Part 3: Significant Errors in the CFPB’s Cost-Based Calculation of the Safe-Harbor Limit
The Consumer Financial Protection Bureau is authorized under the Credit Card Accountability Responsibility and Disclosure (CARD) Act of 2009 to set a safe harbor for credit card late fees such that a fee within this safe harbor is presumed to be permissible under the statute. The current safe harbor limits, $30 for a first violation and $41 for a subsequent violation within the next six billing cycles, derive from rules promulgated in 2010 by the Federal Reserve Board based on analysis of the economic role of card late fees. The Bureau, on Feb. 1, released a Notice of Proposed Rulemaking on Credit Card Penalty Fees that proposes to reduce the safe harbor limit to $8 for any violation.
The proposed, large reduction in the safe-harbor limit rests on a purported, cost-based calculation by which the Bureau determines that an $8 fee suffices to cover the costs associated with late payments. However, the Bureau’s calculation is seriously flawed, due to major gaps and errors in the data used and the procedure applied. When these omissions and errors are accounted for, the proposed $8 limit is revealed to be far too small to cover banks’ costs, as explained below. The Bureau should rescind its proposal, correct its errors and omissions and ensure that any re-proposal is based on sound conceptual and empirical analysis.
The Bureau’s calculation includes four types of significant errors or omissions, each of which may cause substantial upward bias in the calculated ratio of late fees paid to the costs of late payments. These are:
- The data used by the Bureau excludes many of the costs associated with late or missed payments, including many costs related to credit card collections, resulting in an overstatement of the measured payment-to-cost ratio.
- The Bureau’s exclusion of collection costs incurred after an account is charged off is arbitrary and lacks economic justification.
- The Bureau calculates the industry average fees-to-cost ratio weighting each individual bank’s ratio by number of accounts; this can make the calculation sensitive to outliers among individual banks.
- The Bureau does not adequately consider the macroeconomic conditions during its selected period of analysis, which is not representative of a typical economic cycle. As a result, the measured fees-to-cost ratio appears to be higher than its long-run average.
This note, the third in a series, demonstrates how the data and procedure used in this calculation are inadequate and insufficient for determining that an $8 limit would cover issuers’ costs related to late or missed payments.
Also, as described in the previous two BPI notes in this series but not repeated here, by basing the safe-harbor limit solely on this cost calculation, the Bureau does not meaningfully consider the role of late fees in deterring cardholders from making late payments, in violation of the statute.
For its calculation, the Bureau relies primarily on data collected by the Federal Reserve Board for stress testing purposes, the Y-14M data collection. The calculation uses two data items from the Y-14M: “Total Non-Interest Expense – Collections Expense” and “Fee Income – Late Fee Income.” Additionally, the Bureau relies on data collected from a survey of the debt collection and loss mitigation practices of several large issuers, conducted as part of its biennial review of the consumer credit card market.
The calculation proceeds as follows. First, the Bureau excludes from the calculation of costs associated with late payments any costs associated with collection activities that occur after an account has been charged off. The Bureau’s decision to exclude these costs appears to be based on a seemingly arbitrary interpretation of commentary in the current regulation, rather than on an economic argument. Next, the Bureau estimates, using the above two data sources, the share of total collection costs that derive from commissions paid to collection agencies for recoveries on charged-off accounts, as a proxy for costs incurred after charge-off.
The Bureau then applies the latter estimate to Y-14M data and proceeds to calculate the ratio of net assessed fees to before-charge-off collection costs for banks that report these data, by month, for each month in January 2016 through March 2022. The Bureau observes that since August 2021 “late fee income has exceeded the relevant estimated pre-charge-off costs more than fivefold” and that this “resembles” what is observed for the 2016-2019, pre-pandemic period.
Based on these observations, the Bureau infers that the average net assessed fee is at least five times larger than necessary to cover the typical issuer’s pre-charge-off collection costs. Therefore, since most issuers currently charge late fees ranging as high as $40 as permitted under the safe harbor for a subsequent violation, the Bureau concludes that an $8 late fee “would still recover the average issuer’s pre-charge-off collection costs, as that fee represents one-fifth of the maximum late fee amount, which is necessarily greater than average fee income per late payment.”
The discussion below explains each of the aforementioned shortcomings with this analysis in turn. The correct calculation would imply a markedly higher penalty fee to cover the costs associated with late or missed payments compared to the proposed $8 limit.
The Bureau fails to consider the limitations of the Y-14M data items.
For data on collection expenses, the Bureau relies solely on the line item in the Y-14M, “Collections Expense.” The instructions to filers are simply: “Report costs incurred to collect problem credits. Include total collection cost for delinquent, recovery, and bankrupt accounts.”
The Bureau did not investigate what specific costs issuers actually report for this line item or how comprehensive the reporting is in relation to all relevant costs associated with late payments. Moreover, the Bureau did not account for the possibility that reporting of this line item differs across issuers. From informal discussions with BPI member banks, it is clear that the line item is far from comprehensive, with many relevant costs omitted.
Table 1 outlines the different types of costs associated with credit card collections or related activity. In general, however, banks’ Y-14M reporting incorporates only the variable costs borne directly by the collections department, which is the first group in the table. Banks do not report on the Y-14M most of the other types of collections-related costs listed in Table 1, such as fixed costs and supporting services from other areas of the bank, although the specifics vary across banks. Consequently, the line item in the Y-14M typically omits a large share of the overall cost of collection or related activities.
Table 1: Cost Categories for Credit Card Collections
It is worth noting that the Y-14M schedules were designed for supervisory stress testing and related regulatory purposes, and for this context it is not necessary that the reported collection cost item include each cost item in Table 1. For stress testing, it can be appropriate to allocate various such costs to the bank department that bears them or to the general PPNR (pre-provision net revenue) stress testing category. The Y-14M data collection was not designed or intended to be used for consumer compliance regulation.
The specific support services for which a collections department relies on other areas of the bank vary across institutions. For example, some banks have substantial, dedicated IT resources within their collections department, while others rely entirely on outside IT support. Some have their own internal audit and control teams, relying only partially on outside support, while others rely entirely on support services from the bank’s audit department.
Similarly, some collection departments arrange and operate their own training programs for servicing agents, while others rely on the HR department to arrange and operate such programs. The share of collection and recovery costs included in the reported Y-14M item would also depend on the extent to which a bank sells delinquent accounts to companies that specialize in managing troubled debt if these sale transactions are primarily conducted outside of the collections department.
Commissions and fees paid to external agents and vendors typically are included in the Y-14M item. Therefore, a bank that relies more heavily on vendors and outside agents generally will have more complete reporting of collections expenses. However, there are exceptions to the latter rule; for example, not all BPI reporting banks include commissions paid to third party collection agencies after a loan has been charged off. Because the Bureau’s calculation assumes that the Y-14M line item consistently incorporates such commissions and applies an adjustment to the data to net out these costs, the calculation will be incorrect as applied to these banks, further biasing upward the calculated fees-to-cost ratio.
In addition, there are costs associated with late payments—especially those completed within 30 days of the due date—that are borne outside of the collections department. These costs are typically incurred by the customer service area, which handles calls and inquiries from those seeking information about their late payment or requesting fee waivers. Those costs are likely not included in the Y-14M item, because they are not “collection” costs or generally associated with delinquent accounts.
Based on these considerations, the degree to which the Y-14M collection cost item incorporates all relevant costs of late payment will vary across issuers, but in general the reported collection cost item omits large, fixed costs and various support costs and thereby substantially understates overall collection costs. Banks that report a less comprehensive share of costs will have a higher fees-to-cost ratio.
An additional data issue is that the Y-14M item for fee income is the sum of fees assessed during the month minus fee reversals and refunds applied during the month (which include reversals due to charge-off). However, in accordance with banks’ loss mitigation practices, each month some delinquent accounts may be modified through re-aging or converted into fixed payment plans while others may be closed in a debt settlement, without explicit reversal of late fees but with concessions to the borrower. These implicit reversals of fee income are not captured in the Y-14M item for net fees assessed, which therefore overstates realized fee income.
The Bureau is wrong to exclude collection costs incurred after charge-off.
As noted earlier, the Bureau excludes costs incurred after an account has been charged off in calculating the fees-to-cost ratio used to derive the proposed safe-harbor limit. These costs make up a material portion of overall collection costs.
To justify the exclusion of these costs, the Bureau refers to Federal Reserve Board commentary on the original 2010 rulemaking which suggests that “losses and associated costs” should not be factored into the “cost of the violation” for determining whether a penalty fee is reasonable and proportional. The Bureau then interprets the term “losses and associated costs” to include collection expenses incurred after charge-off.
