Posted on RiskCenter.com by Donald van Deventer:
On April 9, 2009, under a firestorm of political pressure, the Financial Accounting Standards Board staff put forth the enticingly titled document "FSP FAS 115-2 and FAS 124-2." If only the "and" were replaced with "&," the FASB staff could have outdone poet e. e. cummings with a written work whose title contained no words. Among this blog's readers is a very very bright risk manager, working in disguise as his bank's chief operating officer, someone we'll call Vandy.
Vandy has gone through the (here comes the real title) "Financial Accounting Standards Board Staff Position Number 115-2 and Number 124-2: Recognition and Presentation of Other-Than-Temporary Impairments" and pointed out that this FSP (FASB Staff Position) document seems to contain some very bad financial analysis, specifically with respect to the discount rate used for valuation. We examine that issue in this post.
In our April 2 post on the biggest mistakes in FAS 157 valuations, mistake number seven was a series of technical mistakes, most notably "double counting credit risk." In that original post, we discuss Bob Jarrow's comment that there are an "old way" and a "new way" of valuation that are totally consistent with each other. The April 9 FASB Staff Proposal is consistent with neither approach. It represents the collision of three irresistable objects: political forces, legacy accounting rules and modern finance. In the end, if politics and legacy accounting rules lead to valuations that are inconsistent with market prices, then modern finance triumphs, because modern finance always starts from market prices and works backwards to insure that the valuation assumptions are consistent with those market prices. Yesterday's post on the Copula method for valuation was a "post mortem" on why that valuation formula was rejected as inaccurate. The same will happen with April 9's FASB staff proposal.
Let's talk about a bond that has a currently observable market price of 68. What Bob Jarrow describes as the "old way" of valuation has two steps. Step 1 is to lay out the cash flows that are scheduled to be paid. Step 2 is to then solve for the yield to maturity (or spread to a floating rate index) that makes the valuation equal 68 plus accrued interest. Note that this methodology puts the full adjustment for credit risk in the discount rate, not the "expected cash flows." See chapter 7 of my book with Mark Mesler and Kenji Imai, Advanced Financial Risk Management (John Wiley & Sons, 2004) for the exact traditional "bond math" calculation.
The "new way" of valuation is the modern approach to derivatives valuation like the Black-Scholes options model, something already enshrined in Generally Accepted Accounting Principles, albeit with a lag of some 20 years after the original paper was published. The process again is a two step process. The random payoffs on the security are laid out in N scenarios, which may include random interest rates. Within each of these N scenarios, the random "risk free" interest rate is used to discount the cash flows for that scenario. The valuation for the security is the average of the values for the N scenarios. Note that this generally will NOT be equal to the discounted value (using the average simulated risk free rate values) of the average cash flows over the N scenarios when the cash flows are highly skewed over the scenarios, like they are in almost all recently issued structured products like CDOs.
Since the "new way" is already enshrined in GAAP, is that what the April 9 FASB Staff Prosition calls for? More than 30 years ago, the chief tax accountant of Security Pacific Corporation patiently explained to me that "logic" and the U.S. Tax Code have nothing to do with each other. Because accounting pronoucements have such a heavy political content, it's disappointing to see that the current proposals seem not to recognize that they are inconsistent with both the "old ways" of valuation in GAAP and the "new ways" of valuation in GAAP, particularly those regarding stock options. The new proposals specify a valuation that cannot possible match market prices when they are observable.
As Vandy describes the FASB Staff Proposal, he says that the FASB is hung up on discounting either at the "effective rate" on the loan or the "current yield" in the case of a beneficial interest like a CDO tranche. What is the "effective rate" on a CDO tranche? Implied in various FASB documents is the fact that it's the "old way" yield that prevailed at the time the tranche was purchased, say 3 years ago. Why would one use a 3 year old discount rate to value the tranche today? In addition to the timing mistake, there is a second more serious mistake. April 9's position requires that credit losses be deducted from cash flows before discounting, but the "old way" valuation method has the full credit adjustment in the discount rate, not the cash flows. By making the credit risk adjustment in both places, the calculation is completely wrong. Moreover, by following the prescribed calculation, there is no way to back into a currently observable market price for a CDO tranche unless credit losses are zero--because the calculation requires you to double count them. The "current yield" discount rate for CDOs is only slightly better, because it's "current," not the yield which prevailed at origination. The double counting mistake remains, and it's such a fundamental flaw that it's hard to understand how it emerged in print, except to recognize that the political pressures for a speedy change to mark to market regulations made mistakes more likely.
What about the "best case" information environment where we are trying to justify valuation for a CDO tranche where there has just been a real trade at 68, but the current bid of 58 and offered of 78 is a spread so wide that no more trades are taking place. Which number should be the focus of valuation? Most finance and market experts would argue that the real trade or mid-market price should be what one is trying to back into. Like the Black-Scholes options valuation where one backs into implied volatility from the observable price, we check to insure that our simulation of losses and discount rates produce a reasonable price. In yesterday's blog, we described at length how it was eventually realized that the copula method for CDO valuation was not able to produce realistic prices in the current environment. For the same reason, double counting credit losses doesn't match market prices either unless losses are zero.
What if there is a quote of 58 bid and 78 offered, but no recent trade at 68? One would still fit the valuation technology to the mid-market price. The spread between bid and offered is a function of lots of things, but primarily it's due to the level of uncertainty in valuation. In the basis swap market, for example, the bid offered spread is very wide for swaps between two rates with an unpredictable relationship. One shouldn't make the mistake that only bid prices are real, because there may not be a bid either.
What does one do in that case? One uses the "new way" of valuation, where very well tested default probabilities (see the KRIS default probabilities on this web site) are driven up and down by macro economic factors, and the resulting post credit loss cash flows are discounted at the risk free rate that prevails in each scenario. When there are no observable market prices, netiher the "effective rate" nor the "current yield" are very helpful. At least the "effective rate" is known, because the yield at origination was observable at the price the investor paid. That doesn't make the flaws of the technique any less. The "current yield" is a useless concept when there is literally no bid/no ask. Why not use the yield on other CDO tranches, one might ask? Because the devil is in the details. If the reference collateral is different, there is no reason at all to expect the "current yields" to be comparable.
What does "other than temporary impairment" mean to the FASB, and what should it mean to market participants? We leave that issue for another day.