Consumer preferences for payment methods: Role of discounts and surcharges

Accepted Manuscript

Consumer Preferences for Payment Methods: Role of Discounts and Surcharges Joanna Stavins PII: DOI: Reference:

S0378-4266(18)30139-0 10.1016/j.jbankfin.2018.06.013 JBF 5373

To appear in:

Journal of Banking and Finance

Received date: Revised date: Accepted date:

4 August 2017 7 June 2018 20 June 2018

Please cite this article as: Joanna Stavins , Consumer Preferences for Payment Methods: Role of Discounts and Surcharges, Journal of Banking and Finance (2018), doi: 10.1016/j.jbankfin.2018.06.013

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Consumer Preferences for Payment Methods: Role of Discounts and Surcharges

CR IP T

Joanna Stavins

ED

M

AN US

Abstract: We use new data from the 2015 Diary of Consumer Payment Choice to analyze the role of consumer preferences for specific payment instruments, and how price discounts and surcharges based on payment method affect payment instrument choice. We test whether consumer demand for payment instruments is price elastic, namely whether consumers are likely to deviate from their preferred methods in order to get a discount or to avoid a surcharge. We find that the occurrence of price incentives is low, but that given cash discounts, the probability that a cash transaction is conducted by a consumer who prefers other payment methods increased by 19.2 percent, after controlling for merchant category and dollar value of the transaction. Payment method choice is affected very strongly by consumer individual preferences, but steering by merchants may be effective under some circumstances. Both merchants’ reluctance to offer price discounts and consumers’ limited response to them lead to the low observed occurrences of such incentives.

PT

Keywords: Consumer payments, consumer preferences, merchant steering, discounts, surcharges

CE

JEL classifications: D03, D14, G02

AC

Joanna Stavins is a Senior Economist and Policy Advisor in the Research Department of the Federal Reserve Bank of Boston. Contact information: [email protected], Research Department, Federal Reserve Bank of Boston, 600 Atlantic Avenue, Boston, MA 02210, 617-973-4217. I am grateful to Scott Schuh and Bob Triest for helpful comments and to Allison Cole, Emily Wu, and Liang Zhang for excellent research assistance. This paper, which may be revised, is available on the web site of the Federal Reserve Bank of Boston at http://www.bostonfed.org/economic/wp/index.htm. The views expressed in this paper are those of the authors and do not necessarily represent the views of the Federal Reserve Bank of Boston or the Federal Reserve System.

AC

CE

PT

ED

M

AN US

CR IP T

ACCEPTED MANUSCRIPT

2

ACCEPTED MANUSCRIPT

I.

Introduction

Payment methods vary in terms of their real and perceived cost and benefits—including the cost of issuing, the cost of conducting transactions for each party involved, the cost of time, security, convenience, among others. Although there is no consensus on which payment methods are least

CR IP T

and most costly, those costs have generated substantial controversy, as illustrated by the longstanding debate on the credit card interchange fees, which constitute just one type of cost of conducting credit card transactions. Despite the differences in cost among payment methods, there are almost no differences in prices faced by consumers. Therefore the effect of price incentives on the use of payment methods is an important economic issue.

AN US

Shy and Stavins (2015) showed that U.S. merchants rarely took advantage of their recent freedom to differentiate prices based on the method of payment in 2012. Using data from the 2015 Diary of Consumer Payment Choice (DCPC), we find that the prevalence of discounts and surcharges based on payment method did not increase from 2012 to 2015. However, a small number of surcharges on credit card transactions and discounts on cash or debit card transactions

M

were recorded by the survey respondents. In this paper, we take advantage of detailed information on individual transactions and on payment preferences of a representative sample of

ED

U.S. consumers and analyze consumer preferences and their behavior in response to such pricing incentives.

PT

We find that consumer preferences are correlated with demographic and income attributes, but that they also vary by the value and type of transaction. We look closely at who

CE

received a discount or paid a surcharge, which types of merchants utilized such pricing incentives, and how the price incentives varied by dollar values of transactions. Finally, we

AC

examine the role of consumer preferences with respect to payment methods and test whether consumers are likely to shift away from their preferred payment method in order to receive a discount or avoid a surcharge. In other words, we test whether consumer demand for payment instruments is elastic with respect to the differentiated cost of making a payment. As was the case in 2012, the occurrence of price incentives based on the payment method continued to be low in 2015. For the majority of transactions, consumers tend to use their preferred payment instruments. They deviate from their preferred payment method choices 1

ACCEPTED MANUSCRIPT

mainly because of the dollar values of transactions—for example, by switching to cash for lowvalue transactions—or because of the merchant type. For consumers who prefer to use cash, they typically shift to another payment method because of insufficient cash on person to conduct a cash transaction. However, we found that consumers are significantly likely to switch to using cash specifically because of cash discounts offered, even after controlling for transaction value

CR IP T

and merchant type: given cash discounts, the probability that a cash transaction is conducted by a consumer who prefers noncash payments increases by 19.2 percent, after controlling for merchant category and dollar value of the transaction. As the amount of cash discount is typically low, the result indicates that consumers’ demand is highly elastic with respect to discounts.1

AN US

The low occurrence of price incentives might be caused by merchants’ reluctance to offer them. Although merchants have been allowed to impose credit card surcharges in order to recover their credit card processing costs since January 27, 2013, they must follow certain rules, such as the following: merchants must notify consumers that they are being charged for using credit cards, both at the register and on the receipt; merchants are allowed to pass on their credit

M

card merchant fees, but are not allowed to make a profit on credit card surcharges; and surcharges are limited to credit card transactions (merchants cannot surcharge debit card

ED

transactions). Because the rules may be difficult to follow, most merchants—including gasoline

PT

station proprietors—prefer to provide cash discounts rather than credit card surcharges. Laws in ten states—California, Colorado, Connecticut, Florida, Kansas, Maine,

CE

Massachusetts, New York, Oklahoma and Texas—prohibit merchants from imposing surcharges on credit card transactions. Merchants are allowed to offer cash discounts, even if the resulting price differences between cash and credit card transactions are the same. In March 2017, the

AC

Supreme Court ruled that telling merchants how to communicate their prices violates the First

1

For example, the difference between the cash and credit prices at gasoline stations is typically about 5 cents per gallon, according to the National Association of Convenience Stores: “Amounts of the discount vary, but most retailers offer approximately 5 cents off per gallon to customers paying by cash.” (www.nacsonline.com)

2

ACCEPTED MANUSCRIPT

Amendment.2 The case is part of a long-running dispute between merchants and credit card companies, and is likely not the final word on the topic. In addition to the legal and regulatory restrictions associated with imposing surcharges, merchants may be reluctant to do so in expectation of losing potential customers or receiving bad publicity. Even offering cash discounts may not protect merchants from criticism. For example,

CR IP T

Consumer Reports warned their readers: “Don’t be tricked by gas station cash discounts. Not paying attention could leave you paying more at the pump” (Consumer Reports, August 9, 2013).

Our dataset is uniquely suitable for analyzing consumers’ response to price incentives.

AN US

We use transaction-level data that also include consumer-specific (demographics and income) and payee-specific (merchant type) information, all needed to conduct the analysis. It is important to note that what we observe is an intersection of supply and demand—for either a discount or a surcharge to be observed in the data, a merchant must offer the price incentive and the consumer must conduct the transaction. Thus, we do not observe cases where consumers

M

ignored the offer of a discount or avoided paying a surcharge because they used a different

ED

payment method.

We compared the pace of credit card surcharging in the United States with that in Australia. Starting in 2003, the Reserve Bank of Australia (RBA) removed the “non-surcharge

PT

rule” and merchants were allowed to impose surcharges on credit card transactions there. 3 Australian merchants were slow and reluctant to take up surcharging, perhaps due to fear of

CE

public backlash and reputation damage. By 2006, only 7 percent of merchants were surcharging. However, the incidence of merchant surcharging increased quickly after that: in 2007, RBA

AC

reported that 15 percent of very large merchants, 9 percent of large merchants, 6 percent of small merchants, and 5 percent of very small merchants surcharged credit card transactions. By 2010— or 7 years after the change in regulation—20 percent of small merchants and 40 percent of large merchants had started to surcharge. It is possible that the number of merchants who surcharge 2

Adam Liptak, “Justices Side With Free-Speech Challenge to Credit Card Fees,” The New York Times, March 29, 2017. https://www.nytimes.com/2017/03/29/business/supreme-court-credit-card-fees.html 3

For more details, see Reserve Bank of Australia (2011) and http://www.fairtrading.nsw.gov.au/biz_res/ftweb/pdfs/About_us/Credit_card_surcharges_part1.pdf.

3

ACCEPTED MANUSCRIPT

credit card transactions will rise in the U.S. as well. However, the fraction of transactions surcharged in Australia was substantially smaller than the fraction of merchants who surcharge, as consumers avoided using credit cards where surcharges were imposed. The rest of the paper is as follows. Section II summarizes the relevant literature and shows how our paper contributes to the existing literature. Section III describes the data. Section

CR IP T

IV discusses consumers’ payment method preferences, including the use of consumers’ preferred methods and the probability of deviating from those preferred methods. Section V focuses on surcharges and discounts, and Section VI estimates the effect of price incentives on the probability of deviating from a preferred payment method. Section VII concludes.

Literature review

AN US

II.

Our study adds to a growing literature on the effect of price incentives on consumer payment choice. In particular, existing papers focus primarily on three different mechanisms: reward or loyalty programs, bank-imposed fees, and merchant-imposed surcharges and discounts.

M

There exists evidence that reward programs can steer consumers toward greater card use. Ching and Hayashi (2010) find that payment card rewards have statistically significant effects on

ED

consumer choice of payment methods. Agarwal, Chakravorti, and Lunn (2010) use actual credit card transaction data from a large U.S. financial institution and find that cash-back rewards lead

PT

to an increase in credit card spending and debt. Using Australian transaction-level diary data, Simon, Smith, and West (2010) note that consumers who participate in loyalty programs and

CE

have access to an interest-free period tend to use credit cards more. Arango, Hyunh, and Sabetti (2015) focus on how reward program incentives and merchant acceptance affect consumer

AC

payment choice.

