You Can Take It To The Bank:

Demographic, Socioeconomic, and Financial Knowledge of the Unbanked and

Underbanked

Draft

Elizabeth Breitbach

University of Nebraska

10/19/2012

Abstract

Checking and savings accounts are frequently used financial instruments by U.S. households. As household

transactions shift from cash toward e-money, these financial accounts become even more important for economic

well-being. Households without these accounts may be able to participate fully in the economy, but may incur

additional costs by using alternative financial instruments. The focus of this research is to study the segment of the

U.S. population who are considered unbanked and underbanked. The study draws on three large, national data

sets: (1) a survey of Financial Capability in the United States by the Financial Industry Regulatory Authority

(FINRA), a National Survey of Unbanked and Underbanked Households by the Federal Deposit Insurance

Corporation (FDIC), and, (3) the Survey of Consumer Finances by the Federal Reserve System. The study

investigates the economic and demographic characteristics of unbanked and underbanked households across the

three surveys to explain which households have a low level of banking participation.

Keywords: Checking, Bank, Consumer, Finance

JEL Code: Personal Finance (D140), Banks; Other Depository Institutions; Mico Finance Institutions; Mortgages

(G210)

2

Introduction:

As with much of the world, individual financial decisions in the U.S. are involved at some level with

the banking industry, even if an individual chooses to avoid banking institutions altogether. These banking

relationships are significant for several reasons. Individuals with traditional transaction accounts are

found to have higher levels of savings than their unbanked counterparts. Not only do bank accounts

promote increased saving, they often offer check cashing and bill paying services at a lower cost than

alternative financial products such as non-bank money orders or bill pay and non-bank check cashing

services. For most households it would seem impossible not to have a transaction account to make day-

to-day financial payments, obtain cash for purchases, or deposit a paycheck and other checks. For

approximately 7.5% of U.S. households, however, access or use of banking for transactions purpose never

occurs because they do not have a checking or savings account.

Holding some form of transaction account is an essential tool of day to day life. Often utility

companies and other billing agencies require payment to be made by check, online, or by cash at the

office location. Having a checking account offers easier payment since it does not require the customer to

drive to the corporate location, during business hours, to pay in cash. If a household has a traditional

checking account, they can mail in a check, sign up for automatic bill pay, or make online payments. The

ease in payment with a transaction account reduces the cost bore by the individual, and with automatic

bill pay, it can be easily and immediately used to prevent late fees from accruing.

Not only is bill paying easier with a transaction account, budgeting and determining areas where

expenses may be cut can be better examined and is more likely to occur if the household has an account

(Hogarth & Anguelov 2004). Creating a budget and tracking how well the household is sticking to their

goals can be easier with a monthly statement of transactions, typically provided with a checking account.

Without a transaction account, keeping receipts and combining income and expenses into a monthly

statement can be difficult and time consuming. Utilizing a bank prepared statement ensures all income

3

and expenses made with the account are included and the totals can be seen with little effort by the

household. Unnecessary or frivolous expenses can be tracked using monthly statements of usage.

Carefully analyzing one’s monthly statement can ensure a household is aware of where their money is

going each month and can influence future purchasing behavior. For example, after examining a bank

statement the household may become aware of the total monthly purchases at coffee shops and could

find ways to decrease this expense, either by purchasing a coffee machine or limiting their purchases to

fewer times per week. This increased awareness is one way statements can help reduce frivolous

spending and, in turn, increase savings.

A transaction account also offers a proof of payment. A check or other form of bank-provided

proof can verify that an individual made a payment and the date it went through the banking system.

Proof can be important in financial transactions that are affected by legality or timing. For example,

verifying the date a check was written and cashed can help avoid late fees, along with offering proof of

payment in the secondary market. Cash payments do not have this benefit, it can be difficult to prove

that a payment was made and when it occurred.

Receiving income also is more efficient with a transaction account. With the increasing popularity

of e-transactions, many employers and other income sources have come to prefer direct deposits to

payroll checks (Anguelov, et al 2004). The Federal Government is also following this trend; the U.S.

Treasury Department enacted a voluntary program called Electronic Transfers Accounts (ETA) for anyone

who receives a payment from the Federal Government. An ETA is a low cost bank account that allows for

the federal payment to be directly deposited each month (Department of the U.S. Treasury 2012).

If a household has a transaction account where income can be deposited, it can be received faster

and with less hassle than going to a bank or other business to cash the check. When an individual cashes a

check at a non-bank business they are likely to pay for the service, either with a flat fee or as a percentage

of the check amount. These transactions costs can become a relatively large percentage of a typical

4

household’s income if the transactions are frequently made (CCCS 2010). These additional costs can

contribute to income constraints that force households to choose between paying bills and other

necessities.

Alternatives have arisen due to individuals lacking traditional bank accounts, whether by choice or

allowance. To pay bills, non-bank money orders and prepaid cards have evolved to meet the needs of

unbanked households. To meet the demand for non-bank check cashing, some grocery and other retail

stores cash checks for a fee, as well as specific businesses that solely cash checks. These services can be

very costly, especially when transactions are completed on a reoccurring basis (Bell 2011). When

managed properly, a traditional checking account can have very low to no fees. These cost savings can

help lower income households meet debt obligations or accumulate funds that accrue interest.

The advantages of a checking account so far have focused on the ease and low cost of this service.

Another reason for using a transaction account is increased safety over holding cash or a prepaid card. It

is not safe to hold a significant amount of cash in a home or on person because the individual generally

assumes the risks associated with loss from theft, fire, and misplacement. Even when insurance does

cover some of these costs, it may only be a portion of the total loss. By contrast, a bank or financial

institutions assume a certain degree of risk for checking accounts that protect the customer. For

example, if a checkbook is stolen or lost it can be closed or flagged so the bank is aware the account

holder is not the individual writing checks. This becomes more difficult when the household primarily

relies on cash or prepaid cards. It is nearly impossible to track the use of cash to make purchases. If a

prepaid card is lost or stolen, it is possible to shut the account down, although a reactivation fee and new

card fee may be incurred.

Another benefit of holding a transaction account at a banking institution is federal insurance

through the Federal Deposit Insurance Corporation (FDIC). Even if a bank goes out of business individuals

will receive their funds up to the insured amount. Most prepaid cards are not federally insured and do not

5

offer the same protection if the corporation goes out of business. This creates an additional risk with

these services as many of them are relatively new entities that rely heavily on partnerships with retail

outlets (Bell 2012).

A traditional bank account used responsibly not only has the advantages discussed above, but it

can also lead to future positive outcomes. Studies have shown that households with a transaction

account are more likely to have savings accounts than their unbanked counterparts (Beverly, et al. 2004).

Additional savings can be a benefit for low income households who have trouble meeting unexpected

expenses. Holding a rainy day fund can help a household avoid having to secure a loan to meet

unexpected expenses, which can be especially costly if the household has a low credit rating. Payday

loans are typically short-term loans that have high annual percentage rates (APRs), which in extreme fees

can be in excess of 400%. Having the funds to address debt obligations when they arise will lead to

further cost savings. Furthermore, an account can also have a positive effect on an individual’s credit

score, increasing household access lower cost loans.

A well-managed transaction account not only leads to greater savings, but other ‘good’ financial

decisions as well. Not only is saving for unexpected expenses important, but retirement savings, savings

for durable goods, and savings to improve one’s education can also be helpful to further increase the

future income of a household. This can improve not only the respondent’s current financial well-being,

but their future wellness as well. Having additional savings can serve as a cushion for unexpected

expenses that arise or can offer a way for households to purchase relatively expensive durable items

typically purchased with credit. Using money to purchase these items can lead to further savings by

avoiding interest payments and other fees incurred when using credit. Future well-being must also be

considered, as savings for additional education or retirement can put the household in a better financial

position. With the uncertain future of Social Security and Medicare, savings may be vital as a household

6

moves toward retirement. Increased wealth can occur not only through current savings, but investing in

one’s human capital to grow future earning potential.

In addition to households who avoid traditional banking services, there are households who use

these services while supplementing them with costly alternatives. Approximately 18% of households fall

into this category of underbanked. Although these households do receive the benefits of a transaction

account, they do not seem to take full advantage of all the services a bank account offers. This behavior

warrants further discussion due to the additional costs incurred by the supplemental services.

These alternatives include frequent use of non-bank money orders or check cashing services,

which is particularly concerning. Most traditional transaction accounts include checks or a debit card as a

form of payment method from the account. The use of non-bank money orders indicates these

households are not fully aware of the services an account offers. As mentioned, the fees on check-cashing

can be a significant portion of a household’s income if used frequently on a reoccurring basis; most

banking institutions offer free check cashing services to customers and direct deposit is encouraged.

The other services that are used in determining whether a household is underbanked is the use of

non-bank short term loans, including payday loans, pawnshops, tax anticipated refund loans, and rent-to-

own services. Compared to both traditional bank loans and credit cards, these loans are less than ideal

due to their high interest rates and service fees.

While there has been research examining who the unbanked are and why they do not have an

account, literature on the underbanked is nearly nonexistent, other than research on a specific

alternative service used to define them. To improve on current literature in this area, multiple data sets

from recent years will be used: the 2009 FINRA National Capabilities Survey, State-by-State (FINRA), the

2009 FDIC Survey of Unbanked and Underbanked Households (FDIC), and the 2010 Federal Reserve’s

Survey of Consumer Finances (SCF). The extensive information included in these data sets is relevant

since previous literature has not had as much information on the financial knowledge of the household or

7

their use of alternative services. This analysis will offer a more complete understanding of what these

households are using to make day-to-day transactions and whether information and knowledge on the

benefits of bank accounts can move these households toward a higher level of banking participation.

It is first crucial to discuss who the unbanked and underbanked are by examining their

demographic and socioeconomic characteristics, along with their financial knowledge. Describing their

characteristics is important in understanding which subgroups of the population hold this level of banking

participation. If these households are not aware of the benefits, marketing tools can be used to move

these households to a higher level of participation. Financial knowledge is also expected to have an

impact on banking participation. If a household does not have a traditional bank account, or supplements

it with alternatives, it may be the case that providing more financial education will help in properly

managing an account and improve the likelihood of utilization.

Households with low levels of banking participation are more likely to be Hispanic or African

American, young, and low income. This analysis will further look at the effect of income by not only

exploring the level of income, but whether the household has experienced a change in income. This

question will offer insight into the effect a recession can have on banking participation and indicate the

potential effect a savings account can have. An indicator for the financial knowledge of a respondent can

be obtained in the FINRA data set. Using this information it can be seen whether it is the demographic

characteristics of a household alone that determine banking participation or if education has an effect. It

is the hope that financial knowledge does have an impact so proper education and information can lead

households to a higher level of banking participation.

There are obvious benefits to having a transaction account, raising the question of why some

households would choose to avoid their services. Understanding which households are choosing to be

unbanked and underbanked is an important piece in comprehending this decision. Exploring which

8

households have a low level of involvement is essential if it is the hope of policy makers and banking

institutions to move these households toward a higher level of banking participation.

Literature Review Unbanked

The first question to be answered is what it means to be unbanked. Grimes, et al. (2010) define

unbanked as not having “any type of commercial bank account.” Hogarth, et al. (2005) define unbanked

as individuals not having a “transaction account.” They cite that a transaction account as including

“checking, savings, money market accounts at depository institutions and brokerage firms and call

accounts.” Rhine and Greene (2006a) and Rhine, et al. (2006b) define unbanked, similar to Grimes, et al.,

as lack of a checking or savings account, whereas Paulson and Rhine (2008) separate checking and saving

accounts and explore banking participation at the individual account level. A review of the literature

presents few discrepancies in defining unbanked, which enables ease in comparing results of different

studies.

Not only is it important to discuss how these papers describe an unbanked household, it is

essential to compare the data sets that are used. Some data sets focus on a subset of the United States

population, like those in the Chicago metropolitan area, or a specific ethnic group, such as Hispanics.

These differences are important to note when making comparisons across surveys. Grimes, et al. (2010)

uses the Council for Economic Education’s National Financial Services Survey, conducted in 2008. This

data set includes 1,759 respondents from the United States. Hogarth, et al. (2005) employs the Survey of

Consumer Finances from 1989 to 2001. When looking at the unbanked over time, the authors use the

surveys separated by year and as a time series to show differences in banking participation across time.

The discussion in this paper will focus on the full sample, which includes information on nearly 16,000

households following the U.S. Census population of the United States.

The focus of Rhine and Greene’s (2006a) study is exploring the banking participation of U.S.

immigrants. To get information on the length of time the immigrant was in the United States and other

9

related information the authors used the Survey of Income Program Participation. This survey includes

information on both U.S. born households and Immigrant households. For better comparison, the results

presented in the literature review will focus on the U.S. born households. Amuedo-Dorantes and Bansak

(2006) also explore banking participation focusing on immigrants, specifically Mexican immigrants. The

primary purpose of the paper is to determine what factors affect the amount of money the immigrant

transfers home. One expected determinant of the aggregate transferred is whether the respondent

opened a bank account while in the United States, and the authors run regressions on banking

participation to better understand the decision to hold an account. The data set used to analyze these

questions is the Mexican Migration Project from 1970 to 2004. For the banking participation regression,

there are 2,978 observations. Rhine, et al. (2006b) explored racial and ethnic differences in banking

participation using the Metro Chicago Information Center and the Federal Reserve Bank of Chicago’s

annual survey. A total of 2,339 respondents were included in the banking analysis. The final data set that

will be extensively explored in this section is by Paulson and Rhine (2008). This paper focuses on an even

more specific ethnic group in the United States, Hmong immigrants living in Minnesota. The data set

includes information on 202 respondents from this subset of the United States and 202 control

respondents from similar neighborhoods.

After defining what it means to be unbanked, it is necessary to explore who is unbanked, including

demographic and socio-economic characteristics, along with the amount of credit the household has

access to and level of assets they hold. Gender has had both mixed results in its significance and its sign.

Rhine and Greene (2006a) found that female immigrants are less likely to be unbanked, relative to

immigrant males. However, in another article, Paulson and Rhine (2008) found that male head of

households were less likely to be unbanked relative to situations where females were the head of

household. While these studies show conflicting significant results, others have found no significant

relationship between gender and banking participation (Amuedo-Dorantes and Bansak 2006, Grimes, et

10

al. 2010, Rhine, et al. 2006b). Hogarth, et al. (2005) explored gender and banking participation over time.

They found that both single males and single females have increased their banking participation between

1989 and 2001. However, over the period males maintained a slightly higher banking participation rate

than females, with a gap of about 4%.

Race and ethnicity are also expected to be determinants of whether or not a household has a bank

account. It has been suggested that if an individual does not speak English as a first language, they may

feel intimidated by the banking system (Rhine and Greene, 2006a). Amuedo-Dorantes and Bansak (2006)

explored Mexican Immigrants who had plans to return to Mexico. While in the U.S., these immigrants

often sent money home through the use of money transfers. When looking at the issue of banked versus

unbanked, the authors found that undocumented workers and those who were in the United States for

only a short time were more likely to be unbanked. Holding a bank account can make this process easier

and less costly. Spader, et al. (2009) also looked at the issue of Hispanics and banking. To increase the

percentage of Hispanics with bank accounts, a television show was developed to create a more favorable

opinion of banks and the services they offer. The show had positive effects on the participant’s opinions

of banks and banking services, but little change was shown in behavior (Spader, et al. 2009). This is an

important result, since a large percentage of the unbanked do not have an account because they dislike

dealing with the institution. Most results have found black and Hispanic households are significantly more

likely to be unbanked, relative to Caucasian households (Hogarth, et al. 2005, Rhine and Greene 2006a,

Rhine et al. 2006b). Grimes, et al. (2010) finds a similar trend but the result is not as significant (significant

at the 10% level).

Grimes, et al. (2010) found that a significantly negative indictor of being unbanked is age. While

this result is significant, .at 08% it is not large in magnitude. This may be due to the fact that the authors

included it as a continuous variable and few changes occur from year to year, whereas more significant

changes take place from decade to decade. Amuedo-Dorantes and Bansak (2006) also included age as a

11

continuous variable; their results were also small in magnitude, but not significant. Paulson and Rhine

(2008) included age as a continuous variable and added age squared to better understand the

relationship. They found that as age increased the respondent was less likely to hold a checking or savings

account, although it was at a decreasing rate. Rhine and Greene (2006a) found that U.S. born individuals

between 18 and 25 are 3% more likely to be unbanked relative to all other age groups. Rhine, et al

(2006b) found a coefficient of similar magnitude and significance. Hogarth, et al. (2005) broke households

in four age cohorts. Relative to 18 to 34 year olds, households in the 50 to 64 years and 65 and over

cohorts are significantly more likely to be banked.

When looking at married versus single households, Rhine and Greene (2006a) and Rhine, et al.

(2006b) found that those who are married are less likely to be unbanked, however, Grimes, et al. (2010)

found it to be insignificant. Hogarth, et al. (2005) separated single male and female households; relative

to married households, single female households are more likely to be banked, while the result for males

is insignificant. Family size, or number of dependents, is another variable that has been included in

analysis of banking participation. Hogarth, et al. (2005) has explored the effect of dependent children in

the household and found that households with dependents are less likely to be banked, but the mean

difference is not significant. Paulson and Rhine (2008) also found that household size did not significantly

impact a households banking participation, while, Rhine and Green (2006a) found that a larger family size

was significantly more likely to be unbanked.

Education is also found to be a significant factor in indicating whether or not an individual has a

bank account. Hogarth, et al. (2005) uses a set of dummy variables to specify the education level of

respondents. They find that, relative to those with a high school degree, individuals with a high school

degree or less are significantly less likely to be banked, and those with some college or more are

significantly more likely to be banked. Grimes, et al. (2010) includes a dummy variable for whether the

respondent has any post-secondary education. Relative to those with a high school degree or less, these

12

individuals are significantly less likely to be unbanked. Rhine and Greene (2006a) and Rhine, et al. (2006b)

both find that those with a high school degree or less are significantly more likely, relative to those with

at least some college education, to be unbanked. Paulson and Rhine (2008) found that, relative to those

with less than a high school degree, respondents with a high school degree, some college experience, and

those with a college degree or higher are significantly more likely to be banked. Amuedo-Dorantes and

Bansak (2006) did not find significant results for education, but this may be due to the fact that education

was included as a continuous variable.

