The downside to simpler bank capital rules?

While Zelmer acknowledges that models are far from perfect. He nevertheless defends their role in setting capital ratios, arguing that the criticism “has been overly simplistic and rather narrow.”

“It has focused on the rules themselves without considering the repercussions for bank risk management and banking supervision more generally,” he says. “The time has come to broaden the debate.”

Indeed, he maintains that models must have some role in setting capital requirements. “There is no getting around the fact that discussions of bank capital requirements need to be grounded in the information provided by well-designed and properly used risk models.

Simplifying the current rules would not eliminate the need for models, he says. “Instead, it would simply relegate them to the private conversations between banks and their front-line supervisors, resulting in higher costs to institutions and more face time with their supervisors,” he says.

Rather, he argues that there needs to be an effort to restore confidence in the models. “More needs to be done if we are to re-establish investor confidence in model-driven risk weights,” he says, noting that improved disclosure is one way to bolster that confidence.

To that end, he says that OSFI is requiring the Big Six banks to adopt the recommendations of the Enhanced Disclosure Task Force (EDTF) regarding their risk exposures and risk management practices, including their modelling practices. “This should help users understand how bank risk models are influenced by data inputs, modelling assumptions, mathematical formulations, manual overrides and point-in-time versus through-the-cycle assumptions. We also expect to see more explanations as to how banks assess their model performance, including how credit risk models perform relative to actual default and loss experience,” he says.

Zelmer notes that many of the EDTF’s recommendations have already been adopted by the Big Six, and that the banks are expected to implement the rest of them over this year and next year.

“Another way of restoring confidence in the markets is by showing more restraint around what models can accomplish and applying some safeguards, including tighter calibration standards,” he says, adding, “…greater use of floors on calculations of specific elements of capital may be a helpful antidote against acute model uncertainty or intentional and unintentional gaming.”

He also suggests that it’s worth considering tying some model-based capital calculations more tightly to the calculations from the standardized capital framework; particularly for exposures that appear to carry very low risk due to the fact that such risk events are rare.

“It is difficult to judge whether such exposures are truly low risk, or there were few adverse events in the data sample used to run the model,” he notes; which means that these types of risks are not well accounted for in the typical models-based approach.

Zelmer maintains that these sorts of responses are preferable to more fundamental reforms that would put a greater reliance on regulators.