So, let’s take a look - with the disclaimer that I’m a macro strategist, not a risk analyst.

When Wall Street talks about risk, it’s usually talking about volatility. That makes a lot of sense: The wider the range of possible outcomes for a given investment, the bigger the potential losses - as well as the potential gains. The first problem is that no measure of volatility can capture all possible outcomes; the second is that many measures of volatility used by Wall Street definitely don’t - instead, they overweight downside movements and underweight equally large movements to the upside.

Take, for instance, the best-known measure of volatility: The VIX, popularly known as the “fear index.” The VIX is based on the pricing of future options on the S&P 500. If traders are willing to pay more to insure against movements in the market, that means they expect more volatility in the future. The catch: the VIX tends to rise only when markets go down, and to drop when markets are going up.

This too makes a superficial kind of sense: Not many traders are going to worry about trouble ahead if the markets are rising, even if indexes jump by leaps and bounds.

So it was in August 2007: When the first major whiff of panic hit the markets, the VIX soared from the teens into the high 30s. But when market indexes recovered to hit new highs after the government’s first palliative measures, the VIX dropped back into the teens.

In retrospect, the volatility index was signaling smooth sailing ahead when the real risk of future losses was nearing its peak (the maximum pain has been inflicted on those who bought when the market hit its all time high in October 2007). Thanks, VIX!

What good is an index that tells you of danger ahead only after you’ve hit the iceberg? Or, to put it another way, what good is a “fear index” that’s blind to emerging bubbles, when fear is most needed?

To be sure, VaR metrics gauge volatility in a variety of ways, many by looking back at historical data. Again, though, embedded optimism confounds the data. For instance, predictions based on past variances will underestimate future risk if the data pool is drawn from periods of calm.