The Financial Accounting Standards Board has been working for decades on accounting for financial instruments. While the urgency associated with the topic has waxed and waned, it has never completely dropped off the agenda. Of course, the financial crisis has pushed it to the forefront again.

A May 2010 FASB proposal would have required virtually all financial instruments to be carried at fair value. This would have provided a common measurement characteristic for financial instruments and could have solved many problems while significantly reducing complexity. As everybody knows, FASB has since changed its course. Many observers have criticized FASB for backing away from the fair-value proposal, describing the change as a retreat and implying that board members caved in to pressure from Wall Street banks and their lobbyists.

Well, I won't be joining that group. I do believe that fair value is the best measurement for financial instruments, and I've grown tired of the baseless predictions of catastrophe that we always hear from financial institutions (and sometimes their regulators) whenever we raise the prospect of telling investors what financial instruments are actually worth. But feedback on FASB's proposal indicated that opposition to reporting loans at fair value goes beyond the financial firms. Financial statement users (analysts, investors, and so forth) don't really want it either.

In other words, FASB's customers told the board they didn't like the proposal. If changing a product to match customers' preferences is caving and retreating, every successful business is as guilty as FASB is.

So, despite being a proponent of fair value for financial instruments, I don't mind the so-called “three-bucket” approach now being considered. But another aspect of the financial instruments project concerns me quite a bit: the recent proposal regarding loan impairment. That proposal would require that loans in so-called open portfolios be separated into a “good book” and a “bad book.” The bad book would consist of loans for which the lender has already identified a problem, and all expected losses on such loans would be recognized.

This isn't significantly different from how such “bad book” loans would be handled today. The change would come in how losses would be recognized on loans in the good book—those loans that haven't been individually identified as being impaired. Of course, you need a reserve to cover losses inherent in those loans that simply can't yet be pinned down. The new proposal would address that by spreading expected losses over the life of the good loans and by imposing a floor on the loss reserve, such that it would always cover at least the losses expected in the “foreseeable future.” I'm not fond of either of those requirements, but that isn't really my biggest concern.

A Risky Proposal

My main concern is with one of the stated objectives of the approach: “ensuring that the amount of the allowance for credit losses is adequate to cover expected credit losses before they occur.” Did I read that right? The objective is to recognize losses before they occur? I understand that losses on loans generally exist before we get to the point of identifying the particular loan as uncollectible. But the proposal doesn't say that the goal is to recognize losses when they occur, rather then when they become known—it says we should try to recognize the losses before they occur.

In that case, what other expected losses should we record before they happen? Maybe we could recognize severance payments before we've fired anybody, because we know we're likely to have a restructuring and will offer severance when we do. Oh wait … we did do that, until FASB prohibited it. Financial Accounting Standard No. 146, Costs Associated with Exits or Disposals, was issued in part to delay the recognition of severance and exit costs because the board was concerned that losses were being recognized before they were incurred.

My real concern is that once we start recognizing the effects of things that are expected but haven't yet happened, we remove a lot of discipline from the way we do accounting.

And now we're proposing a model that has, as an objective, recognizing losses before they happen? This isn't a situation when the portfolio is expected to lose money in the aggregate—the good book would still be expected to earn more in interest than it would lose in credit losses.

Don't get me wrong; I don't love current loan loss accounting, but the objective of recording losses when they happen isn't the problem. If it were, we'd need to change a whole lot of accounting requirements. Instead, I think the problems with the current model are caused by two things.

First, knowing when a loss on a loan actually occurs is not easy. In some cases, the loss is immediate, as there are some loans that never truly stand any chance of being repaid. In other situations loan losses happen sometime later, perhaps when the credit card holder suffers an injury and can no longer work, or when a company's main product is upstaged by new technology. More frequently, however, it's unclear when a loan loss occurred even in hindsight, because lots of events lead up to a loan being uncollectible.

In the recent crisis, for example, many mortgage loans went bad, but it isn't clear whether the loss event was the collapsing resale market, the worsening job market, the individual borrower's pay cut or job loss, or that the loan shouldn't have been made at all. In reality, they all probably factored into the loss.

The second problem with current accounting is that we don't reserve until we think it's “probable” that a loss has happened. That means, for example, that if it is 60 percent likely that a loss as been incurred, a reserve wouldn't be recorded under the current guidance. Of course, this problem is easily rectified. We could simply change the guidance to require recognition of a loss when it's “more likely than not” that a loss has been incurred.

The problem of being unable to identify when a loss occurs, however, is much more difficult. FASB's latest proposal would solve this by specifically trying to build up loss reserves before any loss occurs, mandating that all expected losses be recognized over the life of the loan, with an overlay of a minimum reserve to cover all losses expected in the foreseeable future.

Among the justifications for this is the belief that losses in the recent financial crisis were recognized too late. But I find it difficult to believe that we'd have seen loan loss reserves recognized much earlier if the guidance focused on expected rather than incurred losses. After all, if anything approaching the actual level of losses had been “expected,” we wouldn't have continued to see such free lending go on until immediately before the collapse. The plain fact is that lenders took larger losses than they expected. Telling them to record the expected losses won't help if their expectations are too low.

But the fact that recording “expected losses” wouldn't have helped in the recent crisis isn't the main reason I am concerned about the proposal. My real concern is that once we start recognizing the effects of things that are expected but haven't yet happened, we remove a lot of discipline from the way we do accounting. Things that we believe will happen in the future should be discussed in Management's Discussion and Analysis, or Risk Factors, or even the financial statement footnotes—but they shouldn't be recognized in the financial statements.

A More Reasonable Proposal

There actually could be a place in an incurred loss model, for both a ratable recording of some losses and a floor. Both could be justified as methods of attempting to estimate the losses in a portfolio that have already occurred but haven't become evident yet. Ratable recognition of expected losses could be explained as dealing with the reality that many events taken together typically lead to a loss. Perhaps therefore, the best estimate of when a loss actually occurs is ratably over the life of the loan.

A floor, on the other hand, could be explained by noting that, on average, a period of time elapses between the time a loss on a loan in the good book actually occurs and the time the bank's management flags the loan as being troubled and moves it to the bad book. To account for this delay, we could assert that a 12-month look-forward (or some other timeframe) reserve is necessary to get at the losses that have already happened, but we don't yet know about.

While ratable recognition and a floor wouldn't be my preferred methods of getting at the “incurred but not reported” losses, they could still be explained and used within a model that tries to recognize losses when they occur. Instead, though, FASB proposes them as part of a model intended to recognize losses before they occur. The implications of embedding a principle into any part of the accounting literature to recognize something before it happens are huge. Even if FASB sticks with the mechanics of the recent proposal, we should all hope that it changes the principle. Because a principle of recognizing losses before they occur may be the beginning of an awfully slippery slope.

Topics