The expected life of funded credit card loans is generally estimated by considering expected future payments on the credit card account, and determining how much of those amounts should be allocated to repayments of the funded loan balance (as of the balance sheet date) versus other account activity. This allocation is made using an approach that incorporates the payment application requirements of the Credit Card Accountability Responsibility and Disclosure Act of 2009, generally paying down the highest interest rate balances first.

The estimate of expected credit losses includes expected recoveries of amounts previously charged off or expected to be charged off, even if such recoveries result in a negative allowance.

Collective and Individual Assessments

When calculating the allowance for loan losses and the allowance for lending-related commitments, the Firm assesses whether exposures share similar risk characteristics. If similar risk characteristics exist, the Firm estimates expected credit losses collectively, considering the risk associated with a particular pool and the probability that the exposures within the pool will deteriorate or default. The assessment of risk characteristics is subject to significant management judgment. Emphasizing one characteristic over another or considering additional characteristics could affect the allowance.

•Relevant risk characteristics for the consumer portfolio include product type, delinquency status, current FICO scores, geographic distribution, and, for collateralized loans, current LTV ratios.

•Relevant risk characteristics for the wholesale portfolio include LOB, geography, risk rating, delinquency status, level and type of collateral, industry, credit enhancement, product type, facility purpose, tenor, and payment terms.

The majority of the Firm’s credit exposures share risk characteristics with other similar exposures, and as a result are collectively assessed for impairment (“portfolio-based component”). The portfolio-based component covers consumer loans, performing risk-rated loans and certain lending-related commitments.

If an exposure does not share risk characteristics with other exposures, the Firm generally estimates expected credit losses on an individual basis, considering expected repayment and conditions impacting that individual exposure (“asset-specific component”). The asset-specific component covers modified PCD loans, loans modified or reasonably expected to be modified in a TDR, collateral-dependent loans, as well as, risk-rated loans that have been placed on nonaccrual status.

Portfolio-based component

The portfolio-based component begins with a quantitative calculation that considers the likelihood of the borrower changing delinquency status or moving from one risk rating to another. The quantitative calculation covers expected credit losses over an instrument’s expected life and is estimated by applying credit loss factors to the Firm’s

estimated exposure at default. The credit loss factors incorporate the probability of borrower default as well as loss severity in the event of default. They are derived using a weighted average of five internally developed macroeconomic scenarios over an eight-quarter forecast period, followed by a single year straight-line interpolation to revert to long run historical information for periods beyond the eight-quarter forecast period. The five

macroeconomic scenarios consist of a central, relative adverse, extreme adverse, relative upside and extreme upside scenario, and are updated by the Firm’s central forecasting team. The scenarios take into consideration the Firm’s overarching economic outlook, internal perspectives from subject matter experts across the Firm, and market consensus and involve a governed process that incorporates feedback from senior management across LOBs, Corporate Finance and Risk Management.

The COVID-19 pandemic has stressed many MEVs to degrees not experienced in recent history, which has created additional challenges in the use of modeled credit loss estimates and increased the reliance on management judgment. In periods where certain MEVs are outside the range of historical experience on which the Firm’s models have been trained, the Firm makes adjustments to appropriately address these economic circumstances. The Firm also considers the impact of other events, such as government unemployment benefits or other stimulus programs, when determining whether adjustments are necessary.

The quantitative calculation is adjusted to take into consideration model imprecision, emerging risk assessments, trends and other subjective factors that are not yet reflected in the calculation. These adjustments are accomplished in part by analyzing the historical loss experience, including during stressed periods, for each major product or model. Management applies judgment in making this adjustment, including taking into account uncertainties associated with the economic and political conditions, quality of underwriting standards, borrower behavior, credit concentrations or deterioration within an industry, product or portfolio, as well as other relevant internal and external factors affecting the credit quality of the portfolio. In certain instances, the interrelationships between these factors create further uncertainties.

Throughout 2020, the Firm made adjustments to its quantitative calculation which placed significant weighting on its adverse scenarios, as a result of continued uncertainty related to the COVID-19 pandemic.

