Stress Test Survey Results of BPI Members: Active Deferrals from June 30 to Nov. 30, 2020

BPI is concerned that losses on loans under deferral programs, as of June 30, 2020, could be materially overstated in the stress test results if the Federal Reserve were to assume that all or most of those loans would default as a result of a severe drop in economic activity. Such an assumption would significantly overstate the maximum decline in banks’ capital ratios over the stress planning horizon, which has the potential to undermine public confidence in the overall health and stability of the banking system and could harm the U.S. economic recovery.

BPI conducted a survey of its members to provide an update on the change in loan deferrals between June 30, 2020, and Nov. 30, 2020 (or the most recent date for which data were available).[1] In addition, the survey also collected data on the current status of loans in deferral as of June 30, 2020. Although some of the loans that were in deferral programs back in June and are now performing have been modified, especially for corporate portfolios, it is highly likely that improvements in economic conditions since the summer were the main driver of the reduction in loan deferrals over the past five months. The goal of our survey was to give a realistic and up-to-date view of the performance of loans deferred due to the coronavirus event in the spring.

The survey results are representative of the entire sample of banks subject to the supervisory stress tests. The survey collected responses from 16 banks subject to these tests. Those banks account for about 65 percent of total loans outstanding at stress-tested banks.[2]

The results show a 73-percent reduction in active deferrals—from 4.9 percent of loans outstanding on June 30, 2020, to 1.4 percent on Nov. 30, 2020. Loans to businesses and consumer loans experienced the largest declines in active deferrals (all over 85 percent). For residential real estate loans, active deferrals declined by more than 50 percent. The majority of the loans in deferral as of June 30, 2020—about 61 percent—migrated back to performing loan status, and only 3 percent of previously active deferrals migrated to non-performing. In summary, there has been a large reduction in active deferrals, and most of these loans are now performing.

We would further note that firms now must reserve the full expected lifetime losses on loans and certain commitments. As a result, any concern that ongoing deferrals or performing loans are masking loan losses is less pertinent today than before implementation of the current expected credit loss methodology.[3] In addition, even though banks saw improved economic outlooks in the third quarter, many banks did not meaningfully release reserves at that time due to continuing uncertainties. As a result, those reserves remain available to support potential future losses.

Notably, the Federal Reserve and other banking agencies encouraged banks to prudently work with borrowers in a safe and sound manner to mitigate adverse effects on consumers due to the pandemic. Assigning a high probability of default to these loans seems inconsistent with these statements and could appear to imply that the Federal Reserve is assuming for purposes of the stress tests that banks did not prudently modify the loans on a safe and sound basis.

Description of the Sample

BPI members include 29 of the 33 banks in Categories I–IV (as defined in the Federal Reserve’s tailoring framework) participating in the second round of stress tests. Of these, 16 firms participated in our survey on active deferrals, the results of which are summarized below. The banks that participated in the survey account for 65 percent of total loans outstanding for the 33 banks subject to the supervisory stress tests.

Reduction in Active Deferrals Since the Second Quarter of 2020  

Table 1 shows the reduction in active deferrals from June 30, 2020, to Nov. 30, 2020. The total reduction in active deferrals among all 16 survey participants was 73 percent or $118 billion.

Table 1: Reduction in Active Deferrals from June 30, 2020, to Nov. 30, 2020

Table 1 further breaks out the reduction in active deferrals by major loan type. Commercial real estate loans registered the largest decline in active deferrals (about 90 percent), closely followed by consumer loans, which show a reduction in active deferrals of about 88 percent, and commercial and industrial loans (about 87 percent) over the same period. The decline in loan deferrals for first-lien mortgage loans and HELOC loans was the lowest across all major loan categories but still represented a substantial reduction in active deferrals (about 51 percent) over this time period. The smaller decline in loan deferrals for first-lien mortgage loans and HELOC loans likely reflects the longer deferral tenor of residential real estate loans in that the deferral program has not expired yet.

As Table 1 reports, active deferrals for commercial and industrial loans fell from 2.3 percent to 0.3 percent of total loans outstanding over the past five months, which corresponds to a decline in loan deferrals of about 87 percent.[4] Similarly, active deferrals for commercial real estate loans declined from 7.6 percent to 0.8 percent over the period, a drop of about 90 percent.

Table 1 also reveals that for consumer loans, active deferrals declined from 5.9 percent to 0.7 percent of total loans outstanding, nearly an 88-percent decrease. Finally, active deferrals for first-lien mortgage and HELOC loans declined from 7.4 percent to 3.6 percent over the period, representing about a 51-percent decrease in active deferrals.