Work force participation is another possible determinant of banking participation. Previous

studies that used a dummy variable to indicate if the respondent was in the work force and has a work

commitment have found mixed results. Rhine and Greene (2006a) included work commitment as a

continuous variable equal to the number of hours worked. As work commitment increased respondents

were less likely to be unbanked, but the result was not significant. Grimes, et al. (2010) included a dummy

variable for whether the respondent was full time, part time or self-employed. Being employed lead to a

lesser likelihood of being unbanked, but the result was not significant. Hogarth, et al. (2005) took a more

in depth look at work status. Variables were included for working, retired, unemployed – looking, and

unemployed – not looking. Relative to head of households who are unemployed – not looking, those who

are working and retired are significantly more likely be banked, while unemployed – looking are

significantly more likely to be unbanked.

Many studies have concluded that having low income is not only a significant factor effecting

banking participation, but it is one of the primary determinants in predicting if households are unbanked

(Grimes, et al. 2010, Hogarth et al. 2005, Paulson and Rhine 2008, Rhine, et al. 2006b, Rhine and Greene

2006a). Amuedo-Dorantes and Bansak (2006) include variables for living standards instead of income,

these variables are not significant indicators of banking participation.

13

It is not only income, but a household’s overall financial situation that affects banking

participation. The inclusion of net worth and access to credit as controls for banking participation vary

across studies. Rhine and Greene (2006a) included a set of variables to indicate net worth of the

respondent. Relative to households reporting no net worth, those with positive net worth, and even

those with negative net worth, are significantly less likely to be unbanked. Hogarth, et al. (2005) includes

net worth as a set of dummy variables with similar results. Hogarth also includes variables on whether the

respondent is a home owner and a vehicle owner. Results indicate that homeowners and those owning a

car, both newer and older, are significantly more likely to be banked. Grimes et al. (2010) and Rhine, et al.

(2006b) include a dummy variable for whether the respondent owns their home, results are consistent

with those found by Hogarth, et al. (2005).

Grimes, et al. (2010) includes a dummy variable for whether or not the household owns a credit

card. They find that households holding at least one credit card are significantly less likely to be

unbanked. Hogarth, et al. (2005) includes access to credit as a dummy variable for whether the

respondent has been rejected or obtained a lesser amount of credit than requested. These households

are significantly more likely to be banked, relative to those who have not been rejected. While this result

is not expected, it may be explained by unbanked households not making an attempt to apply for credit.

If this is true, it is mainly banked individuals attempting to access credit, consequently they are the

individuals being rejected.

Underbanked

The majority of current literature is in the area of unbanked individuals, however, there is another

category of individuals who are not unbanked but seem to be making costly financial decisions that

warrant further investigation. The FDIC defines an underbanked household as “those that have a checking

or savings account but rely on alternative financial services. Specifically, underbanked households have

used non-bank money orders, non-bank check-cashing services, payday loans, rent-to-own agreements,

14

or pawnshops at least once or twice a year or refund anticipation loans at least once in the past five

years.” There has been little research done specifically in the area of underbanked, but use of the

alternative services that define the underbanked have been explored to various degrees.

The first two alternative services that make up the underbanked are the use of money orders and

non-bank check cashing services. Schuh & Stavins (2011) found that 11% of check users had also used

money orders. The authors also reported that consumers using money orders paid a higher percentage of

their total transactions in cash rather than bank account deductions or online bill paying. Paulson and

Rhine (2008) explored the use of non-bank money orders and check cashing services together. They

found that low income households are significantly more likely to use these services, and as use

significantly increases so does household size. Rhine, et al. (2006b) explored obtaining financial services

from currency exchanges, including cashing checks, purchasing money orders, paying bills, and wire

transferring money. The authors found that low income households were nearly 40% more likely to use

these services. Those who are 25 years or younger, are black or Hispanic, or have a high school degree or

less, are more likely to use these services as well. An indicator for whether the household was unbanked

was included in the analysis. Households without a bank account were significantly more likely to use

financial services from currency exchanges.

With low income being a significant determinant of usage for both money orders and check

cashing services, a look at the fees associated with these services is warranted. Some check cashing

outlets provide money orders for free when another service, such as check cashing, is purchased. Fox and

Woodall (2006) report that the average fee for a $100 money order was $1.08 and ranged from 50 cents

to $16. They report the fee has increased by 8% since 1997. The authors also compared outlets’ fees to

those at the United Postal Services, which charge 95 cents per money order. Fox and Woodall (2006) also

explored the fees on check cashing services. The cost of cashing the check, and whether the outlet would

even cash the check, depended on the type of check. For a government benefit check, 94% of outlets

15

cashed the checks with an average fee of 2.44%. The authors reported an increase in the cost of cashing

a Social Security check; in 1997 the fee was 2.11% of the check value, increasing to 2.44% in 2006.

Paychecks are also widely accepted in check cashing outlets, 93% were willing to cash them. The average

fee for cashing these checks was 2.52% and ranged from 1% to 5%. The average fee increased if the check

was a hand-written paycheck (4.11%) and ranged from 1% to 10%. The most expensive type of check to

cash is a personal check, with only half of check cashing outlets willing to cash them. The average fee for a

cashing a personal check was 8.77%, with fees ranging from 2% to 15%, a drop from 9.36% in 1997.

Accompanying this decrease was and increase in outlets willing to cash personal checks.

The high cost of check cashing services has provoked some state governments to intervene in the

market, capping the fees that businesses can charge. Governments have also worked to lower the cost of

banking by capping the fees banks charge for minimum balances and number of checks written.

Washington (2006) found that these efforts resulted in a three to four percentage point decrease in the

number of low-income minority households that were unbanked. The results became stronger the longer

the regulation was in place, but due to the lack of immediate success of these programs, some were

cancelled, bringing back the original problem of high costs. Washington cites two years as the time frame

when results of the regulation begin to show, which she attributed to the lack of advertisement by banks

of the new accounts.

Payday loans and other high interest short-term loans are relatively new to the market, so little

research has been done in the area. There were virtually no payday loan businesses in 1990, but by 2001

the number was approximately 12,000 to 14,000 (Consumer Federation of America & U.S. Public Interest

Research Group 2001). Most of the research done in the area of payday loans and other alternative

services explores access to the loans (Melzer 2011, Skiba and Tobacman 2007) and effects of policy and

regulations on the use of these services (Carter 2012, Avery 2011, Peterson 2007, Stoesz 2012, Hill, et al.

1998, Edmiston 2011).

16

One study that investigated payday loan customers is a 2008 study by Lawrence and Elliehausen.

When looking at a mean comparison between payday advance customers and all adults, those who have

used payday loans were more likely to have incomes between $25,000 and $49,999. Individuals under 35

represented a higher percentage of payday users. Both these results were confirmed by Elliehausen

(2009), who also found that payday borrowers had a limited amount of liquid assets, had a high school

diploma or some college education, and had experienced credit limitations in the past 5 years. Chatterjee,

et al. explored the use of high interest loans to meet short term need. The authors found that older

individuals, males, whites and those without children under the age of 18 were less likely to use

alternative banking options. Educational attainment and income were also negatively associated,

specifically with payday borrowing. Stegman and Faris (2003) focused on low income individuals in North

Carolina and found that African American and younger households were more likely to use payday loans.

The authors also included variables for whether the respondent received welfare and whether the

respondent had a savings account.

Fox and Woodall (2006) also explored payday loans. They reported that in order to qualify for a

payday loan a customer needs a bank account and a source of income. The authors found that the

average maximum loan size was $696, ranging from $250 to $5,000. To determine the cost of a payday

loan, the authors inquired about borrowing $300 for two weeks. The average cost of this short term loan

was $46.85, or 406% APR. The highest fees were charged in states that did not have caps on the interest

rate. Under the Truth in Lending Act, payday lenders are required to quote the cost “as an interest rate if

any cost in quoted,” in attempt to help the consumer understand the actual cost of the loan. Fox and

Woodall report that three fourths of the clerks reported the cost for the entire loan amount or the loan

cost per $100 borrowed. Another 8% of the clerks refused to disclose the cost and only 17% reported the

cost as an interest rate.

17

Those households who choose to purchase or rent items from Rent-to-Own (RTO) stores and have

a transaction account are also considered to be underbanked. RTOs can be more expensive than other

forms of short term credit provided by banking institutions. Anderson and Jaggia (2012) explored three

different categories of using RTO. First, the customer can choose to return, meaning “payments cease and

the merchandise is returned to the store, perhaps involuntarily.” The second option is purchase,

“ownership is transferred to the customer, possibly through the exercise of an early payment option.”

Skip is the final category, meaning “payments prematurely stop but the merchandise cannot be recovered

by the store for some reason – ‘the customer skipped with it.’” The authors then combine all these to

explore all RTO customers. The authors found that nearly 60% of RTO customers were under the age of

25, 25% were male, and 25% were married. The authors also found that 45% were repeat customers.

Among the three categories 62.2% of customers returned the good, 20.1% purchased and 3% skipped.

The other category, which is left out for this analysis, is the 14.7% of contracts that remain open. Looking

at the authors’ regression analysis, it was found that older and repeat customers are significantly less

likely to return or skip, and those who are unemployed are significantly less likely to return the item;

whereas married households are significantly less likely to skip.

In an earlier paper, Anderson and Jaggia (2009) explored various types of goods that can be

purchased/rented at a RTO store. They focused on appliances, electronics, and furniture. They found that

older, married customers were more likely to rent furniture. Those receiving a form of government aid

are significantly less likely to rent electronics and furniture. McKernan, et al. (2003) also explored use of

RTOs. They reported that households who do not own their home, have an income of $15,000 to $24,999,

are separated or African American, and those with a high school degree or lower are more likely to use

RTOs. Individuals who were older and retired were significantly less likely to use them. Geography had an

impact on use as well; those in nonmetropolitan areas were more likely to use RTOs, as were households

in the Midwest and South.

18

When check cashing outlets first emerged, the pawn broking business began to decline. It then

rebounded in the mid-1970s and continued to grow rapidly throughout the ‘90s (Caskey 1994). While

there is relatively little current research done on pawn shops in the United States, there was a survey

done in 2010 on pawn broking in the UK. The report finds that pawn broking customers tend to be

women with families. Three-fourths of the customers were between the ages of 20 and 49. A small

percentage, 20% owned their home, either with a mortgage or outright. Work statuses of the customers

were also explored; 25% of the customers were unemployed and looking for work, 27% were full time

employees, and 13% were part time employees. Another interesting finding was that 53% of customers

lived in a household with no one working; however, most of these households were comprised of single

parents or adults living alone. Banking status was also reported, with 11% of customers considered

unbanked. Unbanked was defined as not having a traditional bank account or a Post Office Card Account

(Collard and Hayes 2010).

The final alternative use that makes up the underbanked is consumers who use tax anticipated

refund loans. To be considered underbanked by the FDIC, a household must have used this service at

least once in the past 5 years. Elliehausen (2005) explored the use of refund anticipation loans. The

majority of refund anticipation loans are used by $15,000 to $24,999 and $25,000 to$39,999 income

groups. The users of refund loans also tend to be younger, married with children, and have high

consumer debt payment to monthly income ratio.

Financial Literacy

Financial literacy has been an important topic during the current financial crisis. Many experts

believe it is a lack of financial knowledge that leads individuals to make costly financial decisions

(Bernheim and Garrett 2003, Fox, Bartholomae and Lee 2005). However, the focus of the current financial

crisis on improving financial knowledge has been met with skepticism and criticism (Willis, 2011). There

are many studies that have been published showing little to no improvements in knowledge after an

19

intervention has been made, although these findings may be the result of flaws in the transfer of

knowledge (Cole and Shastry 2008, Hathaway and Khatiwada 2008, Carswell 2009).

The question of what determines banking behavior and other financial behaviors may not be

completely based upon our demographic and socioeconomic characteristics. If our behaviors were solely

determined by characteristics we are unable to alter, changing our behavior or preventing ‘bad’ financial

behavior from occurring would be difficult. It is the hope of researchers and policy makers that something

can be done to modify financial behavior, most believe this is financial literacy and financial knowledge.

Before going more in depth into the discussion of financial literacy and behavior, it is important to

address the question of what financial literacy is and how it is determined if someone is financially

literate. As researchers begin to characterize someone as financially literate or illiterate, distinction

between vocabularies must be clear. In Huston’s 2010 article, she raises issue with the fact that experts

are using terms such as financial knowledge, financial literacy, and financial education interchangeably.

Though these terms have been used synonymously in the past, they may not hold the same meaning to

all experts, especially in different fields. To ensure these terms are used efficiently, common definitions

needs to be established.

Huston offers a suggestion on how we should define financial literacy: “To be financially literate,

individuals must demonstrate knowledge and skills needed to make choices within a financial

marketplace that all consumers face regardless of their particular characteristics.” She stresses two parts

of this definition: knowledge and skills. Huston also points out the difference between financial literacy

and financial education. She defines financial education as targeted toward “improving a person’s level of

knowledge and/or ability, can and should be tailored to suit different demographics, life stages, and

learning styles.” She believes the difference between financial literacy and financial education is that

financial education is used to teach individuals to become financially literate (Huston 2010).

20

An alternative definition for financial literacy comes from Remund in his 2010 article: “Financial

literacy is a measure of the degree to which one understands key financial concepts and possesses the

ability and confidence to manage personal finances through appropriate, short-term decision making and

sound, long-range financial planning, while mindful of life events and changing economic conditions.”

Remund includes the two components that Huston stresses while adding a time horizon and changing

economic conditions. His definition seems to be stronger than Huston’s, but he does not offer clear

descriptions of many ideas he presents. He fails to explain what he believes to be the key concepts and

what determine sound decisions. Huston’s definition allows for knowledge without action while

Remund’s requires knowledge and action (Remund 2010).

Defining financial literacy is an important task not only for providing more consistent research, but

also in allowing for better comparison across studies. A better definition of financial literacy comes from

the Jump$tart Coalition (2007), which defines financial literacy as “the ability to use knowledge and skills

to manage financial resources effectively for a lifetime of financial wellbeing.” They also state that

“financial literacy is not an absolute state…(it) refers to an evolving state of competency that enables

each individual to respond effectively to ever-changing personal and economic circumstances.” This is an

improved definition of financial literacy because it addresses several issues the above definitions do not.

For example, the Jump$tart definition includes a time dimension, stating that financial literacy is a

“lifetime of financial wellbeing.” This is important because it recognizes that financial literacy is not just

knowing about your finances today but also knowing how choices today impact future financial wellbeing.

Another important element that is included in this definition is the recognition that financial

markets change over time. The financial decisions made by an individual’s grandparents are not

necessarily the same decisions that are relevant today. The introduction of new financial instruments and

changing government involvement will introduce new choices and opportunities. Additionally, this

definition is superior to others that have been presented in that it allows for individuals to make decisions

21

that may be seen as less than ideal as long as an individual is “responding effectively” to their personal

circumstance.

It has been stated previously that education has an effect on the banking participation of

respondents. Many studies have found that those with higher levels of education are more likely to be

financially literate (Lusardi and Mitchell 2008, Van Rooij, et al. 2011a, Fonesca, et al. 2012, Worthington

2006). There are two possible reasons for this; first, those who have higher levels of education may learn

financial knowledge in their extra years of schooling. The other possible reason is a self-selection issue,

those who have higher levels of financial knowledge may self-select to attend higher levels of education

because they understand the financial benefits better than those who choose not to attend.

While there has been little done on the impact of financial literacy on banking participation,

literacy has been shown to have an effect on other financial behaviors. High levels of financial knowledge

have been found to lead to more responsible credit card behavior (Robb 2011, Wickramasinghe and

Gurugamage 2012), increased patience (Hastings and Mitchell 2011), planning for retirement and wealth

accumulation (Behrman, et al. 2012, Fernandez et al. 2010, Lusardi and Mitchell 2007, Lusardi and

Mitchell 2008, Van Rooij, et al. 2011a), and participating in the stock market (Abreu and Mendes 2010,

Van Rooij, et al. 2011b). Since many other financial behaviors are impacted by financial literacy, it is

expected that it will also have an effect on banking participation.

One study that explores the effect of financial literacy on banking participation is Grimes, et al.

The authors measured financial literacy in a number of ways. First, the authors explored financial literacy

by using the number of correct answers on a set of financial literacy questions. The set of questions can

be found in Appendix 1. The authors found that, out of 7 questions, the average percentage answered

correctly was 48.16%, or approximately 3 questions.

A second measure of financial literacy was through the use of two dummy variables indicating

whether the respondent had ever taken an economics course, business course, or a personal finance

22

course in high school. The authors used these variables in a broad definition of economics to indicate if

the respondent had any access to financial education. It was reported that 35.58%, 29.67%, and 10.10%

of respondents had taken an economics, business, or personal finance course, respectively. When

combined into the broad definition, 55.63% had taken one or more of the courses.

The authors first completed a mean comparison, comparing financial knowledge of unbanked and

banked respondents. It was found that banked individuals were significantly more likely to take a business

course. The results were not statistically different for the other courses individually; however, the broad

definition of economics was significant, with more banked individuals taking one or more of the courses.

The other financial literacy indicator, score on the set of financial literacy questions, was also statistically

different. Banked individuals scored, on average, 49.60% compared to the unbanked respondents’ score

of 36.04%.

Using a probit regression, the authors explored the unbanked, controlling for a variety of

demographic, socioeconomic, and geographical factors. Four models were used with each differing on

how courses are included; all included the score on the financial literacy questions. Across all regressions

the sign and significance of the score variable remains the same, higher levels of financial knowledge

leads to less likelihood of being unbanked. When the different dummy variables for courses were

included, controlling for exposure to economics, all had a negative effect on being unbanked. However,

taking at least one or more economic, business or finance course (broad definition of economics) and

taking at least one business course were significant.

Data

The first data set used in this paper is the Financial Capability in the United States –State-by-State

Survey created by the Financial Industry Regulatory Authority (FINRA). This data set is part of a set of

three surveys in the Financial Capabilities Study. The others are the National Survey and Military Survey

(FINRA Investor Education Foundation 2012). All data sets collected information from a unique set of

23

respondents. The national survey was a telephone survey where approximately 1,500 households

responded. The survey was completed between May and July of 2009. The military survey included

information on 800 military service members and spouses. Due to the small sample size of these surveys

these data sets will not be included in the analysis. The State-by-State data set was chosen for this

analysis due to large number of observations, approximately 28,146 American adults. A sample of at least

500 respondents from each state and the District of Columbia was obtained by an online survey between

June and October of 2009. The questionnaire is similar across surveys, so comparisons can be easily done.

