Tuesday, April 22, 2008

Economic risk models failed a lot of banks

The credit crunch occurred because a lot of banks, both commercial and investment banks, failed to recognize the risk they were taking when the loaned cheap money to mortgage lenders. They were depending on computerized quantitative risk analysis to prevent that kind of mistake.

For an idea of what went wrong, here is a brief report from NPR. The Brownian motion described in the report means that the motion of tiny pollen grains placed in water is completely random. The direction of movement of the pollen grain is unpredictable. It has no memory of the previous change of direction, so the next move is as likely to be left as right.

That's been found to be true of market prices for stocks and bonds also. In the very short term of the next transaction there is no way to tell if the price will go up or down. The direction of the nest transaction price is completely random. Is it predictable in any way? The risk model says "No" because all biases are known to the market makers and are already reflected in the price of the security. This is the thesis of Burton Malkiel's influential book "A Random Walk Down Wall Street."

It was the assumption that all biases were built into the price that failed. Either the market biases were concealed from market makers, or the market makers ignored the indications of market bias.

The ways the biases were concealed were numerous. Rating agencies that were expected to provide accurate ratings for securities were chosen by how good a rating they would give the security, and they only got paid if they won the bidding war for how high they would rate the security. Mortgage brokers - again, only paid if they successfully sold a mortgage, were allowed by lack of regulation, oversight and lack of transparency to fake documents sent to the banks, who, similarly, were ignoring good underwriting practices when they loaned the money. It didn't matter to either the mortgage broker or the banker how good the loan was because they were going to sell it to investors concealed in large securities that included hundreds of mortgages.

These practices were ignored by the market makers because they assumed that the rating agencies were doing their job, and traditionally banks had not loaned money to people who could not pay because they would lose money if the loan defaulted. The large securities normally have a buy-back provision so that the bank will have to buy back from the investor loans that default. As long as only a few default, this is not a problem. But if a lot default at once, the bank can't afford to buy the defaulted loans back.

Since the risk models look at transactions individually and assume no overall bias, they did not anticipate the factors that were concealed by the mortgage brokers and lenders and ignored by bankers. It is a truism of the discipline of Risk Management that quantitative tools cannot predict and evaluate all risks. Those tools have to be used along with the judgment of human experts. Quantitative risk models, no matter how powerful, are always incomplete.

But the human experts who could have recognized the possibility that market will not always go up learned that fact when they experienced the recession of the early 80's. Since then, except for a few short market blips, markets have always gone up. No real surprise, since when they started to go down, the fed would loosen money and lower interest rates. [*]

The quantitative computerized risk management models have all been developed and adopted since the recession of the early 80's. The older experts do not understand the mathematics of the new models. Instead, they have been adopted and installed by a new generation of math- and computer-savvy whiz kids - whiz kids who do not remember the recession of the early 80's, and if anyone mentioned it to them would just shrug and assume that the federal reserve could prevent the market from going South on them. It always has.

As the NPR article pointed out, not all banks have been caught in the current credit crisis. In some the older hands did keep the decisions from involving too much risk. They added judgment to the conclusions the computer models offered. But there haven't been enough of those to prevent the current economic downturn.

It's in the banks that ignored good human judgment on risks that things went South. The lesson has been learned again. Quantitative risk models cannot be trusted without the judgment of experienced bankers. The bankers who keep their jobs during the next few years of economic difficulty will mostly be those who understand that lesson.

Of course, there is always a new generation of young mathematical whiz kids who think they know everything that matters waiting in the wings for the next time.

[*] It should be noted that the period since the recession of the early 80's has also been dominated by conservative free-market ideology and by the removal of the Depression era limitations on the risks big banks and big business could take. It has also seen the shift of risk from businesses to individual workers and the use of governments - Federal, state and local - to tax individuals and give the proceeds to the wealthy.

There has also been an increase in the financial exploitation of the middle class by the financial industry as is indicated by the orgy of issuing credit cards together with the banker-inspired restrictions on bankruptcy. The working class has similarly been hit with Pay Day Lenders who take advantage of the every greater uncertainty of employment as companies have made workers into the variable costs they cut when the market fails to go their way.

The current credit problems and recession are a repudiation of those conservative theories.

No comments: