ArticlePDF Available

The Economics of a Cashless Society: An Analysis of the Costs and Benefits of Payment Instruments

The Economics of a Cashless Society:
An Analysis of the Costs and Benefits of Payment Instruments
Daniel D. Garcia Swartz, Robert W. Hahn and Anne Layne-Farrar*
Related Publication 04-24
September 2004
* Mr. Garcia Swartz is a Senior Managing Economist with LECG. Mr. Hahn is cofounder and executive
director of the American Enterprise Institute-Brookings Joint Center for Regulatory Studies. Ms. Layne-
Farrar is a Senior Managing Economist with LECG. We would like to thank Howard Chang, David
Evans, David Humphrey, Randy Kroszner, Pia Orrenius, Grete Øwre, Richard Schmalensee, Joanna
Stavins, and Scott Wallsten for helpful comments and Melissa Long and Nese Nasif for excellent research
assistance. The authors would like to thank the AEI-Brookings Joint Center and Visa U.S.A. for
supporting this research. This paper reflects the views of the authors and not necessarily those of their
affiliated institutions.
Executive Summary
Ever since Diners Club introduced the first general-purpose charge card in the early
1950s, pundits have been predicting the emergence of a “cashless society.” Fifty years later, we
still use cash, but it is much less important than it once was. Today, there are a wide array of
payment instruments, including cash, checks, several different types of payment cards, and
electronic transfers. Moreover, there is a heated debate as to whether government should regulate
various kinds of private payment systems, such as credit cards.
As a guide for examining policy questions, this paper provides an economic approach for
assessing the relative costs and benefits of various payment methods. It is the first study to
examine empirically the move toward a cashless society using cost-benefit analysis.
We present three case studies that illustrate the welfare implications of substituting one
kind of payment method for another. We find that when all key parties to a transaction are
considered and benefits are added, cash and checks are more costly than many earlier studies
suggest. In general, the shift toward a cashless society appears to be a beneficial one.
The Economics of a Cashless Society:
An Analysis of the Costs and Benefits of Payment Instruments
Daniel D. Garcia Swartz, Robert W. Hahn and Anne Layne-Farrar
Over the course of history, there have been many different forms of payment systems.
Originally, barter was quite common. Eventually, various forms of money were introduced.1 In
the mid-twentieth century, charge cards debuted.2 Ever since then, pundits have been predicting
the demise of paper instruments and the emergence of a “cashless society.”3 Today, we still pay
with cash and checks, but several other payment instruments, such as credit and debit cards, are
widely used. The use of paper money is declining, but at a slow pace. In this paper, we analyze
the costs and benefits of using different payment instruments that have affected the shift toward
a cashless society.
As more payment systems have been introduced, researchers have begun to critically
examine their costs from both a private and social perspective.4 From a private perspective,
researchers have examined the incentives payers have for choosing a given type of payment
instrument, the incentives retailers may have for accepting such instruments, and why various
payment methods are used in different settings. From a social perspective, researchers have
examined whether economic welfare would increase if certain payment instruments displaced
others––such as, if electronic instruments displaced paper-based instruments. Our study is the
first to examine empirically the move toward a cashless society using a framework that
includes benefits as well as costs.
1 For a history of the evolution of money see, e.g., Glyn Davis, A History of Money: From Ancient Times to the
Present Day, Cardiff, U.K.: University of Wales Press, 2002.
2 Diners Club, introduced in 1950, was the first general purpose (as opposed to store-specific) charge card. David
Evans & Richard Schmalensee, Paying With Plastic: The Digital Revolution in Buying and Borrowing,
Cambridge, MA: MIT Press (forthcoming, 2005).
3 As a 1958 U.S. News & World Report article on the then-new Diners Club credit card stated, “Before long, a
man will be able to live or travel almost anywhere with only a few dollars in his pocket. Instead of paying cash
as he goes, he will simply settle by check for just about everything at the end of the month.” “What You Now
Can Do Without Cash,” U.S. News & World Report, October 24, 1958.
The interest in the economics of payment instruments has been more than academic. For
example, two recent cases centered on whether payment cards are “overused” in a social
welfare sense: the investigation of MasterCard by the Office of Fair Trading in the United
Kingdom, and the Australian central bank investigation of MasterCard and Visa.5 In brief, the
regulators claim that payment card systems charge “unjustifiably high” fees to merchants for
payment cards while the banks issuing payment cards provide consumers with below-cost
services and loyalty rewards.6 Thus, by not charging consumers the full marginal cost imposed
by their card use, the regulators assert that payment card systems encourage consumers to
overuse payment cards, which they can afford to do profitably because merchants shoulder the
In the United States, the scrutiny has been no less intense, but has emphasized different
pricing aspects of the payment card systems. In the widely publicized Wal-Mart case, for
example, the plaintiffs argued that debit cards were illegally tied to credit cards––that is,
merchants accepting credit cards also had to accept debit cards. They claimed that this tie
enabled the payment card systems to charge “exorbitant” fees to merchants, although they did
not argue that the end result was debit card overuse.7 At the other extreme in terms of policy
implications, New York Attorney General Eliot Spitzer is considering opening an investigation
on the commercial bank practice of charging consumers per-transaction fees for PIN debit
4 We define “private” costs and benefits as those faced by only one side of a transaction, such as a merchant payee
or a consumer payer. “Social” costs and benefits are those incurred by the economy as a whole, including all
parties to a transaction.
5 “MasterCard Interchange Fees: Preliminary Conclusions,” Office of Fair Trading (OFT634), February 2003, at 3.
Reserve Bank of Australia, “Reform of Credit Card Schemes in Australia IV: Final Reforms and Regulation
Impact Statement,” August 2002. The Australian central bank’s expert economist argued, “distortions in credit
and charge card prices can lead to excessive card use.” Michael L. Katz, “Network Effects, Interchange Fees and
No-Surcharge Rules in the Australian Credit and Charge Card Industry,” August 2001, at 2, ¶5. The ruling by the
Reserve Bank of Australia, Australia’s central bank, forced Visa and MasterCard to cut their interchange fees by
40 percent. “Visa Throws in Towel on New Card Rules,” New Zealand Herald, October 30, 2003.
6 Merchants pay a fee, referred to as the “merchant discount,” to the banks that process their payment card
transactions. Merchant’s banks must then pass on a portion of that fee, referred to as the “interchange fee,” to
cardholders’ banks. Thus, merchants indirectly pay card issuing banks for each payment card transaction.
7 Second Amended Consolidated Class Action Complaint and Jury Demand, In re Visa Check/MasterMoney
Antitrust Litig. (E.D.N.Y. 2000) (No. CV-96-5238). The case settled before the jury trial began. In the
settlement, MasterCard and Visa agreed to pay several billion dollars in cash and to not require merchants to take
their debit cards as a condition of taking their credit cards.
cards. While it is unclear what Spitzer’s plans are, he appears to be concerned that consumers
pay too much to use PIN debit payment cards.8
While all of these policy issues are worth examining, we would argue that most
attempts at answering them have been incomplete because they do no adequately consider key
parties in a transaction or economic benefits. This paper has two main objectives: first to
develop a method for evaluating payment instruments and second to apply that method. Given
the regulatory activity on this issue, a sensible cost-benefit approach is needed. The empirical
applications we develop illustrate how cost-benefit analysis can be used. They also provide
policy makers with insights on how moves toward a cashless society affect net economic
welfare. In particular, we find that merchant cost studies suggesting that payment cards,
especially credit cards, are very costly are misleading when used to argue that payment cards
are costly for the economy as a whole. When other parties to a transaction are considered and
benefits are included, payment cards appear competitive with other forms of payment available
to consumers. Our results suggest that the slow but inevitable shift toward a cashless society is
beneficial, providing certain groups with clear benefits while not costing more to process in the
aggregate than traditional paper transactions.
The paper proceeds as follows. Section II offers guidelines for assessing the costs and
benefits of payment instruments. In Section III, we evaluate the literature on payment
instruments in light of these guidelines. A significant portion of the literature has focused solely
on the costs of payment instruments. Some studies have counted elements of cost in an
inconsistent manner and others have focused on the costs of payment instruments for just one
party to a transaction, notably merchants. As a result, the literature provides only limited
guidance on the desirability of substituting certain payment instruments for others. To better
understand the economics of the transition from cash and checks to cards, we develop case
studies in Section IV for three retail environments: grocery stores, discount stores, and
electronics specialty stores. Our calculations focus on providing rough estimates of marginal
costs and benefits for an additional transaction by an average consumer. The case studies
8 David Breitkopf, “What Are N.Y. PIN-Debit Fee Probers After?” American Banker, October 6, 2003.
present evidence on the incentives driving the move toward a cashless society. Section V
presents our main conclusions.
This section of the paper is intended as background for our later analysis. Readers with
economics training will find it quite familiar. Our method is based on three fundamental
principles. First, a payer tends to choose a payment system that maximizes her net benefits
from a particular transaction. Her choice is constrained by available instruments, such as a
previous decision to apply for a credit card, and by payees’ decisions about what instruments to
accept.9 Payees make choices by comparing the net benefits of accepting a particular
instrument with those of not accepting it. Interestingly, much of the economic literature on
payment systems focuses solely on payees’ costs, even though this focus can be misleading for
ranking the relative merits of payment systems from both an individual and a social
Second, all parties involved in a transaction should be included in a cost-benefit
analysis. Focusing on the costs and benefits to a payer or payee can help explain why a payer
uses or a payee accepts a particular instrument. But even when separately analyzing the
incentives for use or acceptance it is appropriate to analyze both sides of a transaction. A
payer’s decision to carry a payment instrument will be affected by how many payees accept it,
and a payee’s decision to accept it will be affected by how many payers carry and use it.10
Focusing on the net benefits to all parties in a transaction is necessary when examining the
9 To clarify the concepts necessary for analyzing the costs and benefits of different payment methods, we initially
assume that payers choose among available payment options. That is, the availability of payment instruments is
assumed to be exogenous. An individual payer is largely responsible for determining the payment method,
selecting an instrument from among those she has access to and that the payee accepts. Note that consumers
initiate more than 90 percent of all transactions, so starting from a payer’s choice of instruments and then
factoring in acceptance constraints is a natural approach. Brian Mantel, “Why Don’t Consumers Use Electronic
Banking Products? Toward a Theory of Obstacles, Incentives, and Opportunities.” Federal Reserve Bank of
Chicago, Emerging Payments Occasional Paper Series EPS-2000-1, September 2000, at 8. This simplifying
assumption is dropped later on when we address individual decisions by payees to accept a given method and by
payers to carry a given method.
10 Payment instruments operate in a two-sided market. Payment instrument providers need to coordinate the
demands of two (or more) distinct groups of customers, payers and payees. See the discussion in David S. Evans,
“The Antitrust Economics of Multi-Sided Platform Markets,” Yale Journal on Regulation, Vol. 20, 2003, at 357-
social net benefits from a transaction. In some cases, a payer’s benefit may represent a cost to
payees. Float is an example: the private costs and benefits tend to cancel out in a social net
benefit calculation. It is also important to consider the externalities imposed on other parties by
some payment instruments. For instance, because paper checks tend to take longer to process at
the point of sale than other forms of payment, they impose relatively higher opportunity costs
on anyone waiting in line behind a payer.11
Third, all significant components of costs and benefits should be considered when
comparing payment instruments. This includes both explicit costs (such as bank fees) and
implicit costs (such as opportunity costs). Some costs are certain and unavoidable, such as
envelope and postage costs for mailing a check to pay a bill. Others are not, such as the
possible loss of cash if a person is robbed on the way to a store. Regardless of the underlying
nature, all significant costs and benefits should be included in a reasonable assessment of net
A. The Parties Involved
Consider a simple transaction in which a consumer wants to purchase a good at a store.
There are always at least two parties to this transaction—the consumer and the store. Both
participate directly, not only because the consumer purchases a good that the store supplies, but
also because both choose a given payment instrument to consummate the transaction. The store
chooses to accept a set of payment methods and the consumer chooses among them.
Even the simplest payment mechanisms involve numerous indirect parties, although
some operate behind the scenes. For instance, cash has to be produced and distributed. In fact,
in the late 1990s the American government spent roughly $700 million a year to buy new notes
and coins from the Bureau of Engraving and Printing and to process, store and distribute the
11 The opportunity cost of time is the value of time in the best alternative use. For example, a worker paid $10 an
hour who needs to leave work to spend one hour at a doctor’s appointment would face an opportunity cost of
$10. This represents the money he could have earned if he had not spent the time at the doctor.
12 Ideally, one would want to identify and quantify all costs and benefits to the extent reasonable. Of course, the
identification of significant costs and benefits is a matter of judgment. Moreover, the quantification of certain
costs and benefits is difficult in practice, as we discuss in the case studies.
currency.13 In order to make a cash payment a consumer must first obtain cash, say from an
ATM. This machine is maintained and serviced by a financial institution.14 These institutions
play a role at the beginning of a transaction by distributing cash prior to a sale and at the end of
the process by accepting cash deposits, say in a retailer’s bank account.
Consider what happens if a consumer pays for goods with a check. She first must have a
checking account at a financial institution and must obtain checks from a check-printing firm.
Assuming the store accepts personal checks, it will deposit the consumer’s check in its bank. If
the consumer’s and the store’s accounts are in the same bank, then the check is called “on us”
and stays within that bank.15 Otherwise, the physical check usually travels from the store’s
bank through a financial intermediary (a clearing house or a Federal Reserve Bank) to the
paying consumer’s bank. It is often then returned to the consumer at the end of the month along
with a bank statement.16
The example highlights a simple, but fundamental, point: although the payer and payee
are usually the two primary parties directly involved in a given transaction, many other parties
are involved as well. Many of these parties charge payers and payees directly for their services,
others incur costs that are not fully passed on (at least not directly). Including these other
parties is essential for calculating social costs and benefits––a point illustrated by our case
studies––but it also can be important for calculating private costs and benefits.
13 Paul W. Bauer, “Currency: Time for Change?,” Federal Reserve Bank of Cleveland Economic Commentary,
October 1998. Note that the dollar figures throughout this paper have not been adjusted to current dollars, and
are reported as the original author(s) or study reported them.
14 On the development of the ATM system, see Donald I. Baker & Roland E. Brandel, The Law of Electronic Fund
Transfer Systems: Legal and Strategic Planning, New York: Warren, Gorham & Lamont, 1998, at ch. 6.
15 On-us check transactions comprise approximately 33 percent of all checks processed. Joanna Stavins, “A
Comparison of Social Costs and Benefits of Paper Check Presentment and ECP Truncation,” New England
Economic Review, July/August 1997, at 32.
16 Stavins, “Comparison of Social Costs,” supra note 15; Joanna Stavins, “While More People Are Paying
Electronically, Many of Us Still Cling to Checks,” Federal Reserve Bank of Boston Regional Review, Vol. 11,
2002; and Stuart E. Weiner, “Electronic Payments in the U.S. Economy: An Overview,” Federal Reserve Bank
of Kansas City Economic Review, 1999. Due to recent check-imaging legislation (commonly referred to as the
Check 21 Act), more consumers will receive a statement with an image of their cancelled checks as opposed to
the physical checks themselves. Will Wade, “Gauging the Long-Term Opportunities in Check 21,” American
Banker, October 31, 2003.
B. Private Versus Social Net Benefits
While the distinction between private and social costs may seem obvious, some of the
literature to date has not dealt with the distinction clearly.17 For our purposes, private net
benefits accrue to each of the private parties (or groups) involved in a transaction, such as
merchants or consumers.18 Social net benefits include payers and payees plus indirect parties.19
Understanding private net benefits is a key to understanding the incentives that payers have for
using a particular payment instrument and that payees have for accepting it. Understanding
social net benefits is critical for assessing the overall impact of a given instrument.
Consider again a consumer purchasing a good at a store, but now assume a debit card is
used for payment. The benefits for the payer might include the decreased bulk of debit cards as
compared to cash and a reduced need to obtain cash from an ATM. If the debit card receipt is
signed for authorization (called signature debit) rather than authorized with a personal
identification number (PIN debit), the consumer is guaranteed full reimbursement for any
unauthorized debit charges to her account.20 Furthermore, signature debit, unlike PIN debit,21
benefits the consumer with up to two days of float, as the money is not immediately deducted
from her account. If, on the other hand, the transaction is processed as PIN debit, the consumer
can get cash back from the store and save another trip to the ATM.22 On the cost side, the
consumer will certainly have to wait at the store counter while the card transaction is processed,
and the wait may be slightly longer than if she had paid with cash––although it is likely shorter
than with checks.23 She may also have to pay her bank a transaction fee for using PIN debit.24
17 We discuss this in Section III.
18 Net benefits are defined as the difference between benefits and costs.
19 We focus on economic welfare and do not consider equity or distributional issues.
20 Ann H. Spiotto, “Credit, Debit, or ACH: Consequences and Liabilities. A Comparison of the Differences in
Consumer Liabilities,” Federal Reserve Bank of Chicago, Emerging Payments Occasional Paper Series EPS-
2001-3, 2001.
21 Most physical debit cards allow both PIN and signature debit transactions. How the card is read and authorized
determines the kind of debit transaction. Signature debit transactions are processed by swiping the card through a
traditional credit/charge card reader and requiring the consumer to sign the receipt; PIN debit transactions are
processed by passing the card through a PIN pad and requiring the consumer to enter her PIN for authorization.
22 Ingenico, “Online Debit: A significant opportunity for retailers to reduce their cost of business while improving
customer service,” 2002 (visited March 25, 2003) <>.
23 One of the few U.S. studies that reports time spent at the point of sale is the Food Marketing Institute. See, e.g.,
Food Marketing Institute, “Benchmarking Comparative Payment Methods: Costs and Case Studies,” 1994; Food
She likely considered these costs and benefits when she first decided to carry a debit card. The
decision to carry a debit card is often trivial for consumers, however, since many banks
automatically issue one when a customer opens an account. The more important decision for
individuals is whether to use the card at stores that accept it.
When it first decides whether to accept debit cards, a store also considers the benefits
and costs. Many customers carry and use debit cards so a store could decide to accept them as a
convenience service or to increase the likelihood of a sale. For PIN debit transactions, stores
typically receive payment the following business day; for signature debit, stores get paid within
a couple of days.25 Cash can be counterfeited or stolen from a store’s register and checks can
bounce, but theft and insufficient funds are less likely with debit cards. Processing debit card
transactions may take slightly longer than cash, but is certainly faster than processing checks.
More importantly, with debit cards a store does not have to spend resources in preparing cash
or check deposits or in transporting paper instruments to a bank, which eliminates the deposit
fee that merchants often have to pay for depositing cash or checks.26 On the other hand, a store
would have to pay a fixed per-transaction fee to a processor to clear debit card transactions.27
With signature debit transactions it typically would have to pay an additional fee to a bank, one
that is proportional to the transaction amount.28 Finally, before accepting PIN debit, a store
would have to install a PIN reader, a separate machine from the charge and credit card reader it
probably already has.
