The purpose of this paper is to study the “underbanked” – those who already possess bank accounts but are patrons of alternative financial services (AFS) providers at the same time.
Linking the FDIC unbanked/underbanked surveys of nationally represented households with FDIC bank information and local MSA demographics, demographic and economic profiles of the underbanked households are examined, together with the determinants of their choice of nonbank financial services.
The author finds that bank fees are associated with the likelihood for households to obtain AFS, especially nonbank credit. Households’ attitudes and experience with banks are important in the choice of getting AFS. Furthermore, most underbanked households used AFS temporarily, partly reflecting rather informed and calculated financial decisions.
The results from this paper provide implications for different types of AFS users. For example, the use of transactional AFS responds to the availability of online or mobile banking; meanwhile, it is also sensitive to branch closure. Users of nonbank credits are likely to be price savvy, and these products serve as valuable alternatives for short-term financing, especially during unfavorable economic situation.
Better understanding of the underbanked could help banks tailor to existing clients’ needs, for instance, providing innovative short-term credit products for those with little or impaired credit history. The study also helps policy makers re-evaluate banking regulations since the Great Recession. As regulations squeezed bank profits in certain areas and forced banks to consolidate, come alongside higher bank fees, potential branch closure and loss of service, which ultimately forced banked individuals to the less regulated alternative providers.
The analysis utilizes a comprehensive set of variables, from household social-economic characteristics to local banking industry characteristics, together with households’ subjective opinions of their banking institutions. The focus on the underbanked brings attention to this underserved population and discusses areas where banks can improve. The study contributes to the understanding of AFS users, draws implications for regulation toward banking and shadow banking.
Emerald Publishing Limited
Copyright © 2019, Emerald Publishing Limited
Alternative financial service (“AFS” thereafter) providers refer to the private financial institutions that are not federally insured and do not have to comply with many regulations that govern banking institutions. A variety of businesses such as check-cashing venues, pawnshops, rent-to-own stores belong to this category. The sector had witnessed tremendous growth in the last decade, with an estimated annual volume of transactions exceeding US$320bn (Bradley et al., 2009). Although AFS has been closely linked to the unbanked population, a large number of households who already own bank accounts also use AFS (aka the “underbanked”). In fact, around 20 per cent families with bank accounts have obtained financial service or product outside the banking system, according to the surveys conducted by FDIC supplemental to the current population survey (“CPS” thereafter) (FDIC, 2015).
Nonbank AFS providers have become a viable alternative source for families to obtain financial services, to fill in the need where traditional banking is not available. While banking advocates and policy makers have focused their attention and efforts on bringing unbanked population to mainstream banking, little is known about the underbanked group as to how and why they became underserved by their existing banking institutions. Interviews conducted at a check-cashing facility in New York in 2013 provided some insights as of why unbanked or underbanked customers chose AFS. For instance, the interviews showed that customers made informed decisions when choosing AFS providers over banks, which is different from the common perception of AFS users that they do not have basic financial knowledge to make sound financial decisions. The patrons cited required minimum balances imposed by banks, ATM fees and high account maintenance fees being the culprit of turning bank clients to AFS providers for basic services such as check cashing. While AFS could be costly, according to the interviewees, its predictable high flat fees were easier to manage than uncertain and unexpected charges by banks, especially the overdraft fees; most clients tend to live from check to check and could not always predict when the checks would clear. The customers also referred to “quality of service, trust and respect” they received from AFS providers as reasons for their choice of AFS over traditional banking.
Interviews of such kind are informational but limited to a specific sample – patrons of a check-cashing venue in New York. This study examines the “underbanked” using a large, nationally representative data set from the FDIC unbanked/underbanked survey supplemental to the CPS, covering all states and the District of Columbia. The survey asked the households of any type of AFS being used; therefore, the study is not limited to a particular type of service. I explore the determinants of banked household’s usage of AFS, including the socioeconomic household characteristics used in existing literature, and the factors suggested in the interviews such as service quality and bank service charges. The 2015 FDIC survey also contains survey questions on existing bank customers’ rating of their experience at the banks, such as whether they think banks were interested in serving them. FDIC bank-level data are used to create measure of service charge in the locality of the surveyed households. Other local banking market variables are constructed to measure the accessibility of banking branches in the local area, competitiveness of the local banking market and neighborhood socioeconomics. The regression analysis used in the study, therefore, includes a comprehensive set of variables, from household traits, household’s perception of bank service and experience dealing with banks, to neighborhood demographics and local banking market characteristics.
Another question examined is the dynamics of households’ AFS use. AFS users are often depicted as being trapped in cycles of debt and are more likely to be financially distressed (Birkenmaier and Tyuse, 2005; Melzer, 2009). The availability of earlier waves of FDIC surveys allows to track the same households over several years’ span – whether the households, once AFS users, remain underbanked, become unbanked or migrate to fully banked status later.
The main findings from the study include the following. The underbanked households tend to be less educated, black or Hispanic families with larger number of younger children, disabled, experienced monthly income volatility and earning lower family income, largely similar to the description of AFS users from previous studies (Despard et al., 2015; Goodstein and Rhine, 2017; Gross et al., 2012). Banks’ attitude in serving clients is indeed important determinant of clients’ decision using AFS; clients’ negative feeling about applying for bank credit significantly impacted their decision of getting nonbank credit. The amount of bank-level service charges such as account maintenance fees is found to have some impact on the decision to get credit products from nonbank establishments, somewhat echoing the criticism toward high bank fees from the interviewees. More than two-thirds of the underbanked families became fully banked later, and prior usage of nonbank credits does not predict the future uptake of such products. The results imply that AFS products provided valuable short-term relief to families’ financial needs, when traditional banking service is not available or is cost prohibitive.
This paper adds to the existing literature on financial inclusion and the use of alternative financial service. While discussions of financial inclusion usually focus on the unbanked population – especially how to promote mainstream banking to low-income and minority population, this paper studies the underbanked – those who already have bank accounts yet being underserved by mainstream banking. Compared to previous research on households’ choice of alternative financial services (Despard et al., 2015; Goodstein and Rhine, 2017), this research incorporates a rich set of variables, including local banking characteristics, customers’ rating and experience with banking and regional differences. The scope of study is not limited to a specific type of AFS, compared to, for instance, Goodstein and Rhine (2017) on transactional AFS only. The inclusion of multiple waves of FDIC unbanked/underbanked survey provides some insights on changes in households’ financial decision over time – AFS and particularly, nonbank credits was only a temporary fix for many households.
The results from this paper provide implications for different types of AFS. For example, the use of transactional AFS responds to the availability of online or mobile banking; meanwhile, it is also sensitive to branch closure. Users of nonbank credits are likely to be price savvy, and these products serve as valuable alternatives for short-term financing, especially during unfavorable economic situation. Better understanding of the underbanked could help banks tailor to existing clients’ needs, for instance, providing innovative short-term credit products for those with little or impaired credit history. The study also helps policy makers re-evaluate banking regulations since the Great Recession. As regulations squeezed bank profits in certain areas and forced banks to consolidate, come alongside higher bank fees, potential branch closure and loss of service, which ultimately forced banked individuals to the less regulated alternative providers.
2.1 Household socioeconomic characteristics
Existing research has found that household social-economic characteristics are important determinants of bank account ownership and use of nonbank financial services (Despard et al., 2015; Goodstein and Rhine, 2017; Gross et al., 2012). Despard et al. (2015) surveyed participants from a tax refund program; they find the users of AFS are more prevalent among less educated, lower-income female minorities with dependents and those who experienced financial difficulties. Gross et al.’s (2012) study uses the 2009 FDIC supplement to CPS and describes users of AFS as less educated, minority, middle aged, lower income, unemployed, from large households and unbanked. In particular, low-income consumers and unemployed consumers are more likely to use nonbank credit products because they are easier to obtain than qualifying for a bank loan; these two groups of consumers are also more price-sensitive on fees charged by banks on products such as money order. Using the 2011 FDIC supplement, Goodstein and Rhine (2017) find that race and ethnicity are important determinants for the likelihood to use alternative transactional services compared to other household socioeconomic attributes.