All results presented in this paper were confirmed using the National survey.

The primary purpose of these surveys was to evaluate the financial capabilities of adults in the

United States. The main content areas covered by the FINRA surveys include: financial capabilities,

financial literacy measures, financial behaviors, financial attitudes, and standard demographic

characteristics. For the purpose of this study, variables from all sections will be used (Applied Research

and Consulting LLC 2009).

In order to compare results across studies in this paper the state data set will be weighted to

match 2008 American Community Survey (ACS). Since the sample distributions are initially by state, the

weights will adjust distributions, by age category to match gender, race/ethnicity, and level of education.

There are three stated goals of the FINRA surveys, with the first being to benchmark key measures

of financial capabilities. Key financial capabilities of interest are listed as “banked” status, access and

participation in retirement savings, and debt burden. The next objective is to understand the

characteristics of relevant households, such as demographic characteristics, financial knowledge, and

behavioral traits. The final goal is to inform public policy based on the results of this study. Using the

characteristics found to be significant, it is the hope of the survey creators that public policies promoting

financial capabilities are put in place (Applied Research and Consulting LLC 2009).

24

The State-by-State Survey was chosen as the primary data set due to the financial knowledge

variables that can be obtained. This data set asked respondents a set of five financial literacy questions to

estimate their financial understanding. Since banking participation requires a relatively low level of

financial knowledge, it is expected that the respondents’ score will have a small but significant effect on

participation. While it is expected that financial literacy will have a small effect in general, the magnitude

of the effect may get larger as we move to a comparison of the underbanked and fully banked. For a

household to be underbanked they must use alternative services that are often more costly than their

traditional counterparts. Having a knowledge that these services tend to be more costly may incentivize

the household to choose traditional services. In addition to the respondent’s total score, individual

questions, which vary in difficulty, will be explored. It is predicted that the relatively easy questions will

have the greatest impact on banking participation due to low level of financial involvement. Given the

richness of the financial literacy variables, the FINRA survey will be primary data set used in the analysis.

The second data set used to analyze differences in the characteristics across banking levels is the

FDIC National Survey of Unbanked and Underbanked Households. This survey was a supplement to the

January 2009 Census Bureau’s Current Population Survey (CPS). Since this data set is linked to the CPS,

there is more information concerning the work status of the respondents than with the FINRA data set.

The full CPS data set includes information on 54,000 households, with nearly 47,000 respondents

completing the supplemental FDIC survey. While this was the number of respondents who began the

survey, the number of questions in the survey varied based on the responses given. If the respondent was

not aware whether the household had a checking or saving account, or refused to answer the question,

the survey ended. The survey was also terminated If the respondent reported that they were “not at all”

involved in making financial decisions, or that they did not know or refused to answer their level of

participation in the decision making process. After these drops were made, the number of observations

used was 45,875. All households that reported knowing whether they had a checking or saving account

25

were included in the unbanked analysis. However, when moving to the underbanked analysis, households

for which the survey was terminated, due to their involvement in making financial decisions, were not

included.

The FDIC Executive Summary states the purpose of the survey is to “address a gap in reliable data

on the underbanked and underbanked households in the United States.” Under the Federal Deposit

Insurance Reform Conforming Amendments Act of 2005 (Reform Act) the FDIC must conduct ongoing

surveys to determine the efforts of banks to serve the underbanked, this survey is conducted in order to

comply with that law.

The sampling method of the CPS is complex. The first step, based on the 2000 census information,

created just over 2,000 geographical areas called “primary sampling units” (PSU) for the entire United

States. These PSUs are formed into strata, by themselves and within each state. A total of 842 PSUs are

sampled. The second step was to choose households within these PSUs to survey. Around 72,000

households are chosen each month, but due to unoccupied households and those who do not respond

because they are absent or refuse to answer, the data set usually falls to around 57,000 households. The

CPS then collects data on the members of the household, applying household responses to all members.

In a given month, information is obtained on approximately 112,000 individuals age 15 years or older,

31,000 children (0-14 years of age), and about 450 individuals in the Armed Forces.

As with most national data sets, the CPS does oversample some groups, requiring the use of

weights to complete an analysis. The first weight included in the data set is the “inverse of the probability

of the person being in the sample.” This weight is fairly consistent for individuals living within the same

state but can differ greatly across states. The CPS also includes weight for non-interviewed households

and ratio estimates. The ratio estimate is a weight that accounts for differences between the sample and

the actual population. The characteristics that are considered are “age, race, sex, and state of residence.”

This is primarily purpose of these weights are for analysis of work force participation. When looking at

26

banking participation the household weight will be applied to all descriptive statistics and regression

analysis.

One limitation of the FDIC survey is that it does not include information on the respondents’

financial knowledge. While the FDIC lacks this information, there are several benefits of the survey. The

first is the large sample size, the FDIC data set has nearly 46,000 observations. Another benefit is the

significant number of banking participation questions asked of respondents. The FINRA survey has a

limited number of questions on why a household is unbanked and no questions on why a household is

underbanked. The FDIC survey has many questions concerning both these “why” questions. Knowing why

a household chooses a given level of participation can offer insight into whether the household is

unbanked by choice or result of refusal. These indicators allow for a more in-depth analysis of

demographic and socioeconomic characteristics and differences on why individuals do not hold a bank

account. The FDIC survey also includes information on whether the household ever had a bank account

and if they plan to open one in the near future. These questions offer interesting distinctions within the

subset of unbanked households.

The final data set that will be used in this analysis is the 2010 Survey of Consumer Finances which

is sponsored by the Federal Reserve Board in cooperation with the Department of the Treasury. The data

was collected between May and December and includes a sample size of 6,492. The purpose of the SCF is

to track changes in the financial situations and participation over time. The SCF has been conducted

triennially since 1983, with panel surveys being completed in 1983-1989 and 2007-2009. The purpose of

the most recent panel data set was to explore the effect the current recession has had on consumer

finances.

The SCF also oversamples select segments of the population to obtain a more accurate picture of

the population. The sample design consists of obtaining “a standard, geographically based random

sample and a special oversample of relatively wealthy families.” Keeping consistent with previous

27

literature that has used the SCF, the descriptive statistics will be weighted, but the regression analysis will

use a repeated imputation inference (RII) technique that addresses the issue of missing observations.

While this paper will use this method, an in-depth discussion will not be included. (For more information

on the RII method see (Montalto and Sung 1996, Kennickell 1998)).

The SCF does have some benefits over the previously presented data sets. The first benefit is that

a previous analysis of the unbanked has been completed using the SCF (Hogarth, et al. 2005). This is

particularly important to compare current results to previous studies ensuring that questions are similar

across the data sets. Another benefit of the SCF is the time trend, it has been given every three years

since 1983. These time factors can be used to look at banking participation over time, and of particular

interest, how banking participation has changed during periods of expansions and recessions. One

limitation of the SCF data set is that it does not include information on underbanked households. There is

one question included concerning the use of payday loans. However, due to the small percentage of total

households, particularly banked households, using this service, determining which households are

underbanked is difficult. Since the underbanked cannot be identified using this data set, it will be left out

and only the previous two data sets will be used in the underbanked analysis.

Since there are three data sets that will be used to analyze banking participation a comparison of

variables across surveys is important to ensure that terms are well defined and discrepancies are pointed

out. For a complete comparison of the variables used in this paper see Appendix 2.

The first comparison to discuss is how the survey chose the respondent. For the FINRA survey the

respondent was selected at random, there was no targeting of “heads of households or primary financial

decision makers” (Applied Research and Consulting LLC 2009). As mentioned above, the FDIC survey is a

supplement to the CPS. The “reference person” for the FDIC survey is the “person who owns or rents the

home” (FDIC 2009). The SCF, like the FINRA survey, does not target the head of household, but the

28

respondent is also not chosen at random. The respondent of the SCF is the “most financially

knowledgeable person in the household (Lindamood, et al. 2007).

The variables that warrant the most discussion are the dependent variables unbanked and

underbanked. The definition of unbanked is fairly consistent across the three data sets. The FINRA data

set requires two questions to determine whether the respondent is unbanked. First they are asked if they

or their household has a checking account. The second relevant question is whether they or their

household has a “saving account, money market account, or CDs.” If a respondent answered yes to at

least one of these questions, they are considered to be banked. If a household reported they did not

know/refused to answer one (and did not hold the other account) or both they were dropped from the

analysis1. The FDIC survey asks one question to get at the same unbanked variable; “Do you or does

anyone in your household currently have a checking or saving account?” The SCF treats the banking

questions similar to the FINRA data sets, asking whether the respondent had a checking account, a saving

account of some type. A difference that should be noted is the inconsistency between previous work

using the SCF and this study. Previous work using the SCF has defined unbanked as not having a

transaction account, including a checking account, savings or money market account, or a call account.

Since the comparison to the other data sets used in this analysis is more important than comparing to

previous literature, the call accounts will not be included in the definition of unbanked2.

The next dependent variable is whether the household is underbanked. As previously mentioned,

the FDIC defines underbanked households as “those that have a checking or savings account but rely on

alternative financial services. Specifically, underbanked households have used non-bank money orders,

non-bank check-cashing services, payday loans, rent-to-own agreements, or pawnshops at least once or

twice a year or refund anticipation loans at least once in the past five years.” This is the definition that will

be used to build the underbanked variable. Due to the newness of the definition different questions were

1 Determining whether 373 households or 1.3% of households were unbanked was not possible due to don’t know/refused responses.

2 There are nine households that have call accounts, but no other transaction account. These households would be considered banked under

the Hogarth, et al. (2005) definition.

29

asked about alternative financial services used by the unbanked. This leads to slight inconsistencies in the

definition.

Since the FDIC developed the definition, the FDIC survey includes all relevant questions to

determine whether a household is underbanked. The difference will come in the FINRA definition of

underbanked. The first discrepancy will be use of non-bank money orders and check cashing services. The

FINRA survey did not ask banked households about their usage of these services, which may

underestimate the number of underbanked households3. Another difference is the inclusion of an auto

title loan. While the FDIC data set does not include this service in its definition, it can be considered an

alternative to traditional banking loans so will be included in FINRA definition of underbanked4. The

FINRA data sets do not include information about frequency of use, leading to the final difference

between definitions. Distinguishing between frequent and infrequent users is not possible, so

underbanked households are those that have taken out or used these services5. As previously mentioned,

the SCF does not include enough information to determine if a household is underbanked. The only

question asked concerning alternative financial services is whether the respondent uses payday loan

services. While this variable will not be used for any analysis purpose, the descriptive statistics will be

reported for comparison purposes.

Most demographic and socioeconomic variables were found in all data sets. The data was

combined in a manner that was consistent with the FINRA data sets. For example, age was included as a

categorical variable in the FINRA data set, and was used as such for all data sets. While most of the

controls were found in all surveys, there were a couple variables that were not common across data sets.

Race/Ethnicity variables varied slightly across surveys. The FINRA and FDIC surveys include similar race

3 If households who use solely money orders and/or check cashing services were excluded from the FDIC definition of underbanked the

percentage of banked households would fall from 20.3% to 7.0% (of banked households). 4 Excluding households who only use auto title loans from the underbanked the percentage of underbanked households fall from 23.2% to

20.1% (of banked households). 5 When infrequent users of alternative services are considered underbanked 34.2% of banked households would fall into that category, an

increase of 14%. If those only using money orders and/or check cashing services are excluded from that percentage there is an increase from 7.0% to 10.7%.

30

breakdowns including: Caucasian/White (non-Hispanic), African American/Black (non-Hispanic), Hispanic,

Asian, Native American/Native Alaskan, and other (primarily composed of respondents reporting more

than one race). The SCF includes a slightly different breakdown, White-Caucasian (non-Hispanic), Black or

African American (non-Hispanic), Hispanic, or other. The other category includes Asians, American

Indians/Alaskans, Native Hawaiian, and others. This is not an ideal breakdown of race since the

differences between Asians and Native Americans are significant (Fernanadez 1996).

Employment status is another area with slight discrepancies between the data sets. The FINRA

data set includes information on whether the household is self-employed, full time employed, part time

employed, a homemaker, student, disabled, unemployed or temporarily laid off, or retired. The FDIC has

a similar breakdown but with self-employed households combined with full time or part time employed

based on the number of hours worked. Another difference worth noting is the indicator for whether a

household is a student. The FINRA survey asks the respondent if they are a student as one option for

work status, whereas the FDIC survey asks the respondent their work status, then an additional question

on whether they are enrolled in school full or part time. For the FDIC survey a respondent is considered a

student if they are not in the workforce, retired, or disabled and report being enrolled in school full or

part time. The breakdown for the SCF is similar to the FINRA but defined slightly different. Respondents

of the SCF were given a few more options as work status choices: volunteer and other reason for not

being in the labor force. To keep results as consistent as possible, these responses where combined with

homemakers.

The FDIC also has two important variables that are not able to be determined based on the data

available. The first is if the respondent has a credit card. Controlling for whether or not the respondent

has a credit card is important because it is an indicator of access to credit. If a household has access to

credit, along with a greater need for a transaction account, their access to traditional short term loans

may make them less likely to use alternatives. The second variable that cannot be obtained from the FDIC

31

data set is the drop in income variable. The FINRA and SCF surveys include questions asking whether the

respondent has experienced a drop in income in the past 12 months. If a household has experienced the

drop in income, it may have an effect on their banking participation and use of alternative services. The

need for short term loans may be greater for these respondents due to a lack of funds to meet short term

debt obligations.

While there are some differences across surveys, overall they are very similar and comparisons

can be made with a few notes for the variations. Using the three data sets will create stronger results due

to the individual and combined strengths.

Who are the Unbanked?

To examine who is unbanked, first a mean comparison will be completed using the three data

sets. The comparisons are presented in Tables 1a through 1c, which report the descriptive statistics from

the FINRA, FDIC, and SCF, respectively. All descriptive statistics have been weighted so the results

represent characteristics of the population of the United States.

Previous literature has indicated that the percentage of unbanked is around 10% (Grimes, et al.

2010, Hogarth, et al. 1998). Table 1a presents the FINRA data set, with results indicating a much lower

percentage of unbanked, 5.3%. This percentage is slightly lower than other surveys that will be explored

in this paper. Both the FDIC and the SCF find that 7.5% of households are unbanked. While the FINRA data

set does report a smaller percentage of unbanked household, it is still in line with the other results.

The focus of this paper is to determine the characteristics of unbanked households, and

demographic variables have been found to be different across banking participation levels. Gender is one

variable that will be included in the analysis, but the impact is expected to vary across surveys. This is

because the respondent is chosen in different ways across surveys. The gender variable, in the case

where the head of household or the most financially knowledgeable is interviewed, is more interesting

since it will determine whether males running a household have different banking participation than their

32

female counterparts. As expected, the FINRA data set indicates no difference in banking participation by

gender. However, when the interviewed household member is not chosen at random, the results indicate

that a higher percentage of females fall into the unbanked category.

Age is another demographic variable that was expected to have an effect on whether or not an

individual is unbanked. A set of dummy variables is used to compare the effects of banking participation

for different age groups. The dummies include young adults (18-34 years of age), middle aged (35-54

years of age), and mature adults (55 years of age or older). It was expected that as age increases the level

of banking will increase as well. A mean comparison further confirms this hypothesis. All data sets seem

to follow the trend that as age increases banking participation increases. The strongest results come from

the State survey, the full sample has a nearly 33% breakdown of all age groups. When exploring the age

breakdown within the unbanked category 52% fall into the youngest cohort and only 10% fall into the

oldest.

It was also expected that there will be differences in banking across race and ethnicity. One reason

may be language barriers that are present. If English is not the first language, some individuals may feel

uncomfortable engaging in banking services. It was expected that relative to Caucasian/white non-

Hispanics, African American/Black, Hispanic, Native American/Alaskan, and the “other” category are more

likely to be unbanked, while Asians will be less likely. Race/ethnicity variables vary slightly across surveys.

The FINRA comparison shows strong statistically different results for all race categories with the

exception of the other category. These results indicate that higher percentages of Caucasian/white non-

Hispanics and Asians are banked, while African American/blacks, Hispanics, and other represent a higher

percentage of the unbanked. These trends follow through to the FDIC data. The SCF finds similar trends,

however there is no significant difference in the other category, most likely due to the breakdown issue

stated in the previous section.

33

Marital status is the next demographic variable of interest. It was expected that, relative to

married respondents, single, never married, individuals will be more likely to be unbanked. As the

number of adults in a household increases, it was expected that their finances will become more

complicated and their need for an account will increase. It will also be the case that their income will

grow and they may be more likely to meet minimum balance requirements and avoid various fees that

may be associated with low balances and overdrafts. The expected sign of divorced and widowed

individuals is unknown. Previous results indicate married households are more likely to be banked while

single households are more likely to be unbanked. Initial analysis also appears to indicate a greater

percentage of divorced or separated households fall into the unbanked status, while the reverse is true

for widowed.

It was expected that the presence of children will decrease the likelihood that a household is

banked. The extra expense of having a dependent child will make meeting minimum balances harder and

will be more likely to make bank hours and locations inconvenient. While the data sets have information

on number of children as a continuous variable, it was expected that it is the presence of at least one

child that will have an impact on banking participation. For this reason a dummy variable has been

created to account for at least one dependent child being present in the household. Across all surveys, a

higher percentage of unbanked households have at least one dependent child present. Not only is the

result significant but the magnitude is large. The FDIC survey finds the largest spread with 42% of

unbanked households having a dependent child present, compared to 29% of banked households.

Whether or not an individual is banked will likely be affected by their education level as well. As an

individual’s education increases they are more likely to have a higher paying job, creating a greater need

for an account. As education increases it was also expected that an individual will become more aware of

the additional expenses associated with not having an account, decreasing their use of alternative

services and, in turn, increasing their banking participation. When looking at the mean difference

34

comparison the initial hypothesis is confirmed. Those with a high school degree or less are significantly

more likely to be unbanked, while those with some college or more are significantly more likely to be

banked. These results are consistent across all surveys.

The next variable of interest is employment status. It was expected that work status will have an

effect on banking participation. If an individual is employed, cashing a paycheck is much easier if they

have a bank account. It is predicted that, relative to full time workers, individuals who are part-time

employed, permanently sick or disabled, and unemployed or temporarily laid off will be more likely to be

unbanked. Exploring the results of the three data sets shows consistent results. Across all surveys,

respondents who are employed self-employed, full time or retired represent a higher percentage of

banked households, while those who are employed part time, homemakers, disabled, and unemployed

are unbanked.

