Friday, September 4, 2009

Time to take "procyclicality" out of bank capital requirements

In a recent post called The perverse impact of Value at Risk, we focused on the problem of banking institutions using market based models to set aside capital. For example the same dollar position in the S&P500 (or most other asset classes) during 2007 would have a third of the current (2009) capital requirements (because 2006-07 market volatility was so much lower).

This issue is actually not unique to trading positions. Loan loss reserves follow the same pattern. Banking institutions use historical default experience to assign loss reserves to loan portfolios, but this data tends to be cyclical. Therefore the reserves would drop during a prolonged economic expansion. Instead, however, banks should be doing the opposite - increasing reserves during good times to be able to draw on them during times of economic stress.

The US Treasury finally woke up to this problem and issued a statement yesterday with guidelines to reform bank capital requirements. They focused among other things on what they call "procyclicality of the regulatory capital and accounting regimes". During times of growth, capital requirements drop, promoting extension of more credit and increasing the ability to take larger risks. That in turn feeds growth of credit and liquidity in the system, creating what's called a "positive feedback loop". An example of that would be placing a speaker next to a microphone and connecting the two - creating the unbearable noise and potentially frying the speaker. This is how credit bubbles get built.

Of course this also works in the opposite direction. A "negative perturbation" in the system causes capital requirements to go up and liquidity and credit availability to get sucked out of the markets, forcing capital requirements to increase even more, and so on (violently deflating the bubble). Engineers would call this procyclicality an "unstable equilibrium".

From the US Treasury:
Certain aspects of accounting standards also have procyclical tendencies. For example, during good times, loan loss reserves tend to decline because recent historical losses are low.

... The capital rules should rely less on procyclical Value-at-Risk (VaR) models, point-in-time internal rating systems, and non-stressed risk parameters. A movement toward greater use of longer-horizon, through-the-cycle risk estimates should result in higher capital requirements in the early phases of the credit cycle and more uniform capital requirements throughout the cycle.

...in determining their loan loss reserves, banking firms also should be required to be more forward-looking and consider factors that would cause loan losses to differ from recent historical experience.

The Treasury proposes some creative solutions to even out bank capital requirements or even push them to become "anticyclical" (moving toward a "stable equilibrium"). Some examples include linking capital and loss reserves to economic indicators - the stronger the economy, the more incremental capital (above the minimums required) the banks should reserve. Banks can also issue debt that converts to equity in an economic downturn, effectively recapitalizing the bank.

It's a well thought out proposal - definitely worth a read.