The dramatic reduction in active deferrals demonstrates how important it is that the Federal Reserve disclose its assumptions about performance of loans in deferral on June 30 for purposes of its stress test. For example, an assumption that all or most of those loans would immediately be in default would dramatically overstate the losses and thus the capital needs of the affected banks.

Distribution of Loan Status on Nov. 30, 2020, That Were in Active Deferral on June 30, 2020

Exhibit 1 describes the distribution of loan status on Nov. 30, 2020, that were in active deferral on June 30, 2020, offering insight into what happened to these loans between June 30 and the most current date. We present results in totals across all survey participants who provided data. The key takeaway is that the vast majority of the loans are now performing.

Of the $162 billion in active deferral loans in our sample as of June 30, 2020, about 61 percent are performing loans as of the most current date for which data are available. The share of performing loans is higher for business loans (the sum of commercial and industrial loans and commercial real estate loans) and consumer loans than for first-lien mortgage loans and HELOCs loans.

Next, about 28 percent of total loans considered, or $44 billion, remain in active deferral status, driven in large part by first-lien mortgage loans and HELOCs. As noted above, we believe the higher share of such loans under deferral reflects the nature of the deferral tenor in residential real estate loans (i.e., deferral program has not expired yet). This should not be interpreted as a sign that those loans are experiencing more difficulties in migrating from the deferral status back to the performing status. Indeed, the housing market is holding up well because of increased demand for residential real estate outside of major cities as a result of remote work.

Nonperforming loans and loans accruing past due 90 days or more account for a small percentage of active deferral loans at the end of the second quarter of 2020, or about 3 percent combined based on current data. Finally, loans that have been paid in full, charged off, sold, etc., account for about 8 percent of loans in active deferral as of June 30, 2020. Moreover, any losses on these loans have also already been taken into account in bank regulatory capital levels.

The results in Exhibit 1 demonstrate that most previously active deferrals are now performing loans. However, some caveats are worth noting. First, as described in the Supervision and Regulation Report, “The prevalence of loan modification programs may obscure credit quality issues, as a loan is typically not counted as ‘nonperforming’ while it is covered by a loan modification program.”[5] Second, we are unable to observe the counterfactual, meaning what would have happened to the performance of such loans had we experienced a second severe recession. Nevertheless, the data demonstrate that the degree of uncertainty regarding the size of loan defaults in 2021 tied to deferrals resulting from the coronavirus event in the spring of 2020 has declined significantly over the period. This is because of the performance of these loans to date, largely due to improved economic conditions.


The survey results show a marked improvement in active deferrals over the period. Even if we assumed another severe recession were to occur, it would be unreasonable to assign a high uniform probability of default (PD) to loans under an active deferral as of June 30, 2020, for the following three reasons:

  • Consistency with the two stress scenarios would not be maintained. The Federal Reserve made a significant change to the stress scenarios that will mitigate the procyclicality of the stress tests. Instead of using the 13-percent unemployment rate in the second quarter as the jumping-off point from which the deterioration in the unemployment rate starts, both stress scenarios use the third-quarter baseline forecast for the unemployment rate of 9½ percent. So, if the Federal Reserve allowed the unemployment rate to fall 3½ percent in the third quarter, then it would be more consistent with the stress scenario to use the third-quarter active deferral data, which had already declined considerably between the second and third quarters of 2020.[6]
  • Borrowers in active deferral are not identical. Applying a high PD to all borrowers that benefited from deferral actions is also unreasonable because of the heterogeneity in borrower and loans characteristics as well as the type of deferral actions.[7] Reportedly, the data show that many customers paid during these deferrals, suggesting that the deferrals were driven by precautionary reasons and not because the deferrals were needed. Next, borrower and loan characteristics will be important drivers of default after the expiration of the deferral programs, even if we assume another serious second economic downturn were to occur. Chan and colleagues (2012) showed that borrower and loan characteristics (credit score and loan-to-values) were important in determining which loans cured during the peak of the mortgage crisis period.[8] Also, Calem and colleagues (2020) explain that that re-default depends importantly on credit score, loan-to-values and the generosity of the loan modification.[9] On the consumer side, Calem, Jagtiani, and Lang (2015) argue that homeowners benefiting from mortgage forbearance from the GSEs or FHA would be expected to use the windfall to pay down consumer debt.[10] And Canals-Cerda and Kerr (2015) show that variables like credit score, utilization rate, balance and delinquency history are highly important determinants of defaults on credit cards.[11] In summary, it is more sensible to stress borrowers under active deferral programs by increasing their PD based on borrower and loan characteristics as well as the types of deferral actions, instead of assuming a PD of 100 percent irrespective of those characteristics.
  • Ad-hoc adjustments to the PD of borrowers in deferral programs make the projected losses of non-deferral borrowers more conservative. After singling out borrowers who benefited from deferral actions for special, ad hoc treatment, the stress test models as applied to the remaining borrower population naturally become more conservative. This is a mathematical consequence of the fact that modeled PDs are based on historical data pooling together the higher-PD borrowers, who would have availed themselves of the opportunity for deferral treatment had it been as widely available as now, with lower-PD borrowers. Given that the modeled PDs for the remaining population become more conservative, there is less need to apply extreme conservatism in assigning PDs to the borrowers who benefited from deferral.