As previous literature has indicated, income was expected to be one of the most significant

determinants of being unbanked. Most unbanked respondents describe lack of money as the chief reason

they are unbanked. It was expected that, relative to middle income, lower income households will be

more likely to be unbanked. Across all data sets, the percentage of unbanked households that fall into the

lowest income category is approximately 85%.

If a household has experienced a decrease in income, it is also to be an important indicator of

whether a household is unbanked. It was expected that households who have been confronted with a fall

in income would be less likely to meet minimum balance requirements and therefore, less likely to hold a

transaction account. Both the FINRA and SCF show there is a significant difference between the

percentage of unbanked and banked households who report experiencing a drop in income. The FINRA

survey shows that 56% of unbanked households experience the drop in income, while only 40% of banked

households report the same.

35

The next set of variables that are of interest to this study are indicators for the household’s access

to credit and their level of assets. Access is an important determinant of banking participation since it is

an indicator of involvement with other financial institutions. The first variable of this subset is a dummy

variable for whether or not the respondent owns their residence. It was expected that this variable will be

negatively associated with being unbanked. Homeowners are likely to have, or have had, a mortgage,

therefore a greater need for a transaction account to make payments. The breakdown between levels of

banking indicates a higher percentage of banked households are homeowners. Using the FINRA data, only

17% of unbanked households own a home, compared to 62% of banked households. The mean difference

in homeownership between the unbanked and banked is statistically significant at the 1% level. The FDIC

and SCF report results of similar magnitude and significance.

Credit cards are another example of a respondent’s involvement with financial institutions.

Obtaining a credit card requires credit, which can be built by holding a transaction account. Also, paying

for a credit card is easier if the household has a checking account. While the FINRA data set does offer

information on the number of credit cards the respondent has, it is the act of holding at least once credit

card that was expected to have an impact on banking participation. For this reason, an indicator for

whether the respondent has at least one credit card will be included in the analysis. The SCF analysis will

include a similar dummy variable. As expected, only a small percentage of the unbanked hold at least one

credit card, 20%, compared to 78% of banked households. Results are even stronger using the SCF data,

only 10% of unbanked households report having at least one credit card, while 73% of banked households

do. Both sets of results are significant at the 1% level.

The final set of variables that will be used to determine who the unbanked are is financial

knowledge or financial literacy. While there are mixed results on whether or not financial literacy affects

banking participation, it is important to explore potential differences in the level of literacy between the

banked and unbanked. It was expected that financial literacy will not have a significant effect on banking

36

participation in this study for two reasons. First, the questions asked of respondents were general

financial knowledge and had little to do directly with the choice of having a bank account. Since the

knowledge is not directly associated with bank account holdership, the effect will be smaller than if the

financial questions dealt directly with common misconceptions of bank accounts and alternative financial

services. The second reason that financial literacy is predicted to have little to no effect on account

ownership is that most respondents’ cite the reason they do not have an account as “not enough money”

or “no need or want of an account.” If most respondents did not have an account due to high fees or

other high cost complaints, it may be more likely that financial literacy would have an impact.

The FINRA questions used in this analysis can be found in Appendix 3. Statistics on financial

literacy can be found in Table 2. The average number of questions answered correctly by a respondent

was three, with the difference between the number of questions answered correctly by the unbanked

and banked being significantly different. Unbanked respondents, on average, answered 2.67 questions

correctly, while banked respondents answered 3.20 correct. The number of don’t know/refused

responses are also of interest because it is an indicator of the respondent acknowledging they are not

financially knowledgeable about the specific topic. This should be differentiated from respondents who

answered incorrectly. On average, unbanked households responded don’t know/refused slightly more

often, but the difference is not large in magnitude.

Not only is it important to look at the financial literacy score as an aggregate, but individually as

well. The first two questions inquire about savings and interest rates, with the second adding an inflation

component. These questions are used to determine the numeracy skills of the respondent. Both

questions were answered correctly by 79% and 67% of respondents, respectively, indicating that they

were fairly easy questions. The third question, the bond question, is the most difficult, with only 29% of

respondents answering correctly. It was also the question where the most individuals reported they “did

not know” the answer. This question may differentiate respondents with basic knowledge from those

37

with sophisticated financial knowledge. The fourth question concerns length of a mortgage and the

principal payments and total amount of the loan. This question is also considered relatively easy, and was

answered correctly by 78% of individuals. The final question of the set concerns stock diversification and

risk. This question was fairly difficult in that only 55% of respondents answered the question correctly.

Regression Results

To explore the different levels of banking participation, probit regression will be used. The

dependent variable, unbanked, will be coded 1 if the responded does not have any type of transaction

account and 0 otherwise. The first set of regressions includes controls for demographic and

socioeconomic characteristics, as well as the respondents’ access to credit and assets. This is the basic

model which includes a fairly consistent set of variables included in all three surveys. The results for these

regressions can be seen in Tables 3a through 3c.

The results seem to indicate that females are less likely to be unbanked, relative to males. The

results are significant for all surveys except the FINRA survey. This may be a result of a difference in the

role of the respondent in the household. The FINRA survey asked to speak with the individual in the

household whose birthday was closest, so any significance would be described as differences in reporting;

women respond differently than males to the banking questions. Since the results are not significant, it

appears that women are not more likely to report being unbanked than men. The FDIC focuses on the

head of household and the respondent for the SCF is the most financially knowledgeable, so the fact that

gender does play a significant role in these cases is more noteworthy than the FINRA results. While these

results are significant the results are small in magnitude, with less than a 1% difference.

It was expected that as age increased, finances became more complex and need for an account

increased. The results for this variable are consistent across surveys; the oldest households are

significantly less likely to be unbanked, relative to those in the middle cohort. This result is significant but

38

not as large in magnitude as expected, as those in the oldest cohort are approximately 1% less likely to be

unbanked, relative to those in the middle cohort.

Race was also found to be a significant determinant of banking participation. Consistent across all

data sets, relative to Caucasian/Whites, African American/Black and Hispanic households were more

likely to be unbanked, with the result being significant in most cases. While the results are similar in sign

across the surveys, the magnitude varies slightly. The FDIC analysis indicates that African

Americans/Blacks are 5% more likely to be unbanked and Hispanics are 3%. The FINRA and SCF

percentages are closer to 1%.

The mean difference analysis indicated that a higher percentage of married households were

banked. These results only follow through to the probit analysis, results for all data sets indicate that

single, never married and divorced/separated households are significantly more likely to unbanked. These

magnitudes remain fairly small and hover around 1%.

Including an indicator for dependent children present in the household resulted in a significantly

positive coefficient. Households with children under 18 present are also significantly more likely to be

unbanked, but the result is not significant. This effect may explained by the additional expenses children

bring to a household. With the additional expenses, households may be unable to meet the minimum

requirements to hold a bank account, or simply may not have a need for an account due to lack of

funding. It is also possible that the presence of children makes traditional banking more inconvenient

than the alternatives.

Education was expected to have a strong influence on level of banking participation, it was

predicted that not only do increases in education lead to higher incomes and more complex finances, but

also more knowledge that may lead to increased bank participation. This expectation was confirmed by

all three data sets, which find the levels of education to be significantly associated with banking

participation. Those with less than a high school degree, relative to respondents with a high school

39

degree, were 2% more likely to be unbanked. A respondent with some college education, or greater, is 1

to 2% less likely to be unbanked, relative to a high school graduate.

Based on the mean difference analysis, it was expected that full time workers would be more

likely to be banked. This regression omitted the full time dummy variable, so results are reported relative

to full time workers. Since the variables are coded in this manner, the expected sign is positive, indicating

a greater likelihood of being unbanked for other work statuses. The FINRA and SCF results are the

expected sign, with most positive or near zero. The FDIC survey finds that students are significantly less

likely to be unbanked, though this coefficient is not large in magnitude. The coefficient on students for

the other studies is not significant and is near 0. Respondents who report being homemakers are

significantly more likely to be unbanked, the same is also true of disabled and unemployed households.

A set of income variables are the next controls included in the regression analysis, both income

level and changes in income (when available) were included. It was expected that income would be a

primary motivator of whether a household held a transaction account. This result was confirmed by all

studies, low income households are significantly more likely to be unbanked. Both the FDIC and SCF also

find that those that fall into the highest income bracket are significantly less likely to unbanked, whereas

the result was near 0 and insignificant in the FINRA analysis.

Not only was the level of income expected to have an effect, but the change in income was

included to control for households that felt they no longer needed an account or could no longer afford it

due to the loss of funds. This variable is of particular interest because of the time period in which the

survey was administered, 2009 and 2010. In both surveys that included this variable, the result was in the

expected direction: households who experienced a drop in income were more likely to be unbanked. The

result, however, is not large in magnitude and is only significant in the FINRA analysis. One potential

reason this variable does not have the impact expected is due to its correlation with the unemployed

status of the respondent.

40

The final set of variables included in this regression set is controls for a household’s access to

credit and assets. Homeowners are significantly less likely to be unbanked in both the FINRA and FDIC

analysis. This was expected since homeownership generally requires some interaction with financial

institutions and is an indicator of wealth. A homeowner is 2% less likely to be unbanked; this result is

similar in magnitude and significant across all surveys. It was also expected credit cards would lead to a

less likelihood of being unbanked, since acquiring that form of credit typically requires an account and

creates a greater need for the account. The FINRA and SCF both find that respondents holding at least

one credit card are 4% less likely to be unbanked.

Financial literacy is the final set of variables that will be explored in terms of banking participation.

Financial literacy will be controlled for in several different ways. First, the number of questions answered

correctly will be included. All respondents were given the option to refuse or answer “do not know” to

each question. These responses, along with incorrect responses, were given a value of 0. A correct

response was given a value of 1. The sum was then taken, the highest possible score on this value is 5 and

the lowest is 0. The regression results for this analysis are presented in Column 1 of Table 4.1. The more

questions the respondent answered correctly, the less likely the household was unbanked. While this

result is significant, it is not large in magnitude.

The second way that financial literacy will be analyzed is by including each question individually.

These variables are coded similar to the first regression, but the results are not aggregated, and rather

used as separate indicators. It was expected, since banking is a low level of financial involvement,

relatively easy questions would have the strongest impact on banking participation. All questions, with

the exception of the question on the relationship between bond prices and interest rates, are the

expected signs and most significant. The coefficients are not large in magnitude, but this may be a result

of the correlation between questions.6

6 Correlations between questions range from 0.13 to 0.36.

41

The final set of financial literacy regressions includes each question separately, with results found

in Table 4.2. Indicators for whether the respondent answered the question correctly and whether they

responded “don’t know” are included. Results are presented relative to those who answered the question

incorrectly. The expected sign of answering a given question correctly is negative, meaning those

households are less likely to be unbanked. This result is found across all questions, except the bond price

question which has a coefficient of 0. The results are also significant, answering the question correctly

leads to a nearly 1% decrease in the likelihood the household is unbanked.

Who are the Underbanked?

When exploring banking participation we see several significant differences between households

that are banked and unbanked. However, thinking that these are the only two levels of banking would be

erroneous. There are different levels of participation within the banked category that can be explored.

This portion of the paper will focus on those who have a transaction account and also use alternative

financial services to finance short term loans, referred to as the underbanked. It was expected that the

underbanked are more similar to the unbanked because they use more costly financial services. For this

reason, it was expected many of the results will be similar to the unbanked analysis. In this section of the

paper only households who have a transaction account will be analyzed. The sample size will not match

those who are banked in the previous analysis because the level of participation is unknown for some

banked households. Respondents that don’t know/refused to disclose their level of participation, or for

which the survey was terminated before their status was determined have been dropped.7 For the FINRA

data set, this decreases the sample size from 26,544 to 26,146 households, of which 23% are

underbanked. The FDIC data suggests that 20% of banked households are underbanked. The sample size

using the FDIC sample has decreased to 41,813 households, from 43,514. The SCF will be left out of this

analysis due to the lack of information on the variables that define the underbanked.

7 Households who answered yes to at least one alternative service are considered underbanked, even if they did not know/refused other

services. If the respondent answered they did not use any of the alternatives but didn’t respond or refused one question they were dropped from the analysis.

42

Table 5a and 5b present the descriptive statistics for the FINRA and FDIC full sample of banked and

a breakdown of the underbanked and fully banked. Exploring the demographic variables is the first step in

determining the difference between underbanked and fully banked households. A mean analysis

indicates that there are a slightly larger percentage of females falling into the underbanked category than

the fully banked.

Age is another demographic variable expected to influence banking participation, it is

hypothesized that older respondents are less likely to be underbanked. This was expected as this cohort

may have more complex finances and more credit experience that will allow them to obtain traditional

financial services over the alternatives. Both the FINRA and FDIC data sets show a trend of greater

banking participation as age increases. There are a higher percentage of young and middle age adults

who are underbanked, while the reverse is true for the older cohort.

The mean difference in race variables when comparing the unbanked to the banked was highly

significant. The differences are not as clear for the comparison of the underbanked to the fully banked.

Both the FINRA and FDIC survey indicate that Caucasian/White and Asian respondents represent a

significantly higher percentage of fully banked households. Respondents who are African American/Black,

Hispanic, Native American/Alaskan, and report multiple races/ethnicity compose a higher percentage of

underbanked households. The most drastic result is for African American respondents, which make up

just fewer than 10% of all banked respondents. When looking at the race breakdown by banking

participation, Blacks represent more than 20% of underbanked households and only 7% of fully banked

households.

Both mean analyses also show a significant difference in banking participation based on marital

status. Single, never married, and divorced/separated households represent a significantly greater

percentage of the underbanked. Married households and widowed households therefore make up a

greater percentage of fully banked households.

43

It was expected that households with dependent children are more likely to require alternative

financial services due to the additional expenses children bring to the household. Meeting short term

debt obligations may become more difficult with children present in the household. For this reason, it

was expected that households with children present are more likely to be underbanked. An initial

analysis using the FINRA and FDIC indicates this prediction is true, 53% and 36% of underbanked

households report having at least one dependent child present, respectively.

As with the unbanked, it was expected that underbanked households will be less educated than

fully banked households. As an individual becomes more educated they become more aware of the

additional expenses of alternative financial services, leading to decreased use. The results are as expected

for respondents with a college degree or higher; there are significantly more fully banked compared to

underbanked. The FDIC data indicates that both those with less than a high school and those that hold a

high school degree or equivalent are more likely to be underbanked.

As discussed above, it was expected the reason underbanked households use alternative services

is to meet short term debt obligations. Respondents who are unemployed or disabled may be more likely

to need the additional money to meet these needs. It is also hypothesized that retired households are

less likely to be underbanked because they have more complex financial needs and have established

credit throughout their working career. Both sets of data confirm these results. Unemployed households

make up 4.9% of all banked households in the FDIC survey. When moving to the breakdown of

underbanked and fully banked households, unemployed respondents represent 8.3% of the

underbanked. The reverse is found for retired respondents, the FINRA results show retired respondents

make up 8.3% of the underbanked, but over 20% of the fully banked.

A household’s level of income was expected to be the primary determinant of whether the

household had a transaction account. Unlike the unbanked, it was expected that experiencing a drop in

income will have a larger effect on whether a respondent is underbanked. Both the FINRA and FDIC

44

results show significant differences in the mean percentage of households who are underbanked versus

those fully banked for the lowest and highest income levels in the expected directions. An indicator for

drop in income is included in the FINRA analysis. Results show that 54% of underbanked households have

experienced an unexpected drop in income, while only 35% of their fully banked counterparts have

experienced a similar change in income.

Access to credit and level of assets are also expected to have an impact on whether or not a

household chooses to use alternative financial services in addition to traditional bank accounts. It was

expected that homeowners and those holding credit cards will be less likely to use the alternative

financial services because they are able to obtain traditional forms of credit that are relatively less

expensive. If an individual is a homeowner, they most likely have taken out a mortgage to pay for their

home. This can improve a homeowner’s credit rating, allowing them to obtain other loans, including a

home equity loan. Traditional loans generally allow these individuals to borrow funds at a lower rate. This

is also true of credit cards; obtaining credit cards can increase credit scores and allow users to obtain

other forms of credit. Credit cards can also be seen as substitutes for other short term loan alternatives.

These hypotheses seem to be confirmed by the mean difference analysis. The FDIC data indicates that

53% of underbanked households are homeowners, compared to 77% of fully banked households. These

results are slightly different in magnitude, but confirmed with the FINRA data. The FINRA results also

explore the use of credit cards by the two groups, 61% of underbanked households have at least one

credit card, while 84% of fully banked households do.

The final set of variables that will be used to better understand the underbanked are a set of

financial literacy variables. It was expected that the decision on whether to have a bank account would

not be strongly impacted by the respondent’s financial literacy. Since obtaining an alternative financial

loan or using alternatives to traditional accounts can be very costly decisions and require more

involvement, it is projected that financial knowledge will have a greater effect. The reason most

45

households cite as being unbanked is that they do not have enough money to have an account. It does

not seem, based on this reason, that financial knowledge can significantly influence banking behavior. It is

the lack of funds to meet current debt obligations that was predicted to be the reason a household is

underbanked. If this is a result of mismanagement of money, education can be beneficial.

Table 6 shows the descriptive statistics for the financial literacy variables of all banked households,

along with underbanked and fully banked subsets. Looking at the first indicator of financial literacy, the

number of correct answers given on the set of five questions, a significant difference between the

number answered correctly by the underbanked and the fully banked households is shown. Underbanked

households answer 2.7 questions correctly while fully banked households answer 3.2 correct. While this

difference does not seem large, due to the small number of questions the result is significantly different.

Underbanked respondents were slightly more likely to respond don’t know/refuse a response.

Due to the varying difficulty of questions, an exploration of individual questions to determine the

rates underbanked households answered them correctly is also of interest. Fully banked households

answered all questions correctly significantly more often than their underbanked counterparts. The

inflation question had the largest difference between the percentage of underbanked who answered

correctly compared to fully banked, with a spread of 15.1%. This is particularly concerning since it is a

numeracy question that deals with inflation rates and interest rates. If a respondent does not answer this

question correctly, it may be the case they do not fully understand interest rates. This may be an

indication that financial education would be beneficial in helping the respondent understand what an

interest rate is and the effect it has on income. The second largest spread was found in the stock

diversification and riskiness question. 44% of underbanked households answered the question correctly,

compared to 59% of fully banked households. It may be the case that this question differentiates

between individuals with simple or sophisticated levels of financial involvement. Since an individual with

complex finances would be expected to be fully banked, this may be the effect the question is picking up.

