• No results found

How To Determine Why People Choose Debit Card Over Credit Card

N/A
N/A
Protected

Academic year: 2021

Share "How To Determine Why People Choose Debit Card Over Credit Card"

Copied!
32
0
0

Loading.... (view fulltext now)

Full text

(1)

Why Do People Use Debit Cards?

Evidence From Checking Accounts

Marc Anthony Fusaro College of Business Arkansas Tech University

Russellville AR 72801 Abstract

Debit cards are the fastest growing consumer payment method despite being more expensive and less versatile than credit cards. In this paper, We investigate some of the oft cited explanations for debit card use. Checking account data shows that debit card use is correlated with age, pay frequency, overdrafting and ATM use, but not income, gender, crime, or expenditure. The data contain some signs that debit cards might be used as a method of spending restraint.

Keywords: debit cards, credit cards, payment choice, self control, bank JEL codes: D14, G21, L14

The author would like to thank Don Dutkowsky, Ron Borzekowski, Rick Ericson, Brad Heim, Peter Meyer, Greg Kurtzon, Mark Manuzak, Lester Zeager, Robert Prati and Randy Parker and participants at the ASSA, MEA, Boston Fed Consumer Payments, and Atlanta/New York Fed Consumer Payments conferences for their helpful comments. Please address correspondence to Marc Anthony Fusaro, College of Business, Arkansas Tech University, Russellville, AR 72801.

(2)

1. Introduction

Debit card use has grown faster than any other payment instrument in the U.S. A minor segment of the market used debit cards a decade ago. However, figure 1 shows that due to rapid growth in their use, more transactions are made with debit than with credit cards. The puzzle of debit’s popularity is that explicit costs appear to favor credit over debit. With debit, money immediately comes out of the bank account. Whereas, with credit, the household is billed for purchases at the end of the month and pays within a further grace period. As long as no balance is carried on the credit card, the consumer pays no interest. This float may be minor if evaluated at a savings account interest rate; however, for liquidity constrained households, the opportunity cost – payday loans or bank bounce protection – is quite high. Credit cards also offer the option to convert the balance into an uncollateralized loan with zero transaction cost. Most credit cards offer rewards (cash back, airline miles, gift certificates, warranty on purchase, charitable

donations, etc.) or a below market interest rate. Only 5.7% of banks offer rewards on their signature debit cards (Moebs 2006) and some financial institutions (15%) charge a fee for PIN debit purchases, ranging from ten cents to two dollars per transaction.

The goal of this paper is to investigate the factors that have caused a seemingly

dominated product to become a market leader. Consumer debit card choice has implication for monetary policy as such choices affect money demand and money transmission mechanisms. With $894 billion in credit card debt at the end of 2009 and 1.4 million personal bankruptcies in 2009, Americans’ ability to effectively manage personal finances can have a strong effect on standard of living. Since payment card networks exhibit network externalities, market

concentration and its suppressive effects on competition is of concern to anti-trust regulators. The extent to which these network externalities exist is dependant on the degree of

substitutability of competing payment media.

(3)

(2008), and Borzekowski, Kiser and Ahmed (2008) offer several motivations for choosing debit over credit: cost, convenience, merchant acceptance, time at checkout, purchase tracking, safety, being credit constrained, and spending restraint. Zinman notes that up to 31% of households could use debit for self-restraint. Uniquely, we use household transactions level data, a course of action recommended by Zinman. The sufficiently rich data is consistent with spending control, purchase tracking,

merchant acceptance, and pecuniary motives (paying interest when revolving a credit card balance) as an explanation for some households’ debit use.

Much of the growth in debit card use has come at the expense of checks or cash. From 1995 to 2003, electronic payments grew from 23 percent to 55 percent of all payments.

Borzekowski and Kiser (2008) find that the closest substitutes for debit are cash, checks and credit in that order. However, the debit vs credit comparison is quite clean (debit and credit equate in many dimensions considered by Klee, 2006) and recently available data to track debit transactions as opposed to the anonymity of cash. However, with ATM withdrawals observable, we are able to compare debit and credit to cash.

To speculate whether debit card use might be correlated with spending control motives, we look for evidence that debit card users are paying down a credit card balance. The checking account data contains credit card payments but not credit card balances. While the data do not contain full information about credit card balances, they do contain payments to credit cards. This is enough information to show that debit users are more likely to pay the same amount in multiple months, a sign that they are paying a credit card balance. This fact could indicate that

FIGURE 1

Payment Card Usage, 1998-2006

(4)

debit card users are likely to be paying down a credit card balance or it could indicate that debit card users are at their credit limits.

Section 2 describes the hypothesized motivations for debit card use. Section 3 describes the data, and defines the variables. Regression results are presented and discussed in section 4. Section 5 compares our checking account records to survey results and section 6 concludes.

2. Explanations For Debit Card Use

To explain the determinants of debit card use we examine the relationship between debit card use and demographic, financial and behavioral variables. This section explains these demographic, financial and behavioral variables. Because there are two different types of debit card, we examine the effect on total debit card use, signature debit card use, and PIN debit card use. PIN debit cards use the ATM network infrastructure to transfer data, and money, between customers, merchants and their respective banks. Identity is verified by a four digit PIN number. Signature debit cards use the Visa or MasterCard network to transfer data and identity is verified by the customer’s signature.

2.1 Demographic Variables

In separate surveys, Schuh and Stavins (2010) and Arango, Huynh, and Sabetti (2011) investigated the relationship between debit card use and demographic variables. Both papers find that debit cards are more popular among younger account holders so we measure the effect of age on debit use. We also look at the effect of time since the account was opened. We

measure whether there is any gender component in debit card use as Arango, Huynh, and Sabetti (2011) found. Schuh and Stavins (2010) found that married individuals are more likely to use

(5)

debit than single or divorced individuals. It is possible that married people find it difficult to manage a checking account when two people are charging expenses to it. Rather, they might put expenses on a credit card and then one account manager would pay the bills monthly. The respondents of both surveys as well as Borzekowski and Kiser’s (2008) respondents cited

“safety” as a motivation for using debit. Thus we investigate whether the amount of street crime influences the willingness of a person to carry cash. An individual who lives, works, or plays in a high crime area might use a debit card as a seemingly safer alternative to cash. Again all three surveys asked their respondents about the connection between debit cards and speed,

convenience and record-keeping benefits; thus we look for a correlation between debit card use and opportunity cost of time.

2.2 Financial Variables

The choice to use a debit card could be related to financial variables. These variables seek to measure financial resources and indicators of how finances are managed. Schuh and Stavins’ (2010), and Mantel and McHugh (2001) find an income effect on debit use. However, Stavins (2001) and Borzekowski and Kiser (2008) find no such relationship. The conflicting results on income may derive from income being a proxy for another variable such as expenses or wealth. Our data allows us to control for income, expenses, and wealth.

An important determinant in debit card use, noted by Schuh and Stavins (2010) is education. The checking account records do not contain any information on educational attainment, however, the frequency with which paychecks are received should be correlated to education.

The decision to use a debit card has been found (Arango, Huynh, and Sabetti, 2011) to be related to the decision to use a credit card. A credit card user who carries a balance is charged interest beginning on the day of the purchase; thus a credit revolver would avoid interest

(6)

to ATM use. Surveys of debit users (Schuh and Stavins, 2010, Ching and Hayashi, 2010, and Borzekowski and Kiser, 2008) cite convenience as a motive for debit card use; therefore we measure the correlation between debit use and the use of free ATMs. We relate debit card use to check bouncing activity, and the variation in account balance that might increase the likelihood of bouncing checks.

