Dynamic Provisioning: A Buffer Rather Than a Countercyclical Tool?

Abstract

The global financial storm which started in 2007 and which we are still experiencing is one of the very best examples in recent economic history of how much the financial system can exacerbate real economy cycles. Such pro-cyclicality has triggered a lively debate on which tools can be used to smooth this pattern, focused on macro-prudential policies. There was limited experience in the use of macro-prudential instruments before the crisis. The most prominent examples are the use of Loan-to-Value ratios (LTV) in some Asian countries and dynamic provisions in Spain (see Caruana, 2010). The latter received a lot of attention and in the early stages of the crisis were seen as a model for the then incipient international regulatory reform. But the debate shifted rapidly from provisions to capital and the reform crystallized soon in the adoption of a capital buffer in the context of Basel III, whereas the discussion on provisions languished and seems now in a deadlock. Two reasons explain this decreasing interest: the difficulties for accounting harmonization between the Americans and the Europeans, which partly explains why the Basel Committee took the easier route of capital; and the evidence, as the crisis deepened, that dynamic provisions did not prevent serious problems in certain segments of the Spanish banking system. The analysis of the Spanish case raises now more complex issues than a few years ago: why was dynamic provisioning insufficient? Was it a problem of design or application? Or were the bubble and the crisis too big to be addressed by this tool? Did dynamic provisioning have unintended consequences? In particular, did it delay the solution of the problems of saving banks? Was it a useful buffer, but not a genuine countercyclical tool? This paper focuses in the latter question, but to do so we need to address the previous questions to a certain extent, since they are very much interlinked. The question of whether dynamic provision was a buffer or a dampener is closely related inter alia to the rules versus discretion debate. Under a formula-driven system, the required level of provisions would vary according to some predetermined metric. It would provide a preset discipline independent on judgment. However, its success will depend crucially on the possibility of calibrating the business cycle ex ante, an issue to which we will return later. A rules-based system is superior to a discretionary mechanism in situations where the policy maker faces a problem of lack of credibility of its commitment. However, a rules-based system may face constraints that ultimately lead to discretionary adjustments. In particular, asymmetric market discipline (that fact that markets are too lenient in the good times and too strict in the bad times) may preclude the use of the accumulated buffer in the downturn, thus impeding the anti-cyclical compensation. We provide some evidence that this was a factor in the case of Spain: when liquidity dried-up and funding in the interbank market disappeared, markets required a higher level of own funds, limiting the anti-cyclical impact of dynamic provisions. To be fair, the sheer size of the crisis was also a factor limiting the compensation of rising NPLs, with the final effect that total provisions rose considerably in the bust, contrary to what was intended. Spain is not the only country that adopted dynamic provisions, but it was the pioneer and the only one for which there is experience on a boom and bust cycle. In order to obtain more general conclusions we do not limit ourselves, however, to the analysis of the Spanish case, but compare it to two Latin American countries that adopted dynamic provisions in the late 2000s: Peru and Colombia. The comparison with these countries allows extracting more general conclusions on the pros and cons of alternative designs, and allows also considering whether there are particular aspects that need to be taken into account in applying this tool to emerging market economies.