Some earlier papers also note that bank-imposed fees may negatively affect the use of

debit cards. Humphrey, Kim, and Vale (2001) measure price elasticities of ATM cash withdrawals, checks, and debit cards at the point of sale using aggregate Norwegian data. Borzekowski, Kiser, and Ahmed (2008) study the relationship between an increase in bankimposed fees and a decline in debit card use at the point of sale.

4

ACCEPTED MANUSCRIPT

More closely related to the context of this paper are studies that consider merchantimposed discounts and surcharges. Barron, Staten, and Umbeck (1992) look at trends in cash discounting in retail gasoline during the 1980s. Amromin, Jankowski, and Porter (2007) utilize data on toll payments on the Illinois Tollway and discover that consumers rapidly switched to electronic toll payments when toll fees doubled for cash payers. Using survey data from the

CR IP T

Netherlands, Bolt, Jonker, and van Renselaar (2010) show that surcharges on debit card transactions do steer Dutch consumers away from debit cards to cash. Briglevics and Shy (2014) discuss potential theoretical reasons for the low number of observed instances of merchant steering via cash and debit card discounts.

Our dataset is uniquely suitable for analyzing consumers’ response to price incentives.

AN US

Similar to Klee (2008) and Wang and Wolman (2014), we have transaction-level data with the transaction value, merchant type, and payment method used, but we also have detailed demographic and income data on every consumer, the information that Klee (2008) lacked. Cohen and Rysman (2013) used scanner data with demographic information, but could not separate debit cards from credit cards, and had no information on price incentives. O’Brien

M

(2014) shows that consumers’ stated payment instrument preferences are a very strong predictor of consumers’ payment behavior, but his analysis is limited to cash. As far as we know, this is

ED

the best transaction-level dataset that also includes consumer-specific and payee-specific information, all needed to conduct this kind of analysis.

PT

Agarwal and Qian (2014) and Agarwal, Koo, and Qian (2017) found that liquidity constraints affect the choice of payment instruments by consumers in Singapore. In particular,

CE

they found that consumers switched from credit card to debit card because of liquidity constraints. Our data allows us to control for whether a consumer had a credit card or a debit

AC

card with him at the time of each transaction, and also for whether the consumer had enough cash with him at the time of each transaction to pay for this particular transaction. Therefore we can control for liquidity constraints. Shy and Stavins (2015) analyze data from an earlier version of the Diary of Consumer Payment Choice to investigate whether U.S. merchants are actively steering consumers to certain payment methods post recent legislative changes. Our current paper not only provides an extension to previous work on the effect of discounting and surcharging, but also incorporates 5

ACCEPTED MANUSCRIPT

information from self-reported consumer preferences for certain payment instruments. We test whether consumers deviate from their preferred payment instrument in response to price incentives. This study brings new insights that—to our knowledge—have not yet been fully explored in the existing literature.

Data

CR IP T

III.

The data used in this paper are from two consumer surveys: the Survey of Consumer Payment Choice (SCPC) and the Diary of Consumer Payment Choice (DCPC), both conducted by the Federal Reserve Bank of Boston in the fall of 2015. The SCPC is an annual survey of a representative sample of U.S. adult consumers that collects data on payment instrument

AN US

adoption, use, and assessments of characteristics of major payment instruments: cash, check, money order, debit cards, credit cards, prepaid cards, bank account number payments, and online banking bill payments. Moreover, the SCPC gathers detailed information about the respondents’ bank, cash, virtual currency, credit, prepaid, and non-bank account holdings, as well as

M

demographic and income data. The 2015 SCPC includes responses from 2,040 adults. As a complement to the SCPC, the 2015 DCPC collected detailed transaction-level data

ED

from the same set of respondents. Each respondent filled out the diary for three consecutive specified days. For each transaction conducted during a specified three-day period, respondents recorded dollar value, type of merchant, type of transaction, and the payment instrument used.

PT

Transactions included all retail purchases4 conducted in person and online, bill payments, and person-to-person payments. As in Shy and Stavins (2015), our analysis is limited to non-bill

CE

purchases, because the discount and surcharge questions were asked only in reference to the purchases. The 2015 DCPC includes 6,123 individual purchases5 across 1,585 diaries by

AC

1,076unique respondents. (The reason the number of diaries is greater than the number of respondents is because 509 diarists filled the diary twice, as that was the best way to reach the desired number of diaries and also allowed us to test whether consumers conduct more 4

Purchase transactions include expenditures and account-to-account transfers.

5

The number of transactions refers to the number of purchases. Including bill payments, there were 11,503 transactions in the 2015 DCPC. For consistency, we exclude respondents from an additional survey conducted by another vendor.

6

ACCEPTED MANUSCRIPT

transactions closer to the holiday period.) Excluding extreme values, the median purchase amount was $16.93, and the mean was $33.76 We do not use data from the 2012 DCPC, because that earlier survey was conducted by another vendor and the questionnaire was changed for the 2015 DCPC. In particular, the 2015 DCPC asked for payment preferences for bills and purchases separately, while the 2012 DCPC

CR IP T

asked for overall preferences only. Respondents typically vary their preferred payment methods by type of transaction. Moreover, the 2012 DCPC did not ask follow-up questions as to why the respondent deviated from his or her preferred payment method, while the 2015 DCPC did ask the question. The analysis may be reviewed using the 2016 diary data.

AN US

On the first day of each diary, respondents were asked to provide their payment instrument preferences for purchases and bills, and for in-person and online purchases. For all purchases paid with cash, debit cards, or credit cards, they were asked whether they received a discount or a surcharge because of the payment instrument they used for the transaction. Respondents answered the following questions about each purchase they made:

M

For cash transactions:

ED

Question: Did you receive a discount from the merchant specifically for using cash? [Yes/No] For debit card transactions:

PT

Question: Did you receive a discount from the merchant specifically for using this debit card?

CE

[Yes/No]

For credit card transactions:

AC

Question: Did you pay an extra charge, surcharge, or convenience fee to the merchant specifically for using this credit card? [Yes/No]

Because all DCPC respondents also answered the SCPC, we merged the SCPC data and

added many other variables to the DCPC data using unique respondent identifiers: demographics, household income, and assessments of payment instrument characteristics.

7

ACCEPTED MANUSCRIPT

Table 1 shows the summary statistics for the 2015 DCPC by demographic and income cohorts, both in terms of the number of respondents and the percentage within each variable. For example, looking at the breakdown by income, 22 percent of the sample is in households with annual income below $25K, 23 percent between $25K and $49K, and 3 percent in households

CR IP T

with annual income above $200K.

IV. Payment method preferences

To assess the effect of price incentives on payment choice, we start with an analysis of consumer preferences concerning payment methods. Before the diary period begins, each respondent is asked which payment method he or she most prefers to use for purchases, separately for in-

AN US

person and online transactions (respondents are also asked which payment method they prefer for bills). We assume that the act of specifying the preferred payment method does not affect the respondents’ payment behavior during the subsequent three-day diary period. A.

How do consumers prefer to pay?

M

The survey results show that consumers prefer to use different payment methods for inperson versus online purchases, for bills versus purchases, and small versus large dollar value

ED

transactions. Figures A1 and A2 in Appendix A show the distribution of preferred payment methods. Figure A1 shows preferred payment methods for in-person purchases, broken down by

PT

transaction value, and, separately, for online purchases. The figure shows the percentage of respondents who state that they prefer to use a given payment instrument for transactions of a

CE

given value. Note that as the value of a transaction rises, consumers become less likely to prefer cash and more likely to prefer credit or debit cards for in-person payments. For example, for transactions above $100, 40 percent of respondents prefer to use credit cards, and 39 percent of

AC

respondents prefer to use debit cards. For transactions below $10, those numbers are 11 percent and 20 percent, respectively. Not surprisingly, credit or debit cards are almost universally preferred for online transactions (right-side bar in Figure A1). A very small fraction of consumers prefer to use cash for online transactions. Out of the 13 consumers who prefer to use cash for online transactions, 7 were unbanked, while 6 had a bank account. Figure A2 shows a breakdown of preferred payment method for bills and for purchases (for any value of

8

ACCEPTED MANUSCRIPT

transaction). Payment preferences differ greatly between bills and purchases, one of the reasons why the earlier version of the diary lacks the necessary information for our study. Appendix Figure A3 shows the actual use of payment methods by transaction value for in-person purchases. Although Figure A3 shows the percentage of transactions while Figure A1 shows the percentage of consumers, the actual behavior and the stated preferences display

CR IP T

similar patterns: as the value of a transaction rises, consumers become less likely to use cash and more likely to use credit or debit cards for in-person payments. However, for transactions over $100, 22 percent of in-person transactions are conducted with cash, a higher fraction than for transactions between $50 and $100. As we discussed above, consumers received cash discounts

AN US

for some of those transactions.

Table 2 shows payment instrument adoption and preferred payment instrument, by major demographic and income categories. The left-side panel shows the percentage of consumers within each cohort who adopted cash, credit card, or debit card. The right-hand side panel shows the percentage of consumers within each cohort who stated that they prefer to use cash, credit

M

cards, or debit cards for their purchases. Youngest, lowest-income, and least-educated consumers prefer to pay cash, while older, higher-income, and highly-educated consumers prefer to use

ED

credit cards. At least to some extent those preferences are dictated by supply-side constraints— low-income and low-education consumers are much less likely to own a credit card. Use of preferred payment methods

PT

B.

Above, we discussed consumers’ stated preferences for specific payment methods. Here, we

CE

analyze the actual use of payment instruments and how use relates to preferences. Figures 1a, 1b, and 1c show the probabilities of using cash, debit cards, and credit cards, respectively, by the

AC

dollar value of transaction for purchases, and by the stated preferences. The probability of using cash declines with transaction value for transactions below $50, and the probability of using cards increases with transaction value (at least initially) for all consumers; but for all transaction values consumers have a higher probability of using their preferred payment method. This finding is similar to the finding in O’Brien (2014). Thus, consumers who prefer cash are more likely to use cash even for large-value transactions, and those who prefer debit or credit are more likely to use their preferred payment methods even for low-value transactions. 9

ACCEPTED MANUSCRIPT

The probability of using cash is especially high for transactions below $10. There are several reasons why consumers use cash for low-value transactions. Some merchants impose a minimum amount for card transactions. As part of the Dodd-Frank Wall Street Reform and Consumer Protection Act, merchants can legally set a minimum amount for credit card transactions of up to $10, and credit card companies must allow these minimums. Although the

CR IP T

law does not specify such amount for debit cards, our data show that consumers are also much less likely to use debit cards for low-value transactions, as shown in Figures 1b and 1c.