2.2 Local banking industry characteristics
According to the FDIC surveys, almost 50 per cent of households that do not currently have a bank account had one in the past, and many cite high fees as one of the reasons for leaving the banking system. Bank fees have increased in recent years due to regulatory moves following the Great Recession. The regulations effectively limited large banks’ consumer-related fee incomes; in response, banks have increased monthly fees on their retail accounts, raised minimum balance requirements and overdraft fees, to compensate for the loss in incomes. Banks’ monthly service fees on checking accounts jumped 25 per cent between 2010 and 2011; meanwhile, the per cent of free checking accounts declined from 75 to 39 per cent from 2009 to 2011.
Convenient physical access to bank branches encourages banking versus going to alternatives. As shown in Celerier and Matray (2018), an exogenous increase in bank branch density due to the US interstate branching deregulation increased the likelihood for low-income households to hold a bank account. Goodstein and Rhine (2017) find a negative relationship between transactional AFS use and reasonable geographic access to bank branches (within five miles of residence). For the same reason, we would expect branch closure could force some bank customers out of the banking system or turn them to alternative providers. Nguyen (2019) documented a negative impact on local credit supply due to branch closing; some of the negative impact was persistent – the volume of new small business loans was 13 per cent lower even several years after branch closing, and even in areas with dense branch network.
Consolidation in the US banking industry over the last three decades has drastically changed the landscape of banking. Community banks diminished in importance as their share of the market: the assets controlled by smaller banks (with assets less than US$100m) dropped by 73 per cent between 1984 and 2014; banks with less than US$1bn of assets in 1984, which once controlled nearly a third of banking assets, controlled less than 10 per cent by 2014 (Backup and Brown, 2014). During the same period, institutions with assets greater than US$10bn have seen their total assets grown more than ten-fold. Consolidation could cause higher bank fees or lower deposit rates as a result of diminished competition. For example, Bord (2017) finds that bank consolidation made deposit account fees and minimum required balances higher, hence forced some low-income depositors to exit the banking system; furthermore, areas where small banks being acquired by large ones experienced faster growth in alternative financial service providers. Prager and Hannan (1998) present evidence of bank mergers increasing market power, from the observation that banks where horizontal mergers took place offered lower deposit interest rates to their customers than banks elsewhere. Similarly, Park and Pennacchi (2009) report that large multimarket banks offer lower retail deposit rates than single-market banks operating in the same local market. Consolidation, if results in branch closure, could also lead to reduced access to credit or reduced access to banking services for certain population.
2.3 Local economics and demographics
Neighborhood racial and economic profile were found to be significantly related to the number of AFS providers in the area. In a research conducted by the Urban Institute on three types of alternative providers in eight counties with diverse demographic and regulatory environments, alternative providers are found to be disproportionally located in minority, low-income neighborhoods and often co-exist with banks geographically (Temkin and Sawyer, 2004). In a county-level study for the entire country, Prager (2014) finds that alternative financial service providers are more prevalent in areas with higher concentration of African Americans and people who lack high-school diplomas. The alternative providers are also found to avoid the poorest areas. Nonbank credit providers such as payday lenders and pawnshops are more likely to be in areas with more people without credit scores.
Furthermore, household’s choice of AFS could be impacted by friends or neighbors. Peer effect in financial decision-making has been well documented in literature, for example, in Duflo and Saez (2003) on retirement savings decisions, Ivković and Weisbenner (2007) on individual investor’s stock trading decision being influenced by neighbors.
Existing literature on alternative financial services, therefore, suggests household’s decision to adopt the services could be determined by household socioeconomic characteristics, local banking market and socioeconomics and demographics in the neighborhood.
Descriptive and regression analysis are used to study the underbanked households, their socioeconomic profile and the determinants of their decision to use AFS. Household’s decision to use AFS is modeled to be affected by several categories of factors. The first category includes household socioeconomic factors, such as ethnic background, education and employment status. The second set of variables focuses on local banking industry characteristics, including physical accessibility of local bank branches, bank fees and local banking market competitiveness. The third category contains households’ neighborhood characteristics, such as racial composition and poverty ratio. The remaining variables capture households’ perception and past experiences with mainstream banking.
Main data used in our analysis come from unbanked/underbanked supplement to the CPS. The supplement, sponsored by FDIC, collects information about household’s banking status and use of alternative financial services. Together with the demographic characteristics in the CPS questionnaire, the supplement provides important information of the unbanked and underbanked consumers.
Over the FDIC survey waves since 2009, the percentage of banked households using AFS remained quite stable around 20 per cent. Most of the variation in the use of AFS comes from cross-sectional differences; therefore, the main regression analysis focuses on single-year survey administered by the census in June 2015. Multiple observations for the same household within the year are aggregated at household level using assigned weights provided in the survey.
Bank financial data come from the FDIC income and expenses reports, which contain details about member institutions’ incomes. The variable extracted from the report is service charge imposed upon deposit accounts under the category of “non-interest income”. Bank branch information is extracted from the FDIC’s summary of deposits, which has information on branch office location and amount of branch deposit. MSA-level social-economic characteristics are downloaded from the American Community Survey.
Using location identifier, the 2015 CPS supplement is merged with MSA-level bank variables and social-economic variables in year 2014. Households that reported “unknown” for either MSA or banking status were excluded. The merge results in 23,861 unique households.
3.3.1 Dependent variables.
The main dependent variable is an indicator whether household has used ANY type of alternative financial service within the last 12 months. Indicators for distinct type of alternative financial services used by the households in the prior year – money order, check cashing, credit service, payday loan and pawn shop loan – are also created.
3.3.2 Household socioeconomic variables.
Variables include household income level, number of children under 18 years, education level of the household head, disability, employment status, working full time or part time, income volatility and ethnic background.
3.3.3 Local banking industry characteristics.
There is no direct information on bank fees incurred by the household or the banks used by the household being surveyed. The proxy for bank fees is constructed by taking the average of service charges by bank branches within the MSA area where the household resides. Bank-level information about service charges imposed on deposit accounts is extracted from the FDIC income and expense reports. The bank fee variable is calculated as total service charge divided by total domestic deposits for each bank, then averaged across all FDIC bank branches within the MSA.
Bank branch density is calculated using the number of FDIC bank branches in an MSA divided by total MSA population.
Herfindahl index has been used by the Federal Reserve as numerical guideline in analyzing the effect of bank mergers and serves as a common measure for industry concentration (Rhoades, 1993). Herfindahl index (HHI) used in this study is calculated based on branch’s share of total deposits in the MSA area.
3.3.4 MSA-level variables.
Neighborhood demographic and economic characteristics such as the ratio of Black and Hispanic population and poverty ratio are included in the analysis to control for the potential peer effect and the supply side of alternative financial services.
3.3.5 Attitude, perception and experience of banking.
Quality of service was mentioned by patrons of transactional AFS as one of the reasons choosing nonbank service providers over banking institutions. The FDIC supplement includes information about households’ experience with banking. Three variables are selected to be included in the analysis: an indicator for denial of bank credit during the last 12 months, indicator for fear of being denied bank credit application and rating of banks’ interest level in serving the households being surveyed which could be used as a proxy for the service quality received by bank clients.
3.4 Descriptive statistics
3.4.1 The use of AFS.
Table I shows that, among the 23,861 unique households in 2015, about 93 per cent were banked, and the remaining 7 per cent households were left out of the banking system. Around 21 per cent of the banked households (equivalent to almost 4,600 unique households) have used some form of alternative financial service in the prior year (i.e. “underbanked” households). In comparison, 58 per cent of the unbanked households (equivalent to 960 unique households) reported having used any type of AFS in the previous year. Although the percentage of households using AFS was significantly higher among the unbanked, the number of banked households that were exposed to AFS was substantial.