The application of different inputs into the quantitative calculation, and the assumptions used by management to adjust the quantitative calculation, are subject to significant management judgment, and emphasizing one input or assumption over another, or considering other inputs or assumptions, could affect the estimate of the allowance for loan losses and the allowance for lending-related commitments.

Asset-specific component

To determine the asset-specific component of the allowance, collateral-dependent loans (including those loans for which foreclosure is probable) and larger, nonaccrual risk-rated loans in the wholesale portfolio segment are generally evaluated individually, while smaller loans (both scored and risk-rated) are aggregated for evaluation using factors relevant for the respective class of assets.

The Firm generally measures the asset-specific allowance as the difference between the amortized cost of the loan and the present value of the cash flows expected to be collected, discounted at the loan’s original effective interest rate. Subsequent changes in impairment are generally recognized as an adjustment to the allowance for loan losses. For collateral-dependent loans, the fair value of collateral less estimated costs to sell is used to determine the charge-off amount for declines in value (to reduce the amortized cost of the loan to the fair value of collateral) or the amount of negative allowance that should be recognized (for recoveries of prior charge-offs associated with improvements in the fair value of collateral).

The asset-specific component of the allowance for loan losses for loans that have been or are expected to be modified in TDRs incorporates the effect of the modification on the loan’s expected cash flows (including forgone interest, principal forgiveness, as well as other concessions), and also the potential for redefault. For residential real estate loans modified in or expected to be modified in TDRs, the Firm develops product-specific probability of default estimates, which are applied at a loan level to compute expected losses. In developing these probabilities of default, the Firm considers the relationship between the credit quality characteristics of the underlying loans and certain assumptions about housing prices and unemployment, based upon industry-wide data. The Firm also considers its own historical loss experience to-date based on actual redefaulted modified loans. For credit card loans modified in or expected to be modified in TDRs, expected losses incorporate projected delinquencies and charge-offs based on the Firm’s historical experience by type of modification program. For wholesale loans modified or expected to be modified in TDRs, expected losses incorporate management’s expectation of the borrower’s ability to repay under the modified terms.

Estimating the timing and amounts of future cash flows is highly judgmental as these cash flow projections rely upon estimates such as loss severities, asset valuations, default rates (including redefault rates on modified loans), the amounts and timing of interest or principal payments (including any expected prepayments) or other factors that are reflective of current and expected market conditions. These estimates are, in turn, dependent on factors such as the duration of current overall economic conditions, industry-, portfolio-, or borrower-specific factors, the expected outcome of insolvency proceedings as well as, in certain circumstances, other economic factors. All of these estimates and assumptions require significant management judgment and certain assumptions are highly subjective.

Allowance for credit losses and related information

The table below summarizes information about the allowances for loan losses and lending-related commitments, and includes a breakdown of loans and lending-related commitments by impairment methodology. Refer to Note 10 for further information on the allowance for credit losses on investment securities.

The adoption of the CECL accounting guidance resulted in a change in the accounting for PCI loans, which are considered PCD loans. In conjunction with the adoption of CECL, the Firm reclassified risk-rated loans and lending-related commitments from the consumer, excluding credit card portfolio segment to the wholesale portfolio segment, to align with the methodology applied when determining the allowance. Prior-period amounts have been revised to conform with the current presentation. Refer to Note 1 for further information.


(Table continued on next page)
	2020(e)
Year ended December 31, (in millions)	Consumer, excluding credit card		Credit card		Wholesale		Total
Allowance for loan losses
Beginning balance at January 1,	$	2,538	$	5,683	$	4,902	$	13,123
Cumulative effect of a change in accounting principle	297		5,517		(1,642)		4,172
Gross charge-offs	805		5,077		954		6,836
Gross recoveries collected	(631)		(791)		(155)		(1,577)
Net charge-offs	174		4,286		799		5,259
Write-offs of PCI loans(a)	NA		NA		NA		NA
Provision for loan losses	974		10,886		4,431		16,291
Other	1		—		—		1
Ending balance at December 31,	$	3,636	$	17,800	$	6,892	$	28,328