The probability of using the preferred method also varies across demographic groups, location of transaction, and merchant categories (see Tables 3a and 3b). Looking at the aggregate values in the top row of each table, over one-third of all transactions are conducted with a non-

AN US

preferred payment method (also shown in Figures 1a-c). Regardless of the preferred payment method, consumers use cash for lower-value transactions, and use cards for larger-value transactions. People who prefer cash use noncash payment instruments for larger-value transactions: the median value of their cash transactions is $10, compared to $21.75 as the median value of their noncash transactions. Similarly, people who prefer cards use cash for

M

lower-value transactions and use their preferred method—cards—for larger-value transactions. Table 3a shows a breakdown by demographic attributes. Note that among consumers who prefer

ED

credit cards, those under age 25, those in the lowest-income group, and those with the lowest education are most likely to use non-preferred payment methods. The reason could be supply-

PT

side related: as we discussed above, those consumers are much less likely to have a credit card, compared to consumers in other demographic cohorts, and are likely to have lower credit limits

CE

on their cards. Table 3b shows a breakdown by merchant type. Despite the variation across merchants, cash transactions tend to be of lower value than card transactions everywhere.

AC

Briglevics and Schuh (2014) found that having at least as much cash in wallet or purse as

the transaction value is a strong predictor of whether the consumer used cash for that transaction, especially when the consumer was one who faced higher withdrawal costs and had diverse payment instrument bundles. Below we estimate the probability that a consumer switched to cash when a cash discount was applied. In that model, we control for whether the consumer had sufficient cash with him or her to pay for the transaction. In addition to the overall amount of cash on person, the currency denomination is also important. Note that the probability of using 10

ACCEPTED MANUSCRIPT

cash by those who prefer debit cards spikes at transaction value of $100 (Figure 1a). It is possible that the likelihood of using cash is to some extent dictated by currency denomination—people who have a $100 bill with them might pay cash for a transaction of that amount. C.

Payment preferences and demographics

To understand what affects a consumer’s preference for a given payment method, we

CR IP T

estimate regressions of payment preferences on consumers’ demographic attributes and income. We focus on the payment instruments used most commonly for purchases: cash, debit cards, and credit cards, and we separately analyze cash preferences and card preferences. The sample in model (1) includes all respondents who prefer cash, debit cards, or credit cards, whereas the

AN US

sample in model (2) includes only respondents who prefer cards. Pr(𝐶𝐴𝑆𝐻𝑃𝑅𝐸𝐹𝑖 ) = 𝐶𝐻(𝐴𝑔𝑒𝑖 , 𝐸𝑑𝑢𝑖 , 𝑅𝑎𝑐𝑒𝑖 , 𝐺𝑒𝑛𝑖 , 𝐼𝑛𝑐𝑖 , 𝐸𝑚𝑝𝑖 )

(1)

Pr(𝐶𝐶𝑃𝑅𝐸𝐹𝑖 ) = 𝐶𝐶(𝐴𝑔𝑒𝑖 , 𝐸𝑑𝑢𝑖 , 𝑅𝑎𝑐𝑒𝑖 , 𝐺𝑒𝑛𝑖 , 𝐼𝑛𝑐𝑖 , 𝐸𝑚𝑝𝑖 )

(2)

M

where:

1 if consumer 𝑖 prefers credit cards over debit cards 0 if consumer 𝑖 prefers debit cards over credit cards

PT

𝐶𝐶𝑃𝑅𝐸𝐹𝑖 = {

ED

1 if consumer 𝑖 prefers cash over credit and debit cards 𝐶𝐴𝑆𝐻𝑃𝑅𝐸𝐹𝑖 = { 0 if consumer 𝑖 prefers credit and debit cards over cash

𝐴𝑔𝑒𝑖 , 𝐸𝑑𝑢𝑖 , 𝑅𝑎𝑐𝑒𝑖 , 𝐺𝑒𝑛𝑖 , 𝐼𝑛𝑐𝑖 , 𝐸𝑚𝑝𝑖 = a vector of exogenous attributes of consumer i: age,

CE

education, race, gender, income, and employment. Table 4 presents the results of estimating equations (1) and (2). Both equations were

AC

estimated using probit, and the level of observation is an individual respondent. Most of the attributes significantly affect consumers’ payment preferences, with education having the largest effect on payment preferences. Despite the fact that younger people tend to prefer debit cards over credit cards, age was not significant in the regressions when controlling for education and income. Consumers who have low education, low income, are Latino, and are male are more likely to prefer cash over cards. Employed consumers are less likely to prefer cash. Among consumers who prefer to use cards over cash, less educated consumers tend to prefer debit over 11

ACCEPTED MANUSCRIPT

credit cards, while higher-income and highly-educated consumers prefer credit over debit. In particular, those with household income above $100,000 strongly prefer credit over debit. Consumers who are non-Latino or male are also significantly more likely to prefer credit cards over debit cards. Controlling for income, employed consumers are significantly less likely to prefer credit cards. Note that this model explains stated preferences only and not the actual use of

CR IP T

each payment instrument. Although employed respondents are more likely to prefer debit cards to credit cards than those who are not employed, there is no significant difference in the use of credit cards (measured as the share of all purchases) between consumers who are employed and those who are not employed. Those who are not employed include many different categories: unemployed,

AN US

retired, students, etc. There is no evidence that the unemployed rely more heavily on credit cards than those who are employed. In fact, among consumers who prefer credit cards, there is some evidence that the employed tend to use credit cards more intensively than the unemployed when use is measured as a share of all transactions conducted with a given payment instrument. In

reverse is found for debit cards.

M

addition, unemployed consumers tend to use cash more than employed consumers do, while the

ED

To test whether experiencing financial difficulty affects consumer payment preferences, we ran alternative regressions that include a measure of financial fragility.6 The variable is equal to 1 for consumers reporting that they would have difficulty coming up with $1,000 to address an

PT

unexpected emergency.7 Approximately 29 percent of respondents are financially fragile based on that measure. We find that individuals who are financially fragile are more likely to prefer

CE

cash over cards, and more likely to prefer debit cards over credit cards, controlling for all the other consumer demographic and financial attributes. Additionally, consumers who use payday

AC

loans tend to prefer cash over cards. One explanation could be that financially fragile individuals—who tend to be more credit constrained—try to avoid accumulating further debt and want to commit to spending less and so would be more likely than other consumers to prefer

6

The results of those specifications are available from the authors.

7

The amount was randomized across respondents, but we created an indicator variable for all respondents equal to 1 if a person can come up with at least $1,000 for emergency expenditures.

12

ACCEPTED MANUSCRIPT

cash over cards, and debit over credit. All the other coefficients remain qualitatively unchanged when these measures of financial difficulties are included in the model. D.

Characteristics of payment methods

The survey asks respondents to pick the most important characteristic of their preferred payment method. The characteristics include: acceptance, setup, cost, convenience, records, and security.

CR IP T

Previous research has found that consumers’ perceptions of those characteristics significantly affect adoption and use of payments (Schuh and Stavins 2010, 2013). When we included

characteristics of cash in equation (1) and characteristics of credit cards in equation (2), the

coefficients were positive and significant, as could be expected, but the demographic and income variables remained significant and qualitatively similar to those reported in Table 4.8 Although

AN US

there is some variation among payment methods, convenience and security were most often listed as the two most important characteristics. Demographic attributes explain very little of the variation in consumers’ perceived characteristics: in ordered probit regressions of the ranking of characteristics on demographics, the pseudo R2 in all the regressions were below 0.05. Why do consumers deviate from their preferred method?

M

E.

Table 5a shows the percentage of consumers who prefer each payment instrument—cash, debit,

ED

and credit—and the percentage of transactions for which the preferred payment instrument is used. Consumers use their preferred method for approximately two-thirds of their transactions.

PT

For the transactions where consumers deviate from their preferred payment instrument, they were asked why they deviated. As the table shows, the most common reason why consumers

CE

who prefer cash used another payment was because they did not have enough cash with them. In contrast, the main reason why consumers deviated from using credit or debit was transaction size. Consumers who prefer cards may decide to switch to cash for low-value transactions, while

AC

consumers who prefer cash may have to use cards when they do not have enough cash on their person. Table 5b shows which payment instruments consumers deviated to: For those who prefer either credit or debit cards, if they deviate from their preferred method, it is predominantly to cash rather than to the other type of card.

8

The regression results with the characteristics included are available from the authors.

13

ACCEPTED MANUSCRIPT

We compared the survey responses where consumers stated that they did not have enough cash on person to conduct their transactions in cash with an interpolated measure of how much the person was carrying in their wallet, purse, or pocket at the time of payment (we refer to this amount as “cash in wallet.”). Our cash-in-wallet measure takes into account all relevant cash activities throughout the day, such as bill and cash payments, cash deposits, cash withdrawals,

CR IP T

and prepaid card loadings (see Briglevics and Schuh 2014 for a more detailed discussion). For the most part, the cash-in-wallet amount is consistent with respondents’ indication of whether they had enough cash at the time of the transaction: 90 percent of transactions for which the respondent had enough cash on person were reported consistently, and 91 percent of transactions for which the respondent did not have enough cash on person were reported consistently.9

AN US

Therefore, insufficient cash on person is the main reason why consumers deviate from cash use when that is their preferred payment method.

In order to assess whether price incentives have an effect on consumers’ use of payment instruments, we first investigate the effect of other (non-price) factors on payment choice. Consumers could “deviate” from their preferred payment method and use another method

M

because of supply-side restrictions, or because their preferences vary with other payment-specific factors, such as dollar value, location, or type of merchant (demand side). To test which factors

ED

affect the probability of deviating from a preferred payment method, we estimate a set of “FROM_PI” probit regressions where the dependent variable is the probability of deviating from

PT

the preferred payment instrument. For example, for those respondents who prefer cash we estimate the probability of deviating from cash. Here, the sample includes all consumers who

CE

prefer the given payment instrument, and the dependent variable equals 1 if the transaction is conducted with another payment instrument and 0 otherwise:

AC

Pr  FROM _ CASH ij   f

 AMOUNT , MERCHANT , AMTMERCH

Pr  FROM _ DEBITij   f

Pr  FROM _ CREDITij   f

j

j

j

 AMOUNT , MERCHANT , AMTMERCH j

j

, CASH ij , X i  j

, DCi , X i 

 AMOUNT , MERCHANT , AMTMERCH j

j

9

j

, CCi , X i 

(3) (4) (5)

Note that the question whether the respondent had enough cash was asked only when cash was not used for inperson transactions at retail stores or restaurants.

