The riskiest year in human history was 1962. The year of the Cuban missile crisis, the closest we ever came to a nuclear war. The mother of all tail events, where all prices go to zero. Volatility that year was average — 16.5%
How can market risk be average when tail risk is at its highest?
The following figure shows the annualized S&P 500 volatility from 1928 until today, along with the median, the quarter smallest and the 10% smallest volatility. 1962, is highlighted in red along with the two runners-up.
Volatility in 1962 was average and slightly above the median. The following year Kennedy was assassinated, and then volatility was at its 10% low. The following year, the Vietnam proxy war between the Soviet Union, China and the United States was really heating up, while volatility was at its historically low.
Tail risk in those three years was extremely high, while market risk was very low. How can that be? Elementary. Risk is not a unified concept.
Volatility is the risk of a short-term and frequent change in prices, while a nuclear war resulting from the Cuban missile crisis would be an extreme and once off event. Volatility isn't simply designed to capture such risk. And neither are most other market risk measures, like CDS spreads, VaR, ER, SRISK, CoVaR, and the like.
Market data, and hence market risk measurements, are most informative about the short run and innocuous. High-frequency small events.
As time passes, day-to-day risk market ceases to be relevant. Fist, institution idiosyncratic risk is what matters, and then systemic risk. As time horizons become longer, the macroeconomy is increasingly driving risk, only to be replaced by politics at the very longest horizons.
The most damaging financial crises are invariably caused by politics or the macroeconomy, not by excessive amounts of financial risk.
The relationship between the drivers of risk, time and what matters is shown in the following figure.
What we care about should determine where we look for risk. Day-to-day market risk, like volatility, is usually what matters for the trading floor and microprudential regulations. If we care about solvency of individual institutions, or financial crises, the concern of the macroprudential regulators, market risk is irrelevant.
That is not how most risk methodologies see it. While the drivers of long-term risk are predominantly political, almost all risk models only measure short-term risk.
Why is that? Because it is much easier to measure. All we need is market data and a model, and voilà, we have a risk forecast.
We, therefore, are in the perverse or amusing situation where what we spend most of our time measuring is what matters the least, and what we care most about is unmeasured.
More formally, when we measure risk, we get perceived risk, while we care most about actual risk.
It reminds me of the old joke about the policeman who encounters a drunk man crawling on four legs under a lamp post. The policeman asks what are you doing. The drunk responds by saying he's looking for his keys. The policeman asks why are you looking there. The drunk responds "because that's where the light is."
I was in a risk conference recently where a number of participants maintained that systemic risk was increasing, pointing to a large number of reasons why. I disagree.
Is long-term systemic risk is going up or down? Taking a historical perspective, there are many periods in history where systemic risk was much higher than today. 1962 is one, but there are many others, as you can see on my website extremerisk.org that links significant events to market risk.
So, much of our efforts to contain systemic risk are wasted. Even worse, all the metrics and dashboards might give us the illusion that everything is under control. They got it wrong in 2007 and also today.