46

The bond price question was the most difficult question, and there is nearly a 7% difference in the

number of underbanked and fully banked households that answered the question correctly. It is

hypothesized that relatively difficult questions will be better at differentiating between the underbanked

and fully banked respondents, compared to relatively easy questions.

Regression Results

Table 7a reports results for the FINRA data set and Table 7b for the FDIC data. The dependent

variable of interest is underbanked, with a dummy variable equal to 1 if the household uses at least one

alternative banking service and 0 otherwise. The first regression includes information on the respondents’

demographic and socioeconomic characteristics that are expected to impact banking participation.

Gender is first of the demographic variables included in the underbanked analysis. The FINRA

results show that females are significantly less likely to be underbanked, but the coefficient is not large in

magnitude. This result follows to the FDIC survey, but the result is not significant.

The expectations of age on banking participation were confirmed. Looking at the FINRA data,

those in the youngest cohort are 5% more likely to be underbanked, relative to those in the middle aged

cohort. As was expected, respondents in the oldest cohort were nearly 8% less likely to be underbanked.

The FDIC results tell a similar story, but the coefficients are not as large in magnitude and only the oldest

cohort effect is significant.

The effect of race/ethnicity is much larger than expected. Both the FINRA and FDIC results report

African Americans/Blacks, relative to Caucasian/Whites, are 12% and 21% more likely to be underbanked,

respectively. These results are significant at the 1% level. The Native American/Alaskan effect is also large

in magnitude, the FDIC results indicate this group is 19% more likely to be underbanked. Hispanics are

also significantly more likely to be underbanked, but the result is not as large in magnitude as the

previous race variables. While these races/ethnicities are more likely to be underbanked, the reverse is

47

true for Asians, who are significantly less likely to be underbanked; results ranging from 10% to 4% are

found.

The effect of marital status is not very strong, only divorced/separated households are

significantly more likely to be underbanked in both surveys. These respondents are approximately 3%

more likely to be underbanked relative to married households.

It was expected that those households where dependent children are present will have more

expenses than households where there are no children under the age of 18. With greater expenses there

may be greater need for short term loans to meet debt obligations. This result is confirmed by the

significance of the indicator for children present in the household in both the FINRA and FDIC analysis.

Relative to those with no dependent children, these households are 10% and 4% more likely to be

underbanked.

Education was also predicted to be a significant determinant of whether or not a household was

underbanked. As education increases, an individual was expected to become more aware of the costs

associated with alternative financial services and in turn, use them less. The results found in both the

FINRA and FDIC analysis confirm this hypothesis. Those who have less than a high school degree are more

likely to be underbanked, relative to those with a high school degree or equivalent. However, the result is

only significant in the FIDC regression. For all levels of education above a high school degree, respondents

are less likely to be underbanked. Having some college education decreases the likelihood by 2% and

respondents with a college degree or higher are 10% less likely to be underbanked.

It was anticipated that since the majority of services that define the underbanked are related to

short term loans, those who are unemployed and temporarily disabled may be more likely to fall into this

category. This prediction is confirmed by the FDIC analysis. Relative to respondents employed full time,

those who are unemployed are 6% more likely to be underbanked while those who are disabled are 5%

more likely. The conflicting result comes from the FINRA survey. Results confirm that disabled

48

respondents are 5% more likely to be underbanked, whereas unemployed households are 2% less likely.

While this result was not expected it may be explained by the inclusion of the drop in income variable.

Households who have experienced an unexpected drop in income are 9% more likely to be underbanked.

Due to the correlation between these variables, the change in income variable may take some of the

explanatory power of unemployment.8

While income was expected to have large effect on whether a household had a transaction

account, it was not expected to be as strong of a determinant in whether the household is underbanked.

Although the prediction was income would not have as large of an effect, the results are still significant

and in the expected direction. Households in the lowest income bracket are 1% and 3% more likely to be

underbanked, while those in the highest bracket are 6% and 8% less likely to be underbanked, in the

FINRA and FDIC analysis, respectively.

The final controls in the first set regression are indicators for assets and credit. Homeownership

and holding at least one credit card are both negatively associated with being underbanked. A

homeowner is nearly 10% less likely to be underbanked with both the FINRA and FDIC data. This is the

expected result due to the fact that homeowners have access to more affordable short term loan options,

such as a home equity loan. Households that have at least one credit card are 12% less likely to be

underbanked. This was the hypothesized result since it is assumed credit cards are substitutes for many

of the alternative loans that define the underbanked.

Tables 8.1 and 8.2 show the impact financial literacy has on the banking status of households with

transaction accounts. The first regression includes financial literacy as the number of questions answered

correctly out of five. This result is significant and in the expected direction: the higher number of

questions answered correctly the less likely the household is underbanked. The inclusion of the financial

knowledge variable leads to some changes in the demographic and socioeconomic controls; slightly

8 The coefficient on unemployed becomes positive if the control for a household experiencing a drop in income is not included in the

regression.

49

decreasing the impact of the race/ethnicity variable. While this result was unexpected, it suggests that

race alone is not the main determinant of whether or not a household uses the alternative services. The

change in income variable is still a significant indicator and remains relatively large in magnitude.

The second regression separates the questions into individual controls to determine the impact of

each question on its own. All coefficients are in the expected direction except the mortgage question.

While this question is in the unexpected direction, the result is not significant. The largest effect comes

from the inflation and stock diversification questions. These questions had the largest spread between

the percentages of underbanked and fully banked answering the question correctly. The results show that

if a respondent answered the savings, inflation, and stock diversification questions correctly they are 2%,

3%, and 3% less likely to be underbanked, respectively.

The final set of regressions includes each question separately. An indicator for whether the

respondent answered the question correctly and don’t know/refused to answer are included, with the

omitted group responding incorrectly. Results show that all signs on the coefficients for answering the

questions correctly are significant and in the expected direction. The question with the largest impact is

the stock diversification question. Answering that question correctly leads to an 8% less likelihood the

respondent is underbanked. Another interesting result from these regressions are the signs and

significance of the coefficients on the don’t know/refused responses. Those who didn’t know or refused

to answer are significantly less likely to be underbanked for all individual questions.

Conclusion

The idea that 7% of households in the United States are unbanked and an additional 18% are

underbanked may be hard to believe for most Americans. As nearly all day to day transactions are easier

when a checking account is held. Households who do not use these accounts or underutilize them may

not be minimizing their expenses or theirinvested time. Understanding who these individuals are is

important to ensure they are making a fully informed decision by remaining at a low level of participation.

50

An analysis across three data sets has shown that unbanked individuals tend to be African

American, Hispanic, and Native American/Alaskan. This subset of individuals may not feel as

comfortable/welcomed in banking institutions or have other reasons for avoiding traditional accounts.

Unemployed and disabled individuals were also more likely to not have a transaction account; it is

expected these households choose to avoid banking services due to lack of income, making it difficult to

meet minimum balances, or they may feel there is less need for an account. As with previous research,

income had a significant effect on banking participation. Households earning less than $35,000 were

significantly more likely to be unbanked. Those who experienced an unexpected drop in income were also

more likely to avoid transaction accounts. While these were the expected results, the coefficients were

not as large in magnitude as previous literature. This may be due to the fact that both the level of income

and change in income were included in the same regression. Previously, the change in income variable

had not been included. The final set of variables, homeownership and credit card possession, lead to

significantly less likelihood the household was unbanked. These results were expected since they are

indicators of wealth and access to credit.

The second section dropped the unbanked from the investigation and included only those

households who reported holding a transaction account. Determining which households were

underbanked, relative to fully banked, was the intended objective. While results were in similar directions

as the unbanked analysis, the degree of significance was slightly different. The race/ethnicity effect is still

strongly present in the underbanked analysis; African American/Blacks, Hispanics, and Native

American/Alaskans are significantly more likely to be underbanked. Those with dependent children are

also more likely to be underbanked, which was the expected result since these households may require

short term loans to meet unanticipated expenses. The education coefficients found in the underbanked

analysis were larger in magnitude compared to the unbanked; additional education, beyond high school,

was found to decrease the likelihood a household was underbanked. It was expected the effect on the

51

underbanked would be stronger since it is a more complicated financial decision to employ alternative

financial services. Using short term loans that have high interest rates and using non-bank services

(generally offered for free with a transaction account), indicate these individuals have a lack of

information and financial knowledge. Improving education on budgeting and services offered by

traditional bank accounts may improve banking participation among the low educated.

Work status did not seem to be a significant determinant of whether a household choose to have

a bank account, but was a good indicator for whether they were underbanked. Those who were

unemployed and disabled were particularly more likely to use alternatives to supplement their bank

account. This was the expected result, as these individuals may have experienced a decrease in income

that led to greater need for funds to meet short term debt obligations. This is further confirmed by the

significance of a change in the level of income, experiencing an unexpected drop in income leads to a 10%

greater likelihood a household is underbanked.

The results suggest, demographic and socioeconomic characteristics are not the sole determinant

of a household’s banking participation level, financial knowledge also has an effect. This result was larger

for the underbanked than the unbanked, but the conclusion was as expected. The primary reason

unbanked households do not have an account is lack of money, whereas the need to meet short term

debt obligations is the primary reason the underbanked use alternative services. The strong underbanked

result indicates these households may lack budgeting skills or have a high cost of budgeting. These

households decide to pay a premium, in the form of high fees and service charges on alternative loans, to

avoid budgeting and money management. It may be the case that money management education could

move these households toward a higher level of participation.

The cross study comparison of results yielded very similar results in terms of significance and

magnitude of coefficients. This was the expected result since the surveys were nationally representative

52

and completed by reputable agencies. The slight discrepancies found across the analyses can be mostly

explained by differences in the wording of questions and information available.

Now that these individuals have been defined, a better understanding of why unbanked and

underbanked households partake in particular levels of banking participation and how they make day to

day transactions is appropriate. First, it must be taken into consideration that these households do not

have a bank account because a bank does not allow them to hold one. Some individuals are not allowed

to hold an account after fraud or extensive fees have been accrued. If this is the case, a different

approach must be taken as compared to a situation where the respondent does not have an account by

choice. Understanding these details can lead to better insight into whether education and access to

information would benefit these individuals. The alternatives households use in place of an account is

also important to understand, since it may be the case they are paying a premium to avoid traditional

services. It may be the case they find these services to be more convenient or welcoming, in which

education and information may not lead to a better outcome. However, if households use alternatives

because they believe banks do not offer these services or do so for a higher price, education will lead

consumers to change their behavior toward increased banking participation.

Much understanding of the unbanked and underbanked is necessary to determine whether these

households are making a decision that is least costly to them. It is important to not assume that holding a

transaction account would make them better off, since they may be willing to pay a premium to avoid

banking institutions. Further insight into the reasons households have a low level of banking, and if they

are fully informed in their decision-making, can help both institution owners and policy makers. If bank

managers understand why these households are avoiding their services, they may be able to increase

their customer base by implementing minor changes. Policy makers, who often target low income

individuals and unemployed households (those who avoid traditional services the most), can amend and

modify programs and processes in their attempts to improve the financial situation of thousands.

53

Appendices

Appendix 1: Economic Understanding Questions: National Financial Services Survey

Each question included a Don’t Know/Not Sure and a Refused option. 1. What is the current national unemployment rate?

a) one percent or less b) between 1 percent and 10 percent c) over 10 percent

2. What is the current annual rate of inflation?

a) one percent or less b) between 1 percent and 10 percent c) over 10 percent

3. Is the main purpose of the Federal Reserve:

a) to set interest rates and monetary policy b) to set tax rates and government spending

4. There is a deficit in the Federal Budget when:

a) government spending is greater than tax revenues b) US imports are greater than US exports c) the total demand for money is greater than the total supply of money

5. The purchasing power of people’s income is MOST affected by the:

a) inflation rate b) trade deficit c) balance of payments

6. In a competitive market, the prices of most products are determined by:

a) The government b) business monopolies c) supply and demand d) the Consumer Price Index

7. Does setting quotas on foreign goods imported into the US increase the number of jobs for American

workers in the next 5 to 10 years? a) Yes b) No

54

Appendix 2: Comparison of Variables across Data Sets Categories Variables FINRA State-by-State FDIC SCF

Dependent Variable:

Unbanked Unbanked

(Do you/Does your household) have a checking account

and (Do you/Does your

household) have a saving account, money market

account, or CDs?

(Do you/Does anyone in your household) currently have a

checking or savings account?

Do you (or anyone in your family living here) have any

checking accounts at any type of institution?

and Do you (or anyone in your

family living here) have any savings or money market

accounts?

Underbanked Variables/

Alternative Loans Check Cashing

Do you or your spouse sometimes go to a check

cashing store to cash checks?

AND Do you or your spouse

sometimes cash checks at a grocery store or

supermarket?

Have you or anyone in your household ever gone to a place other than a bank, a savings

and loan or a credit union to cash a check

that was received from someone else?

NA

Money Order Do you or your spouse

sometimes pay your bills with money orders?

Have you or anyone in your household ever purchased a money

order at a place other than a bank, a savings

and loan or a credit union?

NA

Payroll Card NA

Do you/Does anyone in your household receive payment for wages by having the employer

deposit the salary onto a payroll card instead of

paying via cash or check?

NA

Payday Loans In the past 5 years: Have

you taken out a short term “payday” loan?

Have you or anyone in your household ever used payday loan or

payday advance services?

During the past year, have you (or anyone living here) taken out a “payday loan,”

that is borrowed money that was supposed to be repaid in full out of your

next paycheck?

Pawn Shop In the past 5 years: Have you used a pawn shop?

Have you or anyone in your household ever sold items at a pawn

shop?

NA

Tax

Anticipation Loan

In the past 5 years: Have you gotten an advance on

your tax refund? This is sometimes called a

“refund anticipation loan” or “Rapid Refund”

In the past 5 years, have you or anyone in your household taken out a tax refund anticipation

loan?

NA

Rent to Own In the past 5 years: Have you used a rent-to-own?

NA NA

Auto Title Loan In the past 5 years: Have Have you or anyone in NA

55

you taken out an auto title loan?

your household ever rented or leased

anything from a rent-to-own store because you couldn’t get financing

any other way?

Control Variables:

Gender Male, Female Male, Female Male, Female

Age 18-34, 35-54, 55+ 18-34, 35-54, 55+ 18-34, 35-54, 55+

Race/Ethnicity

Caucasian/White, African American/Black, Hispanic,

Asian, Native American/Alaskan, Other

Caucasian/White, African American/Black, Hispanic, Asian, Native

American/Alaskan, Other

Caucasian/White, African American/Black, Hispanic,

Other

Marital Status

Married, Single (never married),

Divorced/Separated, Widow

Married, Single (never married),

Divorced/Separated, Widow

Married, Single (never married),

Divorced/Separated, Widow

Dependents Presence of Children

under 18 Presence of Children

under 18 Presence of Children under

18

Education

Less than High School Degree, High School

Degree or equivalent, Some College Education,

College Degree, Post College Education

Less than High School Degree, High School

Degree or equivalent, College Degree, Post

College Education

Less than High School Degree, High School Degree

or equivalent, College Degree, Post College

Education

Work Status

Self Employed, Full Time, Part Time, Homemaker,

Student Disabled, Unemployed/Laid-off,

Retired

*Full Time, Part Time, Homemaker, Student

Disabled, Unemployed/Laid-off,

Retired

Self Employed, Full Time, Part Time, Homemaker,

Student Disabled, Unemployed/Laid-off,

Retired

Income Level Less than $35,000,

Between $35,000 and $75,000, Over $75,000

Less than $35,000, Between $35,000 and $75,000, Over $75,000

Less than $35,000, Between $35,000 and $75,000, Over

$75,000

Change in

Income

In the past 12 months (have you/has your

household) experienced a large drop in income

which you did not expect?

NA

Is this income unusually high or low compared to

what you would expect in a "normal" year?

Homeowner Do you (or your

spouse/partner) currently own your home?

Are your living quarters (a) owned or being

bought by a household member?

Do you (and your family living here) own this (house

and lot/apartment/ranch/farm)?

Credit Card

How many credit cards do you have? Please include

store and gas station credit cards but NOT

debit cards.

NA

Do you (or anyone in your family living here) have any

credit cards or charge cards?

Financial Literacy

See Appendix 3 NA NA

*Unlike the other surveys, the FDIC survey asks whether the respondent is a student in a question separate from work force participation. For this survey students are defined as those respondents who report not being in the work force, but enrolled in school part or full time. Individuals who are in the work force or retired and report being a student are not considered a student, but are reported as their other work status.

56

Appendix 3: Economic Understanding Questions: FINRA

Each question included a Don’t Know/Not Sure and a Refused option.

1. Suppose you had $100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow?

a) More than $102 b) Exactly $102 c) Less than $102

2. Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After 1 year, how much would you be able to buy with the money in this account?

a) More than today b) Exactly the same c) Less than today

3. If interest rates rise, what will typically happen to bond prices? a) They will rise b) They will fall c) They will stay the same d) There is no relationship between bond prices and the interest rate

4. A 15-year mortgage typically requires higher monthly payments than a 30-year mortgage, but the total interest paid over the life of the loan will be less.

a) True b) False

5. Buying a single company’s stock usually provides a safer return than a stock mutual fund. a) True b) False

57

References

Abreu, Margarida, and Victor Mendes. 2010. "Financial Literacy and Portfolio Diversification."

Quantitative Finance 10, no. 5: 515-528.

Amuedo-Dorantes, Catalina, and Cynthia Bansak. 2006. "Money Transfers among Banked and Unbanked

Mexican Immigrants." Southern Economic Journal 73, no. 2: 374-401.

Anderson, Michael and Sanjiv Jaggia. 2012. “Return, Purchase, or Skip? Outcome, Duration, and

Consumer Behavior in the Rent-to-Own Market.” Empirical Economics 43 no. 1: 313-334.

Anderson, Michael and Sanjiv Jaggia. 2009. “Rent-to-Own Agreements: Customer Characteristics and Contract Outcome.” Journal of Economics and Business 61 no. 1: 51-69.

Anguelov, Christoslav E., Marianne A. Higert, and Jeanne M. Hogarth. 2004. “U.S. Consumers and

Electronic Banking, 1995-2003.” Federal Reserve Bulletin.no. 1: 1-18.

Applied Research and Consulting LLC. Financial Capability in the United States: Initial Report of Research

Findings from the 2009 National Survey. New York City: FINRA, 2009.

Avery, Robert. 2011. "Payday Loans versus Pawnshops: The Effects of Loan Fee Limits on Household Use."

Working Paper.

Behrman, Jere R., Olivia S. Mitchell, Cindy K. Soo, and David Bravo. 2012. “How Financial Literacy Affects

Household Wealth Accumulation.” Working Paper.

Bell, Claes. Bankrate, "Report: Prepaid Cards Risky." Last modified September 7, 2012. Accessed October

1, 2012. http://www.bankrate.com/financing/banking/report-prepaid-cards-risky/.

Bell, Claes. Bankrate, "Check Cashing: still not a good deal." Last modified November 18, 2011. Accessed

October 30, 2012. http://www.bankrate.com/financing/banking/check-cashing-still-not-a-good-deal/.

Bernheim, B. Douglas, and Daniel M. Garrett. 2003. "The Effects of Financial Education in the Workplace:

Evidence from a Survey of Households." Journal Of Public Economics 87, no. 7-8: 1487-1519.

Beverly, Sondra G., Jennifer Tescher, and Jennifer L. Romich. 2004. "Linking Tax Refunds and Low-Cost

Bank Accounts: Early Lessons for Program Design and Evaluation." Journal Of Consumer Affairs 38, no. 2:

332-341.

Carter, Susan Payne. 2012. “Payday Loan and Pawnshop Usage: The Impact of Allowing Payday Loan

Rollovers.” Working Paper.

58

Carswell, Andrew T. 2009. "Does Housing Counseling Change Consumer Financial Behaviors? Evidence

from Philadelphia." Journal Of Family And Economic Issues 30, no. 4: 339-356

Caskey, John P. 1994. Fringe banking: Check-cashing outlets, pawnshops, and the poor. n.p.: New York:,

1994.

Chatterjee, Swarn, Joseph Goetz, and Lance Palmer. 2009. "An Examination of Short-Term Borrowing in

the United States." Global Journal Of Business Research 3, no. 2: 1-8.

Cole, Shawn, and Guri Kartini Shastry. 2008. “If You Are so Smart, Why Aren’t You Rich?” Presented at the

2009 Federal Reserve System Community Affairs Research Conference: Washington DC.

Collard, Sharon and David Hayes. 2010. “Pawn Broking Customers in 2010: A Survey.” A Report to the

National Pawn Brokers Association.

Consumer Credit Counseling Service of San Francisco (CCCS), "High Cost Financial Services. Chapter 3:

Fringe Financial Services." Last modified 2010. Accessed October 30, 2012.

https://www.cccssf.org/credittrack/highcost/ch3.html.

Consumer Federation of America & U.S. Public Interest Research Group. 2001. “Rent-a-Bank Payday

Lending: How Banks Help Payday Lenders Evade State Consumer Protections. The2001 Payday Lender

Survey and Report.

Department of the U.S. Treasury, "ETA and Direct Deposit." Last modified 2012. Accessed October 15,

2012. https://www.eta-find.gov/Index.cfm.

Edmiston, Kelly D. 2011. "Could Restrictions on Payday Lending Hurt Consumers?." Economic Review

(01612387) 96, no. 1: 63-93.

Elliehausen, Gregory. 2009. “An Analysis of Consumers Use of Payday Loans.” Federal Reserve System.

Elliehausen, Gregory. 2005. “Consumer Use of Tax Refund Anticipation Loans.” Credit Research Center.

FDIC. 2009. “FDIC Nation Survey of Unbanked and Underbanked Households.” Federal Deposit Insurance

Corporation.

Fernandez-Lopez, Sara, Luis Otero, Milagros Vivel, and David Rodeiro. 2010. "What Are the Driving Forces

of Individuals' Retirement Savings?." Finance A Uver/Czech Journal Of Economics And Finance 60, no. 3:

226-251.

FINRA Investor Education Foundation. 2012. National Financial Capability Study.

http://www.finrafoundation.org//programs/capability/index.htm (accessed January 2, 2012).

59

FONSECA, RAQUEL, KATHLEEN J. MULLEN, GEMA ZAMARRO, and JULIE ZISSIMOPOULOS. 2012. "What

Explains the Gender Gap in Financial Literacy? The Role of Household Decision Making." Journal Of

Consumer Affairs 46, no. 1: 90-106.

Fox, Jonathan, Suzanne Bartholomae, and Lee Jinkook. 2005. "Building the Case for Financial Education."

Journal Of Consumer Affairs 39, no. 1: 195-214.

Fox, Jean Ann and Patrick Woodall. 2006. “Cashed Out: Consumers Pay Steep premium to “Bank” at

Check Cashing Outlets.” Consumer Federation of America.

GRIMES, PAUL W., KEVIN E. ROGERS, and REBECCA CAMPBELL SMITH. 2010. "High School Economic

Education and Access to Financial Services." Journal Of Consumer Affairs 44, no. 2: 317-335.

Hastings, Justine S. and Olivia S. Mitchell. 2011. “How Financial Literacy and Impatience Shape Retirement

Wealth and Investment Behaviors.” Working Paper.

Hathaway, Ian, and Sameer Khatiwada. 2008. "Do Financial Education Programs Work?" Federal Reserve

Bank of Cleveland Working Paper.

Hill, Ronald Paul, David L. Ramp, and Linda Silver. 1998. "The Rent-to-Own Industry and Pricing Disclosure

Tactics." Journal Of Public Policy & Marketing 17, no. 1: 3-10.

Hogarth, Jeanne M., and Kevin H. O'Donnell. 1998. "Being Accountable: A Descriptive Study of Unbanked

Households in the U.S." Partners In Community and Economic Development: Federal Reserve Bank of

Atlanta.

Hogarth, Jeanne M., and Christoslav E. Anguelov. “Are Families Who Use E-Banking Better Financial

Managers?” prepared for the Association for Financial Counseling and Planning Education.

Hogarth, Jeanne M., Christoslav E. Anguelov, and Jinkook Lee. 2005. "Who Has a Bank Account? Exploring

Changes over Time, 1989-2001." Journal Of Family And Economic Issues 26, no. 1: 7-30.

Huston, Sandra J. 2010. Measuring financial literacy. Journal of Consumer Affairs 44(2): 296–316.

Jump$tart Coalition for Personal Financial Literacy. 2007. “National Standards in K-12 Personal Finance

Education: With Benchmarks, Knowledge Statements, and Glossary.”

Kennickell, Arthur B. 1998. “Multiple Imputations in the Survey of Consumer Finances.” Prepared for the

Joint Statistical Meetings. Dallas, TX. August 1998.

60

Lawrence, Edward C., and Gregory Elliehausen. 2008. "A Comparative Analysis of Payday Loan

Customers." Contemporary Economic Policy 26, no. 2: 299-316.

Lindamood, Suzanne, Sherman D. Hanna, and Lan Bi. 2007. “Using the Survey of Consumer Finances:

Some Methodological Considerations and Issues.” The Journal of Consumer Affairs. 41, no 2: 195-214.

Lusardi, Annamaria, and Olivia Mitchell. 2007. “Baby Boomer Retirement Security: The Roles of Planning,

Financial Literacy, and Housing Wealth. Journal of Monetary Economics. 54, no 1: 205-224.

Lusaridi, Annamaria, and Olivia Mitchell. 2008. “How Ordinary Consumers Make Complex Economic

Decisions: Financial Literacy and Retirement Readiness.” Working Paper.

McKernan, Signe-Mary, James M. Lacko and Manoj Hastak. 2003. “Empirical Evidence on the

Determinants of Rent-to-Own Use and Purchase Behavior.” Economic Development Quarterly. 17 no. 1:

33-52.

Melzer, Brian. 2011. “The Real Costs of Credit Access: Evidence from the Payday Lending Market. The

Quarterly Journal of Economics. 126, no 1: 517-555.

Montalto, Caterine P. and Jaimie Sung. 1996. Multiple Imputations in the 1992 Survey of Consumer

Finances. Financial Counseling and Planning, 7: 133-146.

Peterson, Christopher. 2007. "Usury Law, Payday Loans, and Statutory Sleight of Hand: An Empirical

Anaysis of American Credit Pricing Limits." Working Paper.

Paulson, Anna, and Sherrie L. W. Rhine. 2008. "The Financial Assimilation of an Immigrant Group: Evidence on the Use of Checking and Savings Accounts and Currency Exchanges." Journal Of Family And Economic Issues 29, no. 2: 264-278. Remund, David L. 2010. Financial literacy explicated: The case for a clearer definition in an increasingly complex economy. Journal of Consumer Affairs 44(2): 276–295.

Rhine, Sherrie L. W., and William H. Greene. 2006a. "The Determinants of Being Unbanked for U.S.

Immigrants." Journal Of Consumer Affairs 40, no. 1: 21-40.

Rhine, Sherrie, William Greene, and Maude Toussaint-Comeau. 2006b. “The Importance of Check-Cashing

Businesses to the Unbanked: Racial/Ethnic Differences.” Review of Economics and Statistics. 88, no 1:

146-157.

Robb, Cliff. 2011. “Financial Knowledge and Credit Card Behavior of College Students.” Journal of Family

and Economic Issues. 32. No6: 690-698.

61

Schuh, Scott and Joanna Stavins. 2011. “How Consumers Pay: Adoption and Use of Payments.” Working

paper.

Skiba, Paige Marta, and Jeremy Tobacman. 2007. “Measuring the Individual-Level Effects of Access to

Credit: Evidence from Payday Loans.” Working Paper.

SPADER, JONATHAN, JANNEKE RATCLIFFE, JORGE MONTOYA, and PETER SKILLERN. 2009. "The Bold and

the Bankable: How the Nuestro Barrio Telenovela Reaches Latino Immigrants with Financial Education."

Journal Of Consumer Affairs 43, no. 1: 56-79.

Stegman, Michael A., and Robert Faris. 2003. "Payday Lending: A Business Model That Encourages

Chronic Borrowing." Economic Development Quarterly 17, no. 1: 8-32.

Stoesz, David. 2012. “Payday Loans and the Secondary Financial Market. Working Paper.

Van Rooij, Maarten C.J., Annamaria Lusardi, and Rob J.M. Alesssie. 2011a. “Financial Literacy and

Retirement Planning in the Netherlands.” Journal of Economic Psychology. 32, no 4:593-608.

Van Rooij, Maarten C.J., Annamaria Lusardi, and Rob J.M. Alesssie. 2011b. “Financial Literacy and Stock

Market Participation.” Journal of Financial Economics. 101, no 2: 449-472.

Washington, Ebonya. 2006. "The Impact of Banking and Fringe Banking Regulation on the Number of

Unbanked Americans." Journal Of Human Resources 41, no. 1: 106-137.

Wickramasinghe, Vathsala, and Anurudh Gurugamage. 2012. “Effects of Social Demographic Attributes, Knowledge About Credit Cards and Perceived Lifestyle Outcomes on Credit Card Usage.” International Journal of Consumer Studies, 36 no 1: 80-89. Willis, Lauren E. 2011. "The Financial Education Fallacy." American Economic Review 101, no. 3: 429-434.

Worthington, Andrew C. 2006. "Predicting Financial Literacy in Australia." Financial Services Review 15,

no. 1: 59-79.

62

Tables

Obs. Mean S.D. Obs. Mean S.D. Obs. Mean S.D.

Unbanked 27773 0.053 0.223 - - - - - -

Banked 27773 0.947 0.223 - - - - - -

Gender Female 28146 0.513 0.500 1229 0.514 0.500 26544 0.512 0.500

Age Cohort 18-34*** 28146 0.306 0.461 1229 0.521 0.500 26544 0.293 0.455

35-54 28146 0.379 0.485 1229 0.377 0.485 26544 0.379 0.485

55+*** 28146 0.315 0.465 1229 0.101 0.302 26544 0.327 0.469

Race White, Non-Hispanic*** 28146 0.685 0.464 1229 0.519 0.500 26544 0.697 0.460

Black, Non-Hispanic*** 28146 0.115 0.319 1229 0.227 0.419 26544 0.107 0.309

Hispanic*** 28146 0.134 0.341 1229 0.212 0.409 26544 0.129 0.335

Asian*** 28146 0.046 0.210 1229 0.019 0.138 26544 0.047 0.212

Native American/Alaskan** 28146 0.016 0.127 1229 0.024 0.152 26544 0.016 0.126

More than One Race 28146 0.008 0.092 1229 0.008 0.088 26544 0.009 0.092

Marital Status Married*** 28146 0.534 0.499 1229 0.235 0.424 26544 0.551 0.497

Single*** 28146 0.282 0.450 1229 0.543 0.498 26544 0.267 0.443

Divorced*** 28146 0.140 0.347 1229 0.200 0.400 26544 0.136 0.343

Widow*** 28146 0.044 0.205 1229 0.022 0.146 26544 0.045 0.208

Dependents Dependent Children*** 28146 0.384 0.486 1229 0.425 0.495 26544 0.382 0.486

Education Less than High School*** 28146 0.035 0.183 1229 0.175 0.380 26544 0.027 0.161

High School Degree*** 28146 0.293 0.455 1229 0.462 0.499 26544 0.282 0.450

Some College*** 28146 0.419 0.493 1229 0.305 0.461 26544 0.427 0.495

College Degree*** 28146 0.159 0.365 1229 0.048 0.215 26544 0.166 0.372

Post College Education*** 28146 0.094 0.292 1229 0.009 0.096 26544 0.099 0.298

Work Status Self Employed** 28146 0.081 0.272 1229 0.069 0.253 26544 0.081 0.273

Full Time Employed*** 28146 0.361 0.480 1229 0.172 0.377 26544 0.373 0.484

Part Time 28146 0.098 0.297 1229 0.105 0.307 26544 0.097 0.296

Homemaker*** 28146 0.089 0.285 1229 0.116 0.321 26544 0.087 0.282

Student*** 28146 0.058 0.234 1229 0.076 0.266 26544 0.057 0.232

Disabled*** 28146 0.042 0.201 1229 0.074 0.261 26544 0.040 0.197

Unemployed*** 28146 0.098 0.297 1229 0.339 0.474 26544 0.084 0.278

Retired*** 28146 0.172 0.378 1229 0.049 0.216 26544 0.179 0.384

Income Less than $35K*** 28146 0.407 0.491 1229 0.843 0.364 26544 0.380 0.485

$35K to $75K*** 28146 0.349 0.477 1229 0.123 0.328 26544 0.363 0.481

$75K or more*** 28146 0.244 0.430 1229 0.035 0.184 26544 0.257 0.437

Unexpected Drop in Income*** 27585 0.406 0.491 1180 0.563 0.496 26134 0.397 0.489

Access to Credit and Assets Homeowner*** 27808 0.591 0.492 1205 0.166 0.372 26330 0.615 0.486

Credit Card*** 27369 0.748 0.434 1202 0.204 0.403 25948 0.781 0.413

* significantly different at the 10% level

** 5% level

*** 1% level

Demographics

Bank Participation

Full Sample

Table 1a: Banked versus Unbanked

FINRA State By State Financial Capability Study

Descriptive Statistics

Unbanked Banked

63

Obs. Mean S.D. Obs. Mean S.D. Obs. Mean S.D.

Unbanked 46547 0.075 0.263 - - - - - -

Banked 46547 0.925 0.263 - - - - - -

Gender Female*** 46547 0.489 0.500 3033 0.562 0.496 43514 0.483 0.500

Age Cohort 18-34*** 46547 0.221 0.415 3033 0.372 0.483 43514 0.209 0.407

35-54*** 46547 0.397 0.489 3033 0.423 0.494 43514 0.395 0.489

55+*** 46547 0.382 0.486 3033 0.204 0.403 43514 0.396 0.489

Race Caucasian*** 46547 0.717 0.450 3033 0.320 0.467 43514 0.749 0.433

Black*** 46547 0.116 0.320 3033 0.333 0.471 43514 0.098 0.298

Hispanic*** 46547 0.112 0.315 3033 0.295 0.456 43514 0.097 0.296

Asian*** 46547 0.038 0.191 3033 0.020 0.139 43514 0.039 0.194

Native American/Alaskan*** 46547 0.006 0.077 3033 0.018 0.134 43514 0.005 0.070

Other* 46547 0.011 0.105 3033 0.014 0.117 43514 0.011 0.104

Marital Status Married*** 46547 0.521 0.500 3033 0.278 0.448 43514 0.540 0.498

Single*** 46547 0.205 0.404 3033 0.389 0.488 43514 0.191 0.393

Divorced*** 46547 0.175 0.380 3033 0.259 0.438 43514 0.168 0.374

Widow*** 46547 0.099 0.299 3033 0.074 0.261 43514 0.101 0.301

Dependents Dependent Children*** 46547 0.300 0.458 3033 0.416 0.493 43514 0.290 0.454

Education Less than High School*** 46547 0.125 0.331 3033 0.403 0.491 43514 0.102 0.303

High School Degree*** 46547 0.293 0.455 3033 0.369 0.483 43514 0.287 0.452

College Degree*** 46547 0.195 0.396 3033 0.038 0.192 43514 0.207 0.405

Post College Education*** 46547 0.107 0.309 3033 0.009 0.094 43514 0.115 0.319

Work Status Full Time Employed*** 46350 0.537 0.499 3032 0.354 0.478 43318 0.552 0.497

Part Time* 46350 0.079 0.270 3032 0.086 0.280 43318 0.079 0.269

Homemaker*** 46350 0.071 0.257 3032 0.174 0.379 43318 0.063 0.243

Student 46350 0.006 0.079 3032 0.006 0.074 43318 0.006 0.079

Disabled*** 46350 0.054 0.227 3032 0.165 0.371 43318 0.045 0.208

Unemployed*** 46350 0.055 0.229 3032 0.139 0.346 43318 0.049 0.215

Retired*** 46350 0.197 0.398 3032 0.076 0.266 43318 0.206 0.405

Income Less than $35K*** 39907 0.391 0.488 2594 0.886 0.317 37313 0.351 0.477

$35K to $75K*** 39907 0.334 0.472 2594 0.103 0.304 37313 0.352 0.478

$75K or more*** 39907 0.275 0.447 2594 0.011 0.104 37313 0.297 0.457

Access to Credit and Assets Homeowner*** 46547 0.681 0.466 3033 0.238 0.426 43514 0.717 0.450

* significantly different at the 10% level

** 5% level

*** 1% level

Demographics

Bank Participation

Full Sample

Table 1b: Banked versus Unbanked

FDIC Survey of Unbanked and Underbanked Households

Descriptive Statistics

Unbanked Banked

64

Obs. Mean S.D. Obs. Mean S.D. Obs. Mean S.D.

