Having a bank account is important for the personal economic growth and independence of American individuals and households. Unfortunately, not everyone in America has an account. Those without them are referred to as the unbanked, and they are predominantly poorer Americans. Solving the problem of the unbanked is a major public policy challenge. Fortunately, a solution is materializing: “Bank On” is a national program, being implemented through many local Bank On coalitions in collaboration with the Cities for Financial Empowerment (CFE) Fund, whose goal is to advance financial inclusion through expanded access to low-cost bank transaction accounts.
Some of the core features of the program include offering simple checking accounts with low minimum balances and no overdraft fees. Currently, more than 100 banks, comprising 52 percent of national deposits, offer Bank On certified accounts.
A previous BPI research note presented a descriptive overview of the program and provided a geographic analysis of take-up and usage of Bank On certified accounts, based on a pilot data collection conducted in 2017 by the CFE Fund and the Federal Reserve Bank of St. Louis. This research note focuses on the program’s importance as a conduit for financial inclusion as measured by account take-up rates in relation to financial inclusion metrics across states.
The study relies on data from larger and more formalized (post-pilot) collections in 2018 and 2019 (mostly on the latter), analyzed jointly with data from the Federal Deposit Insurance Corporation (FDIC) on household use of banking products and services. The analysis demonstrates a significant positive correlation between the rate of Bank On account openings (net of closings) and the proportion of unbanked households across states, consistent with Bank On being a successful program to enhance financial inclusion.
In addition, as in the prior BPI note, the results show relatively high take-up rates for Bank-On accounts in areas with high concentrations of low- or moderate-income (LMI) and minority households. For instance, 42 percent of account openings in 2019 occurred in neighborhoods (ZIP codes) with more than 50 percent LMI population (whereas only 21 percent of the branches of the associated banks are in majority-LMI neighborhoods).
The study also finds that take-up of Bank On accounts has been material in relation to the size of the unbanked household population. Specifically, a net addition of more than 750,000 Bank On certified accounts during 2019 was recorded at banks that participated in the 2019 data collection, in states where these banks have a major presence (where their share of bank branches is at least 20 percent). Relative to the 3.3 million unbanked households as of mid-year 2019 in these states, this is a substantial take-up rate. Notably, the study sample understates the full contribution of Bank On. Although the 2018 and 2019 Bank On data collections included four of the largest U.S. retail banks (including the top three), many small and regional banks that participate in Bank On are still in the process of being incorporated into future data collections.
The Federal Reserve Bank of St. Louis and the CFE Fund launched a data collection and pilot study in 2017 to examine take-up and use of Bank On accounts, with participation of four financial institutions: Bank of America, JPMorgan Chase, U.S. Bank and Wells Fargo. The pilot study demonstrated how data on Bank On certified accounts could be collected in a consistent and centralized manner across institutions.
Building on this experience, the CFE Fund and St. Louis Fed in 2018 initiated an annual, standardized data collection to establish and maintain a “National Data Hub” to support analytics that advance understanding of the market for these accounts. The four large banks from the pilot study were joined by six community and regional banks (Carrollton Bank, First Commonwealth Bank, IBERIABANK, Old National Bank, Southern Bancorp, and The First – A National Banking Association) as participants. The same group of institutions participated in a second annual collection for the National Data Hub in 2019.
Both the 2018 and 2019 data, fully disaggregated to the ZIP code level, were made publicly available just in the past two months (subject to exclusion of ZIP codes for which fewer than three of the participating banks are reporting data). These samples constitute the core data for our analysis. Some summary information about the samples is presented in Table 1, alongside corresponding information about the full (non-public) sample from reports published by The Federal Reserve Bank of St. Louis.
Table 1: Summary information from the National Data Hub 2018 and 2019 samples
The 2018 and 2019 public data samples span 11,321 and 15,652 ZIP codes, representing 27 and 37 percent of U.S. ZIP codes, respectively. The restriction to ZIP codes with at least three reporting institutions reduces geographic coverage by more than half compared to the full, non-public samples, but results in the exclusion of no more than 10 percent of accounts by any of the three measures shown in Table 1 (accounts ever opened, accounts currently open, or accounts opened during the year.)
Bank On account take-up activity at the participating banks increased sharply in 2019. As reported for the full sample, about 2.6 million Bank On certified accounts were open at these banks as of year-end 2019, double the number open as of year-end 2018. About 1.9 million new accounts were added in 2019, of which 85 percent were opened by customers new to the institution, while just under 800,000 were added in 2018, with 75 percent of these opened by customers new to the institution.
Data merges. The next step is to merge the Bank On data with information on the demographic and income composition of each ZIP code and with financial inclusion data, by state, from the FDIC’s survey.