Transactional AFS include services such as cashing a check at a local retailer or obtaining money orders from a nonbank institution. At least 86 per cent of the underbanked households have used transactional AFS such as money order. Almost one-third of the underbanked households have used credit type of alternative services; one quarter of the credit-type AFS users got payday loans and another quarter used pawn shop loans. The pattern is somewhat different from unbanked households: much greater percentage of unbanked families used check-cashing services from nonbank venue and obtained pawn shop loans than banked families. Use of AFS also varies greatly among states (state means are not tabulated). Nevada had the highest proportion of households using AFS in 2015 (32.7 per cent), while Wisconsin and Minnesota had the lowest AFS usage (12 per cent and 12.8 per cent, respectively).
3.4.2 Summary statistics.
The descriptive statistics in Table II show that the demographic and economic profile of the underbanked households lied somewhere in between those for the unbanked and fully banked households. The underbanked group also experienced more bank credit denial and felt more discouraged applying for bank credit compared to unbanked and fully banked groups, implying this group was not served by their banks to their financial needs, and therefore turned to alternative providers.
4.1 Household-level analysis for AFS use
The main regression model is a logit regression, where the dependent variable indicates whether the surveyed households have used ANY type of AFS in the last year (i.e. being “underbanked”) and the specific types of AFS being used. The explanatory variables include household-level socioeconomic variables, MSA-level variables that capture local banking market structure and MSA-level demographics and poverty rate. Standard errors are clustered at the MSA level to allow for interdependence among households within the same MSA. Table III summarizes the estimated marginal effects from logit regressions.
4.1.1 Household social-economic characteristics.
The estimates presented in Column 1 show that the association between household socioeconomic characteristics and use of any AFS is largely consistent with existing studies (Despard et al., 2015; Goodstein and Rhine, 2017; Gross et al., 2012). Specifically, the number of children under 18 years in the households, being black or Hispanic are positively associated with the likelihood to obtain AFS. The probability for black households obtaining AFS was 17 percentage points higher, and Hispanic household were 12 percentage points more likely to obtain AFS (both effects are statistically significant at the 1 per cent level). Higher education level suppressed the use of AFS, particularly for those with college degrees (12 percentage points lower probability compared to household heads without high-school diploma). Household income at higher brackets reduced the likelihood of using AFS; the probability for households earning at least US$75,000 to get AFS was 16 percentage points lower than households earning less than US$15,000 (the default bracket). Being unemployed did not have a significant marginal effect on the decision to have AFS; however, the marginal effect of monthly income volatility is significantly positive – a household with volatile income was 7 percentage points more likely to use AFS. Having a disabled household head also increased the tendency to get AFS while controlling for all other household characteristics.
Column 2 presents the estimated marginal effects of the logit regression where the dependent variable is indicator for banked households’ use of transactional AFS. The associations between household characteristics and the use of transactional AFS are qualitatively similar to those in Column 1. Family ethnic background, education level, household income level and income volatility show strong and statistically significant impact on the decision to acquire nonbank transactional services. Columns 3-4 show that, although most household socioeconomics (college education, income level, disability and ethnicity) exerts the same direction of impact on the tendency to use check-cashing or money-order services, their effects on using check-cashing are more attenuated than on money-order usage. Unemployment seems to be an important determinant in the decision to acquire check-cashing AFS for banked households, with a significantly positive marginal effect.
Regression results for credit-type AFS usage are listed in Columns 5-7. Note that only the highest education level (college degree) and higher income brackets (greater than US$50,000) were significantly negatively associated with the probability of getting nonbank credit products. Hispanic households could not be linked to greater chance of using credit-type AFS (except for payday loan).
4.1.2 Local banking characteristics and neighborhood socioeconomics.
The regressions in Table III include three local banking industry characteristics: number of bank branches relative to population within an MSA, banking market concentration in the local area measured by the HHI and the average level of service charge on depository accounts among local banks. Higher branch density is expected to discourage the use of AFS for existing customers, market concentration suppresses competition, and therefore makes banking more expensive or deposit rates less attractive, in a direction of pushing banking clients away; higher bank fees are also expected to increase the likelihood for AFS use.
Regressions results shown in Table III do not support that bank density or market concentration were important determinants for the decision to take up AFS. The only bank characteristics variable that had some effect on AFS is the level of bank fees. The marginal effect is positive and statistically significant (at 1 per cent level) on the decision to acquire nonbank credit. The size of the impact is around 0.6 percentage point increase in probability for a one standard-deviation increase in service charge (0.315 * 0.019 = 0.006). It is small on an absolute scale, but not negligible considering the overall smaller presence of nonbank credit product users (7 per cent for the sample). A positive relation between bank fees and credit AFS could be because bank clients were using nonbank credit service as a substitute to overdrafts for short-term financing (Zinman, 2010 showed former payday borrowers turned to bank overdrafts due to restricted access to payday loans). And, as overdraft fees increased, overdraft financing became more expensive; clients were more likely to go for nonbank credit.
Lastly, neighborhood demographics and poverty rate have only weak effect on household’s decision to obtain AFS. The percentage of white population in the local MSA area has a small negative marginal effect on AFS, while the presence of black population is negatively related to credit AFS. Poverty has a significantly negative effect on transactional AFS, but meanwhile, slightly increases the tendency for credit AFS; therefore, the overall effect on AFS is muted.
4.1.3 Attitude, perception and experience of banking.
Customers were asked to rate their experience with banking in the FDIC surveys. The first variable included in the regression is perceived banks’ interest in serving clients like the household being surveyed, to proxy for service quality. Column 1 in Table IV shows that the estimated marginal effect of interviewee’s rating of bank’s interest in serving has a negative and significant impact on the likelihood to use transactional AFS – a bank that shows “very interested” in serving reduces the likelihood for existing clients going to alternative providers by 2 percentage points (significant at 1 per cent level). Rating of bank’s interest is not a significant factor for bank clients’ decision on choosing a nonbank service provider over banks for check-cashing. However, customers were more likely to turn to nonbank providers for money orders if they felt that banks were not interested in serving them.
For households’ use of credit AFS, additional variables reflecting their experience with bank credit are included in the regressions. One variable is specifically on whether they have been denied bank credit in the last year, and the other variable captures whether they felt discouraged applying for bank credit. Negative experience or feeling with bank credit seems to be an important factor in the decision to get nonbank credit products, even after controlling for household characteristics such as family income level. As shown in Columns 4-6, the indicators for being denied bank credit or feeling discouraged applying for bank credit have significantly positive marginal effects on households’ decision to get nonbank credit. Particularly, if the household felt discouraged applying for bank credit, the chance to get nonbank credit was 6 percentage points higher.
4.2 Robustness check
The results so far show that some household socioeconomic factors remain important determinants in the decision to acquire AFS. Other important external factors include bank fees in the local market and the service quality offered by banks. Neighborhood traits such as poverty and presence of minority have only modest effect. The observed relationship between AFS use and household characteristics, local banking market conditions and demographic traits could be driven by regional differences such as state-specific financial regulations or other unobserved factors. The emergence of online or mobile banking offered an alternative to traditional banking, which could attenuate some of the impact due to local banking markets, such as physical accessibility of branches. In this section, alternative explanations are tested.
4.2.1 Regional differences.
Regulations toward AFS vary by state. For instance, as of 2011, 17 states and the District of Columbia capped payday interest rate at 36 per cent APR or lower (Morgan-Cross and Klawitter, 2011). Such bans were shown effective in restricting access to expensive nonbank credit, while welfare implications of the regulation remained unclear (Weaver and Galperin, 2014; Zinman, 2010).