Marketing Institute, “A Retailer’s Guide to Electronic Payment Systems Costs,” 1998; and Ingenico, supra note
24 See, e.g., David Breitkopf, “Pulse: Processing Tab Main Reason for Hike,” American Banker, July 7, 2003;
Lavonne Kuykendall, “Debit Cards Gain in Popularity, Not on Bottom Line,” American Banker, June 13, 2002;
and W.A. Lee, “Can Debit Cards’ Popularity Support Reward Programs?,” American Banker, September 18,
25 Ingenico, supra note 22.
26 Retailers report paying deposit fees in the following empirical studies: Food Marketing Institute,
“Benchmarking Comparative Payment Methods,” supra note 23; Food Marketing Institute, “A Retailer’s
Guide,” supra note 23; and Coopers & Lybrand, “Survey of Retail Merchants to Determine Costs of Payment
Acceptance,” 1994.
27 The banks that handle payment card processing for retailers are typically referred to as “acquirers” (as opposed
to “issuers” that issue payment cards to consumers). Acquiring banks often sign on a merchant for a particular
payment card brand, say MasterCard, and then sometimes outsource the actual card processing to a third party.
28 Banks belonging to the MasterCard and Visa networks process signature debit cards. Evans & Schmalensee,
supra note 2.
Debit card transactions at a store impose two types of costs on indirect parties––an
important consideration for social net benefits. First, waiting for a debit transaction to be
processed at the counter generates an opportunity cost of time for the payer as well as the
people in line. As an externality, time waiting in line is not included in the private net benefit
calculations for a payer, but should be included in calculations of social net benefits. Of course,
all payment instruments generate analogous externalities at the point of sale. As such, retailers
may account for them in the price of goods sold if the externalities necessitate hiring additional
clerks, for instance. A key to determining social net benefits is assessing the relative magnitude
of the externality for each payment method; that way, incremental social net benefits can be
In addition, debit card transactions impose other costs on indirect parties that are passed
through to consumers or stores. A bank processing a debit transaction incurs processing costs,
but passes these on to the merchant through the so-called “merchant discount.”30 While
including processing costs in social net benefits of debit cards is important, note that the
merchant discount is designed, at least in part, to cover these costs. Understanding the role that
different fees play is essential in fully capturing social net costs while avoiding double
C. Careful Accounting
While the importance of considering net benefits as opposed to costs or benefits in
isolation seems apparent, in practice it is difficult to do. As a result, the debate over payment
instruments has frequently only counted costs. But an analysis that examines only one side of
29 This issue suggests the need to set a baseline for comparison. If all payment mechanisms impose externalities,
such as waiting time in line, then each payment instrument could be compared to a given option, such as cash. In
the case studies evaluated in Section IV, we compare the net costs of payment cards to the net costs of paper
30 The merchant discount is the price a merchant pays to an acquirer for the service of payment card processing.
For Visa and MasterCard transactions, the merchant discount includes the “interchange fee,” a transfer payment
that acquiring banks make to issuing banks (that is, banks that issue cards to consumers). Issuers incur a greater
proportion of costs of the system and the interchange fee serves to balance costs and demand between the issuing
and acquiring sides. Closed systems, such as American Express do not have separate issuers and acquirers. See
Richard Schmalensee, “Payment Systems and Interchange Fees,” Journal of Industrial Economics, Vol. 50,
the equation can be misleading for both private and social decision-making, a point our case
studies clarify.
Consider the private perspective in the consumer purchase example. When a consumer
decides whether to pay cash or write a check, it is clear that the relative costs of these
instruments matter. Writing a check will likely impose a higher opportunity cost of time for the
payer as well as for other people in line. But it is also clear that relative benefits matter. If, say,
having proof of payment were important to a payer, then the fact that returned checks provide
this benefit while cash does not will matter. The fact that we see checks heavily used in some
venues but not in others suggests that proof-of-payment plays a key role only in certain
circumstances. Paying with cash, on the other hand, can maintain a customer’s anonymity and
thus privacy. Without employing some other means of tracking, such as a customer loyalty
card, retailers are unable to capture and analyze a consumer’s cash purchases.
Now consider the social perspective. If we were trying to determine whether replacing
checks with debit at the point of sale increased welfare, we would have to consider net benefits
for the payer, net benefits for the store, and costs and benefits other parties incur from the use
of one payment instrument versus the other. In this calculation, some items are a benefit for one
party, but impose a cost on another participant. When a consumer writes a check at a store, she
gets the benefit of float for a few days. The merchant, however, incurs a cost from float. Even
though float tends to cancel out in social calculations, it is an important element of the
calculation because it affects private choices.31
All benefits should be considered, even though some are difficult to quantify. The
proof-of-payment component of checks, for example, may be trivial in some settings––say,
convenience stores––but is undoubtedly important in many other circumstances, such as paying
when utility bills. In fact, survey research shows that people are willing to pay to receive their
cancelled checks from their banks.32
31 To the extent that different interest rates are available to consumers as opposed to merchants, float may be
valued differently by the various parties and therefore would not completely cancel out in a net social cost
calculation. Also, if individuals expend resources to influence float (either to increase or decrease it), those
expenditures will represent a cost to society.
32 Stavins, “Comparison of Social Costs and Benefits,” supra note 15, at 38.
To the extent possible, implicit costs should be included in the cost-benefit analysis. For
example, the time spent at the point of sale while a transaction is processed represents an
opportunity cost for payers. Probabilistic costs can be important as well. For instance, there is
some chance that cash will be stolen from a consumer on the way to a store or by employees
once paid to a store. Although some costs are easier to estimate than others, seemingly second-
order costs can matter when it comes to both private decision-making and economic welfare
evaluations. One study found, for example, that the crime rate plays an important role in
determining the intensity of cash use.33 In fact, one important obstacle to developing the ATM
system in the U.S. in the late 1980s and early 1990s was increased ATM-related crime.34
D. How to Pay, How to Get Paid
Before deciding which instrument to use, a payer has to choose which instruments to
carry. On the other side of the transaction, a payee has to choose whether to accept a payment
instrument or not. This raises the issue of the difference between the acceptance decision and
the usage decision, which is linked to the distinction between marginal and total costs and
Consider once again a consumer purchasing a good at a store. Assume that at the end of
the month the store calculates the cost of accepting debit cards. The total cost for a payee to
accept debit cards is comprised of some fixed and some variable components.35 For example, in
order to accept PIN debit cards, a store has to purchase PIN pads for customers to enter their
PINs and it has to train its staff to use these machines properly. These costs are fixed—they
have to be paid in order to accept PIN debit at all, but are paid only once. Each PIN debit card
transaction also imposes a variable cost—merchants pay a per-transaction fee to the bank that
processes its payment card transactions.
33 David B. Humphrey, Lawrence B. Pulley, & Jukka M. Vesala, “Cash, Paper, and Electronic Payments: A Cross-
Country Analysis,” Journal of Money, Credit & Banking, Vol. 28, 1996.
34 Baker & Brandel, supra note 14, at 1-19ff. Bank customers sued banks seeking compensation for injuries from
crimes related to ATM use, an interesting example of payers trying to pass through to third parties one portion of
the cost of transacting in cash.
35 Variable costs are components that vary with the number of transactions or with the size of a transaction.
How much does it cost a store to process an additional PIN debit transaction? In
answering this question, fixed costs are irrelevant. The only components of cost that matter are
the variable ones, in particular the fee paid by a merchant to its processor. Thus the marginal
cost of debit is the relevant measure for questions of payment instrument use.
The distinction between marginal and total costs has implications for analyzing the net
benefits of payment instruments. The private marginal cost of using a given payment
instrument for a payee may be the lowest among all payment instruments. If, however, the
fixed costs of initial acceptance are relatively large, payees may be reluctant to adopt this new
payment method.36 Questions of payment method acceptance thus require analyzing total, not
just marginal, costs and benefits. Therefore, when we talk about costs and benefits it is worth
clarifying whether fixed costs are included and whether a given cost is fixed or variable at the
time of evaluation.
The availability decision can be more significant for one party versus another depending
on the payment instrument. First, consider cash. Merchants have been accepting cash basically
since cash came into being as a payment instrument, although not always at par.37 Since cash is
legal tender, American merchants have given little thought to the issue of accepting it. For
present-day U.S. payers, however, deciding whether or not to carry cash, or deciding how much
to carry, is far from trivial. They face a fixed cost of obtaining cash: explicit ATM fees, the
opportunity cost of the trip to the ATM, and, prior to the expansion of the ATM system, the
larger time cost of going to a bank to obtain cash. Now, consider payment cards. History shows
that convincing merchants to accept various payment cards has been a far more substantial
issue than convincing consumers to carry and use them. Merchants have had to make
investments in card-processing technology and staff training, whereas consumers have often
obtained cards for free from issuers and have even received rewards for using them.38
36 If some parties face higher costs or fewer benefits from switching than others, they can prevent the entire system
from changing. Electronic check processing may suffer this fate. Stavins, “Comparison of Social Costs,” supra
note 15.
37 That is, at various points in time certain currencies have traded at less than their face value.
38 This is not say that the card availability decision has been totally costless for payers—some credit cards, for
example, charge annual fees, clearly a fixed cost of paying with cards. And using debit cards instead of checks
In our case study analysis, we focus on marginal costs and benefits. That is, we examine
the costs and benefits of an incremental transaction and do not include fixed costs.
E. Different Instruments Serve Different Needs
In examining payment instrument use across different transaction sizes and in different
environments, it is clear that many different methods are used for very similar purchases.
Figure 1 shows the percentage of 2001 sales volume for consumer-to-business transactions paid
for by payment cards and paper instruments, broken out by transaction size.39 As the chart
illustrates, several different payment instruments are employed at each transaction size.
has required consumers to relinquish some float. For more detail on the evolution of payment cards, see Evans &
Schmalensee, supra note 2.
39 The Visa U.S.A. survey data employed here tracks ownership and usage of payment cards and other payment
mechanisms for households over time. Between 4,000 and 6,000 households are interviewed each quarter, and
the results are representative of households with incomes over $10,000 that also own at least one plastic card
(store or general purpose). Households are surveyed once per quarter for their spending patterns and card
ownership for the three months of the quarter. Thus, each household will have at most four units of observation
each year (and some will have less than four, due to some respondents not providing survey information in a
particular quarter). Publicly available data, such as statistics in The Nilson Report, do not provide the same level
of detail (i.e., no figures by transaction size or by merchant). However, once aggregated, the Visa U.S.A. data are
consistent with The Nilson Report.
Figure 1. Transaction Shares of Various Payment Instruments, 2001
$5-$10 $10-$20 $20-$40 $40-$60 $60-$80 $80-$100 $100-$500 $500+
Transaction Size
Share of Sales Volume (%)
Credit Card
Debit Card
Source: Visa U.S.A., 2001 data.
A number of factors explain the diversity in consumer payment instrument use. For a
given dollar amount, a wide range of transaction types exists depending on, among other
things, who is making a payment to whom. For example, consumers pay landlords rent–a
recurring, fixed amount charge; consumers purchase goods and services from businesses, either
in person or remotely; and consumers pay utilities, another recurring charge, though typically
not of a fixed amount. Differing levels of trust between payer and payee will influence the
preferred payment method,40 as will payee constraints on what is accepted. Moreover,
consumers vary in income and wealth. Individuals who command higher labor incomes may
have a higher opportunity cost of time.41 They may be willing to pay more for a transaction that
is faster. Thus, variation in income can generate different valuations across consumers for
various payment instruments.
Transactions themselves are also diverse. They may be small or large on average; they
may be occasional or frequent; they may be in person or conducted remotely. Monthly rent, for
instance, may be several hundred dollars while newspapers cost just a few dollars a week. Even
at the same merchant, transaction size varies. At the supermarket, one could purchase eggs and
milk for cash or use a check to buy a week’s worth of groceries.
Heterogeneity among merchants plays a role in determining payment instrument
acceptance. For example, a discount retailer with multiple checkout lanes will face different
costs and benefits from accepting personal checks or online debit cards than a boutique retailer
with only one payment register. Merchants also deal in goods with different prices. A retailer
dealing in low-price items may look at the relative merits of PIN and signature debit very
differently than a retailer dealing in higher-price items since merchants typically pay a fixed
per-transaction fee for PIN debit and a percentage of the transaction value for signature debit.42
Merchants must also consider what payment instruments their clientele are most likely to want
to use. As a result of all of these factors, for example, consumers may pay for $10 worth of
goods at a department store with cash or a credit card, but newsstands typically require cash.
40 Jeffrey K. MacKie-Mason & Kimberly White, “Evaluating and Selecting Digital Payment Mechanisms,”
Working Paper, 1996.
41 Irrespective of labor income, people in a hurry to do something else will face a higher opportunity cost of time.
42 Evans & Schmalensee, supra note 2.
Even when consumers’ choices are not constrained, payment methods can differ for the
same purchase size because consumers view transaction types differently. For example,
consumers are more likely to pay for $100 worth of groceries with a check or debit card, but
will often pay for $100 worth of clothes with a credit card.43 The fact that we observe several
instruments used leads us to believe that payers with different personal characteristics have
different assessments of costs and benefits. Some payers, for example, may place a high
valuation on transaction speed whereas others may place a higher valuation on expense
tracking and the possibility of getting cancelled checks back. Private net benefits of a given
instrument vary across people and, therefore, private usage in a given environment and for a
given transaction size vary too.
These points raise an interesting issue in regard to evaluating social net benefits, which
requires combining costs and benefits of payment instruments as perceived by heterogeneous
agents. This is a standard problem in cost-benefit analysis, one that is typically resolved by
considering an average or representative value.44 But heterogeneity implies that the average
value may not be relevant for particular transactions, consumers, or merchants.45 Likewise, it
would be inappropriate to assess general merchant costs by focusing on, say, supermarkets.
While we use an average agent approach, we address three elements of diversity in our
analysis––retail venue, transaction size, and consumer heterogeneity.
F. What to Hold Constant
A final point to consider in analyzing the net benefits of a payment method is the
underlying assumption that a given transaction will occur regardless of the payment method
decision. The consumer purchase example above illustrates a “conditional” approach. The
assumptions here are that the transaction will take place with certainty, that it will take place at
a given size and that it will take place in a specific context. Under these assumptions a
consumer is faced with choosing a payment instrument based on the net benefits of each
accepted instrument.
43 Evans & Schmalensee, supra note 2.
Imagine a slightly different situation, one that allows the transaction to vary, say by
purchase size or place. This unconditional approach raises different issues, assessing whether
the transaction would have taken place in the absence of certain payment instruments,
evaluating potential changes in the characteristics of the transaction, and gauging the economic
The second approach is harder to implement, but it recognizes the possibility that
different payment instruments can, in general, increase economic welfare by increasing the
number and type of transactions that take place or by affecting transaction costs. Consider, for
example, certain impulse purchases. In arguing the need for stored value “smart” cards,46 the
Chairman of Intuit, Scott Cook, recounted a tale of driving through Santa Cruz on a hot day
looking for a soft drink.47 He stopped at two separate stores but encountered long checkout
lines and vending machines that took coins only (which he did not have) or rejected the bills he
did have. He never did buy a drink.
A new transaction may not take place due to restrictions on payment methods that are
accepted. Of course, a payment instrument’s impact on the transactions taking place has limits.
If one store does not accept personal checks, a store down the street may. That the second store
accepts checks does not raise the number of transactions taking place, it merely shifts the
venue. Even here, though, economic welfare is affected because searching for a store that
accepts the desired payment method involves transaction costs. Nor is there perfect substitution
over time for consumption––Scott Cook searched unsuccessfully for his preferred payment
method. This simple illustration highlights two possible complementary effects from
broadening the payment methods available: first, more consumption may occur because
44 One can also use weights to calculate an aggregate value. See, e.g., Daniel T. Slesnick, “Empirical Approaches
to the Measurement of Welfare,” Journal of Economic Literature, Vol. XXXVI, 1998.
45 We explore this issue below in considering the sensitivity of our results to changes in the wage rate.
46 Stored value smart cards are essentially electronic wallets. Holders transfer money to the card, which can then
be used instead of cash or change at vending machines, parking meters, and other places where small bills and
coins are used. Stored value smart cards are already well established in Asia (Japan and Hong Kong in
particular) and in public transit systems in the United States. “Octopus Reaches Out Beyond Hong Kong,” Card
Technology, February 1, 2003; Donald Davis, “JCB’s Chip Card Growth Strategy,” Card Technology, February
1, 2003; and Burney Simpson, “Public Sector Takes the Lead in Adopting Chips,” Card Marketing, March 1,
transaction costs are lower; second, the distribution of transactions can differ with consumers
benefiting from getting closer to their preferences.
The conditional perspective on costs and benefits can produce results that look very
different from the unconditional approach. If a transaction is going to happen at a store with
certainty, then the opportunity cost of time spent traveling to the store is irrelevant; the trip is
made regardless of the payment instrument used. But once you consider the possibility of
purchasing goods from home via the internet, say, time spent in getting to the store is no longer
irrelevant in comparing marginal costs and benefits. If, thanks to the availability of certain
instruments, some consumers make the transition from in-store purchase to internet purchase,
they clearly benefit from the time saved by avoiding the trip. Instead, they now have to account
for time spent consummating a transaction online and the inconvenience of waiting for goods to
be delivered. In sum, evaluating net benefits of payment instruments changes considerably in a
conditional versus an unconditional approach. In practice, the conditional approach is far easier
(and thus is the one we focus on in our analysis), but the unconditional approach can illuminate
key issues.
Several scholars have attempted to evaluate the costs of various payment instruments.
When considered in light of the approach outlined above, it becomes clear that few papers
present a comprehensive analysis of the payment methods they review. Most ignore the
benefits side of the equation altogether, and some do an incomplete job in examining costs.
Our review focuses on the United States. Given the historical and institutional
differences across countries, it is difficult to translate cost and benefit studies from one country
to another. Even payment instruments that look similar across countries on the surface may be
quite different operationally and, therefore, in terms of their costs and benefits.48 Nonetheless,
47 He presents this anecdote in the foreword to Smart Cards, Seizing Strategic Business Opportunities, edited by
Catherine Allen & William J. Barr, New York: McGraw-Hill, 1997, at xi.
48 Consider check clearance in two seemingly similar countries, the United States and Canada. Canadian banks
developed a system where checks are cleared on a same-day basis, but U.S. banks have no such system in place
and paper checks generally take around 5 days to clear. Due to historical differences in banking laws, Canada’s
commercial bank industry is highly concentrated: in 1990 the six largest banks in Canada accounted for 90
studies from abroad can offer insights on how to approach payment instrument evaluation. We
therefore review several non-U.S. papers that make points relevant to our analysis.
A. The Academic Literature
We begin our review of the academic literature with some of the early and influential
studies that compare and contrast multiple payment instruments. These studies attempt broad
cost evaluations, covering paper and electronic payment methods. None of them address
benefits in a systematic way, although a few of them mention benefits in passing. From these
studies, we move on to another group of papers that attempt broad cost comparisons across
nations. The cross-country studies attempt to use variations in costs at the national level to help
explain payment instrument use. As noted above, it is difficult to compare payment systems
across nations due to legal and historical differences. Even so, comparing different rates of
payment instrument use can illustrate how payers and payees react to varying incentive
structures. As with the first group of papers, these studies also generally ignore benefits. We are
aware of only one U.S. paper that considers both costs and benefits in a comprehensive manner.
We conclude our academic review with that paper, as well as a few others that look solely at
benefits and ignore costs.