14

ACCEPTED MANUSCRIPT

where: 1 if consumer i prefers cash and uses not cash for transaction j FROM _ CASH ij   0 if consumer i prefers cash and uses cash for transaction j

CR IP T

1 if consumer i prefers debit and uses not debit for transaction j FROM _ DEBITij   0 if consumer i prefers debit and uses debit for transaction j 1 if consumer i prefers credit and uses not credit for transaction j FROM _ CREDITij   0 if consumer i prefers credit and uses credit for transaction j AMOUNT j = dollar value of the transaction j

AN US

MERCHANT j = merchant category for transaction j

AMTMERCH j = AMOUNT j * MERCHANT j interaction term of dollar amount and merchant

type

CASH ij = 1 if consumer i had enough cash on person for transaction j

M

DCi = 1 if consumer i has a debit card on person

ED

CCi = 1 if consumer i has a credit card on person. Even though we do not control for which payment instrument the consumer switched to,

PT

we know that he switched away from his preferred payment. We find that all consumers— regardless of their preferred method—are more likely to deviate from cash and less likely deviate

CE

from debit cards for higher-value transactions (Table 6a). Our results indicate that a $1 increase in transaction amount raises the probability of switching from cash by 0.3 percent, and lowers

AC

the probability of switching from debit cards by 0.1 percent. Surprisingly, the dollar amount does not significantly affect the likelihood of deviating from credit cards. Carrying the payment instrument on one’s person is a necessary condition for using that

payment method for a transaction and is therefore a significant predictor of behavior: Consumers who had sufficient cash on person to conduct the transaction were significantly less likely to deviate from using cash (or more likely to pay cash), and consumers who carried their debit or credit card were less likely to deviate from using that method. To focus on demand-side effects, 15

ACCEPTED MANUSCRIPT

for the cash regression we restricted the sample to in-person transactions only. For the cards regressions, there was no significant difference between in-person and online transactions. Controlling for having the payment instrument on person, older and less-educated consumers are less likely to deviate from cash. Less-educated consumers are more likely to

CR IP T

deviate from credit cards, while men are less likely to deviate from credit cards. Table 6b shows the results separately for the probability of switching from cash, debit, or credit specifically to one of the other two payment methods. Because most of the transactions where consumers switch from cards are to cash (rather than to the other type of card), the results of the “From Debit to Cash” regression are qualitatively similar to those in the “From Debit” those in the “From Credit” regression.

AN US

regression, and the results of the “From Credit to Cash” regression are qualitatively similar to the

Surprisingly, we find that high-income consumers (those with household income over $100K) are more likely to deviate from credit cards than other consumers. In Appendix B we analyze the transactions where high-income consumers who prefer credit cards use other

M

payment methods. Most of those transactions are below $10 (Figure B1), the most frequent merchant category is food (Figure B2), the most frequently cited reason for the deviation is the

ED

transaction amount (Figure B3), and the payment method used most frequently for those transactions is cash (Figure B4). Therefore high-income consumers tend to prefer credit cards

PT

overall, but are likely to use cash when making small-value food purchases. When comparing the effect of merchant categories on the probability of deviating from

CE

preferred methods, consumers who prefer cash are more likely to deviate from cash for highvalue auto-related related transactions (coefficient on the AUTO * AMOUNT interaction term is

AC

positive), while consumers who prefer credit cards are less likely to deviate from credit cards for auto-related high-value transactions (coefficient on the AUTO * AMOUNT interaction term is negative). Both effects indicate that consumers are less likely to use cash and more likely to use credit cards for higher-value auto-related transactions regardless of their payment preferences. The evidence suggests that relatively low-value transactions such as gasoline purchases and tolls are paid with cash, but auto repairs or parts are paid with cards. As we found in other

16

ACCEPTED MANUSCRIPT

specifications, for higher-value transactions, consumers are more likely to deviate from cash and less likely to deviate from cards. Figures 2a, 2b, and 2c show the reasons consumers gave for deviating from their preferred payment method. Consumers deviated from cash either because they did not have cash on hand or because of the transaction size (especially for transactions over $40). Consumers

CR IP T

deviated from using debit or credit cards mainly because of the transaction size, especially for transactions under $10, for which they used cash. Receiving a discount was rarely mentioned as a reason for deviating, and surcharges were mentioned even less frequently. We analyze surcharges and discounts in greater detail below.

AN US

Appendix Figure A3 shows the payment method used by the value of transaction,

including both purchases and bill payments. Interestingly, 22 percent of all transactions above $100 were paid in cash. Over half of those cash transactions—51.8 percent—are conducted by consumers who prefer noncash payment methods. We analyzed those transactions more closely, and found that for 11 percent of them the consumer received a cash discount. So even though

M

consumers may state another reason for why they deviated from their preferred payment methods

V.

ED

to cash, in some cases they received a cash discount.

Discounts and surcharges

PT

As was the case in 2012 (Shy and Stavins 2015), discounts and surcharges based on the payment instrument continued to be very rare in 2015. Tables 7a and 7b show the percentage of

CE

transactions that received a discount for using cash or debit card, and the percentage of transactions that received a surcharge for using a credit card. In 2015, consumers reported receiving cash discounts on 1.6 percent of in-person cash transactions, debit card discounts on

AC

2.0 percent of in-person debit card transactions, and credit card surcharges on only 1.0 percent of in-person credit card transactions (all purchases). In 2012, the comparable numbers were 1.7 percent, 1.8 percent, and 1.2 percent, respectively. We have no information on the percentage of transactions or merchants where discounts or surcharges were offered but not received or paid, only the percentage where the discount or surcharge was actually applied. Therefore, it is possible that a consumer decided not to use a credit card because he or she would have had to pay a surcharge, but it is also possible that a consumer decided to use a debit card because of the 17

ACCEPTED MANUSCRIPT

discount and that otherwise he or she would have used another payment method instead. Below we investigate these potential shifts in payment instrument choices induced by price incentives. Figure 3 plots the percentage of transactions with discounts and surcharges by transaction value. We observe only the intersection of supply and demand—that is, only when a merchant decides to offer a discount or impose a surcharge and when a consumer uses the payment method

CR IP T

associated with the price incentive. We do not see discounts that were not received or surcharges that were not paid. As Figure 3 shows, cash discounts are more likely to be observed at higher transaction values—the probability rises for transactions over $15, but starts declining for transactions over $30. The supply of either discounts or surcharges by merchants may be independent of the transaction value. The fact that the observed occurrence of discounts or

AN US

surcharges varies with value may be caused by the choices made by consumers, and not by the merchants’ pricing strategy. For example, merchants may impose surcharges on credit card gasoline sales regardless of the amount, but consumers typically spend $40 on filling their tank. As a result, we observe a spike in credit card surcharges on gasoline sales at $40. There is relatively little variation in the observed occurrence of discounts or surcharges

M

by consumers’ demographic attributes (Table 7a). Looking at the breakdown by merchant sector (Table 7b), the percentage of cash transactions for which discounts are observed is higher in the

ED

auto or transportation sectors. In particular, 7.3 percent of cash transactions at gas stations received cash discounts, compared with 1.0 percent of cash transactions elsewhere (all

PT

weighted). Debit card discounts were more likely to be received in general merchandise transactions, and credit card surcharges were most likely to be observed in medical/education

CE

services and transportation. The transportation transactions where credit card surcharges were observed are likely the same types of transactions that would receive cash discounts if the consumers paid cash instead of credit cards. Probit regressions of the probability of receiving a

AC

discount or paying a surcharge on consumers’ demographic and income attributes (not reported here) show no significant differences in the probability of receiving debit card discounts among the various demographic groups, while cash discounts were more likely to be received by older consumers and credit card surcharges were more likely to be paid by white consumers.

18

ACCEPTED MANUSCRIPT

VI. Price incentives and payment preferences In this section, we address the question whether payment preferences are related to discounts and surcharges. In particular, do people deviate to a non-preferred payment method because of such price incentives? In other words, is consumer demand for payment instruments elastic with

CR IP T

respect to price incentives? When we examine respondents’ explanations for their deviating to other (non-preferred) payment methods, summary statistics show very low occurrence of discounts or surcharges as the reported reason, and most common reasons do not include discounts or surcharges (Table 5a). In only between 2 and 4 percent of transactions did consumers deviate from their preferred

AN US

payment method because they received a discount, and in only 1 percent of transactions did they deviate from using a credit card to avoid a surcharge (Table 8). Furthermore, there were no statistically significant differences in the probability of receiving a discount or paying a surcharge between consumers who used their preferred method and those who did not use their preferred method.

M

Although consumers report some instances when they received a discount for using a credit card, follow-up cognitive interviews revealed that those were cases when consumers used

ED

store credit cards, such as a Sears or Macy’s card, and the individual stores issued discounts. In contrast, credit card surcharges apply to cases when a consumer might use any type of general-

PT

purpose credit card.

We model consumer payment choice as a function of discounts and surcharges.

CE

Consumer i selects payment instrument (PI) k among his adopted set of payment instruments, based on whether there is a discount or a surcharge, on the amount of transaction, merchant type,

AC

and on characteristics of i, including demographics and income: Pr  PI ikj   f

 DISCOUNT

kj

, SURCHARGEkj , AMOUNT j , MERCHANT j , X i  (6)

where: Pr  PI ikj  = probability that consumer i selects payment instrument k for transaction j

k = cash, debit card, credit card. 19

ACCEPTED MANUSCRIPT

DISCOUNTkj = 1 if there is a discount for using k on transaction j (for cash or debit card only); if there is a discount for choosing payment instrument k, then the probability of selecting k is expected to increase. SURCHARGEkj = 1 if there is a surcharge for using k on transaction j (for credit card only); if there is a surcharge for choosing payment instrument k, then the probability of selecting k is expected to decrease.