Other considerations not directly tied to the severity of the stress scenario are worth noting:

  • Assuming deferred loans default creates disincentives for banks to grant deferrals during a period of stress to support customers and the economy. An assumption in stress tests of high uniform PD runs counter to statements by the Federal Reserve and other agencies that have encouraged banks to work prudently with borrowers who are or may be unable to meet their payment obligations because of the effects of COVID-19. These statements indicate that such proactive measures are in the best interest of financial institutions, their borrowers and the economy.[12] The agencies also said they would not criticize institutions for working with borrowers in a safe and sound manner, so assigning a high probability of default to these loans seems inconsistent with these statements. It also implies that the banks did not prudently modify the loans on a safe and sound basis.
  • Banks must reserve for lifetime losses under CECL. The new accounting standard, which replaces the incurred loss methodology for financial assets measured at amortized cost, requires banking organizations to immediately recognize lifetime expected credit losses and to incorporate reasonable and supportable forecasts in developing the estimate of lifetime expected credit losses. As a result of this standard, firms now have to reserve the full lifetime losses on loans, so any concern that ongoing deferrals are masking loan losses is less pertinent today than before CECL. Whether the loan is in deferral or not, banks have reserved for the full lifetime loss of the loan.
  • The economic outlook has improved. When the Federal Reserve released the results of the June sensitivity analysis, it cited “material uncertainty about the trajectory for the economic recovery and corresponding uncertainty related to its effects on the financial health of banking organizations” to take additional actions to preserve capital at banks and ensure they “remain a source of strength in the future.”[13] The marked improvement in active deferrals over the last five months reflects ongoing improvements in the economic outlook relative to June and the expected availability of several reliable COVID-19 vaccines in 2021.
  • Procyclicality of the capital framework could be exacerbated. An extreme and empirically unjustified overlay applied to loans in deferral, modification or forbearance clearly exacerbates the procyclicality of stress tests. During a stress period, many balance sheet assets are recalibrated to have a higher probability of default or loss-given-default, which in turn implies an increase in projected losses under stress scenarios. For example, loan-to-value ratios increase for mortgages as house prices decline; credit scores of individuals and credit ratings of firms tend to decline; and borrower delinquency and default rates increase. The attendant increase in stressed losses and capital requirements can incentivize banks to tighten lending requirements and credit supply, delaying economic recovery. Also, showing higher peak-to-trough decreases in the common equity Tier 1 ratios of banks could lead market participants to use the results of the second stress tests to infer or create a “shadow” SCB and effectively subject banks to higher capital requirements, thereby further increasing the procyclicality of the capital framework.
  • There could be outsized public reaction to stress test result disclosures. Making unduly severe assumptions in the treatment of deferred loans could mislead the public about the true state of U.S. bank balance sheets. This would needlessly undermine the strength of and confidence in an otherwise resilient banking system.

Concluding Thoughts

Loans in deferral as of June 30, 2020, have performed well; a substantial majority of those loans are now performing. This result highlights the importance of the Federal Reserve disclosing what assumptions it made about these loans in its upcoming stress tests, and in a future determination about capital distributions.

In particular, an assumption that these loans would be in default with 100-percent certainty would be unreasonable. This would be true even if the stress tests seek to evaluate the worst possible case and some loans that were subject to a deferral program and are now performing would have defaulted under a severe drop in economic activity. At a minimum, the Federal Reserve should use the loans in active deferral as of the end of the third quarter to align with the unemployment and GDP assumptions embedded in the two macroeconomic scenarios. The Fed could also consider differences in the risk characteristics of borrowers that have benefited from deferral programs and the size of such deferrals to differentiate among PDs more representatively and accurately.