1. Multiple Payment Instrument Comparisons
Humphrey and Berger (1990) present one of the earliest attempts to comprehensively
estimate payment instrument costs.49 Using 1987 data, the authors calculate private and social
costs for nine separate payment instruments—cash, checks, credit cards, travelers checks,
money orders, Automated Clearing House transfers (ACH), wire transfers, point of sale (POS)
percent of total banking assets and for more than 75 percent of payment volume, whereas this level of coverage
in the U.S. would have required more than 3,000 institutions. The relative concentration in Canada allowed
banks to negotiate same-day check clearance (achieved through post-dating), while U.S. checks take several days
to clear and are handled, on average, by almost 4 separate banks. David B. Humphrey & Allen N. Berger,
“Market Failure and Resource Use: Economic Incentives to Use Different Payment Instruments,” in The U.S.
Payment System: Efficiency, Risk and the Role of the Federal Reserve, edited by David B. Humphrey, 1990, at
49 Humphrey and Berger, supra note 48. For an earlier attempt along the same lines, see David B. Humphrey, The
U.S. Payments System: Costs, Pricing, Competition, and Risk, New York: Monograph Series in Finance and
Economics, 1984.
bill payments, and ATM bill payments.50 They find that from a social cost perspective cash is
the cheapest payment instrument, followed by ACH, POS bill payment, and ATM bill payment.
From a private perspective, however, checks emerge as the cheapest payment method, followed
by cash, ACH and POS bill payment.51
According to Humphrey and Berger, float introduces a wedge between social and
private cost. With cash, users make float payments to the government, the cash issuer; with
checks, payees make float transfer payments to payers; with credit cards, payees receive their
money quickly while the card system provides float to payers, who do not pay for their
purchases until their monthly bill arrives. Since payers using credit cards and checks do not
face the full cost of their instrument choice, Humphrey and Berger argue that payers tend to
“overuse” these payment methods.52 That is, the authors maintain that float imposes an
externality and thus introduces a market failure, which should be corrected through some
means, possibly including government intervention.53
The Humphrey-Berger analysis focuses exclusively on costs. Float is counted as a cost
only; the benefits of float for consumers are not included. Although the authors acknowledge
that “convenience and acceptability” may play a role in payment instrument choice, they do not
include these benefits in their calculations. Their consideration of government intervention is
premature given that they have not calculated the net benefits of payment instruments.54
50 Humphrey and Berger calculate private cost as the residual of social cost once float is removed. This definition
differs from our definition of private cost.
51 Strictly speaking, Humphrey and Berger find that, from a private perspective, consumer checks are more costly
than cash, ACH and POS, whereas business and government checks are less costly than cash.
52 This point does not necessarily follow, though, since payment card systems are two-sided markets. The
asymmetric fee arrangement can help to get both of the crucial parties to the transaction to participate. For an
overview of two-sided markets, see Evans, supra note 10. For a theoretical treatment of two-sided market issues,
see Jean-Charles Rochet & Jean Tirole, “Platform Competition in Two-Sided Markets,” Journal of the European
Economic Association, Vol. 1, 2003; and Geoffrey G. Parker & Marshall W. Van Alstyne, “Information
Complements, Substitutes and Strategic Product Design,” Working Paper, November 8, 2000. Because the
platform (the payment card system) needs to balance overall demand, the prices charged to any one side will not
generally equal the marginal cost on that side, if indeed marginal costs can even be separated by sides. See
Schmalensee, supra note 30.
53 Humphrey & Berger, supra note 48, at 47.
54 The authors note that “speed, security, availability, and acceptability” may be important determinants of choice,
but do not count these features in their accounting because benefits of this type are difficult to quantify.
Humphrey & Berger, supra note 48, at 52.
Humphrey and Berger also do not include some important cost elements. For example,
a payer’s cost of cash is excluded from their calculations; yet, this is likely one of the key
factors driving individuals away from paper instruments and toward payment cards for point of
sale transactions.55 Nor do their calculations include a payer’s opportunity cost of time for
check writing, which could be one of the elements driving individuals away from checks
toward payment cards and automatic bill payment. Finally, the paper does not include the
payee’s cost of handling and processing checks, which available evidence indicates is a
significant cost for retailers.56
Wells (1996) builds on the Humphrey-Berger study.57 For checks and ACH payments,
she updates the 1987 data to 1993 dollars and then presents her own calculations side-by-side.58
She finds that between 1987 and 1993 the value of check float declined considerably, due to
both a decline in interest rates and improvements in check processing. She points out that
despite the fall in the value of float, check use did not decline, but instead total check use
increased by 20 percent.59 Wells thus questions the Humphrey-Berger hypothesis that float
explains the use of checks.
Wells also includes an estimate of the payee cost of checks, omitted from the
Humphrey-Berger analysis. This addition makes a significant difference in the final numbers.
Her estimate of the social cost of checks, for example, is significantly higher than Humphrey
and Berger’s.60
55 Humphrey and Berger acknowledge that the payer cost of using cash should include the cost of lost cash and
theft, along with the cost of obtaining cash from a bank, but the calculations do not do so. Humphrey & Berger,
supra note 48, at Table 2-A2.
56 See, e.g., Food Marketing Institute, “A Retailer’s Guide,” supra note 23.
57 Kirstin E. Wells, “Are Checks Overused?,” Federal Reserve Bank of Minneapolis Quarterly Review, Vol. 20,
58 Wells points out that an important difference between her study and the Humphrey-Berger analysis is that she
accounts for the payer cost of checks and ACH. Wells, supra note 57.
59 As a fraction of all transactions, however, checks declined by 3 percent in that time. Payment Systems in Eleven
Developed Countries, Bank for International Settlements: Central Banks of the Group of Ten Countries and
Switzerland, May 1989; and Statistics on Payment Systems in the Group of Ten Countries, Bank for International
Settlements: Committee on Payment and Settlement Systems, December 1996.
60 Wells calculates the total social cost of checks is between $2.78 and $3.09, whereas the total private cost is
between $2.69 and $3.00. In Humphrey-Berger’s approach, as updated by Wells, the total social cost of checks is
only $1.00, and the total private cost is slightly negative.
Wells does not include benefits in her calculations, but she also acknowledges that they
matter, particularly in her discussion of why checks are used more frequently than less costly
alternatives. She suggests that users do not treat checks and electronic payment instruments as
close substitutes for one another: “Checks may be used more…simply because users prefer
them to other ways of making payments.”61 In particular, benefits may stem from the control
that is associated with having a physical payment instrument.62
In a paper that emphasizes the low cost of electronic payment instruments, Gresvik and
Øwre (2002) study how much it costs Norwegian banks to process various payment
instruments, including certain forms of paper and electronic giro,63 checks, and payment cards
used both at ATMs and at the point of sale.64 They base their results on an activity-based
costing analysis survey.65 The survey contains data on direct costs that vary with the number of
transactions––costs related to individual services provided by a bank such as the purchase of
notes and coins, the purchase of card services, and inter-bank charges. It also covers indirect
costs, including personnel costs, computer system operations, the cost of buildings, machines,
office supplies, and marketing. Indirect costs are allocated across payment services based on a
bank’s activities in supplying that service.
The authors conclude that in general paper-based payment methods are more costly than
electronic methods. One exception holds, however––internet giro costs more than mail giro,
probably because internet giro is relatively new in Norway and start-up costs are high. The
authors identify payment cards, which are more akin to PIN debit cards in the United States, as
61 Wells, supra note 57, at 5.
62 Wells suggests that “the primary objection to ACH payments may be not that they are electronic, but rather that
they are automatic.” Wells, supra note 57, at 5.
63 Giro payments, very significant in several European countries, are essentially credit transfer payments
equivalent to preauthorized bill payments in the United States. In Europe, postal savings institutions, commercial
and savings banks, and credit cooperatives set up their own separate giro systems in the late 1800s and early
1900s. After a period of competition between these various systems, cooperative agreements between postal and
bank giros were established. The clearing systems for credit transfer giro payments started making the transition
to electronics in the early 1970s. David Humphrey, Setsuya Sato, Masyoshi Tsurumi, & Jukka Vesala, “The
Evolution of Payments in Europe, Japan, and the United States: Lessons for Emerging Market Economies,” The
World Bank, Policy Research Working Paper No. 1676, October 1996.
64 Olaf Gresvik & Grete Øwre, “Banks’ Costs and Income in the Payment System in 2001,” Norges Bank
Economic Bulletin, 2002.
65 Activity-based cost accounting, frequently referred to as ABC, is a method of assigning costs to business
processes and activities. It attempts to determine which activities “cause” which costs.
among the cheapest payment instruments. In Norway, cards may be used to make payments at
store terminals and to withdraw cash at either store terminals or ATMs.66 In fact, payment cards
used at in-store terminals are the most popular payment instrument in the country as well as the
cheapest. The average cost of in-store card use has declined considerably, the authors explain,
due to economies of scale achieved through significant growth in the number of transactions.
Payment cards used for cash withdrawals at ATMs cost considerably more since the
transactions involve cash replenishment, maintenance and security costs. Using checks for cash
withdrawals is even more expensive, however, costing around three times as much as payment
card cash withdrawals at ATMs.
2. Cross Country Comparisons
Humphrey, Pulley, and Vesala (1996) analyze patterns in the use of cash and other
paper and electronic payment instruments in 14 developed countries, including the U.S.67 Using
payment data for 1987-1993, the authors address what determines payment instrument usage.
Treating payment instruments as if they were traditional goods, the authors construct measures
of the costs (analogous to prices) of various payment methods in order to study whether
differences in payment instrument use across countries and over time can be explained by
differences in their relative prices.
The authors attempt to collect systematic information on the prices of five payment
instruments—checks, paper giros, electronic giros,68 credit cards and debit cards. They
calculate the payer price of checks, for example, as the opportunity cost of holding a minimum
balance in a checking account plus mailing costs minus the value of the float.69 They calculate
prices for other payment instruments along similar lines.
One interesting aspect of this paper is that the calculated prices of payment instruments
fail to explain differences in instrument use across countries and over time.70 The authors offer
66 Similar to PIN debit cash-back within the United States.
67 Humphrey, Pulley & Vesala, supra note 33.
68 In the study, electronic giro payments include direct deposits, direct debits, and other automated clearing house
69 Humphrey, Pulley & Vesala, supra note 33, at 938.
70 Humphrey, Pulley & Vesala, supra note 33, at 936.
various explanations for the lack of explanatory power, one of which is particularly appealing.
Differences among the calculated prices are generally small and simple differences in the
“convenience” of using a particular instrument—a factor that is not measured—may outweigh
the price differences that users face.71 This is an acknowledgement that net benefits, not just
costs, matter for determining usage.
The authors find that what they call “institutional variables”––in particular the violent
crime rate––explain some non-trivial portion of the observed differences in usage across
countries.72 Crime could be interpreted as an element of cost, such as the expected cost of using
cash. Unlike payment cards and checks, which bear the owner’s name and therefore require a
degree of sophistication to use without authorization, stolen cash is perfectly transferable. The
authors’ finding confirms the importance of considering probabilistic costs, an issue raised
In another study comparing costs across nations, De Grauwe, Buyst and Rinaldi (2000)
study the costs of cash and payment cards in Iceland and Belgium.73 The authors select these
two countries because they provide a clear contrast. Among the developed countries, Iceland
has one of the lowest rates of cash use while Belgium is at the other extreme.
The authors focus on social cost, defined as the resource cost to the economy of
operating a cash-based system and a card-based system. They measure social cost by
identifying all the participants in each system and counting their costs net of revenues. For the
card-based system, they examine the card companies, commercial and savings banks,
cardholders, merchants, and clearing and settlement institutions (where they exist). For the
cash-based system, they study the role of the central bank, commercial banks, consumers and
merchants. For the cash payment system in Iceland, for example, the authors estimate the cash
production and distribution costs incurred by the central bank and then subtract the revenues
that this institution collects through interest foregone on cash in circulation. They then add the
costs incurred by commercial banks, merchants and consumers.
71 Humphrey, Pulley & Vesala, supra note 33, at 932-33.
72 Humphrey, Pulley & Vesala, supra note 33, at 934, 936.
73 Paul De Grauwe, Erik Buyst & Laura Rinaldi, “The Costs of Cash and Cards Compared: The Cases of Iceland
and Belgium,” mimeo, February 2000.
From a social perspective, the authors conclude a card-based system is considerably
more efficient than a cash-based system. For Iceland, the per-transaction cost of cash is about 5
times higher than for cards. In Belgium, the per-transaction cost of cash and cards is similar,
but cash is about 5 times more expensive than cards if cost is expressed as a percentage of
transaction value.
As payment card use increases relative to cash for a given amount of economic activity,
the authors argue that the average cost of cash increases relative to cards for two reasons. First,
diseconomies of scale in the supply of cash rise as cards displace cash, while economies of
scale improve for cards. Second, the displacement relegates cash to smaller transactions.
Because smaller transactions must cover the fixed costs of the cash system, the cost of cash
expressed as a percentage of transaction size rises.
The authors find that distribution of costs and revenues change dramatically as an
economy shifts from a cash-based system to a card-based system. The central bank is the
biggest loser in the process since it foregoes an important source of revenue. Consumers are the
primary winners because the transition to a card-based system lowers the resource cost of the
payment system. This has the same effect as any reduction in the cost of inputs in the
production system—it can lower the price of goods and services for consumers.
Surprisingly, the authors find that cards are significantly less costly than cash even
when only merchants are considered. They explain that this contradicts the perception that
many merchants have because merchants often fail to include the resources they spend in
handling and transporting cash.
Further supporting the cost savings involved in moving to a non-paper payment system,
Humphrey, Willesson, Lindblom, and Bergendahl, survey the payment cost accounting
literature, including many of the papers reviewed above (and below).74 The authors discuss the
limited data on payment method costs in various countries and conclude that electronic
payment instruments, such as electronic giro, ACH, and debit cards, cost nations between one-
third and one-half as much as paper-based instruments, such as cash and checks. They suggest
that a country may save 1 percent of its GDP annually by shifting from a fully paper-based to a
fully electronic-based system.75
In addition to the startling cost savings suggested from moving to a cashless economy,
this survey makes a couple of other interesting points. First, scale economies have a dramatic
impact on the costs of operation. Since the level of use varies across countries for a given
payment instrument, so do the costs. For example, bank check processing costs are over ten
times higher in Norway than in Spain. The authors posit that processing economies of scale are
to blame, as only 0.3 percent of all payments in Norway are made by check. Second, some
paper-based methods appear to be less costly to process than electronic methods. For instance,
in Germany checks cost retailers less than debit cards or electronic giro. The authors note that
this finding is somewhat misleading, though, since check processing in Germany is entirely
electronic. In essence, a check in that country is a paper initiated electronic instrument.
The Humphrey, et al. survey does have some limitations. The costs presented are not
calculated for a given transaction size, but instead each cost is specific to the average
transaction size of a given payment instrument within a given country. Nor are all parties to a
transaction included: missing from many of the comparisons are government and consumer
costs. Furthermore, retailer costs representing U.S. merchants come from the Food Marketing
Institute’s study of supermarket and grocer costs.76 Given the unique characteristics of
grocers,77 it seems unlikely that grocer payment processing costs are indicative of other
retailers. Finally, because the many studies reviewed in the paper do not count benefits, the
survey does not include benefits in its calculations.
74 David Humphrey, Magnus Willesson, Ted Lindblom, & Göran Bergendahl, “What does it Cost to Make a
Payment?,” Review of Network Economics, Vol. 2, June 2003.
75 A later paper by the same authors confirms this finding in an empirical analysis based on data from 12 European
countries. See David Humphrey, Magnus Willesson, Göran Bergendahl, and Ted Lindblom, “Cost Savings from
Electronic Payments and ATMs in Europe,” working paper, August 2003.
76 This study is discussed in detail below.
77 See the first case study below.
3. Comparisons Including Benefits
A paper by Stavins (1997)78 addresses different ways of processing one payment
instrument––the transition from traditional check clearing to electronic check “presentment and
truncation.”79 The analysis systematically accounts for both costs and benefits to all parties
involved in check processing. As the author observes concerning benefits, “even though one
payment instrument may be cheaper than another, it is more efficient only if its net social
benefits are higher than those of the alternatives. To determine whether that is the case, benefits
must also be taken into account.”80
Stavins examines costs and benefits arising from electronic check presentment and
truncation for the depositing customer (the check recipient), the bank of first deposit, an
intermediary (such as the Federal Reserve), the paying bank, and the paying customer. If the
bank of first deposit truncates the check, for example, the check writer faces two additional
costs. First, she loses float because truncation expedites check processing. Second, she loses
whatever value she places on receiving the canceled check. To estimate the latter cost, Stavins
relies on market surveys finding that consumers are willing to pay for their canceled checks.81
The author concludes that electronic check processing would raise net social benefits by a
substantial amount.82 Despite the savings, however, several obstacles stand in the way of easy
adoption. Transition costs, network externalities, and an uneven distribution of the savings are
some of the key factors that might prevent or slow electronic check processing.
Carow and Staten (2000) do not evaluate costs at all and instead focus on benefits. The
authors examine why consumers might choose to pay with general-purpose payment cards.83
78 Stavins, “Comparison of Social Costs,” supra note 15.
79 With electronic check truncation, the physical transfer of the original paper check to the paying bank is stopped
at some point, such as at the bank of first deposit. Instead, the check is sent through the processing route
electronically, presented to the paying bank (and all the other processors in between) via an electronic image of
the check.
80 Stavins, “Comparison of Social Costs,” supra note 15, at 28.
81 Stavins, “Comparison of Social Costs,” supra note 15, at 38-42.
82 She assumes that all on-us checks would be kept by the paying bank while all others would be truncated either
by the bank of first deposit or by an intermediary. Under these assumptions, electronic check processing
increases net social benefits by 2.39 cents per check, or $1.4 billion per year. Stavins, “Comparison of Social
Costs,” supra note 15, at 37.
83 Kenneth A. Carow & Michael E. Staten, “Plastic Choices: Consumer Usage of Bank Cards vs. Proprietary
Credit Cards,” Working Paper, April 2000.
Using survey data from retail and gasoline cardholders, the authors consider the substitution of
general-purpose (bank) cards for proprietary retailer cards. They find that the convenience of
using a single payment card and the potential to earn rebates are among the primary reasons for
using a bankcard. They also find that consumers use proprietary gasoline cards to keep
purchase records and use proprietary retail cards to obtain better service.
Mantel (2000) summarizes the literature on consumer payment decision-making, also
focusing on payer benefits.84 He proposes a framework in which three factors explain the
private use of electronic banking: consumer wealth; personal preferences, including
convenience, control, budgeting, privacy, security, and personal involvement; and transaction-
specific factors, such as transaction size and fixed-amount vs. variable-amount bills.
Also examining payment instrument decisions, MacKie-Mason and White (1996)
summarize the properties of 10 electronic payment mechanisms according to 30 criteria.85 They
show how a decision maker may follow a systematic approach to selecting a payment
mechanism. According to the authors, certain characteristics of payment methods may be so
important that they play a fundamental role in the selection process. For example, mechanisms
with low overhead costs are ideal for small transactions that occur frequently.
B. Retail Industry Studies
Another set of studies explore retailer costs of processing various payment instruments.
