Saturday, May 03, 2008

Persaud on Fixing Bank Regulation

Avinash Persaud has a neat article on how to amend our banking regulation. It's pretty thoughtful when put in the context of congressional suggestions (eg, buy back all mortgages at 85% face value, ban foreclosures), but ultimately, I don't think it's the framework of anything that would be better. Here's why:

1) First, he brings up his old idea that we should not use market prices in regulation, because 'markets' get us into trouble, ergo, 'market prices' is just doubling down on the problem. His model, explained here, is that if we view VaR type regulation as the key regulation of banks , then banks will necessarily heard into the same low VaR assets. Thus, eventually, everyone is in emerging market debt, or subprime real estate, and when one tries to sell, it causes a stampede via the positive feedback loop of lower prices, higher correlations, higher volatility --> worse VaR --> more selling --> etc.

My main problem with this model is that VaR is a pretty minor tool in the regulation of banks, and a pretty minor determinant of bank capital allocation. 10-day, 99%, VaR for UBS was about $420MM, which annualizes to $2B. And they wrote down $37B in one year. Now clearly their VaR was calculated too low, but it was also not covering much of the assets that caused the problem. It's like watching the Maginot Line while the real risk just crept around it. Whatever makes it into the 'banking book', and is not market to market, is not subject to the VaR, and that is generally the largest risk of banks. To think that VaR is the major influence on bank capital allocation is simply wrong.

VaR is mainly a device used to placate regulators, not to make real business decisions. That is why this device tends to focus on trading desks that make most of the money off of customer flow--bid-ask spread, commissions, trade impact--so in those cases the pnl/VaR is ridiculously high. In this case, it really doesn't matter what the VaR is, because your currency desk will have a Sharpe of around 10 using an annualized VaR, and a more conservative VaR might move that to 7, but either way, who cares? Whatever is subject to VaR is always so high above the hurdle rate (the stock market Sharpe is 0.4), this metric is clearly highly selectively applied. For the banks, its great, because the answer doesn't matter, and lots of PhDs get to talk to each other, and make each other feel like there's some really cutting-edge top down regulation going on. VaR is an essential tool for managing a desk, but as a top-down risk metric, it's a sideshow.

2) He suggests 3 pillars to a new approach.

a) First, countercyclical bank charges based on some asset price, so that bank pay in good times, and perhaps receive rebates in bad times. This seems at some level be a practical way to tax and subsidize the industry, given that, with periodic crises and the central bank response, this is what is happening in practice (they get guarantees, liquidity, lower funding, in bad times). However, the tax will in general be either too high or too low by a factor of at least 2; there are too few cycles to calibrate this correctly. This will cause lots of political haggling, and targeting of other objective (ie, special exemptions for 'good' investments), I think the deadweight losses would outweigh any gain.

b) Maturity mismatches. Now, the old days when you could measure interest rate sensitivity via maturity of assets and liabilities are long gone. Further, since 1994, most banks realize the folly of merely betting on the yield curve. So I think this misses its objective--sensitivity to interest rates--and misses the priorities. We need to be vigilant on interest rate sensitivity, and they should have to publish the results from stress tests that look at what happens when the curve shifts or tilts by certain amounts, and these should be analyzed by regulators and rating agencies. But thinking that maturity mismatch is a symptom of risk, or that maturity non-mismatch is a symptom of safety, is too easy to game.

c) Banks should pay more insurance premiums. Clearly, if the Fed is expected to step in in times of crisis, this 'option' is worth something to banks, and they should pay for it. But this option value is subject to moral hazard and adverse selection. The ignorant and greedy get the best bang for the buck in this type of scenario, hardly an incentive towards long-term stability.

I think fundamentally, we must recognize that whatever we do, there will be periodic financial crises. As we don't fully understand the mechanism, perhaps its a soluble problem, but perhaps not (remember, we've been studying business cycles since at least Jevon's sunspot cycle in the 1860's). I think a better solution would require better disclosure, in that, while Persaud notes that subprime was well labeled, no bank was required to show investors how much, or what kind, of subprime they had on their books. They knew what they were buying, but outsiders did not. Bank balance sheets are a black box, the famously non-transparent balance sheet of lenders, where everything 'current' looks suspiciously the same. It was news to me that Bear and UBS had tens of billions in RMBS on their balance sheets--they had no obligation to tell us, so they didn't until after the horse left the barn. Any stock analyst who saw that would have crucified them. So, disclosure is still a good option: how much of various types of assets, stratified by Agency rating, type (eg, corporate, residential mortgage, etc.), and maturity.

Secondly, regulators should protect against a contagion, but also should make certain to zero the equity holders out. That is, we need a law that prevents equity owners from clogging up the courts with suits that make any settlement that pays them little difficult, or the ability to threaten such suits. Equity owners hold an option, and so it is worth it to spend million in court, in order to get, say, a price of $10 instead of $2, in the current legal framework. The regulators should be able to give them $0 without fear of a litigation mess (if my case is taking 2 years--and concerning a mere 'hundreds of millions of dollars'--just think how long the bailout of Bear will take). That creates good incentives for equity owners.

No comments: