Monday 2017-01-23

Not unsurprisingly after AveragingDown Bronte Capital has a follow-up post asking risk managers what they would have done given an example real portfolio from March 2008.

Culling from the comments:

a risk manager or analyst might like to sit the manager down in the proverbial quiet room, and ask just exactly what was the macro and micro rationale behind making these big sector bets.
-- January 12, 2017 at 1:15 AM
I also cannot imagine how the March 08 return sensitivities for credible operating scenarios would be so strongly skewed to the upside (say >10% IRR) to justify such concentration. Any sensible analysis (in early 2008) would allow for a high probability of a sharp recession in which case a lot of the potential scenarios would look pretty bad for these companies on a 1-3 year view.
-- January 12, 2017 at 11:18 PM

These two comments are presumably why Hempton says "the trader is smarter than the risk manager", i.e. the trader can always game the model of reality the risk manager has, and tell a story about the portfolio that the risk manager will agree to.

It seems that Hempton is asking "Is there a risk management system that an Evil Sabotage Trader cannot subvert?" If so, then he wants it because he knows that one day he might accidentally construct an Evil Sabotage portfolio.

However, the answer is No; we haven't found a fool-proof system yet. Which is why risk management requires a backup plan:

I would expect that the evolution of the VaR during 2007 and 2008 as correlations began to converge would have sent off serious warning signals, but I would have to simulate the portfolio at stock level to demonstrate that.
-- January 11, 2017 at 12:07 AM

This view can be improved by making it execution oriented; i.e. the trader gets a VaR budget, and for each position they need to provide a stop-loss execution program in advance. There's no magic avoidance of pain, just planned exits that limit losses to predetermined amounts.

Ultimately, we try to reduce the odds of accidental Evil by working with a risk manager to create additional thesis tests tied to price, events, etc. However, we can't reduce them to zero, and in those cases when the fit hits the shan, we're happy that we already have the exit planned.