On December 12, the Bank Policy Institute (BPI) organized a symposium on the regulatory capital implications of the Financial Accounting Standards Board’s new current expected credit loss (CECL) accounting standard for loan losses which becomes effective on January 1, 2020. The symposium covered four topics related to the regulatory capital impact of CECL on bank balance sheets: (i) the likely day-one impact; (ii) whether CECL will be more or less procyclical than the incurred loss methodology; (iii) potential adjustments to the regulatory capital framework to offset unintended consequences of CECL; and (iv) integration of CECL into the Federal Reserve’s stress tests. The symposium, which was held under the Chatham House rule, was attended by representatives from federal bank regulatory agencies, staff of BPI’s member banks, market participants, and academics.
The first panel discussion focused on the day-one impacts of CECL on loan-loss reserves and regulatory capital. The panelists generally agreed that the day-one consequences would vary across banks and importantly would depend on portfolio composition (e.g., credit card portfolios will experience higher reserves, while other portfolios could experience a decline in reserves) and risk composition within the portfolio (e.g., prime versus subprime). In addition, panelists noted that differences in methodology will also contribute to a wide range of projected day-one impacts. For example, one of the panelists stated that differences in estimates of the probability of default and prepayment estimates are materially based on the results of a recent benchmarking study. Another panelist observed that regulators do not intend to dictate reserve methodologies to banks but acknowledged that significant variability in reserving of credit card loans that reflect differences in payment assumptions and estimated recoveries. A third panelist pointed out that the day-one impact on reserves is less significant for projected earnings, but that the opaqueness underlying the main reasons driving the changes in provisions over time—i.e., due to changes in exposures, scenarios, and models—will create uncertainty and could dampen comparability and bank valuations.
The second panel discussed the potential procyclical effects of CECL on bank regulatory capital and credit supply over the business cycle. One panelist argued that the existing incurred-loss provisioning framework is also highly procyclical, citing the decline in loan-loss reserves during the period leading up to the Great Recession, before these reserves rose sharply during the recession itself. While the panelists did agree that reserves under CECL will be procyclical, they expressed sharply different views on whether the degree of procyclicality is likely to be greater or less than under the current incurred-loss methodology. It was noted during the discussion that the cyclicality of CECL reserves is likely to vary across asset classes. For example, loans with longer maturities (e.g., mortgages) may display less procyclicality, because a greater proportion of reserves are determined by the historical long-run average (i.e., are required for outside the reasonable and supportable period and therefore not based on macroeconomic forecasts).
One recurring theme of the second panel was that CECL reserves will be highly sensitive to modelling choices and forecasting. Panelists cited several examples of how small changes in modelling assumptions (e.g., the length of the reasonable and supportable period over which losses are based on projections) had significant effects on the path of expected losses, particularly at longer forecast horizons. All panelists agreed that CECL is significantly less procyclical if the macroeconomic forecasts used as inputs can accurately predict turning points in the business cycle, although macroeconomic forecasts have not previously offered much (if any) advance warning of macroeconomic downturns. Put another way, CECL will be procyclical because forecasts are unable to accurately predict turning points in a business cycle. Panelists also discussed the “right” level of procyclicality. For example, one participant highlighted that some degree of procyclicality would be expected from any reasonable framework for loan-loss provisioning, given that realized credit losses are highly concentrated during and after economic downturns.
The third panel pivoted the symposium discussion from the likely reserve effects of CECL to its policy implications. The three panelists were tasked with proposing and making the case for specific adjustments to how reserves under CECL are treated in regulatory capital to mitigate its procyclical effects. The first panelist noted that the regulatory treatment of reserves was established in Basel I and needs to be updated. He argued that, because reserves are loss absorbing, they should count toward the highest form of capital: common equity Tier 1 (CET1), instead of Tier 2 capital currently. Alternatively, he proposed that reserves in excess of expected 1-year credit losses as measured under the U.S. Basel 3 framework be included in CET1 and the remainder added to Tier 2 capital. The panelist further argued that the inclusion in capital of reserves in excess of 1-year credit losses would level the playing field for U.S. institutions relative to other jurisdictions that essentially limit the reserves to a 1-year horizon.
The second panelist observed that the procyclicality of CECL is made worse by the material losses banks are required to book each time they make a new loan. That loss is the result of CECL forcing banks to immediately recognize expected lifetime losses on the loan, but not allowing them to recognize expected lifetime interest earnings. The loss booked on origination is higher for both riskier loans and loans with longer maturities, as well as being higher during economic downturns. He proposed that loss provisions made on loan originations be credited back to regulatory capital and then amortized over the expected life of the loans.
The third panelist noted that Federal Reserve leadership has observed on several occasions that the total amount of loss absorbency in the banking system is currently about right. He argued that CECL will result in an unintended increase in total loss absorbency because of how it likely will be incorporated into the stress tests. In particular, reserves for the entire 9-quarter stress horizon would be booked at the beginning of the horizon. However, net charge-offs occurring after the quarter that the bank’s projected capital ratio reached its nadir would be permanently “trapped”, since they do not affect the amount of capital the bank needs to pass the test (which is related to the discussion on the last panel). He suggested that net charge-offs occurring beyond that quarter be deducted from the bank’s projected capital need under stress.
The fourth panel discussed the eventual incorporation of CECL into the stress tests. The panelists focused the discussion around four topics: (i) the challenges with the perfect-foresight assumption under the stress tests; (ii) adaptation of CECL business-as-usual models into the stress tests; (iii) use of qualitative factors in stress; and (iv) originations of assets under stress in the context of the flat balance sheet assumption. Panelists generally agreed that shortening the reasonable and supportable period and quickly reverting loan-loss projections to their mean values after that would be a viable way of addressing the unrealistic jump in reserves resulting from the perfect-foresight assumption under stress.