While industry groups conducted or sponsored these studies, at least one of them is frequently
cited in the academic literature. As noted above, the Humphrey, Willesson, Lindblom, and
Bergendahl analysis relies on the Food Marketing Institute’s study of grocery stores and
supermarkets to represent merchant costs for all American retailers.
The reliance on industry studies is not surprising given the dearth of merchant cost
information. Government data sources, used for many payment instrument costs, do not cover
merchants’ costs. In fact, it appears that retailers themselves often do not track their costs at the
84 Mantel, supra note 9.
85 MacKie-Mason & White, supra note 40.
level of detail needed to evaluate one payment instrument against another.86 As a result, the few
industry studies that focus exclusively on specific kinds of merchants and on processing costs
for point of sale transactions are an important source for academic studies. That said, it is
difficult to determine whether these studies accurately portray merchant costs. Surveys may not
reach the proper decision makers within the company, or even if they do, if retailers are not
accurately tracking costs then survey responses may be little more than the merchant’s best
guess. Even if individual merchants’ responses are accurate, it is difficult to know whether the
respondents are representative of their industry. More fundamentally, since so few industries
are covered it is doubtful that the available studies represent the full range of retailer costs.
1. Grocery Store Costs
In 1994, the Food Marketing Institute (FMI) conducted a study on the cost to U.S.
grocers to accept various payment instruments.87 The numbers were updated in 1998 following
the same research method and again in 2000, although the 2000 method does not match that of
the earlier two studies.88 All three reports are based on surveys of FMI members, each with a
limited sample size. The reports do not provide enough information to determine whether the
samples are representative of grocers nationwide.89
86 The 2000 Food Marketing Institute study, as well as the 1994 Coopers & Lybrand study report, both complain
of this. See Food Marketing Institute, “It All Adds Up: An Activity Based Cost Study of Retail Payments,” 2000,
at Section 2.1; and Coopers & Lybrand, supra note 26.
87 Food Marketing Institute, “Benchmarking Comparative Payment Methods,” supra note 23. In particular, FMI
conducted a survey of member companies.
88 Food Marketing Institute, “A Retailer’s Guide,” supra note 23; and Food Marketing Institute, “It All Adds Up,”
supra note 86. Note that at the time of both the 1998 and the 2000 studies, the FMI was a plaintiff in a lawsuit
filed against Visa and MasterCard regarding the payment card associations’ “honor-all-cards” rules. See Second
Amended Consolidated Class Action Complaint and Jury Demand, In re Visa Check/MasterMoney Antitrust
Litig. (E.D.N.Y. 2000) (No. CV-96-5238).
89 Surveys were sent to all FMI member companies. In 1994, FMI received 42 responses, in 1998 they received 48
responses, and in 2000 estimates they received 36 responses. The 1998 responses “account for 27 percent of the
[food retailer] industry sales.” Food Marketing Institute, “A Retailer’s Guide,” supra note 23, at 1. We do not
have sufficient information to determine whether the responding grocers are representative of the industry as a
whole. We were unable to audit the report’s findings as we do not have access to individual survey responses.
Note that the surveys were sent to FMI’s “key contact” at the member company, who may or may not have been
someone familiar with payment systems and their costs. See Defendants’ Memorandum in Support of Motion in
Limine to Exclude from Trial Plaintiff Food Marketing Institute’s Reports of Surveys on Payment Costs,
Redacted Version, In re Visa Check/MasterMoney Antitrust Litig. (E.D.N.Y. 2003) (No. CV-96-5238), at 7.
The FMI studies focus on the “direct” costs to grocers of accepting cash, non-verified
and verified checks, food stamps, credit or charge cards, debit cards, EBT, and WIC.90
Consistent with the focus on processing costs, allocations of corporate overhead and equipment
costs are excluded from the analysis. Surprisingly, float costs are excluded, despite the
significant role they appear to play in the academic literature. Included are the costs of
processing transactions at the checkout counter (“tender time”), the costs of preparing and
transporting bank deposits (“deposit preparation time”), the costs of financial institution
services (“bank charges”), and costs specific to certain payment instruments (“other direct
The FMI studies calculate costs both for an average transaction size, which differs
across payment instruments, and scaled to $100 of sales.92 The report does not clarify whether
transaction sizes include cash back to customers for PIN debit and checks.93 If included, the
reported costs could be considerably distorted. Consider a typical $40 PIN debit grocery sale
with reported merchant processing costs of 29 cents.94 Scaled to $100 of sales, PIN debit
processing costs are 70 cents. If, however, $20 of the $40 was cash given back to the consumer,
which has no processing cost and does not represent a sale or any profit for the merchant, then
the true purchase size is only $20. Scaling the same 29 cents of processing costs to $100 of
sales, but starting from a $20 average purchase implies $1.45 in PIN debit processing cost.
Thus, processing cost per dollar of sales is dramatically understated if cash back is erroneously
90 EBT is the electronic transfer of government benefit funds (e.g., food stamps). Recipients access benefits
through plastic cards at retail point of sale or ATM terminals. See Food Marketing Institute, “Benchmarking
Comparative Payment Methods,” supra note 23, at 117. WIC is the Supplemental Food Program for Women,
Infants, and Children. It provides government assistance to low-income, pregnant and post-partum women and
their infants. Benefits are delivered either by paper voucher or EBT. See Food Marketing Institute, “A Retailer’s
Guide,” supra note 23, at 28.
91 Food Marketing Institute, “Benchmarking Comparative Payment Methods,” supra note 23, at 1.
92 This measure simply scales up the cost per average transaction. One hundred dollars worth of sales can be
comprised of a single $100 transaction, two $50 transactions, ten $10 transactions, or any other combination.
Generally, the number of transactions making up $100 worth of sales will vary depending on the payment type.
Cash purchases are typically smaller than credit card purchases, for example, so $100 worth of cash sales is
likely to include more transactions than $100 worth of credit card sales.
93 PIN debit cards allow for ATM-like withdrawals at retail point of sale. The transaction, referred to as “cash
back”, is dependent on a sale but is frequently free of charge. Many groceries also allow customers with
checking privileges to write checks in excess of the purchase in order to receive cash back. The FMI survey
question (the full survey is included as Appendix B in the 1998 report) is ambiguous in regards to including
cash-back in the purchase.
included. Unfortunately, it is unclear whether or not cash back amounts are counted in the
purchase size.
With these caveats in mind, the 1998 study95 finds that on an average transaction basis
cash is the cheapest payment instrument for the grocery industry, followed (in ascending cost
order) by PIN debit cards, ACH debit,96 and verified checks. Credit is found to be the most
costly for grocery retailers to accept. The ordering changes when costs are scaled to $100 of
sales: PIN debit is the cheapest payment instrument for grocers, followed by verified checks,
ACH debit, and non-verified checks, with WIC by far the most costly.97 In addition to mean
processing costs, FMI reports high and low costs based on survey responses. The variation is
substantial, even considering that the study covers only one retail segment. For instance, the
highest cost for processing cash transactions is twice the mean and almost ten times the lowest
2. Selected U.S. Retailer Costs
Similar to FMI, Coopers & Lybrand conducted a study using 1994 data to determine the
cost U.S. merchants face in accepting various payment instruments: cash, checks, store cards,
Visa credit, MasterCard credit, Discover and American Express credit or charge cards.99 Unlike
the FMI study, the Coopers & Lybrand project covers four merchant segments: department
94 These are, in fact, the numbers reported by FMI. Food Marketing Institute, “A Retailer’s Guide,” supra note 23.
95 Our analysis is primarily based on the 1998 study. The 2000 study appears to omit cost categories that are
clearly important in the 1994 and 1998 studies. It also combines payment categories, reporting all checks
together (as opposed to splitting verified and non-verified) and reporting credit cards combined with signature
debit cards. Moreover, the costs reported in the 2000 study are dramatically different from those in the 1998
study, even though the 2000 study collected cost data from the 1998-1999 timeframe, just one to two years after
data collection for the 1998 study. For example, the “other direct costs” category for cash transactions was just
under 14 cents in the 1998 study, but falls to 0.1 cent in 2000; bank charges for check deposits were 5 cents a
check in the 1998 study and $0 in the 2000 one. (See FMI, It All Adds Up, supra note 88.) These discrepancies,
along with sparse details and backup information in the 2000 report––which prevent a careful comparison of the
studies––led us to report findings from the 1998 study instead.
96 ACH debit involves supermarket loyalty cards linked to a customer’s bank account; the cards act as automated
clearing house debit cards. (See “A Retailer’s Guide,” supra note 23, p. 13.)
97 See Food Marketing Institute, “A Retailer’s Guide,” supra note 88, p. 3.
98 We explore this variation further in the first case study.
99 Coopers & Lybrand, supra note 26. Note that the Coopers & Lybrand study was prepared for Visa U.S.A. Visa
was a defendant in the Wal-Mart case, revolving around “Honor-all-card” rules imposed by payment card
systems. Second Amended Consolidated Class Action Complaint and Jury Demand, In re Visa
Check/MasterMoney Antitrust Litig. (E.D.N.Y. 2000) (No. CV-96-5238).
stores, apparel stores, discount stores, and specialty electronics stores. It excludes fixed costs
from the calculations, relies on one “representative” merchant from each segment,100 and
estimates payment instrument costs both for the average transaction size (which varies by
payment instrument) and scaled to $100 of sales.101 Float is included among the variable cost
Echoing the FMI findings, the Coopers study finds that cash is the cheapest payment
instrument for each of the merchant segments, for an average-sized transaction. For $100 worth
of sales, cash is again the cheapest in all but one merchant segment—checks are cheaper for the
department store. Checks are the second cheapest instrument in both types of calculations in all
segments but one—the Discover card is second cheapest for the apparel store.
The Coopers & Lybrand analysis highlights the importance of two elements of cost.
First, average transaction size affects the relative cost rankings of the various payment
instruments. Second, the study illustrates that relative payee costs vary depending on the
merchant segment considered. For $100 worth of sales at the department store, for example,
American Express was about 1.73 times as expensive as cash, while it was 6.93 times as
expensive at the discount store.102 The Coopers and FMI studies tend to agree that cash is the
cheapest instrument for merchants to accept as payment while credit cards are the most
3. Overseas Retailer Costs
A Dutch retail trade association, Hoofdbedrifjschap Detailhandel (HBD), published a
report a few years ago on the costs of various payment instruments to retailers. The study found
that the cost of an average cash transaction was half as much as a typical debit card payment.
The HBD commissioned a second study to further analyze the results of the first report. The
100 The same caveat that applied to the FMI study certainly applies here. The Coopers & Lybrand calculations are
based on just one retailer survey in each category. They state that the chosen retailers are “representative” of the
industry, but it is unlikely that any one retailer could accurately represent an entire retail segment.
101 As with the FMI study, the Coopers study multiplies the cost for the average transaction size by the constant
necessary to reach $100 of sales. Thus, costs are assumed to increase linearly.
102 Coopers & Lybrand, supra note 26, at 32.
second report, which adopted a more comprehensive approach to analyze payment instrument
costs, was published in March 2001, with the main results summarized by Van Hove.103
According to Van Hove, the study authors do not concentrate on the cost of an average
transaction—rather they estimate how the costs incurred by retailers vary as a function of
transaction size. Van Hove notes that cash is cheaper than debit for retailers for transaction
amounts up to NLG 67 (around US $37).104 Debit, however, is cheaper for higher amounts. His
main explanation is that the fixed costs associated with debit card payments are substantial.
Van Hove then discusses the estimates of the social costs involved, defined as the
private costs to both merchants and consumers plus the costs incurred by commercial banks and
the central bank (to the extent that these costs are not paid for by merchants and/or consumers).
He reports that the social cost of cash is larger than its private cost, but the private and social
costs of debit are essentially the same since the prices charged by the Dutch electronic funds
transfer operator cover costs. The bottom line: for amounts between NLG 29 and NLG 67
(around $16 to $37 US), cash payments may be cheaper for retailers but are more expensive for
the economy as a whole. For amounts greater than NLG 67, debit is cheaper than cash both
privately and socially.
C. Lessons Learned From The Literature
As the above review illustrates, the literature has approached payment instrument
evaluation from several different angles. The variety of methods and differing degrees of
completeness provide several important lessons. First, the benefits that individuals obtain when
using payment instruments are at least as important as the costs when it comes to choosing an
instrument. Second, some of these benefits are rather difficult to measure directly. Third,
103 See Leo Van Hove, “The Price of Cash Revisited,” Epso Newsletter, (July 2001) (visited Mar. 23, 2003)
<>; and Leo Van Hove, “Electronic Money and Cost-Based Pricing,”
Wirtschaftpolitische Blätter, Vol. 49(2) (April 2002).
104 The Dutch Guilder (NLG) used to be the official currency of The Netherlands. On January 1, 1999 eleven of
the countries of the European Economic and Monetary Union, including The Netherlands, decided to give up
their own currencies and adopt the Euro. The NLG ceased to be legal tender on February 28, 2002. At the time
of this writing, the conversion of NLG to Euros took place at the rate of (NLG/2.20371) = 1 EUR. The exchange
rate between U.S. dollars (USD) and Dutch Guilders (NLG) at the time of the HBD analysis was about 1.00
USD = 2.06296 NLG. See Universal Currency Converter (visited August 24, 2004)
different parties to a transaction may assess benefits and costs very differently. Fourth, the
same party may evaluate benefits and costs differently depending on the transaction type or
size. Finally, the fact that a payment instrument is cheaper for one party (such as retailers) does
not mean that it is cheaper for all parties.
Consider first the industry studies, each of which limits its analysis to the cost of
accepting various payment instruments at the point of sale. They do not consider payers’ costs
or costs imposed on other parties to the transaction. The studies do not even mention the
benefits that both payers and payees receive from using various payment instruments. Narrowly
focusing on one side of the transaction, however, can offer an in-depth accounting of the direct
cost to merchants in accepting various payment instruments, and thus their private incentives
for accepting certain payment methods while discouraging or rejecting others. The
disadvantage of this approach is its potential for conveying the misleading impression that what
is more costly for one side of the transaction is necessarily more costly for the parties as a
whole. Moreover, merchants themselves must consider more than cost because in practice
many do accept payment cards. By revealed preference, many retailers find that the benefits of
accepting payment cards outweigh the costs.105
Another misleading aspect of the merchant studies lies in their assumptions about
transaction size. The cost comparisons presented in retailer studies, while meaningful for
merchants, can be deceptive for policy makers. In particular, the costs are calculated for the
average transaction size specific to a given instrument. Each instrument has a different average
transaction size so each cost is calculated for a different base. Asking what it costs to process
the average purchase is a relevant retailer question, but social questions of optimal use imply
changes in behavior. If we consider, say, whether payment cards are overused relative to cash,
we want to know what it costs to process these two instruments at the same transaction size.
That way, we can determine if moving more transactions (of all sizes) from one payment
method to another saves resources. Some components of processing cost vary nonlinearly by
105 Some merchants may accept credit cards because they feel they must––once competitors accept plastic, it may
be a necessary service to maintain customers. Other merchants, however, may accept cards because of perceived
benefits. For instance, some fast food chains that began accepting credit cards within the last few years maintain
transaction size––bank charges, for example––while others do not vary at all, so simply scaling
to $100 of sales does not solve the comparison problem. To make an “apples-to-apples”
comparison needed for policy, costs need to be reconstructed from their elements, varying only
those components that are sensitive to transaction size.
Despite their deficiencies in understanding private or social choices, the industry studies
do identify important components of processing costs for different payment instruments. For
example, the FMI data imply that three items of cost combined (tender time, deposit
preparation time, and other direct costs) represent the bulk of payee cost for paper-based
instruments––about 98 percent for cash, 91 percent for non-verified checks, and 71 percent for
verified checks. Bank charges, which are the fourth cost item and include the merchant
discount fee in the case of credit, charge and debit cards, represent between 45 and 55 percent
of payee cost for PIN debit cards and ACH debit, respectively. In contrast, the FMI data
indicate that bank charges dominate the payee cost of accepting credit, charge and signature
debit cards, ranging from 78 to 81 percent. The differences in the relative importance of cost
items underscore the importance of careful accounting. If bank charges were omitted from the
FMI study, the payment instrument ordering the study reports would no longer hold. Instead
credit and charge cards would appear cheaper for grocers than the cheapest paper instrument
From a social perspective, bank charges represent a cost to merchants but provide
revenues to acquiring and issuing banks and help to pay for benefits, including low to zero
annual fees for consumers and fraud prevention and dispute resolution systems for both
merchants and consumers. If electronic payment instruments are cheaper for merchants when
bank charges are excluded, then an important question is whether bank charges provide benefits
for other parties to the transaction such that, when all parties’ costs and benefits are taken into
account, net social benefits are positive.
that the average purchase amount is larger for credit card customers than for cash customers. See Bruce
Horovitz, McDonald's Joins Pay-With-Plastic Trend, USA Today July 23, 2004.
106 Without counting bank charges, credit ($0.2005) is slightly more expensive than signature debit ($0.1783), PIN
debit ($0.l592) and ACH debit ($0.1592), but still cheaper than cash ($0.2119), which is in turn cheaper than
verified checks ($0.3165) and non-verified checks ($0.5319). See Food Marketing Institute, “A Retailer’s
Guide,” supra note 23, at 3.
The majority of the studies––including the academic ones––focus either exclusively or
mainly on costs. This is likely due to the difficulty in defining and measuring the benefits
various payment instruments provide. While some studies hint that benefits such as
convenience matter, they do not develop the implications of this insight. Stavins is the only
U.S. study, to our knowledge, that makes a systematic attempt to measure costs and benefits for
all parties involved. Regarding the dangers of looking only at costs and disregarding benefits,
Stavins observes: “…even though bicycles are cheaper than cars, one would not recommend
that bicycles be substituted for cars without evaluating the relative benefits of each mode of
transportation.”107 We concur that cost considerations are necessary but not sufficient for
making sound policy recommendations. The challenge is to produce a study on the costs and
benefits of various payment methods taking as many costs and benefits as possible into account
for all parties involved––a task we attempt in the next section.
To demonstrate the importance of moving beyond a narrow focus on costs, we develop
three case studies based on existing cost analyses, but expanded according to our guidelines. To
fully understand how one payment instrument compares to another, it is necessary to delve into
the details. We therefore conduct a careful cost-benefit assessment, including as many
individual costs and benefits as can be quantified. We focus our analysis on the shift from using
paper instruments toward greater use of payment cards. This payment transition is occurring
throughout the retail sector. While the three case studies do not cover the entire sector, they do
shed light on the move to a cashless economy and underscore the lessons we discuss above.
Over the last decade, payment cards such as credit and debit have tended to replace cash
and checks in a wide variety of transactions, both large and small. Table 1 illustrates this shift
by calculating the percentage point change from 1994 to 2001 in the consumer-to-business
sales volume paid for using several different payment methods.108 As the table shows, the use
of cash in all but one transaction size category ($500 and over) fell substantially. The use of
107 Stavins, “Comparison of Social Costs and Benefits,” supra note 15, at 28.
108 1994 and 2001 represent the endpoints of our dataset. The decline in cash and check use is fairly steady
throughout the time period, as is the rise in credit and debit card use.
checks fell substantially in all categories. In contrast, the use of credit, charge and debit cards
increased in all categories.