There are other problems associated with an unreasonably high assumed default rate on deferred loans in the stress tests conducted in the middle of an economic downturn. For example, it is highly procyclical; could result in a higher “shadow” SCB; and could lead to an outsized negative market reaction, undermining the confidence of the public in the banking sector. Moreover, it creates a disincentive for banks to grant deferrals the next time the Federal Reserve encourages them to do so, undermining a financial policy that may otherwise be supportive of banks accommodating their customers in periods of stress.

[1] For simplicity, throughout this post, we refer to Nov. 30, 2020. Most firms provided data around that date, and the results include three survey responses as of Sept. 30, 2020. Although the survey uses each bank’s own definition of what constitutes a loan deferral, “active deferrals” generally includes payment relief in the form of payment deferral programs, maturity extension programs, mortgage forbearance and payment plans. It excludes non-financial relief such as covenant waivers, loans classified as a troubled debt restructuring, and loan modifications made after expiration of the deferral program.

[2] The responses to BPI’s survey were received from each bank. However, the individual responses have not been shared among banks and are only being reported in an aggregate, anonymized form. Although we collected the data ourselves under a tight timeframe, firms can also provide data to the Federal Reserve directly, if requested, through supervisory channels, to give the Federal Reserve a clearer view into results for individual institutions. We expect any such data would offer the Federal Reserve a strong basis to return to the normal capital framework of the stress capital buffer to govern capital distributions for stress-tested banks.

[3] We note this is true regardless of the incorporation of CECL into the Federal Reserve’s stress tests. See Federal Reserve Board, Statement on the current expected credit loss methodology (CECL) and stress testing (December 21, 2018), available at

[4] Almost all small business loans are included in the consumer portfolio. Some banks reported small business loans in commercial and industrial loans.

[5] Federal Reserve Board, Supervision and Regulation Report, at 7–8 (November 2020), available at (citing section 4013 of the Coronavirus Aid, Relief, and Economic Security Act, Pub. L. No. 116-136, 134 Stat. 281 (March 27, 2020); Interagency Statement on Loan Modifications and Reporting for Financial Institutions Working with Customers Affected by the Coronavirus (Revised) (April 7, 2020), available at; and SR Letter 20-18/CA 20-13, Joint Statement on Additional Loan Accommodations Related to COVID-19 (August 3, 2020), available at

[6] For instance, JPMorgan reported an approximately 50-percent reduction in payment deferrals for consumer loans between 2Q20 and 3Q20 based on its third-quarter public earnings results, while our data show a 73-percent reduction between 2Q20 and Nov. 30, 2020, in aggregate across all 16 banks that participated in our survey. See JPMorgan Chase & Co., 3Q2020 Financial Results at 4 (October 13, 2020), available at

[7] For example, we recognize the Federal Reserve has previously assumed a PD of 100 percent for troubled loans. See Federal Reserve Board, Dodd-Frank Act Stress Test 2019: Supervisory Stress Test Methodology at 27 (March 2019), available at

[8] Chan, Sewin, Claudia Sharygin, and Andrew Haughwout, “Pathways after Default: What Happens to Distressed Mortgage Borrowers and Their Homes?” Journal of Real Estate Finance and Economics, Vol. 48, No. 2, 2014, available at

[9] Calem, Paul, Julapa Jagtiani, and Raman Maingi; “Redefault Risk in the Aftermath of the Mortgage Crisis: Why Did Modifications Improve More Than Self-Cures?” forthcoming in Journal of Housing Research, available at

[10] Calem, Paul, Julapa Jagtiani, and William Lang, “Foreclosure Delay and Consumer Credit Performance,” FRB of Philadelphia Working Paper No. 15-24/R, available at

[11] Canals-Cerda, José J., and Sougata Kerr; “Forecasting Credit Card Portfolio Losses in the Great Recession: A Study in Model Risk,” Journal of Credit Risk, 2015, FRB of Philadelphia Working Paper No. 14-10, available at

[12] See Agencies provide additional information to encourage financial institutions to work with borrowers affected by COVID-19 (March 22, 2020), available at; Interagency Statement on Loan Modifications and Reporting for Financial Institutions Working with Customers Affected by the Coronavirus (Revised) (April 7, 2020), available at; SR 20-15: Interagency Examiner Guidance for Assessing Safety and Soundness Considering the Effect of the COVID-19 Pandemic on Institutions (June 23, 2020), available at; SR 20-18 / CA 20-13: Joint Statement on Additional Loan Accommodations Related to COVID-19 (August 3, 2020), available at

[13] Federal Reserve Board, Assessment of Bank Capital during the Recent Coronavirus Event at 1 (June 2020), available at