Neighborhood demographic information is obtained, by census tract, from the 2010 Decennial Census dataset. The demographic measures utilized are share of the population that identifies as non-white (Black or African American, Hispanic, Asian-American, or Native American or Native Hawaiian). These tract-level variables are allocated to ZIP codes using the tract-to-ZIP mapping available from the U.S. Department of Housing and Urban Development (HUD).
The share of a ZIP code’s population that is low- or moderate-income (LMI) is based on low- or moderate-income population share by census tract, which is derived from the 2011-2015 American Community Survey (ACS), based on Census 2010 geography, and obtained from HUD. These tract-level LMI shares are then allocated to ZIP codes by again using the HUD mapping.
Selected information on financial inclusion status of the population is merged in, by state, to complete the study sample. Financial inclusion metrics by state are obtained from the biennial FDIC Survey of Household Use of Financial Services. This survey gathers data on the number of U.S. households that are unbanked and underbanked, their demographic characteristics, and their reasons for being unbanked and underbanked. Our analysis utilizes data from the 2013 through 2019 survey samples.
An unbanked household in the survey is one in which no one in the household has a checking or savings account at a bank, thrift or credit union. The survey also collects information on whether a household has (during the previous 12 months) relied on a type of nonbank financial service company that typically is associated with providing high-cost products or services, such as payday loans or check-cashing. For our analysis, we characterize a household as underbanked if it is either unbanked or has utilized alternative providers of transactions services; specifically, check-cashing or money orders. Note that in other contexts, individuals utilizing alternative providers of high-cost credit also are generally classified as “underbanked,” but the focus here is on the degree to which households have access to affordable deposit or transactions accounts.
Table 2 shows estimated counts of unbanked and underbanked households based on the 2019 FDIC survey, the percentages of U.S. households these correspond to and the estimated proportions of unbanked and underbanked households in the areas covered by our (merged) study sample. The survey results indicate that 5.4 percent of U.S. households are unbanked, and another 12 percent tend to utilize alternative providers of transactions services and therefore are included in our definition of underbanked. Nearly the same proportions are observed for our study sample.
Table 2: Financial inclusion statistics
Bank On Take-up Activity in Relation to Financial Inclusion and Other Metrics
The Bank On program as a conduit for financial inclusion can be assessed by quantifying account take-up activity in relation to the proportion of unbanked households across states or regions. To the extent that the program has been effective at expanding financial inclusion, take-up activity should be greater in areas with larger financial inclusion gaps.
This section examines Bank On account take-up activity in 2019 in relation to the proportion of unbanked households across states. The focus on 2019 is motivated by the greatly accelerated pace of account openings occurring that year.
An intuitive measure of account take-up activity is the rate of account openings net of closings, calculated as number of accounts opened minus the number closed during a year, divided by the beginning-of-year number of open accounts. A higher net rate of opening of Bank On accounts in areas where there is a higher concentration of unbanked households would suggest a favorable effect on financial inclusion.
Chart 1 presents a scatter plot of the net rate of account opening in 2019 in relation to the proportion of unbanked households by state, for the 45 states (inclusive of the District of Columbia) where the group of participating banks has a material branch presence (each dot represents a state). As indicated by the positively sloped trendline, on average the net rate of account opening is higher in states with a larger concentration of unbanked households, reflected in a correlation coefficient of 31 percent.
Table 3 presents results from a regression equation for the relationship in Chart 1 and several related regression equations. Summary statistics for the state-level variables used in these regressions are shown in Table 4.
Table 3: Estimated regression equation for account take-up activity
Dependent variable: net rate of account opening
* Statistically significant at 10 percent level ** Statistically significant at 5% level
Table 4: Summary statistics for the state-level data
Source: National Data Hub 2019 public data.
The regression equation in column 1 of Table 3 provides a statistical representation of the data in Chart 1. The estimated coefficient on proportion of unbanked households is positive and statistically significant, consistent with the visual relationship observed in the chart.
Chart 1: Net rate of Bank On account opening in relation to unbanked share of the population in 2019
The regression equation in column 2 appends lagged (2018) account take-up activity to the column 1 equation. Inclusion of the states lagged take-up rate controls for factors that might cause atypically high (or low) take-up rates in some states to persist over time or, conversely, to moderate (return to a more typical rate) over time. The results confirm a strong, positive association across states between the net rate of account openings and percentage of unbanked households, as indicated by statistical significance (at the 5 percent level) of the estimated coefficient. The estimated coefficient on lagged take-up activity is negative and statistically significant (at the 10 percent level), consistent with a reversion-to-mean effect.