Given the ubiquitous shift away from paper (both cash and checks), we examine
whether the increased use of payment cards has generated welfare gains. In particular, we
examine the net benefit of replacing one additional consumer paper currency transaction with a
payment card transaction. We therefore analyze marginal use among consumers who already
have payment cards at merchant locations that already accept them. Since upwards of 73
percent of all U.S. households currently have a general purpose credit or charge card,109 the
marginal cost of one more card transaction is quite relevant.
By revealed preference, the fact that consumers are choosing to pay more often with
payment cards when they could continue to pay with checks or cash implies that they are better
off––increased use of payment cards must make consumers better off or they would not alter
their behavior. The question is whether there are net gains to the economy as a whole, including
the merchants receiving those payments and the banks processing them. If consumers do not
109 Survey of Consumer Finances, 2001; and Evans & Schmalensee, supra note 2.
Table 1. The Percentage Point Change in Transaction Shares by Transaction Size:
$5-10 $10-20 $20-40 $40-60 $60-80 $80-100 $100-500 $500+
Credit/Charge 3 5 8 6 7 5 4 5
Debit 7 12 12 11 10 9 7 2
Card Total 10 17 21 17 17 14 11 8
Cash -6 -9 -7 -6 -6 -4 -2 1
Checks -5 -7 -13 -11 -12 -11 -9 -5
Paper Total -10 -17 -21 -17 -17 -15 -12 -4
Notes: Debit combines signature and PIN debit transactions. Due to rounding, card and paper
totals may not sum exactly. ACH, EFT, and other miscellaneous payment methods are
excluded, and thus the card total and paper totals do not net to zero.
Source: Visa U.S.A., 1994-2001.
pay the full price of using a payment card, an argument made by several scholars,110 then what
is good for consumers does not necessarily imply net social benefits.111
We attempt to answer the question of net social benefits on a per transaction basis
relying on the guidelines developed in Section II. In particular, because the location of a
transaction influences the cost and benefit calculations, we analyze three specific venues:
grocery stores, discount stores, and specialty electronics stores. These three venues cover a
range of goods and average prices, allowing us to explore how the context of a transaction can
influence the costs and benefits of using a particular payment instrument. We begin with the
merchant cost estimates presented in the two U.S. industry cost studies reviewed above. Then
we expand on those studies by fixing transaction size, broadening the analysis to include all
parties to a transaction, and adding in benefits where possible. We highlight the dramatic
changes that these modifications bring. Those instruments that appear expensive to merchants
are not expensive for the economy as a whole. In general, the steady shift toward a cashless
society appears to be a beneficial one.
As a cautionary note, remember that calculations like the ones presented below can
convey a false sense of precision. In order to highlight the often small differences in costs and
benefits across payment instruments, the figures in the tables below are reported at the nearest
cent. Most of these numbers are rough estimates. All of the numbers should be interpreted as
suggestive but not definitive. Nonetheless, our assumptions are reasonable and supported by the
available evidence and, as such, the calculations resting on them are informative. While they do
not provide precise differences in costs and benefits among various payment methods, they do
110 Determining whether payment card use is actually subsidized is complicated by the fact that cards are used in
two-sided markets (see footnote 10). What some scholars call a subsidy may in fact be a pricing mechanism
intended to balance the two sides of the market. Moreover, removing the cash/check to payment card “subsidy”
may not improve social welfare. In a recent paper, Rochet and Tirole examine card pricing and social welfare in
a theoretical model where consumers may pay with either a card or cash. The authors examine the effects of
lifting the no-surcharge ban, which prevents merchants from charging cardholders a higher price than customers
paying in cash. Rochet and Tirole find that merchants would raise the price for cardholders and lower it for cash
users, but the social welfare implications are “ambiguous.” In other words, forcing card using consumers to pay
a higher price for goods than cash using consumers does not necessarily raise social welfare and in fact could
reduce it. Jean-Charles Rochet and Jean Tirole, “Cooperation Among Competitors: Some Economics of Payment
Card Associations,” Rand Journal of Economics, Vol. 33, Winter 2002.
111 In fact, this point was made clear by the Dutch retailer study reviewed earlier, which found that what was best
for merchants was not best for the economy as a whole.
reveal relationships and patterns. Thus, the analysis presented here is useful in examining
general welfare issues and goes considerably beyond the existing literature.
A. Replacing Cash and Checks at the Grocery Counter
Our first case study examines payment instrument use in grocery stores.112 The same
pattern we observe in Table 1 for consumer-to-business transactions as a whole is apparent in
supermarket transactions: payment cards are replacing cash and checks. Figure 2 reports the
share of grocery store purchases allocated to different payment methods from 1994 through
2001. While checks fall from around 50 percent to near 30 percent and cash declines from 40
percent to below 30 percent, debit rises from under four percent to over 20 percent and
credit/charge increases from around 5 percent to almost 15 percent of grocery purchases.
We start with the Food Marketing Institute (FMI) cost study of 1998 (see Table 2
below), which reports the costs that grocers face in accepting various payment instruments.113
112 See the Appendix for full details on the calculations presented below.
113 Food Marketing Institute, “A Retailer’s Guide,” supra note 23, at 3. As explained earlier, the FMI 2000 study
did not include as much detail as the 1998 study nor were as many payment categories included. As a result, we
rely on the 1998 figures with the exception of the 2000 time observations, as explained below. Given the paucity
of retailer cost data estimates, we feel the publicly available FMI data are our best option.
Figure 2. Share of Sales Volume by Payment Method at Grocery Stores: 1994–2001
1994 1995 1996 1997 1998 1999 2000 2001
Share of Merchant Volume (%)
Credit Card
Debit Card
Source: Visa U.S.A., 1994-2001.
Costs vary considerably among grocers due to size, geographic location, age and sophistication
of the equipment used, among other things. We employ FMI’s average cost figures in the bulk
of our calculations, but examine the high and low figures later in sensitivity analysis. The
variable “tender time” captures the cost of the time that register attendants must spend to
process payment for a transaction, not counting the time to ring up individual items. “Deposit
preparation” measures the labor costs of preparing a typical paper bank deposit for a
merchant’s bank account, such as counting cash and reconciling the register drawer. Electronic
payment methods are cleared online at the point of sale and therefore do not involve any
deposit preparation. “Bank charges” are explicit fees, such as a deposit fee for cash and checks
or bank processing fees for payment cards. “Other direct costs” capture miscellaneous costs
associated with each type of payment, such as check losses and collection fees, credit card
losses,114 and armored car costs for transporting cash to a bank. Finally, the cost per $100 of
sales translates the reported costs, which apply to the average transaction size for that payment
type, to $100 worth of sales.
114 This category covers “charge backs” for merchants. Merchants are only responsible for these refused charges
under certain circumstances. When a card issuer or consumer questions a transaction because of a stolen card,
dispute over receipt of goods or cashier error, retailers have to gather documentation from stores as proof of
purchase. If they do not deliver this evidence within a set timeframe the disputed amount is charged back to the
In order to provide accurate and current numbers, most of the figures in the above table
are modified from the original FMI report.115 First, the 1998 “tender time” cost FMI reported
was based on time observations taken in 1994. Because processing time for payment card
transactions has fallen considerably since then, we use the tender times (in seconds) reported by
FMI in 2000. We then adjust the tender time and deposit preparation costs for all payment
types to reflect wage changes since 1997, the year FMI calculated the cost figures. We also
update the bank charges for cash and checks to reflect inflation and for payment cards to reflect
increases in processing fees.116 Finally, we update other direct costs for cash (which includes
armored car transport) to reflect inflation.
A couple of items in Table 2 should be highlighted. First, following FMI, the figures in
each column of the table are calculated for a different transaction size––the typical size for each
payment instrument. For example, the average cash transaction was $11.52 and the average for
115 See the Appendix for a replication of the 1998 FMI table.
116 Note that PIN debit bank processing fees have been escalating over time. The 1998 FMI study reports average
PIN debit transaction fees of 13 cents while the 2000 study reports 20 cents. Several PIN networks recently
Table 2. Per Transaction Processing Costs for Various Payment Instruments
Grocery Store Merchants ($)
Check Credit/
Charge Signature
Debit PIN Debit
Tender Time 0.11 0.26 0.24 0.18 0.18 0.17
Deposit Preparation 0.004 0.03 0.03 0 0 0
Bank Charges 0.004 0.06 0.16 0.94 0.56 0.41
Other Direct Costs 0.16 0.27 0.03 0.02 0 0
Per Transaction Total Cost 0.27 0.63 0.46 1.14 0.75 0.57
Average Purchase for
Payment Type 11.52 54.24 54.24 44.50 33.00 41.05
Cost Scaled to $100 of Sales 2.35 1.16 0.85 2.56 2.26 1.39
Notes: Numbers may not add due to rounding.
1Non-verified checks are manually authorized at the checkout counter. Verified checks are
electronically authorized to have sufficient funds in the account to cover the purchase.
Source: Food Marketing Institute, “A Retailer’s Guide to Electronic Payment Systems Costs,” 1998,
updated by authors as detailed in the Appendix.
checks was $54.24.117 Certain costs, such as some bank fees, are sensitive to the purchase
amount, so costs presented this way are difficult to compare, as discussed earlier. Second, the
relative costs of the various payment instruments change considerably when the costs are
translated to $100 of sales. When costs are scaled-up in this fashion, a credit card transaction is
only around 9 percent more expensive than a cash transaction for grocers.118
The costs reported in Table 2 say nothing about whether the economy benefits when
payment cards replace cash and checks for grocery purchases. In keeping with our cost-benefit
approach, three adjustments are needed to analyze welfare. First, to facilitate comparisons we
need to recalculate costs at some fixed transaction size. We choose the typical cash transaction
size and the typical check transaction size because these are the transactions that are being
replaced by payment cards. Moreover, for grocery purchases these two transaction sizes
represent the lowest and highest average purchase sizes. Second, we need to consider other
parties involved in the transaction. Finally, where possible, we need to calculate the benefits
accruing to the parties involved. We present the results of the first step, setting the transaction
size, in Table 3.
1. Setting the Transaction Size
To make comparisons across payment instruments meaningful, we calculate costs for
two fixed transaction sizes.119 Several of the cost categories collected by FMI vary with
transaction size. Bank charges for cash, credit, charge and signature debit cards are dependent
on purchase size, for instance. And check and credit card losses are reported as a percentage of
raised their rates (including NYCE, Visa’s Interlink, and MasterCard’s MoneyLink). (See David Breitkopf,
“Study: PIN Fees Are Up But Won’t Become Norm,” American Banker, Monday August 19, 2002.)
117 This problem is especially acute at the grocery store, where buying a gallon of milk and a loaf of bread is quite
a different transaction than shopping for an entire cart full of groceries. Supermarkets, in fact, sometimes devote
a separate checkout lane to purchases of 15 items or less.
118 Even this comparison is not accurate, however, since not all costs increase with transaction size, thus scaling
can misrepresent actual cost.
119 The fact that the average transaction size for cash differs from that for checks (and payments cards, as well)
indicates that consumer preferences for payment mechanisms vary by transaction size. William Whitesell
develops a model that explains some of the preference differences (see William C. Whitesell, “The Demand for
Currency versus Debitable Accounts,” Journal of Money, Credit, and Banking, Vol. 21, No. 2 (May 1989)). See
also Anthony M. Santomero and John J. Seater, “Alternative Monies and the Demand for Media of Exchange,”
Journal of Money, Credit, and Banking, Vol. 28, No. 4 (November 1996, Part 2).
the sales value.120 Table 3 presents the results of fixing the transaction size across payment
instruments, employing both the average cash size and the average check size.
Choosing a fixed transaction size induces a dramatic change in terms of relative costs,
even before considering all parties involved in a transaction. Taking the typical cash transaction
size in the second row, the costs of credit cards and non-verified checks fall considerably as
compared to row one. Examining the typical check transaction size in the last row, on the other
hand, credit and charge cards are twice as expensive for merchants as checks are. Cash remains
the cheapest payment method for grocers at each transaction size.
2. Adding Other Parties to the Transaction
Next we add other key parties to the transaction. In particular, we add the costs of using
or processing each payment instrument for consumers, the central bank, and commercial banks.
Table 4a adds these costs based on the average cash transaction size of $11.52, while Table 4b
uses the average transaction size for checks of $54.24.
120 Food Marketing Institute, “A Retailer’s Guide,” supra note 23, at 20.
Table 3. Per Transaction Processing Costs for Various Payment Instruments
Grocery Store Merchants, Fixed Transaction Size ($)
Check Credit/
Charge Signature
Debit PIN Debit
Merchant Cost, Avg.
Transaction Size
(Repeated from Table 2)
0.27 0.63 0.46 1.14 0.75 0.57
Merchant Cost, Cash
Transaction Size $11.52 0.27 0.41 0.44 0.61 0.68 0.57
Merchant Cost, Check
Transaction Size $54.24 0.28 0.63 0.46 1.21 0.81 0.57
Source: Authors’ calculations as detailed in the Appendix.
Table 4a adds several elements to the analysis. For merchants, the FMI study omits at
least two components of the marginal cost of payment instruments: theft/counterfeit loss for
Table 4a. Per Transaction Processing Costs for Various Payment Instruments
Grocery Store Cash Transaction $11.52 ($)
Check Credit/
Charge Signature
Debit PIN Debit
Theft/Counterfeit 0.03 0 0 0 0 0
Float 0.001 0.002 0.002 0.002 0.002 0.001
Costs From Table 3 0.27 0.41 0.44 0.61 0.68 0.57
a) Merchant Marginal Cost 0.30 0.42 0.44 0.61 0.68 0.57
Processing Time 0.14 0.33 0.31 0.23 0.23 0.21
Queue Time 0.14 0.33 0.31 0.23 0.23 0.21
Explicit Price 0.03 0.04 0.04 0 0 0.13
Implicit Price 0.28 0 0 0 0 0
Seigniorage 0.07 0 0 0 0 0
b) Consumer Marginal Cost 0.65 0.70 0.65 0.46 0.46 0.55
Central Bank
Production 0.001 0 0 0 0 0
Processing 0.002 0.03 0.03 0 0 0
c) Central Bank Marginal Cost 0.004 0.03 0.03 0 0 0
Commercial Banks
ATM Maintenance 0.06 0 0 0 0 0
Production 0 0 0 0.01 0.01 0.01
Processing 0.004 0.12 0.12 0.29 0.26 0.26
Card Rewards 0 0 0 0.01 0 0
d) Commercial Bank Marginal Cost 0.07 0.12 0.12 0.32 0.28 0.28
e) Sum of Marginal Costs (double count) 1.02 1.27 1.24 1.39 1.42 1.40
f) Social Marginal Cost (no double count) 1.01 1.18 1.05 0.96 0.92 1.00
Notes: Numbers may not add due to rounding. Sum of MC, e), is equal to a + b + c + d. Due to transfers
across parties (such as merchant payments to banks), this figure double counts some cost
elements. Social Marginal Cost, f), eliminates any transfers that give rise to double counting.
Source: Authors’ calculations as detailed in the Appendix.
cash and float loss for each instrument.121 Card issuers bear the brunt of any payment card float
cost. They pay acquirers for any purchases within one to two days of the transaction and
acquirers then pass those payments on to merchants. The cardholder does not pay the card
issuer until the monthly bill arrives, meaning the issuing bank funds credit and charge card float
for an average of 25 days.122 An issuer’s float cost is included, however, in the interchange fee
that issuers charge acquiring banks, and that acquirers then pass on to merchants in the form of
the merchant discount. As a result, card issuer float costs are already counted in the bank fees
in Table 2. The merchant still faces a one- to two-day wait for credit and charge card
reimbursements, however. We include this cost in Tables 4a and b. For checks and signature
debit cards, the merchant typically waits two days before receiving funds from the customer’s
bank; with PIN debit the delay is typically one day.123 We also assume merchants take a day on
average to deposit cash receipts.
The above calculations also include other parties to the transaction, starting with payers.
The opportunity cost of the time that consumers spend at the register while a transaction is
processed is captured by “Consumers: Processing Time.” Consumers waiting in line also face
an opportunity cost for their time while the current payer’s transaction is being conducted,
which is captured by “Queue Time.” Both of the time costs likely will vary with payers’ wage
rates; our calculations rely on typical earnings.124 Next we include the explicit price that
consumers pay for an instrument. For cash, we include the average within network (“on us”)
ATM fee for cash withdrawals, scaled by the percentage of people facing such charges. To the
extent that consumers access off-network ATMs, we are understating the explicit price for cash
121 Recall that check loss and credit card charge-backs are already included in FMI’s estimate for Other Direct
Costs. Thus, adding the marginal cost of cash theft and counterfeit loss here provides a more even treatment of
payment instruments and makes the costs more comparable.
122 “Survey of Credit Card Plans,” Federal Reserve Board (credit terms as of July 31, 2003) (visited Nov. 10,
2003) <>.
123 For checks, see Stavins, “Comparison of Social Costs,” supra note 15. For debit cards, see Ingenico supra note
22. For credit cards, see “Survey of Credit Card Plans,” supra note 122.
124 There is some debate within the economic literature on whether wage rates accurately reflect the opportunity
cost of time. (See, for instance, W. Douglass Shaw, “Searching for the Opportunity Cost of an Individual’s
Time,” Land Economics, Vol. 68, February 1992.) In particular, this calculation assumes that work and non-
work hours can be substituted at will, which is not always the case. As a result, wage rates may overstate or
understate the opportunity cost of time. Recognizing this, we conduct sensitivity analysis on the wage measure
used to calculate time costs.
faced by consumers.125 In the case of checks, this is the price consumers pay for each paper
check. Because credit and charge cards typically only charge an annual fee for rewards cards
(where the rewards will offset the fee to some extent), and finance charges only apply to credit
services and not transaction services, the explicit marginal cost to consumers of using these
cards for another purchase is zero.126 Many issuers do, however, charge per transaction fees for
PIN debit.127 We scale the average PIN fee by the percentage of institutions assessing a fee.
As the last cost item for consumers we add the implicit cost of obtaining cash.128 The
first component of this cost is comprised of the time that consumers spend in making a trip to
an ATM.129 Seigniorage is the second implicit cost for cash. Seigniorage can be defined in
many ways, but the definition most closely matching our cost-benefit exercise is fiscal
125 In 2001, only 10 percent of banks charged a fee for own-customer ATM access. However, over 78 percent of
banks charged a fee for withdrawals out-of-network and that fee averaged $1.17. Over 88 percent of the
operators of the “foreign” network tacked on an additional fee, which averaged $1.32. Thus the typical charge
for getting cash from an off-network ATM was around $2.50 in 2001. Timothy H. Hannan, Retail Fees of
Depository Institutions, 1997-2001, Federal Reserve Bulletin, September 2002. The rise in ATM fees appears to
be one of the factors driving the popularity of debit cards. See Joanna Stavins, “Effect of Consumer
Characteristics on the Use of Payment Instruments,” New England Economic Review, Issue Number 3, 2001, p.
126 Only 15 percent of the 410 million bankcard accounts in 2002 had an annual fee. (See James J. Daly, “Issuers
Shed Some Profitability,” Credit Card Management, May 2003 Vol. 16 No. 2, p. 36.) Moreover, annual fees
represent fixed, not marginal costs. Using a credit card for transactions implies that the balance is paid in full
when the bill arrives. We consider credit purchases below when we add benefits for each payment method.
