Observations from the NCUA’s New Data on Non-sufficient Funds and Overdraft Fees

NCUA Office of the Chief Economist: Research Note¹

Background

Beginning with the first quarter of 2024 Call Report, NCUA began collecting information on year-to-date revenues from non-sufficient funds (NSF) and Overdraft (OD) fees. Credit unions with assets in excess of $1 billion have been required to submit their year-to-date revenues from OD and NSF fees on separate Call Report accounts.² The FDIC began requiring submission of such fee data in 2015. The FDIC collects the total of OD and NSF fees but does not provide a breakout of each fee type.

This Research Note analyzes statistics for OD and NSF fees, with a particular focus on evaluating OD and NSF revenues as a fraction of total revenues. Using revenues for the first three quarters of 2024, credit unions are put into categories based on the share of their revenue derived from these two sources. While other ratios could be used, this ratio measures the relative importance of such fees to the credit union’s overall revenue generation.³

It should be noted that supervisory-related inferences should not be made from the observations in this Research Note. “Higher” and “lower” levels of OD and NSF fee income in this Research Note are simply meant to express relative levels. As outlined in a recent Letter to Credit Unions,⁴ while certain types of practices and product features may raise supervisory concerns, in general OD and NSF fees do serve legitimate business purposes. Given varying business models and diverse consumer preferences and, in some instances, needs, it is reasonable to expect that some credit unions will charge such fees and that the reliance on such fees will vary.

This Research Note provides two particular observations on the relationship between OD and NSF fees and other revenues. The first looks at the relationship between OD and NSF fees and other types of fees collected by the credit union. The second evaluates the relationship between such fees and interest rates, as it analyzes net interest margins and interest rates on mortgages.⁵

Summary Statistics

For the 444 credit unions that supplied OD and NSF fee data to NCUA in 2024Q3, the table below provides summary statistics for the share of revenues accounted for by these fees. In general, the table indicates that such fees make up about 2 to 5 percent of total revenues for the majority of these credit unions, although some deviate significantly from this range. Federal credit unions (FCUs) and federally insured, state-chartered credit unions (FISCUs) are not significantly different in their reliance on such fees, with nearly identical medians and similar distributions. Credit unions that are Minority Depository Institutions (MDIs) and those with low-income designations (LIDs) tend to have slightly higher combined OD and NSF fee revenues as a share of total revenue relative to those without those designations.

Overdraft + NSF Fees as a Share of Total Revenue

Credit Union Type	Minimum	25th Percentile	Median	75th Percentile	Maximum
All Filers (> $1 Billion in Assets)	-2.5%	2.0%	3.5%	4.9%	18.2%
FCUs	0.0%	1.8%	3.5%	4.8%	15.2%
FISCUs	-2.5%	2.3%	3.5%	4.9%	18.2%
Neither MDI nor LID	0.0%	1.2%	2.8%	4.2%	15.2%
MDIs	-2.5%	1.1%	4.5%	7.4%	9.4%
LIDs	0.0%	2.7%	3.9%	5.2%	18.2%

The statistics shown do not control for other factors. A more in-depth analysis would be necessary to assess whether the relationships still hold after accounting for other factors (such as, credit union size, location). It is also important to recall that the collection of the OD and NSF data is recent, and more quarters of data are needed to fully understand the reliability of the reported information.

Observation 1: Credit unions with higher combined OD and NSF fees per member do not seem to have lower fees per member for other services.

As credit unions have numerous pricing “levers” to adjust to cover costs, some might ask whether credit unions with higher OD and NSF fees have correspondingly lower fees for other services. The chart below graphically evaluates that hypothesis by comparing OD and NSF revenues to other fees. Specifically, for the first three quarters of 2024, the scatter plot compares combined OD and NSF fees per member against other fees collected per member.⁶ The relative size of the dots is proportionate to the credit union assets as of 2024Q3.

