy early, sell in the middle, and (if possible) short the stock as the price declined towards the end.

This experiment clearly shows that even with a finite period and fully predictable returns, stocks still exhibit bubble behaviour.

And the monetary corollary? Money could be worth something even if history were to have a known end-point. And this indicates that there's something missing in neoclassical monetary theory. Part of the value of money (and other assets) is given by a phenomenon of self-justifying expectations about other people, which cannot be expressed in a rational model.

The Kiyotaki and Moore lecture is very well-written, very elegant and unfortunately misses a crushingly fundamental point. People do not make a rational decision to trust money. It's much more subtle than that.

Thanks to this article by Nick Rowe and numerous interesting comments on it for indirectly provoking my thoughts on this point.

No comments: