Technology Adoption and Consumer Payments: Evidence from Survey Data

Fumiko Hayashi and Elizabeth Klee1

8 November 2002
WP02-01

Preliminary: Do not cite or quote without authors permission.

Abstract
Consumers pay for hundreds of goods and services each year, but across
households and across goods, consumers do not choose to pay the same way. This paper
posits that these differences depend in part on consumers propensity to adopt new
technologies, and depend in part on the nature of the transaction. In order to test these
hypotheses, this paper offers comparisons of payment instrument use at the point of sale
and for bill payment from a sample of consumers surveyed in 2001, drawn primarily
from users of the Internet. The results indicate that consumers who use technology or
computers are more likely to use electronic forms of payment. In addition, use of direct
deposit is a significant predictor of use of electronic payments. Furthermore, payment
choice depends on the characteristics of the transaction. By analyzing these hypotheses
together, consumer payment choice may lend insight into consumer technology adoption
behavior more generally.

Hayashi: Payments System Research, Federal Reserve Bank of Kansas City, 925 Grand Boulevard,
Kansas City, MO 64198. Tel: (816) 881-6851, Fax: (816) 881-2425, Email: [email protected].
Klee: Mail Stop 188, Division of Reserve Bank Operations and Payment Systems, Board of Governors of
the Federal Reserve System, Washington, DC 20551. Tel: (202) 721-4501, Fax: (202) 872-7533, Email:
[email protected]. This paper reflects the views of the authors and not necessarily those of the
Federal Reserve Bank of Kansas City or the Board of Governors of the Federal Reserve System.

Technology Adoption and Consumer Payments: Evidence from Survey Data

8 November 2002
Preliminary: Do not cite or quote without authors permission.
Abstract
Consumers pay for hundreds of goods and services each year, but across
households and across goods, consumers do not choose to pay the same way. This paper
posits that these differences depend in part on consumers propensity to adopt new
technologies, and depend in part on the nature of the transaction. In order to test these
hypotheses, this paper offers comparisons of payment instrument use at the point of sale
and for bill payment from a sample of consumers surveyed in 2001, drawn primarily
from users of the Internet. The results indicate that consumers who use technology or
computers are more likely to use electronic forms of payment. In addition, use of direct
deposit is a significant predictor of use of electronic payments. Furthermore, payment
choice depends on the characteristics of the transaction. By analyzing these hypotheses
together, consumer payment choice may lend insight into consumer technology adoption
behavior more generally.

-1-

1. Introduction
Consumers pay for goods and services every day, yet consumers do not always
choose to pay the same way. At the point of sale, U.S. consumers have a choice of four
payment instruments: cash, check, credit card and debit card. To pay for bills, U.S.
consumers generally have a choice of two payment instruments: check and the automated
clearing house (ACH), the name for direct electronic credit transfers or direct electronic
debit transfers from a bank account. In 2000, checks represented the largest share of the
volume of retail noncash payments (59.5%), followed by credit cards (28.9%), debit
cards (11.6%) and retail ACH payments (5.6%). Recent estimates of the purpose of
consumer check transactions indicate that 29% of consumer check payments are at the
point of sale, 36% are for bill payments, and 13% may be either point of sale payments or
bill payments.2 These statistics show that more than one payment instrument is used in
equilibrium, and the same payment instrument may be used for more than one purpose.
Hence differences in payment choice.
There are a few competing hypotheses to explain differences in payment choice.
Some posit that differences in payment choice stem from the supply of payment
instruments. The choices available to consumers may depend on historical accident and
the evolution of the banking system.3 Other theories regarding payment choice focus on
demand factors. Differences in payment choice by consumers may depend on the
characteristics of the consumer, 4 the characteristics of the goods bought, 5 or the
characteristics of the payment instrument itself.6 In general, these studies, ranging from
the 1980s to today, cover a large part of the time period when electronic payments
became more widespread. Each provides evidence that consumer payment choice is
2

Gerdes and Walton (2002).

For example, the check system developed in the late 1800s and early 1900s is evidence of a paper based
system that was widely adopted due in part to the spread of the railroads, mandatory par collection of
checks by banks that were members of the Federal Reserve system, and the prohibition of interstate bank
branching. See Baxter (1983) and Spahr (1926). In addition, Humphrey, Pulley and Vesala (1996) offer an
international perspective on this hypothesis.
4
Daniels and Murphy (1994), Mantel (2000) and Stavins (2001).
5
Shy and Tarkka (2002) and Klee, Guerin and Marquardt (2002). Schreft (2002) examines these questions
in the context of payments made on the Internet.
3

-2-

dependent on demographics and income. Previous studies consistently indicate that

consumers who are younger, wealthier, and technologically proficient more often adopt
electronic forms of payment.
This paper focuses on the demand factors of payment choice. It makes two
important contributions. As a baseline, the results confirm those from earlier studies,
using a new set of consumers. We find that younger, wealthier consumers are more
likely to use electronic forms of payment for point of sale transactions and for bill
payment. We also address whether adopting new technologies indicates more frequent
use of electronic forms of payment. The results show that technology adoption is
significant in predicting the probability of using an electronic form of payment. Finally,
we find that the characteristics of a transaction significantly affect the payment method
chosen.
Consumer payment behavior has direct policy implications. Payment systems are
network goods.7 Characteristics of payment systems that are classic attributes of network
goods include complementarities and economies of scale for supply, and consumption
externalities, switching costs and lock-in for demand.8 Identifying the factors that
contribute to adopting electronic forms of payment, which tend to be less resource
intensive than paper-based forms of payment, is important for society.9 In addition,
household payment behavior has broader economic implications for example, how
people choose to pay affects checking account balances, cash holdings, which in turn has
the potential to affect monetary aggregates.10
Thus understanding consumer payment behavior is important to achieve the above
goals. Section 2 presents the model. Section 3 describes the data and presents summary
statistics. Section 4 discusses the model, presents the model estimates and discusses the

Another strand of payments literature uses a general equilibrium monetary economics framework to study
payments. For an overview, see Green (2002).
7
McAndrews (1997).
8
Shy (2001), p. 187-213.
9
For example, although significant economies of scale can be achieved in check processing, some costs of
check processing are difficult to reduce, including physically transporting checks from one place to
another. In contrast, once an electronic payments network is built, barring capacity constraints for
computing, the marginal cost of an additional payment is negligible.
10
See Board of Governors of the Federal Reserve System (1998), Ferguson (2001), and Avery et al. (1986).

-3-

implications of these estimates. Section 5 presents conclusions and avenues for further
research.
2. Theoretical Framework
The analysis focuses on consumer characteristics, goods and transaction
characteristics, and payment instrument characteristics to explain payment choice.
Information on technology adoption behavior acts as a proxy for unobserved
characteristics that may affect payment choices. The structure of the data set allows us to
test how the nature of the transaction determines payment instrument choice. We discuss
each of these factors in turn.
To start, most of the literature on consumer payment choice focuses on consumer
characteristics. Previous research indicates that certain demographic groups are more
likely to adopt electronic forms of payment. In studies that address payment choice at the
point of sale where the goods bought with the payment instrument are unspecified,
wealthier and more educated individuals are more likely to use electronic forms of
payment.11 These results are consistent with those from studies that consider only one
good. For example, when making gasoline purchases, more educated, higher income,
and individuals who are not middle-aged are more likely to use electronic forms of
payment.12 Evidence exists that electronic bill payers are more educated, have higher
incomes, and younger than individuals who pay by other means.
Likewise, evidence exists that adoption of certain payment types relies on
availability, which may be a direct function of location or the consumers banking
relationship. For example, especially in their infancy, credit cards were used primarily by
a higher income population. Throughout the 1980s and 1990s, more consumers gained
access to credit cards, and thus they became a payment choice for a larger segment of
society.13

11

Kennickell and Kwast (1997), Mantel (2000) and Stavins (2001).

Carow and Staten (1999).
13
Black and Morgan (1999).
12

-4-

Previous research concludes that the propensity to adopt one new technology may
be correlated with the propensity to adopt another new technology.14 This paper presents
a test of this hypothesis in two ways. First, the survey instrument asks respondents
whether they use certain devices a cell phone, a computer, or the Internet. Second, the
sample construction of the data set provides a structure to test whether consumers who
choose to respond via the web behave differently than consumers who choose to respond
via paper. With these two features, we will be able to test whether technology adoption is
correlated with use of electronic payment instruments.
It may be the case that the nature of the goods or of the transaction affects the
payment instrument used, all other things being equal. For example, consumers may be
less likely to buy groceries on credit than other items, as groceries are generally
considered to be necessities. Or, for bills, it may be the case that bills that arrive at
regular intervals and are for the same amount of money each month for example,
mortgage or rent payments are more likely to be paid directly than bills that may vary.
Including the characteristics of the goods or the transaction in the estimation is therefore
important, as these characteristics affect the utility of a particular payment instrument.
Note that the goods and transaction characteristics are correlated, but ultimately
independent of total expenditure. The amount of money a consumer spends in a trip to
the grocery store may be equal to the amount of money that a consumer spends in a trip
to a discount store; however, payment behavior may differ. Similarly, the payment due
on a monthly credit card bill may be similar to the amount paid in rent or for a mortgage,
but the nature of the bill dictates how the consumer chooses to pay. Thus, it makes sense
to connect the goods bought with the payment chosen, and not rely completely on total
expenditure.15
Finally, payment instrument characteristics may affect payment choice. Paying a
mortgage payment in cash may not be desirable to the consumer, as this would require a
large wad of bills and potential loss or theft. Paying for an ice cream cone with a
14

Huffman and Mercier (1991) study the determinants of farmers adoption of two types of computer
systems. They find that characteristics predicting the adoption of one technology also predict the adoption
of another technology.
15
Other models that examine payment choice are based solely in total expenditure. However, many of
these models answer different policy questions than those posed here. See Whitesell (1992) and Tarkka
and Shy (2002).

-5-

prearranged ACH payment may be similarly inconvenient. Although these examples are
extreme, they point out the potential for the characteristics of the payment instrument to
influence payment choice.
The data analysis that follows will illustrate how consumer characteristics, goods
and transaction characteristics, and payment instrument characteristics interact for
consumer payment choice. But first, we specify our model.
Consumer i purchases good j with payment instrument k, the combination of
which maximizes the consumers utility, Uijk. This utility depends on the characteristics
of the consumer, xi, the characteristics of the goods bought and the transaction, yj, and the
characteristics of the payment instrument itself, zk. The dependence of utility on all of
these simultaneously incorporates all factors that affect payment choice.
The characteristics of the consumer can be divided further into two parts:
demographic characteristics, di, and technology adoption characteristics, ti. Testing to
see whether one or both sets of consumer characteristics affect payment choice is a
critical part of our analysis.
Utility is linear in the characteristics of the consumer, of the goods bought, and
the characteristics of the payment instrument, with weights , and , respectively. The
consumer chooses goods and payment instruments to maximize utility subject to a budget
constraint, which is a function of the prices of the goods obtained and the prices of the
different payment instruments.
Not all characteristics of the payment instruments, goods and transaction, or
consumers can be observed. These unobserved characteristics are incorporated into an
error term ijk, which is assumed to follow a generalized extreme value distribution.16
This implies that the utility can be written as
U = U ijk (x i , y j , z k ) = + x' i + y ' j + z ' k + ijk ,
where indicates the intercept.

16

Payment choice is a classic discrete choice problem: the consumer has a limited set of options from
which to choose, the options are mutually exclusive, and the set of options is the only set available. See
Train (1986).

-6-

In addition, we only observe whether the consumer uses the payment instrument.
Thus let the variable u*ijk indicate whether consumer i purchases good j with payment
instrument k, where

u * ijk = 1 if u ijk > u ijk ' , k ' k ,

= 0 otherwise.
With our assumption on the distribution of the error term, we can form the
probability Pijk that consumer i purchases good j with payment instrument k as
Vijk

Pijk =

Vijl

lK ij

where Vijk represents the observed portion of the representative utility and Kij is the set of
payment instruments available to consumer i to purchase good j.
However, because of the construction of the data set, we do not observe whether a
particular payment instrument and good combination is the most preferred option.
Rather, we only observe whether the particular payment instrument and good
combination is one of the portfolio of options chosen for a similar situation. This must
necessarily imply that any particular payment instrument and good combination is the
most preferred option in some situation, and the equation above still holds. For all of our
estimations, then, the parameters should be interpreted as reflecting whether a particular
payment instrument and good combination is chosen relative to all others in some
situation, knowing that unobserved factors influence when this particular instrument is
chosen relative to another instrument.
This data construction also determines our estimation procedures. In order to try
to capture all of the factors that may affect payment choice, we report results from two
separate estimation procedures. The first examines how consumer characteristics affect
payment choice. We estimate binomial logits for each payment choice, across all types
of establishments. The second examines how the attributes of the purchase affect
payment choice. We estimate the probability that a consumer uses a particular payment
instrument, given the consumers portfolio of payment instruments.
The computational procedure is to use maximum likelihood techniques to obtain
estimates of the parameter vectors , , and .

-7-

3. Data and Summary Statistics

The primary analysis is performed on a sample of U.S. consumers surveyed in

2001 on behalf of the American Bankers Association, drawn primarily from users of the
Internet.17 The American Bankers Association (ABA) contracted with Dove Consulting,
Inc. to conduct a survey of U.S. households on payment instrument use. The survey
addressed in-store payment preferences, bill payment preferences and Internet payment
preferences specifically, as well as payment use more generally.
The data are well suited for testing hypotheses regarding technology adoption and
payments. First, respondents were asked questions regarding technology use. The data
contain information on consumer use and adoption of three new technology devices:
computers, the Internet and cell phones. Of course, due to the sampling structure of the
survey, all of the respondents who answered the survey on the web necessarily use a
computer and the Internet. However, some paper-based respondents may use the
Internet, and in addition, cell phone use is independent of the sampling structure.
Second, survey respondents completed survey forms either on paper or on the
web. Comparing and contrasting the behaviors of both populations lends insight into
overall payment adoption behavior.
Third, respondents were asked questions that addressed both payments at the
point of sale and payments made for bills. Thus, the data contain information on whether
consumers use certain payment instruments for certain types of payments. The store
types included in the survey were hardware stores, gas station/convenience stores,
department stores, discount stores or warehouse clubs, drug stores, restaurants, and fast
food establishments. The bill types included in the survey were rent, loans, credit card
payments, insurance, membership, tuition and utility bills.
Table 1 gives definitions for the variables in the Dove sample. The first four
variables address bank product use: bank accounts (BKACCT), credit cards (CREDIT
CARD), debit cards (DEBIT CARD) and direct deposit (DIRECT DEPOSIT). The four
variables COMPUTER, INTERNET, PURCHINT and CELL capture the propensity for a
17

See the Appendix for a discussion of sampling and data issues.

-8-

consumer to adopt a new product. The next sets of categories income, education,
gender and age capture demographic effects that may be significant determinants of
payment choice.18 The final set of regional variables capture supply side effects that may
be present at the census division level, as different regions of the country may have
different availability of payment instruments.19
Table 2 presents summary statistics for our sample, plus a comparison to the
samples of two national surveys of consumers. The first comparison survey is the Survey
of Consumer Finances (SCF).20 The SCF surveys a cross-section of U.S. households. It
is conducted triennially by the Federal Reserve in conjunction with the National Opinion
Research Center at the University of Chicago (NORC). The second comparison survey is
the Current Population Survey (CPS) Computer and Internet Use Supplement of August
2000, which outlines computer use by different segments of the population.21
On certain dimensions, the Dove data differ from national averages; on others, the
data are quite similar. Turning first to bank product holdings, the Dove sample has a
higher percentage of bank account holders, credit card holders and debit card holders than
the SCF.22,23 However, the Dove sample has the same proportion of direct deposit users
as the SCF. One potential explanation for this similarity is that in many cases, the
employer, and not the employee, decides whether to use direct deposit.
Turning now to new product adoption behavior, the data show that the Dove
Survey population is much more likely to have used a computer, used the Internet, or
18

The education variable is defined as the highest level of education attained.

Admittedly, this is an imperfect way to capture supply-side effects. A full structural approach would
require data on the availability of payment instruments and the extent to which payments are widespread in
the area.
20
For details on the structure of the SCF, see Kennickell (1998, 2000), and Kennickell, Starr-McCluer and
Surette (2000).
21
For details on the structure of the CPS and the supplements for the CPS, see U.S. Department of Labor,
Bureau of Labor Statistics (2000), Chapters 3 and 11. Also, our estimates are on a subsample of the
Supplement sample. From an original sample of 134,986 observations, we eliminated respondents under
19 years old or with zero income, to arrive at a final sample of 62,938 observations.
22
In both the SCF and the Dove samples, credit card summary statistics include respondents who were not
approved for a credit card. There are 53 respondents in the Dove POS sample and 55 respondents in the
Dove bill payment sample for whom this is the case. Similarly, the debit card summary statistics include
respondents who do not have a debit card because they do not have a bank account. There are 40
respondents in the Dove POS sample and 39 respondents in the Dove bill payment sample for whom this is
the case.
23
The SCF and Dove questions on debit cards are not exactly the same. The SCF asks whether respondents
use a debit card, while the Dove survey asks whether respondents have a debit card. This should be noted
when interpreting the summary statistics and the estimation results reported later in the paper.
19

-9-

made a purchase on the Internet than the population sampled for the CPS supplement.
The effect of this difference on payment choice is a critical part of our analysis.
Demographically, the Dove sample has a higher percentage of women, a higher
percentage of middle-aged people, and a higher educated sample than the nationally
representative data sets.24 At the same time, the income and employment distributions in
the three data sets are fairly similar.25
Table 3A presents statistics for payment choice at different establishments.
Respondents were asked whether they pay using a particular payment instrument at each
type of establishment. We report the distribution of payment preferences for respondents
who indicated preferences for each type of establishment.26 Except for department stores,
overall, cash has the highest percentages for use at each type of establishment. Almost
98% of respondents report using cash at fast food restaurants, but only about 55% of
respondents report using cash at department stores. Although this is perhaps correlated
with the dollar amount of the sale, it is also important to note that approximately 72% of
respondents report using cash at the grocery store, but only 63% report using cash at a
discount store. It is possible that the dollar amounts of transactions at these types of
establishments may be similar.
The largest percentage of respondents report using checks at discount stores.
However, the distribution of these statistics is more uniform than the cash statistics.
Credit cards are used by the largest percentage of respondents at department stores and
restaurants, which also have the highest percentage of respondents for using signature
debit transactions. Overall, however, the highest percentage of respondents report using
any type of debit at grocery stores.
Table 3B presents statistics for payment choice for different types of bills.
Respondents were asked whether they pay using a particular payment instrument for each
type of bill. We report the distribution of payment preferences for respondents who
indicated preferences for each type of establishment. Overall, checks have the highest
24

Note that comparing samples with respect to gender may be misleading. The unit of observation in the
Dove survey is a respondent, and the unit of observation in the SCF is a household. If a household in the
SCF has both a male and a female, it is coded as a male headed household. For details, see Kennickell
(2000).
25
Some of the percentages presented in the tables may not add to 100 due to rounding.
26
This implies that these percentages should be interpreted as conditional probabilities.

- 10 -

percentage for respondents paying bills. However, it is not the only type of payment used
to pay the same type of bill. For example, almost 33% of respondents report using cash
to make tuition payments, and 26.84% report using cash to make rent payments.
Although this may seem implausible, the survey was constructed so that respondents
could choose more than one payment type for each type of establishment or bill. It is
interesting to note that respondents seem to choose more than one for many types of bills.
Perhaps because these payments go through the mail or by computer, only 17% of
respondents use cash to pay a credit card bill. Credit cards and debit cards were most
frequently used to pay membership payments. Direct payments were used by the highest
percentage of respondents to pay loans and insurance, and online payments were used
most frequently to pay a credit card bill. This conforms to our hypothesis stated above as
to why consumers do not use cash to pay a credit card bill.
Table 4 reports correlation coefficients across these sets of variables. The simple
correlation coefficients seem to indicate that higher income is correlated with use of all of
the products listed: direct deposit, computers, the Internet, purchases on the Internet, and
cell phones. Education level is also positively correlated with use of these products.
Being female is negatively correlated with use of these products. Interestingly, the
correlation coefficient of direct deposit and female is significantly different from zero,
but it is not significantly different from zero for the other products.
Age plays an interesting role with regards to use of these products. None of the
correlation coefficients for younger individuals is significantly different from zero. In
contrast, except for direct deposit, all of the correlation coefficients for middle aged
individuals are significantly different from zero, and moreover, are positive. The
correlation coefficients for older individuals are positive for direct deposit potentially
due to social security payments and negatively correlated with use of computers, the
Internet, purchases on the Internet, and cell phones.
The summary statistics show that there are differences in payment choice with
respect to consumer characteristics and establishment characteristics. The correlations
show that consumers with different characteristics have different behaviors regarding
adoption of new technologies. With these summary statistics and correlations in mind,
we now turn to the estimation results.

- 11 -

4. Estimation and Results

Table 5A reports the parameter estimates from the baseline model, which
estimates the probability that a consumer uses a particular payment instrument.27 At the
point of sale, respondents in the middle of the income distribution are more likely to use a
check or a debit card than respondents at the high or low end of the income distribution.
Middle-aged and older respondents are more likely to use checks and less likely to use
debit cards at the point of sale than younger respondents. Respondents who live in New
England are less likely to use checks and more likely to use debit cards at the point of
sale, and respondents who live in the Pacific West are more likely to use debit cards at
the point of sale.
Turning to the bill payment results, all income levels are more likely to use a
check to pay for bills than lower income respondents are. Furthermore, individuals in the
middle of the income distribution are more likely to use direct payment for bills than
other individuals, and the highest income respondents are more likely to use online bill
payment than other respondents are. In addition, higher educated respondents are more
likely to use online bill payment. Middle-aged respondents are more likely to use checks,
direct payment or online payment than younger respondents, and older respondents are
more likely to use checks, and less likely to use debit cards or online payment than
younger individuals. All other things equal, respondents who live in the Mid-Atlantic
region are more likely to use checks than the baseline to pay bills, and respondents in the
South or the West are more likely to use debit cards to pay bills.
As a point of comparison, table 5B presents results from estimating similar
models using the SCF. Because the primary purpose of the SCF is to understand the
wealth holdings of households, and not payment choice, we estimate on a subsample of
households and eliminate weights for our logit estimations. We report results from
estimating on the first imputation only.28
27

In all of our estimations, the excluded income category is less than $20,000, and the excluded census
region is Midwest: East North Central.
28
We eliminated households with over $3 million in stocks, bonds and liquid assets, households with over
$375,000 in annual income, households with over $10,000 in credit card debt, and less than a high school

- 12 -

The results for using debit cards at the point of sale are similar for both data sets.
All of the income coefficients have the same signs and are significant for the same
variables. In the SCF, the coefficient on education is statistically significant, but keeps
the same sign as in the Dove survey. The age coefficients are both the same sign, and are
significant in the estimations on both data sets. There is a little more variation in the
regional variables. The coefficient on NEMA changes sign but remains insignificant, and
the coefficient on MWWNC and WM are significant.
The results for direct payment, however, differ for the two samples. All of the
income variables have statistically significant coefficients in the SCF sample, while in the
Dove sample they do not. The education variable also has a significant coefficient in the
SCF, but not in the Dove sample. MIDDLE changes sign and significance in the SCF
sample, and some of the regional variables are significant in the SCF sample, while none
is significant in the Dove sample. It is important to note these differences when trying
how consumers behave with respect to payment choice.
Table 6A reports results from the same baseline estimation as table 5A, but
includes direct deposit as an independent variable. Direct deposit is not a significant
predictor of the use of checks at the point of sale. Furthermore, no coefficients change
sign or significance from the baseline model. Interestingly, however, the results do
change for the use of debit cards at the point of sale. In contrast to check use at the point
of sale, direct deposit is a significant predictor of debit card use at the point of sale. From
the baseline results, INCOME2 and INCOME3 become insignificant (although the
coefficients remain positive) and the constant terms becomes significant (again,
remaining negative).
Turning to bill payment, we find that the coefficient on direct deposit is positive
and significant in all of the estimations. For checks, the coefficient on old becomes
insignificant (but is still positive) and the coefficient on New England becomes negative,
although in both cases, one cannot reject the hypothesis that the coefficient equals zero.
For paying a bill with a debit card, we find that the coefficient on the highest income
becomes significant, but remains negative.

education. For details on multiple imputation and estimation of models, see Montalto and Sung (1996,
1998).

- 13 -

Perhaps, however, the most interesting change is in the coefficients on income for
direct payment.29 In the baseline results, the coefficients on the income variables are
significant and positive. When direct deposit is included in the estimation, these
coefficients become insignificant, and on the INCOME2 variable, changes sign to
negative. Our correlations in table 4 suggest why this may be the case. Respondents
whose income levels are in the INCOME1 or INCOME2 ranges are less likely to use
direct deposit, and respondents whose income levels are in the INCOME4 and
INCOME5 ranges are more likely to use direct deposit.30 Moreover, we cannot reject the
hypothesis that the correlation between direct deposit and INCOME3 is equal to zero.
Because the income variables are significant in the baseline model but insignificant once
direct deposit is included suggests that direct deposit has a significant effect on a
respondents decision to use direct payment.
For online payment, only the coefficient on the second income class changes sign;
however, in both cases one cannot reject the hypothesis that these coefficients equal zero.
And finally, for the electronic results, although the coefficients on two of the income
variables, the EDUCATION variable and the OLD variable change sign, none is
statistically significant, and we find that using direct deposit is a significant predictor of
the use of electronic payment.
Table 6B presents the results of estimating similar models using the SCF. Similar
to the Dove sample, the results indicate that direct deposit use is a significant predictor of
using debit cards and direct payment. In contrast to the Dove sample, INCOME2 and
INCOME3 have significant coefficients for debit card use, and the coefficient on NENE
reverses sign and is insignificant for the SCF sample. For direct bill payment, the
coefficients on the income variables are significant and positive, and there are some
regional coefficients that change sign and become significant. Nevertheless, the results
are the same across both data sets that direct deposit use is a significant predictor of
using electronic forms of payment.

29

The coefficient on NEMA also changes sign, but in both estimations, one cannot reject the hypothesis
that the coefficients equal zero.
30
Employers, not employees, choose whether to use direct deposit for payroll. It may be the case that
companies who pay higher wages also encourage direct deposit.

- 14 -

Table 7 presents results from including new product adoption characteristics from
the Dove sample. Again, the coefficient on direct deposit is significant and positive in
the estimations involving electronic payment. Although using a computer is not a
significant predictor in any of the estimations, the coefficient on PURCHINT is
significant and positive for paying bills with a debit card at the point of sale or for bills,
and for using online payment. While the coefficient is positive on cell phone use in all of
the estimations, it is not significant.
It is interesting to note that, except for the check estimation, none of the income
variable coefficients are significant. This points to the same result we had for the direct
deposit estimation: once we control for new product adoption, the effect of income on
choosing electronic payment is diminished, and the new technology adoption
characteristics become significant. From our correlation tables, we know that these
variables are positively correlated with one another; hence all income effects that were
present before are potentially better explained by new product adoption effects.
Although these estimations do not show causality, the correlation between technology
adoption in one sector and technology adoption in another sector is interesting from a
consumer behavior perspective.
Finally, table 8 presents results from estimations of whether the respondents to the
Dove survey used the Internet or used the paper to fill out the survey. Respondents who
used the Internet were less likely to use checks at the point of sale or to pay bills, and
more likely to pay bills online. Although the coefficients on this variable are
insignificant in the other estimations, again, this negative coefficient for check payments
is the contrapositive of our hypothesis, and continues to support our hypothesis that
adopting technology in one sector is correlated with adopting technology in another
sector.
Our final set of results, reported in tables 9A and 9B, examine the characteristics
of the point of sale or bill payment that determine payment choice. We estimate the
probability of using a payment instrument at an establishment, conditional on using that
payment instrument at a certain number of establishments.31 By doing this, we can

31

The estimation procedure follows Chamberlain (1984).

- 15 -

control for the unobserved fixed effects of the consumer, and estimate the effects of the
store type on the probability of using a certain payment instrument.
Two examples will clarify our estimation procedure. Suppose consumer i uses
payment instrument k at two establishments, j and j. The conditional probability that
consumer i shopping at establishment j uses payment instrument k is
J ik

P u *ijk = 1, u *ij 'k = 1 | uipk = 2 =

p =1

e ijk e ij 'k
,
J ik
J ik
Vipk
Vip 'k
e e

p =1
p '= p +1
where Jik is the number of stores or bills where consumer i may use payment instrument
k.
Similarly, suppose consumer i uses payment instrument k at three establishments,

j, j and j. Then our conditional probabilities take the following form:

J ik

P u * ijk = 1, u * ij 'k = 1, u * ij ''k = 1 | u * ipk = 3 =

p =1

e ijk e
J ik 2

e
p =1

Vipk

Vij ' k

Vij '' k

J ik 1 Vip 'k J ik Vip ''k

e e

p '= p +1
p ''= p +1

In this way, we eliminate the potential unobserved fixed effects of the consumer
characteristics, and we can estimate the effects of the nature of the transaction, which also
proxies for the goods bought.
The seven types of establishments for POS transactions are listed in table 3A, and
the seven types of bills are listed in table 3B. There are three POS transaction
characteristics: cashier presence, value, and self-service. The excluded categories for
these categories are cashier present, mid-range value and no self service. Three POS
establishments satisfy all of these criteria grocery stores, discount stores, and drug
stores. Hence, the coefficients should be interpreted as relative to these baseline
transactions.
We find that the characteristics of the transaction itself strongly affect payment
choice. For point of sale payments, conditional on using cash at two establishments,
consumers use cash if there is no cashier, or if it is a low-value transaction. No cashiers
have a negative effect on the probability of using a check or a debit card, but a positive
effect on the probability of using a credit card.

- 16 -

Perhaps the most striking result from this table is that the coefficient on high
value is positive only for credit transactions. Thus the nature of the transaction high
value determines the payment choice. Note that our estimation procedure eliminates
unobserved consumer characteristics that may determine credit demand. Hence, we do
not need to know whether the consumer uses the credit line for this transaction. The
fixed effect of the credit certainly affects payment choice, but our estimation gives a clear
indication that the characteristics of the individual transaction affect payment choice.
Turning to the bill payment results, there are two general categories of transaction
characteristics: frequency and variation in dollar value. The coefficients should be
interpreted as relative to a transaction that is infrequent and is of variable dollar value
charitable contributions.
We see that direct payments are more likely to be used if the bill is either in a
fixed amount or is paid frequently. If a bill is frequent or fixed, consumers are less likely
to use a check. For bills, if a bill is frequent, a respondent is less likely to use debit cards,
but if the bill is fixed, a respondent is more likely to use debit cards.
Consumers are more likely to use direct and online bill payment for frequent bills.
The difference in use of these two instruments stems from whether the dollar amount of
the bill is fixed. Consumers use direct bill payment for bills that have a fixed dollar
amount, while consumers use online bill payment for bills that have a variable dollar
amount. Part of this difference could be in the nature of the setup for bill payment. For
direct bill payments, consumers do not necessarily authorize the transaction each time.
On the other hand, consumers do authorize each transaction for online bill payment. This
difference potentially explains why consumers would be more willing to use direct
payment for fixed bill amounts than for variable bill amounts.
Overall, these results point to the fact that the nature of the transaction also
determines in part payment choice. Thus it is important to investigate technology
adoption and goods characteristics jointly in order to understand fully payment choice.

- 17 -

5. Conclusions and Suggestions for Further Research

The model presented in this study has some limitations. The framework used is a
demand framework only, and takes the supply of payment instruments as given. To form
a clearer picture of payment choice, both demand and supply would need to be included.
This supply is not uniform. For example, not every bank offers its customers the ability
to make a direct payment from an account. This affects the supply of payment
alternatives available to its customer base, and thus its consumers are limited by the
technology options offered by the bank. Stored value, or other products that did not
achieve success are not widespread options. Indeed, other consumer payments studies
have found that there is regional variation in consumer choices of payment.32
Furthermore, the demographic characteristics of a consumer may determine his
access to bank accounts or credit cards, thus limiting his choice set. However, including
demographic characteristics (including income) and the payment instrument will help to
control for the potential availability of payment instruments to consumers. According to
the Survey of Consumer Finances, in 1998, approximately 68% of consumers had a
general purpose credit card (or a bank credit card), and approximately 34% used a debit
card for making purchases at the point of sale. It is not a surprise that having declared
bankruptcy may limit a consumers access to credit cards. In 1998, only approximately
58% of households who had declared bankruptcy had a bank credit card, in contrast to
approximately 69% percent of those households who had not declared bankruptcy.33
Even more critical, and is a fault of most of payment choice studies, is missing
data on the price of these different payment instruments to consumers.34 Few data sets
contain both fee and price data in conjunction with demographic and income data.
Evidence exists that consumers may be price sensitive. In a study of direct payment
behavior, 23% of non-users claimed that they would switch to an electronic form of

32

Gowrisankaran and Stavins (2002).

Approximately 8.5% of the SCF sample households have declared bankruptcy.
34
An exception to this is Humphrey, Kim and Vale (2001), who estimate aggregate price elasticities with
respect to payment instruments using Norwegian data.
33

- 18 -

payment if offered a 5% discount on the bill to be paid.35 Further research using price
data may lend greater insight into the determinants of payment choice.
The results in this paper point to three things. First, we note that the
characteristics of the consumer affect payment choice. Demographic characteristics are
significantly correlated with payment choice, but also, new product adoption
characteristics are also significantly correlated with payment choice. Second, use of
direct deposit is a significant predictor of using other forms of electronic payment. And
third, the attributes of the transaction significantly affect the payment choices made.
We cannot say that adopting one type of technology necessarily leads to adopting
another type of technology. However, our results do indicate that use of one technology
is significantly correlated with use of another, independent technology. From a payments
perspective, because we find that technology adoption does significantly predict
electronic payment use, it may be the case that as the Internet, cell phones and direct
deposit spread to more households in the U.S., the checkless society may arrive
eventually. From a broader perspective, as technology adoption spreads in one sector,
other sectors may also experience technology adoption. It is interesting that this
common-sense idea presents itself in the results of this article.

35

See Federal Reserve Financial Services (1998), p. 13.

- 19 -

Bibliography

Avery, Robert B., Elliehausen, Gregory E., Kennickell, Arthur B. and Spindt, Paul A.
The Use of Cash and Transaction Accounts by American Families, Federal Reserve
Bulletin, vol. 72, no. 2, February 1986, p. 87-108.
Baxter, William F. 1983. Bank Interchange of Transactional Paper: Legal and
Economic Perspectives, Journal of Law and Economics, 26, October, pp. 541-588.
Black, Sandra and Morgan, Donald. Meet the New Borrowers, Current Issues in
Economics and Finance, Federal Reserve Bank of New York, February 1999.
Board of Governors of the Federal Reserve System. The Federal Reserve in the
Payments Mechanism. January 1998.
Board of Governors of the Federal Reserve System. When is Your Check Not a Check?
Electronic Check Conversion, Available for download at
http://www.federalreserve.gov/pubs/checkconv/default.htm
Carow, Kenneth A. and Staten, Michael E. Debit, Credit or Cash: Survey Evidence on
Gasoline Purchases, Journal of Economics and Business, vol. 51, September-October
1999, p.409-421.
Chamberlain, Gary. Panel Data, in Handbook of Econometrics, vol. II, Zvi Grilliches
and Michael D. Intriligator, eds. Amsterdam: Elsevier Science Publishers BV, 1984, p.
1247-1318.
Daniels, Kenneth N. and Murphy, Neil B. The Impact of Technological Change on the
Currency Behavior of Households: An Empirical Cross-section Study, Journal of
Money, Credit and Banking, vol. 26, no. 4, November 1994, p. 867-874.
Deaton, Angus. The Analysis of Household Surveys: A Microeconometric Approach to
Development Policy. Baltimore: Johns Hopkins University Press. The International
Bank for Reconstruction and Development/World Bank, 1997.
Federal Reserve Financial Services. A Summary of Consumer and Business Attitides on
Direct Deposit and Direct Payment. Available for download at
http://www.stls.frb.org/banking/ach. 1997.
Ferguson, Roger W. Jr. The Evolving Financial and Payments System, Remarks at the
Federal Reserve Bank of Philadelphia, September 4, 2001.
Foderaro, Lisa W. A Test in Cashless Spending Turns Out to be a Hard Sell, New York
Times, July 27, 1998, p. B4.

- 20 -

Gerdes, Geoffrey R. and Walton, Jack K II. The Use of Checks and Other Noncash
Payment Instruments in the United States, Federal Reserve Bulletin, vol. 88, no. 8,
August 2002, p. 360-374.
Green, Edward J. Payment Arrangements and Inflation, American Economic Review,
vol. 92, no. 2, May 2002, p. 51-57.
Gowrisankaran, Gautam and Stavins, Joanna. Network Externalities and Technology
Adoption: Lessons from Electronic Payments, Federal Reserve Bank of San Francisco
Working Paper 2002-10, May 3, 2002.
Halbfinger, David M. West Siders Testing a Plastic Substitute for Pocket Change, New
York Times, October 7, 1997, p. B1.
Hancock, Diana and Humphrey, David B. Payment Transactions, Instruments and
Systems: A Survey, Journal of Banking and Finance, vol. 21, no. 11-12, December
1997, p. 1573-1624.
Hannan, Timothy and McDowell, John M. The Determinants of Technology Adoption:
The Case of the Banking Firm, Rand Journal of Economics, vol. 15, no 3, Autumn
1984, p. 328-335.
Hansell, Saul. Got a Dime? Citibank and Chase End Test of Electronic Cash, New
York Times, November 4, 1998, p. C1.
Hausman, Jerry. Mismeasured Variables in Econometrics Analysis: Problems from the
Right and Problems from the Left, Journal of Economic Perspectives, vol. 15, no. 4, Fall
2001, p. 57-67.
Hausman, Jerry A., Abrevaya, Jason, and Scott-Morton, Fiona M. Misclassification of
the Dependent Variable in a Discrete-Response Setting, Journal of Econometrics, vol.
87, no. 2, September 1998, p. 239-269.
HSN Consultants, Inc. The Nilson Report. Oxnard, CA. No. 753, December 2001.
Huffman, Wallace E. and Mercier, Stephanie. Joint Adoption of Microcomputer
Technologies: An Analysis of Farmers Decisions, The Review of Economics and
Statistics, vol. 73, no. 3, August 1991, p. 541-546.
Humphrey, David B., Kim, Moshe and Vale, Bent. Realizing the Gains from Electronic
Payments: Costs, Pricing and Payment Choice, Journal of Money, Credit, and Banking,
33, no. 2 (May 2001, Part 1) pp. 216-234.
Humphrey, David B., Pulley, Lawrence B. and Vesala, Jukka A. Cash, Paper, and
Electronic Payments: A Cross-Country Analysis, Journal of Money, Credit, and
Banking, vol. 28, no. 4, Part 2, November 1996, p. 914-939.

- 21 -

Kennickell, Arthur B. Codebook for 1998 Survey of Consumer Finances. Board of

Governors of the Federal Reserve System, February 15, 2000.
Kennickell, Arthur B. Multiple Imputation in the Survey of Consumer Finances,
mimeo prepared for the August 1998 Joint Statistical Meetings, Dallas, TX, September
1998.
Kennickell, Arthur B. and Kwast, Myron L. Who Uses Electronic Banking? Results
from the 1995 Survey of Consumer Finances, Proceedings from the 33rd Annual
Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, 1997,
p. 56-75.
Kennickell, Arthur B., Starr-McCluer, Martha, and Surette, Brian J. Recent Changes in
U.S. Family Finances: Results from the 1998 Survey of Consumer Finances, Federal
Reserve Bulletin, vol. 86, no. 1, January 2000, p. 1-29.
Klee, Elizabeth, Guerin, Thomas and Marquardt, Jeffrey. Payment Choice: Evidence
from the Survey of Consumer Finances, mimeo, Board of Governors of the Federal
Reserve System, January 9, 2002.
Mantel, Brian. Why do Consumers Pay Bills Electronically? An Empirical Analysis,
Economic Perspectives, Federal Reserve Bank of Chicago, 4th Quarter 2000, p. 32-47.
McAndrews, James J. Network Issues and Payment Systems, Federal Reserve Bank of
Philadelphia Business Review, November/December 1997, p. 15-25.
Montalto, Catherine Phillips and Sung, Jaimie. Multiple Imputation in the 1992 Survey
of Consumer Finances, Financial Counseling and Planning, vol. 7, 1996, p. 133-146.
Montalto, Catherine Phillips and Sung, Jaimie. Estimating Nonlinear Models with
Multiply Imputed Data, Financial Counseling and Planning, vol. 9, 1998, p. 97-101.
Saloner, Garth and Shepard, Andrea. Adoption of Technologies with Network Effects:
An Empirical Examination of the Adoption of Automated Teller Machines, Rand
Journal of Economics, vol. 26, no. 3, Autumn 1995, pp. 479-501.
Schreft, Stacey L. Clicking with Dollars: How Consumers Can Pay for Purchases on the
Internet, Federal Reserve Bank of Kansas City Economic Review, First Quarter 2002, p.
37-64.
Shy, Oz. The Economics of Network Industries. New York: Cambridge University
Press, 2001.
Shy, Oz and Tarkka, Juha. The Market for Electronic Cash Cards, Journal of Money,
Credit, & Banking, vol. 34, no. 2, May 2002, p. 299-314.

- 22 -

Spahr, Walter Earl. The Clearing and Collection of Checks. New York: The Bankers
Publishing Company, 1926.
Stavins, Joanna. Effect of Consumer Characteristics on the Use of Payment
Instruments, New England Economic Review, Federal Reserve Bank of Boston, 2001,
No. 3, p. 19-31.
Train, Kenneth. Qualitative Choice Analysis: Theory, Econometrics, and an Application
to Automobile Demand. Cambridge: MIT Press, 1986.
U.S. Department of Labor, Bureau of Labor Statistics, and U.S. Department of
Commerce. Current Population Survey: Design and Methodology. Technical Paper No.
63, March 2000.
Whitesell, William C. 1992. Deposit Banks and the Market for Payment Media,
Journal of Money, Credit and Banking, 24:4, pp. 483-498.

- 23 -

Table 1: Variable Definitions

Variable Name
BKACCT
CREDIT CARD
DEBIT CARD
DIRECT DEPOSIT
COMPUTER
INTERNET
PURCHINT
CELL
INCOME1
INCOME2
INCOME3
INCOME4
INCOME5
EDUCATION
FEMALE
YOUNG
MIDDLE
OLD
NENE
NEMA
SSA
SESC
SWSC
MWENC
MWWNC
WM
WP
N

Definition
Has a bank account
Has a credit card
Has a debit card
Uses direct deposit
Owns a computer
Has Internet access at home
Has made a purchase on the Internet
Owns a cell phone
Under $20,000 annual household income
$20,000-$39,999 annual household income
$40,000-$59,999 annual household income
$60,000-$99,999 annual household income
Over $100,000 annual household income
Equals 0 if high school graduate, 1 if college graduate, 2 if graduate school
Female respondent
Younger respondent (19-34 years old)
Middle aged respondent (35-64 years old)
Older respondent (over 64 years old)
Northeast: New England
Northeast: Middle Atlantic Division (NY, NJ, PA)
South: South Atlantic Division (DE, DC, FL, GA, MD, NC, SC, VA, WV)
South: East South Central Division (AL, KY, MS, TN)
South: West South Central Division (AR, LA, OK, TX)
Midwest: East North Central Division (IL, IN, MI, OH, WI)
Midwest: West North Central Division (IA, KS, MN, MO, NE, ND, SD)
West: Mountain Division (AZ, CO, ID, MT, NV, UT, WY, NM)
West: Pacific Division (AK, CA, HI, OR, WA)
Number of observations

- 24 -

Table 2: Summary Statistics and Comparing Data Sets

POS

BILL

SCF

CPS
Supplement

Variable Name
95.64*
95.84*
92.81
BKACCT
*
88.00
87.37*
79.71
CC
63.06*
63.69*
38.50
DC
64.69*
64.55*
67.78
DIRECT DEPOSIT
82.70**
83.53**
64.03
COMPUTER
81.68**
81.88**
52.13
INTERNET
70.30**
70.75**
18.91
PURCHINT
61.57
60.94
CELL
12.16*,** 12.31*,**
20.24
17.60 (a)
INCOME1
*,**
*,**
26.03
26.35
28.93
23.61 (a)
INCOME2
**
**
22.29
21.88
20.06
19.53 (a)
INCOME3
22.45
22.12
20.69
23.67 (a)
INCOME4
*
*
17.07
17.33
10.08
15.60 (a)
INCOME5
*,**
*,**
42.32
43.14
35.08
37.45
EDUCATION = 1
39.98*,** 39.37*,**
49.76
52.81
EDUCATION = 2
17.69*,** 17.49*,**
15.15
9.73
EDUCATION = 3
59.00*,** 59.45*,**
26.82
52.22
FEMALE
19.88*,** 19.76*,**
23.02
32.42
YOUNG
68.90*,** 69.57*,**
57.34
55.30
MIDDLE
11.22*,** 10.67*,**
19.64
12.23
OLD
7.64*,**
7.69*,**
5.243
5.07
NENE
13.72
13.41
14.96
13.52
NEMA
16.45
16.71
17.82
18.32
SSA
5.61
5.49
6.24
6
SESC
7.79**
7.84
9.04
10.08
SWSC
18.00
17.57
17.61
16.49
MWENC
7.87
8.16
7.32
7.52
MWWNC
7.17
7.45
7.79
6.73
WM
15.74
15.69
13.98
16.23
WP
1283
1275
2625
62,938
N
All figures are in percent.
Bank account statistics for SCF exclude credit unions.
*
Indicates statistic is significantly different from SCF at the 95% confidence level.
**
Indicates statistic is significantly different from CPS Supplement at the 95% confidence level.
(a)
The CPS Supplement is income-topcoded at $100,000. These statistics reflect the CPS, and thus may not
be exactly comparable.

- 25 -

Table 3A: Point of Sale Statistics

GROCERY GAS DEPARTMENT DISCOUNT
DRUG
RESTAURANT
71.64 76.26
54.73
63.55
71.65
81.20
CASH
45.80 17.86
43.70
50.32
45.59
19.54
CHECK
26.68 51.16
58.82
41.60
35.50
56.30
CREDIT
32.98 21.32
14.60
18.80
21.01
7.98
PIN DEBIT
14.29 17.47
17.86
13.24
15.44
19.75
SIGNATURE
DEBIT
44.64 36.55
30.99
30.25
33.93
27.10
DEBIT
N = 952.
All figures are in percent.
Only respondents who indicated payment preferences for all stores are included in sample.