This attempt to expand the definition of “losses and associated costs” to include collection costs incurred after charge-off is inconsistent with business practice and logic. Collections activity is an ongoing business activity of a credit card lending institution. The debt collections process commences shortly after a payment due date is missed, typically as an account nears 30-day delinquency, and continues after charge-off in much the same manner as before, as is well documented in the Bureau’s biannual Consumer Credit Card Market Report.
Collections activity remains an ongoing business expense of the bank if and when a bank contracts with an outside collection agency to pursue recovery on credit card debt, whether or not the debt has already been charged off. Such third-party agents act on behalf of the bank and are paid commissions for the amounts recovered. There is no defined period after which the cost of collecting on delinquent or charged-off debt ceases to be an operating cost of the bank.
Nor do lending institutions automatically transition from internal to third-party collection activity when a charge-off occurs, and some banks continue to rely on their internal collections function for a material portion of charged-off card debt. Some banks also rely on third-party collectors for a material portion of debt not yet charged off.
An analogy can be made to companies that specialize in loan servicing. Collecting on delinquent and defaulted debt is an integral part of the business model of such companies, and the costs of collection activities are an ongoing cost of their business operations, with no discontinuity arising from a charge-off. Costs of continuing to collect on loans that have charged off are ongoing operating costs for these companies.
Another important factor to consider is that when a consumer misses a payment date, from the point of view of the card issuer (who cannot know the reason for the missed payment), the probability that the account may eventually go to charge-off increases. Economic principles imply that issuers should update their risk assessments and risk pricing when consumers pay late, including by setting late fees to cover potential future costs, including costs after charge-off.
Moreover, the Bureau’s decision to exclude costs incurred post charge-off has potentially adverse implications for consumers. These costs won’t disappear simply because the Bureau places them off-limits to reimbursement through late fees. Banks will seek other ways to offset them.
One likely response of lenders would be to increase up-front interest rates or make credit cards less available, especially for consumers who are seen as less creditworthy due to having lower credit scores. Thus, while consumers with prime credit scores who on occasion pay late (and by doing so indicate an increased risk of default) now share substantially in these costs, the Bureau’s proposal will shift more of that share onto consumers with lower credit scores that may never pay late.
In sum, in deciding whether a late fee is reasonable and proportional, the Bureau provides no rationale for excluding post-charge-off (collection) costs.
The Bureau fails to consider the disadvantages of weighting by number of accounts.
The natural and logical way to calculate an overall, industry average fees-to-cost ratio would be to divide total net assessed fees of all reporting institutions to total costs; that is, to calculate an aggregate fees-to-cost ratio. However, the Bureau calculates an industry average ratio not as an aggregate ratio, but by first calculating each individual bank’s fee-to-cost ratio and next weighting each institution according to the number of credit card accounts it has. The resulting weighted average of individual banks’ ratios can be overly sensitive to the presence of outlier banks with relatively large fee-to-cost ratios, potentially resulting in upward bias relative to the aggregate ratio.
Figure 1 compares the Bureau’s account-weighted average ratio to the Y-14M aggregate ratio, where the latter equals the sum of net assessed fees across all Y-14M reporting institutions divided by the sum of before charge-off collection costs of these institutions. The account-weighted ratio is relatively volatile and is also systematically larger, consistent with the concern that it may be overly influenced by a few outlier institutions with unusually high and volatile fee income relative to costs. Moreover, it seems plausible that measurement error tied to excluded costs may contribute to the presence of such outliers—a few large institutions may have a relatively large, missing cost component, resulting in large overstatement of their ratios and greater sensitivity of their ratios to variation in fee income over time. Overall, the Figure 1 comparison raises a concern that weighting by number of accounts yields a less reliable estimate of the ratio due to enhancing the influence of outliers, including by augmenting the upward bias caused by the measurement error in reported collection costs.
Figure 1: Weighted Average versus Y-14M Aggregate Fees-to-Cost Ratio
Ratio of net assessed fees to the sum of before charge-off collection costs.
