Wednesday, April 08, 2009

Regulatory Capital Arbitrage

I was in charge economic capital allocations at KeyCorp. We allocated 'economic' or 'risk' capital within the bank. This exercise was used for reporting, pricing, compensation, and risk management. One issue that often arose was what to do when the economic risk capital allocation was different than the regulatory capital allocation. I wrote an article in Bank Accounting and Finance in 1997 on this issue. As there were relatively few buckets for regulatory capital--6% for most assets, 2% for OECD debt--the issue was always present. But 'arbitraging' regulatory capital was only recommended when the estimated risk capital was lower than the regulatory capital. That is, people don't increase leverage by securitizing unless they believe, and their debtors believe, they do not need the capital required by regulators. You arbitrage it when it is too high, not too low. Banks don't intentionally take on too much risk.

Felix Salmon notes one can reduce one's regulatory capital requirement by about 1% by securitizing. That's fine, but in general banks that could avail themselves of this strategy were not increasing leverage over the past 20 years. Most estimates of the required risk capital were much lower than regulatory capital, around 2-3%, because they were assumed to have their historical credit loss rates, which were de minimus. That was a mistake, to be sure, and people like Peter Schiff were spot on in seeing that, but it was the best estimate of academics, regulators, investors, and bankers.

Thus, the idea that we need to prevent regulatory arbitrage presumes this is because risk capital is greater than regulatory capital requirements in certain circumstances. With hindsight, there was too little capital allocated for mortgages originated between 2005 and 2007. Then again, with hindsight, they had a negative NPV, and should not have been issued by any rational profit maximizing firm. But that's hindsight. The market risk capital for those assets is probably much higher than any regulatory requirement right now. This is a problem that does not need fixing, because bankers do not take on too much risk on purpose. If the regulatory capital requirement is 2%, and your assessment is 6%, you allocate 6% unless you have excess capital from other investments. The idea that bankers knew the asset required a 10 to 1 leverage ratio, but chose 30 to 1 because regulations allowed it, is like assuming that people drink 12 shots of scotch in 30 minutes because it is legal to do so. The key reason is the willingness, not the legal ability. It is impractical to design a legal system that will capture the myriad nuances involved in allocating capital to literally thousands of different bank liabilities. Currently there are 3 different regulatory capital buckets (OECD government debt, A rated investments, and other). In a few years, this might change to 5.

The bottom line is this crisis was not caused by regulatory arbitrage, but rather, everyone assuming mortgages had much less risk than they did, or do. You should not give a mortgage to someone who can't pay for it on the presumption that collateral never declines in value. We now know mortgages are much riskier than we thought in 2006. Bankers do not lever investments more than they should merely because it is feasible under the regulatory requirements. Like the Space Shuttle disaster, fixing the specific cause of the last crash is easy (more resilient O-rings). Having a mechanism that better anticipates the next unknown failure is not so simple.


Anonymous said...

i njoy reading what u write because it is obvious that u r an intelligent and articulate person. what makes it even more interesting is to note that it is possible for a person like myself (moderately intelligent and articulate) to spot how limited ur sort of reasoning can be ...
we dont need mechanisms to anticipate the next unknown failure - we need mechanisms to prevent the next KNOWN FAILURE LIKE THE CURRENT ONE:

... all we need are:
1) NOT the moral hazard as currently
2) compensation with long term claw-back provisions

Anonymous said...

I think your experience is just evidence that KeyCorp was not a "too big to fail" bank. (The assets of the top 5 bank holding companies are more than 50 times KeyCorp's assets.)

Of course, there are thousands of banks that have to worry about a safe level of capital requirements, rather than regulatory capital requirements. Unfortunately those thousands of banks account for only a small fraction of bank assets in this country.

Andrew said...

Can someone help me by pointing to some meaningful evidence that it was impossible for the collateral to lose value?

I grew up in the UK, and purchased my first appartment in London in 1990 for 56,000 pounds. The couple I bought it from had purchased it for 59,000 pounds two years earlier.

Three years later, I sold the flat for 42,000 pounds. I am not the world's best negotiator, but I can't see how anyone could have done better than a 10,000 pound loss (as opposed to my 14,000 loss) in the same period.

It's possible that some economists were aware that in markets like Japan, the UK, and elsewhere, house prices had declined in nominal terms in the past (attempt at irony).

Anonymous said...

I am so glad you choose the O-rings! The lesson to learn from that incident is not that more resilient o-rings should be used. The lesson to be learned is that you must not use them outside their specs, whatever their resilience is. And, as the great Feynman noted, you must not distort the facts in your analysis, because you can not fool nature. The same is true with a modification for human beings: Some of the people all the time, all the people some of the time, but not all people all the time.