FAST FOOD
97.58
16.18
15.97
7.14
7.04
13.76

Table 3B: Bill Payment Statistics

RENT
CASH
CHECK
CREDIT

26.84
86.50
4.60

LOAN
20.55
83.59
6.90

CREDIT
CARD BILL
17.02
86.66
2.45

INSURANCE
20.40
83.90
8.44

MEMBERSHIP

TUITION

15.64
86.35
26.23

3.53
5.06
5.37
4.60
9.05
DEBIT
19.33
23.93
13.19
21.78
7.36
DIRECT
PAYMENT
5.52
6.13
13.80
7.06
5.98
ONLINE
N = 652.
All figures are in percent.
Only respondents who indicated payment preferences for all stores are included in sample.

- 26 -

UTILITY

32.82
85.58
17.02

24.54
85.89
9.36

4.14
4.29

6.13
15.18

2.76

9.51

Table 4: Dove Survey Correlations

INCOME1
INCOME2
INCOME3
INCOME4
INCOME5
EDUCATION
FEMALE
YOUNG
MIDDLE
OLD

DIRECT
DEPOSIT
-0.1596**
(<.0001)
-0.1064**
(0.0001)
0.0098
(0.7227)
0.0920**
(0.0009)
0.1488**
(<.0001)
0.2060**
(<.0001)
-0.0618**
(0.0254)
-0.0326
(0.2385)
-0.0426
(0.1232)
0.1307**
(<.0001)

COMPUTER

INTERNET

PURCHINT

CELL

-0.0407
(0.1413)
-0.0958**
(0.0005)
-0.0097
(0.7265)
0.0585**
(0.0345)
0.0946**
(0.0006)
0.0739**
(0.0075)
-0.0245
(0.3769)
-0.0202
(0.4646)
0.1234**
(<.0001)
-0.1943**
(<.0001)

-0.0872**
(0.0016)
-0.0663**
(0.0164)
-0.009
(0.745)
0.0408
(0.1405)
0.1180**
(<.0001)
0.0845**
(0.0022)
-0.0039
(0.8869)
-0.0280
(0.3114)
0.1093**
(<.0001)
-0.1560**
(<.0001)

-0.1272**
(<.0001)
-0.0929**
(0.0008)
0.0033
(0.9063)
0.0811**
(0.0033)
0.1248**
(<.0001)
0.1368**
(<.0001)
0.0039
(0.8871)
0.0414
(0.1343)
0.0597**
(0.0308)
-0.1761**
(<.0001)

-0.1741**
(<.0001)
-0.1407
(<.0001)
-0.0028
(0.9198)
0.0942**
(0.0006)
0.2140**
(<.0001)
0.1156**
(<.0001)
-0.0273
(0.3241)
-0.0418
(0.1311)
0.0759**
(0.0061)
-0.0723**
(0.0089)

N = 1323.
**
Significant at the 5% level.

- 27 -

Table 5A: Dove Survey Baseline Results

POS
CHECK

POS
DEBIT

BILL
CHECK

Independent
Variable
INCOME2

BILL
DEBIT

BILL
DIRECT

0.4246**
0.4297**
0.8161**
-0.1423
0.0126
(0.2015)
(0.2104)
(0.2648)
(0.2085)
(0.2012)
0.7041**
0.4707**
1.5028**
0.1308
0.4800**
INCOME3
(0.2105)
(0.2157)
(0.3184)
(0.2131)
(0.2074)
0.9099**
0.6748**
2.0158**
-0.0093
0.4299**
INCOME4
(0.2212)
(0.2232)
(0.3955)
(0.2227)
(0.2151)
0.3638
1.8129**
-0.4608
0.4042
0.4286
INCOME5
(0.2402)
(0.2455)
(0.4325)
(0.2532)
(0.2364)
-0.0173
0.1245
0.1851
-0.1334
0.1588
EDUCATION
(0.0889)
(0.0869)
(0.1682)
(0.0913)
(0.0855)
0.0537
0.0818
-0.1114
-0.0500
0.1234
FEMALE
(0.1285)
(0.1258)
(0.2373)
(0.1317)
(0.1247)
0.3418**
-0.5908**
0.8855**
0.1040
0.4658**
MIDDLE
(0.1480)
(0.1481)
(0.2254)
(0.1536)
(0.1484)
0.9467**
-1.3849**
1.2677**
-0.8462**
0.2550
OLD
(0.2342)
(0.2371)
(0.4191)
(0.2649)
(0.2235)
-0.8309**
0.5652**
0.0761
0.1045
-0.1435
NENE
(0.2522)
(0.2510)
(0.4609)
(0.2690)
(0.2480)
-0.1436
0.8086**
0.1642
-0.0051
-0.6978**
NEMA
(0.2089)
(0.2106)
(0.4117)
(0.2214)
(0.2066)
-0.1719
0.3233
0.1620
0.4201**
-0.1609
SSA
(0.2021)
(0.1967)
(0.3429)
(0.2055)
(0.1948)
0.2116
0.2415
0.2798
0.2912
0.1734
SESC
(0.2983)
(0.2783)
(0.5015)
(0.2910)
(0.2789)
0.1354
0.1623
0.4418
0.0557
-0.3912
SWSC
(0.2597)
(0.2472)
(0.4381)
(0.2645)
(0.2492)
-0.0755
-0.2818
0.2586
-0.1026
-0.0212
MWWNC
(0.2542)
(0.2557)
(0.4191)
(0.2662)
(0.2417)
0.4001
-0.0249
0.5213**
0.2767
-0.1312
WM
(0.2597)
(0.2532)
(0.4023)
(0.2572)
(0.2502)
-0.0037
0.4344**
0.1041
0.2978
0.1165
WP
(0.2075)
(0.1996)
(0.3588)
(0.2121)
(0.1986)
-0.2232
-0.5306
0.1139
-0.5275**
-1.0201**
Constant
(0.2913)
(0.2911)
(0.4550)
(0.2997)
(0.2880)
-822.547
-845.163
-339.184
-794.821
-862.617
Log-Likelihood
787
577
1163
431
617
N
N (Total, POS) = 1283.
N (Total, Bill) = 1275.
N (Specific estimation) = Number of respondents who use the payment instrument.
Standard errors are in parentheses.
**
Significant at the 5% level.

- 28 -

BILL
ONLINE

0.0348
(0.2682)
0.2876
(0.2682)
0.3942
(0.2726)
0.7951**
(0.2872)
0.1655
(0.1005)
-0.1180
(0.1472)
-0.0717
(0.1718)
-0.6247**
(0.2950)
0.7468**
(0.2677)
0.3093
(0.2393)
-0.2060
(0.2429)
-0.3192
(0.3698)
-0.4578
(0.3278)
-0.5079
(0.3333)
0.1219
(0.2935)
0.3254
(0.2273)
-1.6924**
(0.3524)
-658.463
300

Table 5B: SCF Baseline Results

POS
DEBIT
Independent
Variable
INCOME2
INCOME3
INCOME4
INCOME5
EDUCATION
FEMALE
MIDDLE
OLD
NENE
NEMA
SSA
SESC
SWSC
MWWNC
WM
WP
Constant
Log-Likelihood
N

0.3487**
(0.1423)
0.4439**
(0.1570)
0.4151**
(0.1606)
0.2011
(0.1707)
0.1808**
(0.0670)
-0.0180
(0.1134)
-0.7880**
(0.1094)
-1.7727**
(0.1529)
-0.3116
(0.2152)
0.0976
(0.1472)
0.2137
(0.1440)
0.1824
(0.2123)
-0.0794
(0.1754)
-0.4047**
(0.2028)
0.3776**
(0.1842)
0.3660**
(0.1467)
-0.5272**
(0.1933)
-1616.723
954

BILL
DIRECT
0.6166**
(0.1441)
1.0873**
(0.1566)
1.1204**
(0.1605)
1.0703**
(0.1682)
0.2590**
(0.0644)
-0.0088
(0.1103)
-0.0514
(0.1113)
0.0175
(0.1350)
-0.0835
(0.2007)
0.0010
(0.1441)
0.3180**
(0.1397)
0.6772**
(0.2056)
0.2300
(0.1688)
0.5838**
(0.1830)
0.5588**
(0.1816)
0.1073
(0.1441)
-1.8378**
(0.1977)
-1704.578
1082

N (Total) = 2625.
N (Specific estimation) = Number of respondents who use the payment instrument.
Standard errors are in parentheses.
**
Significant at the 5% level.

