U.S. banking regulators recently released a Notice of Proposed Rulemaking (NPR) that would make changes to regulatory capital requirements. The proposed changes, known as the Basel III Endgame, are expected to increase capital requirements for large banks. The proposal runs 1,087 pages and covers multiple areas. One proposed change would revise the capital treatment of residential mortgages. The banking industry has predictably reacted with wailing, gnashing of teeth, and doomsday scenarios. However, it’s important to put the proposal in its proper context. How are residential mortgages treated under current rules? Do the current rules accurately reflect risk? What would change under the proposal?
The Current Environment
The largest and most systemically important U.S. banks are subject to two sets of capital approaches. The Standardized Approach follows the basic risk weighting approach that has existed since 1990, which assigns classes of assets a risk weight, roughly corresponding to their perceived credit risk. Although 100% is supposed to be the standard risk weighting, most banks have risk-weighted assets (RWAs) well below the amount of assets reported on their balance sheets. The largest and most complex banks also need to calculate risk weighted assets under the Advanced Approach. Banks use internal models, subject to certain parameters established by regulators, to calculate capital charges for credit risk, operational risk, and the market risk of trading activity.
The Collins Amendment required AA banks to calculate RWA under both Standardized and Advanced approaches, with capital requirements based on the stricter approach. For most banks, the Standardized Approach is the binding constraint. That, if anything, understates how favorable internal models are to credit risk compared with standardized risk weights. Credit risk charges are consistently lower under the AA, partially offset by operational risk charges, which don’t apply under the Standardized Approach.
Consider the following comparison between credit risk weighting under Standardized and Advanced approaches for the largest banks (10-Q data as of March 2023):
Internal models are supposed to be more risk-sensitive and more reflective of an individual bank’s credit risk. Consistently lower risk weights under the Advanced Approach suggest that if left to their own devices, banks will calculate lower capital requirements for themselves. The difference is even more extreme for residential mortgages. A review of March 2023 FFIEC 101 data for AA banks shows the following effective risk weights for residential mortgages under the Advanced Approach:
For reference, under the Standardized Approach, the risk weight for “prudently underwritten” residential mortgages is 50% and is 100% for other residential mortgages. For FNMA and FHLMC guaranteed MBS, (which have never experienced a credit loss), the risk weight is 20%.
How did they come up with those numbers? The Advanced Approach boils down to three elements: exposure at default (EAD), probability of default (PD), and loss given default (LGD). These three parameters are then plugged into a supervisory formula as follows:
It’s easy to see how this approach might go off the rails. The formula is supposed to forecast a thousand-year event (0.999) with perhaps twenty years of data. The weighted average AA RWA for residential mortgages is also nearly ten percentage points lower than in Q1 2018, suggesting that the impact of the 2008 Financial Crisis on these estimates has sharply diminished over the past few years. I suppose this time is different. The supervisory formula also assumes that mortgage losses are relatively uncorrelated (R = 0.15), though the breakdown in prudent mortgage underwriting practices prior to 2008 suggests that something systematic was going on.
Each bank’s approach is subject to an initial and ongoing supervisory non-objection requirement. Unfortunately, supervisors didn’t have a lot to go on. It’s one thing to be skeptical that the bank’s PD and LGD assumptions accurately reflect risk. It’s quite another to prove the bank is wrong. That might require another financial crisis.
These RWA estimates are out of line relative to risk weights under other regimes and to recent historical experience. Delinquencies on single family residential mortgages peaked at 11.48% in Q1 2010. Loss given default during the Financial Crisis averaged about 35% or more. That implies a stressed loss of about 4.00%, which corresponds to a 50% risk weight. The Collins Amendment applies to capital requirements under Standardized and Advanced approaches in aggregate. Low AA risk weights for mortgages did not only benefit those where the AA was the binding constraint. An operational risk charge only applies under the Advance Approach. The relatively low RWA for credit more than offset capital requirements for operational risk, which meant an effective operational risk capital charge of zero for most systemically important banks.
Proposed Changes
The NPR would ditch models-based approaches for both credit risk and operational risk. At the same time, it introduces an operational risk component to the Standardized Approach and makes the SA more risk-sensitive for residential mortgages. The proposal would tie risk weights for residential mortgage to loan-to-value (LTV) ratios at origination as shown below:
This approach is more granular but not necessarily more conservative than the current Standardized Approach. The current definition of “prudently underwritten” is squishy but essentially assumes an LTV of 90% or less. The NPR reduces that threshold to 80%, but allows lower risk weights for loans with LTVs of 60% or less. It also reduces risk weights below 100% for loans with LTVs above 90% and even for LTVs above 100%, where the banks are practically begging borrowers to hand them the keys. A separate, higher set of risk weights apply to first mortgages on income-generating properties.
Higher risk weights would generally apply for mortgages in the 80-90% LTV range. It’s useful to place the new 80% threshold in perspective. GSEs continue to use an 80% LTV threshold for conforming mortgages. This threshold also applied to risk-based capital requirements for federally regulated thrifts until 2002, when it was raised to 90%. The disastrous performance of mortgages during the Great Recession might cause reconsideration the 90% LTV threshold, but it didn’t.
The capital proposal does not treat private mortgage insurance as a risk mitigant for higher LTV mortgages. Both the GSE conforming mortgage and the prudent underwriting definitions recognize PMI. Although there is evidence that PMI mitigated credit losses at GSEs, it also exposes the financial system to tremendous systemic risk. Insurance tends to work best for small, idiosyncratic events. Mortgage defaults and losses tend to come in bunches. Only six firms offer PMI and two other mortgage insurers failed during the Great Recession.
Some worry that higher risk weights for mortgages will make home financing less available to first-time home buyers and lower income borrowers. Pages 71-73 of the NPR discuss these concerns and express a willingness to consider alternative approaches. I’m also for promoting home ownership but regulators should base risk-based capital requirements on … risk! Allowing too big to fail banks to leverage based on understated risk figures doesn’t make a lot of sense.