Evan Soltas
Jul 9, 2012

A Theory of Efficient Bubbles?

Cookies are not baked one at a time. Bakers make them in batches. They do so even though it is well known that cookies go stale; there are depreciation costs to deferred consumption. And they do so because the fixed costs of cookie production, such as prep time, are high enough that the point of minimum average total cost in the short run, which determines the level of economic profit, occurs somewhere near a dozen cookies.

Therefore a time series of an efficient daily production of cookies by a household baker may look something like this: 0 0 0 0 0 12 0 0 0 0 12 0 0 0. Since cookie consumption is smoothed, it follows that the accumulation of cookie inventories can also be efficient, despite depreciation of cookie assets.

This story may seem like an absurd use of economic terminology on cookie production, a process almost everyone understands. Yet I think it captures some of the essential facts of why bubbles may be not only rational, but also efficient. Theories of rational speculative bubbles may explain why bubbles happen; here I try to imagine may the conditions under which bubbles could be efficient.

Note that I do not seek to imply that all bubbles are efficient -- just that there exists a plausible set of conditions for which a large, temporary, inventory-creating, price-driven increase in production could be an optimal outcome for a problem of allocating scarce resources. My analysis here focuses almost exclusively on the real side of bubbles, and by extension those bubbles which have a real side -- that is, those bubbles which occur in industries where supply is highly inelastic, such as farmland, are generally  incompatible with the conditions which make bubbles efficient. Given this focus on the behavior of production during a bubble, rather than the asset price -- which has been more extensively studied -- I do not consider the existence or magnitude of costs which come directly from the movements in asset prices, nor the degree to which they would offset the benefits on the real side. In general, a theory of efficient bubbles must look at the movements in asset prices as a mechanism for encouraging the clustering of production over spans of time.

To return to the cookie-production analogy, cookie production is efficiently clustered because of high fixed costs and a long stretch of low marginal costs, such that the average total cost function is falling for a large range of quantities moving from zero to N cookies. That is the primary condition for an efficient bubble. Under such conditions, if microeconomic demand for cookies can be "pooled" into a brief period of time -- such as would be generated by a speculative price bubble -- then there could be a cost savings because of a more efficient scale of production, even though it is one which cannot be sustained for the long-term given the lower long-term level of demand.

Without the implication that the housing bubble was efficient, there is evidence that the speculative change in price does induce such a "demand pooling" situation in which production is therefore clustered.In this graph, the level of new completed homes on the market seems to be propelled forward by the increase in prices. There is an apparent lag, but looking more carefully, it appears that the acceleration of the real variable is determined by the level of the nominal variable -- most conspicuously, note that the real 10-city Case-Shiller index of housing prices and the acceleration of new completed homes entering the market had their maximums at exactly the same moment. Prices acting as a physical "force" acting on new completed homes is consistent with the efficient-bubbles theory of prices acting as the demand- and production-pooling mechanism.

The second condition which would make a bubble efficient is the presence of positive externalities such that total returns for a society are often higher then the private return. If private individuals can be "convinced" by an asset-price bubble to make those marginal investments which eventually have a negative private return but a positive total return, then a bubble may be an efficient outcome for a society.

I think I actually got this idea from, out of all places, Thomas Friedman's The World Is Flat:

Overinvestment is not necessarily a bad thing--provided it is eventually corrected. I'll always remember a news conference that Microsoft chairman Bill Gates held at the 1999 World Economic Forum in Davos, at the height of the tech bubble. Over and over again, Gates was bombarded by reporters with versions of the question, "Mr. Gates, these Internet stocks, they're a bubble, right?" Finally an exasperated Gates said to the reporters something to the effect of, "Look, you bozos, of course they're a bubble, but you're all missing the point. This bubble is attracting so much new capital to the Internet industry, it is going to drive innovation faster and faster."
As Friedman proceeds to describe, what happened was that investments in fiber-optic cables proved far less privately lucrative than was hoped, leading to the bust of the speculative boom for these Internet companies. But the overinvestment generated a publicly-usable infrastructure -- our 21st-century communications grid -- which conferred very large positive externalities in excess of the private losses. Given that laying fiber-optic cable fits the cost structure I described as the first condition, it is possible that this aspect of the tech bubble may have been efficient.

(Note: I first put forward my efficient-bubbles theory in a comment here to a post by Scott Sumner.)