- 29 -

Table 6A: Dove Survey Direct Deposit Results

POS
CHECK

POS
DEBIT

BILL
CHECK

Independent
Variable
DIRECT
DEPOSIT
INCOME2

BILL
DEBIT

BILL
DIRECT

0.0326
0.6681**
0.8594**
0.4496**
0.8505**
(0.1319)
(0.1325)
(0.2287)
(0.1376)
(0.1314)
0.4210**
0.3594
0.7573**
-0.1937
-0.0836
(0.2021)
(0.2124)
(0.2685)
(0.2100)
(0.2059)
0.6974**
0.3389
1.3326**
0.0331
0.3106
INCOME3
(0.2123)
(0.2191)
(0.3237)
(0.2160)
(0.2130)
0.9003**
0.4876**
1.7778**
-0.1474
0.1833
INCOME4
(0.2245)
(0.2280)
(0.4009)
(0.2277)
(0.2224)
-0.6114**
0.1390
0.4182
0.1542
1.5222**
INCOME5
(0.4411)
(0.2583)
(0.2438)
(0.2439)
(0.2513)
-0.0202
0.0671
0.1058
-0.1786
0.0822
EDUCATION
(0.0897)
(0.0885)
(0.1682)
(0.0931)
(0.0878)
0.0547
0.1064
-0.0831
-0.0402
0.1469
FEMALE
(0.1286)
(0.1273)
(0.2395)
(0.1324)
(0.1266)
0.3407**
-0.6273**
0.8484**
0.0895
0.4604**
MIDDLE
(0.1512)
(0.1481)
(0.1502)
(0.2283)
(0.1541)
0.9365**
-1.6055**
0.9017
-0.9984**
0.0047
OLD
(0.2378)
(0.2429)
(0.4314)
(0.2701)
(0.2290)
0.5005**
-0.0374
0.0557
-0.2328
-0.8343**
NENE
(0.2526)
(0.2538)
(0.4649)
(0.2704)
(0.2511)
-0.6972**
-0.1348
0.8606**
0.1714
0.0005
NEMA
(0.2089)
(0.2126)
(0.4165)
(0.2221)
(0.2102)
-0.1703
0.3595
0.2314
0.4460**
-0.1246
SSA
(0.2023)
(0.1989)
(0.3462)
(0.2065)
(0.1984)
0.2105
0.2190
0.2514
0.2715
0.1432
SESC
(0.2983)
(0.2814)
(0.5054)
(0.2927)
(0.2843)
0.1364
0.1834
0.4571
0.0582
-0.3926
SWSC
(0.2597)
(0.2504)
(0.4424)
(0.2658)
(0.2532)
-0.0746
-0.2781
0.3173
-0.0988
-0.0104
MWWNC
(0.2543)
(0.2585)
(0.4244)
(0.2673)
(0.2467)
0.2437
-0.1336
0.3598
-0.0637
0.5040**
WM
(0.2585)
(0.2544)
(0.2599)
(0.2558)
(0.4089)
-0.0024
0.4675**
0.1180
0.3055
0.1353
WP
(0.2076)
(0.2018)
(0.3613)
(0.2130)
(0.2021)
-0.2316
-0.7111**
-0.0570
-0.6342**
-1.2610**
Constant
(0.2933)
(0.2967)
(0.4596)
(0.3033)
(0.2961)
-822.5161 -832.1849 -331.9607 -789.3762
-841.1114
Log-Likelihood
787
577
1163
431
617
N
N (Total, POS) = 1283.
N (Total, Bill) = 1275.
N (Specific estimation) = Number of respondents who use the payment instrument.
Standard errors are in parentheses.
**
Significant at the 5% level.

- 30 -

BILL
ONLINE
0.5917**
(0.1634)
-0.0361
(0.2703)
0.1618
(0.2717)
0.2208
(0.2779)
0.6123**
(0.2926)
0.1140
(0.1025)
-0.0986
(0.1482)
-0.0780
(0.1731)
-0.7696**
(0.2977)
0.6951**
(0.2693)
0.3128
(0.2410)
-0.1821
(0.2443)
-0.3466
(0.3713)
-0.4570
(0.3290)
-0.5095
(0.3347)
0.1003
(0.2948)
0.3385
(0.2286)
-1.8776**
(0.3603)
-651.6330
300

Table 6B: SCF Direct Deposit Results

POS
DEBIT
Independent
Variable
DIRECT
DEPOSIT
INCOME2
INCOME3
INCOME4
INCOME5
EDUCATION
FEMALE
MIDDLE
OLD
NENE
NEMA
SSA
SESC
SWSC
MWWNC
WM
WP
Constant
Log-Likelihood
N

0.5272**
(0.0920)
0.2967**
(0.1429)
0.3501**
(0.1583)
0.3338**
(0.1620)
0.1549
(0.1718)
0.1500**
(0.0678)
-0.0404
(0.1143)
-0.7819**
(0.1103)
-1.9051**
(0.1555)
-0.3152
(0.2155)
0.1274
(0.1480)
0.2020
(0.1449)
0.2017
(0.2138)
-0.0690
(0.1766)
-0.4462
(0.2043)
0.3505
(0.1858)
0.3874**
(0.1477)
-0.7347**
(0.1982)
-1599.948
954

BILL
DIRECT
1.0239**
(0.0931)
0.5515**
(0.1467)
0.9704**
(0.1598)
1.0346**
(0.1638)
1.0622**
(0.1717)
0.2062**
(0.0662)
-0.0495
(0.1128)
-0.0114
(0.1147)
-0.1901
(0.1386)
-0.0936
(0.2049)
0.0557
(0.1477)
0.3200**
(0.1429)
0.7494**
(0.2132)
0.2751
(0.1733)
0.5515**
(0.1868)
0.5291**
(0.1865)
0.1471
(0.1478)
-2.3575**
(0.2108)
-1640.736
1082

N (Total) = 2625.
N (Specific estimation) = Number of respondents who use the payment instrument.
Standard errors are in parentheses.
**
Significant at the 5% level.

- 31 -

Table 7: Dove Survey New Product Adoption Results

POS
CHECK
Independent
Variable
DIRECT
DEPOSIT
COMPUTER

POS
DEBIT

BILL
CHECK

BILL
DEBIT

BILL
DIRECT

0.0384
0.6523**
0.8388**
0.4339**
0.8373**
(0.1325)
(0.1335)
(0.2304)
(0.1385)
(0.1317)
-0.3158
0.1578
0.0269
0.2815
0.0398
(0.1825)
(0.1810)
(0.2902)
(0.1931)
(0.1771)
-0.1073
0.4299**
0.8265**
0.3329**
0.1607
PURCHINT
(0.1498)
(0.1494)
(0.2323)
(0.1546)
(0.1465)
0.0195
0.1882
0.1485
0.1544
0.2039
CELL
(0.1321)
(0.1316)
(0.2268)
(0.1363)
(0.1304)
0.3128
0.6737**
-0.2218
-0.1202
0.4201**
INCOME2
(0.2038)
(0.2147)
(0.2750)
(0.2130)
(0.2076)
0.7126**
0.2444
1.2167**
-0.0416
0.2382
INCOME3
(0.2159)
(0.2230)
(0.3326)
(0.2213)
(0.2168)
0.9306**
0.3342
1.5774**
-0.2711
0.0762
INCOME4
(0.2314)
(0.2353)
(0.4142)
(0.2369)
(0.2296)
0.4678**
-0.0663
1.2582**
-0.7919**
-0.0088
INCOME5
(0.2555)
(0.2637)
(0.4604)
(0.2723)
(0.2555)
-0.0105
0.0460
0.0518
-0.1990**
0.0737
EDUCATION
(0.0902)
(0.0893)
(0.1701)
(0.0939)
(0.0882)
0.0689
0.0802
-0.1477
-0.0681
0.1336
FEMALE
(0.1292)
(0.1285)
(0.2421)
(0.1335)
(0.1270)
0.3398**
-0.6137**
0.8849**
0.0980
0.4656**
MIDDLE
(0.1487)
(0.1514)
(0.2329)
(0.1549)
(0.1519)
0.7998**
-1.3673**
1.2771**
-0.7718**
0.1210
OLD
(0.2472)
(0.2513)
(0.4568)
(0.2782)
(0.2390)
-0.8006**
0.4543
-0.1708
0.0086
-0.2516
NENE
(0.2531)
(0.2556)
(0.4739)
(0.2717)
(0.2519)
-0.7025**
-0.1223
0.8759**
0.1754
0.0072
NEMA
(0.2095)
(0.2144)
(0.4255)
(0.2233)
(0.2110)
-0.1759
0.3597
0.1890
0.4540**
-0.1281
SSA
(0.2026)
(0.2003)
(0.3523)
(0.2078)
(0.1990)
0.2236
0.2260
0.1622
0.2682
0.1467
SESC
(0.2993)
(0.2827)
(0.5071)
(0.2941)
(0.2849)
0.1426
0.1716
0.3918
0.0511
-0.4000
SWSC
(0.2603)
(0.2516)
(0.4484)
(0.2669)
(0.2538)
-0.0842
-0.2554
0.3079
-0.0685
0.0142
MWWNC
(0.2554)
(0.2605)
(0.4278)
(0.2694)
(0.2478)
-0.1251
0.3855
-0.0775
0.5179**
0.2661
WM
(0.2605)
(0.2580)
(0.4135)
(0.2600)
(0.2555)
0.0186
0.4367**
-0.0095
0.2773
0.1258
WP
(0.2085)
(0.2036)
(0.3677)
(0.2141)
(0.2028)
-0.4340
-1.0892**
-1.4434**
0.0585
-1.1292**
Constant
(0.3322)
(0.5080)
(0.3446)
(0.3297)
(0.3251)
-820.0010 -824.5609 -324.9113 -783.4376
-838.9459
Log-Likelihood
787
577
1163
431
617
N
N (Total, POS) = 1283.
N (Total, Bill) = 1275.
N (Specific estimation) = Number of respondents who use the payment instrument.
Standard errors are in parentheses.
**
Significant at the 5% level.