CR IP T

Because we cannot test whether a consumer’s probability of selecting a payment instrument is significantly affected by a discount or a surcharge being offered (we do not observe all potential discounts and surcharges, only when they actually occurred), we use an alternative approach, incorporating the survey’s information on consumers’ preferred payment instrument.

To estimate the marginal effect of a discount or a surcharge on the probability of

AN US

switching away from one’s preferred payment instrument, we estimate “TO PI” regressions, the probability of shifting away from a preferred payment method to PI, where the sample includes all transactions conducted with the given payment instrument PI, and the dependent variable equals 1 if the consumer prefers another payment instrument and 0 otherwise:

 DISCOUNT

Pr  TODEBITij   f

 DISCOUNT

kj

, AMOUNT j , MERCHANT j , AMTMERCH j , X i 

(7)

, AMOUNT j , MERCHANT j , AMTMERCH j , X i 

(8)

ED

kj

M

Pr  TOCASH ij   f

where:

PT

1 if consumer i uses cash for transaction j , and he prefers not cash TOCASH ij   0 if consumer i uses cash for transaction j , and he prefers cash

CE

1 if consumer i uses debit for transaction j , and he prefers not debit TODEBITij   0 if consumer i uses debit for transaction j , and he prefers debit

AC

Equation (7) estimates how much the probability that a cash transaction is conducted by a noncash-preferring consumer is affected by cash discounts and by other factors, such as transaction amount and merchant type. Equation (8) estimates a similar effect for debit cards. Table 9 shows the results of estimating equations (7) and (8). We do not estimate an equation for the Pr  TOCREDITij  , because we assume that consumers minimize their cost and would not switch to using a credit card because of a surcharge. Among cash transactions, it is more likely 20

ACCEPTED MANUSCRIPT

that a consumer switched to using cash from his preferred payment method if there was a cash discount. The dependent variable equals 1 if a respondent used cash for the transaction, but preferred another payment instrument, and equals 0 if a respondent used and preferred cash. The results presented in Table 9 show that if there is a cash discount, the probability that a cash transaction is conducted by a consumer who prefers noncash payments increases by 19.2

CR IP T

percent.10 Thus, there is a significant impact of cash discounts on noncash-preferring consumers, even after controlling for the dollar amount, merchant sector, interaction of dollar amount and merchant sector, consumers’ demographic attributes, and income. The result suggests that discounts can steer people away from other payment methods to cash. We did not find any evidence that a debit card discount significantly affected the probability of using a debit card if

AN US

the consumer preferred another payment method. Note that even though we cannot determine for sure whether a consumer decided not to use a credit card because of a surcharge on credit card transactions, very few consumers cited surcharges as a reason for deviating from credit cards. The result has potentially important policy implications. Merchants were not allowed to

M

differentiate their prices based on payment instruments, and their legal disputes with credit card companies have a long history. But we find that even if credit card companies let merchants

ED

impose surcharges on credit card transactions, there may be other reasons for merchants to offer cash discounts instead. Namely we find evidence that cash discounts are a more effective tool to influence consumers’ decision of which payment method to use at checkout. So even without

PT

any legal or regulatory restrictions, merchant might choose a carrot rather than stick as a way to steer consumers’ decision-making. This might be a rare case when merchants and credit card

CE

companies are in agreement.

Consumers who have reward credit cards might be less likely to switch payment methods

AC

when facing credit card surcharges, because their net cost of surcharges is smaller than the cost for those without rewards. Although we did not estimate “TO CREDIT” regression, we did include a dummy variable for having a reward credit card in equation (5), estimating the probability of switching “FROM CREDIT.” The coefficient on the variable indicating whether the consumer had a reward credit card was not statistically significant, indicating that reward 10

We thank an anonymous referee for a clarification of the interpretation of this result.

21

ACCEPTED MANUSCRIPT

credit card holders were not less likely to switch from credit cards than were non-reward credit card holders. Even though we do not observe whether a credit card surcharge was imposed and avoided by switching to another payment method, we do observe that a consumer who prefers credit cards used another (non-preferred) method for some reason. Having a reward credit card

CR IP T

did not affect that probability of switching away from credit cards.

VII. Conclusion

Despite the fact that U.S. merchants have been allowed to price differentiate based on payment method for a few years, such price incentives—discounts on cash and debit card transactions, and surcharges on credit card transactions—continued to be very rare in 2015, as was found in

AN US

2012 (Shy and Stavins 2015). We analyze in detail who receives discounts and surcharges, and whether such price incentives induce consumers to deviate from their preferred payment methods. We find evidence that consumers switch from their preferred payment methods to cash if a discount is offered. However, most of the switching from a preferred payment method continues to take place for low-value auto-related transactions, such as gasoline purchases, where

M

cash discounts have been offered for a long time, compared with other retail sectors. More research is needed to determine whether the infrequent occurrence of price incentives is caused

ED

by merchants’ reluctance to offer them or by consumers’ inelastic demand for payment

AC

CE

PT

instruments.

22

ACCEPTED MANUSCRIPT

References Agarwal, S., S. Chakravorti, and A. Lunn. 2010. “Why Do Banks Reward their Customers to Use their Credit Cards?” Federal Reserve Bank of Chicago Working Paper WP 2010-19.

CR IP T

Agarwal, S. and W. Qian. 2014. Consumption and debt response to unanticipated income shocks: evidence from a natural experiment in Singapore, American Economic Review 104(12): 42054230. Agarwal, S., K. Mo Koo and W. Qian. 2017. “Consumption Response to Temporary Price Shock: Evidence from Singapore’s Annual Sale Event. Georgetown McDonough School of Business Research Paper No. 3035787. Available at SSRN: https://ssrn.com/abstract=3035787 Amromin, G., C. Jankowski, and R.D. Porter. 2007. “Transforming Payment Choices by Doubling Fees on the Illinois Tollway.” Economic Perspectives 31(2): 22-47.

AN US

Arango, C., K. Hyun, and L. Sabetti. 2015. “Consumer Payment Choice: Merchant Card Acceptance Versus Pricing Incentives.” Journal of Banking & Finance 55(c): 130-141. Barron, J., M. Staten, and J. Umbeck. 1992. "Discounts for Cash in Retail Gasoline Marketing." Contemporary Economic Policy 10(4): 89-102.

M

Bolt, W., N. Jonker, and C. van Renselaar. 2010. “Incentives at the Counter: An Empirical Analysis of Surcharging Card Payments and Payment Behavior in the Netherlands.” Journal of Banking & Finance 34(8): 1738-1744.

ED

Briglevics, T., and S. Schuh. 2014. “This Is What's in Your Wallet...and Here's How You Use It.” Federal Reserve Bank of Boston Working Paper 14-5.

PT

Briglevics, T., and O. Shy. 2014. “Why Don’t Most Merchants Use Price Discounts to Steer Consumer Payment Choice?" Review of Industrial Organization 44(4): 367-392. Borzekowski, R., E. Kiser, and S. Ahmed. 2008. “Consumers' Use of Debit Cards: Patterns, Preferences, and Price Response.” Journal of Money, Credit and Banking 40(1): 149-172.

CE

Ching, A., and F. Hayashi. 2010. “Payment Cards Reward Programs and Consumer Payment Choice.” Journal of Banking & Finance 34(8), 1773-1787.

AC

Cohen, M., and M. Rysman. 2013. “Payment Choice with Consumer Panel Data.” Federal Reserve Bank of Boston Working Paper 13-6. Humphrey, D., M. Kim, and B. Vale. 2001. “Realizing the Gains from Electronic Payments: Costs, Pricing, and Payment Choice.” Journal of Money, Credit and Banking 33(2): 216-234. Klee, Elisabeth. 2008. “How People Pay: Evidence from Grocery Store Data.” Journal of Monetary Economics 55(3), 526–541. O’Brien, S. 2014. “Consumer Preferences and the Use of Cash: Evidence from the Diary of Consumer Payments Choice.” Federal Reserve Bank of San Francisco Working Paper.

23

ACCEPTED MANUSCRIPT

Reserve Bank of Australia. 2011. Review of Card Surcharging: A Consultation Document. http://www.rba.gov.au/publications/consultations/201106-review-card-surcharging/pdf/201106review-card-surcharging.pdf Schuh, S., and J. Stavins. 2010. “Why Are (Some) Consumers (Finally) Writing Fewer Checks? The Role of Payment Characteristics.” Journal of Banking and Finance 34(8): 1745–1758.

CR IP T

Schuh, S., and J. Stavins. 2013. “How Consumers Pay: Adoption and Use of Payments,” Accounting and Finance Research 2(2). Shy, O., and J. Stavins. 2015. “Merchant Steering of Consumer Payment Choice: Evidence from a 2012 Diary Survey.” Journal of Behavioral and Experimental Economics 55(c): 1-9. Simon, J., K. Smith, and T. West. 2010. “Price Incentives and Consumer Payment Behavior.” Journal of Banking & Finance 34(8): 1759-1772.

AC

CE

PT

ED

M

AN US

Wang, Z., and A. Wolman. 2014. “Payment Choice and the Future of Currency: Insights from Two Billion Retail Transactions.” Federal Reserve Bank of Richmond Working Paper No. 14-09.

24

ACCEPTED MANUSCRIPT

Table 1 Number and percentage of DCPC diarists by demographic groups, weighted

Gender Ethnicity

Latino Not Latino White Black Asian Other

CE

Race

Employment Status(2) Employed Not Employed Total

AC

CR IP T

AN US

Education

ED

Income

Under 25 25-34 35-44 45-54 55-64 Over 64 < 25k 25-49k 50-74k 75-99k 100-124k 125-199k > 200k Less than High School High School Some College College Graduate Male Female

PT

Age

DCPC Diarists Number of Respondents Percent(1) 69 6% 232 22% 170 16% 214 20% 181 17% 199 19% 229 22% 249 23% 210 20% 128 12% 99 9% 107 10% 36 3%

Groups

M

Categories

95

9%

346 199 283 142 511 555

32% 19% 27% 13% 48% 52%

139 927 807 144 50 64

13% 87% 76% 14% 5% 6%

674 392 1,066

63% 37%

Source: 2015 Diary of Consumer Payment Choice and authors' calculations. Statistics displayed are all weighted. Notes: (1) (2)

Percent adds up to 100% within each category. Employment status variable = 1 if currently working, = 0 otherwise.