Source: Revenue and Collection Costs at Large Bank Holding Companies (consumerfinance.gov)
The Bureau does not discuss these potential limitations of using the account-weighted average for deriving the proposed safe harbor limit or explain why it chose to calculate it this way. Nor does the Bureau provide any information on the distribution of the measured ratios across institutions; why they appear to be correlated with institution size (number of accounts); the reasons for volatility of the account-weighted average ratio; or whether and how outliers may have been addressed. Because the account-weighted average might be sensitive to outliers and exacerbate effects of measurement error in the data, it would seem preferable for the Bureau to utilize the Y-14M aggregate ratio (although, of course, this would not resolve the overall significant shortcomings of using the Y-14M data for the analysis).
Moreover, use of the aggregate ratio generally is more appropriate even if the gap between the account-weighted average and Y-14M aggregate ratios is not solely due to measurement error. Even with the individual ratios measured accurately, if a minority of relatively large institutions have comparatively large ratios, the Y-14M aggregate ratio still would be the more appropriate measure for the purpose of setting a safe-harbor limit applicable to all institutions.
Two hypothetical examples, presented in Table 2 illustrate why this is so. Each of these examples posits banks with differing ratios of total fee income to total collection cost, with a weight representing their market share of accounts. The gap between the account-weighted average and the Y-14M aggregate ratios is similar to that observed in the actual data in Figure 1 since June 2021, and reflects the influence of a single large bank with a relatively large fees-to-cost ratio (Bank 1).
The last two columns show the result of a fee cap based on the weighted average ratio (dividing current fee income by this ratio), which is essentially what the Bureau has proposed doing. Each bank’s fee income would decline to what is shown in the next to last column, resulting in the new fees-to-cost ratio shown in the last column.
Table 2: Examples illustrating effects of using the account-weighted average
Because of the outsized influence of Bank 1 on the account-weighted average ratio, Banks 2 and 3 are unable to recoup their costs, and aggregate fee income falls short of aggregate costs. This outcome is neither sensible nor fair. To the extent that Bank 1’s high ratio is due to measurement error, this outcome would be even less desirable.
The Bureau does not adequately consider the macroeconomic context around its analysis.
The Bureau’s analysis of credit card late fee income and collection costs uses monthly data from January 2016 through March 2022, with an emphasis on the sub-period since August 2021. However, this period is not representative of the longer-term performance of the credit card market. It was relatively benign in terms of credit card delinquency and loss, and particularly so in the sub-period since August 2021. Consequently, the Bureau’s analysis is flawed.
This limitation of the analysis is evident from Figure 2, which depicts the historical time series of credit card charge-off rates since January 1990. During the Bureau’s analysis period (post-2015), charge-off rates were well below the long-term average of about 4.5 percent, with no major downturn affecting household debt and repayment performance occurring during this period. After 2020, charge-off rates plummeted to unprecedentedly low levels, reflecting the drop in consumer spending alongside the income transfers and other government support for households that occurred during the COVID-19 pandemic.
Because of this benign environment, the data very likely understate quite materially the long-run average fees-to-cost ratio. In particular, during a period of stress, collection costs would greatly increase as banks expand their collection activities in line with rising delinquency. Late fee income would likely decline or at least remain flat (with the increase in assessed fees being offset by increased reversals due to the spike in charge-offs).
Figure 2: Quarterly Charge-Off Rate on Credit Card Loans—Q1 1990 through Q3 2022
All commercial banks, seasonally adjusted.
Source: Federal Reserve Economic Data (https://fred.stlouisfed.org)
Note: The Bureau’s analysis period commences in January 2016, indicated by the blue vertical line.
Although the Bureau’s analysis period is relatively benign for credit losses, there is enough variation in the data to shed some light on the relationship among repayment performance, net assessed fees and collection costs. Figure 3 depicts the quarterly, 90-day-plus balance delinquency rate (delinquent as a share of total balances) for this period. Figure 4 shows total collection costs and total net assessed fees for this period relative to (normalized by) their January 2016 values.
Figure 3: Quarterly Delinquency Rates on Credit Card Balances—Q1 2016 through Q1 2022
Source: Federal Reserve Bank of Philadelphia, Large Bank Consumer Credit Data
Figure 4: Total Collection Costs and Total Net Assessed Fees: January 2016 through March 2022
Data are expressed as a ratio to the January 2016 value.
Source: Revenue and Collection Costs at Large Bank Holding Companies (consumerfinance.gov)
Collection costs trend upward through April 2021, consistent with the rising delinquency rate into the first quarter of 2021. After January 2021, collection costs trend downward, consistent with the sharp drop-off in delinquency, providing more evidence that collection costs are influenced by macroeconomic factors.