Finally, another important determinant in debit card use, noted by Schuh and Stavins (2010), Bolt, Jonker and Renselaar (2010), Rochet and Wright (2010), Simon, Smith and West (2010), Ching and Hayashi (2010), and Carbo-Valverde and Linares-Zegarra (2011) is cost (in the form of rewards or surcharges), however, the checking account records do not show any variation in this dimension as they are derived from the same depository institution.

2.3 Spending Restraint

The literature references spending control as a potential motive for using a debit card (Arango, Huynh, and Sabetti 2011, Borzekowski, Kiser and Ahmed 2008, Zinman 2004). The idea behind this is as such. Some consumers seem to have difficulty matching a long term spending plan with their current spending. This causes them to overspend, accumulate debt, patronize high cost lending sources, overdraft, and generally suffer other consequences. This phenomenon is modeled by Thaler and Shefrin (1981), Bertaut and Haliassos (2002), and Prelec and Loewenstein (1998) and empirically investigated by Prelec and Simester (2001), Laibson, Repetto and Tobachman (2007), Ashraf, Karlin and Yin (2006) and Ausubel (1991). This literature attributes the disconnect between current spending and long term spending management to behavioral causes. For example, dual-self models posit that the household consists of two decision makers, a planner responsible for solving an intertermporal optimization problem and a shopper responsible for current consumption. Thus the planner will take actions in order to constrain the shopper’s choices. One example of this type of action is taking small amounts of cash out of the ATM for the shopper to spend. Another example is taking away the

(7)

shopper’s credit card and replacing it with a debit card. Thus using a debit card can be seen as a form of self-control, constraining ones current spending potential to current liquid assets even if a creditor would be willing to extend credit to the individual.

Alternatively, households could suffer from imperfect information. Fusaro (2008) models a debit card as being a lower cost method of calculating affordable spending than

acquiring full information and optimizing. Similar models are presented by Ameriks, Caplin and Leahy (2004) and Kocherlakota (1998). Also, Zinman (2004) refutes the behavioral spending restraint explanation in favor of rational motives.

Whether consumers used a debit card because of spending restraint or because it conveys information, it will appear the same in their behavior as viewed through their bank accounts. Herein is presented three tests which, if confirmed in the data, would be consistent with both behavioral spending restraint and information acquisition. Thus, it is impossible to distinguish empirically between the two theories.

First, if some consumers’ spending control stems from impulse purchases, then we would expect to see them shopping in a greater variety of stores than those who do not suffer from such an impulse buying problem. If these impulse buyers choose to use debit cards as a form of spending control, to limit their impulse shopping, then we would see a correlation between debit card purchases and store variety.

Second, often a household will accumulate a high level of credit card debt before realizing that its level of spending is unsustainable given its income. At some point the

household will take action to change its spending habits (control its level of spending) and pay down its debts. If cutting up1 its credit cards and paying via a debit card is one of the measures it takes to restrain its spending, then debit card use would be correlated with revolving a credit card

1Though some consumers get out the scissors, many methods are employed to restrain their use of credit cards. The author’s personal favorite is a customer who froze his credit card, literally. He put it in a tub of water which he placed in the freezer. The author has read of this story multiple times and has seen the suspended credit card on one occasion.

(8)

balance.

Third, one example of a method to control spending is to carry small amounts of cash. A second method is to use a debit card instead of a credit card for purchases. If consumers use both methods of spending control, then we would see a correlation between debit card use and visiting ATMs frequently withdrawing small amounts each time. These three tests are similar to those used in Fusaro (2008).

3. Data and Variable Definitions

In this section, we describe the data set and define the variables listed above. The empirical analysis employs a sample of transaction records from 2,312 checking accounts obtained from a small depository institution in the Midwest. Two accounts are discarded due to lack of customer information, leaving us with 2310 accounts. The data include customer

information and all transactions with associated balances from June, July, and August 2003. For each transaction, the data contain the date, amount, balance and information on the source or destination. For ATM or debit card transactions, the address and/or store name is listed. The data also contain customer information including age, account tenure, gender, and a joint account indicator. Table 1 shows summary statistics for the data. The median customer was 44 years old. The average checking account balance was $1843. The average annual income for customers was $31,000. This was below the national average, although this figure represents only that portion of income received via direct deposit into this account.

3.1 Demographic Variables

(9)

joint account and crime. The birth date of the account holder was provided with the account transaction records. Likewise, whether the account is a joint account was included from the data provider (joint). For data security purposes, account numbers and full names were not included in any of the information provided to us. While gender is not included in the institution’s database, they were able to provide us with the account holder’s first (given) name. From this we were able to determine that 46% of account holders were male, 45% were female, and the remaining 9% had dual-gender or otherwise uninformative names.

Crime is measured through the account records and law enforcement crime statistics. In our bank account transaction records the address and/or business name is provided for ATM and debit card transactions. Crime statistics for every city in the state were obtained from the FBI uniform crime reports. Crime statistics for every zip code in the nearest large city were obtained from the city police department. Each transaction for which an address is identified, and the address is in a city covered in our crime data, can then be connected with a per capita crime rate. The maximum crime rate for each account observed each day is calculated to determine how unlikely the account holder might be to carry cash that day. The average of these daily crime maxima is taken to provide a measure for the average crime rate in the neighborhoods in which the account holder travels. The average crime rate across the locales in which an account holder travels is identified for 74% of accounts. Among this group, on average, a location can be identified on 20% of days and a location can be identified for 78% of ATM withdrawals.

Three proxies for opportunity cost of time are: having a non-working adult in the

household, working multiple jobs, and being a single parent. Having a non-working adult in the household is proxied by the interaction term (Joint*1p) where Joint, as defined above is the indicator for a joint account (Joint) and 1p is an indicator for a household that receives only one paycheck. Ceteras paribus, a family with x kids and two working parents will have less time to balance the checkbook than a family with x kids, one working parent and one full-time home-keeping parent. Likewise a family with x kids and a single working parent has less time to

(10)

balance the checkbook than a family with x kids, one working parent and one full-time home-keeping parent. The potential problem with this variable is that we can not effectively control for the number of children in the household and households with more children are more likely to have a full-time homekeeper. Define the second variable (Jobs), as the number of paychecks from distinct employers. The more jobs one works, the higher the opportunity cost of time, ceteras paribus. The final variable (Divorce) captures single parent households and is an

indicator for a household which receives child support payments. These households are assumed to have a relatively higher opportunity cost of time. Only 29 households in the sample received child support payments.

3.2 Financial Variables

Financial variables entering the regressions include income, paycheck frequency,

variation in total expenditure, wealth, precautionary balances, bounced checks, proximity to free ATMs, credit card use, and ATM use.

Income is observed through direct deposit of paychecks, where they are identifiable. Define income as annualized income calculated from paychecks. Define payfreq as the number of days between paychecks. Most are weekly (7 days), bi-weekly (14 days), semimonthly (15.25 days), or monthly (30.5 days). Expenses are measured via Baumol-Tobin style transactions balances, i.e. the difference between the peak checking account balance achieved just after a paycheck and the low checking account balance observed just before a paycheck is received. This transaction balance is calculated for every pay cycle and then define transaction balances (BalTr) for the customer as the average across the pay cycles.

