Jean Tirole stopped here in Chicago this week to present his new paper, an explanation of how mechanism design can aid government policy when credit markets freeze. Since I’ve got Jean at number 3 on my Fantasy Econ Nobel draft board (you don’t have one? Our group of old Fed colleagues goes eight rounds deep!), that’s not a presentation I could miss.
The model is simple. Let firms all hold an asset which pays xR, where R is a constant, and x is a variable in [0,1] representing the asset’s type. Financial institutions know the type of their own asset, but no one knows anyone else’s type. Let each firm want to invest in a new project with positive net present value. If the project is financed at cost I, it returns more than I if the seller “behaves”, but returns nothing but a private benefit to the seller if he “misbehaves”. As in a standard mechanism design problem, the seller will need to be a given a stake in the new project in order to be coerced into behaving. Though sellers own assets of varying quality, they all are proposing identical new projects.
The problem in the market at large is that some shift in beliefs about asset quality has caused the distribution of x to be such that, because of asymmetric information, no one is able to sell their asset, and because of this, no one is able to finance their new project. This occurs even though people with high quality assets would be able the finance the new project should the market know they had high quality assets. But assume that the government can take some action first, then allow the market for assets to open, with every seller deciding whether to participate in the government mechanism or to wait for the market. Is there any welfare-enhancing way to “unfreeze” the market? Note the two problems, from a theoretical standpoint. First, the participation constraint in the government mechanism is endogenous, since market outcomes depend on who joins the government mechanism. Second, all of the standard informational issues from a signaling/adverse selection problem are present.
Tirole shows the optimal government policy is as follows. First, the government buys the assets (or some share of the assets) only of people with low quality assets; of course, government cannot see asset quality, but they are able to make such purchases by offering a sufficiently low price. Second, sellers with “medium-quality” assets only have a share of their assets bought by the government. Third, the government purchases must be of a sufficiently large scale in order to “unfreeze” the market. Fourth, the government will always lose money when they buy assets, even though sellers are desperate to sell assets for less than their worth in order to finance the new positive net present value project.
Why? Buying bad assets instead of good is optimal because it leaves only good assets to the market, which will then be able to make money by financing the remaining sellers. Only a portion of low quality assets are bought in order to keep as many of the assets sold in the market as possible. The government must buy a sufficiently large number of bad assets in order for the market to earn at least I, in expectation, from the assets they buy. And the government always loses money because, even though the sellers are eager to sell their old assets, they were also eager to sell even before the government intervention. The government thus is paying more than what the sellers could have gotten in the market without intervention, and since the market after intervention is competitive (has a zero-profit condition), the government will be the one losing money. This does not suggest intervention should not happen: the government can still lose money even when intervention is socially optimal.
Finally, Tirole points out that this policy is optimal given a credit freeze because of adverse selection (e.g., a shock to the return of assets of varying quality currently on financial institution balance sheets). Of course, alternative policies (such as government-provided liquidity) might be better if the reason for credit market seize-up is different. In any case: mechanism design is a quality tool.
A Fine Theorem 