127 According to a recent survey conducted by Dove Consulting Group Inc., 26 percent of financial institutions
charge their debit cardholders a per-use fee, with fees ranging from 25 cents to one dollar. Joan Goldwasser,
“Check Card Changes Ahead,” Kiplinger's Personal Finance, Oct. 20, 2003.
128 For non-cash payment methods, the POS time factor (already captured in processing time) represents the bulk
of any implicit costs. Checks involve carrying identification and cumbersome checkbooks––which do not fit into
the average wallet. Thus women, who usually carry purses, write a greater proportion of checks than men. Jackie
Spinner, “The Check’s Last Writes,” Washington Post, Feb. 9, 2003; Gregory Richards, “Check Use Continues
to Decline as People Favor Credit, Debit Cards,” Florida Times-Union, Mar. 31, 2003. We were unable to
quantify these implicit costs, however.
129 Some consumers at some times will not need to make a special trip to the ATM, but will instead obtain cash
while taking care of other errands. At other times, consumers may need to go considerably out of their way to
visit an ATM. Note that only 20 percent of supermarkets contain an in-house cash machine (Food Marketing
Institute, “2003 Security and Loss Prevention Issues Survey,” 2003, at 2). Moreover, non-banks (called “ISOs”)
operate at least thirty percent of all ATMs located outside of banks (e.g., in groceries and airports, for instance).
(David Gosnell, “ATM portfolios grow through acquisition,” ATM&Debit News, 25 June 2002.) ISOs charge for
each and every withdrawal. Even if a bank were operating the ATM in a particular grocery, many of the
grocery’s customers would have accounts at different banks and thus would be subject to off-network ATM fees.
Since the typical off-network charge is $2.50 per extraction, obtaining cash at the grocery store would save the
time cost of driving to an ATM but would cost most consumers significantly more in fees. On the whole, our
ATM time assumptions seem reasonable for the implicit cost of obtaining cash, especially considering that we
assume no waiting time to use an ATM (i.e., no line) and do not count gasoline costs for driving to an ATM.
seigniorage––the profit from printing currency, net of expenses, that the central bank actually
has available for budgeting purposes.130 Cash users pay seigniorage to the Federal Reserve as
an implicit transfer. We therefore list this item as a cost for consumers.131
The central bank is the second party added to the calculations.132 The Federal Reserve
Bank incurs processing costs for handling currency and removing unfit bills. The central bank
subsidizes the use of cash in coordinating cash supplies for banks and maintaining a high level
of cash quality without charge to consumers or banks.133 The Federal Reserve also acts as an
intermediary in check processing for the majority of checks that cross commercial banking
networks, although it passes these costs on to commercial banks.134 Finally, the central bank
incurs production costs for cash. Unlike metal coins, paper currency has a short lifespan. With
each transaction, a bill becomes more soiled and worn. The Federal Reserve typically destroys
130 For technical definitions of fiscal and other kinds of seigniorage, see Manfred J.M. Neumann, “Seigniorage in
the United States: How Much Does the U.S. Government Make from Money Production?,” Review – Federal
Reserve Bank of St. Louis, Mar/Apr 1992. Note that the central bank operates on a non-profit basis, turning all of
its profits over to the Treasury department.
131 It is listed again in later tables as a revenue (benefit) to the central bank. Seigniorage therefore cancels out at the
social level, although it affects private estimates.
132 Some economists might object to including the central bank in a cost-benefit calculus, on the grounds that it is
not an individual. Any costs incurred by the government, or even quasi-government entities like the Federal
Reserve, are ultimately borne by tax paying individuals. (See, for example, Steven E. Landsburg, The Armchair
Economist: Economics and Everyday Life, New York: Free Press (1993), at chapter 10.) We agree, in principle.
In practice, however, it is impossible to trace these costs to their final destination. We choose instead to consider
costs at their first instance.
133 See Jeffrey M. Lacker, “Should We Subsidize the Use of Currency?,” Federal Reserve Bank of Richmond
Economic Quarterly, Vol. 79/1 Winter 1993. As we noted earlier, the question of subsidization has been raised
for several different payment instruments. Humphrey and Berger (1990) argue that checks and credit cards are
both subsidized (by commercial banks in the case of checks and indirectly by merchants in the case of credit
cards). It appears that commercial banks subsidize PIN debit, as well. To provide merchants with an incentive to
lease and install special PIN pad equipment, bank associations initially set the PIN debit price structure with
relatively low interchange fees as compared to signature debit and credit cards. Signature debit transactions, on
the other hand, are processed using credit card equipment already in place at most merchants. Now, however,
after many merchants have already installed PIN pad equipment, debit-processing banks appear concerned they
are not covering costs. (The number of PIN pads installed increased from 53,000 in 1990 to 529,000 in 1995 to
more than 4 million in 2002. Faulkner & Gray, 1996 Debit Card Directory, at 28; The Nilson Report, No. 785,
April 2003.) For example, Corus Bank in Chicago reportedly loses around 2.5 cents on every PIN debit
transaction its customers make. (See David Breitkopf, “PIN Debit Fees Becoming A Math Problem For Banks,”
American Banker, July 23, 2003.) The issue of subsidies is clearly an important one when considering the net
social benefits of shifting to a cashless society. This issue, and its policy ramifications, is an interesting topic for
future research.
134 By law, the Federal Reserve charges its cost plus an adjustment factor intended to simulate a normal economic
profit. “Monetary Control Act of 1980,” at §11A (available at
the most frequently used 5-dollar bills after just 15 months in circulation.135 We therefore
interpret cash production costs as a marginal cost for each transaction.136
Commercial banks are the third and final party added to the calculations. Just as with
paper currency, the magnetic stripe on plastic cards wears out over time and must be replaced.
Only about a tenth of the magnetic stripes on credit, charge and debit cards fail within two
years, but most banks replace their plastic cards within this time period to preempt any
problems.137 Commercial banks must also maintain ATM machines and refill them with cash
periodically.138 We allocate the monthly maintenance costs for ATMs across monthly grocery
transactions to obtain a per transaction cost. Commercial banks also expend resources in
processing checks.139 In addition to in-house processing costs, commercial banks must pay the
Federal Reserve (and any other intermediary) for any checks it processes as an intermediary.
Commercial banks expend resources to process, sort, and handle paper currency as well.
Banks incur processing costs for payment cards as well. For example, issuers fund
consumer credit card float for an average of 25 days and signature debit float for around 2 days.
Acquirers facilitate transaction clearing between card issuers and merchants.
Finally, card issuers expend resources to provide consumers with rewards, such as
airline miles or Discover’s “cash back” promise. Around twelve percent of credit or charge
cardholders have cards that provide rewards of some kind, counting American Express,
135 The average lifespan of both $1 and $10 bills is only 18 months. “Dollars and Cents, The Circulation of
Money,” Federal Reserve Bank of Atlanta, (visited February 9, 2004)
136 For paper checks, the marginal cost nature of production is obvious: each check is used only once.
137 Rebecca Cox, “Cost-Conscious Banks Seek Ways to Prolong Life of Plastic Cards,” American Banker, May 3,
1989. The Nilson Report, No. 759 (March 2002); The Nilson Report, No. 760 (March 2002); The Nilson Report,
No. 784 (March 2003); The Nilson Report, No. 785 (April 2003).
138 This is the regular upkeep cost of the machine, not the initial installation cost, which is considered overhead
and not a per transaction cost. Maintenance includes cash replenishment, servicing, telephone costs, rent, and so
on. See 2003 American Bankers Association ATM Fact Sheet, available online,
139 See Stavins, “Comparison of Social Costs,” supra note 15 for a description of the tortuous path that the typical
personal check travels before satisfying its payment function. The high cost of check processing is one reason
behind bank support of debit cards. As one bank executive observed, “Our philosophy around PIN debit is that,
even without making money off of it, it saves us money because it’s one less check that we need to process.” (As
quoted in David Breitkopf, “Study: PIN Fees Are Up But Won’t Become Norm,” American Banker, August 19,
Discover, MasterCard, and Visa. We estimate the cost of providing loyalty rewards, scaled by
the percentage of cardholders with cards of this type.
We present two summary cost calculations at the bottom of Table 4a. The first simply
sums the four private parties. Because several elements represent transfer payments, such as
commercial bank payments to the Federal Reserve for check processing and merchant
payments to commercial banks for payment card processing, this sum double counts certain
costs. In order to obtain social marginal cost, transfer payments must be subtracted, which is
done in the last line of the table. The social marginal costs in Table 4a are relatively close to
one another. In Table 2, the spread between the most and least expensive instruments was 87
cents; in Table 4a that spread is only around 26 cents. Moreover, the cheapest instrument for
merchants (cash) is not the cheapest instrument for the economy. Counting all parties, signature
debit cards are cheapest, followed closely by PIN debit and credit cards. Paper instruments,
cash plus verified and non-verified checks, emerge as the most costly forms of payment. Thus
adding in the other parties to the transaction has changed the relative cost situation
considerably––at least for small transactions. We turn next to a larger transaction size, the
average check purchase.
Table 4b is analogous to 4a, but highlights how transaction size matters. For instance,
compared to Table 4a, ATM maintenance fees are higher since the costs are spread over fewer
Table 4b. Per Transaction Processing Costs for Various Payment Instruments
Grocery Store Check Transaction $54.24 ($)
Cash Non-
Check Credit/
Charge Signature
Debit PIN Debit
Theft/Counterfeit 0.14 0 0 0 0 0
Float 0.004 0.01 0.01 0.01 0.01 0.004
Costs From Table 3 0.28 0.63 0.46 1.21 0.81 0.57
a) Merchant Marginal Cost 0.43 0.64 0.47 1.22 0.82 0.57
Processing Time 0.14 0.33 0.31 0.23 0.23 0.21
Queue Time 0.14 0.33 0.31 0.23 0.23 0.21
Explicit Price 0.03 0.04 0.04 0 0 0.13
Implicit Price 1.29 0 0 0 0 0
Seigniorage 0.33 0 0 0 0 0
b) Consumer Marginal Cost 1.92 0.70 0.65 0.46 0.46 0.55
Central Bank
Production 0.01 0 0 0 0 0
Processing 0.002 0.03 0.03 0 0 0
c) Central Bank Marginal Cost 0.01 0.03 0.03 0 0 0
Commercial Banks
ATM Maintenance 0.30 0 0 0 0 0
Production 0 0 0 0.01 0.01 0.01
Processing 0.02 0.012 0.12 0.49 0.35 0.34
Card Rewards 0 0 0 0.05 0 0
d) Commercial Bank Marginal Cost 0.31 0.12 0.12 0.55 0.36 0.36
e) Sum of Marginal Costs (double counts) 2.67 1.49 1.27 2.23 1.64 1.48
f) Social Marginal Cost (no double count) 2.66 1.40 1.08 1.22 1.01 1.08
Notes: Numbers may not sum due to rounding. Sum of MC, e, is equal to a + b + c + d. Due to transfers
across parties (such as merchant payments to banks), this figure double counts some cost elements.
Social Marginal Cost, f, eliminates any transfers that give rise to double counting.
Source: Authors’ calculations as detailed in the Appendix.
higher dollar transactions ($54.24 versus $11.52).140 Commercial bank payment card
processing costs rise as well, since the cost of funds for the grace period increases with the
transaction size. The implicit cost of cash increases dramatically as well. At the average ATM
cash extraction of $60, consumers can make only one grocery purchase of $54.24. While
consumers might adjust their cash withdrawal amount to reflect the higher expected purchase,
consumers would have to bear the inconvenience of carrying bulky cash and the increased
value of potential losses or theft. Recall from Figure 1 that consumers use cash to pay for
purchases in the $40-$60 range only one-quarter as often as they do to pay for purchases
costing $5-$10. Clearly consumers perceive either a significant cost to using cash for relatively
large purchases or significant benefits to using other methods like checks and credit cards. We
were unable to estimate these carrying and use costs directly, so as an alternative we assume
that consumers do not adjust ATM withdrawal amounts and thus face a relatively high time
cost when using cash for higher priced purchases.
As with the cash purchase size, adding other parties to the transaction changes the
relative-cost picture significantly as compared to Table 3 at the average check purchase size.
Once other parties are considered, payment cards do not look nearly as expensive vis-à-vis cash
and checks as indicated by merchant costs alone. Signature debit is again the least costly; PIN
debit and verified checks follow. Cash and non-verified checks are the most expensive
3. Quantifying Benefits
We now turn to our third adjustment: quantifying the benefits of payment cards relative
to cash and checks. Tables 5a and 5b present the results.
140 Recall that this cost element is estimated by allocating maintenance costs among potential grocery transactions
generated from the cash dispensed at the ATM. If each grocery transaction costs more, maintenance costs will be
allocated among fewer transactions.
Table 5a adds estimates of per-transaction benefits associated with each of the payment
instruments at the grocery cash transaction size. Even more than the cost estimates, these
figures are approximations given the difficult nature of quantifying benefits. Unfortunately, we
do not have estimates for merchant-specific benefits aside from relative cost savings across
payment instruments, something captured in the cost estimates already.
Table 5a. Adding Selected Benefits
Grocery Store Cash Transaction $11.52 ($)
Check Credit/
Charge Signature
Debit PIN Debit
Float 0 0.002 0.002 0.02 0.002 0.001
Credit Option 0 0 0 0.06 0 0
Record Keeping 0 0.04 0.04 0.04 0.04 0.04
Cash Back Option 0 0.15 0.15 0 0 0.13
Signature Debit 0 0 0 0 0.13 0
Reward Cards 0 0 0 0.01 0 0
Discover Cards 0 0 0 0.004 0 0
Privacy 0.12 0 0 0 0 0
Consumer Marginal Benefits 0.12 0.19 0.19 0.13 0.16 0.16
Central Bank
Processing Revenue 0.002 0.03 0.03 0 0 0
Seigniorage 0.07 0 0 0 0 0
Central Bank Marginal Benefits 0.07 0.03 0.03 0 0 0
Commercial Bank
Processing Revenue 0.003 0.06 0.16 0.42 0.50 0.41
Sum of Marginal Benefits 0.19 0.28 0.38 0.55 0.66 0.57
Sum of Marginal Costs (Table 4a) 1.02 1.27 1.24 1.39 1.42 1.40
Net Social Marginal Cost 0.83 0.99 0.87 0.84 0.76 0.83
Notes: Numbers may not add due to rounding. Net Social Marginal Cost equals the Sum of Marginal Costs
minus the Sum of Marginal Benefits, which removes cross-party transfers.
Source: Authors’ calculations as detailed in the Appendix.
Consumer benefits are listed first. All non-cash payment instruments provide consumers
with some level of float.141 Even PIN debit cards, which withdraw payments directly from a
consumer’s bank account, take a day to process on average. Credit and charge cards provide
consumers with the greatest amount of float. On average, consumers have 25 days from the
date of purchase to the date their card bill is due.142 Card issuers fund this float, recovering the
cost of doing so in the interchange fee.143
Credit cards also give consumers the option to borrow.144 Uncertainty over income or
expenses can lead a consumer to rely on credit cards for payment, even if she intends at the
time of purchase to pay the charge in full when her bill arrives. This benefit is difficult to
measure because it varies with personal finances and available credit options. For instance,
consumers who own their home and have accumulated some equity may qualify for a home
equity loan through their bank. This is a relatively low-cost credit option, but involves
transaction costs to set up and is unavailable to renters. All secured loans involve this
constraint—for many consumers these loans are not available. Credit cards, on the other hand,
offer a convenient source of unsecured credit to a broad cross section of consumers.145 Another
unsecured credit alternative is the payday loan, but this is far more costly than credit card
141 Note that check float has declined considerably over the last ten years as check processing has improved.
According to our review of the literature, however, “the limit to mechanical processing of checks has effectively
been reached. While reader/sorter speeds could be increased, it would just rip the checks and jam the machines.”
(David Humphrey et al., “What does it Cost to Make a Payment?,” Review of Network Economics, Vol. 2, Issue
2, June 2003, at 170 n28.) Faster check processing will likely require moving beyond traditional paper checks, to
point of sale check imaging or truncation/electronic presentment, for instance. The recently passed Check 21 Act
aims at just this sort of advance in check processing. See supra note 21.
142 “Survey of Credit Card Plans,” supra note 122.
143 Thus, as discussed above, the merchant bank charge costs plus merchant float costs will generally offset the
float benefit received by consumers. In effect, float is a transfer payment.
144 Here we consider only the option value of obtaining credit, not the benefit of the credit itself net of the cost of
funds. As Brito and Hartley observe, credit cards “[provide] insurance against unanticipated shocks to
expenditure or income.” Dagobert L. Brito and Peter R. Hartley, “Consumer Rationality and Credit Cards,”
Journal of Political Economy, 1995, vol. 103, no. 2, at 402.
145 See, e.g., Joseph Nocera, A Piece of the Action: How the Middle Class Joined the Money Class, New York:
Simon & Schuster, 1994, at 93-94; Scott B. MacDonald & Albert L. Gastmann, A History of Credit & Power in
the Western World, New Brunswick, NJ: Transaction Publishers, 2001, at chapter 11; Glenn B. Canner &
Charles A. Luckett, “Developments in the Pricing of Credit Card Services,” Federal Reserve Bulletin, September
1992, at 653.
debt.146 An alternative relatively similar to credit cards is a personal line of unsecured bank
credit. These credit lines typically have a minimum amount of several thousand dollars,147 but
the interest rate terms are similar to credit cards.148 Once an unsecured line of credit is
established, consumers must pay an annual fee to maintain it. This fee can be thought of as the
option price of borrowing for the year. We therefore use the typical unsecured loan annual fee
spread over the minimum loan amount and multiplied by the transaction size as a rough
estimate of the per-transaction value of having the option to place a transaction on credit. While
not all consumers exercise the option for credit, at the time of purchase it is likely to have some
value for the majority of consumers.149
All payment methods except for cash provide consumers with a record keeping
mechanism useful for budgeting, planning, and income tax preparation.150 Rather than having
to keep track of each purchase’s paper receipt, checks and payment cards provide itemized
monthly statements. Most payment cards also have online statements accessible anytime.
Using checks and debit cards for grocery purchases can save consumers time outside of
the checkout lane. Grocers commonly offer a cash back option that is free of charge for these
payment instruments. As a result, consumers can reduce their need to make a separate trip to an
146 A typical payday loan charges $17.50 on every $100 increment borrowed for each 15-day period. Thus a person
borrowing $200 for one month would pay $70 in fees, which translates into an APR of 457 percent. PIRG and
CFA report for 2000, see
147 Brito and Hartley quote a bank officer observing that the costs to the bank of processing a loan are so high that
the bank could not afford to make a loan of less than $3,000 for 1 year, except at interest rates that exceed those
charged on credit cards. Brito & Hartley, supra note 144, at 402. Thus, credit cards appear to be the lowest cost
means for borrowing small amounts without security.