The data is not rich enough to observe total wealth, however, some measures of cash balances are observable. If the account holder has a savings account at the same institution, the balances of that account are observed. Define the average savings account balance to be BalSav. Further, Baumol-Tobin style precautionary checking account balances is observed. Define

(11)

BalPr as the average low point observed just before an account’s paychecks. While having total wealth would be ideal, these measures of liquid wealth are not without value.

In the bank account data, non-sufficient funds (NSF) transactions can be identified. The data provider also included the number of NSFs since the account had been opened. Define the variables NSFnow and NSFhist respectively to be the number of pay periods in which the account overdraws during the three month sample and the number of NSFs per year since opening the account. Define an indicator variable, ATM, for whether the account holder uses ATMs which is generated from the transaction records. In order to measure the usage of free ATMs, a list of ATMs in the same network, which would be free to the customers was cross referenced with the transaction data. Define ATMfree to be the percentage of ATM withdrawals made at ATMs that are free to the account holder.

Over 60% of account holders used debit cards. The average customer made 125 withdrawal transactions in the three month sample, and the average debit card user made 49 debit card transactions. This rate of debit usage is comparable to the Survey of Consumer Finances (SCF) and the survey reported in Borzekowski, Kiser and Ahmed (2008). Three variables measure debit use. We measure the total number of debit purchases (debittrx), the number of PIN debit transactions (PINdebittrx), and the number of signature debit transactions (sigdebittrx).

Credit card users are elusive because the data allow us to observe only the checking account. Define a credit user (credit) as an account holder who made a credit card payment. On average, credit users made four payments. The mode was three payments – generally to the same card – over the three month sample. These payments are identified through the descriptive field of ACH2 transactions in the data. For example, ACH payments made to “American

Express” or “Discover” or “Cred Crd Pmt” are identified as payments toward credit card

2Automated Clearing House is the electronic money transfer system through which direct deposit and direct billing operates.

(12)

balances. (See the Appendix for a full list of the descriptions identifying credit card payments.) Thus credit card payments are identified when either made electronically (ACH) by the

customer, or converted to an ACH transaction by the credit card issuer in a process known as

All Accounts Credit Usersa

Obs Mean Std

Dev All UsersDebit

Debit Non-Users p-value b Units Debit User Transactions Debittrx Pin Debit Signature 2310 2310 1517 819 1478 .606 125.3 47.9 7.96 44.7 – 87.1 37.3 9.73 35.9 .778 167.6 55.1 7.89 51.8 – 188.1 57.4 8.16 53.1 – 95.8 – – – – .000** – – – Indicator Countc Countc Countc Countc Age Female Joint Crime Joint*1p Jobs Divorce 2310 2107 2310 1730 2167 1483 2167 45.0 .497 .506 3.77 .261 1.27 .013 13.5 – – 1.62 – .576 – 42.7 .479 .567 3.80 .316 1.27 .009 41.5 .481 .554 3.85 .318 1.28 .009 46.9 .473 .610 3.44 .309 1.27 .010 .000** .001** .321 .097 .874 .974 .886 Years Indicator Indicator Percentage Indicator Countc Indicator Income PayFreq Acct Bal BalTr BalPr BalSav NSFnow NSFhist ATM ATMfree Credit 1483 1483 2310 1483 1483 355 2161 2310 2310 1340 2310 31.4 14.96 1.84 1.24 1.58 4.98 0.47 25.0 .644 21.2 .195 22.3 7.04 4.39 1.09 4.92 17.7 1.52 71.4 – 28.7 – 38.0 13.4 1.59 1.39 1.08 6.52 0.54 32.3 .800 21.3 – 37.5 12.9 1.35 1.33 0.79 7.12 0.59 30.7 .911 20.6 – 40.0 15.6 2.45 1.62 2.23 2.70 0.35 37.8 .410 30.0 – .474 .000** .000** .013* .000** .624 .182 .437 .000** .087 – Thousand $ Days Thousand $ Thousand $ Thousand $ Thousand $ Countc Countc Indicator Percentage Indicator Disp ATM locations CC Payments Round Same Revolver ATMvisits ATMcash 1340 1340 450 300 450 450 1340 1340 .500 6.13 3.94 2.86 .333 .482 13.5 .036 .275 5.08 4.02 2.46 – – 12.1 .077 .512 6.89 – – – – 14.4 .033 .516 7.13 4.12 3.02 .374 .517 14.9 .033 .471 4.15 3.33 2.16 .190 .360 8.93 .034 .348 .006** .083 .016* .001** .006** .007** .935 Ratio Countc Countc Countc Indicator Indicator Countc Ratio

a Figure reported are the means when sample is restricted to credit users and the indicated subgroups. b P-value of a test for independence between debusers and non-debit users. If the values in the previous two

columns are means, then it is the p-value of a t-test, if they are indicator variables then it is the p-value of a χ2

test. Significance is indicated by ** for the 1% level, and * for the 5% level.

c All count variables are for the entire 3 month sample.

(13)

check truncation whereby the receiving party of a check, in this case the card issuing bank, scans the check and it becomes an electronic transaction. In the data, nearly 20% of accounts made at least one identified payment toward a credit card balance as compared to 72% in the SCF. What accounts for the 52 percentage point discrepancy? Two types of credit card payments can not be identified: those that were processed as a paper check through the entire payments system, and those that have an ambiguous or misleading description. The latter should be random variation; however, the former could introduce bias into this measure of credit card users.

In order to assess the representativeness of the data, we compare it with the SCF. In terms of debit use, age, gender, joint account status, and child support, credit users are similar to the general population while credit users have 30% higher income than non-credit users

(according to the SCF). This was true in the transactions data as well, albeit with only a 20% income premium for credit-users. Also, the identified credit revolvers in the transactions data were similar to credit revolvers in the SCF (48% and 44% respectively), indicating that little if any bias is introduced by the inability to detect all credit users. The exception to this lack of bias is that debit users are over represented among identified credit users. In the SCF, 58% of people are debit users, whether looking at the whole population or just credit users. In the transactions data, however, debit users were more likely than debit non-users to be identified as credit users – more likely to pay their bills online. Indeed, 78% of credit users also use debit cards compared to 56% of credit non-users.

3.3 Variables That Might Correlate with Spending Control

Above, we highlighted three characteristics of people who may be in need of spending restraint. Impulse purchasers may find spending restraint necessary. Those who revolve a credit card balance may be trying to pay it down (as opposed to those who pay their credit card balance in full each month, who appear to be successful at spending management). Those carrying small volumes of cash at one time, relative to total cash spent have signaled a need to control their

(14)

spending. If these indicators are correlated with debit card use, then we would have some evidence that debit cards are potentially being used as a form of spending control.

We might choose to identify an impulse shopper as someone who has a greater variety of purchase locations. This pattern appears less disciplined than someone who returned to the same stores repeatedly. The ideal metric would be the amount of distinct purchase locations as a percent of all purchases; however, the data contains purchase locations for debit card purchases but not for credit card or cash purchases. Instead, let us shift our attention to a subset of

purchases. Consumers use cards (debit or credit) for high-value transactions and use cash for low-value transactions (Klee, 2006). With an incomplete picture of high-value (card)

transaction, we turn to low-value (cash) transactions. Cash purchase locations are unseen; however, ATM withdrawal locations are observable. Since consumers often withdraw money when they need it, the number of ATM withdrawal locations is a reasonable proxy for the number of cash purchase locations. Define purchase dispersion (disp) as the ratio of distinct ATM locations used to the total number of ATM withdrawals. The mean of disp is 0.5, indicating that the average customer visits each ATM twice in three months. If debit

transactions are correlated with impulse shopping (as evidenced by a high disp ratio) then we might see this as evidence that debit cards are being used to restrain impulse purchasing.