BILL
ONLINE
0.5420**
(0.1682)
0.4032
(0.2779)
1.7325**
(0.2554)
0.2252
(0.1660)
-0.1478
(0.2831)
-0.0556
(0.2867)
-0.1123
(0.2959)
0.1788
(0.3180)
0.0498
(0.1060)
-0.1568
(0.1545)
-0.0019
(0.1779)
-0.0890
(0.3186)
0.6149**
(0.2782)
0.3381
(0.2504)
-0.1775
(0.2519)
-0.3127
(0.3802)
-0.5070
(0.3376)
-0.4707
(0.3450)
0.1612
(0.3060)
0.2488
(0.2356)
-3.4713**
(0.4763)
-609.1433
300

Table 8: Dove Survey Internet Results

POS
CHECK

POS
DEBIT

BILL
CHECK

Independent
Variable
INTERNET
RESPONSE
DIRECT
DEPOSIT
COMPUTER

BILL
DEBIT

BILL
DIRECT

-0.9627**
-0.2382
-0.7850**
0.2111
0.0826
(0.1484)
(0.1406)
(0.2952)
(0.1466)
(0.1401)
-0.0035
0.6453
0.7893**
0.4429**
0.8412**
(0.1351)
(0.1337)
(0.2326)
(0.1388)
(0.1319)
-0.0580
0.2173
0.2716
0.2257
0.0179
(0.1891)
(0.1845)
(0.3063)
(0.1970)
(0.1810)
0.1653
0.4994
0.9935**
0.2755
0.1375
PURCHINT
(0.1581)
(0.1554)
(0.2415)
(0.1596)
(0.1516)
0.0134
0.1875
0.1332
0.1564
0.2045
CELL
(0.1342)
(0.1318)
(0.2290)
(0.1365)
(0.1304)
0.3366**
0.2897
0.6135**
-0.1999
-0.1117
INCOME2
(0.2071)
(0.2153)
(0.2776)
(0.2136)
(0.2080)
0.5554**
0.2017
1.0796**
-0.0009
0.2540
INCOME3
(0.2200)
(0.2246)
(0.3364)
(0.2232)
(0.2185)
0.6574**
0.2581
1.3945**
-0.2025
0.1023
INCOME4
(0.2381)
(0.2398)
(0.4199)
(0.2415)
(0.2337)
0.1355
-0.1526
1.0531**
-0.7170**
0.0211
INCOME5
(0.2647)
(0.2688)
(0.4693)
(0.2773)
(0.2604)
-0.0581
0.0349
-0.0027
-0.1885
0.0783
EDUCATION
(0.0922)
(0.0896)
(0.1726)
(0.0943)
(0.0885)
0.2537
0.1265
0.0301
-0.1097
0.1181
FEMALE
(0.1354)
(0.1314)
(0.2521)
(0.1369)
(0.1297)
0.3575**
-0.6138
0.8831**
0.0924
0.4640**
MIDDLE
(0.1515)
(0.1518)
(0.2343)
(0.1550)
(0.1519)
0.6788**
-1.4027
1.1192
-0.7437**
0.1327
OLD
(0.2506)
(0.2528)
(0.4600)
(0.2789)
(0.2398)
-0.8373**
0.4539
-0.1497
0.0067
-0.2517
NENE
(0.2585)
(0.2558)
(0.4781)
(0.2721)
(0.2520)
-0.7224**
-0.1214
0.8841**
0.1736
0.0069
NEMA
(0.2137)
(0.2147)
(0.4289)
(0.2236)
(0.2110)
-0.1236
-0.2309
0.3472
0.1716
0.4668**
SSA
(0.2081)
(0.1991)
(0.2067)
(0.2007)
(0.3558)
0.2064
0.2180
0.1446
0.2755
0.1488
SESC
(0.3041)
(0.2831)
(0.5090)
(0.2943)
(0.2850)
0.1231
0.1682
0.3424
0.0556
-0.3985
SWSC
(0.2635)
(0.2517)
(0.4486)
(0.2673)
(0.2539)
-0.0926
-0.2593
0.3276
-0.0650
0.0152
MWWNC
(0.2599)
(0.2611)
(0.4319)
(0.2694)
(0.2477)
-0.1042
0.3905
-0.0823
0.5152
0.2651
WM
(0.2649)
(0.2584)
(0.4170)
(0.2601)
(0.2556)
0.0552
0.4461
0.0109
0.2701
0.1231
WP
(0.2124)
(0.2038)
(0.3697)
(0.2144)
(0.2030)
0.4138
-1.0419**
-0.0990
-1.1698**
-1.4754**
Constant
(0.3350)
(0.3363)
(0.5276)
(0.3493)
(0.3342)
-797.985
-823.120
-321.176
-782.398
-838.772
Log-Likelihood
787
577
1163
431
617
N
N (Total, POS) = 1283.
N (Total, Bill) = 1275.
N (Specific estimation) = Number of respondents who use the payment instrument.
Standard errors are in parentheses.
**
Significant at the 5% level.

- 33 -

BILL
ONLINE
0.4590**
(0.1685)
0.5482**
(0.1688)
0.3070
(0.2805)
1.5993**
(0.2586)
0.2242
(0.1664)
-0.1084
(0.2838)
0.0298
(0.2883)
0.0332
(0.3007)
0.3490
(0.3247)
0.0730
(0.1069)
-0.2369
(0.1582)
-0.0220
(0.1786)
-0.0340
(0.3194)
0.6211**
(0.2795)
0.3356
(0.2513)
-0.1506
(0.2531)
-0.3023
(0.3813)
-0.4955
(0.3382)
-0.4664
(0.3459)
0.1370
(0.3079)
0.2322
(0.2367)
-3.6510**
(0.4822)
-605.363
300

Table 9A: POS Transaction Characteristics

Independent
Variable
NO
CASHIER
HIGH AVG.
VALUE
LOW AVG.
VALUE

CASH

CHECK

CREDIT

DEBIT

1.2046**
(0.1161)
-1.2985**
(0.1109)
4.8809**
(0.3644)

-2.9699**
(0.1370)
-0.3845**
(0.1127)
-3.4760**
(0.1479)

2.0565**
(0.1176)
2.3318**
(0.1229)
-1.8803**
(0.1241)

-1.2083**
(0.1266)
-0.7023**
(0.1261)
-2.9226**
(0.1428)

-3.2136**
1.5426**
0.0377**
0.6843**
SELF(0.1087)
(0.1420)
(0.1104)
(0.1268)
SERVICE
-1020.190
-904.054
-1101.636
-865.317
Loglikelihood
530
523
590
416
N
N (Specific Estimation) = Number of observations used in estimation.
Base = (Grocery store, drug store, discount store)
Log-likelihood is conditional on using payment instrument at one of the three stores.
Standard errors are in parentheses.
**
Significant at the 5% level.

Table 9B: Bill Transaction Characteristics

Independent
Variable
FREQUENT
FIXED
AMOUNT
Log-likelihood

CHECK

DEBIT

DIRECT

ONLINE

-0.2676**
(0.1313)
-0.3322**
(0.1321)
-506.852

-0.1891
(0.1741)
0.1263
(0.1769)
-271.939

2.2721**
(0.1503)
0.8248**
(0.1143)
-665.919

1.9963**
(0.2277)
-0.9920**
(0.1812)
-241.876

184
92
275
106
N
N (Specific Estimation) = Number of observations used in estimation.
Base = (Charitable contribution)
Log-likelihood is conditional on using payment instrument at one of the three stores.
Standard errors are in parentheses.
**
Significant at the 5% level.

- 34 -

Appendix: Data Issues and Sampling Discussion

Table A1 gives a summary of data problems with the Dove survey. Due to
missing values and observations, of the 1,499 responses received, 1,336 may be used for
estimating models related to payment choice at the point of sale, and 1,327 may be used
for estimating models related to bill payments. Moreover, as presented in table 2, a few
of the responses were inconsistent.36 We also dropped observations if the respondent is
under 18 years old (1 observation) or resides in Canada (41 observations).
The missing values and inconsistent replies are theoretically inconsequential if the
incidence of missing values and inconsistent replies is uncorrelated with the hypotheses
we are trying to test. This is not the case in this sample. A higher incidence of missing
values occurs in the older and retired subset of the Dove sample, in the subset who do not
use new products (including a computer or the Internet), or have not made a purchase
over the Internet. This presents particular problems regarding the answers to questions in
the bill payment portion of the survey. However, we tested to see whether possible
misclassification biases our results.37 The results of our tests indicate that the bias due to
misclassification is not significantly different from zero.

36

An example of an inconsistent reply is if a respondent indicated that he did not have a credit card in one
question, but indicates that he uses a credit card in another.
37
See Hausman, Abrevaya and Scott-Morton (1998) for the method used.

- 35 -

Table A1: Dove Survey Response Rate

Paper
5,754
646
11.2%
43.1%

Number Surveyed
Number of Respondents
Response Rate
Percent of Total Responses

Web
9,893
853
8.6%
56.9%

Total
15,637
1,499
9.6%
100%

Table A2 (a): Dove Survey Missing Responses POS

Paper

Web

12
80
88
86.4%
41.8%

Dependent
Independent
Total
Percent Valid Responses
Percent of Total Responses

Total
35
41
75
91.2%
58.2%

47
121
163
89.1%
100.0%

Table A2 (b): Dove Survey Missing Responses Bill Payment

Paper
36
80
105
83.7%
40.8%

Dependent
Independent
Total
Percent Valid Responses
Percent of Total Responses

- 36 -

Web

Total
27
41
67
92.1%
59.2%

63
121
172
88.5%
100.0%

Table A3 (a): Dove Survey Inconsistent Responses Bill Payment (N=1327)

Paper

Web
6
0
1
17
5

Check
Credit
Debit
Direct
Online

Total
8
1
0
18
15

14
1
1
35
20

Table A3 (b): Dove Survey Inconsistent Responses POS (N=1336)

Paper

Web
3
0
3
2
7

Check
Credit
PIN debit
Signature debit
Prepaid card

- 37 -

Total
1
0
6
0
2

4
0
9
2
9