The column 3 equation swaps in unbanked plus underbanked share as the dependent variable for the column 2 equation. The equation in column 3 is identical to that in column 2 but is estimated using the full sample of states. Together, these results demonstrate robustness of the relationship between rate of account opening and financial inclusion.
Take-up by neighborhood classification. Since unbanked individuals and households are concentrated among lower-income and minority populations, another way to assess the reach of Bank On into the unbanked population is to examine the distribution of account take-up at the neighborhood level. This was the approach utilized in the previous research note, and we update that analysis here.
Table 5 presents the updated findings on 2019 account take-up activity by neighborhood classification. Three take-up metrics are considered alongside two comparison benchmarks.
Table 5: Account take-up activity as of year-end 2019 by neighborhood category, relative to benchmarks
Source: National Data Hub 2019 public data and 2010 Decennial Census
The take-up metrics are the ZIP code category’s share of ever-opened accounts; the category’s share of accounts open as of year-end; the category’s share of new accounts opened in 2019. In addition, the table shows the percentage of new accounts within each category that are associated with customers new to the institution.
The first comparison benchmark is based on the study sample’s ZIP code distribution: the percentage of sample ZIP codes within each category. The second benchmark is based on the overall branch location distribution of the 10 reporting institutions: the percentage of total branches (including branches with ZIP codes that are not in the analysis sample) within each ZIP code category.
Consistent with the findings based on the pilot sample discussed in the previous BPI research note, Bank On certified accounts disproportionately are associated with areas that have predominantly minority populations. About 57 percent of accounts ever opened, currently open (as of year-end 2019) or opened in 2019 were in ZIP codes that are more than 50 percent minority, more than double the share of participating institutions’ branches within this neighborhood category.
Likewise, Bank On certified accounts are disproportionately associated with areas that are predominantly LMI. About 44 percent of ever-opened accounts, and about the same percentages of accounts currently open and of those opened in 2019, were in ZIP codes with a greater than 50 percent LMI population, compared to 21 percent of the participating institutions’ branches.
Materiality of account take-up activity. The preceding analysis suggests that Bank On account take-up activity is greater where there is a larger concentration of unbanked households, consistent with the program’s goal of reaching the unbanked. But is the program sufficiently scalable to have a material impact on financial inclusion?
Because many of the Bank On institutions have yet to participate in the data collection, the study sample will understate the program’s full reach in states where the 10 participating banks have a relatively minor presence. Therefore, to address the question of materiality, we first restrict attention to the 18 states where the banks that participated in the 2019 data collection together account for at least 20 percent of bank branches statewide.
In these states, there were about 3.3 million unbanked households as of mid-year 2019 based on the FDIC survey, and the 2019 study sample records a net addition of about 763,000 Bank On accounts during the year at the participating banks. Thus, the scale of the program appears substantial, with a take-up rate in 2019 equal to 20 percent of the unbanked population in these states.
The Bank On program, coordinated by the CFE Fund, seeks to connect traditional banks with unbanked and underbanked individuals and families. This note highlights a strong correlation between Bank On account take-up activity and the percentage of households that are unbanked across states. Also, along with a previous research note, it shows that take-up activity has been particularly strong in neighborhoods with predominantly minority and lower-income populations.
As the analysis is based on data from just 10 of the more than 100 banks currently offering Bank On accounts, it provides only a partial view. As the program builds on its current momentum, and as more of the Bank On institutions join the National Data Hub effort, a fuller picture will emerge. Nonetheless, the findings are suggestive of a material role of the Bank On program in fostering financial inclusion. These findings argue for policymakers to take further action to publicize and encourage even broader customer usage of Bank On accounts, and for more banks to consider joining the Bank On movement.
 A list of the 23 account activity metrics included in the annual data collection was included in the previous BPI research note on Bank On. There are three categories of metrics: account opening and closing; account usage and consistency; and online or digital access. These are collected from each institution at the ZIP code level.
 For each ZIP code, the weighted average of each measure is calculated using the weights provided by the HUD mapping. This mapping, which weights each Census tract fully or partially contained within a given ZIP code by the tract’s share of the total residential population in that ZIP code, is available here.
 It is conducted by the U.S. Census Bureau as a special supplement to the Current Population Survey (CPS).
 The survey collects information on both credit and transactions dimensions of underbanked households.
 We exclude six states where their share of total bank branches is less than 5 percent.
 For these calculations, 266 ZIP codes associated with 5,474 open accounts are dropped from the sample due to missing demographic information.
 For additional context, note that about 11 percent of U.S. ZIP codes are majority minority and about 17 percent are majority low- or moderate-income.