Just a quick post to record an opinion on the idea of dealing with banks' troubled assets by writing insurance against their most extreme losses, a solution which has been applied in the UK, and is being considered in the US and EU as an alternative to buying the troubled assets outright. So far, proposed troubled asset purchase schemes like the TARP have been stymied by the problem of valuing the assets, which are typically mortgage-based complex structured products, at the right level to induce the banks to sell them while avoiding losses to the taxpayer.
RebelEconomist has never tried to value structured products (as a fund manager, he always avoided them), but he would imagine that the key to realistic valuation of them is their left tail (extreme loss) risk, for two reasons. First, he suspects that the structuring was often designed to shape the securities' return distributions to minimise the probability of default as defined by the rating agencies, while accepting the maximum risk given default (not to mention accepting the maximum risk short of default too) as far as such risks attracted a return premium. In other words, structured products were built to arbitrage the naïve reliance of risk management on credit ratings. The left tail of their return distribution may well be very different from regular fixed income securities. Second, by definition returns in the tails of a return distribution are seldom observed, so the tails are the most uncertain parts of the distribution. And unfortunately, in securities involving credit risk, the left tail is typically long. It is the left tail of their return distributions where the difficulty of valuing structured products resides.
This view of the troubled asset valuation problem suggests that pricing insurance against their worst losses, which essentially means buying the left tail of the return distribution only, does not represent a significantly easier task, and that writing such guarantees is not necessarily a better deal for taxpayers than buying the assets. Guarantees look cheaper (ought to generate fee income up front, in fact) because they do not involve buying the bulk of the return distribution, over which the average return should be clearly positive and which will have a relatively well-defined value like a regular bond. The danger is that this minimal initial cost makes an insurance scheme enticing to politicians who can be seen to "do something" without asking voters to make an immediate tangible sacrifice, even if there is some attempt by the official accountants to estimate the expected outlay on the guarantees. Moreover, the fact that the cost arises later, as and when the guarantees are called upon, perhaps even under a different administration, reduces the government's incentive to drive a hard bargain on the fee.
RebelEconomist still believes that something like the PRAT scheme is the right approach.