The second variable which might indicate someone trying to restrain spending is revolving a credit card balance – paying toward a balance rather than just paying for current period expenses. Two variables, Round and Same, indicate consumers who used their credit card not for payments but rather for paying toward a previously accumulated balance. A consumer paying a balance in full would pay the whole balance right down to the penny; whereas someone paying part of a high balance is likely to pay in round increments. Define round as the number of credit card payments made in multiples of $10.3 Two thirds of credit

3Regression results are robust to altering the definition of

(15)

card users made at least one round payment, and they made an average of three round payments in three months. Also, when paying down balances people likely pay the same amount each month, whereas those paying for recent purchases are very unlikely to have equal payments across months. Define same as an indicator for a consumer who made two credit card payments of identical amounts. One third of credit users made two identical payments. Finally, define a credit card revolver (revolver) as the union of round and same – one who either made two payments of the same amount or made more than one round payment. Nearly half (48.2%) of credit card users are revolvers, similar to the percentage of those who identified themselves as “revolvers” in the SCF.

The third characteristic which may be interpreted to indicate someone trying to restrain spending is carrying less cash. Such a consumer, ceteris paribus, makes lower value cash

Table 2: Empirical Variable Definitions

Debit number of debit card transaction in the 4 month sample Age age of account holder in years

Female indicator for account holder with a female given name Joint indicator for a joint account

Crime average crime rate at the ATM locations of used by account holder 1p indicator for only one pay source

Jobs number of paychecks from distinct sources Divorce indicates receiving a child support payment

Income average paycheck size * days between paychecks / 365 Paycycle days between paychecks

BalTr transactions balances (average of account peak minus account traugh) BalPr precautionary balances (average account traugh)

BalSav average savings account balance

NSFnow proportion of paycycles in which the account goes into a negative balance NSFhist Number of NSF written since account was opened

ATMuser indicator for ATM use

free ATM proportion of ATMs used which do not charge a fee

Credit indicates a payment made toward a credit card balance; see Appendix A Disp ratio of distinct ATM locations used to total ATM transactions

Round number of credit card payments which are multiples of $10

Same indicates customer making two credit card payments for the same amount Revolve household makes more than one round payment or same payments ATM visits number of ATM withdrawals

(16)

withdrawals but needs to visit the ATM more often. This does not imply that a customer withdraws more or less total cash, merely that cash acquisition is more diffuse, conditional on total cash withdrawn. Define the variable ATMcash as the average withdrawal amount as a fraction of total expenditure. Define ATMvisits as the number of ATM withdrawals. The average withdrawal was three percent of quarterly expenditure, and the average customer visited the ATM once a week.

4. Empirical Results

This section presents the results of tobit regressions of the number of debit transactions on the demographic, financial, and spending control variables discussed in section 2 and defined in section 3 above. Table 3 reports the regression which includes all of the variables discussed above. The dependant variable in the first column is the total number of debit card transactions, in the second column is the number of PIN debit transactions, and in the third column is the number of signature debit transactions. In the final column the dependant variable is the number of ATM transactions, which is presented for comparison purposes. Notice that 1033

observations can be utilized in these regressions. Variables calculated from ATM activity can not be calculated for a subset of the population – those account holders who do not use ATMs. Table 4 reports the same four regressions this time excluding any variable which is restricted to ATM users. In these regressions we are able to utilize 1483 observations. Income information is available only where paychecks are directly deposited. Thus if we remove any income related variables we can expand the data to include those who do not receive direct deposit into the observed checking account providing 2161 observations in table 5.

(17)

4.1 Demographic Variables

Our results show that younger individuals are more likely to use debit cards. The effect is strong and highly significant. According to the results of the regressions, the average 20 year old makes between 25 and 31 more debit card purchases than the average 60 year old. Most of the regressions do not reveal any relationship between gender and debit card use. The exception is the effect of total debit card transactions in table 5; although the two components of that total – PIN debit transactions and signature debit transactions – are not related to gender. Having a joint account increases the likelihood that a debit card is used to access the account.

The regressions provide no evidence that individuals are more likely to use debit cards if they frequent higher crime areas. By way of comparison, those who frequent high crime cities and zip codes, are more likely to conduct more ATM withdrawals, probably, so that they do not have to take out – or carry – as much cash at any given time.

Consider the variables that proxy for the opportunity cost of time. One variable in one regression (the interaction between joint and one paycheck in the PIN debit regression of table 4) shows a statistically significant relationship: that two-parent/one-income families conduct fewer debit purchases than others. However, table 3 shows no such relationship; neither signature debit nor total debit transactions bear any such relationship with two-parent/one-income households; and the other two proxies for opportunity cost of time (jobs and divorce) are not correlated with any kind of debit card use. The lack of a connection between debit use and the time value proxies could be due to a lack of connection between debit use and the opportunity cost of time or due to the variables used being poor proxies. Only 29 observations are receiving child support (the variable divorce). With respect to the interaction between joint and one paycheck (1p), the proxy’s lack of significance could be due to our inability to control for the number of children in the household.

(18)

4.2 Financial Variables

In the regressions reported in table 5, household income is positively related to debit card usage. However, in the presence of the expanded list of control variables contained in tables 3 and 4, this effect disappears and even turns negative for PIN debit reported in table 3. Most likely, the table 5 result is due to a scale effect (higher income households conduct more transactions of all types) since table 5 does not control for expenditure (BalTr). This lack of a relationship between debit transactions and income is surprising given that other electronic payments tend to be more common among higher income individuals as does new technology in

dep var debit

transactions transactionsPIN debit signature debittransactions transactionsATM Coef St. Err Coef St Err Coef St Err Coef St Err age female joint crime joint*1p jobs divorce incomea paycycle BalTra BalPra BalSava NSFnow NSFhist free ATM credit disp round same revolver ATMvisits ATMcash constant –0.617** –1.422 21.05** –2.341 –0.162 4.142 9.902 –0.052 –1.068** 0.617 –0.631 –0.259 –3.594** –0.020 –16.71** 5.065 19.08** –0.144 17.08* –9.503 0.571** –56.94* 62.23** 0.127 2.701 6.110 82.39 6.446 4.110 8.315 0.861 0.296 2.250 0.382 0.155 0.898 0.025 4.277 3.961 5.211 1.227 7.466 7.462 0.120 20.57 8.846 –0.026 1.200 2.294 37.54 –1.321 1.583 0.891 –0.090** –0.323** 1.380 0.157 0.100 –0.200 –0.010 –12.19** 0.301 9.636** –0.147 5.055 –4.686 0.232** –17.63* –2.311 0.043 0.905 2.043 27.71 2.160 1.400 2.808 0.032 0.107 0.796 0.127 0.067 0.303 0.008 1.650 1.346 1.785 0.409 2.647 2.643 0.040 8.929 3.042 –0.634** 0.635 18.87** –39.07 1.528 3.743 9.008 –0.015 –1.002** –0.061 –0.724 –0.199 –3.487** –0.014 –11.65** 5.015 15.49** 0.076 15.43* –8.303 0.431** –64.95** 61.59** 0.126 2.648 6.076 82.45 6.404 4.085 8.296 0.086 0.294 2.240 0.377 0.154 0.891 0.025 4.218 3.928 5.187 1.219 7.414 7.411 0.120 23.26 8.834 0.003 0.191 –0.170 18.28** 0.054 0.032 –0.026 .0079 –0.035 –.0048 0.025 –.0060 –0.162* –.0021 –2.819** 0.018 2.821** .0021 0.055 0.217 0.959** –1.536 –1.931** 0.010 0.212 0.488 6.493 0.514 0.328 0.669 .0068 0.023 0.178 0.030 0.012 0.072 0.002 0.340 0.315 0.413 0.089 0.596 0.595 0.010 1.743 0.706 σ 40.12 0.962 12.40 0.396 39.88 0.972 3.213 0.075

Notes: 1033 observations; * significant at 5%; ** significant at 1%.

a Figures are denominated in thousands of dollars.

