Diamond-Dybvig (1983) and the Financial Crisis
| Peter Klein |
I started writing a really clever post about the famous Diamond paper (with Philip Dybvig) on financial intermediation and bank runs, its relevance for the financial crisis, and its elevated status in light of Monday’s Nobel announcement. Then I remembered that the author is Douglas Diamond, not Peter Diamond. Doh!
So I’ll try a different framing. “Speaking of guys named Diamond. . . .” The Diamond-Dybvig model, presented in a 1983 JPE article, has become famous enough to spawn an extensive secondary literature (and even sports its own Wikipedia entry). In a nutshell, it models fractional-reserve banks as intermediaries transforming illiquid assets into liquid liabilities and depicts the relationship among depositors as a coordination game with two Nash equilibria, one in which nobody tries to withdraw his funds because he believes no one else will try to withdraw his funds, and one in which everyone tries to withdraw their funds because they believe everyone else will try to withdraw their funds. Bank runs, in other words, constitute a Pareto-inferior Nash equilibrium. This framework led to extensive discussions about deposit insurance, option clauses, and other mechanisms to prevent the bad equilibrium by affecting depositors’ beliefs about solvency. (My former colleague Larry White devotes nearly a chapter of his Theory of Monetary Institutions to Diamond-Dybvig 1983.)
This is a hugely influential article, and I’m surprised it hasn’t been gotten more attention in the last two years. The essential fragility of a complex, interdependent, highly leveraged, fractional-reserve, implicitly government guaranteed system is at the heart of the financial crisis, so you’d think the Diamond-Dybvig framework would play an important role in the debate. But I can’t find much literature on this. The Richmond Fed devoted a special 2010 issue of its Financial Quarterly, guest edited by Ed Prescott, to the DD model, but it attracted little attention. Writes Prescott in his introduction:
Until recently, bank runs were not considered a major problem in the United States. The introduction of deposit insurance in the 1930s was considered to have essentially solved this problem. There had been very few bank runs since then. Much of the academic literature instead focused on the sizeable costs of moral hazard that can come with a deposit insurance system, as was seen in the savings and loan crisis of the 1980s (see, for example, White ).
What the academic and policy worlds missed was just how much some of the newer (since the 1970s) financial arrangements were starting to resemble banks in that they performed maturity transformation and financed assets with liabilities that resembled demand deposits. Many of these arrangements ran into trouble during the financial crisis when they could not roll over their short-term debt. Whether these episodes match the DD equilibrium in which a solvent bank is run because of a panic is still a topic of debate. After all, a run on a bank is also perfectly consistent with a bank being insolvent.
What we do have now, however, are data that are much higher quality than are available on historical runs. Furthermore, as we will see, these financial arrangements differ along dimensions such as how excess short-term withdrawals are managed. My conjecture is that these sources of variation along with the data will provide an important source of information for not only evaluating the DD model, but also evaluating methods for dealing with a potential run.
I’m not sure what Peter Diamond would say about all this — financial intermediation is a search process, after all — but Doug Diamond was written some important popular and scholarly pieces on the crisis. Another paper (with Raghu Rajan) echoes the Austrian theory of the business cycle: “[W]hen household needs for funds are high, interest rates will rise sharply, debtors will have to shut down illiquid projects, and in extremis, will face more damaging [bank] runs. Authorities may want to push down interest rates to maintain economic activity in the face of such illiquidity, but intervention may not always be feasible, and when feasible, could encourage banks to increase leverage or fund even more illiquid projects up front. This could make all parties worse off.” Add Diamond and Rajan:
We are certainly not the first to place the emphasis for contraction and crises on the mismatch between the long duration before investment produces consumption goods, and the temporal pattern of consumption in an expansion. This dates back at least to Von Mises (1949) and the Austrian School. Von Mises placed the emphasis, though, on an artificially low initial rate of interest, induced by bank credit expansion, which makes the process of creating new goods excessively long compared with the tolerance of consumers to postpone consumption. While it is difficult to map his theory precisely to a rational expectations general equilibrium model, it would appear that Von Mises (1949) places the blame for crises squarely on the heads of overly optimistic, excessively aggressive, bankers (and on central bankers who encourage aggressive credit expansion). . . .
Summarizing our analysis, a central bank that promises to cut interest rates conditional on stress, or that is biased towards low interest rates favoring entrepreneurs, will induce banks to promise higher payouts or take more illiquid projects. This in turn can make the illiquidity crisis more severe and require a greater degree of intervention, a view reminiscent of the Austrian theory of cycles. . . .
Our model suggests that the crisis of 2007-2009 may not be unrelated to the actions of the Federal Reserve earlier in the decade, not only in convincing the market that interest rates would remain low for a sustained period following the dot-com bust because of its fears of deflation, but also in promising to intervene to pick up the pieces in case of an asset price collapse — the so-called Greenspan put.