I’ve been involved in financial reporting for a large amount of my career, so even though I’ve never had an accounting course in my life, I’ve had to work with some of the most arcane accounting rules out there as an actuary, and later as an investor. Over the years, the direction that the FASB and IASB have gone is in the direction of presenting the statement of financial position (balance sheet) on more and more of a fair market value basis. (Please ignore the treatment of goodwill, advertising, R&D, you get the idea though…) To soften the blow on the income statement, changes in the value of many balance sheet items don’t get run through net income, but through accumulated other comprehensive income, so that income can reflect sustainable earnings power, in theory. Now, I agree with Marty Whitman’s critique on these accounting issues. We may be getting more accurate on individual companies (if the accounting is done by angels, for humans we are granting too much freedom), but we are losing comparability across companies. What an item means on the balance sheet of one company may be considerably different than the value at another company.
The hot topic today is SFAS 157 and 159. I would point you to Dr. Jeff’s article this evening on the topic. I would like to give my perspective on this, becaue I have had to work with these accounting rules, and ones like them.
At one company that I managed money for, I originated a bunch of long duration high quality assets that did not trade. At year end, our incentive payment was based on the total return that we generated. Interest rates had fallen through the year, and so my high quality illiquid assets had yields well in excess of where new money could be deployed. What were those assets worth? Historic cost? The cash flow streams were fixed. As a conservative measure, though spreads over Treasury yields had fallen for those instruments, we kept the spreads from the issue, and accounted for the price change due to the move in Treasury yields. (If spreads had risen, I would have argued that we move the spreads up as a conservative gesture.) Now this was prior to SFAS 157, but it illustrates the point. How do you calculate the value of illiquid instruments? Worse, under SFAS 157, you can’t be conservative; you have to try to be realistic.
Now, that was a simple example. Almost every moderate-to-large life insurance company has a variety of illiquid privately placed bonds for which there is no market. What is the fair value? Who can tell you? Well, the broker(s) that brought the deal are supposed to provide continuing “color” on the bonds, and what few trades might transpire. Typically, they don’t move the prices much as the interest rate and spread environments change, and third party pricing services are loath to opine on anything too illiquid. Though the rating agencies night give a rating at issue, they might not update it for some time. What’s the fair value? The life insurer has a hard time determining that for that small minority of assets.
Now let’s take it to a yet more difficult level. If we are talking about many asset-backed securities, they are generic enough that pricing models can determine a spread to Treasury or Swap yields for tranches with a given vintage, maturity, originator, and rating. Yes, there will be many assets that “trade special,” but those are deviations from the model that the traders feel out.
With CDOs, things get more difficult, because aside from indexed CDOs, there is no generic structure. The various tranches are bought and held. They rarely trade. Projecting the cash flows is a difficult talk, because there are many different bonds in the trust, with many different scenarios for how many will default, and what recoveries will be obtained. The best a good simulation model can do is to illustrate what a wide variety of possibilities could be, and look at the average of those possibilities. Even then, the modeler has an expected cash flow stream. What’s the right discount rate to use?
There is no good answer here. One can try to infer a rate from what few trades have happened in the market with similar instruments, but that can be unreliable as well. During a bear market, the sellers will be more incented than the buyers, particularly if they are trying to realize tax losses. One can try to look at the scenarios across the tranches, and see which tranches have cash flows that behave like bonds, equities, and warrants, and apply appropriate discount rates like 6%, 20%, and 40% respectively. Some explanation:
- Bonds: pays interest regularly, and principal within a narrow window. Few deferrals of interest.
- Equities: high variability of payoffs. Pays something in almost all scenarios, but the amounts vary a lot. Timing and existence of principal repayment varies considerably. Interest deferrals are common, but rarely last long.
- Warrants: many scenarios have very low or zero payoffs. Some scenarios have significant payoffs. Interest deferrals last a long time, many never end. Principal payments are rare.
Estimating fair value in a case like this is tough, if not impossible. But a fair value must be estimated anyway. Management teams may try to make the third party estimator come to a certain value that fits their accounting goals. Given the squishiness of what the discount rate ought to be, management teams could say that once the market normalizes a low discount rate will prevail, and our models should reflect normalized, not panic conditions.
Well, good, maybe. The thing is, once we open Pandora’s box, and allow for flexibility in valuation methods, subject to auditor sign-off (now, who is paying them?), our ability as third party investors to evaluate the value of illiquid assets and liabilities declines considerably. There’s a great argument here for avoiding companies that own/buy complex assets in an era where fair value accounting reigns. There is too much room for error, and human nature tells us that the errors are not likely to yield positive earnings surprises for investors.