Unbanked 6482 0.075 0.263 - - - - - -

Banked 6482 0.925 0.263 - - - - - -

Gender Female*** 6482 0.271 0.445 461 0.382 0.486 6021 0.262 0.440

Age Cohort 18-34*** 6482 0.210 0.407 461 0.308 0.462 6021 0.202 0.401

35-54** 6482 0.393 0.488 461 0.438 0.496 6021 0.389 0.488

55+*** 6482 0.397 0.489 461 0.254 0.435 6021 0.409 0.492

Race Caucasian*** 6482 0.708 0.455 461 0.366 0.482 6021 0.736 0.441

Black*** 6482 0.138 0.345 461 0.352 0.478 6021 0.121 0.326

Hispanic*** 6482 0.108 0.310 461 0.240 0.427 6021 0.097 0.296

Other 6482 0.046 0.210 461 0.041 0.198 6021 0.047 0.211

Marital Status Married*** 6482 0.505 0.500 461 0.243 0.429 6021 0.526 0.499

Single*** 6482 0.210 0.408 461 0.409 0.492 6021 0.194 0.396

Divorced*** 6482 0.193 0.394 461 0.275 0.447 6021 0.186 0.389

Widow* 6482 0.092 0.289 461 0.074 0.262 6021 0.093 0.291

Dependent Children** 6482 0.435 0.496 461 0.480 0.500 6021 0.432 0.495

Education Less than High School*** 6482 0.120 0.325 461 0.360 0.480 6021 0.100 0.300

High School Degree*** 6482 0.276 0.447 461 0.338 0.473 6021 0.271 0.445

College Degree*** 6482 0.242 0.428 461 0.024 0.154 6021 0.260 0.438

Post College Education*** 6482 0.127 0.333 461 0.009 0.096 6021 0.136 0.343

Work Status Self Employed*** 6482 0.114 0.318 461 0.079 0.270 6021 0.117 0.321

Full time Employed*** 6482 0.490 0.500 461 0.329 0.470 6021 0.504 0.500

Part Time Employed*** 6482 0.047 0.212 461 0.096 0.295 6021 0.043 0.203

Homemaker** 6482 0.015 0.122 461 0.031 0.173 6021 0.014 0.117

Student 6482 0.016 0.124 461 0.013 0.113 6021 0.016 0.125

Disabled*** 6482 0.070 0.255 461 0.205 0.404 6021 0.059 0.236

Unemployed*** 6482 0.069 0.254 461 0.191 0.393 6021 0.060 0.237

Retired*** 6482 0.193 0.395 461 0.083 0.277 6021 0.202 0.402

Income Less than $35K*** 6482 0.398 0.490 461 0.866 0.340 6021 0.360 0.480

$35K to $75K*** 6482 0.325 0.469 461 0.124 0.329 6021 0.342 0.474

$75K or more*** 6482 0.276 0.447 461 0.010 0.099 6021 0.298 0.457

Drop in Income*** 6482 0.253 0.435 461 0.372 0.483 6021 0.244 0.429

Access to Credit and Assets Homeowner*** 6482 0.601 0.490 461 0.165 0.371 6021 0.636 0.481

Credit Card*** 6482 0.680 0.467 461 0.095 0.293 6021 0.727 0.446

* significantly different at the 10% level** 5% level

*** 1% level

Demographics

Table 1c: Banked versus Unbanked

Survey of Consumer Finances

Bank Participation

Descriptive Statistics

BankedUnbankedFull Sample

65

Obs. Mean S.D. Obs. Mean S.D. Obs. Mean S.D.

Unbanked 27773 0.053 0.223 - - - - - -

Banked 27773 0.947 0.223 - - - - - -

# of Questions Correct*** 28146 2.989 1.443 1229 2.026 1.445 26544 3.062 1.412

Don't Know/Refused*** 28146 1.280 1.420 1229 1.982 1.669 26544 1.221 1.376

Savings Question*** 28146 0.777 0.416 1229 0.619 0.486 26544 0.791 0.407

Inflation Question*** 28146 0.645 0.478 1229 0.408 0.492 26544 0.662 0.473

Bond Price Question*** 28146 0.276 0.447 1229 0.195 0.396 26544 0.283 0.450

Mortgage Question*** 28146 0.756 0.429 1229 0.498 0.500 26544 0.775 0.417

Stock Diversification Q.*** 28146 0.534 0.499 1229 0.306 0.461 26544 0.550 0.497

* difference at the 10% level

** the 5% level

*** the 1% level

Table 2: Banked versus Unbanked subset of Financial Literacy Variables

Actual Knowledge

FINRA State By State Financial Capability Study

Bank Participation

Descriptive Statistics

Full Sample Unbanked Banked

66

Dependent Variable

Coefficient P-Value *

Gender (Male) Female -0.001 0.647

Age Cohort (35-55 years) 18-34 0.001 0.683

55+ -0.007 0.002 *

Race (White Non Hispanic) Black 0.008 0.000 *

Hispanic 0.004 0.140

Asian -0.005 0.246

Native American/Alaskan -0.003 0.454

Multiple Races -0.001 0.818

Marital Status (Married) Single 0.010 0.000 *

Divorced 0.011 0.000 *

Widow -0.002 0.625

Presence of Children Dependent Children 0.003 0.047 *

Education (High School Degree) Less than High School 0.025 0.000 *

Some College -0.009 0.000 *

College Degree -0.010 0.000 *

Post College Education -0.011 0.000 *

Work Status (Full Time) Self Employed 0.011 0.003 *

Part Time 0.004 0.186

Homemaker 0.015 0.000 *

Student 0.001 0.785

Disabled 0.009 0.019 *

Unemployed 0.030 0.000 *

Retired 0.000 0.892

Income ($35K to $75K) Less than $35K 0.015 0.000 *

$75K or more 0.000 0.928

Change in income Unexpected Drop in Income 0.003 0.089 *

Homeowner -0.016 0.000 *

Credit Card -0.046 0.000 *

Observations 26585

Pseudo R2 0.305

Observed P 0.051

Predicted P 0.013

Access to Credit and Assets

Table 3a: Unbanked Households Relative to All Banked

Demographics

Variables

Probit Regression - Marginal Effects Reported

FINRA State By State Financial Capability Study

Unbanked

67

Dependent Variable

Coefficient P-Value *

Gender (Male) Female -0.004 0.001 *

Age Cohort (35-55 years) 18-34 0.001 0.528

55+ -0.011 0.000 *

Race (White Non Hispanic) Black 0.050 0.000 *

Hispanic 0.031 0.000 *

Asian 0.004 0.382

Native American/Alaskan 0.077 0.000 *

Other 0.008 0.133

Marital Status (Married) Single 0.013 0.000 *

Divorced 0.013 0.000 *

Widow 0.006 0.034 *

Presence of Children Dependent Children 0.008 0.000 *

Education (High School Degree) Less than High School 0.020 0.000 *

Some College -0.012 0.000 *

College Degree -0.017 0.000 *

Post College Education -0.016 0.000 *

Work Status (Full Time) Part Time 0.004 0.042 *

Homemaker 0.029 0.000 *

Student -0.012 0.001 *

Disabled 0.032 0.000 *

Unemployed 0.024 0.000 *

Retired -0.004 0.055 *

Income ($35K to $75K) Less than $35K 0.033 0.000 *

$75K or more -0.015 0.000 *

Access to Credit and Assets Homeowner -0.025 0.000 *

Observations 39731

Pseudo R2 0.355

Observed P 0.074953

Predicted P 0.015675

Table 3b: Unbanked Households Relative to All Banked

Demographics

Variables

Probit Regression - Marginal Effects Reported

FDIC Survey of Unbanked and Underbanked Households

Unbanked

68

Dependent Variable

Coefficient P-Value *

Gender (Male) Female -0.005 0.017 *

Age Cohort (35-55 years) 18-34 -0.002 0.222

55+ -0.004 0.075 *

Race (White Non Hispanic) Black 0.022 0.002 *

Hispanic 0.015 0.013 *

Other 0.014 0.144

Marital Status (Married) Single 0.007 0.077 *

Divorced 0.005 0.134

Widow 0.001 0.734

Presence of Children Dependent Children 0.002 0.296

Education (High School Degree) Less than High School 0.013 0.009 *

College Degree -0.008 0.008 *

Post College Education -0.006 0.075 *

Work Status (Full Time) Self Employed 0.005 0.229

Part Time 0.007 0.181

Homemaker 0.016 0.265

Student -0.005 0.175

Disabled 0.016 0.021 *

Unemployed 0.016 0.012 *

Retired 0.000 0.914

Income ($35K to $75K) Less than $35K 0.014 0.004 *

$75K or more -0.007 0.072 *

Change in income Unexpected Drop in Income 0.003 0.183

Homeowner -0.013 0.001 *

Credit Card -0.041 0.000 *

Observations 6482

Pseudo R2 0.426

Observed P 0.075

Predicted P 0.009

Access to Credit and Assets

Table 3c: Unbanked Households Relative to All Banked

Federal Reserve Survey of Consumer Finances

Probit Regression - Marginal Effects Reported

Unbanked

Variables

Demographics

69

Dependent Variable

Coefficient P-Value * Coefficient P-Value *

# of Questions Correct -0.003 0.000 * - -

Savings Question - - -0.004 0.026 *

Inflation Question - - -0.003 0.088 *

Bond Price Question - - 0.002 0.356

Mortgage Question - - -0.006 0.000 *

Stock Diversification Q. - - -0.001 0.434

Gender (Male) Female -0.002 0.174 -0.002 0.200

Age Cohort (35-55 years) 18-34 0.000 0.951 0.000 0.856

55+ -0.007 0.003 * -0.007 0.003 *

Race (White Non Hispanic) Black 0.007 0.002 * 0.006 0.003 *

Hispanic 0.003 0.198 0.003 0.250

Asian -0.006 0.196 -0.006 0.166

Native American/Alaskan -0.002 0.542 -0.002 0.564

Multiple Races -0.001 0.781 -0.001 0.732

Marital Status (Married) Single 0.009 0.000 * 0.009 0.000 *

Divorced 0.011 0.000 * 0.011 0.000 *

Widow -0.003 0.517 -0.003 0.503

Presence of Children Dependent Children 0.003 0.058 * 0.003 0.065 *

Education (High School Degree) Less than High School 0.022 0.000 * 0.022 0.000 *

Some College -0.007 0.000 * -0.007 0.000 *

College Degree -0.009 0.000 * -0.009 0.000 *

Post College Education -0.010 0.001 * -0.010 0.001 *

Work Status (Full Time) Self Employed 0.011 0.002 * 0.011 0.002 *

Part Time 0.004 0.177 0.004 0.178

Homemaker 0.013 0.000 * 0.013 0.000 *

Student 0.001 0.718 0.001 0.815

Disabled 0.008 0.033 * 0.008 0.032 *

Unemployed 0.028 0.000 * 0.028 0.000 *

Retired 0.000 0.941 0.000 0.937

Income ($35K to $75K) Less than $35K 0.014 0.000 * 0.014 0.000 *

$75K or more 0.000 0.931 0.000 0.939

Change in income Unexpected Drop in Income 0.003 0.056 * 0.003 0.051 *

Homeowner -0.015 0.000 * -0.014 0.000 *

Credit Card -0.044 0.000 * -0.044 0.000 *

Observations 26585 26585

Pseudo R2 0.309 0.310

Observed P 0.051 0.051

Predicted P 0.013 0.013

Access to Credit and Assets

Table 4.1: Unbanked Households Relative to All Banked - Controlling for Financial Literacy

Probit Regression - Marginal Effects Reported

Unbanked

Demographics

Variables

Knowledge

Unbanked

FINRA State By State Financial Capability Study

70

Dependent Variable

Coefficient P-Value * Coefficient P-Value * Coefficient P-Value * Coefficient P-Value * Coefficient P-Value *

Savings Question Correct -0.005 0.029 * - - - - - - - -

Don't Know/Refused 0.002 0.387 - - - - - - - -

Inflation Question Correct - - -0.004 0.044 * - - - - - -

Don't Know/Refused - - 0.002 0.403 - - - - - -

Bond Price Question Correct - - - - 0.000 0.872 - - - -

Don't Know/Refused - - - - -0.001 0.583 - - - -

Mortgage Question Correct - - - - - - -0.009 0.000 * - -

Don't Know/Refused - - - - - - -0.001 0.730 - -

Stock Diversification Q. Correct - - - - - - - - -0.007 0.014 *

Don't Know/Refused - - - - - - - - -0.004 0.138

Gender (Male) Female -0.001 0.375 -0.002 0.298 -0.001 0.687 -0.001 0.427 -0.001 0.516

Age Cohort (35-55 years) 18-34 0.001 0.677 0.000 0.945 0.001 0.690 0.001 0.648 0.000 0.792

55+ -0.007 0.002 * -0.007 0.003 * -0.007 0.002 * -0.007 0.002 * -0.007 0.003 *

Race (White Non Hispanic) Black 0.008 0.000 * 0.007 0.001 * 0.008 0.000 * 0.007 0.002 * 0.008 0.000 *

Hispanic 0.003 0.150 0.003 0.171 0.003 0.145 0.003 0.228 0.003 0.162

Asian -0.006 0.198 -0.005 0.232 -0.005 0.245 -0.006 0.183 -0.005 0.248

Native American/Alaskan -0.002 0.523 -0.002 0.487 -0.003 0.451 -0.002 0.501 -0.003 0.475

Multiple Races -0.001 0.785 -0.001 0.824 -0.001 0.813 -0.001 0.750 -0.001 0.775

Marital Status (Married) Single 0.010 0.000 * 0.010 0.000 * 0.010 0.000 * 0.009 0.000 * 0.010 0.000 *

Divorced 0.011 0.000 * 0.011 0.000 * 0.011 0.000 * 0.011 0.000 * 0.011 0.000 *

Widow -0.002 0.575 -0.002 0.590 -0.002 0.631 -0.003 0.526 -0.002 0.602

Presence of Children Dependent Children 0.003 0.057 * 0.003 0.061 * 0.003 0.049 * 0.003 0.049 * 0.003 0.047

Education (High School Degree) Less than High School 0.023 0.000 * 0.024 0.000 * 0.025 0.000 * 0.024 0.000 * 0.024 0.000 *

Some College -0.008 0.000 * -0.008 0.000 * -0.009 0.000 * -0.008 0.000 * -0.008 0.000 *

College Degree -0.010 0.000 * -0.010 0.000 * -0.010 0.000 * -0.010 0.000 * -0.010 0.000 *

Post College Education -0.011 0.000 * -0.010 0.000 * -0.011 0.000 * -0.010 0.000 * -0.011 0.000 *

Work Status (Full Time) Self Employed 0.011 0.003 * 0.011 0.002 * 0.011 0.003 * 0.011 0.003 * 0.011 0.002 *

Part Time 0.004 0.164 0.004 0.187 0.004 0.189 0.004 0.211 0.004 0.171

Homemaker 0.014 0.000 * 0.014 0.000 * 0.015 0.000 * 0.014 0.000 * 0.014 0.000 *

Student 0.001 0.790 0.001 0.758 0.001 0.790 0.001 0.839 0.001 0.740

Disabled 0.008 0.027 * 0.008 0.023 * 0.009 0.019 * 0.008 0.025 * 0.009 0.020 *

Unemployed 0.029 0.000 * 0.029 0.000 * 0.030 0.000 * 0.028 0.000 * 0.030 0.000 *

Retired 0.000 0.939 0.000 0.917 0.000 0.901 0.000 0.903 0.000 0.898

Income ($35K to $75K) Less than $35K 0.015 0.000 * 0.015 0.000 * 0.015 0.000 * 0.014 0.000 * 0.015 0.000 *

$75K or more 0.000 0.993 0.000 0.982 0.000 0.918 0.000 0.991 0.000 0.935

Change in income Unexpected Drop in Income 0.003 0.070 * 0.003 0.077 * 0.003 0.095 * 0.003 0.057 * 0.003 0.089 *

Homeowner -0.015 0.000 * -0.015 0.000 * -0.016 0.000 * -0.015 0.000 * -0.016 0.000 *

Credit Card -0.046 0.000 * -0.046 0.000 * -0.046 0.000 * -0.045 0.000 * -0.046 0.000 *

Observations 26585 26585 26585 26585 26585

Pseudo R2 0.307 0.307 0.305 0.309 0.306

Observed P 0.051 0.051 0.051 0.051 0.051

Predicted P 0.013 0.013 0.013 0.013 0.013

Access to Credit and Assets

Unbanked

Table 4.2: Unbanked Households Relative to All Banked - Controlling for Financial Literacy

FINRA State By State Financial Capability Study

Probit Regression - Marginal Effects Reported

Variables

UnbankedUnbankedUnbanked

Demographics

Knowledge

Unbanked

71

Obs. Mean S.D. Obs. Mean S.D. Obs. Mean S.D.