(19)

general. These results are similar to Stavins’ (2001), Schuh and Stavins’ (2010), Simon, Smith, and West’s (2010) and Borzekowski and Kiser’s (2008) findings, but contradicts the Ameriks, Caplin and Leahy (2004) model in which the monitoring technology is decreasing in income. Mantel and McHugh (2001), however, find the opposite result, that debit usage increases with income. Carow and Staten (1999) find that debit usage is inverse “u”-shaped (highest among middle-income). Further, Fusaro (2010) points out that income is unrelated to overdrafting, another symptom of absent mindedness.

Pay cycle has a strong negative relationship with debit card use, a result that may be a proxy for education. The regression results from table 3 (table 4 respectively) indicates that someone paid monthly will conduct 17 (10.4 respectively) fewer debit card transactions than someone paid biweekly. Transactions balances – a measure of expenditure – is not related to debit use. Precautionary balances, though, are related to signature card debit and total debit transactions according to table 4. It is not related in the expanded regression which includes features of ATM use. Savings account balances are related to debit card use only in table 5 with the shortest variable list.

In table 3, contemporaneous overdrafts seem to be negatively related to debit use. This should not be surprising since debit transactions did not cause overdrafts at the time of data collection. Account holders who prefer debit cards, will therefore not overdraft as much as those who prefer writing checks, or those who intentionally overdraft will need to pay with a check. The count of past overdrafts is not related to debit use. These overdraft results hold true for table 4 as well. In table 5, though, the relationship reverses – contemporaneous overdrafts are not statistically significant, while past overdrafts are negatively related at the amount of debit use at the 5% level.

The variable freeATM is statistically significant and negatively related to debit use. For every six percentage points increase in the percentage of ATMs that do not surcharge the user, the average account holder conducts one fewer debit transaction. Or to put it another way, an

(20)

individual who uses only free ATMs will conduct 16.7 fewer debit transactions than someone who pays surcharges for all ATM withdrawals. This result may indicate a substitutability between cash and debit. Those who use debit more and cash less are not averse to paying a surcharge on occasion when they visit an ATM. Conversely, those who use cash more and debit less, know where the free ATMs are located and use them to avoid accumulating surcharges.

Finally, consider the relationship between credit use and the number of debit card

transactions. Some evidence suggests that credit and debit are substitute payment cards (Arango, Huynh, and Sabetti, 2011) which would lead us to expect a negative coefficient. However, some people are resistant to technology or wary of the identity theft potential from card use; this so called technology effect would imply a positive coefficient. The results in table 3 show that credit use is not related to the number of debit card transactions made. Potentially, the

dep var debit

transactions transactionsPIN debit signature debittransactions transactionsATM Coef St. Err Coef St Err Coef St Err Coef St Err age female joint joint*1p jobs divorce incomea paycycle BalTra BalPra BalSava NSFnow NSFhist ATM user credit round same revolver constant –0.770** –1.458 13.42* 1.241 6.889 9.012 0.070 –0.648** 0.386 –1.010** –0.194 –2.419** –0.036 49.51** 6.049 –0.272 22.55** –8.197 16.35* 0.116 2.520 5.533 5.835 6.446 8.549 0.075 0.861 1.510 0.361 0.166 0.918 0.032 3.211 3.909 1.263 7.617 7.628 7.380 –0.149** 0.920 0.042 –12.19** 2.148 –1.231 –0.022 –0.238** 0.731 –0.004 –0.082 0.049 –0.005 19.22** 0.872 –0.239 6.223* –4.656 –13.37** 0.042 0.897 1.989 2.100 1.314 2.906 0.026 0.091 0.488 0.118 0.073 0.309 0.008 1.516 1.349 0.421 2.698 2.700 2.831 –0.687** 0.684 12.04* 2.369 6.869 9.284 0.078 –0.581* 0.112 –1.070** –0.155 –2.435** –0.031 45.04** 5.231 0.018 19.65** –5.801 14.57* 0.115 2.491 5.465 5.765 3.539 8.455 0.075 0.238 1.530 0.359 0.164 0.906 0.023 3.184 3.871 1.247 7.513 7.528 7.305 –0.361 –2.050 –4.167 2.243** 2.435 –1.156 0.164** –0.372 –0.016 –0.401 0.012 1.246 –0.010* 1.646 0.066 0.908** 0.283 20.37 0.037 0.807 1.814 1.911 1.181 2.758 0.024 0.078 0.471 0.116 0.053 0.289 0.007 1.243 0.405 2.427 2.428 2.079 σ 43.97 1.000 12.40 0.396 43.33 1.002 13.90 0.334 Notes: 1483 observations; * significant at 5%; ** significant at 1%; a See Table 3 footnote.

(21)

substitutes effect and the technology effect offset each other causing the lack of a detectable effect in these data.

4.3 Variables Potentially Indicating Spending Restraint

As discussed in section II above, survey evidence shows that one of the motives for debit card use is as a tool of spending restraint. Although, this motive can not be tested directly in these data, some results might indicate that debit cards may be used by some as a tool of

spending restraint. This section reports results concerning the relationship between debit use and proxies for impulse shopping, the relationship between debit use and evidence of debt, and the relationship between debit use and the level of cash holding.

A proxy for impulse purchasing was discussed in section III above. The coefficient on this proxy, disp , is positive and significant at the 1% level for total debit transactions, PIN debit use and signature debit use. This result suggests that frequent debit card users shop at a greater variety of locations than those who use debit cards less often. The average customer has a disp ratio of 0.5 (they use the same ATM twice during the sample period). A customer with a disp ratio of 0.25 (uses the same ATM four times) will make 4.77 fewer debit card purchases than the average customer, possibly because that customer is more disciplined and less impulsive.

This estimate may be a biased downward. Since credit cards have higher average purchases than debit cards (Klee, 2006), it is possible that the cutoff level for switching from cash to a card is lower for those who prefer debit than for those who use credit rather than debit. Hence, debit users would use cash for fewer purchases than credit users would. If this is true, this effect could bias the coefficient on disp downward. The true relationship between debit card transactions and the disp ratio might be higher than 4.77 transactions for a 0.25 decrease in the ratio.

