Believe those who are seeking the truth. Doubt those who find it. Andre Gide

Sunday, May 2, 2010

Liquidity and Debt

The empirical evidence shows that throughout history, money (broadly defined to constitute assets that circulate widely as media of exchange) takes the form of debt. In the past we saw banknotes redeemable in coin with senior claim to bank assets. Today we have government-insured demandable bank liabilities redeemable in government fiat. More broadly, it is notable that the collateral objects that "circulated" in the repo market were debt instruments (e.g., US treasuries and AAA rated MBS).

Why do we pay for things using bank debt rather than bank equity? To put things another way, why is debt more "liquid" than equity? What is the link between liquidity and the optimality of debt?

There is, of course, a literature that explains the optimality of debt (e.g., Townsend, Hart and Moore). The focus of this literature, however, is on the primary market and not on the secondary market; that is, they do not link the optimality of debt to liquidity. There is also a literature that studies liquidity provision, but simply assumes the optimality of debt (e.g., Diamond and Dybvig).

Some good news: there is now a nice paper that makes some progress in linking liquidity and debt. The paper is by Tri Vi Dang, Gary Gorton, and Bengt Holmstrom: Opacity and the Optimality of Debt for Liquidity Provision.

I am not an expert in corporate finance, but the theory of debt laid out in this paper seems somewhat different from, say, Townsend's costly state verification model. In Townsend, the security issuer has private information over the realization of the project. The security purchaser has an ability to discover this information at some cost. Costly audits are a social waste, so it is desirable to minimize this expense. Equity finance is terrible in this regard because the purchaser must audit all the time. Debt, on the other hand, triggers an audit only when the security issuer reports terrible outcomes.

Dang, Gorton and Holmstrom (henceforth DGH) have a setup where project realizations are publicly verifiable ex post. The twist in their setup is that prior to exchange, agents have an opportunity to discover the true project realization at some fixed cost. Consider two extremes: the audit cost is zero or infinity. Question: which world would agents rather live in? Answer: ex ante welfare is maximized under ignorance. The intuition for this is similar to Hirchleifer's (AER 1971) classic result; although agents in DGH are risk-neutral.

So their first result is that symmetric ignorance dominates symmetric information (even if such information is costless). Imagine that the security issuer and purchaser both face the same cost (k) of discovery. Is it possible to design a security that gives both agents the incentive not to acquire private information?

Let x denote the project outcome, distributed according to cdf F(x). Let s(x) denote the security's promised payout in state x, and let p denote the price of the security. Since the purchaser is risk-neutral, he is willing to purchase the security (without scrutiny) if p = E[s(x)].

The problem here is that the purchaser will view himself as having "overpaid" for the security in bad states of the world; i.e., for all x such that p - s(x) > 0. The expected cost of overpaying in this manner is given by v = E[max{p-s(x),0}]. DGH show that v is the value of information to the purchaser (they also show that the corresponding value for the seller is always smaller, so we may, without loss focus on the buyer). If v < k, then no party has an incentive to acquire information (preserving symmetric ignorance).

Note to reader: this program does not seem to handle mathematical formulae well. After several attempts at editing without the desired result, I will try to explain as much as possible in words only.

DGH demonstrate that because debt is senior, it reduces the probability that one will overpay (ex post) for the asset, hence it reduces the incentive to acquire information to make sure that one is not overpaying.

If v(debt) < k < v(equity), then the existence of leverage will support trades that would not otherwise have occurred. Leverage lifts the economy higher. But of course, when something is created, one automatically creates the risk that it might collapse. DGH have in mind the arrival of bad news in the form of a public signal (e.g., declining real estate prices). The effect of such a shock may be to render k < v(debt); which is to say, it suddenly pays to acquire information on debt. Such information acquisition, while privately optimal, is socially suboptimal -- it has the effect of generating asymmetric information and adverse selection -- which further hampers the liquidity of debt beyond the effect of the fundamental shock (the bad news). It is in this sense that a fundamental shock can lead to a systemic event.

If you are motivated to read the paper, please let me know what you think about it. I think that the authors are on to something; but I'm not absolutely sure whether the details all hang together. I'd also appreciate references to related literature.


  1. This comment has been removed by a blog administrator.

  2. Hi David,

    It is an interesting paper. I've scanned it to get a sense of the issues. I don't see the application to corporate finance, though. Corporate finance's fundamental motivation is asymmetric information, so mutual ignorance seems rather an impossible sort of circumstance.

  3. "So their first result is that symmetric ignorance dominates symmetric information (even if such information is costless)."

    I think its more an artefact of their model that symmetric information, when costless, has a negative value. I don't see it as central to their results; more of an accidental by-product of some assumptions they made to keep the model simple.

    Leaving that slightly peculiar result aside, their model would say that information has zero social value. And the intuition that information on debt has a lower private value than information on equity seems clear.

    I am reminded of a paper someone (sorry) gave decades ago. About auctions where you bid on the contents of a sealed container without being allowed to look inside. Or de Beers selling closed bags of diamonds that buyers weren't allowed to inspect. Since the buyers would be buying the diamonds anyway, and information was costly to collect, and only affected the price, not the quantities of trade, it had no social value.

  4. Prof J: Well let's think about it a bit. The MBS that was used as collateral in repo (until recently) seemed to serve its purpose well. Nobody really knew the underlying fundamentals of the paper, beyond the fact that it was rated AAA (and hence senior, though not risk free). That no one has the incentive to perform due diligence on a monetary asset is (arguably) a good property for any high velocity asset to have.

    Nick: you are right, the "news" about future events in their model has no social value. It may, however, have private value. To the extent this is true, agents may seek to acquire it. Wealth may be redistributed, but inefficiently.

    And thanks for the reminder about de Beers. I recall reading that article too (Yoram Barzel may have been the author...or perhaps Alchian...can't remember now either!). If I recall, the rationale for the practice is that it economizes on redundant measurement costs. De Beers promises to deliver a bag with a given distribution of diamonds; reputational concerns ensure that they do not exploit the buyers.

  5. David,

    Your point is quite correct vis-a-vis short-term financing. Since some banks were financing their long-term assets with short-term liabilities (repos), there is much less incentive to investigate the quality of the collateral given it is held overnight or a few days at most.

    I thought about it some more with regard to non-financial services corporations. This relates most strongly to commercial paper, and credit card receivables. Often the latter are packaged up into asset-backed securities (ABS). Not sure about the commercial paper... it's already a pretty standardized market.

  6. Nick, if information only affects the price of something but not the volume of trade this does not mean there is no (potential) social loss in general equilibrium. If a price of a good is affected this means its price relative to other goods is also affected and hence there is potential for mis-allocation.