Yesterday I was discussing some risk analysis with a colleague. Specifically, quantitative risk scores for fixed income funds. The details of the scoring are not important for the purposes of this post. Suffice it to say it is along the lines of sorting funds into quartiles based on statistics such as rolling volatility and drawdown.
The point of our discussion was that soon the financial crisis period will roll off the risk scores, penalizing more conservative portfolios in the ranking system. The scores will implicitly reward excessive risk-taking.
This is a great metaphor for the state of markets today.
Collectively, we have forgotten what it means to be afraid. Today, it is all about squeezing as much return as possible out of a portfolio. Fear has rolled off our collective memory. And what’s worse our lack of fear is justifiable according to the trailing 10-year data.
This at a time when:
- Credit spreads are tight.
- Covenants are weak.
- Leverage is high.
- Oh, and interest rates are rising.
There is an inherent tension in risk management between simple statistical measures (which people prefer) and the true nature of risk (which is nuanced and difficult to quantify). In fixed income in particular, the payoffs are negatively skewed. As an extreme example: “I have a 94% chance of earning a 6% yield on my $100 principal investment and a 6% chance of losing the $100 of principal.” Only in real life we don’t know the probability of default in advance.
The standard deviation of a high yield bond fund does not do a great job of describing its risk. In the absence of defaults the volatility will be fairly mild. If defaults tick up in a recession, losses could be catastrophic–particularly if liquidity dries up and twitchy investors decide to redeem en masse. None of the most significant risks to a high yield investment are properly captured by its standard deviation.
But people do not want to talk about conditional probability and expected losses given default and the uncomfortable fact that their financial lives are non-ergodic.
People want simple, black and white answers.
Statistical risk analysis is popular because it uses straightforward inputs and is easy to run at scale. Looking at a portfolio and puzzling out how it might behave in future states of the world, without relying on correlation and volatility statistics, takes a lot of time and energy. It is a “squishy” process. Your peers might think you are a bit of a crank because you aren’t sufficiently “data-driven.”
Oh, and much of the time things will run smoothly anyway.
Diligent risk management is a thankless task. No one pats you on the back for the things that didn’t go wrong. In fact, in a market environment like this one, a little extra prudence can get you fired.
This is why cycles happen. People forget that a little fear is healthy. Or, more precisely, the market environment conditions people to invest more aggressively. They overreach (their backward-looking risk analyses encourage it!) Then when the cycle rolls over they get slaughtered.
Whenever I am looking at an investment one of the things I think long and hard about is under what conditions it might explode spectacularly like a dying star. Excluding fraud, these kinds of blowups are generally caused by leverage (too much debt or financial derivatives with embedded leverage a.k.a convexity) and asset/liability mismatches.
It should be fairly straightforward for a competent analyst to identify and control these risks. More importantly, as analysts we should be getting these things right more often than not as they are triggers for catastrophically bad outcomes. What’s more, none of them is captured by backward-looking statistical measures.