Just when you thought financial firm accounting couldn’t get more dubious…it gets worse.
Deux Ex Macchiato (hat tip FT Alphaville) tells of the disconcerting changes to what was formerly called FAS Rule 157, which brought us Level 1, 2, and 3 accounting. A brief recap:
Readers may recall that the Financial Standards Accounting Board implemented Statement 157, which required financial firms to identify how they arrived at the “fair value” for their assets. Level 1 are ones where there is a market price. Level 2 are those where there may not be much of a market, but they can nevertheless be priced in reference to similar assets that have a market price, or their price can be derived from “observable inputs” which are presumably from financial markets (the use of sunspots, skirt lengths, the Mayan calendar, or a model using, say, a ratio of bullish versus bearish stories on Bloomberg presumably does not qualify).
Level 3 assets are priced using “unobservable inputs,” and is therefore colloquially called “mark to make believe.”
Almost as soon as this regime was in place, the officialdom started giving waivers. The refinements to these valuation rules that become ever-more bank flattering. For instance, consider this item from March 2008 (post Bear, natch), which discussed a Floyd Norris article, “If Market Prices Are Too Low, Ignore Them”:
Norris, who is usually pretty understated, disapproved of one of the items in the SEC letter, as do we…
In the last couple of months, there has been increased worry that mark-to-market accounting leads to the operation of a destructive “financial accelerator.” As prevailing values go down, banks have to lower the value of their holdings. This leads to a direct hit to their net worth, which will lead them to contract their balance sheets, either by withholding credit or selling assets. More sales in a weak market lead to further declines in the prices of financial instruments, leading to more writedowns and sales of inventory.
Funny how no one had a problem with mark-to-market when asset prices were rising. The process in reverse leads to mark-to-market gains, higher net worths fueling balance sheet growth and credit expansion, which led to more demand for financial assets. That gives you higher securities prices which least to more mark-to-market gains. Sounds like a bubble, doesn’t it?
The SEC’s solution for the contractionary version of this dynamic is simple: ignore those market prices if they are too ugly. From the release:
Fair value assumes the exchange of assets or liabilities in orderly transactions. Under SFAS 157, it is appropriate for you to consider actual market prices, or observable inputs, even when the market is less liquid than historical market volumes, unless those prices are the result of a forced liquidation or distress sale (boldface ours).
Quite a few observers had argued that the windups of SIVs and the failure of hedge funds, and even Bear Stearns, would be a good thing because they would force price discovery of assets that are normally illiquid and/or hard to value. That in turn would resolve a great deal of uncertainty of what bank and hedge fund positions were really worth.
But now the SEC has given banks and brokers a huge out. No matter how small or easily absorbed by the market a forced sale might be (think of a hedge fund hit by a margin call), a financial institution can ignore the price realized. In fact, they get to determine what trades constitute a forced sale. As Norris dryly notes:
Some people on Wall Street think that nearly every sale today is a forced sale. There are entire categories of collateralized debt obligations where most, if not all, of the trades, occur because a holder has received, or expects, a margin call.
Back to our current post. The FAS changed the name of the rule from Rule 157 to Topic 820 last year, and Deux Ex Machhiato reports on the continuing devolution on this front. He finds a change he likes (FAS text first, his commentary second):
Financial statement users indicated that information about the effect(s) of reasonably possible alternative inputs [to level 3 valuation models] would be relevant in their analysis of the reporting entity’s performance.
So, with a reasonable amount of luck, 820 will require firms not just to state the value of their level 3 assets, but also to assess uncertainty in that value. This would be a major step forward in accounting disclosures for financial instruments, and I commend the standard setters for it.
Yves here. I hope he is right, but his interpretation looks optimistic. First, this is not a done deal yet, and second “reasonably possible alternative inputs” would appear to give some companies the latitude to define ‘reasonably possible” pretty narrowly, or simply contend that using other “reasonably possible alternative input” made an adverse difference only in a trivial percentage of scenarios. After all, VaR-based models computed the odds of the most two extreme moves in the Dow in October 2008 as being possible only once every 73 to 603 trillion billion years. Since our universe is maybe 20 billion years old, we’d need to wait at least a trillion universes to expect to see one month like that. Weren’t we all lucky?
Here’s the part Deux Ex Macchiato does not like:
I used to think that level 2 assets were things valued using a model, but where all the model inputs were current market observables. In other words, a swap valued using a discounted cashflow model calibrated to the quoted libor rates is level 2, but a quanto option valued using historic correlation isn’t, as correlation is not a current market observable (but rather an historic property). In fact anything valued using a model where one input is an historic property – historic vol, historic prepayment rates, etc. – should be level 3.
Unfortunately the text of 820 now includes the clarification that anything based on a market input is in level 2. And since historical volatility is based on a price history, an option priced using historic rather than implied is in level 2. This is not good. There is a crucial difference between a current price used as an input (or equivalently a convention for quoting prices, like implied vol) and anything else. Level 2 should be kept for purely price based model inputs. That, of course, would also make the level 3 uncertainty disclosures much more useful.
Eeek. Having crawled in the bowels of some financial firms, I’ve been skeptical of whether an investor can make head or tails of the performance of a financial institution of any complexity. This move should give the skeptics even more cause for pause.