, Editorial Director
The subject of risk seemed so simple in the early '70s, when beta was becoming accepted. Then it seemed easy to control investment risk -- just control the beta of the portfolio. And Merrill Lynch Pierce Fenner & Smith's consulting arm and O'Brien Associates were both selling services allowing institutions to calculate the betas of their portfolios. But it wasn't long before the concept of risk got more complicated. And as Jack Gray, quantitative research strategist at Grantham, Mayo, Van Otterloo & Co., Boston, told the attendees at the Association for Investment Management and Research annual conference in Orlando last month, "there is no agreed-upon universal definition of risk."
Because of the advent of concepts such as beta, quantitative concepts of risk dominate investor thinking, Mr. Gray said. And quantitative methods promise a scientific approach to investing. While beta and other quantitative concepts of risk are robust, he said, they are not complete.
One reason is that the human and organizational factors contributing to risk are difficult to control, and they greatly affect investments because of the ability to arbitrage. There is a feedback loop, Mr. Gray said. There are many other risks of which investors ought to be aware, including some that arise from efforts to use quantitative tools for investment insights, he said.
One of these is data mining -- the use of computers to examine masses of historical data in search of new insights. Data mining risks include seeing patterns where none exist, or revealing meaningless ones.
An example Mr. Gray provided was of research showing that between 1977 and 1997, bond yields and earnings and dividend yields were highly correlated. However, other research showed that between 1948 and 1968, the correlations between bond yields and earnings and dividend yields were highly negative. Over the whole period, 1948 through 1997, the correlation was essentially zero.
Other institutional risks Mr. Gray included:
* Uncritically challenged world views.
* Hubris -- a belief that since yesterday's risks have been tamed, all risks have been tamed.
* Moral hazard -- the concept that when a risk is insured against, the investor is likely to engage in risk taking.
* Model risk -- the risk that the model developed from observation is "very wrong."
* Simplicity risk -- the risk that the model is too simple.
* Complexity risk -- the risk of breaking the problem into too many small pieces for the manager to understand what's going on, e.g., having too many managers or too many portfolios each divided into too many segments.
The ultimate and non-quantifiable form of risk control might be challenging, informed, critical, open, ongoing debate, Mr. Gray said.
Unfortunately, few institutions, whether in politics or investment management, welcome informed, critical debate. Few institutions are willing to be open enough to give those not invested in the process enough information for informed debate. And too few chief executives or chief investment officers are willing to expose themselves to challenging and critical debate. Perhaps these institutional weaknesses were the core of the Long Term Capital Management disaster. And probably there will be other such collapses because there cannot be informed, critical, open debate about black-box products.
It seems that the biggest risk is ignorance, though, Mr. Gray adds, too much knowledge might also be a source of great risk.