Menu

Risk Weighted

Should Internal Bank Models Be Used for Capital?

Large U.S. banks may be forced to employ standardized, regulator-driven models to calculate capital in the future. But, for a variety of reasons, forcing these institutions to move away from internal methodologies for loss projections may not be advisable.

Friday, September 22, 2023

By Tony Hughes

Advertisement

A few weeks back, U.S. regulators proposed a suite of new rules designed to strengthen capital requirements for large banks. The most consequential proposal was a dramatic scaling back of banks’ ability to use internal models to calculate capital charges. In its place, banks will be required to use standardized risk measures determined by models controlled by regulators. 

Unsurprisingly, the proposal was heavily criticized by the bankers who will be most impacted. JPMorgan Chase CEO Jamie Dimon was especially scathing. "If they want to put all mortgages and small businesses loans out of the banking system, so be it, but they should tell that to the American public. It will have a real effect on consumers," he warned, elaborating that the Fed should be “cautious” about their own models, which in his estimation are “not perfect.”

So, who's right? 

We will start with the ridiculously ingenuous assumption that the proposal is based only on modeling sensibilities. Will the regulator be able to produce more accurate estimates of expected loss than the banks can on their own? To consider this question, we need to further assume that banks and regulators are actually motivated to produce accurate predictions and are not trying to game the system for their own nefarious purposes.

You can easily see why these assumptions are naive.

Tale of the Tape: Regulators vs. Big Banks 

The advantage the regulators hold is that they can draw relevant data from across the entire banking system. In simple statistical terms, they can use a much larger database to build their models, which should enable them to produce more precise model estimates than individual banks. Having access to a deeper dataset is especially important when considering behavior in the critical tails of the distribution, which can be sparsely populated even for large banks.

What’s more, one can argue that regulators are better placed to develop accurate models when, say, mortgages or small business loans are homogeneous in nature.

Bankers will riposte this contention by pointing out that loans are not homogenous – and that banks seek to find competitive advantage through their ability to identify, market to and manage underserved borrowers.

Suppose, for example, that there is a small business with a blemished credit history – but which is now profitable and likely to remain so. One bank may focus on the blemishes and reject the application; a rival bank, on other hand, may look beyond past behavior, correctly identifying the viability of the business.

tony-hughesTony Hughes

Bankers using internal models will be incentivized to capture the proprietary secret sauce in their capital calculations; regulators will not. So, if capital charges are based on the regulator’s model, and if this model is blind to the virtues that the bankers have identified, capital costs will be set too high and the example loan may be scuppered.

Here's another scenario: suppose that there are two banks, both of which manage large small business loan portfolios. Assume that one bank not only has developed sophisticated and accurate scoring models capable of identifying underserved populations but also has a dedicated and professional program in place to help manage troubled loans. The second bank, in contrast, takes a lackadaisical approach to all these issues.

For a given set of basic borrower characteristics, we would expect the first bank to experience lower credit losses than the second. In such circumstances, the regulator’s model will tend to overstate the risk faced by the more attentive firm and understate the risk posed by their more phlegmatic rival. For banks that deviate from an industry average level of performance, the model’s errors will have a nonzero mean .  

Regulator models, in short, will be misspecified, because they will fail to account for heterogeneity in the composition and treatment of borrowers across the industry. Nonetheless, should the proposal that’s now on the table be implemented, regulators could use their models to identify banks that perform consistently better than average, and offer associated capital relief to these companies.

If regulators do not recognize quality in this way, bankers will have no incentive to find hidden gems, and some of the dire outcomes suggested by Mr. Dimon will become more likely. 

Barriers to Entry

One of the major problems with the IRB approach is that it puts significant barriers in the path of smaller banks. To qualify to use internal models, banks must have (1) access to a large dataset that enables them to build accurate representations of expected loss; (2)  talented, experienced modeling teams in place that can produce models capable of passing a rigorous examination process; and (3) the infrastructure necessary to implement the models and monitor their ongoing performance. 

I’m aware of many long-established midsized banks that have struggled to navigate this process. Small banks, moreover, can’t even make it to first base – if you don't have enough data, the IRB project is a nonstarter.

These smaller institutions are required to use standardized capital charges (which are often ridiculously high), while big banks can currently pick and choose, using internal models whenever it suits them. This provides an enormous competitive advantage for larger banks.

In the spirit of “too big to fail,” regulators should be looking for ways to smooth the path for smaller banks. Allowing them to tap into the same regulator-owned models as larger banks would be a great way to spread risk more effectively across the industry. 

Parting Thoughts

At the end of the day, of course, proposals like this have little to do with optimal modeling principles. IRB-qualified large banks have a strong incentive to game internal models to reduce associated capital charges. So, a talented modeler will often propose scores of plausible models capable of passing validation exercises.  

Bankers will always favor the specifications that produce the most optimistic forecasts. And regulators will always try to circumvent these practices.

Your views on whether regulators are too heavy-handed with capital rules will depend on how you balance unfettered commerce with the threat of banking system failure. Currently, to me, the balance looks reasonable, which makes the recent regulatory proposal seem rather austere and extreme.

Regulators' end game is focused on curbing big banks rather than spurring small banks.

If the regulators’ proposal had included opening regulator models to all banks, it would have been much easier to support, because it would have reduced the value of IRB by scaling back the advantage it bestows on bigger banks.

As it's currently constructed, it looks like a straightforward and unnecessary capital pile-on.

 

Tony Hughes is an expert risk modeler. He has more than 20 years of experience as a senior risk professional in North America, Europe and Australia, specializing in model risk management, model build/validation and quantitative climate risk solutions.




Advertisement

We are a not-for-profit organization and the leading globally recognized membership association for risk managers.

weChat QR code.
red QR code.

BylawsCode of ConductPrivacy NoticeTerms of Use © 2024 Global Association of Risk Professionals