Earlier today at RealMoney, I responded to a question in the Columnist Conversation. It was a longish post that tried to be complete, so I reprint it here:

		David Merkel
		Mark-to-Management Assumptions
		9/18/2007 1:50 PM EDT

Bob, Joe is essentially correct, but I’d like to add a little. From the Office of Thrift Supervision Examination Handbook (pages 137 & 138):

Observable Inputs – market participant assumptions developed based on market data obtained from sources independent of the reporting entity

Level 1 Inputs – Unadjusted quoted prices in active markets for the identical assets and liabilities that the reporting entity has the ability to access at the measurement date. Examples: Treasury bonds and exchange traded securities.

Level 2 Inputs – Other than quoted prices included in Level 1 that are observable for the asset or liability, either directly or indirectly. Level 2 inputs include quoted prices for similar assets or liabilities in active markets and quoted prices for identical or similar assets or liabilities in markets that are not active. Examples: loans traded within the secondary market and plain vanilla interest rate swaps.

Unobservable Inputs

Level 3 Inputs – Entity specific inputs to the extent that observable inputs are unavailable. Because there is little to no market activity, these inputs reflect the entity’s supposition about the assumptions of market participants based on the best information available in the circumstances. In those situations, the reporting entity need not undertake all possible efforts to obtain information about market participant assumptions. However, the reporting entity must not ignore information about market participant assumptions that is reasonably available without undue cost and effort. Examples: credit enhancing I/O strips and private equity securities.

Another way to phrase it is this:

Level 1 – publicly observable data

Level 2 – derived almost entirely from publicly observable data, and a commonly-used model

Level 3 – significant use of private firm-specific data, or public data not derived from the markets (think of a life insurance industry standard mortality table)

Now, I’m not a fan of SFASs 157 & 159, or any of the current statements dealing with intangibles. Even level 1 is subject to problems when markets are less liquid. I’ve known of situations where a bond manager found himself holding a disproportionate share of the market of a publicly tradable bond, where it almost never trades because he owns so much of the issue. Where do you mark that? That’s just level one!

Aside from AAA securities, most asset backed bonds never trade. Level 2 comes into play here, because the dealers estimate a pricing grid from what few transactions take place. with “fair value” accounting, there is no way to avoid mark-to-model, but there are significant possibilities for error.

The classic case of level 3 is how one estimates the changing value of private equity investments over time. Discounted cash flow anyone?

As a result of the changes, we have to be a lot more careful in how we interpret the financial statements of financial companies. The game just got a lot more complex given the new fair value accounting rules.

Position: none

After I wrote that, a friend of mine e-mailed me saying that Private Equity accounting was for the most part conservative at present, but that there was some pressure to use fair value accounting to smooth results. He also thought the use of these methods wouldn’t make private equity correlate more closely with public equities. I think he is onto something there, and that could affect that amount allocated by pensions and endowments to private equity. On the flip side, if the returns are smoothed through these accounting methods, the standard deviation of returns would drop, which is a bigger effect than the correlation effect. So allocations might go up, and some Private Equity managers, believing the smoother returns, might be tempted to lever up more.

One other note: I expect that companies with high percentages of level 3 assets will trade at discounts to relative to their peers. Accounting complexity and opaqueness always have valuation discounts. I see it in insurance for financial insurers, reinsurers, and long-tailed commercial lines. Uncertain assets and liabilities should always get lower valuations. Thus, aggressive users of fair value will wonder why their P/Es and P/Bs are so low. It’s because of the lack of ability of investors to verify the asset and liability figures used.