Relationship between OD + NSF Fees ($ Per Member) and Other Fees ($ Per Member)

For each credit unions with Overdraft and NSF fee information, the scatterplot plots the sum of Overdraft and NSF fees per member (x-axis) against other fees per member (y-axis). The size of each dot is proportional to the assets of the credit union in 2024Q3. A trendline is included that reveals a slight upward slope, indicating a slight positive relationship between Overdraft and NSF fees and other types of fees.

The graph indicates that, at least in the first three quarters of 2024, there was little evidence of such an inverse relationship between the two types of fees. Indeed, estimating a simple linear model to the data (reflected as the dotted line in the graph), one finds a somewhat positive correlation: higher OD and NSF fees were associated with higher fees for other services.

Observation 2: Credit unions with higher combined OD and NSF fee revenues do not seem to be using those fees to “subsidize” better interest rates.

One might also ask whether credit unions with higher OD and NSF fees use those fees to facilitate offering more attractive interest rates. The graphs below suggest that may not be the case. Comparing OD and NSF fees to net interest margins for FCUs and FISCUs separately, the graphs indicate that higher combined OD and NSF fees are associated with higher net interest margins (NIMs). The relationship was much more pronounced for FCUs, where median NIMs increased 72 basis points from 2.59 percent for the credit unions with the lowest OD and NSF fees to 3.31 percent for those with the highest OD and NSF fees. For FISCUs, the range was 28 basis points (from 2.85 percent to 3.13 percent).

Relationship Between OD + NSF Fee Intensity and Net Interest Margin (FCUs)

The bar chart shows the median Net Interest Margins for five groups of Federally Chartered Credit Unions that filed overdraft and NSF fee information: All FCUs and four categories based on the share of total revenue comprised by OD+NSF fees. Those four categories include: OD+NSF fees comprise less than 2% of all revenues, OD+NSF fees comprise 2-3.5% of revenues, OD+NSF fees comprise 3.5-5% of revenues, and OD+NSF fees comprise more than 5% of total revenue. For all FCUs, the median Net Interest Margin is 3.01%. For the four categories, the median Net Interest Margins were 2.59% (<2% of Revenues), 3.03% (2-3.5% of Revenues), 3.1% (3.5-5% of Revenues), and 3.31% (>5% of Revenues).

Relationship Between OD + NSF Fee Intensity and Net Interest Margin (FISCUs)

The bar chart shows the median Net Interest Margins for five groups of Federally Insured, State-Chartered Credit Unions (FISCUs) that filed overdraft and NSF fee information: All FISCUs and four categories based on the share of total revenue comprised by OD+NSF fees. Those four categories include: OD+NSF fees comprise less than 2% of all revenues, OD+NSF fees comprise 2-3.5% of revenues, OD+NSF fees comprise 3.5-5% of revenues, and OD+NSF fees comprise more than 5% of total revenue. For all FISCUs, the median Net Interest Margin is 2.97%. For the four categories, the median Net Interest Margins were 2.85% (<2% of Revenues), 2.82% (2-3.5% of Revenues), 2.97% (3.5-5% of Revenues), and 3.13% (>5% of Revenues).

Analyzing net interest margins provides a high-level perspective on credit union pricing, but the relationship shown could simply be the result of differences in the overall credit risk of membership; higher average margins could merely reflect higher average credit risk. To pursue a more informative measure of the relationship between interest rates and OD and NSF fees, one might want to look at risk-adjusted interest margins. Such a comparison might be useful for assessing whether higher OD and NSF fees were subsidizing preferred pricing elsewhere (like, through lower loan rates).

While no comprehensive data are available for wholistic evaluation of risk-adjusted pricing, loan-level data submitted under the Home Mortgage Disclosure Act (HMDA) provide a readily available option for mortgages. Using detailed loan data from credit union HMDA filers for 2023, OCE has estimated the degree to which each credit union’s mortgage rates deviated from what would be expected given the institution’s loan characteristics.⁷ Those deviations might be correlated with risk-adjusted margins and thus can be compared to OD and NSF fees.

