As promised, here is a constructive approach to managing risk. I start with the least trustworthy (but still useful) metrics and move down to the most trustworthy.
4. Current market information - for most asset classes, we can look to the collective wisdom of the market for the short-term risks. For example, the current prices of options reveal the market's consensus on the probability distribution of equities over the next few months. Most of the time the market is very accurate and does a fine job of predicting the day to day volatility of a portfolio.
3. Over a very long time frame, historical data shows us what is likely to happen in similar circumstances and the great extremes that are possible. For example, we should note that in previous recessions, P/E ratios frequently dropped below 12; given current estimates for next year's corporate earnings, a P/E ratio of 12 would entail the S&P 500 dropping another 35% (from 840 to 550). Contrary to current market information, historical data suggests that risks are highest at the peak of booms when implied volatility is lowest.
2. Consider "game changers." Any pricing of risk makes assumptions about how the future will resemble the past. What assumptions could be proven wrong? For example, credit default swap contracts priced in almost no risk of the issuer defaulting, and that changed overnight. Must US treasuries sell off in an inflationary environment or is that just consensus? In investing, consensus frequently ends up being wrong, current market pricing of volatility changes, and historical patterns continue until they don't. We need to consider that the risks themselves may change.
1. Absolute exposure - ultimately the only metric of risk in which we can have full confidence is the total dollar exposure. A leveraged portfolio will always have a greater risk of going to 0 than a portfolio with 20% in cash. As leverage increases, so does the exposure to mistaken risk analysis. A 20x leveraged portfolio can only weather a slight miscalculation by management. No management is perfect, we all make mistakes despite our best efforts. An unleveraged (or only modestly leveraged) portfolio allows for a necessary margin of error.
You'll notice I ommitted recent historical data like "beta." I think focusing on concepts like "beta" and "alpha" are more misleading than helpful. As Alpha noted, big losses usually come from uncertainty not risk. As risk managers, we should focus on constantly seeking out sources of uncertainty. We need to construct a portfolio that can tolerate our inevitable misjudgements and the unpredictable "black swan" events that will test us.