Alternatively, greater merchant acceptance for credit could explain why debit users use more (disperse) ATMs, i.e. merchant non-acceptance of debit cards causes debit users to use

(22)

cash more frequently (at more locations) than credit users do. However, this explanation is only valid for individuals who use PIN debit exclusively because merchant acceptance of credit and signature debit were identical, during the data sample period. Until 2004, Visa and MasterCard enforced the honor-all-cards rule, which stated that any merchant accepting Visa or MasterCard credit cards must likewise accept Visa or MasterCard debit cards, respectively. In fact, other types of debit were available in some locations where credit was not accepted.

This strong relationship between debit transactions and ATM diversity is an interesting empirical feature of debit card use, however, it is difficult to interpret. If one accepts the premise that diversity of ATM locations is related to diversity of cash purchase locations, and one

accepts the premise that diversity of cash purchase locations is a sign of impulse shopping, then this result could be seen as evidence of a positive relationship between impulse purchasing and debit card use which could be an attempt by those impulse purchasers to control their spending. This peculiar result indicates that further study of the possible connection between debit use, impulse buying, and purchase dispersion is warranted.

dep var debit

transactions transactionsPIN debit signature debittransactions transactionsATM Coef St. Err Coef St Err Coef St Err Coef St Err age female joint incomea BalSava NSFnow NSFhist ATM user divorce credit round same revolver constant –0.625** –4.342* 11.96** 0.282** –0.180** –1.398 –0.010* 52.10** 14.34 6.241 –0.176 20.48** –7.838 –0.332 0.082 2.041 2.106 0.044 0.144 0.804 0.019 2.529 8.247 3.270 1.146 6.727 6.785 4.725 –0.123** 1.379 0.726 0.034* –0.036** 0.169 –0.006 18.94** 0.359 1.556 –0.321 4.862* –3.582 –16.55** 0.030 0.718 0.742 0.015 0.056 0.265 0.006 1.176 2.757 1.110 0.380 2.317 2.337 1.867 –0.608** 3.505 11.42** 0.269** –0.156* –1.439 –0.036* 47.60** 14.23 5.316 0.049 18.61** –5.893 –0.549 0.081 2.012 2.075 0.043 0.141 0.791 0.018 2.497 8.131 3.229 1.128 6.622 6.680 4.662 –0.414** –2.377** –2.532** 0.187** –0.021 1.027** 0.005 0.734 2.204* 0.143 1.709 0.426 18.72** 0.027 0.673 0.694 0.014 0.044 0.260 0.006 2.739 1.071 0.379 2.203 2.218 1.285 σ 43.33 0.836 12.96 0.351 42.56 0.836 14.11 0.292 Notes: 2161 observations; * significant at 5%; ** significant at 1%; a See Table 3 footnotes.

(23)

The second indication of spending restraint discussed in section II is a connection between debit card transactions and debt – in particular credit card debt. Section III discussed two variables which may indicate credit card debt. The data allow us to observe some credit card use, however, they do not directly differentiate revolvers (those who carry credit card debt) from transactors (those who pay the full bill each month). In order to indicate revolvers we include two variables in the regressions: round and same as discussed in section III.

The results indicate a relationship between debit transactions and same but no

relationship between debit and round. This result is consistent across the tables and across the dependant variables. In addition, the variable revolve which is the union of the other two variables is not related to debit card transactions. However, when running a probit regression (not shown) where the dependant variable is an indicator of debit card use, the two variables round and revolve are positive and statistically significant. Using the Survey of Consumer Finances, Zinman (2009) found a similar connection between debit use and credit card revolving.

One interpretation of this connection – to which Zinman attributes this result – is that credit revolvers pay interest on credit card purchases, whereas credit transactors pay no interest. Thus a credit card revolver has an incentive to use another method of payment for purchases, including another credit card, cash, checks or a debit card. A second interpretation of this result is that the credit card debt is a sign of financial mismanagement – past overspending. If this is the case, then the debit card use could be a reaction to the financial mismanagement – a spending control device. It would take further investigation to discern whether one or both of these

explanations is behind this observed relationship between revolving and debit use.

Finally, consider the multiple methods that households use to limit liquidity as a method of spending restraint. Debit cards reduce liquidity; low cash holding also reduces liquidity. A correlation between debit use and low cash holding might further suggest that debit use, like low cash holding, is related to spending restraint. ATMvisits and ATMcash are significant at the 1%

(24)

and 5% levels, respectively. This result suggests that debit users carry less cash (i.e., withdraw more often but less each time; ATM visits>0, ATM cash<0). It is quite striking that debit users have more frequent withdrawals since cash is a substitute for debit.

To see the magnitude of this difference consider an example. The average number of ATM withdrawals is 13.5 and the average size of an ATM withdrawal is 3.6% of expenditure for a total of 48.6% of expenditure in ATM withdrawals. Consider two examples, a person with 25.6 withdrawals, each 1.9% of expenditure (for a total of 48.6% of expenditure) and a person with 4.6 withdrawals, each 11.3% of expenditure (also for a total of 48.6% of expenditure). These two examples are chosen to represent an individual who is one standard deviation above the mean and an individual who is one standard deviation below the mean. The table 3

regression results indicate that the former individual will make 17.5 more debit transactions than the later person.

Alternatively, households might carry less cash due to liquidity constraints. Debit card users make all payments out of liquidity. Therefore, it is unlikely that debit card users also make small ATM withdrawals due to a lack of liquidity. Further, consider that debit users can get cash from the cashier at a supermarket, an option that non-debit users do not have. The measured coefficient on ATM withdrawal reflects only ATM withdrawals, not cash back from a

supermarket cashier. Therefore some of the demand for cash among debit users is diminished due to the substitute cash source. This availability would bias the coefficient on ATM visits downward. Quite possibly, the true relationship between debit use and cash holding is even stronger than table 3 reflects.

Taken together, the results of this section (a correlation between debit card transactions and ATM dispersion, a correlation between debit card transactions and repeat credit card

payments, and a correlation between debit card transactions and carrying small amounts of cash) are consistent with the spending control motive for debit use. However, the fact that those with longer pay periods are less likely to use debit runs contrary to the spending control motive.

(25)

5. Comparison With Survey Evidence

5.1 Demographic Variables

Our results concerning age are consistent with US survey evidence (e.g., Carow and Staten, 1999, Stavins, 2001, Klee, 2006, and Schuh and Stavin, 2010), Canadian survey evidence (Arango, Huynh, and Sabetti, 2011), and Australian survey evidence (Simon, Smith, and West, 2010). Consistent with Carow and Staten, (1999), Klee (2006) and Schuh and Stavin (2010), the transaction data does not reveal any relationship between gender and debit card use. This result, however contradicts Arango, Huynh, and Sabetti (2011) whose Canadian respondents reported higher debit use among women.

Married individuals reported to Carow and Staten (1999) and to Schuh and Stavin (2010) that they are less likely to use debit than non married survey respondents (although Klee, 2006, and Arango, Huynh, and Sabetti, 2011, found no such statistical relationship). In our data, having a joint account – a proxy for marriage – increases the likelihood that a debit card is used to access that account. The difference could lie simply in the different units of observation – the individual for the survey, and the account for the checking account data. Consider a simple example where all women use a debit card, all married men do not use a debit card, and unmarried men split 50/50. Thus 50% of married people use a debit card while 75% of unmarried people use a debit card, matching the survey results. However, when looking at accounts, all joint accounts (married) show debit use (because the women are doing so), while only 75% of individual accounts (all unmarried women and half of the unmarried men) use a debit card.