The first regression in Table V controls for whether a state had payday rate cap regulation. The estimates show that, in states with payday lending regulation, households were less likely to get payday loan compared to those in states without such regulation; the probability was 1 percentage point lower. Comparing to the overall usage of payday loans in the sample at 1.7 per cent, the regulation proved to be effective in restricting access to the borrowing.
Instead of controlling for state-specific variables, Columns 2-4 in Table V add state fixed effects to the regressions for the use of any AFS, transactional and credit type of AFS. The marginal effects of household characteristics, rating of bank service and experience with bank credit are not much different with the inclusion of state fixed effects. Bank branch density, after controlling for state fixed effects, shows a significantly negative marginal effect on AFS, especially for transactional AFS, with a one standard-deviation increase in branch density reducing the AFS likelihood by 1.4 per cent. Local banking market concentration becomes positive and significant (at 10 per cent level) on the decision of credit AFS. These findings are consistent with previous studies, for example, Goodstein and Rhine (2017) find a negative relationship between transactional AFS use and reasonable geographic access to bank branches, and Nguyen (2019) reports bank consolidation had a negative impact on access to bank credit. However, the strong effect of local bank fees on the decision to acquire nonbank credits disappears after controlling for state fixed effects.
4.2.2 The alternative of online or mobile banking.
Recent years have witnessed growth of online or mobile banking as alternative to traditional banking for transactional services. The FDIC survey evaluated the use of online or mobile banking by asking interviewees the most common way to access a bank account. If a bank customer is literate in online or mobile banking, his/her access to bank accounts will be less affected by physical locations or operation-hour limits by traditional brick-and-mortar banking; therefore, these customers’ demand for nonbank transactional service becomes smaller. Meanwhile, the density of bank branches in the neighborhood should not be as important for an online banking patron. Regressions in Table VI confirm this presumption. The indicator for online or mobile banking as the most common way to access account takes on a significantly negative marginal effect on any AFS and on transactional AFS, as shown in the first two columns. The effect of bank branch density remains muted.
4.2.3 Impact of branch closure.
Literature shows that bank consolidation could have negative impact on deposit account retail customers and small businesses (Bord, 2017; Nguyen, 2019; Prager and Hannan, 1998). As a result, these customers may turn to nonbank services due to higher bank fees or branch closure. However, the results so far do not show strong supporting evidence for the impact of branch density or competitiveness on the use of AFS. Probably, using the existing number of branches and measure of market concentration in a single year do not sufficiently capture the changes in local banking markets. With multiple years of FDIC bank branch data, over-time net change in the number of branches within an MSA is incorporated into the analysis. Areas that have experienced branch closure may show larger effects on AFS usage. FDIC bank branch data are obtained for 2011 and 2015; an indicator is created whether an MSA has experienced loss of branches in net during the time period. The estimated marginal effects of branch closure on the use of AFS are listed in Columns 4-6 in Table VI. Specifically, households in MSAs which had a net loss of bank branches were more likely to get AFS and transactional AFS in particular (4 percentage points more likely); the decision on credit AFS was not impacted by bank branch closure. The regressions already take into account whether the household used online or mobile banking.
The results in Table VI show that changes in the banking industry had significant impact on the demand for transactional AFS: loss of physical branches increased the demand for transactional AFS, while online/mobile banking decreased the demand.
4.3 Analysis using multiple waves of FDIC surveys
A growing concern is the use of AFS, despite providing short-term financial relief, could leave households in escalating debt due to its high cost and vulnerability of the users (Birkenmaier and Tyuse, 2005; Melzer, 2009). To examine whether banked households used AFS as a temporary solution, or rather relied on AFS over a longer term, FDIC surveys from the previous two waves (2011 and 2013) are included. Households are tracked using household ID, and over-time change in household’s banking status is identified. Families that made more than one appearance in the surveys were identified. This results in a total of 18,000 household-year observations. Around 15 per cent of the observations show a switch from underbanked status to fully banked status by the next survey, 14 per cent went from fully banked to underbanked, 6 per cent remained underbanked and the remaining stayed as fully banked. The statistics reveal that AFS served as a rather temporary financial solution for existing banking clients.
I then examine the differences among the groups that underwent different changes in AFS usage, by comparing mean values of key household variables, including change in unemployment status, change in part-time working indicator and ethnicity background. The comparisons are summarized in Panel A of Table VII. Change in unemployment takes on a value of 1 if the household became unemployed between two surveys, and –1 if the households moved from unemployment to employment status. The subsample of households that moved from fully banked to underbanked had a slightly positive mean value of change in unemployment, whereas the group that once used AFS then became fully banked later had a –4.6 per cent change in “unemployment” indicator (meaning that there were more households moving from unemployment to employment in net). Change in part-time work status has the same construct as change in unemployment – equals 1 if household became working part time between two surveys, and –1 if the households moved from part- to full-time work status. The group that later became fully banked shows an average switch to full-time work. The comparisons imply that change in employment/work status over time could be an important determinant for families to get off AFS.
An additional test includes a lagged AFS use indicator into the regressions, to evaluate how likely a household continued to use alternative services into the next survey period. Panel B exhibits the estimated marginal effects, including a lagged dependent variable into the logit regressions. Unemployment status indicator is replaced by a change in unemployment over time, to reflect the observation from Panel A. Most household characteristics, local banking and demographic characteristics tend to be stable between surveys. The results show that being underbanked in the previous period helps explain underbanked status in the current survey, with a 4.7 percentage points significantly higher probability. However, using nonbank credit products in the previous survey did not help explain using it the current survey period. Change in unemployment status contributes to the probability of getting nonbank credits – becoming unemployed increased the likelihood by 1 percentage point. Average bank fee in the local market remains significantly positively associated with the tendency to get nonbank credits, showing price sensitivity for credit products.
With very limited time-series data provided by the FDIC surveys, it shows that households were more likely to stay underbanked into the next period, but households did not stay with nonbank credit products for prolonged period of time, unless employment status has changed negatively.
5. Discussion and conclusion
Using the FDIC unbanked/underbanked supplement to the CPS, this study documents the percentage of underbanked households and the types of nonbank financial services chosen by these households. The socioeconomic profile of the underbanked lies in between that of the unbanked and fully banked. The cross-sectional determinants of AFS usage are studied, with a comprehensive set of variables that include not only household characteristics which were used in previous studies but include new information such as households’ rating of how banks served them and their experience with banking. Effects of local banking industry structure and demographic characteristics within the boundary of the MSA are also studied.
The estimated relationships between household characteristics and the use of AFS are largely consistent with existing literature: in general, households that were from minority background, having lower education and lower income, with disabled household head and experienced greater income volatility were more likely to take on alternative services. The variables of perception and attitudes toward banking provide some new insight. The results show that clients’ complaints about service quality is one of the reasons they turned to nonbank venues. Past denial of bank credit, or negative feeling about applying for bank credit, also proves to be important in banked households’ decision to get nonbank credit. Although negative experience of getting bank credit or feeling discouraged of applying for bank credit could reflect poor credit score or lack of credit, they are still informative given the fact that these variables show significant effect even after controlling for household social-economic characteristics, especially household income which is argued to be the most important determinant of credit scores (Albanesi et al., 2017).
Recent banking regulations caused deposit account maintenance costs to increase, which could have imposed disproportionally larger negative welfare impact on the low-income population (Bord, 2017). Meanwhile, substantial banking consolidation could disturb the local banking market due to branch closures or diminished competition. The findings in this paper show that areas with higher average bank costs were associated with bank clients turning to nonbank institutions for credit type of services, possibly because clients substituted nonbank credit for the more expensive bank overdraft as a source of short-term financing. The price sensitivity confirms that the decisions of choosing these products were rather calculated ones. Areas that have lost branches in net tend to increase the likelihood of households’ taking on alternative transactional services. Households that used online or mobile banking were found less likely to use transactional AFS.