The graph below indicates that higher OD and NSF fee revenues were generally not associated with more favorable mortgage rates for members. Specifically, the scatterplot suggests that credit unions with higher OD and NSF fees (measured on a dollar-per-member basis) generally did not evidence lower mortgage rates in 2023; generally, a fitted line shows no clear relationship between the two. To be sure, this is a less-than-ideal analysis because it only looks at mortgages rates (instead of all interest rates) and only includes HMDA-filing credit unions. Also, the mortgage rates are from 2023, whereas the OD and NSF data are for 2024. Those imperfections aside, the qualitative result does not contrast sharply with what was shown in the NIM graphs: There is no obvious evidence of better interest rates “offsetting” higher OD and NSF fees.

Relationship between OD + NSF Fees ($ Per Member) and Mortgage Pricing

For each credit unions with Overdraft and NSF fee information, the scatterplot plots the sum of Overdraft and NSF fees per member (x-axis) against a measure of how the credit union's mortgage interest rates deviated from what would have been expected given its borrower characteristics (y-axis). Using loan-level data from the Home Mortgage Database (HMDA), the measure of deviation is the average difference between the observed interest rate and the rate expected rate produced by a statistical model that uses loan and borrower characteristics. Values above 0 indicate that rates tended to be above what would be predicted. In general, the scatterplot—and an embedded trendline--reveal no clear positive or negative relationship between NSF and OD fees and mortgage interest rates.

Conclusion

Although the analyses in this Research Note provide some high-level information about how credit union OD and NSF fees differ for various types of credit unions, the data collection will be especially valuable in tracking such fees over time. With only three quarters of OD and NSF data currently available, evaluating meaningful time trends is not yet possible. OCE plans to analyze evolving trends in those revenues as more data become available. To the extent that those trends—or other cross-sectional analyses of such fees—reveal notable observations, those findings may be made available in future Research Notes.

¹ The views and opinions expressed here do not necessarily reflect the views, opinions, or policies of the National Credit Union Administration or its Board.

² The Call Report collects those data in Accounts IS0048 (OD) and IS0049 (NSF).

³ Some observers have expressed fees relative to a credit union’s net income. (See, for instance, Do credit unions have an overdraft problem? | Credit Union Journal | American Banker). Such an approach has drawbacks, however. For accounting and other reasons, net income can vary sharply over time and thus the “importance” of such fees to a credit union’s overall strategy may appear to fluctuate dramatically. Relatedly, because credit unions are not-for-profit entities, their net incomes may be lower than they would otherwise be and thus the ratio may appear elevated. Take, for instance, a hypothetical multi-billion-dollar credit union that intentionally aims to earn de minimis net income each year. In such a situation, by “accounting for” a large fraction of net income, even a trivial amount of fee income could give the impression that fee income is critical to the credit union’s operation.

⁴ See Consumer Harm Stemming from Certain Overdraft and Non-Sufficient Funds Fee Practices | NCUA.

⁵ Other comparisons, while interesting, are beyond the scope of this paper. A simple assessment of net worth, for instance, suggests that Net Worth Ratios are not correlated with relative reliance on OD and NSF fees.

⁶ The prior analysis of fees as a proportion of revenue would not be appropriate here because, by construction, higher OD and NSF fee shares will tend to be associated with lower shares for other fees. Both types of fees are included in total revenues and thus there is some intrinsic endogeneity.

⁷ The “expected” rates are calculated using a simultaneous equation model, which jointly predicts interest rates and other mortgage pricing elements (like, loan costs). The model, which was described in Observations on Credit Unions’ Mortgage Lending to Minority Borrowers, deploys credit scores, loan-to-value ratios, geography indicators, and other credit-relevant factors. The estimates were calculated using 30-year fixed-rate loans.