The account results find no relationship between debit card use and proximity to crime, but find a positive relationship between ATM use and crime. This contradicts Schuh and Stavin

(26)

(2010) who found that those who value “safety” are significantly realated to debit card use. However, the parallel is not exact; survey respondents likely interpret the term “safety” to mean more than just protection from street crime. Indeed, their result matches that of Arango, Huynh, and Sabetti (2011) who used the term “fraud” instead of “safety” and got a similar result.

Our proxies for opportunity cost of time show very little evidence of a connection with debit card use which is consistent with Arango, Huynh and Sabetti’s (2011) finding a lack of effect among those who value “ease”. However, both Schuh and Stavins (2010) and

Borzekowski and Kiser (2008) find that debit cards are used by those who value “speed” and “convenience”. And Arango, Huynh, and Sabetti (2011) contradict all of the above finding that valuing “speed” has a negative relationship to debit card use. Our failure to find a relationship between opportunity cost of time proxies and debit use may contradict their results or may be simply do to our measure being poor proxies for opportunity cost of time.

In the account data, pay cycle has a strong negative relationship with debit card use, a result that may be a proxy for education. This result is new as Carow and Staten (1999) find a positive relationship between debit use and education while Schuh and Stavins (2010) find no relationship between debit use and level of education in their survey.

5.2 Financial Variables

Concerning income, our mixed results do help clarify a very mixed picture gleaned from survey evidence. Schuh and Stavins (2010), and Mantel and McHugh (2001) find a positive relationship between income and debit use. Stavins (2001), and Borzekowski and Kiser (2008) find no relationship. Carow and Staten (1999) and Klee (2006) find that debit usage is inverse “u”-shaped (highest among middle-income). Simon, Smith, and West’s (2010) Australian

respondents show some resemblance to an inverse “u”-shape also. Arango, Huynh, and Sabetti’s (2011) Canadian respondents show no income relationship except a mild negative effect above $80,000.

(27)

Zinman (2009) and Arango, Huynh, and Sabetti (2011) find credit use to be a substitute for debit. In particular as the price of credit drops, quantity of credit use increase while quantity of debit falls. Our account data show no such tradeoff between credit and debit.

5.3 Variables Potentially Indicating Spending Restraint

As noted above, the account data show some signs that might be consistent with the spending control motive for debit use. In the Survey of Consumer Finances, Zinman (2009) found 31% of debit users to be ripe for potential spending control explanations. Borzekowski, Kiser and Ahmed (2008) reports that 5.8% of debit card users report using debit cards for spending restraint. However, since they report results of an open ended survey, their figure represents a lower bound. When we restrict attention only to those who reported their preference for debit over credit, the proportion is much higher – 23.5%.

Spanish consumers agreed with the statement “Payment cards offer control of domestic spending.” at a rate of only 1.53 on a 5 point Likert (Carbo-Valverde and Linares-Zegarra (2011). This is consistent with other survey evidence that only a minority of people see debit as a tool of spending restraint. These survey respondents also indicate that, among those who have a debit card but not a credit card, those who answer this question in the affirmative are less likely to use a payment card. However, when credit card holders are included, this variable becomes insignificant.

Canadian survey respondents report (Arango, Huynh and Sabetti, 2011) that “fear of overspending” has a negative effect on both debit card use and credit card use. The effect is much stronger for credit card use. Thus, credit carries the most risk of overspending, followed by debit, and cash carries the lest risk of overspending.

Taken together, survey results on the spending restraint motives for debit use range from 6% to 30%. This wide range may be due to many factors such as respondents’ unwillingness to admit to be in need of spending restraint, and the fact that neither of these surveys were

(28)

specifically intended to address the spending control motive for debt use.

6. Conclusion

The literature (e.g. Borzekowski, Kiser and Ahmed, 2008; Stavins, 2001; Zinman, 2004) posits several motivations for debit card use: the inability to qualify for a credit card,

convenience, time at checkout, merchant acceptance, record keeping, rewards, desire for cash back at checkout, safety, and spending restraint. A data set containing transaction records for 2310 checking accounts including debit card purchases and credit card payments is employed to evaluate the motives for using debit cards. If debit users were credit constrained, then we would expect a negative coefficient on credit, round, same, and revolve, none of which is the case in any of the regressions, although several are negative but not statistically significant. Since credit constraints implies a negative coefficient on these variables and spending restraint implies a positive coefficient on these variables, it is possible that each explanation is true for a group of consumers. The two effects could offset each other leading to the results we see here.

Maybe people use debit because it is more convenient than its alternatives. If debit users seek convenience then it is certainly curious that they do not minimize their trips to the ATM, but rather maximize them (reference the results on ATMcash and ATMvisits). We may also compare the two forms of debit. While they provide nearly identical convenience, the one difference is that PIN debit is easier to use at the checkout, however this is the much less commonly used (45 PIN debit transactions vs 8 signature debit transactions). Likewise, signature debit is more popular than PIN debit despite time-at-checkout being lower for PIN debit than for signature debit. The explanation for signature being more common is likely merchant acceptance, which is higher for signature debit. But merchant acceptance can not explain much else. As noted above, merchant acceptance was identical for credit and signature

(29)

debit.

Debit cards and credit cards provide similar record keeping. However, debit provides a distinct record keeping advantage over cash or check. The positive relationship between a joint account and debit use may be due to an increased need for centralized record keeping when two people are using the account. Debit cards do not offer rewards as do credit cards. A benefit of debit which was developed by other banks at a time subsequent to our data collection is the forced savings programs. The lack of a correlation between saving balances and debit card use, provide little guidance on the potential market for forced savings products.

As noted above, the positive coefficient on ATMvisits disputes the argument that people use a debit card in order to get cash back and avoid ATMs. The lack of a relationship to higher crime areas dispels the theory that debit cards are carried because they are thought to be safer than cash or other forms of payment. While results such as the coefficients on ATM dispersion, repeat credit card payments, and cash holding suggest a possible spending restraint motive for debit use, they are by no means conclusive.

One much noted predictor of payment method is purchase size. Large purchases are more likely to be credit transactions. Larger purchases, however, are more likely to be one-time purchases; thus, are better candidates for intentional borrowing. This work abstracts away from issues of intentional borrowing or purchase size. Rather, we look at the data aggregated to the household level.

While our results do not confirm the existence of debit cards as a tool of spending control, they do leave the possibility open. The logical implication of consumers spending less with debit cards is that they spend more with credit cards. This fact is defended in stark terms by retail, banking, and credit card executives. However, academic work has not yet sorted out whether people spend more when they have credit cards, or people choose to use credit cards when they are spending more. If the former is the case, then businesses are justified in accepting the high interchange fees of credit cards. If the latter is true, then businesses have a much greater

(30)

ability to steer purchases toward less costly methods of payment. It will be the task of future work to sort that out.