FDIC surveys started only recently which made time-series analysis difficult. With the few waves, I was able to link a small percentage of the families over time to identify changes in their financial service status. It shows some degree of persistence in underbanked status, except for nonbank credits. Change in unemployment status increased the likelihood of using credit type of AFS. This means that alternative credit products were mostly used to fulfill short-term needs; the availability of such short-term credit likely softens financial shocks brought by work status change and facilitates job search.
While nonbank financing provides a valuable alternative to bank financing and helps support real economic activity, regulators fear that increased reliance on nonbank funding could give rise to new risks. As they vowed to protect the consumers by installing new regulations on shadow banking, policymakers must remember the reasons that turn them to alternative providers and make sure affordable lifeline access to capital are not being cut off. For instance, without a variety of alternative forms of short-term borrowing, restricting payday loans may limit a household’s ability to access credit (McKernan et al., 2010; Prager, 2014; Zinman, 2010).
What are the opportunities for banks? In an effort to bring the unbanked and underbanked groups back to mainstream banking, banks need to find how they can better serve their clients for various financial needs. One thing that could help the unbanked is low-cost depository services, for example, that provide no checking but offer debit card and with very low level of required account balance. Given that many underbanked customers also use transactional AFS, it may help to make bank fees transparent, straightforward to understand and more predictable. Meanwhile, the results in this paper support that enhancing banks’ online or mobile account accessibility could help clients with their transactional account needs.
Another area is innovative short-term credit products suitable for the low- and moderate-income households and those who lose their jobs. Due to regulatory change since the recent financial crisis, banks were discouraged from lending but to the creditworthy; families and businesses have, therefore, turned to the less regulated albeit more expensive alternative providers to obtain credit. Credit-building products, such as credit cards secured with minimal balance from savings account, can help customers with little or impaired credit to develop or repair credit history. For customers with impaired or little credit history, different metrics could serve as alternative credit analysis, such as rent payments, utility payments, use of payday loan products and stability of employment (Frumkin, 2006). This can help customers get loans that may not be possible to obtain from traditional channels.
Banks need to be actively engaged in community outreach and financial education, to build relation with customer, establish or reinforce clientele’s trust in the banks and change their perception toward banks. Banks also need to be aware of the potential left-out clientele due to branch closing and loss of certain services amidst the consolidation trend. Retaining customers by providing the solutions needed will be a challenging task.
AFS use for households in FDIC surveys
|Variables||All (N = 23,861)||Banked (N = 22,212)||Unbanked (N = 1,649)||p-value|
|AFS (any type)||0.229||0.207||0.582||0.000|
|Pawn shop loan||0.018||0.015||0.065||0.000|
This table presents the mean values of indicator variables if the household surveyed has used any type of AFS, or specific types of AFS in the prior year. The sample includes unique households from the FDIC unbanked/underbanked surveys conducted in 2015. Households that reported “unknow” for MSA and banking status were excluded
(N = 23,861)
(N = 1,649)
(N = 4,598)
(N = 17,614)
|Number of child < 18||0.594||1.026||0.918||0.740||0.530|
|Education: high school||0.246||0.428||0.352||0.277||0.227|
|Education: some college||0.288||0.449||0.233||0.320||0.284|
|Education: college degree||0.369||0.478||0.073||0.272||0.423|
|Working part time||0.097||0.294||0.117||0.113||0.097|
|Family income [<15]||0.127||0.330||0.489||0.159||0.085|
|Family income [15-30]||0.157||0.361||0.284||0.186||0.136|
|Family income [30-50]||0.192||0.391||0.152||0.231||0.186|
|Family income [50-75]||0.182||0.382||0.044||0.187||0.194|
|Family income [75+]||0.341||0.470||0.031||0.236||0.400|
|Attitudes toward banking|
|Bank’s interest in serving||0.374||0.480||0.082||0.345||0.426|
|Denied bank credit||0.025||0.155||0.014||0.054||0.020|
|Discouraged bank credit||0.055||0.227||0.093||0.132||0.035|
|MSA-level banking variables|
|Bank branches density||27.873||6.868||27.278||27.511||28.073|
|Bank concentration (HHI)||0.069||0.105||0.063||0.068||0.070|
|Average service charge||0.518||0.315||0.558||0.542||0.510|
|MSA social-economic variables|
The sample includes unique households from the FDIC unbanked/underbanked surveys conducted in 2015. Households that reported “unknown” for MSA and banking status were excluded. “Underbanked” refers to the households with bank accounts and reported have used AFS in the prior year. “Fully banked” refers to the households with bank accounts and reported have not used AFS in the prior year. “Unbanked” refers to the households that reported having no bank accounts in the prior year
Determinants of being underbanked – household- and MSA-level characteristics
|ANY AFS||Transactional AFS||Credit type AFS|
|Variables||(1) AFS||(2) Trans. AFS||(3) Check cashing||(4) Money order||(5) Credit AFS||(6) Payday loan||(7) Pawn shop loan|
|Number of children under 18||0.020*** (0.003)||0.012*** (0.003)||0.005*** (0.001)||0.009*** (0.002)||0.013*** (0.002)||0.003*** (0.001)||0.002*** (0.001)|
|Education: high school||−0.050*** (0.014)||−0.049*** (0.012)||−0.012** (0.005)||−0.017 (0.011)||0.003 (0.006)||0.001 (0.003)||−0.000 (0.002)|
|Education: some college||−0.064*** (0.015)||−0.062*** (0.012)||−0.012** (0.005)||−0.022** (0.010)||0.001 (0.007)||0.003 (0.003)||0.001 (0.002)|
|Education: college degree||−0.120*** ( 0.016)||−0.106*** (0.013)||−0.032*** (0.006)||−0.066*** (0.012)||−0.034*** (0.008)||−0.008** (0.004)||−0.008*** (0.003)|
|Disability||0.050*** (0.010)||0.033*** (0.009)||0.007* (0.004)||0.038*** (0.008)||0.024*** (0.006)||0.006** (0.003)||0.006*** (0.002)|
|Unemployed||0.024 (0.017)||0.021 (0.015)||0.020*** (0.007)||0.023* (0.014)||0.021** (0.008)||0.005 (0.004)||0.003 (0.003)|
|Income volatility||0.074*** (0.008)||0.067*** (0.007)||0.024*** (0.004)||0.038*** (0.005)||0.027*** (0.004)||0.009*** (0.002)||0.006*** (0.001)|