Allowance for lending-related commitments
Beginning balance at January 1,	$	12	$	—	$	1,179	$	1,191
Cumulative effect of a change in accounting principle	133		—		(35)		98
Provision for lending-related commitments	42		—		1,079		1,121
Other	—		—		(1)		(1)
Ending balance at December 31,	$	187	$	—	$	2,222	$	2,409

Total allowance for credit losses	$	3,823	$	17,800	$	9,114	$	30,737

Allowance for loan losses by impairment methodology
Asset-specific(b)	$	(7)	$	633	$	682	$	1,308
Portfolio-based	3,643		17,167		6,210		27,020
PCI	NA		NA		NA		NA
Total allowance for loan losses	$	3,636	$	17,800	$	6,892	$	28,328

Loans by impairment methodology
Asset-specific(b)	$	16,648	$	1,375	$	3,606	$	21,629
Portfolio-based	285,479		142,057		511,341		938,877
PCI	NA		NA		NA		NA
Total retained loans	$	302,127	$	143,432	$	514,947	$	960,506

Collateral-dependent loans
Net charge-offs	$	133	$	—	$	76	$	209
Loans measured at fair value of collateral less cost to sell	4,956		—		188		5,144


Allowance for lending-related commitments by impairment methodology
Asset-specific	$	—	$	—	$	114	$	114
Portfolio-based	187		—		2,108		2,295
Total allowance for lending-related commitments(c)	$	187	$	—	$	2,222	$	2,409

Lending-related commitments by impairment methodology
Asset-specific	$	—	$	—	$	577	$	577
Portfolio-based(d)	37,783		—		426,871		464,654
Total lending-related commitments	$	37,783	$	—	$	427,448	$	465,231

(a)Prior to the adoption of CECL, write-offs of PCI loans were recorded against the allowance for loan losses when actual losses for a pool exceeded estimated losses that were recorded as purchase accounting adjustments at the time of acquisition. A write-off of a PCI loan was recognized when the underlying loan was removed from a pool.

(b)Includes modified PCD loans and loans that have been modified or are reasonably expected to be modified in a TDR. Also includes risk-rated loans that have been placed on nonaccrual status for the wholesale portfolio segment. The asset-specific credit card allowance for loans modified, or reasonably expected to be modified, in a TDR is calculated based on the loans’ original contractual interest rates and does not consider any incremental penalty rates.

(c)The allowance for lending-related commitments is reported in accounts payable and other liabilities on the Consolidated balance sheets.

(d)At December 31, 2020, 2019 and 2018, lending-related commitments excluded $19.5 billion, $9.8 billion and $8.7 billion, respectively, for the consumer, excluding credit card portfolio segment; $658.5 billion, $650.7 billion and $605.4 billion, respectively, for the credit card portfolio segment; and $22.4 billion, $24.1 billion and $24.8 billion, respectively, for the wholesale portfolio segment, which were not subject to the allowance for lending-related commitments.

(e)Excludes HTM securities, which had an allowance for credit losses of $78 million and a provision for credit losses of $68 million as of and for the year ended December 31, 2020.


(table continued from previous page)
2019								2018
Consumer, excluding credit card		Credit card		Wholesale		Total		Consumer, excluding credit card		Credit card		Wholesale		Total

$	3,434	$	5,184	$	4,827	$	13,445	$	3,892	$	4,884	$	4,828	$	13,604
NA		NA		NA		NA		NA		NA		NA		NA
902		5,436		472		6,810		977		5,011		361		6,349
(536)		(588)		(57)		(1,181)		(827)		(493)		(173)		(1,493)
366		4,848		415		5,629		150		4,518		188		4,856
151		—		—		151		187		—		—		187
(378)		5,348		479		5,449		(121)		4,818		188		4,885
(1)		(1)		11		9		—		—		(1)		(1)
$	2,538	$	5,683	$	4,902	$	13,123	$	3,434	$	5,184	$	4,827	$	13,445



$	12	$	—	$	1,043	$	1,055	$	12	$	—	$	1,056	$	1,068
NA		NA		NA		NA		NA		NA		NA		NA
—		—		136		136		—		—		(14)		(14)
—		—		—		—		—		—		1		1
$	12	$	—	$	1,179	$	1,191	$	12	$	—	$	1,043	$	1,055

$	2,550	$	5,683	$	6,081	$	14,314	$	3,446	$	5,184	$	5,870	$	14,500


$	75	$	477	$	295	$	847	$	143	$	440	$	350	$	933
1,476		5,206		4,607		11,289		1,503		4,744		4,477		10,724
987		—		—		987		1,788		—		—		1,788
$	2,538	$	5,683	$	4,902	$	13,123	$	3,434	$	5,184	$	4,827	$	13,445


$	5,961	$	1,452	$	1,123	$	8,536	$	6,665	$	1,319	$	1,459	$	9,443
268,675		167,472		480,555		916,702		305,077		155,297		475,561		935,935
20,363		—		—		20,363		24,034		—		3		24,037
$	294,999	$	168,924	$	481,678	$	945,601	$	335,776	$	156,616	$	477,023	$	969,415


$	46	$	—	$	36	$	82	$	16	$	—	$	29	$	45
2,053		—		87		2,140		2,076		—		206		2,282










$	—	$	—	$	102	$	102	$	—	$	—	$	99	$	99
12		—		1,077		1,089		12		—		944		956
$	12	$	—	$	1,179	$	1,191	$	12	$	—	$	1,043	$	1,055


$	—	$	—	$	474	$	474	$	—	$	—	$	469	$	469
30,417		—		392,967		423,384		26,502		—		374,996		401,498
$	30,417	$	—	$	393,441	$	423,858	$	26,502	$	—	$	375,465	$	401,967

Discussion of changes in the allowance during 2020

The increase in the allowance for loan losses and lending-related commitments was primarily driven by an increase in the provision for credit losses, reflecting the deterioration in and uncertainty around the future macroeconomic environment as a result of the impact of the COVID-19 pandemic.

As of December 31, 2020, the Firm’s central case reflected U.S. unemployment rates of approximately 7% through the second quarter of 2021 and remaining above 5% until the second half of 2022. This compared with relatively low levels of unemployment of approximately 4% throughout 2020 and 2021 in the Firm’s January 1, 2020 central case.

Further, while the Firm’s January 1, 2020 central case U.S. GDP forecast reflected a 1.7% expansion in 2020, actual U.S. GDP contracted approximately 2.5% in 2020. As of December 31, 2020, the Firm’s central case assumptions reflect a return to pre-pandemic GDP levels in the fourth quarter of 2021.

Due to elevated uncertainty in the near term outlook, driven by the potential for increased infection rates and related lock downs resulting from the pandemic, as well as the prospect that government and other consumer relief measures set to expire may not be extended, the Firm has placed significant weighting on its adverse scenarios. These scenarios incorporate more punitive macroeconomic factors than the central case assumptions, resulting in weighted average U.S. unemployment rates remaining elevated throughout 2021 and 2022, ending the fourth quarter of 2022 at approximately 6%, and in U.S. GDP ending 2022 approximately 0.9% higher than fourth quarter 2019 actual pre-pandemic levels.

The Firm’s central case assumptions reflected U.S. unemployment rates and U.S. real GDP as follows:


	Assumptions at January 1, 2020
	2Q20	4Q20(b)	2Q21
U.S. unemployment rate(a)	3.7%	3.8%	4.0%
Cumulative change in U.S. real GDP from 12/31/2019	0.9%	1.7%	2.4%


	Assumptions at December 31, 2020
	2Q21	4Q21	2Q22
U.S. unemployment rate(a)	6.8%	5.7%	5.1%
Cumulative change in U.S. real GDP from 12/31/2019	(1.9)%	0.6%	2.0%

(a)Reflects quarterly average of forecasted U.S. unemployment rate.

(b)4Q20 actual U.S. unemployment rate (quarterly average) was 6.8%. 4Q20 actual cumulative change in U.S. real GDP from 4Q19 was (2.5%).

Subsequent changes to this forecast and related estimates

will be reflected in the provision for credit losses in future

periods.