Relying on highly-paid experts working from billion-dollar models, I went out to walk the dog this morning without a raincoat. Storms would not, I was told, hit D.C. until mid-morning, and it was only 6:30. Of course, we got wet – which brings me to the Basel expected-shortfall model for setting risk-based capital for big-bank trading books. Like the value-at-risk model it is to supplant, expected-shortfall is the combined wisdom of finance Ph.D.s and quants across the globe. Will it keep banking dry in the next systemic storm? I too like to tote the latest-model umbrella, but fancy though this one is, I still fear it won’t work.

I will spare all of us the low-down on expected shortfall (ES) for now – clients, watch your inbox for FedFin’s in-depth analysis of Basel’s new consultative paper. Suffice it to say that ES is a new formula for trading-book risk drawn up to remedy the flaws laid bare in value-at-risk (VaR) by the financial crisis. We needn’t relive them now either, nor describe anew the London-Whale case that has become the VaR poster-child for ES advocates. In short, ES is supposedly better at calculating tail risk – that is, how far markets can really fall under stress. ES is also supposed to take the fun out of arbitraging between the trading and banking books, as well as to capture the type of trading liquidity risk that contributed to so much crisis-era grief. And, for good measure, it whacks securitizations in the trading book, continuing the regulatory jihad against secondary markets also seen as sowing the seeds of systemic risk.

So, what’s wrong with ES? Perhaps nothing as a new and improved model of how much money traders can lose if markets behave as expected – the name says it all. The problem isn’t ES, nor indeed was it with VAR for all its long-obvious flaws. The real issue is that regulatory capital isn’t supposed to absorb expected loss, but rather handle the hits that drain expected-loss set-asides like reserves and other rainy-day funds. The difference between expected and unexpected loss is well understood in the banking book – take, for example, the leverage rule, where capital is hugely higher than any expected loss for holdings like U.S. Treasuries. But, when we get around to the trading book, we forget this and model on into the abyss of predicting just how wet we’ll get when the rains come in as planned.

I have no quibble with setting more capital for bank trading books. After all, traders at banks play with other people’s money. Whether it’s VaR or ES, elaborate models are very useful ways to set boundaries around how much traders can put on the table as the market’s wheels spin round. They are also very effective ways for senior management and boards of directors to assess likely risk exposures.

But, are these models really the risk tolerances by which banks should be run by their boards and judged by their supervisors? Of course not. The problem with VaR, just like Basel II’s fancy models – isn’t that any of them are wrong as starting points for judgment of how risky a bank may be. I wasn’t wrong this morning in taking heed of the meteorologist’s forecast of when it would rain – this was a reasonable assumption gained by reliance on decades of research with a high probability of success. But, had I looked out the window – or, maybe better, listened to my dog, who knew better than to go out – I would have at least brought along the slicker and, thus finished my walk.

If the models don’t look right, they aren’t right. There is no way that all of the risk JPMorgan took on the London Whales’ billion-dollar book for highly-correlated, very-complex structured synthetic bets was basis points of the total book. It just can’t be even if all the models say it is, and it for surely isn’t right when the first drops of rain start to fall even if traders insist they’re still on the sunny side of the street.

So, ES may be better than VaR in many ways, but it’s just as dangerous because it’s at least as complex. Indeed, it may be even more dangerous because, chastened by VaR, regulators now want their new toy to drive capital judgments across the industry and around the globe. We will have converted trading risk at individual banks into models risk across the entire industry. If ES is always right, banking may be safer; if it isn’t, we will get very, very wet because we all will be out without an umbrella the next time it starts to pour.