148 Wells Fargo Loans & Lines of Credit–Product Comparison, (visited August 7, 2003)
<>; Wells Fargo Loans & Lines of Credit–Products
(visited August 8, 2003) <>; Wells Fargo Loans &
Lines of Credit–Rate Calculator, (visited August 8, 2003)
149 Just as with stock options, a credit option need not be exercised in order to have value at the time of purchase.
We suspect the majority of credit card users value this option.
150 Many consumers value cancelled checks as payment receipts. (See ATM & Debit News, “Star Survey Finds
Even Debit Card Users Still Write Checks At The Point Of Sale,” March 6, 2003.) However, the recently passed
Check 21 Act makes it legal for banks to provide only digital images of processed checks, allowing them to
refuse to return the physical checks to customers or to charge those customers that request returned checks. (See
Lauri Giesen, “The Enduring Check,” Credit Card Management, Nov. 2002; Lynn Koller, “Making a Clear Case
for Online Digital check Imaging,” Bank Technology News, Apr. 2002. The text of the Check Clearing for the
21st Century Act (P.L. 108-100), or Check 21, is available at Thomas: Legislative Information on the Internet
ATM in order to obtain cash for other purchases. We assume this benefit is equal to the implicit
cost of obtaining cash from an ATM, scaled by the average cash-back amount and the
percentage of debit transactions involving cash-back.151
According to survey results, around 30 percent of debit cardholders prefer signature to
PIN; another 30 to 40 percent prefer PIN to signature.152 Like all debit cards, signature cards
offer enforced budgeting. Unlike PIN, signature debit also offers familiar use patterned after
credit cards, dispute resolution procedures, and more extensive merchant acceptance.153 Survey
results indicate that those consumers preferring signature to PIN debit place a significant dollar
value on that preference (from 25 cents a transaction up to 50% of the transaction value).154
However, the survey methodology was distinctively different from the accounting based
estimates we employ for other payment instrument benefits.155 We therefore chose to treat
signature and PIN debit symmetrically and estimate signature-specific benefits at the PIN-
specific cash-back benefit amount. 156
The benefit to consumers from reward cards is calculated analogously to its cost for
issuers. Again, this benefit is scaled by the percent of cardholders with a reward card.
<>.) As more banks take this cost saving route, the value of checks as record keeping
instruments will likely decline.
151 Consumers could prefer other payment mechanisms (cash, credit) for the majority of their transactions and use
PIN or checks for cash-back only when they need to replenish their cash supplies.
152 “Debit Card Study,” BAI Global: 4316, June 2, 1999, submitted as Expert Report of David W. Stewart, In re
Visa Check/MasterMoney Antitrust Litig. (E.D.N.Y. 2000) (No. CV-96-5238).
153 Signature debit is by far the more prevalent of the two debit forms. By number of transactions signature debit is
used over twice as often as PIN and by sales volume use is nearly three times as large as PIN. The Nilson Report,
No 784 (March 2003); The Nilson Report, No. 785 (April 2003).
154 If signature debit benefit estimates based on the survey instrument are used in the calculations above, signature
debit emerges as significantly less costly than all other forms of payment.
155 The survey calculates the value of a consumer’s entire preference for signature over PIN debit, as opposed to
estimating the value of the individual attributes a consumer may like about signature debit. Preference values for
those consumers preferring PIN to signature debit were not included in the survey.
156 Preferences for one debit card may be stronger than those for the other, but showing this would require
comparable data for both payment instruments. Lacking that data, we opt for symmetric treatment. Consumers
evidently perceive a number of differences between signature and PIN debit, most notably over the value of cash
back and the actual level of security provided. Some debit users appear to believe that PIN debit is more secure
than signature, since merchants rarely check signatures and would-be thieves do not know their PINs. Others,
however, believe that signature is more secure because individual signatures are difficult to forge while a PIN
can be observed as the card owner enters it on a PIN pad. See, for example, David Breitkopf, “Customers Get
Savvier in Debit Debate,” American Banker, May 6, 2003. The realities of payment card security are not relevant
in this decision process––to a very large extent consumers select payment instruments to use and thus their
beliefs are what matters.
The last item listed for consumers is privacy. Some consumers worry that their
purchases will be tracked if they pay with a check or plastic card that displays personal
information. This can be a particular concern at grocery stores, where loyalty cards help
merchants capture customers’ buying habits and influence stocking decisions. Cash offers
consumers an anonymous method of paying. We estimate this benefit using loyalty card
discounts. At least in some retail venues, supermarkets included, shunning a loyalty card means
forgoing product discounts. We can view these discounts, then, as the implicit benefit of
providing personal information. Consumers paying with cash (and not providing a loyalty card)
must value their privacy at least as much as the discount forgone.
Two items are listed for the central bank. First, it earns a profit on the currency it
produces (seigniorage). Second, it recoups its expenses in processing checks. Both of these
revenues must be included in the social calculation to obtain net marginal costs. For the same
reason, commercial bank processing revenues are included as well.
Note that the sum of marginal costs, as opposed to social marginal costs, is used in
Table 5a. This is necessary so as not to double count transfer payments as revenues.157
157 The Appendix describes this issue in greater detail.
Table 5b is analogous to 5a, presenting estimates for the various benefits at the typical
check-size grocery purchase. Record keeping and cash-back, as calculated here, do not depend
on purchase size.158
158 It is possible that the amount of cash-back varies with transaction size but we do not have sufficiently detailed
information on this benefit to make purchase-size-specific calculations.
Table 5b. Adding Selected Benefits
Grocery Store Check Transaction $54.24 ($)
Check Credit/
Charge Signature
Debit PIN Debit
Float 0 0.01 0.01 0.10 0.01 0.004
Credit Option 0 0 0 0.27 0 0
Record Keeping 0 0.04 0.04 0.04 0.04 0.04
Cash Back Option 0 0.15 0.15 0 0 0.13
Signature Debit 0 0 0 0 0.13 0
Reward Cards 0 0 0 0.06 0 0
Discover Cards 0 0 0 0.02 0 0
Privacy 0.54 0 0 0 0 0
Consumer Marginal Benefits 0.54 0.19 0.19 0.48 0.17 0.16
Central Bank
Processing Revenue 0.002 0.03 0.03 0 0 0
Seigniorage 0.33 0 0 0 0 0
Central Bank Marginal Benefits 0.33 0.03 0.03 0 0 0
Commercial Bank
Processing Revenue 0.02 0.06 0.16 1.01 0.63 0.41
Sum of Marginal Benefits 0.89 0.28 0.38 1.49 0.80 0.57
Sum of Marginal Costs (Table 4b) 2.67 1.49 1.27 2.23 1.64 1.48
Net Social Marginal Cost 1.79 1.21 0.89 0.74 0.85 0.92
Notes: Numbers may not add due to rounding. Net Social Marginal Cost equals the Sum of Marginal Costs
minus the Sum of Marginal Benefits, which eliminates cross-party transfers.
Source: Authors’ calculations as detailed in the Appendix.
Taking all of the above benefit estimates into account, the position of payment cards
relative to cash and checks improves considerably, as Tables 5a and 5b illustrate. For either the
average cash or check transaction size, a payment card is the cheapest instrument. Cash and
non-verified checks are among the most costly instruments.
Compared to the above estimates, merchant cost studies do indeed present a misleading
picture for policy makers: the ordering of payment instrument costs reversed entirely after
multiple parties were added and benefits included. Table 6 illustrates how the relative rankings
shift as other parties to a transaction are included and as benefits enter into the calculations.
Cash falls from the cheapest instrument (rank = 1) to last place. Credit cards, in contrast, move
from the most expensive (rank = 6) to the least. Signature debit cards drop from fifth place to
Note that the preceding calculations disregard the value of many other benefits that are
difficult to quantify. First, merchants are not included in the benefits calculations. While many
of the relative benefits of one payment instrument versus another are captured in labor cost
savings (reduced POS labor) or in lower processing fees already captured in these calculations,
it is likely that some direct benefits to merchants are ignored here. For example, if the reduction
in transaction costs for consumers due to a better match between preferred and accepted
payment methods discussed at the beginning of this paper translates into increased
consumption, then payment cards could result in somewhat higher sales for merchants
(although the amount might be insignificant for low price purchases like groceries). We were
unable to estimate any direct merchant benefits.
Table 6. Payment Instrument Cost Rankings
Grocery Store Check Transaction $54.24
Cash Non-Verified
Check Verified
Check Credit/
Charge Signature
Debit PIN
Grocer Marginal Costs 1 4 2 6 5 3
Social Marginal Costs 6 5 3 4 1 2
Net Social Marginal Costs 6 5 3 1 2 4
Source: Rows indicate the relative cost rankings in Tables 3, 4b, and 5b, respectively.
Many of the instruments offer benefits that are difficult to value. For instance, payment
cards offer the option of consolidating payments––everything can be charged on one card with
one bill to pay at the end of the month. For both consumers and merchants, checks and payment
cards provide improved theft and loss prevention as compared to cash, as well as easier dispute
resolution in the event of problems.159 Credit, charge and signature debit cards provide added
protection on this front because the card systems take an active role in dispute resolution.160
PIN debit does not provide this benefit as the money is immediately withdrawn from the
consumer’s bank account and any dispute resolution is left to the consumer and the merchant,
just as with a cash purchase.161 All of these benefits are important and help to explain the
pattern of observed payment instrument use. The difficult nature of estimating their value
implies that we cannot explicitly account for them in Tables 5a and 5b. Despite the many items
omitted from the calculations––and the rough nature of the estimates that are included––our
calculations track observed payment instrument use at grocery stores fairly well. Payment cards
are increasingly used to make grocery purchases and these are the instruments offering
consumers the lowest net cost. While the numbers are only indicial, the relative rankings appear
to be in line with behavior.
B. The Decline of Cash and Checks at Discount Stores
For our second case study, we examine discount stores. Just as with grocery stores,
although less dramatically, discount stores have seen a shift in customer payment methods over
time. Figure 3 presents the share of sales volume for credit, debit, cash, and checks from 1994
through 2001. The proportion of sales paid for with credit cards is mostly unchanged,
remaining between 25 and 30 percent for most of the period. Debit cards, however, rise from
essentially zero to over 20 percent––while cash and checks each decline by at least 10 percent.
159 As one author states “…the decision to use a check or credit card versus other payment options can be viewed
as the purchase of a low cost insurance contract, which limits future potential payment problems at the cost of
some marginal inconvenience and expense [as compared to other electronic methods of paying].” Mantel, supra
note 9, at 26.
160 With signature debit, however, the money is often deducted before the consumer is aware of the fraud, making
the resolution process somewhat more time consuming because reimbursements can be involved.
161 For a discussion of consumer dispute resolution concerns, see Mantel, supra note 9.
To evaluate discount store payment processing costs and benefits, we start with the
merchant costs reported by Coopers & Lybrand (1995), discussed above in Section III. Recall
that the Coopers study relied on cost data for only one “representative” merchant. We present
this case study as a point of comparison to the grocer study. Discount stores are similar to
grocery stores in several ways. Average transaction sizes are quite close, often within a few
dollars for a given payment type (as Table 7 below demonstrates). The point of sale structure is
similar as well––both typically have multiple checkout lanes positioned near the store exit.
Thus, we would expect discount store payment processing costs to be similar to grocery costs.
The cost estimates presented here can therefore be used as a check on the sensibility of the FMI
numbers. The methods used to calculate non-merchant costs are identical to the grocer case,
varying only because of slight differences in transaction sizes and reported differences in point
of sale time. As a result, we present only the initial merchant cost table and the final net
marginal social cost table, which incorporates benefits.162 Table 7 presents discount merchant
store costs, based on Coopers & Lybrand.163
162 The original costs reported in the Coopers’ study, as well as the intermediate cost tables we calculate, are all
presented in the Appendix.
163 The numbers reported are updated using the same approach as the grocery case study.
Figure 3. Share of Sales Volume by Payment Method at Discount Stores: 1994–2001
1994 1995 1996 1997 1998 1999 2000 2001
Share of Merchant Volume (%)
Credit Card
Debit Card
Source: Visa U.S.A., 1994-2001. Credit volume includes charge cards.
The Coopers & Lybrand study includes a slightly different set of marginal processing
costs. “POS Time” is equivalent to grocery tender time. “Drawer/Office,” however, includes
the marginal cost portion of back office cashier/drawer balancing, which FMI excludes from its
numbers. The Coopers study also includes merchant float costs and losses for cash, not just for
checks and payment cards as with FMI.164
The Coopers study provides greater detail on payment cards than FMI did for grocery
stores. While FMI reported processing fees for the different payment card systems in one of the
backup tables,165 grocery store costs were not reported separately by card system. The FMI
study groups all credit and charge cards together, obscuring the processing cost differences
apparent in Table 7. According to the Coopers’ data, American Express costs discount
merchants twice as much as Discover and fifty percent more than Visa/MasterCard.
164 Recall that we add float costs and cash loss/theft to grocery store processing costs in Table 4a.
165 Food Marketing Institute, “A Retailer’s Guide,” supra note 23, at 20.
Table 7. Per Transaction Processing Costs for Various Payment Instruments
Discount Store Merchants ($)
Cash Check
Credit Discover American
Express Signature
Debit* PIN Debit*
POS Time 0.04 0.12 0.07 0.07 0.07 0.07 0.07
Drawer/Office 0.01 0.11 0.01
0.01 0.01 0.01 0.01
Bank Charges 0.01 0.09 0.80 0.54 1.25 0.75 0.44
Losses 0.003 0.16 0.01
0.01 0.01 0.01 0
Float 0.001 0.01 0.01
0.01 0.01 0.01 0.003
Other Costs 0.02 0.001 0 0 0 0 0
Per Transaction Total
Cost 0.08 0.49 0.90
0.64 1.35 0.85 0.52
Average Receipt for
Payment Type 15.49 43.93 41.21 37.60 48.43 41.92 41.07
Cost Scaled to $100 of
Sales 0.49 1.11 2.18 1.70 2.79 2.03 1.26
Notes: Numbers may not add due to rounding.
* PIN and signature debit are imputed as described in the text and the Appendix.
Source: Coopers (1995), updated by authors as detailed in the Appendix.
Because the Coopers study collected data from 1994, debit use was reported as zero by
the discount store that was surveyed. As Figure 3 illustrates, this is a key payment instrument
with increasing use at discount stores, so we impute the costs for signature and PIN debit based
on reported costs for Visa and MasterCard credit cards, debit merchant discount fees, and
information about the relationship of debit to credit in the FMI study.166 Float is calculated
directly in the same manner as for grocery stores.
While some of the individual cost elements are different, discount store processing costs
are quite similar to grocery store costs. Cash is the one instrument with a significantly different
processing cost overall: 30 cents for grocery stores versus 8 cents for discount stores. Two
marginal cost elements appear to drive the difference. First, time costs at the point of sale are
far lower for discount stores as compared to grocery stores. Grocery tender time for a cash
transaction costs around 11 cents; for a discount store merchant that cost is roughly 4 cents.
Second, other direct costs are lower for discount than for grocery stores, costing discount store
merchants around 2 cents but costing grocery merchants almost 16 cents. In terms of payment
instrument rankings, though, the two retail venues are in agreement: cash is the cheapest
instrument to accept, checks come next, and credit cards are among the most costly.
Table 8a below moves beyond discount store merchants by including costs and benefits
to key parties to the transaction. It also fixes the transaction size at the cash average.
166 The vast majority of debit cards are Visa Check cards, a signature debit card. The Nilson Report, No. 784
(March 2003); and The Nilson Report, No. 785 (April 2003).
Following the same pattern as grocery stores, the relative ranking for discount store
payment processing costs changes considerably when parties other than the merchant are
considered and when some benefits are included in the calculation. Signature debit is once
again the cheapest instrument at the cash transaction size; cash, credit cards and PIN debit
come next and are all quite close in net cost. Checks have the highest social cost.
Table 8a. Adding Selected Benefits
Discount Store Cash Transactions, $15.49 ($)
Cash Check
Credit Discover American
Express Signature
Debit PIN Debit
Float 0 0.002 0.03 0.03 0.03 0.03 0.001
Credit Option 0 0 0.08 0.08 0.08 0 0
Record Keeping 0 0.04 0.04 0.04 0.04 0.04 0.04
Cash Back Option 0 0 0 0 0 0 0.13
Signature Debit 0 0 0 0 0 0.13 0
Reward Cards 0 0 0.02 0.01 0.002 0 0
Privacy 0.16 0 0 0 0 0 0
Consumer Marginal Benefit 0.16 0.04 0.16 0.15 0.14 0.19 0.16
Central Bank
Processing Revenue 0.002 0.03 0 0 0 0 0
Seigniorage 0.09 0 0 0 0 0 0
Central Bank Marginal Benefit 0.10 0.03 0 0 0 0 0
Comm. Bank
Processing Revenue 0.01 0.09 0.30 0.22 0.40 0.43 0.44
Sum of Marginal Benefits 0.26 0.16 0.46 0.37 0.54 0.62 0.60
Social Marginal Cost 0.82 1.00 0.75 0.75 0.75 0.71 0.80
Sum of Marginal Costs 0.83 1.12 1.05 0.97 1.15 1.14 1.23
Net Social Marginal Cost 0.57 0.96 0.60 0.60 0.61 0.52 0.64
Notes: Numbers may not add due to rounding. Net Social Marginal Cost equals the Sum of Marginal Costs
minus the Sum of Marginal Benefits, which eliminates cross-party transfers. Social Marginal Cost is
included for reference.
Source: Authors’ calculations as detailed in the Appendix.
At the typical discount store check transaction size, payment cards are considerably less
costly than paper currency, following the same pattern seen for grocery stores.
C. The Rise of Debit Cards at Electronics Specialty Stores
In our third and final case study, we consider electronics specialty stores. As with
grocery and discount stores, consumers shopping at electronics stores have shifted their
Table 8b. Adding Selected Benefits
Discount Store Check Transactions, $43.93 ($)
Cash Check
Credit Discover American
Express Signature
Debit PIN Debit
Float 0 0.01 0.08 0.08 0.08 0.08 0.003
Credit Option 0 0 0.22 0.22 0.22 0 0
Record Keeping 0 0.04 0.04 0.04 0.04 0.04 0.04
Cash Back Option 0 0 0 0 0 0 0.13
Signature Debit 0 0 0 0 0 0.13 0
Reward Cards 0 0 0.04 0.02 0.005 0 0
Privacy 0.44 0 0 0 0 0 0
Consumer Marginal Benefit 0.44 0.04 0.38 0.35 0.34 0.24 0.16
Central Bank
Processing Revenue 0.002 0.03 0 0 0 0 0
Seigniorage 0.26 0 0 0 0 0 0
Central Bank Marginal Benefit 0.27 0.03 0 0 0 0 0
Commercial Bank
Processing Revenue 0.01 0.09 0.85 0.63 1.13 0.77 0.44
Sum of Marginal Benefits 0.71 0.16 1.23 0.98 1.47 1.01 0.60
Social Marginal Cost 2.25 1.10 0.90 0.90 0.89 0.77 0.85
Sum of Marginal Costs 2.26 1.22 1.75 1.53 2.02 1.54 1.29
Net Social Marginal Cost 1.55 1.06 0.53 0.55 0.55 0.53 0.69
Notes: Numbers may not add due to rounding. Net Social Marginal Cost equals the Sum of Marginal Costs
minus the Sum of Marginal Benefits, which eliminates cross-party transfers. Social Marginal Cost is
included for reference.
