I did not wish to imply that it was the common use of the Barra or any other risk model that got equity quants into trouble in August 2007. What got quants into trouble in August 2007 was the mindset of the only active risk that should be taken is stock selection, or residual risk, and that we should control for industry risk and common factor risk such as size, style, etc.
Dan diBartolomeo of Northfield wrote (italics are mine):
…it is common practice to measure risk using the assumptions of uncorrelated residuals, while using stock selection strategies that can only work if equation (1) does not hold (no factor bets). Since active mangers will be concentrating their portfolios in those securities expected to act alike in providing superior returns, the average residual covariance will be positive…Hence, if we use a risk model which assumes uncorrelated residuals we will have a downward biased estimate of the risks not identified by the factors of our risk model.
He went on to say:
One should also be particularly careful with multi-factor selection strategies where some but not all of the selection factors are in common with the risk model. In such cases, the risk model will tend to neutralize exposure to the risk factors, which will lead to the selection of a security set which are weak on the selection factors which are in common with the risk model, while having extremely high exposure to the selection factors which are presumed independent of the risk model. As such, it is easy for a multi-factor selection strategy to become dependent on a single component, defeating the purpose of the multifactor construct.(I am grateful to George Wolfe for pointing this out to me.)
If the “stock selection alpha is the only acceptable source of alpha” mindset is becoming a crowded trade, then it’s time to move outside of that box. Goldman Sachs published a research report entitled “A Stockpicker’s Reality: Part III, Sector strategies for maximizing returns to stockpicking” in 2002 where they “examine the degree to which risk management strategies can be used to increase returns. This question differs substantially from the normal application of risk control, which is focused on tracking error rather than returns.” Their conclusion was:
- Value-driven methods are most compatible with quantitative risk management of benchmark-driven portfolios.
- Growth-driven methods are far less compatible with strict quantitative risk limits and are more effective in relatively more concentrated, less risk-controlled portfolio construction applications where risk management is handled at the asset allocation level by diversifying across managers.
In other words, different strokes for different folks. (If you would like a copy of that Goldman Sachs research report please email me. )