25

ACCEPTED MANUSCRIPT

Table 2 Payment instrument adoption and preferred payment instrument, by major demographic and income categories Adoption Rate (Percentage) Cash Debit Credit 99.9 80.2 77.2 Under 25 25-34 35-44 45-54 55-64 Over 64

100.0 99.4 100.0 100.0 100.0 100.0

76.6 81.6 86.0 78.8 80.8 76.1

47.3 71.0 74.6 74.2 81.8 96.2

45.6 28.4 27.0 24.6 29.6 23.9

47.1 48.3 44.5 45.6 44.4 34.1

0.0 21.5 24.1 24.5 21.4 37.9

< 25k 25-49k 50-74k 75-99k 100-124k 125-199k > 200k

100.0 99.5 100.0 100.0 100.0 100.0 100.0

58.5 84.0 89.3 85.4 91.1 86.1 84.0

45.7 72.2 90.9 91.1 89.3 97.4 86.0

45.4 30.9 22.0 22.3 22.3 13.3 2.3

40.9 44.6 53.3 47.2 37.2 35.5 35.8

5.7 18.4 23.2 29.9 38.0 49.1 55.8

100.0 100.0 100.0 99.5 100.0

55.7 80.2 75.3 88.2 87.9

35.4 72.6 77.2 88.2 94.7

67.9 36.5 27.9 14.9 5.8

28.5 44.1 46.4 48.1 40.7

2.7 13.1 22.2 33.3 50.0

78.7 81.7

77.2 77.3

31.3 24.7

37.7 49.4

27.5 21.1

100.0 99.9

80.1 80.3

61.0 79.7

46.4 25.1

41.5 44.1

8.2 26.6

99.8 100.0 100.0 100.0

81.2 79.5 72.6 76.3

79.6 64.4 89.7 66.6

26.8 30.2 24.5 38.5

42.8 60.5 21.8 35.1

26.3 6.9 41.3 22.8

99.8 100.0

85.2 71.8

80.1 72.4

22.9 36.3

49.1 34.6

23.2 25.8

CE

AC

M

ED 99.7 100.0

PT

Education Less than High School High School Some College College Graduate Gender Male Female Ethnicity Latino Not Latino Race White Black Asian Other Employment Status(1) Employed Not Employed

AN US

Income

CR IP T

Total Age

Purchase Preference Rate (Percentage) Cash Debit Credit 27.9 43.8 24.2

Source: 2015 Diary of Consumer Payment Choice and authors' calculations. Preference rate is defined as the percentage of consumers within each cohort who state that a given payment instrument is their most preferred instrument for purchase transactions. Both the adoption rate and the preference rate are weighted. Note: (1) Employment status variable = 1 if currently working, = 0 otherwise.

26

ACCEPTED MANUSCRIPT

Percent NonPreferred 35.7

Total

Cash Preference Median Amount NonPreferred Preferred $ 21.75 $ 10.00

Percent NonPreferred 44.2

Debit Preference Median Amount NonPreferred Preferred $ 10.58 $ 20.38

Age 32.5 50.9 43.1 30.4 29.0 27.3

$ $ $ $ $ $

15.50 10.00 27.10 23.40 31.18 38.88

$ $ $ $ $ $

7.50 10.00 9.89 10.00 10.00 9.00

42.8 42.1 36.9 41.8 47.1 57.4

< 25k 25-49k 50-74k 75-99k 100-124k 125-199k > 200k

18.9 39.5 40.7 29.1 48.5 34.0 61.1

$ $ $ $ $ $ $

27.10 20.00 19.74 24.47 16.97 64.84 9.50

$ $ $ $ $ $ $

9.51 10.00 10.00 10.00 8.76 7.01 2.77

55.5 53.3 41.4 36.4 39.7 45.4 27.0

Less HS High School Some College College Graduate

20.5 36.7 28.9 49.4 46.9

$ $ $ $ $

38.30 18.29 21.28 23.40 20.95

$ $ $ $ $

10.00 10.00 10.00 9.66 7.97

44.1 40.0 50.1 45.0 43.9

Male Female

38.7 30.9

$ $

21.90 21.38

$ $

8.63 11.01

Latino Not Latino

46.3 32.9

$ $

19.00 22.00

White Black Asian Other

36.6 38.0 22.6 28.7

$ $ $ $

22.31 18.66 12.49 26.30

Income

Ethnicity

$ $ $ $ $ $

12.84 19.49 21.00 21.44 22.30 23.60

$ $ $ $ $ $ $

10.00 11.00 10.00 9.19 12.99 19.96 9.50

$ $ $ $ $ $ $

$ $ $ $ $

16.58 12.32 10.00 10.00 10.69

$ $

Credit Preference Median Amount NonPreferred Preferred $ 10.00 $ 25.28

100.0 38.8 32.1 39.4 55.5 49.2

$ $ $ $ $ $

67.00 9.83 12.94 10.71 8.50 11.05

$ $ $ $ $ $

28.00 23.77 23.47 23.19 31.00

17.93 22.15 19.79 20.00 23.91 21.75 25.42

49.8 52.5 45.2 39.6 47.1 38.0 38.1

$ $ $ $ $ $ $

12.00 10.59 10.00 9.33 10.81 10.00 9.35

$ $ $ $ $ $ $

19.86 21.55 23.92 29.99 31.52 25.00 25.00

$ $ $ $ $

22.71 25.00 20.00 20.00 20.00

62.0 48.0 48.2 43.9 39.4

$ $ $ $ $

40.00 12.00 8.31 10.34 10.00

$ $ $ $ $

353.24 32.14 22.06 25.11 25.53

10.00 11.00

$ $

21.39 20.00

41.9 45.4

$ $

9.82 10.75

$ $

25.00 26.00

$ $

10.00 10.00

40.5 44.9

$ $

18.38 10.00

$ $

20.00 20.55

31.8 44.1

$ $

15.25 10.00

$ $

19.00 25.42

$ $ $ $

10.00 7.50 7.30 12.96

43.1 48.6 56.5 37.7

$ $ $ $

10.52 10.00 32.62 11.00

$ $ $ $

20.42 18.77 27.29 21.46

44.0 56.8 34.0 43.0

$ $ $ $

10.00 10.34 10.57 12.00

$ $ $ $

25.52 21.19 22.73 23.44

PT

Race

42.1 45.9

ED

Gender

17.48 15.00 11.25 9.59 8.87 11.64

M

Education

$ $ $ $ $ $

Percent NonPreferred 43.6

AN US

Under 25 25-34 35-44 45-54 55-64 Over 64

CR IP T

Table 3a Occurrence and median transaction size for using preferred and non-preferred payment method, by demographic groups

AC

CE

Source: 2015 Diary of Consumer Payment Choice. Note: Purchases only. Percent columns indicate percentage of transactions where non-preferred payment method was used. Median Amount columns refer to the amount of the transaction. Means are weighted and medians are unweighted.

27

ACCEPTED MANUSCRIPT

Cash Preference Percent NonPreferred Total

Debit Preference

Median Amount NonPreferred Preferred

Percent NonPreferred

CR IP T

Table 3b Occurrence and median transaction size for using preferred and non-preferred payment method, by merchant category and location Credit Preference

Median Amount NonPreferred Preferred

Percent NonPreferred

Median Amount NonPreferred Preferred

35.7

$

21.75

$

10.00

44.2

$

10.58

$

20.38

43.6

$

10.00

$

25.28

Food

32.9

$

19.29

$

8.80

39.0

$

8.37

$

18.06

42.6

$

9.07

$

19.73

Auto and Vehicle Related

37.8

$

21.00

$

13.00

37.9

$

17.74

$

20.04

25.6

$

14.00

$

25.38

General Merchandise

41.5

$

27.90

$

6.86

40.7

$

12.85

$

25.00

35.6

$

9.83

$

33.59

Entertainment and Transportation

11.7

$

29.36

$

16.00

65.7

$

10.00

$

20.00

59.5

$

14.00

$

20.00

Housing Related

47.6

$

54.19

$

6.40

26.7

$

1.50

$

64.59

41.9

$

5.58

$

11.09

Medical, Education, Person Services

37.4

$

25.00

$

10.00

37.2

$

20.00

$

23.80

92.6

$

19.00

$

27.91

Financial, Professional, Miscellaneous Services

0.0

$

$

50.00

47.2

$

150.00

$

25.98

44.5

$

100.00

$

119.62

Government and Nonprofit

57.2

$

4.60

$

7.50

74.9

$

20.00

$

25.20

68.0

$

12.65

$

40.00

Gifts and Transfers to People

3.5

$

10.00

$

20.00

95.8

$

20.00

$

69.24

99.0

$

20.00

$

32.00

I don't know

50.4

$

In Person

32.9

$

Not In Person

88.7

$

M

ED 3.75

$

5.25

87.3

$

3.00

$

27.44

73.7

$

8.00

$

6.50

22.36

$

10.00

43.3

$

10.00

$

20.00

44.4

$

9.82

$

24.45

$

5.56

50.5

$

38.95

$

30.00

36.2

$

37.25

$

38.50

PT

Location

-

AN US

Merchant Type

19.37

AC

CE

Source: 2015 Diary of Consumer Payment Choice. Note: Purchases only. Percent columns indicate percentage of transactions where non-preferred payment method was used. Median Amount columns refer to the amount of the transaction. Means are weighted and medians are unweighted.

28

ACCEPTED MANUSCRIPT

Table 4 Diarist-level probit regressions estimating the marginal probability of preferring cash over cards (column 1) and credit cards over debit cards (column 2), demographics only

Income [$50K-$75K omitted] Employment Status(3)

Employed Number of Observations Pseudo R-squared

* p<0.10 ** p < 0.05 *** p < 0.01

Credit Preference(2) -0.002 0.000 -0.339 *** -0.160 *** -0.051 0.149 *** -0.178 *** 0.036 0.103 *** -0.078 0.027 0.144 ** 0.217 ***

CR IP T

Ethnicity Race Gender

Cash Preference(1) 0.002 0.000 0.338 *** 0.202 *** 0.110 *** -0.112 *** 0.113 ** -0.052 0.083 *** 0.105 ** -0.015 -0.101 *** -0.076 **

AN US

Education [College omitted]

Variables Age Age * Age Less than High School High School Some College Graduate Degree Latino White Male < $25,000 $25,000-$49,999 $75,000-$99,999 >= $100,000

-0.056 * 1043 0.152

M

Categories Age

-0.130 *** 798 0.126

Source: Authors’ calculations based on the 2015 Diary of Consumer Payment Choice.

