A tag-team of my favorite blogs-the wonderfully sardonic (and insightful) Rhymes With Cars and Girls and the less sardonic (but equally insightful) Deus ex Macchiato-discuss the fact that model-based risk-weighted capital calculations on the same assets differ substantially between models/banks.
From this, DEM concludes the need to centralize capital calculations:
he answer is clear. Centralize and standardise capital calculation. Throw all those internal models away, and use one common, regulator developed approach. Now that a lot of progress has been made on trade reporting, the data infrastructure exists to do this — or least it wouldn’t be too hard to extend what does exist to do it. Developing the models would be a huge undertaking, but compared to having each large bank do it individually, a central infrastructure would be cheaper and more reliable, and anyway you could pick the best of individual banks’ methodologies. You could even spin out the var teams from four or five leading banks into the new central body — just don’t pick the bank whose IRC is less than 10% of the average answer…
Crimson Reach at Rhymes is more cautious, repeating his theme that All Large Calculations are Wrong. He concludes:
In other words, it’s a Large Calculation. But as RWCG readers know, all large calculations are wrong. So, I’m skeptical that the answer to this is to ‘centralise’ the calculation – i.e., make it even larger.
I’m on Crimson’s side on this one. (Sorry, David.)
Centralize capital calculations? No. No. No. No. No fucking no.
Crimson is right: all Large Calculations Are Wrong. Including those-and I would suggest, especially those-undertaken by some Centralized Authority, e.g., The Wizard of Financial Oz.
What would you rather have? Randomly distributed mistakes, or perfectly correlated ones?
The answer, it seems to me, is gob-smackingly obvious. Randomly distributed. Perfectly correlated errors are a recipe for systemic risk. All banks will crowd the risks that the model underprices and avoid the risks that it overprices. If an adverse shock hits the crowded investment, every bank is screwed simultaneously. All will need to liquidate their positions in that asset, or in other assets. Every bank will have fire sales at the same time.
That’s what we want to avoid. That’s what creates a systemic risk.
We want to diversify model risk, not concentrate it. Centralized calculation concentrates it. I actually see an evolutionary benefit in the wide range of model risk weights that DEM bewails. That represents a diverse ecosystem of entities with divergent views that is less vulnerable to a single shock. Yeah, that shock will crater some banks, but not all of them. When I think of any-ANY-centralized calculation, I think of the Socialist Calculation Debate. I also think of monocultures that are dangerously vulnerable-systemically vulnerable-to a single shock. Think of the devastation that smallpox wreaked on native American populations. A single shock can crater everybody all at once.
If that happens, who is the buyer in the fire sale? If there are no buyers in the fire sale, how low can prices go? Pretty damn low.
Crimson is right. All Large Calculations Are Wrong. The answer to this problem is exactly the opposite that DEM suggests. Diversify this risk. Don’t concentrate it. Don’t sort on it.
We need less centralized control-freakishness. Not more. A lot less. Mistakes are inevitable. Create a system that induces them to make uncorrelated mistakes (or at least, mistakes with low correlation). Don’t induce perfect correlation in errors. I say again: don’t induce perfect correlation in errors. Create a diverse ecosystem, not a monoculture. Imposing uniformity because disagreement is unseemly and reveals the fundamental uncertainty with assessing risk is exactly-exactly-the wrong answer. It makes errors systematic. Systematic errors create systematic risk.
Diversified, diverse populations are far more robust than uniform, monocultural ones. Disagreement is a feature, not a bug. Enforced agreement is the worst bug of all. Sort of like smallpox.