|Family income [15-30]||−0.021 (0.015)||−0.017 (0.014)||−0.008 (0.005)||−0.018* (0.011)||−0.002 (0.007)||0.000 (0.003)||−0.005*** (0.002)|
|Family income [30-50]||−0.051*** (0.012)||−0.048*** (0.011)||−0.017*** (0.005)||−0.038*** (0.009)||0.000 (0.006)||0.005 (0.003)||−0.008*** (0.002)|
|Family income [50-75]||−0.084*** (0.013)||−0.070*** (0.012)||−0.015*** (0.005)||−0.066*** (0.009)||−0.017** (0.007)||−0.001 (0.003)||−0.011*** (0.002)|
|Family income [75+]||−0.164*** (0.013)||−0.135*** (0.012)||−0.033*** (0.005)||−0.117*** (0.011)||−0.051*** (0.007)||−0.009** (0.003)||−0.018*** (0.003)|
|Black||0.166*** (0.012)||0.153*** (0.011)||0.016*** (0.005)||0.119*** (0.009)||0.034*** (0.005)||0.014*** (0.002)||0.004* (0.002)|
|Hispanic||0.119*** (0.013)||0.115*** (0.010)||0.008* (0.005)||0.052*** (0.009)||0.006 (0.006)||0.004** (0.002)||0.002 (0.002)|
|MSA level banking variables|
|Bank branches density||−0.000 (0.001)||−0.001 (0.001)||−0.000 (0.000)||−0.001 (0.001)||0.001 (0.000)||−0.000 (0.000)||−0.000 (0.000)|
|Bank concentration (HHI)||0.008 (0.046)||−0.007 (0.041)||−0.013 (0.017)||−0.016 (0.043)||0.028 (0.023)||−0.001 (0.011)||0.008 (0.007)|
|Average service charge||0.021 (0.021)||0.005 (0.018)||−0.005 (0.007)||0.007 (0.012)||0.019*** (0.007)||0.003 (0.003)||0.003 (0.002)|
|MSA social economic variables|
|% White||−0.001* (0.001)||−0.001 (0.000)||0.000* (0.000)||−0.000 (0.001)||−0.000 (0.000)||0.000 (0.000)||0.000 (0.000)|
|% Black||0.000 (0.001)||0.001 (0.001)||0.000** (0.000)||0.001* (0.001)||−0.001* (0.000)||−0.000 (0.000)||0.000 (0.000)|
|% poverty||−0.003 (0.002)||−0.003** (0.002)||−0.000 (0.001)||−0.000 (0.001)||0.002** (0.001)||0.000* (0.000)||0.000* (0.000)|
|Number of observations||15,522||15,522||15,522||15,522||15,522||15,522||15,522|
The sample includes banked households from the FDIC Unbanked/Underbanked surveys conducted in 2015. The dependent variable is an indicator if the surveyed households have used any type of AFS, transactional AFS, non-bank check-cashing and non-bank money order, credit type AFS, payday loans, and pawn shop loans respectively. For education level, the omitted category is “no high school diploma”. For family income, the omitted category is “less than $15,000”. Logit regression models are estimated, with marginal effects at mean reported. Standard errors are clustered at the MSA level;
*p < 0.1;
**p < 0.05;
***p < 0.01
The impact of attitudes and banking experience
|Transactional AFS||Credit type AFS|
|Variables||(1) Trans. AFS||(2) Check cashing||(3) Money order||(3) Credit AFS||(5) Payday loan||(6) Pawn shop loan|
|Number of children under 18||0.012*** (0.003)||0.005*** (0.001)||0.009*** (0.002)||0.011*** (0.001)||0.002*** (0.001)||0.001*** (0.001)|
|Education: high school||−0.049*** (0.012)||−0.012** (0.005)||−0.017 (0.011)||−0.000 (0.006)||−0.000 (0.003)||−0.001 (0.002)|
|Education: some college||−0.062*** (0.012)||−0.012** (0.005)||−0.021** (0.010)||−0.003 (0.006)||0.002 (0.003)||−0.000 (0.002)|
|Education: college degree||−0.104*** (0.013)||−0.032*** (0.006)||−0.064*** (0.012)||−0.034*** (0.007)||−0.008** (0.004)||−0.008*** (0.003)|
|Disability||0.033*** (0.009)||0.007 (0.004)||0.037*** (0.008)||0.019*** (0.005)||0.004* (0.002)||0.005*** (0.002)|
|Unemployed||0.020 (0.015)||0.020*** (0.007)||0.022 (0.014)||0.013* (0.008)||0.002 (0.003)||0.001 (0.003)|
|Income volatility||0.067*** (0.007)||0.024*** (0.004)||0.038*** (0.005)||0.020*** (0.004)||0.006*** (0.002)||0.004*** (0.001)|
|Family income [15-30]||−0.018 (0.014)||−0.008 (0.005)||−0.018* (0.011)||−0.003 (0.007)||−0.000 (0.003)||−0.005*** (0.002)|
|Family income [30-50]||−0.047*** (0.011)||−0.017*** (0.005)||−0.038*** (0.009)||0.003 (0.006)||0.006** (0.003)||−0.007*** (0.002)|
|Family income [50-75]||−0.068*** (0.012)||−0.014*** (0.005)||−0.064*** (0.009)||−0.014* (0.007)||0.001 (0.003)||−0.009*** (0.002)|
|Family income [75+]||−0.132*** (0.012)||−0.033*** (0.005)||−0.114*** (0.011)||−0.041*** (0.007)||−0.005 (0.003)||−0.015*** (0.003)|
|Black||0.152*** (0.011)||0.016*** (0.005)||0.119*** (0.009)||0.030*** (0.004)||0.012*** (0.002)||0.003* (0.002)|
|Hispanic||0.115*** (0.010)||0.008 (0.005)||0.051*** (0.009)||0.002 (0.006)||0.003 (0.002)||0.001 (0.002)|
|Attitudes and experience|
|Bank’s interest in serving||−0.020*** (0.006)||−0.002 (0.003)||−0.018*** (0.006)||−0.011** (0.004)||−0.002 (0.002)||−0.003** (0.001)|
|Denied bank credit||0.040*** (0.007)||0.012*** (0.003)||0.006** (0.002)|
|Discouraged bank credit||0.060*** (0.005)||0.016*** (0.002)||0.011*** (0.002)|
|MSA level banking variables|
|Bank branches density||−0.001 (0.001)||−0.000 (0.000)||−0.001 (0.001)||0.001* (0.000)||−0.000 (0.000)||−0.000 (0.000)|
|Bank concentration (HHI)||−0.004 (0.042)||−0.012 (0.017)||−0.014 (0.043)||0.035 (0.021)||0.002 (0.010)||0.009 (0.006)|
|Average service charge||0.005 (0.018)||−0.005 (0.007)||0.007 (0.012)||0.019*** (0.007)||0.003 (0.002)||0.003 (0.002)|
|MSA social economic variables|
|% White||−0.001 (0.000)||0.000* (0.000)||−0.000 (0.001)||−0.000 (0.000)||0.000 (0.000)||0.000 (0.000)|
|% Black||0.001 (0.001)||0.000** (0.000)||0.001* (0.001)||−0.001* (0.000)||−0.000 (0.000)||0.000 (0.000)|
|% poverty||−0.003** (0.002)||−0.000 (0.001)||−0.000 (0.001)||0.002*** (0.001)||0.001** (0.000)||0.001** (0.000)|
|Number of observations||15,522||15,522||15,522||15,522||15,522||15,522|
The sample includes banked households from the FDIC Unbanked/Underbanked surveys conducted in 2015. The dependent variable is an indicator if the surveyed households have used specific types of AFS in the past year (transactional in general, non-bank check-cashing and non-bank money order, credit type in general, payday loans, and pawn shop loans respectively). Logit regression models are estimated, with marginal effects at mean reported. Standard errors are clustered at the MSA level;
*p < 0.1;
**p < 0.05;
***p < 0.01
Robustness check-controlling for regional differences
|Variables||(1) Payday loan||(2) ANY AFS||(3) Trans. AFS||(4) Credit AFS|
|Number of children under 18||0.003*** (0.001)||0.020*** (0.003)||0.012*** (0.003)||0.011*** (0.001)|
|Education: high school||0.001 (0.003)||−0.048*** (0.014)||−0.048*** (0.012)||0.001 (0.006)|
|Education: some college||0.003 (0.003)||−0.062*** (0.014)||−0.061*** (0.012)||−0.003 (0.006)|
|Education: college degree||−0.008** (0.004)||−0.120*** (0.016)||−0.104*** (0.013)||−0.032*** (0.007)|
|Disability||0.006*** (0.002)||0.049*** (0.010)||0.032*** (0.009)||0.018*** (0.005)|
|Unemployed||0.005 (0.003)||0.026 (0.017)||0.021 (0.015)||0.015* (0.008)|
|Income volatility||0.009*** (0.002)||0.074*** (0.008)||0.067*** (0.007)||0.019*** (0.003)|
|Family income [15-30]||0.000 (0.003)||−0.021 (0.015)||−0.019 (0.014)||−0.003 (0.006)|
|Family income [30-50]||0.005* (0.003)||−0.050*** (0.012)||−0.046*** (0.011)||0.004 (0.006)|