Underbanked 26146 0.232 0.422 - - - - - -

Fully Banked 26146 0.768 0.422 - - - - - -

Gender Female* 26146 0.512 0.500 5894 0.520 0.500 20252 0.510 0.500

Age Cohort 18-34*** 26146 0.290 0.454 5894 0.413 0.492 20252 0.252 0.434

35-54*** 26146 0.381 0.486 5894 0.421 0.494 20252 0.369 0.483

55+*** 26146 0.329 0.470 5894 0.166 0.372 20252 0.379 0.485

Race White, Non-Hispanic*** 26146 0.698 0.459 5894 0.593 0.491 20252 0.730 0.444

Black, Non-Hispanic*** 26146 0.107 0.309 5894 0.181 0.385 20252 0.085 0.279

Hispanic*** 26146 0.128 0.334 5894 0.163 0.370 20252 0.118 0.322

Asian*** 26146 0.047 0.211 5894 0.034 0.180 20252 0.051 0.219

Native American/Alaskan** 26146 0.016 0.126 5894 0.027 0.163 20252 0.013 0.112

More than One Race* 26146 0.008 0.092 5894 0.010 0.099 20252 0.008 0.089

Marital Status Married*** 26146 0.552 0.497 5894 0.475 0.499 20252 0.575 0.494

Single*** 26146 0.265 0.441 5894 0.327 0.469 20252 0.247 0.431

Divorced*** 26146 0.137 0.344 5894 0.163 0.369 20252 0.129 0.336

Widow*** 26146 0.046 0.209 5894 0.036 0.185 20252 0.049 0.215

Dependent Children*** 26146 0.382 0.486 5894 0.525 0.499 20252 0.339 0.473

Education Less than High School*** 26146 0.026 0.160 5894 0.045 0.207 20252 0.021 0.142

High School Degree*** 26146 0.282 0.450 5894 0.348 0.476 20252 0.262 0.440

Some College*** 26146 0.427 0.495 5894 0.450 0.498 20252 0.420 0.494

College Degree*** 26146 0.166 0.372 5894 0.114 0.318 20252 0.181 0.385

Post College Education*** 26146 0.100 0.299 5894 0.043 0.203 20252 0.117 0.321

Work Status Self Employed 26146 0.081 0.274 5894 0.078 0.269 20252 0.082 0.275

Full Time Employed** 26146 0.374 0.484 5894 0.384 0.486 20252 0.370 0.483

Part Time 26146 0.096 0.295 5894 0.096 0.295 20252 0.096 0.295

Homemaker*** 26146 0.087 0.282 5894 0.108 0.311 20252 0.081 0.272

Student*** 26146 0.056 0.230 5894 0.063 0.243 20252 0.054 0.226

Disabled*** 26146 0.041 0.198 5894 0.062 0.241 20252 0.034 0.182

Unemployed*** 26146 0.084 0.278 5894 0.126 0.332 20252 0.072 0.258

Retired*** 26146 0.181 0.385 5894 0.083 0.276 20252 0.210 0.407

Income Less than $35K*** 26146 0.378 0.485 5894 0.505 0.500 20252 0.340 0.474

$35K to $75K 26146 0.363 0.481 5894 0.365 0.482 20252 0.362 0.481

$75K or more*** 26146 0.259 0.438 5894 0.130 0.336 20252 0.298 0.457

Unexpected Drop in Income*** 25779 0.397 0.489 5821 0.541 0.498 19958 0.353 0.478

Homeowner*** 25988 0.616 0.486 5855 0.414 0.493 20133 0.677 0.468

Credit Card*** 25619 0.782 0.413 5813 0.608 0.488 19806 0.835 0.372

* difference at the 10% level

** the 5% level

*** the 1% level

a. 398 households responded that they did not know or refused to report if they used alternative financial services

Underbanked Fully Banked

Table 5a: Underbanked versus Fully Banked

Bank Participation

Access to Credit and Assets

Demographics

FINRA State By State Financial Capability Study

Descriptive Statistics

All Bankeda

72

Obs. Mean S.D. Obs. Mean S.D. Obs. Mean S.D.

Underbanked 41813 0.203 0.402 - - - - - -

Fully Banked 41813 0.797 0.402 - - - - - -

Gender Female*** 41813 0.484 0.500 8110 0.507 0.500 33703 0.478 0.500

Age Cohort 18-34*** 41813 0.208 0.406 8110 0.287 0.452 33703 0.188 0.390

35-54*** 41813 0.395 0.489 8110 0.443 0.497 33703 0.383 0.486

55+*** 41813 0.397 0.489 8110 0.270 0.444 33703 0.429 0.495

Race Caucasian*** 41813 0.753 0.431 8110 0.601 0.490 33703 0.792 0.406

Black*** 41813 0.098 0.297 8110 0.206 0.404 33703 0.070 0.256

Hispanic*** 41813 0.095 0.293 8110 0.150 0.357 33703 0.081 0.273

Asian*** 41813 0.038 0.192 8110 0.016 0.125 33703 0.044 0.205

Native American/Alaskan*** 41813 0.005 0.071 8110 0.010 0.100 33703 0.004 0.061

Other*** 41813 0.011 0.104 8110 0.018 0.132 33703 0.009 0.096

Marital Status Married*** 41813 0.540 0.498 8110 0.452 0.498 33703 0.563 0.496

Single*** 41813 0.189 0.392 8110 0.257 0.437 33703 0.172 0.378

Divorced*** 41813 0.169 0.375 8110 0.221 0.415 33703 0.156 0.363

Widow*** 41813 0.101 0.301 8110 0.070 0.255 33703 0.109 0.311

Dependent Children*** 41813 0.291 0.454 8110 0.361 0.480 33703 0.273 0.445

Education Less than High School*** 41813 0.101 0.301 8110 0.155 0.362 33703 0.087 0.281

High School Degree*** 41813 0.286 0.452 8110 0.340 0.474 33703 0.273 0.445

College Degree*** 41813 0.208 0.406 8110 0.121 0.326 33703 0.231 0.421

Post College Education*** 41813 0.116 0.320 8110 0.052 0.222 33703 0.132 0.339

Work Status Full Time Employed*** 41625 0.551 0.497 8064 0.567 0.495 33561 0.547 0.498

Part Time 41625 0.079 0.270 8064 0.081 0.273 33561 0.079 0.269

Homemaker*** 41625 0.063 0.243 8064 0.077 0.266 33561 0.059 0.236

Student 41625 0.006 0.077 8064 0.006 0.079 33561 0.006 0.076

Disabled*** 41625 0.045 0.208 8064 0.076 0.266 33561 0.037 0.190

Unemployed*** 41625 0.049 0.215 8064 0.083 0.276 33561 0.040 0.196

Retired*** 41625 0.207 0.405 8064 0.109 0.312 33561 0.232 0.422

Income Less than $35K*** 36194 0.351 0.477 7365 0.483 0.500 28829 0.315 0.464

$35K to $75K 36194 0.352 0.477 7365 0.353 0.478 28829 0.351 0.477

$75K or more*** 36194 0.298 0.457 7365 0.164 0.370 28829 0.334 0.472

Homeowner*** 41813 0.718 0.450 8110 0.530 0.499 33703 0.765 0.424

* difference at the 10% level

** the 5% level

*** the 1% level

a. 1,029 households responded that they did not know or refused to report if they used alternative financial services. An additional

672 households did not answer these questions

Table 5b: Underbanked versus Fully Banked

Bank Participation

Access to Credit and Assets

Demographics

FDIC Survey of Unbanked and Underbanked Households

Descriptive Statistics

All Bankeda Underbanked Fully Banked

73

Obs. Mean S.D. Obs. Mean S.D. Obs. Mean S.D.

Underbanked 26146 0.232 0.422 - - - - - -

Fully Banked 26146 0.768 0.422 - - - - - -

# of Questions Correct*** 26146 3.078 1.403 5894 2.673 1.356 20252 3.200 1.393

Don't Know/Refused*** 26146 1.203 1.360 5894 1.387 1.401 20252 1.147 1.342

Savings Question*** 26146 0.794 0.404 5894 0.734 0.442 20252 0.812 0.390

Inflation Question*** 26146 0.666 0.472 5894 0.550 0.498 20252 0.701 0.458

Bond Price Question*** 26146 0.285 0.451 5894 0.232 0.422 20252 0.301 0.459

Mortgage Question*** 26146 0.779 0.415 5894 0.719 0.450 20252 0.797 0.402

Stock Diversification Q.*** 26146 0.554 0.497 5894 0.439 0.496 20252 0.589 0.492

* difference at the 10% level

** the 5% level

*** the 1% level

Table 6: Underbanked versus Fully Banked subset of Financial Literacy Variables

Bank Participation

Actual Knowledge

FINRA State By State Financial Capability Study

Descriptive Statistics

All Banked Underbanked Fully Banked

74

Dependent Variable

Coefficient P-Value *

Gender (Male) Female -0.014 0.035 *

Age Cohort (35-55 years) 18-34 0.051 0.000 *

55+ -0.076 0.000 *

Race (White Non Hispanic) Black 0.121 0.000 *

Hispanic 0.007 0.530

Asian -0.036 0.056 *

Native American/Alaskan 0.073 0.002 *

Multiple Races 0.017 0.470

Marital Status (Married) Single -0.015 0.122

Divorced 0.025 0.012 *

Widow 0.041 0.025 *

Presence of Children Dependent Children 0.099 0.000 *

Education (High School Degree) Less than High School 0.033 0.134

Some College -0.021 0.007 *

College Degree -0.071 0.000 *

Post College Education -0.088 0.000 *

Work Status (Full Time) Self Employed -0.021 0.092 *

Part Time -0.043 0.000 *

Homemaker -0.014 0.238

Student -0.067 0.000 *

Disabled 0.048 0.006 *

Unemployed -0.022 0.061 *

Retired -0.052 0.000 *

Income ($35K to $75K) Less than $35K 0.009 0.296

$75K or more -0.075 0.000 *

Change in income Unexpected Drop in Income 0.091 0.000 *

Homeowner -0.102 0.000 *

Credit Card -0.120 0.000 *

Observations 25174

Pseudo R2 0.149

Observed P 0.234

Predicted P 0.198

Access to Credit and Assets

Demographics

Table 7a: Underbanked Households Relative to Fully Banked

FINRA State By State Financial Capability Study

Probit Regression - Marginal Effects Reported

Variables

Underbanked

75

Dependent Variable

Coefficient P-Value *

Gender (Male) Female -0.002 0.734

Age Cohort (35-55 years) 18-34 0.008 0.251

55+ -0.049 0.000 *

Race (White Non Hispanic) Black 0.209 0.000 *

Hispanic 0.062 0.000 *

Asian -0.097 0.000 *

Native American/Alaskan 0.192 0.000 *

Other 0.116 0.000 *

Marital Status (Married) Single 0.007 0.343

Divorced 0.036 0.000 *

Widow -0.009 0.388

Presence of Children Dependent Children 0.035 0.000 *

Education (High School Degree) Less than High School 0.044 0.000 *

Some College -0.015 0.016 *

College Degree -0.101 0.000 *

Post College Education -0.099 0.000 *

Work Status (Full Time) Part Time -0.013 0.130

Homemaker -0.005 0.619

Student -0.075 0.005 *

Disabled 0.051 0.000 *

Unemployed 0.064 0.000 *

Retired -0.075 0.000 *

Income ($35K to $75K) Less than $35K 0.025 0.000 *

$75K or more -0.055 0.000 *

Access to Credit and Assets Homeowner -0.105 0.000 *

Observations 36024

Pseudo R2 0.123

Observed P 0.214

Predicted P 0.185

Demographics

Table 7b: Underbanked Households Relative to Fully Banked

FDIC Survey of Unbanked and Underbanked Households

Probit Regression - Marginal Effects Reported

Variables

Underbanked

76

Coefficient P-Value * Coefficient P-Value *

# of Questions Correct -0.015 0.000 * - -

Savings Question - - -0.018 0.044 *

Inflation Question - - -0.029 0.000 *

Bond Price Question - - -0.009 0.225

Mortgage Question - - 0.012 0.154

Stock Diversification Q. - - -0.025 0.001 *

Gender (Male) Female -0.023 0.001 * -0.024 0.000 *

Age Cohort (35-55 years) 18-34 0.047 0.000 * 0.045 0.000 *

55+ -0.074 0.000 * -0.073 0.000 *

Race (White Non Hispanic) Black 0.113 0.000 * 0.114 0.000 *

Hispanic 0.004 0.703 0.005 0.678

Asian -0.040 0.035 * -0.038 0.041 *

Native American/Alaskan 0.073 0.001 * 0.073 0.001 *

Multiple Races 0.016 0.478 0.017 0.457

Marital Status (Married) Single -0.017 0.085 * -0.016 0.104 *

Divorced 0.025 0.014 * 0.025 0.013 *

Widow 0.040 0.028 * 0.041 0.022 *

Presence of Children Dependent Children 0.097 0.000 * 0.097 0.000 *

Education (High School Degree) Less than High School 0.028 0.202 0.028 0.204

Some College -0.014 0.085 * -0.013 0.096 *

College Degree -0.061 0.000 * -0.060 0.000 *

Post College Education -0.078 0.000 * -0.076 0.000 *

Work Status (Full Time) Self Employed -0.020 0.111 -0.019 0.121

Part Time -0.043 0.000 * -0.042 0.000 *

Homemaker -0.017 0.164 -0.016 0.176

Student -0.066 0.000 * -0.064 0.000 *

Disabled 0.045 0.010 * 0.045 0.010 *

Unemployed -0.023 0.049 * -0.022 0.062 *

Retired -0.052 0.000 * -0.051 0.000 *

Income ($35K to $75K) Less than $35K 0.005 0.549 0.006 0.476

$75K or more -0.072 0.000 * -0.071 0.000 *

Change in income Unexpected Drop in Income 0.092 0.000 * 0.091 0.000 *

Homeowner -0.099 0.000 * -0.100 0.000 *

Credit Card -0.117 0.000 * -0.118 0.000 *

Observations 25174 25174

Pseudo R2 0.151 0.151

Observed P 0.234 0.234

Predicted P 0.197 0.197

Table 8.1: Underbanked Households Relative to Fully Banked - Controlling for Financial Literacy

FINRA State By State Financial Capability Study

Probit Regression - Marginal Effects Reported

Variables

Access to Credit and Assets

Demographics

Knowledge

Underbanked Underbanked

77

Dependent Variable

Coefficient P-Value * Coefficient P-Value * Coefficient P-Value * Coefficient P-Value * Coefficient P-Value *

Savings Question Correct -0.054 0.000 * - - - - - - - -

Don't Know/Refused -0.047 0.000 * - - - - - - - -

Inflation Question Correct - - -0.059 0.000 * - - - - - -

Don't Know/Refused - - -0.036 0.001 * - - - - - -

Bond Price Question Correct - - - - -0.032 0.000 * - - - -

Don't Know/Refused - - - - -0.035 0.000 * - - - -

Mortgage Question Correct - - - - - - -0.024 0.038 * - -

Don't Know/Refused - - - - - - -0.033 0.016 * - -

Stock Diversification Q. Correct - - - - - - - - -0.084 0.000 *

Don't Know/Refused - - - - - - - - -0.058 0.000 *

Gender (Male) Female -0.016 0.017 * -0.018 0.008 * 0.007 0.511 -0.014 0.044 * -0.017 0.011 *

Age Cohort (35-55 years) 18-34 0.051 0.000 * 0.044 0.000 * 0.009 0.289 0.051 0.000 * 0.048 0.000 *

55+ -0.076 0.000 * -0.074 0.000 * 0.009 0.330 -0.077 0.000 * -0.075 0.000 *

Race (White Non Hispanic) Black 0.118 0.000 * 0.114 0.000 * 0.012 0.106 0.120 0.000 * 0.115 0.000 *

Hispanic 0.006 0.603 0.004 0.731 0.011 0.128 0.007 0.542 0.005 0.626

Asian -0.037 0.049 * -0.038 0.042 * 0.018 0.046 * -0.035 0.063 * -0.037 0.047 *

Native American/Alaskan 0.071 0.002 * 0.071 0.002 * 0.025 0.016 * 0.072 0.002 * 0.072 0.002 *

Multiple Races 0.016 0.495 0.016 0.503 0.024 0.008 * 0.016 0.501 0.014 0.536

Marital Status (Married) Single -0.016 0.107 -0.016 0.093 * 0.010 0.262 -0.015 0.126 -0.016 0.101 *

Divorced 0.025 0.013 * 0.025 0.013 * 0.010 0.138 0.025 0.014 * 0.025 0.012 *

Widow 0.041 0.022 * 0.042 0.021 * 0.019 0.046 * 0.040 0.027 * 0.040 0.028 *

Presence of Children Dependent Children 0.098 0.000 * 0.097 0.000 * 0.008 0.385 0.099 0.000 * 0.098 0.000 *

Education (High School Degree) Less than High School 0.032 0.142 0.031 0.158 0.023 0.026 * 0.033 0.136 0.032 0.152

Some College -0.019 0.018 * -0.018 0.027 * 0.008 0.427 -0.021 0.007 * -0.017 0.033 *

College Degree -0.068 0.000 * -0.065 0.000 * 0.008 0.167 -0.070 0.000 * -0.065 0.000 *

Post College Education -0.085 0.000 * -0.082 0.000 * 0.009 0.101 * -0.088 0.000 * -0.082 0.000 *

Work Status (Full Time) Self Employed -0.021 0.090 * -0.020 0.106 0.012 0.082 * -0.021 0.095 * -0.020 0.111

Part Time -0.043 0.000 * -0.043 0.000 * 0.011 0.096 * -0.043 0.000 * -0.042 0.000 *

Homemaker -0.014 0.227 -0.014 0.226 0.012 0.086 * -0.014 0.233 -0.015 0.217

Student -0.067 0.000 * -0.065 0.000 * 0.013 0.055 * -0.067 0.000 * -0.066 0.000 *

Disabled 0.047 0.007 * 0.047 0.007 * 0.018 0.040 * 0.049 0.006 * 0.048 0.006 *

Unemployed -0.022 0.060 * -0.022 0.062 * 0.011 0.084 * -0.022 0.058 * -0.021 0.068 *

Retired -0.052 0.000 * -0.051 0.000 * 0.011 0.181 -0.051 0.000 * -0.052 0.000 *

Income ($35K to $75K) Less than $35K 0.008 0.318 0.007 0.371 0.008 0.375 0.009 0.297 0.006 0.497

$75K or more -0.073 0.000 * -0.073 0.000 * 0.008 0.261 -0.075 0.000 * -0.073 0.000 *

Change in income Unexpected Drop in Income 0.090 0.000 * 0.091 0.000 * 0.007 0.398 0.091 0.000 * 0.090 0.000 *

Homeowner -0.102 0.000 * -0.102 0.000 * 0.008 0.616 -0.102 0.000 * -0.101 0.000 *

Credit Card -0.120 0.000 * -0.120 0.000 * 0.009 0.784 -0.120 0.000 * -0.118 0.000 *

Observations 25174 25174 25174 25174 25174

Pseudo R2 0.150 0.151 0.150 0.149 0.151

Observed P 0.234 0.234 0.234 0.234 0.234

Predicted P 0.198 0.198 0.198 0.198 0.197

Variables

Knowledge

Demographics

Access to Credit and Assets

Table 8.2: Underbanked Households Relative to Fully Banked - Controlling for Financial Literacy

FINRA State By State Financial Capability Study

Probit Regression - Marginal Effects Reported

Underbanked Underbanked Underbanked Underbanked Underbanked