ED

Note: All diarists who prefer cash, debit cards, or credit cards for purchases (Cash Preference column) and all diarists who prefer credit cards or debit cards (for Credit Preference column). Marginal effects are calculated at the means. (1) Cash Preference = 1 if prefer cash over credit or debit cards for purchases

PT

(2) Credit Preference = 1 if prefer credit cards over debit cards for purchases (3) Employment status variable = 1 if working now, = 0 otherwise.

AC

CE

Model specifications: Pr(CASHPREFi) = f(Xi) Pr(CCPREFi) = f(Xi), where Xi are demographic variables.

29

ACCEPTED MANUSCRIPT

Table 5a Summary statistics for preferences and reason for deviation for purchases Cash

Debit

Credit

% who prefer PI % who used preferred PI % who deviated from preferred PI

23.9%

46.1%

27.3%

64.1% 35.9%

55.6% 44.4%

56.4% 43.6%

17.6% 17.1% 32.0% 9.4%

32.0% 20.8% 4.4% 9.4%

33.4% 18.0% 3.1% 17.4%

Top reasons for deviating(1) Transaction size Merchant type Payment Instrument not on hand Speed of payment

AN US

Source: 2015 Diary of Consumer Payment Choice. Purchases only.

CR IP T

Payment Instrument (PI)

Note:

M

(1) Respondents were asked to choose a reason if they deviated from their preferred payment instrument. These statistics are based solely on respondents' answers to this question.

Table 5b Number of transactions, by deviation from preferred to another payment method Number of Transactions

ED

Prefer cash but deviated to … credit card debit card Prefer debit card but deviated to … cash credit card Prefer credit card but deviated to … cash debit card

PT

148 153

CE

810 227

AC

587 138

Source: 2015 Diary of Consumer Payment Choice. Purchases only.

30

ACCEPTED MANUSCRIPT

Table 6a Transaction-level probit regressions estimating the marginal probability of deviating from the preferred payment method.

(2)

Payment on hand Age

Education [College omitted] Ethnicity Race Gender Income [$50K-$75K omitted]

-0.417 ***

PT

Employment Status

(3)

-0.535 ***

From Credit 0.000 0.000 -0.184 *** -0.038 -0.276 *** -0.003 *** -0.019 *** 0.000 * 0.073

CR IP T

Location

From Debit -0.001 *** 0.000 *** -0.248 *** -0.270 *** -0.269 *** -0.001 *** -0.001 0.001 -0.026

-0.024 *** 0.000 **

0.003 0.000

AN US

Merchant * Amount

From Cash 0.003 ** 0.000 *** 0.038 -0.030 0.115 * 0.000 0.012 *** 0.000

M

Merchant

Variables Amount Amount * Amount Food Auto General Food * Amount Auto * Amount General * Amount In person Enough Cash(1) Debit Card Credit Card Age Age * Age Less than High School High School Some College Graduate Degree Latino White Male < $25,000 $25,000-$49,999 $75,000-$99,999 >= $100,000 Employed Number of Transactions Pseudo R-squared

ED

Categories Amount

-0.172 -0.101 -0.134 -0.017 -0.034 0.070 0.044 -0.200 0.048 -0.015 0.125 -0.013 1023 0.260

*** ** ***

*

***

**

-0.073 -0.061 * -0.015 -0.030 0.004 0.025 0.023 0.036 0.019 -0.065 ** -0.010 0.002 2659 0.075

-0.327 *** 0.008 0.000 0.376 0.073 0.050 -0.017 -0.122 -0.007 -0.048 0.062 0.096 0.051 0.096 -0.027

*

* ** ***

2000 0.102

CE

* p < 0.10 ** p < 0.05 *** p < 0.01

Source: 2015 Diary of Consumer Payment Choice and authors' calculations.

AC

Note: Purchases only. For example, “From Cash” regression includes all transactions associated with diarists who prefer cash. Dependent variable = 1 if transaction is conducted with payment method different from cash and = 0 otherwise. The numbers represent marginal effects. Marginal effects are calculated at the means. (1) Enough Cash variable = 1 if the amount of the transaction is smaller than or equal to the respondents' cash balance at the specific time of this transaction. Respondents reported their cash balance at the start of every diary day. That amount is adjusted throughout the day by subtracting the amount of any cash payments or deposits, and by adding the amount of any cash additions, such as ATM withdrawals. (2) Each diary day respondents were asked what payment instruments they carried. (3) Employment status variable = 1 if working now, = 0 otherwise. Model specifications: Pr(FROM_CASHij) = f (AMOUNTj, MERCHANTj, AMTMERCHj, ENOUGH_CASHij, Xi) Pr(FROM_DEBITij) = f (AMOUNTj, MERCHANTj, AMTMERCHj, DCij, Xi) Pr(FROM_CREDITij) = f (AMOUNTj, MERCHANTj, AMTMERCHj, CCij, Xi)

31

ACCEPTED MANUSCRIPT

From Cash

Variables

to Credit

Location Payment on hand(2) Age

Education [College omitted] Ethnicity Race Gender Income [$50K-$75K omitted] Employment Status(3)

Enough Cash(1)

-0.028

Debit Card Credit Card Age Age * Age Less than High School High School Some College Graduate Degree Latino White Male < $25,000 $25,000-$49,999 $75,000-$99,999 >= $100,000 Employed

-0.017 0.000 -0.056 -0.048 -0.036 0.085 -0.052 0.035 0.028 -0.066 0.034 0.000 0.050

Number of Transactions Pseudo R-squared

CE

*** ***

* ***

0.035 1023 0.237

to Cash

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

From Credit

to Credit

-0.003 0.000 -0.205 -0.189 -0.155 -0.001 -0.004 -0.002 0.266

*** *** *** *** ***

to Cash

*** ***

** *** ***

-0.362

***

-0.058

*

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

0.011 0.000 -0.032 0.041 -0.007 -0.027 0.066 0.012 0.013 0.004 0.045 -0.064 -0.017

***

-0.008 0.000 0.003 -0.041 -0.012 0.006 -0.006 0.012 0.003 -0.028 -0.022 -0.025 0.002

*** ***

0.000

0.024

-0.013

-0.017

-0.021

1023 0.338

2659 0.166

2659 0.130

2000 0.179

2000 0.079

*** **

* **

-0.001 0.000 -0.143 0.061 -0.209 -0.003 -0.023 0.000 0.245

*** *** ***

-0.138 0.019 0.000 0.457 0.036 0.017 0.001 -0.002 0.033 -0.028 0.042 0.068 0.066 0.061

** *** *** *

to Debit

0.001 0.000 0.033 0.176 0.055 0.000 0.000 0.000 -0.050

*** *** *** ***

0.000 0.000 0.038 0.073 0.050 0.000 0.000 0.000 0.020

**

-0.152 -0.006 0.000 0.011 0.035 0.039 0.021 -0.038 -0.010 -0.001 0.015 0.013 -0.020 0.005

** * *

*** *** *** *** * * *** * * ***

*

*

* **

***

** ** **

* * **

* p < 0.10 ** p < 0.05 *** p < 0.01

Source: 2015 Diary of Consumer Payment Choice and authors' calculations. Notes: Purchase transactions only. For example, “From Cash to Credit” regression includes all transactions conducted by diarists who prefer cash but used credit card instead. Dependent variable = 1 if transaction is conducted with credit card and = 0 otherwise. The numbers represent marginal effects. For other notes, see notes under Table 6a.

AC

* *

-0.001

M

Merchant * Amount

0.001 0.000 0.045 -0.005 0.084 0.000 0.004 0.001

ED

Merchant

Amount Amount * Amount Food Auto General Food * Amount Auto * Amount General * Amount In person

PT

Amount

From Debit to Debit

AN US

Categories

CR IP T

Table 6b Transaction-level probit regressions estimating the marginal probability of deviating from the preferred payment method to another specific payment method