|Family income [50-75]||−0.001 (0.003)||−0.081*** (0.013)||−0.067*** (0.012)||−0.011* (0.007)|
|Family income [75+]||−0.008** (0.003)||−0.163*** (0.013)||−0.131*** (0.012)||−0.037*** (0.007)|
|Black||0.014*** (0.002)||0.163*** (0.010)||0.149*** (0.010)||0.028*** (0.004)|
|Hispanic||0.004** (0.002)||0.120*** (0.012)||0.117*** (0.009)||0.002 (0.006)|
|Attitudes and experience|
|Bank’s interest in serving||−0.019*** (0.007)||−0.011*** (0.004)|
|Denied bank credit||0.038*** (0.007)|
|Discouraged bank credit||0.057*** (0.005)|
|MSA level banking variables|
|Bank branches density||−0.000 (0.000)||−0.002* (0.001)||−0.002** (0.001)||0.000 (0.001)|
|Bank concentration (HHI)||0.000 (0.010)||0.016 (0.047)||0.003 (0.046)||0.036* (0.019)|
|Average service charge||0.002 (0.002)||−0.001 (0.021)||−0.006 (0.021)||−0.000 (0.009)|
|MSA social economic variables|
|% White||0.000 (0.000)||−0.001 (0.001)||−0.001* (0.001)||0.000 (0.000)|
|% Black||−0.000 (0.000)||−0.000 (0.001)||0.000 (0.001)||−0.000 (0.001)|
|% poverty||0.000 (0.000)||−0.005** (0.002)||−0.005** (0.002)||0.000 (0.001)|
|State payday rate cap||−0.009*** (0.003)|
|State fixed effects||No||Yes||Yes||Yes|
|Number of observations||15,522||15,518||15,518||15,520|
The sample includes banked households from the FDIC Unbanked/Underbanked surveys conducted in 2015. Regressions 2-4 include state fixed effects. Logit regression models are estimated, with marginal effects at mean reported. Standard errors are clustered at the MSA level;
*p < 0.1;
**p < 0.05;
***p < 0.01
Robustness check-the impact of online or mobile banking, and bank branch closure
|Control for online/mobile banking||Impact of branch closure|
|Variables||(1) AFS||(2) Trans. AFS||(3) Credit AFS||(4) AFS||(5) Trans. AFS||(6) Credit AFS|
|Number of children under 18||0.020*** (0.003)||0.013*** (0.003)||0.013*** (0.002)||0.020*** (0.004)||0.012*** (0.003)||0.013*** (0.002)|
|Education: high school||−0.045*** (0.014)||−0.043*** (0.012)||0.003 (0.006)||−0.044*** (0.015)||−0.043*** (0.012)||0.003 (0.006)|
|Education: some college||−0.057*** (0.015)||−0.055*** (0.012)||0.000 (0.007)||−0.057*** (0.015)||−0.054*** (0.012)||0.000 (0.007)|
|Education: college degree||−0.111*** (0.016)||−0.097*** (0.014)||−0.035*** (0.008)||−0.107*** (0.016)||−0.094*** (0.014)||−0.032*** (0.008)|
|Disability||0.047*** (0.010)||0.030*** (0.009)||0.023*** (0.006)||0.044*** (0.010)||0.029*** (0.009)||0.022*** (0.006)|
|Unemployed||0.024 (0.017)||0.022 (0.016)||0.020** (0.009)||0.021 (0.017)||0.023 (0.016)||0.018** (0.009)|
|Income volatility||0.074*** (0.008)||0.068*** (0.007)||0.026*** (0.004)||0.078*** (0.008)||0.071*** (0.007)||0.028*** (0.004)|
|Family income [15-30]||−0.018 (0.015)||−0.016 (0.014)||−0.002 (0.007)||−0.020 (0.015)||−0.016 (0.014)||−0.002 (0.007)|
|Family income [30-50]||−0.047*** (0.012)||−0.045*** (0.011)||0.001 (0.006)||−0.051*** (0.012)||−0.047*** (0.011)||−0.001 (0.007)|
|Family income [50-75]||−0.078*** (0.013)||−0.066*** (0.012)||−0.017** (0.007)||−0.080*** (0.013)||−0.066*** (0.012)||−0.018** (0.007)|
|Family income [75+]||−0.158*** (0.013)||−0.131*** (0.012)||−0.050*** (0.007)||−0.159*** (0.014)||−0.131*** (0.012)||−0.050*** (0.007)|
|Black||0.163*** (0.011)||0.150*** (0.011)||0.034*** (0.005)||0.165*** (0.011)||0.151*** (0.010)||0.034*** (0.005)|
|Hispanic||0.118*** (0.012)||0.114*** (0.010)||0.006 (0.006)||0.124*** (0.011)||0.119*** (0.009)||0.006 (0.006)|
|Online/mobile banking||−0.019** (0.009)||−0.019** (0.009)||−0.001 (0.004)||−0.021** (0.009)||−0.020** (0.009)||−0.003 (0.004)|
|MSA level banking variables|
|Bank branches density||−0.000 (0.001)||−0.001 (0.001)||0.001 (0.000)||−0.000 (0.001)||−0.001 (0.001)||0.001 (0.000)|
|Bank concentration (HHI)||0.004 (0.045)||−0.006 (0.041)||0.024 (0.022)||0.002 (0.043)||−0.009 (0.040)||0.022 (0.022)|
|Average service charge||0.021 (0.022)||0.006 (0.018)||0.020*** (0.007)||0.020 (0.020)||0.005 (0.017)||0.020*** (0.007)|
|MSA with net branch closure||0.040*** (0.013)||0.040*** (0.012)||0.010 (0.007)|
|MSA social economic variables|
|% White||−0.001* (0.001)||−0.001 (0.000)||−0.000 (0.000)||−0.001** (0.000)||−0.001** (0.000)||−0.000 (0.000)|
|% Black||0.000 (0.001)||0.001 (0.001)||−0.001* (0.000)||−0.000 (0.001)||0.000 (0.001)||−0.001* (0.000)|
|% poverty||−0.003 (0.002)||−0.004** (0.002)||0.002 (0.001)||−0.003 (0.002)||−0.003** (0.002)||0.002** (0.001)|
|Number of observations||15,348||15,348||15,348||14,987||14,987||14,987|
The sample includes banked households from the FDIC Unbanked/Underbanked surveys conducted in 2015. “MSA with net branch closure” is extracted from FDIC bank branch data from 2011 and 2015, that indicates if an MSA has experienced bank branch closure (in net) between the two years. Logit regression models are estimated, with marginal effects at mean reported. Standard errors are clustered at the MSA level;
*p < 0.1;
**p < 0.05;
***p < 0.01
Analysis using multiple waves of FDIC surveys
|Variables||Fully banked to underbanked (N = 2,578)||Underbanked to fully banked (N = 2,790)||Stayed underbanked (N = 1,153)||Stayed fully banked (N = 11,669)|
|Panel (A) Preliminary analysis of household financial service status change over time|
|Avg. change in unemployment||0.003||−0.046||−0.051||−0.007|
|Avg. change in part-time work status||0.023||−0.034||0||0.002|
|Panel (B) A lagged latent model of being underbanked|
|(1) AFS||(2) Trans. AFS||(3) Credit AFS|
|Lagged AFS indicator||0.047*** (0.007)||0.045*** (0.007)||0.023 (0.017)|
|Number of children under 18||0.016*** (0.003)||0.013*** (0.003)||0.009*** (0.001)|
|Education: high school||−0.055*** (0.013)||−0.054*** (0.013)||−0.005 (0.005)|
|Education: some college||−0.059*** (0.014)||−0.060*** (0.013)||−0.002 (0.005)|
|Education: college degree||−0.128*** (0.014)||−0.113*** (0.014)||−0.037*** (0.006)|
|Disability||0.040*** (0.011)||0.030*** (0.010)||0.015*** (0.004)|
|Change in unemployed status||0.011 (0.010)||0.005 (0.009)||0.012** (0.005)|
|Family income [15-30]||0.006 (0.013)||−0.005 (0.013)||0.010* (0.005)|
|Family income [30-50]||−0.033** (0.013)||−0.038*** (0.012)||0.002 (0.006)|
|Family income [50-75]||−0.083*** (0.015)||−0.083*** (0.013)||−0.019*** (0.006)|
|Family income [75+]||−0.151*** (0.015)||−0.130*** (0.013)||−0.053*** (0.007)|
|Black||0.160*** (0.013)||0.147*** (0.012)||0.026*** (0.004)|
|Hispanic||0.088*** (0.013)||0.089*** (0.012)||−0.003 (0.006)|
|MSA level banking variables|
|Bank branches density||0.002* (0.001)||0.006 (0.005)||0.000 (0.000)|
|Bank concentration (HHI)||0.018 (0.039)||−0.025 (0.272)||0.027* (0.016)|
|Average service charge||0.042** (0.020)||0.179 (0.116)||0.019*** (0.005)|
|MSA social economic variables|
|% White||−0.001 (0.001)||0.001 (0.001)||0.000 (0.000)|
