Cold Blooded Risk
It finally happened.
Today, I got the Risk Talk.
By that I mean a senior person sat us down and dispensed his worldly wisdom about Risk and how it is managed:
- The only variable we care about is VaR.
- P&L losses in excess of VaR have to be explained, even those that statistically have to happen at least 2.5 times a year.
- Kurtosis and skew? What’s that? Expected shortfall? What planet are you from?
- Capital requirements for Bank trading operations are calculated on VaR.
- VaRs across asset classes are linearly additive.
- Regulators approve banks to use internal VaR models at their whim.
- Banks do these models with the singular purpose of reducing capital requirements.
- Banks deliberately smooth out the data used in the models and regulators let them.
- The people tasked with doing these models simply do not have the quantitative chops to make sense of what they are doing.
By the end of it I was reduced to a mixture of bemusement and directionless anger. Do these people take this seriously or what? Do they take themselves seriously? Is this all just a big fat joke?
One quote kept replaying in my mind throughout the farce. It was from The Quants:
At Deutsche Bank, risk wasn’t fucking managed. Risk was bitch-slapped, risk was tamed and told what to do.
It felt like ECB’s Inflation Monster, kept all bottled up by wonderful bureaucratic wizardry:
ECB http://www.youtube.com/watch?v=tLYfsN1Uc_E
Everything was so textbook, I didn’t know where to start. But starting by giving the benefit of doubt is usually the most socially desirable option. According to this line of thinking: you know and I know that these things are flawed, but they are required of us, so let’s quit whining and do them anyway.
Because regulators are the closest things to Immovable Objects we have in finance, that assumption alone just about ruled out my objections to all but one of the bullet points above – the last one.
In particular, the bit about the smoothing.
See, in twisted Regulator Logic, “breaches”, or losses in excess of VaR, are a bigger deal than not having enough capital to withstand them. (If I sound light-hearted, I am not – this is wrong, this is twisted, this is EVIL and anyone who has a choice and yet takes part in it is a FRAUD.) Breaches need to be analyzed, explained away, within a 48 hour period of their occurrence. Bank internal models and the capital needed for them are examined once a quarter, if that.
And yet, Bank internal models are smoothed. I don’t know how to express how idiotic this whole endeavor is. Let’s put it this way: how would you like me to estimate your chances of dying in a horrible car crash by first assuming that most horrible car crashes in history were just flukes?
“Oh, those? Those happened to people who deserved it. The bad stuff doesn’t happen to good people like us.”
Yes: I get that it serves the bank to use minimal capital and therefore to understate VaR. But it is also true that realtime P&Ls are mark to market P&L’s, entirely unsmoothed.
Remember Regulator Logic: Breaches are worse for the bank’s risk managers than having too little capital.
Under this perverse surreality, I reasoned that this must happen:
- Banks smooth data used in estimating VaR models
- Regulators approve these internal models
- Banks use the models
- Banks suffer mark to market losses, causing Breaches
- Banks suffer more Breaches than Regulators allow
- Banks lose approval for the models
And when I pointed this out, all I got was flat denial and condescension.
- I wasn’t familiar with the data
- The guy had been doing this for 5 years
- There was no way to know we would get more breaches than allowed unless we actually tried it
Even my junior analyst (no, it doesn’t get old), who was also there for The Talk, understood my reasoning faster than the guy did. I agree it is not a fair comparison given she was an actuary and I am doing a master’s. But there we were, getting The Talk all the same.
If there is one thing I detest more than government perversion of logic, it is human denial of reason and reality.
the title is a reference to Aaron Brown’s new book, is getting great reviews from people I respect.
Comments are closed.
.
