Mark-to-Market vs Mark-to-Model in the Insurance Industry

How do you turn low-rated mortgages into high-rated bonds? There are basically two ways: overcollateralization and insurance. The spotlight has been on the former of late, as people wonder whether the overcollateralization models reflect the reality of the subprime market over the past year or two. And now the other shoe is dropping: David Reilly, today, wonders what the exposure of insurers in general, and AIG in particular, might be to the subprime meltdown.

AIG, of course, has its own models of how much it stands to lose on insurance it's written against subprime losses. But models, as we've seen, don't always match with market reality.

The company also said it didn't see problems related to a kind of insurance contract, or derivative, it has written against financial instruments that include some subprime debt. AIG based its all-clear signal for those derivatives on the fact that its internal models show that losses are extremely remote in the portions of the investment vehicles it's insuring. No likely losses means no reason to worry, the company reasoned.
Yet the company's valuation models seem to ignore the fact that those derivatives would likely take a haircut if sold in today's depressed market. "There's no way these aren't showing a loss," says Janet Tavakoli, president of Tavakoli Structured Finance Inc., a Chicago research firm. That's simply a market reality, she adds, that should be showing up in AIG's results.

This raises an interesting question. Increasingly, there are market-based alternatives to insurance products: you can issue catastrophe bonds in the market, for instance, rather than insuring against a catastrophic event. In this case, AIG's insurance is pretty much identical to the credit default swaps that are being traded in the market and which have seen huge price volatility of late.

But when insurance companies have been in their business for decades or longer, while the open market in such things is very young, does it really make sense for insurance companies to use market valuations rather than their in-house models? Yes, models can be wrong – but so can the market, especially, as now, when there's a lot of stress and volatility.