32

* *

ACCEPTED MANUSCRIPT

Percent Yes 1.6

Total

Cash Discount Median Amount Yes No $ 20.30 $ 7.66

Percent Yes 2.0

Debit Discount Median Amount Yes No $ 29.32 $ 20.38

Age 0.0 1.0 2.8 1.9 1.5 1.6

$ $ $ $ $ $

29.00 20.00 30.00 21.53 15.00

$ $ $ $ $ $

8.84 8.93 7.01 7.45 7.00 8.12

1.1 1.5 4.8 0.9 1.7 0.9

< 25k 25-49k 50-74k 75-99k 100-124k 125-199k > 200k

2.2 1.8 1.3 0.9 2.5 1.2 0.4

$ $ $ $ $ $ $

21.53 20.00 29.50 30.00 21.00 22.06 3.50

$ $ $ $ $ $ $

7.51 8.67 8.00 6.00 8.10 7.06 7.03

1.9 0.8 1.9 3.8 1.1 4.0 1.2

Less HS High School Some College College Graduate

1.3 1.3 1.1 1.7 3.2

$ $ $ $ $

19.50 25.00 21.53 20.60 9.75

$ $ $ $ $

10.00 10.00 7.54 7.76 6.73

3.4 1.2 1.7 2.4 2.8

Male Female

1.3 1.9

$ $

20.30 20.77

$ $

7.06 8.00

Latino Not Latino

4.1 1.2

$ $

22.80 20.00

White Black Asian Other

1.3 2.8 1.8 4.1

$ $ $ $

20.00 10.00 25.00 22.00

Income

Ethnicity

$ $ $ $ $ $

12.84 19.12 21.59 20.08 22.75 24.85

0.0 1.1 1.8 0.4 0.4 0.7

$ $ $ $ $ $

42.00 94.50 23.68 15.49 58.28

$ $ $ $ $ $

20.00 28.00 24.99 25.50 27.40 31.00

4.42 23.00 65.81 19.90 50.96 68.01 148.73

$ $ $ $ $ $ $

20.00 21.71 18.96 20.36 23.55 21.39 25.42

1.5 0.4 0.3 1.0 0.0 2.2 0.0

$ $ $ $ $ $ $

54.56 75.56 43.00 50.14 23.68 -

$ $ $ $ $ $ $

20.00 23.59 26.00 30.53 30.36 28.49 25.00

$ $ $ $ $

44.36 15.14 20.00 21.00 66.45

$ $ $ $ $

31.50 24.55 20.00 20.02 20.00

0.0 2.0 0.2 0.8 0.6

$ $ $ $ $

40.00 28.85 42.00 62.94

$ $ $ $ $

44.72 29.62 26.30 27.40 26.75

$ $

22.86 33.21

$ $

20.58 20.05

0.7 1.0

$ $

26.27 43.00

$ $

26.18 28.00

$ $

8.78 7.52

1.3 2.2

$ $

32.24 29.31

$ $

20.00 20.55

6.1 0.4

$ $

40.00 48.78

$ $

22.50 27.60

$ $ $ $

7.50 7.29 7.45 10.00

1.4 4.0 4.4 4.9

$ $ $ $

29.33 14.28 33.50 17.76

$ $ $ $

20.38 18.81 27.29 21.71

1.0 0.0 0.0 0.3

$ $ $ $

41.50 42.00

$ $ $ $

28.00 24.00 25.66 22.50

PT

Race

1.0 3.0

ED

Gender

20.00 44.35 27.51 17.76 87.25 37.31

Credit Surcharge Median Amount Yes No $ 42.00 $ 27.15

$ $ $ $ $ $ $

M

Education

$ $ $ $ $ $

Percent Yes 0.8

AN US

Under 25 25-34 35-44 45-54 55-64 Over 64

CR IP T

Table 7a Occurrence and median transaction size for receiving a discount or paying a surcharge, by demographic groups

AC

CE

Source: 2015 Diary of Consumer Payment Choice. Note: Purchases only. Percent columns indicate percentage of transactions with a discount or a surcharge. Median Amount columns refer to the amount of the transaction. Percent means are weighted and median amounts are unweighted.

33

CR IP T

ACCEPTED MANUSCRIPT

Table 7b Occurrence and median transaction size for receiving a discount or paying a surcharge, by merchant category and location Percent Yes Total

Cash Discount Median Amount Yes No

1.6

$ 20.30

$

7.66

Food

0.7

$ 33.62

$

8.00

Auto and Vehicle Related

8.0

$ 20.00

$ 10.00

General Merchandise

1.3

$ 22.50

$

Entertainment and Transportation

0.0

$

-

$ 13.00

Housing Related

0.0

$

-

$

Percent Yes

Debit Discount Median Amount Yes No

2.0

$ 29.32

0.4

$ 24.00

0.0

$

Government and Nonprofit

0.4

$

Gifts and Transfers to People

0.0

$

I don't know

1.5

In person

1.6

Not In Person

6.0

$ 11.00

$ 42.00

$ 27.15

$ 17.26

$ 18.52

1.3

$ 40.00

$ 20.64

1.5

$ 23.00

$ 20.02

1.2

$ 58.28

$ 25.18

4.3

$ 37.60

$ 24.99

0.3

$ 23.68

$ 39.21

0.0

$

-

$ 25.32

0.0

$

-

$ 20.00

0.0

$

-

$ 39.58

0.0

$

-

$ 15.38

0.0

$

-

$ 21.38

15.1

$ 95.50

$ 29.22

$ 75.00

0.0

$

-

$ 22.99

0.0

$

-

$ 462.31

$ 10.00

0.0

$

-

$ 27.60

0.0

$

-

$ 18.00

$ 20.00

0.0

$

-

$ 69.24

0.0

$

-

$ 32.00

2.00

$

5.00

0.0

$

-

$ 23.72

0.0

$

-

$

$ 20.60

$

7.67

2.0

$ 25.36

$ 20.00

1.0

$ 42.00

$ 25.53

$ 20.00

$

8.00

2.4

$ 83.02

$ 29.25

0.0

$

$ 39.96

3.50 -

ED $

PT

Location

-

4.67

0.9

1.0

M

Medical, Education, Person Services Financial, Professional, Miscellaneous Services

5.44

$ 20.38

AN US

Merchant Type

Credit Surcharge Percent Median Amount Yes Yes No

-

6.50

AC

CE

Source: 2015 Diary of Consumer Payment Choice. Note: Purchases only. Percent columns indicate percentage of transactions with a discount or a surcharge. Median Amount columns refer to the amount of the transaction. Percent means are weighted and median amounts are unweighted. Observations with missing merchant information are excluded.

34

ACCEPTED MANUSCRIPT

Table 8 Deviating from preferred payment method because of a discount or a surcharge (percentage of transactions) Received discount

Avoided surcharge

Cash

348

3.45%

0.00%

Debit

1035

3.86%

1.05%

Credit

741

2.14%

AC

CE

PT

ED

M

AN US

Source: 2015 Diary of Consumer Payment Choice. Purchases only.

CR IP T

Number of transactions

Deviated from

35

0.97%

ACCEPTED MANUSCRIPT

Table 9 Transaction-level probit regressions estimating the marginal probability of deviating to a specific payment instrument from a preferred payment instrument

Merchant

Merchant * Amount Location Age

Education [College omitted]

0.000 0.000 -0.065 ** -0.193 *** -0.069 ** -0.001 -0.001 0.000 0.001 0.000 -0.441 -0.237 -0.129 0.153 -0.131 0.028 -0.123 -0.108 -0.022 0.107 0.058

Income [$50K-$75K omitted]

Employed Number of Transactions Pseudo R-squared

*** *** *** *** *** *** *** *** *

0.043 *

PT

Employment Status(1)

ED

M

Ethnicity Race Gender

To Cash 0.192 ***

0.072 -0.001 0.000 -0.019 0.058 -0.017 0.001 -0.001 0.000 -0.003 -0.005 0.000 0.096 0.090 0.034 0.024 -0.006 -0.024 0.084 0.051 0.143 -0.023 0.089

***

*** *** ***

-0.092 *** 1837 0.062

CE

2150 0.151

To Debit

CR IP T

Amount

Variables Cash Discount Debit Discount Amount Amount * Amount Food Auto General Food * Amount Auto * Amount General * Amount In person Age Age * Age Less than High School High School Some College Graduate Degree Latino White Male < $25,000 $25,000-$49,999 $75,000-$99,999 >= $100,000

AN US

Categories Discounts/Surcharges

* p < 0.10 ** p < 0.05 *** p < 0.01

AC

Source: 2015 Diary of Consumer Payment Choice and authors' calculations.

Note: Purchases only. For example, “To Cash” regression includes all cash transactions. Dependent variable = 1 if transaction was conducted with cash by a respondent who does not already prefer cash and = 0 otherwise. Marginal effects are calculated at the means. (1) Employment status variable = 1 if working now, = 0 otherwise. Model specifications: Pr(TOCASHij) = f (DISCOUNTkj, AMOUNTj, MERCHANTj, AMTMERCHj, Xj) Pr(TODEBITij) = f (DISCOUNTkj, AMOUNTj, MERCHANTj, AMTMERCHj, Xj) k = cash or debit card.

36

ACCEPTED MANUSCRIPT

CR IP T

Figure 1a Probability of cash use, by payment preference as a function of transaction size

ED

M

AN US

Figure 1b Probability of debit card use, by payment preference as a function of transaction size

AC

CE

PT

Figure 1c Probability of credit card use, by payment preference as a function of transaction size

Source: 2015 Diary of Consumer Payment Choice. Purchases only. Note: The lines indicate estimated probabilities. Confidence intervals are available from the authors.

37

ACCEPTED MANUSCRIPT

AN US

CR IP T

Figure 2a Why do people deviate from cash?

AC

CE

PT

ED

M

Figure 2b Why do people deviate from debit cards?

Source: 2015 Diary of Consumer Payment Choice. Purchases only.

38

ACCEPTED MANUSCRIPT

AN US

CR IP T

Figure 2c Why do people deviate from credit cards?

Source: 2015 Diary of Consumer Payment Choice. Purchases only.

AC

CE

PT

ED

M

Figure 3 Fraction of transactions with discounts and surcharges, by transaction value

Source: 2015 Diary of Consumer Payment Choice. Purchases only. $50 includes all transactions greater than $50.

39

ACCEPTED MANUSCRIPT

Appendix A: Preferences for in-person vs. online purchases, and purchases vs. bills

ED

M

AN US

CR IP T

Figure A1 Preferences: in-person by transaction value and online purchases

Source: 2015 Diary of Consumer Payment Choice.

AC

CE

PT

Note: Other payment instruments includes check, prepaid/gift/EBT card, bank account number payment, online banking bill payment, money order, and mobile phone payment.

40

ACCEPTED MANUSCRIPT

M

AN US

CR IP T

Figure A2 Preferences: bills vs. purchases

ED

Source: 2015 Diary of Consumer Payment Choice

AC

CE

PT

Note: Other payment instruments includes prepaid/gift/EBT card, money order, and mobile phone payment.

41

ACCEPTED MANUSCRIPT

M

AN US

CR IP T

Figure A3 Actual use: in-person by transaction value

ED

Source: 2015 Diary of Consumer Payment Choice

AC

CE

PT

Note: Other payment instrument includes check, prepaid/gift/EBT card, bank account number payment, online banking bill payment, money order, PayPal, account-to-account transfer, mobile phone payment, and deduction from income.

42

ACCEPTED MANUSCRIPT

Appendix B: Why do high-income consumers deviate from credit cards? All figures in this section are generated using non-credit card purchases made by credit cardpreferring respondents with annual household income of $100,000 or more.

AN US

CR IP T

Figure B1 Distribution of transaction sizes.

Source: 2015 Diary of Consumer Payment Choice. The $50 bucket includes all transactions greater than $50.

AC

CE

PT

ED

M

Figure B2 Percentage of transactions across merchant categories.

Source: 2015 Diary of Consumer Payment Choice.

43

ACCEPTED MANUSCRIPT

AN US

CR IP T

Figure B3 Reasons for deviating from credit cards (preferred payment method).

M

Source: 2015 Diary of Consumer Payment Choice.

AC

CE

PT

ED

Figure B4 Payment method used to complete transaction.

Source: 2015 Diary of Consumer Payment Choice.

44