As a consequence of the global crisis, there are worries that many countries will slide into a Japanese-style decade of lost growth. One of the main problems in Japan during the 1990s was that insolvent "zombie" banks decided it was a good idea to persist in lending to insolvent "zombie" firms (Peek and Rosengren 2005), a practice sometimes referred to as forbearance lending. But why would banks want to engage in such lending?
Although cultural factors may have played a role in Japan, insolvent banks in general have incentives to delay the realisation of losses and hide their problems in order to hope for salvation in the form of an economic recovery that improves the situation of their currently insolvent borrowers as a form of gambling for resurrection. And one way to delay the realisation of losses and to hide problems is to roll over loans of insolvent borrowers, instead of foreclosing.
The pain in Spain
A specific example of a country in which such worries seem warranted right now is the country in which we work, Spain. In a country with unemployment currently at 20% and in which a huge real estate bubble has just burst, reported non-performing loan ratios have not actually increased by much, and lending to real estate developers has actually increased. Forbearance lending seems to be a plausible explanation.
The misallocation of credit towards the zombies rather than healthy firms that ensues from forbearance lending has catastrophic effects on overall economic growth (Caballero et al. 2008). For policymakers, then, the prevention of forbearance lending may be an important goal.
What to do about forbearance lending?
Gambling for resurrection and similar distortions such as debt overhang, and possible solutions, have been discussed in the corporate finance literature since at least the 1970s. These issues arise when entities with limited liability are close to insolvency. Solutions revolve around restoring solvency, and include many implicit or explicit forms of recapitalisation.
A basic practical problem that regulators face, though, is that the extent of solvency problems at a given bank is likely to be known by the bank, but not by the regulator. The regulator might know that in the current crisis, all banks have taken a hit, but looking at a given bank, the regulator cannot tell whether it is part of the "walking wounded" and still solvent, or part of the "living dead" and insolvent. In recapitalisation schemes, this private information is likely to generate windfall gains for some bank equity holders, that is, information rents. Such rents are very undesirable – they constitute a reward to banks that have taken unnecessary risks, they may distort ex-ante incentives, and they are socially costly because of the taxation necessary to finance them.
There is a small, developing literature using ideas from mechanism design that looks at how schemes can structured in order to reduce or eliminate such rents in bank bailouts (e.g. Philippon and Schnabl 2009, Bhattacharya and Nyborg 2010). In a recent working paper (Bruche and Llobet 2010), we contribute to this literature by reconsidering the idea of asset buybacks. We show that it is possible to structure these in such a way that completely eliminates information rents that can arise from banks being the only party that knows what fraction of their loans have gone bad.
Many people think that out of all of the ways to implicitly recapitalise a bank, offering banks the chance to sell its bad loans or toxic assets in an asset buyback is one of the worst.
If a bank wants to roll over bad loans as a form of gambling for resurrection, the value that it attaches to them is above the fundamental value obtained when such loans are foreclosed. So overpaying for the bad loans in an asset buyback is unavoidable. The more bad loans a bank has, the more insolvent it is, and the higher the price one would have to pay.
Why do we think, then, that asset buybacks can be a good idea?
- First, the willingness of bank to part with bad loans under an asset buyback can reveal private information on the quantity of bad loans on that bank's balance sheet.
- Second, price discrimination can sometimes be used to obtain the bad loans and, at the same time, to claw back some of the increase in the equity value that the purchase generates. For example, one could try a two-part tariff: Suppose you offer banks a choice to either pay a low participation fee, and then sell bad loans at a relatively low price, or to pay a high participation fee, and then sell loans at a relatively high price. Banks with many bad loans to get rid of would choose to pay the high fee in order to get the higher price, and banks with few bad loans would accept the lower price, so as to pay only the lower fee.
To what extent can a careful design of the menu of prices and fees reduce information rents? The surprising conclusion here is that it is always possible to completely eliminate them. This is possible because one can play off two opposing incentives of a bank to lie about the quantity of bad loans on its balance sheet. On the one hand, banks with a larger quantity of bad loans benefit more from the scheme since they receive a payment for each of these loans. A regulator could charge such banks a higher participation fee without discouraging them from participating. Banks therefore have an incentive to understate their quantity of bad loans, in order to be charged a lower participation fee. On the other hand, banks with a larger quantity of bad loans are more insolvent and have stronger incentives to gamble, and weaker incentives to participate. A regulator would have to charge such banks lower participation fees so as not to discourage them from participating. Banks therefore also have an incentive to overstate their quantity of bad loans, in order to be charged lower participation fees.
In our setup, it is possible to design a menu of prices and fees that exactly balances both incentives. This means that the regulator can get banks to truthfully report the quantity of bad loans on their balance sheet, without having to bribe them with information rents. But how general is this? We show that the properties of the model that make this exact balancing possible are the very same properties that lead to the gambling behaviour in the first place, namely, limited liability and the fact that rolling over bad loans delays the resolution of uncertainty. This suggests that in models in which banks have incentives to engage in forbearance lending as a gamble for resurrection, the type of mechanism that we consider would work.
Would this work in practice?
While sometimes reality is less neat than a model, we believe that the ideas here could at the very least be used to improve on schemes that are already used in practice. For example, in Ireland, a scheme was set up under the umbrella of the National Asset Management Agency. Under this scheme, banks receive a price for each bad loan that they sell to a special purpose vehicle. As argued above, banks are very likely to benefit from this. In fact, the Irish scheme has been heavily criticised, precisely because it is likely to generate very large windfall gains for some banks. Prominent critics include Joseph Stiglitz, who thinks the Irish scheme is something more fitting of a "banana republic". Our question would be: Why not make banks pay for the privilege of participating in the scheme, in the form of participation fees? Even small fees will reduce the size of the windfall gains.
Bhattacharya, S, and KG Nyborg (2010), “Bank Bailout Menus”, Swiss Finance Institute Working Paper 10-24.
Bruche, M and G Llobet (2010), “Walking Wounded or Living Dead? Making Banks Foreclose Bad Loans”, Working Paper.
Caballero, RJ, T Hoshi, and AK Kayshap (2008), “Zombie Lending and Depressed Restructuring in Japan”, American Economic Review, 98:1943-1977.
Philippon, T and P Schnabl (2009), “Efficient Recapitalization,” NYU Working Paper FIN-09-014.
Peek, J and ES Rosengren (2005), “Unnatural Selection: Perverse Incentives and the Misallocation of Credit in Japan”, American Economic Review, 95:1144-1166.