Net assessed fees are quite stable through this period (subject to seasonal variation), consistent with the increase in assessed fees being offset by increased frequency of reversal due to charge-off. Net assessed fees are volatile since the start of the pandemic, reflecting the various unusual factors at play during this period, including forbearance and expiration of forbearance toward skipped payments.
Bottom line assessment
The Bureau’s finding that card issuers’ late fee income has, on average, exceeded the costs associated with late payment by a multiple of no less than five is based on an inaccurate and upward biased calculation of the ratio of fees to costs.
One major problem with the calculation is that it relies on data from 2016 through early 2022, which was relatively benign and does not incorporate a stress period with respect to credit card delinquency and losses. Thus, the calculation cannot be considered representative of the long-term average relationship between fee income and costs. Another problem is that in deriving an industry average fees-to-cost ratio, the Bureau weights each institution by number of accounts, which appears to exacerbate effects of measurement error and outliers.
Looking again at Figures 3 and 4, note that the delinquency rate reaches its maximum level in the first quarter of 2020, and collection costs likewise. Thus, the first quarter of 2020 is closest to “normal” credit card market conditions that can be observed during the analysis period. For this quarter, the industry aggregate ratio of fee income to collection costs prior to charge-off is 3.95 (see Figure 1).
Thus, adjusting only partially for the influence of the benign market conditions and swapping in the industry aggregate ratio for the account-weighted average brings the calculated ratio down from 5 to about 4. However, this still excludes the Bureau’s estimated amount of collection costs incurred after charge-off. The Bureau has not provided a compelling justification for excluding these costs and their exclusion conflicts with business practice and economic reasoning. Reincluding these costs brings the corrected ratio down to 3. Yet, this is still a substantial overestimate, due to the remaining problems: the omission of many relevant and important cost categories from the Y-14M collection cost item and the lack of data from a stress period. Factoring those considerations into the calculation would reduce the ratio much further.
The Bureau’s estimate that an $8 late fee suffices to cover the costs of late payment is grossly inaccurate. The Bureau at present lacks sufficient information to conduct an appropriate, cost-based calculation. If the CFPB had better or more complete information, and also corrected for the omissions and errors this note has highlighted, it’s not unreasonable to think a revised calculation could even confirm the current safe harbor limits ($30-$41). And of course, this calculation would still only reflect the costs associated with collecting or servicing late or missed payments. As discussed in previous BPI blog posts, the Bureau also failed to meaningfully consider the statutorily mandated factors of deterrence and consumer conduct, casting further doubt on the justification for an $8 safe harbor limit.
 More specifically, a fee within the safe harbor is presumed to be reasonable and proportional to the customer’s violation or omission with respect to the card agreement.
 Shortly thereafter, the Bureau was created and assumed responsibility for administering safe harbor.
 The Bureau notes that as part of its biennial review, it “surveys several large issuers to better understand practices and trends in credit card debt collection.” Data collected in response to data filing orders for the 2021 review were used as the basis for estimating the share of total collection costs that are incurred after charge-off.
 For further details, see the Bureau’s summary of the calculation in “Credit Card Late Fees: Revenue and Collection Costs at Large Bank Holding Companies”, Consumer Financial Protection Bureau (February 2023), available at: Revenue and Collection Costs at Large Bank Holding Companies (consumerfinance.gov)
 See page 43 of the NPR.
 On page 44 of the NPR, the Bureau states: “Given the finding that, in the most recent data, late fee income is greater than five times estimated pre-charge-off costs, the Bureau expects that an $8 late fee would still recover the average issuer’s pre-charge-off collection costs, as that fee represents one-fifth of the maximum late fee amount, which is necessarily greater than average fee income per late payment.”
 See the discussion on page 29 of the NPR.
 Another factor likely contributing to recent volatility of net assessed fees as reported in the Y-14M is that this line item nets out reversals taking place in the current month from fees assessed in the current month. Since reversals in the current month to a large extent are tied to late fees assessed in prior months, they may not accurately approximate future reversals of current assessed fees. Because many late fee reversals occur if and when an account is charged off, the Y-14M line item will overstate the realized net assessed fees when rates of transition from delinquency to charge-off are increasing and understate them when these rates are declining. This procyclical error can contribute to volatility of the measure.