I first became aware of this issue when I participated in a process improvement project to review preparation of the quarterly loan loss reserve at a major bank. This project was intended to evaluate the process of estimating the loan loss reserve from multiple angles and recommend ways to make it more efficient, more reliable or both. In this case, however, we spent months debating the very definition of what the loan loss reserve was supposed to be. In fact, we never did resolve the question, we simply took the nature of what we currently did as reflecting some normative indication of the appropriate definition.
Different agencies, different agendas
In the US, besides banks themselves, there are several entities with a role in this debate. These include the Financial Accounting Standards Board, the Securities and Exchange Commission and three different banking regulators.1 The problem is that these entities demand very different standards and banks are left trying to harmonise the irreconcilable. Industry frustration over this issue was expressed very clearly five years ago by America’s Community Bankers, a trade association. In congressional testimony2 they said: “All the parties involved in the controversy, including the banks that have been caught in the middle, have stated that GAAP should be followed in the calculation of loan loss reserves. However, GAAP is not always easy to follow.”
Formal application of Generally Accepted Accounting Principles (GAAP) forbids inclusion in the loan loss reserves of losses that are “more likely than not”, “reasonably possible” or “likely” to occur in future periods. Under GAAP, loan loss reserves are only permitted for “probable losses” and for losses currently “inherent” in the portfolio (such as anticipated future charge-offs based on current repayment behaviour). Some argue that GAAP calls for booking the “most likely” amount of loss on a loan, which certainly sounds like the mode of the loss distribution. This leaves considerable room for judgement, but fundamentally it is a very restrictive definition. In all but extreme cases, the most likely loss on a loan tends to be zero. Only when default looms as a clearly identifiable near-term prospect is some non-zero loss statistically more likely than full repayment.
I firmly believe that a strict application of these rules would result in significantly smaller loan loss reserves than currently exist in the banking system. The reason this does not occur is largely pressure from banking supervisors who want to minimise claims on deposit insurance funds. For that purpose, higher reserves are clearly better, since they act as a buffer to absorb losses without affecting reported capital and earnings.
One harmful consequence of this lack of clear definition is that it opens considerable leeway for managing earnings. Banks can always appeal to the authority of convenience in justifying any quarter’s decision on setting the reserve. More pressing, however, are the problems created by lack of international consensus on the issue. All modern internal systems for capital allocation and risk-adjusted return calculation focus on unexpected losses. Expected losses are priced into the normal terms and conditions of doing business and are absorbed out of current revenues.
At the eleventh hour, this has become an issue in finalising the Basel II capital rules. This is because the consultative papers to date base regulatory capital on expected plus unexpected losses. Banks that are heavily into consumer lending argue, quite correctly, that this places them at an unfair disadvantage because their large but highly predictable losses should imply a fairly modest regulatory capital assessment. One possible, although perhaps not ideal, solution to this would be to permit all loan loss reserves to be included in the regulatory definition of capital. Unfortunately, inconsistent reserve conventions across countries present an obstacle to this approach.
I would make the following modest proposal. Loan loss reserves should be explicitly recognised by all authorities as a portfolio-based concept, not a loan-by-loan concept. This is consistent with modern risk management practice and allows discussion to move away from the largely binary perspective of default/no-default that necessarily surrounds evaluating a single loan. Furthermore, the loan loss reserves should be defined as the expected shortfall (in the technical statistical sense) of future repayments and recoveries relative to the historical book value of all obligations. Hence the loan loss reserve would be the mean of a probability distribution of eventual future losses for the entire portfolio.
Such an estimate would still leave considerable room for judgement. At least, however, disagreements would be over appropriate estimation procedures and relevant data, not over the very definition of what is being estimated.
1 These are the Federal Reserve Board, the Office of the Comptroller of the Currency and the Federal Deposit Insurance Corporation
2 Testimony of America’s Community Bankers on Loan Loss Reserves before The Subcommittee on Financial Institutions and Consumer Credit of The Committee on Banking and Financial Services of The US House of Representatives, June 16, 1999
David Rowe is group executive vice-president for risk management at SunGard Trading and Risk Systems.