Sunday, February 7, 2010

A Proposed Fat-Tail Risk Metric

Posted on by Peter Conti-Brown:

Abstract: This paper argues that the financial regulatory reform currently debated in the U.S. Congress misses a key opportunity to address one of the central causes of the financial crisis: the failure of risk models to account for high impact, low probability events. The paper proposes a legal solution that will create a more robust metric: require mandatory disclosure of a firm’s exposure to contingent liabilities, such as guarantees for the debts of off-balance sheet entities, and all varieties of OTC derivatives contracts. Such disclosures - akin to publicly traded corporations’ filing of 10-Ks with the SEC - will allow regulators and researchers to approximate an apocalyptic, black-out, no-bankruptcy-protection and no-bailout scenario of a firm's implosion; force firm’s to maintain daily record-keeping on such obligations, a task which has proved difficult in the past; and, most importantly, will open up a crucial subset of data that has, until now, been opaque or completely invisible. With such data, researchers, over time, will develop analytical and econometric tools that better assess the consequences of remote events for individual firms and, more importantly, the economy as a whole.

Felix Salmon's comments:

The paper itself is flawed, and the details of how it’s constructed would need to be reworked from scratch. But conceptually, the FTRM I think is a good idea.

Conti-Brown’s method for coming up with the FTRM involves adding up a firm’s total netted derivatives exposure; the size of its off-balance-sheet vehicles; and its liabilities. That gives a total-risk measure; the FTRM itself is the log of that figure.

There are lots of problems here. For one thing, netting derivatives exposure effectively eliminates an enormous amount of counterparty risk. For another thing, it’s impossible to calculate: if I write a call option on a stock, there’s no limit to how much my contingent liability might be, because there’s no limit to how far that stock can rise. And off-balance-sheet vehicles are just one of a potentially infinite line of entities which remove a company’s legal liability, but where the firm can still end up paying out a lot of money in practice. Think, for instance, the money which Bear Stearns threw at its failed hedge funds, or the money which banks used to make whole the people who invested in auction-rate securities. Those things don’t look like bank liabilities, or even contingent liabilities, until it’s far too late.

But put all that to one side: one thing which doesn’t currently exist, and which would be very useful indeed, is some kind of measure of the total amount of risk in the financial system. A lot of people had a conception, pre-crisis, of some kind of law of the conservation of risk: that tools like mortgage-backed securities simply moved risk from banks’ balance sheets to investment accounts, and therefore, at the margin, actually dispersed risk and made the system safer. What was missed, however, was the fact that total risk was increasing fast, especially as house prices rose and the equity in those houses was converted into financial assets through the magic of second mortgages, cash-out refinancings, and home-equity lines of credit.

Some types of risk are more dangerous than others, of course: if there’s a stock-market bubble, then it’s easy to see that the total value of the stock market, which is the total amount that can be lost in the stock market, has risen a lot. But stock-market investments are a little bit like houses without mortgages: where there’s very little leverage, there’s also relatively little in the way of systemic risk. It’s rare to suffer great harm from the value of your house falling if you don’t have a mortgage.

Still, stock-market bubbles can cause harm, and it’s worth including equities as part of the total risk in the system, along with bonds and loans. That’s one metric which macroprudential regulators should certainly keep an eye on; Conti-Brown’s idea is then basically to try to disaggregate that risk on a firm-by-firm basis, to see which companies have the most risk and to see how concentrated the risk is in a small number of large institutions.

It won’t be easy to do that — indeed, it will be impossible to do it with much accuracy. But even an inaccurate measurement will be helpful, especially if it becomes a time series and people can see how it’s been changing over time. It’s good to know how much risk is out there — and it’s better to know that financial institutions themselves are keeping an eye on that number, and trying to measure it as part of their responsibilities to their regulator.

No comments: