AMERICAN EXPRESS CO

Refer to Note 1 to the “Consolidated Financial Statements” for a summary of our significant accounting policies. Certain of our accounting policies requiring significant management assumptions and judgments are as follows:

RESERVES FOR CARD MEMBER CREDIT LOSSES

Reserves for Card Member credit losses represent our best estimate of the expected credit losses in our outstanding portfolio of Card Member loans and receivables as of the balance sheet date. The CECL methodology requires us to estimate lifetime expected credit losses by incorporating historical loss experience, as well as current and future economic conditions over a reasonable and supportable period (R&S Period) beyond the balance sheet date.

In estimating expected credit losses, we use a combination of statistically based models and analysis of the results produced by these models to determine the quantitative and qualitative components of our total balance sheet reserves for credit losses. These quantitative and qualitative components entail a significant amount of judgment. The primary areas of judgment used in measuring the quantitative components of our reserves relate to the determination of the appropriate R&S Period, the modeling of the probability of and exposure at default, and the methodology to incorporate current and future economic conditions. We use these models and assumptions, combined with historical loss experience, to determine the reserve rates that are applied to the outstanding loan or receivable balances to produce our reserves for expected credit losses for the R&S Period. The qualitative component is intended to capture expected losses that may not have been fully captured in the quantitative component. Through an established governance structure, we consider certain external and internal factors, including emerging portfolio characteristics and trends, which consequentially may increase or decrease the reserves for Card Member credit losses.

The R&S Period, which is approximately three years, represents the maximum time-period beyond the balance sheet date over which we can reasonably estimate expected credit losses, using all available portfolio information, current economic conditions and forecasts of future economic conditions. Card Member loan products do not have a contractual term and balances can revolve if minimum required payments are made, causing some balances to remain outstanding beyond the R&S Period. To determine expected credit losses beyond the R&S Period, we immediately revert to long-term average loss rates. Card Member receivable products are contractually required to be paid in full; therefore, we have assumed the balances will be either paid or written-off no later than 180 days past due.

Within the R&S Period, our models use past loss experience and current and future economic conditions to estimate the probability of default, exposure at default and expected recoveries to estimate net losses at default. A significant area of judgment relates to how we apply future Card Member payments to the reporting period balances when determining the exposure at default. The nature of revolving loan products inherently includes a relationship between future payments and spend behavior, which creates complexity in the application of how future payments are either partially or entirely attributable to the existing balance at the end of the reporting period. Using historical customer behavior and other factors, we have assumed that future payments are first allocated to interest and fees associated with the reporting period balance and future spend. We then allocate a portion of the payment to the estimated higher minimum payment amount due because of any future spend. Any remaining portion of the future payment is then allocated to the remaining reporting period balance.

CECL requires that the R&S Period include an assumption about current and future economic conditions. We incorporate multiple macroeconomic scenarios provided to us by an independent third party. The estimated credit losses calculated from each macroeconomic scenario are reviewed each period and weighted to reflect management’s judgment about uncertainty surrounding these scenarios. These macroeconomic scenarios contain certain variables, including unemployment rates and real GDP, that are significant to our models.

Macroeconomic Sensitivity

To demonstrate the sensitivity of estimated credit losses to the macroeconomic scenarios, we compared our modeled estimates under a baseline scenario to that under a pessimistic downside scenario. As of December 31, 2025, for every 10 percentage points change in weighting from the baseline scenario to the pessimistic downside scenario, the estimated credit losses increased by approximately $220 million.

The modeled estimates under these scenarios were influenced by the duration, severity and timing of changes in economic variables within each scenario and these macroeconomic scenarios, under different conditions or using different assumptions, could result in significantly different estimated credit losses. It is difficult to estimate how potential changes in specific factors might affect the estimated credit losses, and current results may not be indicative of the potential future impact of macroeconomic forecast changes.

In addition, this sensitivity analysis relates only to the modeled credit loss estimates under two scenarios without considering management’s judgment on the relative weighting for those and other scenarios, including the weight that has been placed on the downside scenario at the balance sheet date, or any potential changes in other adjustments to the quantitative reserve component or the impact of management judgment for the qualitative reserve component, which may have a positive or negative effect on the results. Thus, the results of this sensitivity analysis are hypothetical and are not intended to estimate or reflect our expectations of any changes in the overall reserves for credit losses due to changes in the macroeconomic environment.

Refer to Note 3 to the “Consolidated Financial Statements” for further information on the range of macroeconomic scenario key variables used, in conjunction with other inputs described above, to calculate reserves for Card Member credit losses.

The process of estimating these reserves requires a high degree of judgment. To the extent our expected credit loss models are not indicative of future performance, actual losses could differ significantly from our judgments and expectations, resulting in either higher or lower future provisions for credit losses in any period.

LIABILITY FOR MEMBERSHIP REWARDS

The Membership Rewards program is our largest card-based rewards program. Card Members can earn points for purchases charged on their enrolled card products. A significant portion of our cards, by their terms, allow Card Members to earn bonus points for purchases at merchants in particular industry categories. Membership Rewards points are redeemable for a broad variety of rewards, including, but not limited to, travel, shopping, gift cards, and statement credits. Points typically do not expire, and there is no limit on the number of points a Card Member may earn. Membership Rewards expense is driven by charge volume on enrolled cards, customer participation in the program and contractual arrangements with redemption partners.

We record a Membership Rewards liability that represents our best estimate of the cost of points earned that are expected to be redeemed by Card Members in the future. The Membership Rewards liability is impacted over time by enrollment levels, attrition, the volume of points earned and redeemed, and the associated redemption costs. We estimate the Membership Rewards liability by determining the URR and the weighted average cost (WAC) per point, which are applied to the points of current enrollees. Refer to Note 9 to the “Consolidated Financial Statements” for additional information.

The URR assumption is used to estimate the number of points earned by current enrollees that will ultimately be redeemed in future periods. We use statistical and actuarial models to estimate the URR of points earned to date by current Card Members based on redemption trends, card product type, enrollment tenure, card spend levels and credit attributes. The WAC per point assumption is used to estimate future redemption costs and is primarily based on redemption choices made by Card Members, reward offerings by partners, and Membership Rewards program changes. The WAC per point assumption is derived from 12 months of redemptions and is adjusted as appropriate for certain changes in redemption costs that are not representative of future cost expectations and expected developments in redemption patterns.

We periodically evaluate our liability estimation process and assumptions based on changes in cost per point redeemed, partner contract changes and developments in redemption patterns, which may be impacted by product refreshes, changes in redemption options and mix of proprietary cards-in-force.

The process of estimating the Membership Rewards liability includes a high degree of judgment. Actual redemptions and associated redemption costs could differ significantly from our estimates, resulting in either higher or lower Membership Rewards expense.

Changes in the Membership Rewards URR and WAC per point have the effect of either increasing or decreasing the liability through the current period Membership Rewards expense by an amount estimated to cover the cost of all points previously earned but not yet redeemed by current enrollees as of the end of the reporting period. As of December 31, 2025, an increase in the estimated URR of current enrollees of 25 basis points would increase the Membership Rewards liability and corresponding rewards expense by approximately $229 million. Similarly, an increase in the WAC per point of 1 basis point would increase the Membership Rewards liability and corresponding rewards expense by approximately $244 million.

GOODWILL RECOVERABILITY

Goodwill represents the excess of acquisition cost of an acquired business over the fair value of assets acquired and liabilities assumed. Goodwill is not amortized but is tested for impairment at the reporting unit level annually or when events or circumstances arise, such as adverse changes in the business environment, that would more likely than not reduce the fair value of the reporting unit below its carrying value. Our methodology for conducting this goodwill impairment testing contains both a qualitative and quantitative assessment.

We have the option to initially perform an assessment of qualitative factors in order to determine whether it is more likely than not that the fair value of a reporting unit is less than its carrying amount. The qualitative factors may include, but are not limited to, economic conditions, industry and market considerations, cost factors, overall financial performance of the reporting unit and other company and reporting unit-specific events. If we determine that it is more likely than not that the fair value of a reporting unit is less than its carrying amount, we then perform the impairment evaluation using a more detailed quantitative assessment. We could also directly perform this quantitative assessment for any reporting unit, bypassing the qualitative assessment.

Our methodology for conducting the quantitative goodwill impairment testing is fundamentally based on the measurement of fair value for our reporting units, which inherently entails the use of significant management judgment. For valuation, we use a combination of the income approach (discounted cash flows) and market approach (market multiples) in estimating the fair value of our reporting units.

When preparing discounted cash flow models under the income approach, we estimate future cash flows using the reporting unit’s internal multi-year forecast, and a terminal value calculated using a growth rate that we believe is appropriate in light of current and expected future economic conditions. To discount these cash flows we use our expected cost of equity, determined using a capital asset pricing model. When using the market method under the market approach, we apply comparable publicly traded companies’ multiples (e.g., earnings, revenues) to our reporting units’ operating results. The judgment in estimating forecasted cash flows, discount rates and market comparables is significant, and imprecision could materially affect the fair value of our reporting units.

We could be exposed to an increased risk of goodwill impairment if future operating results or macroeconomic conditions differ significantly from management’s current assumptions.

We are subject to the income tax laws of the United States, its states and municipalities and those of the foreign jurisdictions in which we operate. These tax laws are complex, and the manner in which they apply to the taxpayer’s facts is sometimes open to interpretation. In establishing a provision for income tax expense, we must make judgments about the application of inherently complex tax laws.

Unrecognized Tax Benefits

We establish a liability for unrecognized tax benefits, which are the differences between a tax position taken or expected to be taken in a tax return and the benefit recognized in the financial statements.

In establishing a liability for an unrecognized tax benefit, assumptions may be made in determining whether, and the extent to which, a tax position should be sustained. A tax position is recognized only when it is more likely than not to be sustained upon examination by the relevant taxing authority, based on its technical merits. The amount of tax benefit recognized is the largest benefit that we believe is more likely than not to be realized on ultimate settlement. As new information becomes available, we evaluate our tax positions and adjust our unrecognized tax benefits, as appropriate.

Tax benefits ultimately realized can differ from amounts previously recognized due to uncertainties, with any such differences generally impacting the provision for income tax.

Deferred Tax Asset Realization

Deferred tax assets and liabilities are determined based on the differences between the financial statement and tax bases of assets and liabilities using the enacted tax rates expected to be in effect for the years in which the differences are expected to reverse.

Since deferred taxes measure the future tax effects of items recognized in the Consolidated Financial Statements, certain estimates and assumptions are required to determine whether it is more likely than not that all or some portion of the benefit of a deferred tax asset will not be realized. In making this assessment, we analyze and estimate the impact of future taxable income, reversing temporary differences and available tax planning strategies. These assessments are performed quarterly, taking into account any new information.

Changes in facts or circumstances can lead to changes in the ultimate realization of deferred tax assets due to uncertainties.