7. Appendix

Transaction Descriptions which denote credit card payment: AMERICAN EXPRESS

BANK ONE IC PAYMENT BANKCARD

BANKCARD PAYMENT BK OF AM CRD ACH BK OF AMER VI/MC

CAPITAL ONE CRCARDPMT CAPITAL ONE ONLINE PMT CAPITAL ONE PHONE PYMT CHASE CARD SERV PAYMENT CHASE CREDIT CAR BILL PAY CHASE MANHATTAN ONLINE PMT CHASE MANHATTAN RE PAYMENT CITI-CLICK 2 PAYCITIBANK CRDT CD ONLINE PMT

CITIBANK-AUTOPAY PAYMENT CREDIT CARD

CROSS COUNTRY BA CC PAYMENT DISCOVER

FIRST NATIONAL PAYMENT FIRST PREMIER BK

FIRST USA CARD

FIRSTCONSUMERS CARD PMT

FLEET CREDIT CRD GM CARD

HOUSEHOLD CREDIT ONLINE PMT JUNIPER BANK CREDITCARD MBNA AMERICA

NATIONAL CITY PAYMENT NEXTCARD

PAY CARD SERVICE PAY-BY-PHONE PYM PEOPLES BANK PREMIER CR CARD

PROVIDIAN PAYMT CREDITCARD RNB - TARGET BILL PAY

SAM'S CLUB ONLINE PMTSEARS ROEBUCK PHONE PMT

SEARS WEB PAY TARGET VISA UCS - AUTOPAY UCS - CLICK TO PAY" VISA

WELLS FARGO

WELLS FARGO CARD WELLSFARGO CARD

8. References

Ameriks, J., A. Caplin, J. Leahy, and T. Tyler. “Measuring Self-Control Problems.” American Economic Review, 97(3), 2007, 966-72.

(31)

Ameriks, J., A. Caplin, and J. Leahy. “The Absent-Minded Consumer.” National Bureau of Economic Research Working Paper No. 10216, 2004.

Arango, C., K. Huynh, L. Sabetti. “How Do You Pay? The Role of Incentives at the Point-of-Sale”, Working Paper 2011-23, Bank of Canada, 2011.

Ashraf, N., D. Karlan, and W. Yin. “Tying Odysseus to the Mast: Evidence from a Commitment Savings Product in the Philippines.” The Quarterly Journal of Economics, 121(2), 2006, 635-72.

Ausubel, L. M. “The Failure of Competition in the Credit Card Market.” American Economic Review, 81(1), 1991, 50-81.

Bertaut, C. C., and M. Haliassos. “Debt Revolvers for Self Control.” Working Papers in Economics No. 0208, University of Cyprus, 2002.

Bolt, Wilko, Nicole Jonker, and Corry Von Renselaar, “Incentives at the Counter: Empirical Analysis of Surcharging Card Payments and Payment Behaviour in the Netherlands”, Journal of Banking and Finance, 34(8), 2010, 1738-1744.

Borzekowski, R., and E. Kiser. “The Choice at the Checkout: Quantifying Demand Across Pay-ment InstruPay-ments.” International Journal of Industrial Organization, 26(4), 2008, 889-902. Borzekowski, R., E. Kiser and S. Ahmed. “Consumers’ Use of Debit Cards: Patterns,

Prefer-ences, and Price Response.” Journal of Money Credit and Banking, 40(1), 2008, 149-72. Buddle, S. L., ed. Card Industry Directory: The Blue Book of the Credit and Debit Card Industry

in North America. 19th. ed. New York: SourceMedia, 2007.

Carbo-Valverde, Santiago and Jose M. Linares-Zegarra, “How Effective Are Rewards Programs in Promoting Payment Card Usage? Empirical Evidence”, Journal of Banking and Finance, 35(12), 2011, 3275-3291.

Carow, K. A., and M. E. Staten. “Debit, Credit, and Cash: Survey Evidence on Gasoline Prices.” Journal of Economics and Business, 51(5), 1999, 409-22.

Ching, Andrew T. and Fumiko Hayashi, “Payment Card Rewards Programs and Consumer Payment Choice”, Journal of Banking and Finance, 34(8), 2010, 1773-1787.

Fusaro, Marc Anthony, “Debit vs Credit: A Model of Self-Control with Evidence from Checking Accounts” mimeo, Arkansas Tech University, April 2008.

Fusaro, Marc Anthony, “Are Bounced Check Loans Really Loans? Theory, Evidence and Policy” Quarterly Review of Economics and Finance, 50(4), 2010, 492-500.

Klee, E. “Paper or Plastic? The Effect of Time on the Use of Check and Debit Cards at Grocery Stores.” FEDS papers No. 2006-2, Federal Reserve Board, 2006.

(32)

Kocherlakota, N. R. “Money is Memory.” Journal of Economic Theory, 81(2), 1998, 232-51. Laibson, D., A. Repetto and J. Tobachman. “Estimating Discount Functions from Lifestyle

Consumption Choices.” Manuscript, Harvard University, 2007.

Mantel, B. and T. McHugh. “Competition and Innovation in the Consumer e-Payments Market? Considering Demand, Supply, and Public Policy Issues.” Emerging Payments working paper No. EPS-2001-4, Federal Reserve Bank of Chicago, 2001.

Moebs $ervices Inc. Survey of Retail Fees of Banks and Credit Unions. Lake Bluff IL, 2005. Moebs $ervices Inc. Survey of Retail Fees of Banks and Credit Unions. Lake Bluff IL, 2006. Prelec, D., and G. Loewenstein. “The Red and the Black: Mental Accounting of Savings and

Debt.” Marketing Science, 17(1), 1998, 4-28.

Prelec, D., and D. Simester. “Always Leave Home Without It: A Further Investigation of the Credit Card Effect on Willingness to Pay.” Marketing Letters, 12(1), 2001, 5-12.

Rochet, Jean-Charles and Julian Wright, “Credit Card Interchange Fees”, Journal of Banking and Finance, 34(8), 2010, 1788-1797.

Schuh, Scott and Joanna Stavins, “Why are (Some) Consumers (Finally) Writing Fewer Checks? The Role of Payment Characteristics”, Journal of Banking and Finance, 34(8), 2010, 1745-1758.

Simon, John, Kylie Smith, and Tim West, “Price Incentives and Consumer Payment Behaviour”, Journal of Banking and Finance, 34(8), 2010, 1759-1772.

Stavins, J. “Effect of Consumer Characteristics on the Use of Payment Instruments.” New England Economic Review, 20(3), 2001, 19-31.

Thaler, R. H. and H. M. Shefrin. “An Economic Theory of Self Control.” Journal of Political Economy, 89(2), 1981, 392-406.

Figure

Table 1: Summary Statistics
Table 2: Empirical Variable Definitions
Table 3: Tobit Regressions – All Variables
Table 4: Tobit Regressions: Sample Expanded to Include ATM Non-Users
+2

References

Related documents

First, hypermarket shoppers shop for personal satisfactions obtained from shopping such as the opportunity to enact a culturally prescribed role; diversion from daily

This exam- ple demonstrates position compensation during motion of a slave axis coupled to a master. The procedure for saving and installing the example program is similar to the

It can also happen when you do not properly dispose of documents like credit card bills, debit card statements, or receipts showing your personal account information.. You

If, for example, you earn a commission from the hotel and do not charge a booking fee, or similar, to the customer, the card fee is standard‐rated, according to HMRC, as it is

The next step in setting up credit and debit card processing using the Payment Terminal Setup Wizard is to ensure pay method codes are properly created and ready for use (Step 7 of

A Merchant account makes it possible for a business to accept a credit card or debit card as a form of payment.. All transactions made with a debit or credit card involve the

&#34;I/We agree and acknowledge that usage of the Debit card, Debit card Personal Identification Number (PIN), Phonebanking Personal Identification Number (PIN), Credit Card,

For any subsequent payments from a previously used card, you need only select the card in question from the “Card” drop-down, specify the To account, the Currency and the Amount –