What should be in the fundamental review of the trading book? Part 3: something consistent March 26, 2012 at 5:30 pm
This is another in my continuing series on the upcoming Basel fundamental review of the trading book.
Today I want to talk about models, and consistency. Now while I have a lot of sympathy with A foolish consistency is the hobgoblin of little minds, it has to be admitted that perceived lack of consistency between different banks’ capital models has caused a lot of comment which has damaged the reputation of the regulatory process. I say ‘perceived’ because few people really understand the materiality of the differences between different banks’ models. That has not stopped commentators from Jamie Dimon to Andy Haldane suggesting that there is too much room for banks to reduce capital by changing their models, and too little supervisory consistency, though, and I can see their point.
All the same, there’s no getting away from models. Only a model can hope to measure risk accurately. Standard rules – like capital = 8% of notional – are just bad models.
So what should we do?
The answer is that the fundamental review of the trading book should introduce a benchmark portfolio regime. All systemically important banks should be required to calculate capital using their own models on a number of standard portfolios every quarter. These portfolios would be updated regularly. Bank results would be compared, and those that were low would have higher capital multipliers imposed. Anyone who was too far out would not be permitted to use their model for that asset class at all. The process could be run by the financial stability institute in Basel, and results should be public. That would certainly enhance confidence in the financial system.
(The cynical among you might note it would also generate wonderful opportunities for folks like, ahem, me, to help banks ‘improve’ their models before the process started. That might be true. But just because something would be a consulting goldmine doesn’t make it bad. Does it?)