Christian Leuz emailed me with a link to his recent paper with Christian Laux, entitled “Did Fair-Value Accounting Contribute to the Financial Crisis.”  He thought it might be of interest to our readership, and having spent some time with the paper this morning, he is right. In short, the paper’s answer is “no,”  which is consistent with the SEC’s Study on Mark-To-Market Accounting.  The paper is not a full-throated endorsement of Fair Value — it is far more positive than normative, and notes that banks may have used the leeway of FV measurement to overvalue assets (rather than undervalue them).

I have not spent enough time with the paper to assess the validity of their calculations. But I like some of the distinctions the authors make, which are all too often forgotten in debates about FV.

The paper begins by addressing concerns about destabilization, and remarks:

Furthermore, downward spirals can arise because contracts (e.g., margin and collateral requirements, haircuts) are based on market values or because banks use market values to manage their business (e.g., Value at Risk techniques). It is easy to confuse problems that stem from using market prices in private arrangements with problems from using market values in accounting. Thus, it is important to be specific about the links through which write-downs under fair-value accounting can create problems, be it capital regulation, contracts, a fixation on accounting numbers by managers or investors, or inefficient markets. [My emphasis]

Hear, hear!  Critics of fair value need to be clear on the mechanism by which it causes problems.  To the extent the mechanism is through banking regulation, the solution is more likely to be decoupling of regulatory accounting from GAAP accounting.  And to the extent it is ‘fixation,’ efficient market theorists have some explaining to do.  (I should point out that the paper does a very nice job of discussing banking regulations, so this makes a good primer for non-banking experts).

Laux and Leuz also distinguish clearly between assets and inputs.  All too often, people refer to “level 1 assets” and “level 3 assets,” thinking that level 1 assets are those for which there is an observable market price for identical assets, and level 3 assets are those for which management must use their own estimates of value (“mark-to-model”).  Not quite.  Instead, management can use level 1 inputs to measure the fair value of an asset (by looking at market prices for identical assets) or can use level 2 or 3 inputs (similar assets or internal models) to measure the fair value.

This is important in understanding the role of fair value measurement in the financial crisis, for, as the authors point out,

Fair-value accounting as stipulated by the FASB has several safeguards against marking to potentially distorted market prices and hence against accounting-induced downward spirals and contagion.

First, the rule explicitly states that prices from a forced liquidation or distress sale should not be used in determining fair value. Thus, if fire sales occur, banks should not mark their assets to these prices, which amounts to a “circuit breaker” in the downward spiral. In practice, it can of course be difficult to identify prices that stem from fire sales but the rule gives banks a legitimate reason to discard certain prices.

Third, as markets become inactive, FAS 157 explicitly allows banks to use valuation models to derive fair values. That is, banks are not forced to use dealer quotes that are distorted by illiquidity. As the financial crisis deepened, banks used this option. Of all the assets reported at fair value in the first quarter of 2007, bank holding companies used Level 1 inputs (quoted prices) for 34 percent of them; by the first quarter of 2009, this fraction decreased to only 19 percent. For bank holding companies, most of the decline in Level 1 assets appears to be compensated by an increase in Level 2 assets, although Level 3 assets increase from about 9 to 13 percent (Table 3). For investment banks, Level 3 assets also increase to 14 percent, mirroring a decrease in Level 1 assets from 27 to 22 percent.

Keep in mind that these are the same assets, but the banks are choosing whether to use level 1 or level 3 inputs, as the level 1 inputs become tainted by illiquidity (disorderly transactions).

One minor quibble:  if I had my way, the authors would replace all uses of the term ‘fair value accounting’ with ‘fair value measurement.’  I know that the first term is more common, and even the FASB uses it.  But fair value accounting suggests that any number of things might be changing.  But the only issue here is how assets are measured, not whether assets and liabilities are being recognized or derecognized, but how they are measured.  For example, we often hear concerns like “fair value accounting will allow firms to recognize profit on products they have just begun designing, because their internal models indicate it is a positive NPV project.”  Not true — fair value standards govern only the measurement of assets that have already been recognized, and I don’t see much chance that we will be recognizing assets for products in the design phase.