Source: Authors’ calculations.
payment behavior over time. Figure 4 presents the share of sales volume from 1994 through
2001 at electronics stores. Credit has a far higher share of electronics stores’ sales volume than
of either grocery or discount store sales volume; it is roughly 60 percent over the period. Debit
cards rise from near zero to 10 percent and checks decline to less than 20 percent. Cash is
generally flat over the period at 10 percent. As was suggested by Figure 1, paper instruments
play less of a role in payments for relatively high priced items such as electronics. Clearly,
consumers find that cash is either more costly to use for more expensive purchases or provides
fewer benefits as compared to other payment mechanisms, such as credit cards.
As a contrast to grocery and discount stores, the estimates presented in the electronics
store case study illustrate the effect that retail venue can have on payment instrument cost and
benefit analysis. First, average transaction sizes are far higher in electronics stores (as Table 9
below illustrates). Second, the point of sale structure is often quite different. The typical
electronics store does not have multiple checkout lanes positioned near the store exit, although
the large chainstores like Best Buy and Circuit City do. Moreover, most electronics stores do
not accept PIN debit, as they require separate card readers.167 Thus, we would expect
167 Even Best Buy, however, processes PIN debit transactions in only a handful of its stores.
Figure 4. Share of Sales Volume by Payment Method
Electronics Specialty Stores, 1994–2001
1994 1995 1996 1997 1998 1999 2000 2001
Share of Sales Volume (%)
Credit Card
Signature Debit
Source: Visa U.S.A., 1994-2001.
electronics store payment processing costs to be very different from discount and grocery store
As with discount stores, we start with the merchant costs reported by Coopers &
Lybrand for one “representative” electronics specialty merchant. Also as before, the methods
used to calculate non-merchant costs are identical to the discount and grocer cases. We again
present only the initial merchant cost table and the final net social marginal cost table, which
incorporates benefits.168 Table 9 presents electronics merchant costs, as reported by Coopers &
Lybrand, updated as with the other case studies.
Many cost elements are dramatically different than those for grocery store merchants,169
some higher and some lower. POS time costs are well below those for grocery stores—for
example, the time cost for checks are lower by a factor of eight. Drawer/office costs are
considerably higher for cash, but nearly identical for checks. Due to back office marginal costs,
payment cards have non-zero drawer costs here. Bank charges are significantly higher as well,
largely driven by the high average transaction sizes. Grocers reported an average cash purchase
of around $11.50, but cash purchases at electronics stores are six times as high. Unlike grocery
stores where checks were used for the largest average transactions, credit and charge cards have
the highest averages here at roughly $150. In contrast to discount store costs, cash is the only
payment instrument with an overall cost similar to grocery stores. Checks are around three
times as costly in electronics stores; signature debit cards are over twice as costly. That said,
cash is still the cheapest instrument for electronics stores to accept and checks are second.
168 The original 1995 costs reported by Coopers as well as the intermediate tables we calculate are all presented in
the Appendix.
169 We could easily compare electronics store costs to discount stores as well, but we would gain few additional
Net social marginal costs for electronics store transactions are presented below in
Tables 10a and 10b. Electronics store merchants may have a benefit that is not applicable to the
other case studies. Before credit cards, merchants selling relatively high-priced items often
provided in-store installment credit plans. Appliances in particular were often bought on a time
plan.170 By accepting credit cards, electronics stores do not have to bear the expense of
providing in-store credit. We do not include the marginal elements of this benefit in our
calculations, as we do not have a reasonable means of estimating it. If included, it would lower
the relative cost of credit cards.
170 See, e.g., Martha L. Olney, “Avoiding Default: The Role of Credit in the Consumption Collapse of 1930,”
Quarterly Journal of Economics, pp. 319-335, Feb. 1999, at 323; and Lendol Calder, Financing the American
Dream: A Cultural History of Consumer Credit (Princeton University Press, 1999), at 199-201.
Table 9. Per Transaction Processing Costs for Various Payment Instruments
Electronics Store Merchants ($)
Cash Check
Credit Discover American
Express Signature
POS Time 0.06 0.15 0.12 0.11 0.11 0.12
Drawer/Office 0.05 0.03 0.06 0.06 0.06 0.06
Bank Charges 0.05 1.38 2.94 1.71 3.44 1.86
Losses 0.08 0.07 0.03 0.04 0.04 0.03
Float 0.005 0.02 0.02
0.02 0.02 0.01
Other Direct Costs 0.07 0.02 0 0 0 0
Per Transaction Total
Cost 0.32 1.67 3.16
1.95 3.67 2.08
Average Receipt for
Payment Type 64.98 124.66 150.39 149.87 148.15 99.30
Cost Scaled to $100 of
Sales 0.49 1.34 2.10 1.30 2.48 2.09
Notes: Numbers may not add due to rounding.
* Signature debit is imputed as described in the text and the Appendix.
Source: Coopers (1995), updated by authors as detailed in the Appendix.
Despite the differences in many of the individual cost and benefit elements, Table 10a
presents a relative net social marginal cost comparison qualitatively similar to the other two
case studies. Payment cards are about half as costly as cash at the typical electronics store cash
purchase size. One surprising result is the relative standing of checks. Unlike grocery or
discount store transactions, here checks emerge as cheaper than either cash or cards. That
checks are cheaper than payment cards is a reversal in standing compared to the other two case
Table 10a. Adding Selected Benefits
Electronics Store Cash Transactions, $64.98 ($)
Cash Check
Credit Discover American
Express Signature
Float 0 0.01 0.12 0.12 0.12 0.01
Credit Option 0 0 0.33 0.33 0.33 0
Record Keeping 0 0.04 0.04 0.04 0.04 0.04
Cash Back Option 0 0 0 0 0 0
Signature Debit 0 0 0 0 0 0.13
Reward Cards 0 0 0.06 0.02 0.01 0
Privacy 0.65 0 0 0 0 0
Consumer Marginal Benefit: 0.65 0.04 0.54 0.50 0.48 0.17
Central Bank
Processing Revenue 0.002 0.03 0 0 0 0
Seigniorage 0.39 0 0 0 0 0
Central Bank Marginal Benefit 0.39 0.03 0 0 0 0
Commercial Bank
Processing Revenue 0.05 1.38 1.27 0.74 1.51 1.31
Sum of Marginal Benefits 1.10 1.45 1.81 1.24 1.99 1.48
Social Marginal Cost 2.78 0.69 1.30 1.29 1.27 1.11
Sum of Marginal Costs 2.84 2.09 2.57 2.04 2.78 2.42
Net Social Marginal Cost 1.74 0.64 0.76 0.79 0.79 0.95
Notes: Numbers may not add due to rounding. Net Social Marginal Cost equals the Sum of Marginal Costs
minus the Sum of Marginal Benefits, which eliminates cross-party transfers. Social Marginal Cost is
included for reference.
Source: Authors’ calculations as detailed in the Appendix.
studies. It is driven by the high bank charge assessed merchants for check deposits as compared
to the bank processing cost, charges that are netted out in the net social marginal cost
calculations. It is unlikely that commercial banks are truly profiting more from checks used at
electronics stores. Rather, it could be that banks charge a higher fee to electronics stores
because the higher purchase amounts imply a greater fraud or non-sufficient fund risk for the
bank, costs not captured in our processing cost estimates. Again, our numbers are not meant to
be absolute, but only relative. Table 10b, below, for the average check purchase size reveals a
similar pattern, although check and credit card net social marginal costs are roughly identical.
Thus, for all three case studies, once other parties to the transaction are considered and
some of the benefits for various payment instruments are counted, cash no longer appears as
attractive as was the case when merchants alone were considered. Our calculations do not
reveal the precise marginal costs and benefits for each payment instrument, but they do suggest
that economic welfare has increased as consumers have increased their use of payment cards at
grocery stores, discount stores, and electronics stores.
Table 10b. Adding Selected Benefits
Electronics Store Check Transactions, $124.66 ($)
Cash Check
Credit Discover American
Express Signature
Float 0 0.02 0.22 0.22 0.22 0.02
Credit Option 0 0 0.62 0.62 0.62 0
Record Keeping 0 0.04 0.04 0.04 0.04 0.04
Cash Back Option 0 0 0 0 0 0
Signature Debit 0 0 0 0 0 0.13
Reward Cards 0 0 0.12 0.04 0.01 0
Privacy 1.25 0 0 0 0 0
Consumer Marginal Benefit: 1.25 0.05 1.00 0.93 0.90 0.18
Central Bank
Processing Revenue 0.002 0.03 0 0 0 0
Seigniorage 0.75 0 0 0 0 0
Central Bank Marginal Benefit 0.75 0.03 0 0 0 0
Commercial Bank
Processing Revenue 0.05 1.38 2.43 1.43 2.89 2.26
Sum of Marginal Benefits 2.05 1.46 3.44 2.35 3.79 2.44
Social Marginal Cost 3.55 0.71 1.62 1.61 1.57 1.24
Sum of Marginal Costs 3.61 2.12 4.05 3.03 4.47 3.50
Net Social Marginal Cost 1.56 0.66 0.61 0.68 0.68 1.06
Notes: Numbers may not add due to rounding. Net Social Marginal Cost equals the Sum of Marginal Costs
minus the Sum of Marginal Benefits, which eliminates cross-party transfers. Social Marginal Cost is
included for reference.
Source: Authors’ calculations as detailed in the Appendix.
D. Sensitivity Analysis
In this section, we consider the variation and the uncertainty in the estimates by
presenting alternative values for key parameters. Even merchants within a single industry can
face different payment processing costs due to size, sophistication, and a host of other factors.
On the consumer side there is considerable heterogeneity in preferences and perceived costs. In
addition to this kind of inherent variation, there is also uncertainty in several of the cost
estimates. While many of the individual cost and benefit components have wide ranges, our
conclusions above appear to be robust to changes in the specific cost and benefit elements.
Given that we do not have information on the joint distribution of costs and benefits, we
focus here on those individual elements that show significant variation or that rely on uncertain
assumptions. We present alternative calculations for grocery store transactions only. The Food
Marketing Institute reports a range of costs (high, low, and average) that we use as the basis for
our analysis. Nonetheless, the sensitivity discussion below likely applies to all three retail
We consider four categories of alternatives. First, we examine the sensitivity of our
calculations to variations and assumptions regarding the opportunity cost of time––an element
that affects three cost components. Second, we briefly discuss the natural variation in the
processing fees that commercial banks charge merchants. Third, we explore the role that
consumer heterogeneity plays in determining which payment instruments are used. Finally, we
investigate the affect of altering the conditional nature of the transaction––the assumption that
it takes place in a given venue for a given dollar amount.
1. Opportunity Cost of Time
As Table 2 shows, time spent at the point of sale counter differs considerably across
payment instruments.171 Our analysis could be sensitive to the use of average POS times, to our
assumptions about point of sale queues, and to our use of average hourly wages in computing
the cost of time. We explore how net social marginal costs vary with changes in these three
To investigate the first effect, we recalculate the net social marginal cost estimates
presented in Tables 5a and 5b using FMI high and low point of sale times. The results are
presented in Figure 5 below. Since point of sale time does not vary by transaction size, we
present the cash transaction calculations only. As can be seen from the chart, time at the
checkout counter does affect the overall cost estimates, but even if we were to compare the
highest times reported for payment cards with the lowest times for paper instruments, we would
still conclude that payment cards generally are not more costly than paper currency.
To test the second component of the opportunity cost of time, we alter our assumption
about the number of people waiting in line behind a payer. Recall that in the calculations above,
we assume one person waits in line for each grocery transaction, imposing an externality cost.
While we were unable to find data on this issue, anecdotal evidence indicates that different
kinds of stores are likely to have differences in the average number of people in line. We expect
that queue size varies by store type. For example, discount stores are likely to have longer lines
on average than electronics specialty stores. Supermarkets are likely to have longer lines than
smaller, typically higher priced specialty grocers or local mom-and-pop grocers. Given the
uncertainty in line length, we explore the sensitivity of our results to our queue assumption.
171 POS time varied considerably across the three case studies as well.
Figure 5. Net Social Marginal Cost Estimates, Varying POS Time
Grocery Store Cash Transaction ($11.52)
Cash Non-
Check Credit Signature
Debit PIN Debit
Payment Instrum ent
Marginal Cost
Source: Authors’ calculations as detailed in the Appendix, based on FMI reported high, mean, and low
POS time costs.
Figure 6 portrays a range of assumptions on queue size for grocery store cash size
transactions. Queue assumptions clearly affect the net social marginal cost, especially for a
payment instrument that takes more time to process at the point of sale, such as checks. With
no one in line, credit cards are somewhat more costly than checks, but still less expensive than
cash. Signature debit remains the cheapest of the payment instruments. With three people in
line, all payment cards are less costly than checks, but cash is cheaper than either credit/charge
or debit. While taking the highest queue length for payment cards and the lowest queue for
paper instruments would change the standings considerably there is little reason to think that
the number of people standing in line depends on the payment instrument used by one payer.
Grocery stores do sometimes have cash only lanes––but these may have longer lines than the
full service lanes as they typically restrict the number of items purchased as well.172 Moreover,
cash-only lanes are evidently becoming more rare, as grocers move to self-checkout lanes
172 Remember that our queue comparison only counts the time spent in processing payments, not the overall time
in line, which would also include the time for the cashier to scan the grocery items.
173 By 2002, about 20 percent of supermarket chains had self-checkout lanes. Peter Lucas, “Self Service at the
Point of Sale,” Credit Card Management, March 2002. A recent survey conducted by the FMI of 40 major
supermarket companies owning more than 6,000 stores found that over half of them are either planning or
already have installed self-checkout centers. At self-checkout stations, about 50 percent of the transactions are
with cash, 40 percent are with plastic cards, and about 10 percent are with checks. David Gosnell, “Going
Automatic with Cash-Back,” Credit Card Management, August 2003.
Figure 6. The Impact of Queue Size on Marginal Cost Estimates
Grocery Cash Transaction ($11.52)
Cash Non-
Check Credit Signature
Debit PIN Debit
Payment Instrument
Marginal Cost ($)
no people
1 person
2 people
3 people
Source: Authors’ calculations as detailed in the Appendix.
As a final test on the sensitivity of our results to opportunity costs, we calculate the
combined consumer time elements (processing time plus queue time) for different consumer
wage levels. Figure 7 reports net social marginal costs for grocery cash transactions, measuring
the cost of time at the tenth wage percentile for U.S. workers, the average wage, and the
ninetieth wage percentile. The average wage repeats the cost of time assumed in the case
Ideally, we would want to examine payment instrument use by income percentile, but
this information is unavailable. Even with our simple approach though, we can interpret Figure
7 as indicating that consumers with higher wages face a higher level of cost for those
instruments with relatively longer point of sale times, such as non-verified checks. While all net
social marginal costs rise with wages, those with longer processing times do so more
2. Bank Charges
Referring back to Table 2, bank charges are another obvious candidate for sensitivity
analysis.174 The merchant discount assessed on payment card purchases plays a pivotal role in
174 There is little variation in deposit preparation costs. While other direct costs do vary by payment instrument,
much of this variation is driven by: 1. FMI’s exclusion of cash theft, counterfeiting and float costs, and 2. the
catchall nature of the category (for example, armored car costs are not applicable to payment cards). For these
reasons, we do not consider variation in these two remaining categories.
Figure 7. Net Social Marginal Cost, Varying Payer Wage,
Grocery Cash Transaction ($11.52)
Cash Non-
Check Credit Signature
Debit PIN Debit
Payment Instrument
Marginal Cost ($)
Source: Authors’ calculations as detailed in the Appendix.
the expense of these cards for retailers. Further, merchant discounts exhibit considerable natural
variation depending on the store type and size as well as on the payment card system.
Table 11 below presents FMI’s reported high, mean, and low bank charges. The
marginal cost tables above are based on the mean charges. One factor behind payment card
bank fee variation is FMI’s decision to report all card systems together. Processing charges
vary substantially across systems, with American Express generally regarded as the most
expensive and Discover generally considered the least.175 A different proportion of sales
attributed to each card system will result in different average bank costs for different
merchants. The table illustrates that bank charges vary considerably among grocery stores for
payment cards as well as verified checks.
Table 11. Bank Fee Variation, Grocery Store Transactions (1998 $)
(Per $1000) Non-Verified
Check Verified
Check Credit/
Charge Signature
Debit PIN Debit
Highest 0.01 0.07 0.37 2.01 0.82 0.24
Average 0.003 0.05 0.13 0.87 0.41 0.19
Lowest 0.001 0.04 0.06 0.42 0.27 0.15
Notes: Fees reported are for instrument-specific average transaction sizes and are therefore not comparable
across payment instruments.
Source: FMI (1998).
The bank charges reported in Table 11 highlight the diversity among grocery store
costs. Even within one narrow retail segment, bank processing fees vary considerably.
Variation of this sort is a key driver behind merchant acceptance decisions. Keep in mind,
though, that the charges in Table 11 are for 1998––as we discussed above, charge card and
signature debit fees have fallen since then, while PIN fees have risen (as reflected in our
updated case study figures). Moreover, since bank charges represent transfer payments from
merchants to commercial banks (or to independent card systems such as American Express)––
not resource costs––the variation in fees does not translate necessarily into variation in net
social marginal cost. Ideally, we would use similar data to specify a range of commercial bank
175 Leslie Beyer, “The Interchange Stranglehold (of Visa and MasterCard),” Grocery Headquarters, October 1,
processing costs, which are actual resource costs and would thus affect our net social marginal
cost estimates, but this information is unavailable.
3. Accounting for Consumer Differences
Several of the benefits included in our net social marginal cost calculations are likely to
exhibit substantial variation across individuals. For instance, consumers most concerned about
a payment mechanism that helps them to spend within a budget are not likely to view the
option value of credit cards as a benefit. In point of sale transactions in particular, consumers
choose the payment instrument, although merchants can limit and/or influence their choices.176
Thus, consumer preferences can dictate to a large degree which instruments are used more
In this section, we construct several scenarios to demonstrate how cost-benefit
calculations can change when heterogeneity among consumers is introduced. Following
consumer research, we develop profiles that stress different aspects embodied in the payment
instruments considered here.177 We then calculate net consumer costs (consumer costs minus
consumer benefits), illustrating how differences in perceived costs and benefits can drive
payment choices. We consider five consumer profiles, described in more detail below, each
driven by one key benefit: payment control, budget enforcement, low cost, security, or
In the control profile, consumers’ primary concern is the ability to control the timing of
a payment.178 That is, these consumers want control over when they actually have to pay the
charge, which is different from when they pay the merchant for every payment instrument
except for cash. Control also covers the ability to stop payments and to dispute charges if
problems arise after a purchase is made. Checks, credit and charge cards all meet these criteria.
While signature debit cards do provide dispute resolution benefits, automatic payment
176 Recall that consumers initiate more than 90 percent of all transactions. Mantel, supra note 9, at 8.
177 Mantel reports the results of focus group interviews. Based on those interviews, he discusses a number of
customer preferences, noting that “Some consumers might value more than one preference, but it appeared that
most consumers were primarily driven by just one or two preferences across the different payments they were
making.” Mantel, supra note 9, at 15.
178 Mantel, supra note 9, at 16.