|% Black||0.000 (0.001)||−0.004 (0.039)||−0.000 (0.000)|
|% poverty||−0.000 (0.002)||0.026 (0.017)||0.001 (0.001)|
|Number of observations||13,510||13,510||13,510|
The sample includes households who had bank accounts and made more than one appearance in the FDIC Unbanked/Underbanked surveys conducted in 2011, 2013 and 2015. Panel A presents the mean values of variables for the households that had changes in AFS usage over time. Change in unemployment is the difference in the indicator for unemployment between two surveys, change in part-time work status is the difference in the indicator for part-time work between two surveys. Panel B presents estimated marginal effects from a logit model with lagged dependent variable, conditional on households also banked in the previous survey period. Standard errors are clustered at the MSA level;
*p < 0.1;
**p < 0.05;
***p < 0.01
“Why do the Unbanked use Alternative Financial Services”, Federal Reserve Bank of Philadelphia, available at: www.philadelphiafed.org/community-development/publications/cascade/84/04_why-do-the-unbanked-use-alternative-financial-services
In 2010, the Federal Reserve began to regulate banks from automatically enrolling new checking-account customers in overdraft coverage programs. The Durbin Amendment to Dodd-Frank Wall Street Reform and Consumer Protection Act, which took effect in 2011, capped the interchange fees that large banks can charge on debit card transactions.
“Are banks too expensive to use?” by Lisa Servon, The New York Times, Oct 29, 2014, available at: www.nytimes.com/2014/10/30/opinion/are-banks-too-expensive-to-use.html
The 2009 survey is not included for the analysis, as observations for the AFS variable (if household has used AFS in the most recent year) were missing. The 2015 survey is used in the main analysis because it has observations on several variables which were not observed in the earlier surveys (such as income volatility, banking quality of service and experience with banking).
Albanesi, S., DeGiorgi, G. and Nosal, J. (2017), “Credit growth and the financial crisis: a new narrative”, NBER Working Paper No. 23740. National Bureau of Economic Research, Boston.
Backup, B.R. and Brown, R.A. (2014), “Community banks remain resilient amid industry consolidation”, FDIC Quarterly, Vol. 8 No. 2, pp. 33-43.
Birkenmaier, J. and Tyuse, S.W. (2005), “Affordable financial services and credit for the poor: the foundation of asset building”, Journal of Community Practice, Vol. 13 No. 1, pp. 69-85.
Bord, V.M. (2017), “Bank consolidation and financial inclusion: the adverse effects of bank mergers on depositors”, Unpublished Manuscript, Harvard University.
Bradley, C., Burhouse, S., Gratton, H. and Miller, R. (2009), “Alternative financial services: a primer”, FDIC Quarterly, Vol. 3 No. 1, pp. 39-47.
Celerier, C. and Matray, A. (2018), “Bank-branch supply, financial inclusion and wealth accumulation”, SSRN Working Paper, available at: http://dx.doi.org/10.2139/ssrn.2392278
Despard, M.R., Perantie, D.C., Luo, L., Oliphant, J. and Grinstein-Weiss, (2015), “Use of alternative financial services in low- and moderate-income households: evidence from refund to savings”, CSD Research Brief, pp. 15-57.
Duflo, E. and Saez, E. (2003), “The role of information and social interactions in retirement plan decisions: evidence from a randomized experiment”, Quarterly Journal of Economics, Vol. 118 No. 3, pp. 815-842.
Federal Deposit Insurance Corporation (2015), “National survey of unbanked and underbanked households”, available at: www.economicinclusion.gov/surveys/2015household/banking-status-findings/
Frumkin, S. (2006), “Reaching minority markets: community bank strategies”, Community Development Insights, Community Affairs Department, Office of the Comptroller of the Currency.
Goodstein, R.M. and Rhine, S.L.W. (2017), “The effects of bank and nonbank provider locations on household use of financial transaction services?”, Journal of Banking and Finance, Vol. 78, pp. 91-107, available at: https://doi.org/10.1016/j.jbankfin.2017.01.016
Gross, M.B., Hogarth, J.M., Manohar, A. and Gallegos, S. (2012), “Who uses alternative financial services, and why?”, Consumer Interests Annual, Vol. 58, pp. 2012-2057, available at: www.consumerinterests.org/cia2012
Ivković, Z. and Weisbenner, S. (2007), “Information diffusion effects in individual investors’ common stock purchases: covet thy neighbors’ investment choices”, Review of Financial Studies, Vol. 20 No. 4, pp. 1327-1357.
McKernan, S., Ratcliffe, C.E. and Kuehn, D. (2010), “Prohibitions, price caps, and disclosures: a look at state policies and alternative financial product use”, Journal of Economic Behavior and Organization, Vol. 95 No. Nov, pp. 207-223.
Melzer, B. (2009), “The real cost of credit access: evidence from the payday lending market”, The Quarterly Journal of Economics, Vol. 126 No. 1, pp. 517-555.
Morgan-Cross, C. and Klawitter, M. (2011), “Effects of state payday loan price caps and regulation”, Evans School of Public Affairs, University of Washington, DC.
Nguyen, H. (2019), “Are credit markets still local? Evidence from bank branch closings”, American Economic Journal: Applied Economics, Vol. 11 No. 1, pp. 1-32.
Park, K. and Pennacchi, G. (2009), “Harming depositors and helping borrowers: the disparate impact of bank consolidation”, Review of Financial Studies, Vol. 22 No. 1, pp. 1-40.
Prager, R.A. (2014), “Determinants of the locations of alternative financial service providers”, Review of Industrial Organization, Vol. 45 No. 1, pp. 21-38.
Prager, R.A. and Hannan, T. (1998), “Do substantial horizontal mergers generate significant price effects? Evidence from the banking industry”, Journal of Industrial Economics, Vol. 46 No. 4, pp. 433-452.
Rhoades, S.A. (1993), “The Herfindahl-Hirschman index”, technical note, Federal reserve bank of St. Louis, available at: https://fraser.stlouisfed.org/files/docs/publications/FRB/pages/1990-1994/33101_1990-1994.pdf
Temkin, K. and Sawyer, N. (2004), “Analysis of alternative financial service providers”, The Urban Institute research report, available at: www.urban.org/research/publication/analysis-alternative-financial-service-providers/view/full_report
Weaver, A. and Galperin, R.V. (2014), “Payday lending and the demand for alternative financial services”, Community Development Discussion, Paper No. 2014-01, Federal Reserve Bank of Boston.
Zinman, J. (2010), “Restricting consumer credit access: household survey evidence on effects around the Oregon rate cap”, Journal of Banking and Finance, Vol. 34 No. 3, pp. 546-556.
The author of this article has not made their research data set openly available. Any enquiries regarding the data set can be directed to the corresponding author.