the economics of payment essays on the impact of payment ...file/dis4697.pdf · essays on the...
TRANSCRIPT
The Economics of Payment
Essays on the Impact of Payment Innovationson Individual Payment Behavior
D I S S E R T A T I O N
of the University of St. Gallen,School of Management, Economics, Law,Social Sciences, and International Affairs
to obtain the the title of
Doctor of Philosophy in InternationalAffairs and Political Economy
submitted by
Tobias Trutsch
from
Unteriberg (Schwyz)
Approved on the application of
Prof. Dr. Beat Bernet
and
Prof. Dr. Monika Butler
Dissertation no. 4697
Difo-Druck GmbH, Bamberg 2017
The University of St. Gallen, School of Management, Economics, Law, SocialSciences, and International Affairs hereby consents to the printing of thepresent dissertation, without hereby expressing any opinion on the viewsherein expressed.
St. Gallen, May 31, 2017
The President:
Prof. Dr. Thomas Bieger
For my beloved ones
Acknowledgements
When I started this project in 2011, the topic of individual payment behavior
was neither on top of the agenda of academic research nor in the collective
public consciousness. Now, in an economically extraordinary environment
accompanied by historically low and even negative interest rates, an unprece-
dented amount of central bank money floating financial markets as well as
experienced spying affairs by governments, it has become part of daily dis-
cussions either led by scholars, newspapers or the public. This is because
there is an actual debate ongoing about the restriction of cash payments and
lastly about phasing out paper currency. However, cash is suggested to still
be the sole payment instrument that represents vital freedom enabling to
protect everyone’s savings from negative interest rates, the amputation of
bank deposits and data robbery, amongst others. During this changing envi-
ronment and within these years of writing, I have always enjoyed academic
freedom that made it possible to pursue my research interests.
Therefore, first and foremost, I want to thank my supervisor Beat Bernet
who granted me complete academic freedom in preparing and accomplishing
this thesis. I always could benefit from his valuable inputs during the process
of writing. I am also very grateful to my co-supervisor Monika Butler for
her willingness to support me as my second referee. Her comments helped
to substantially improve my work. Similarly, I thank Stefan Jaeger for his
commitment as my third referee. I extend my gratitude to Franz Jaeger
who provided me a sound and flexible working atmosphere at the Executive
School. His constructive advice on how to proceed was always very help-
ful. In addition, my colleagues at the institute, especially Carolin Gussow,
deserve credit for having frequently offered me open-minded and fulfilling
conversations that made the research process more pleasant. This thesis
also benefited from many comments by my fellow doctoral students, Carolin
Gussow and my friend Florin Botschi, who actually spared no effort to proof-
read parts of it. I am furthermore deeply indebted to my parents Manfred
and Agatha Trutsch who made my education possible. They always sup-
ported me in what I was willing to do. Without their assistance this thesis
would have never been written. I also want to express my gratitude towards
my uncle Hanspeter Trutsch, who made my time as a Bachelor and Master
student sharing his flat a life-time experience. Lastly, I thank Ana Lopez for
her love and encouragement.
St. Gallen, March 2017 Tobias Trutsch
Contents
List of Figures v
List of Tables vii
Abstract xi
Zusammenfassung (Summary in German) xiii
1 Introduction 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Determinants of Payment Behavior . . . . . . . . . . . . . . . 5
1.2.1 Models of Technology Acceptance . . . . . . . . . . . 6
1.2.2 Payment Instrument Attributes . . . . . . . . . . . . . 11
1.2.3 Physical Factors . . . . . . . . . . . . . . . . . . . . . 18
1.2.4 Psychological Aspects . . . . . . . . . . . . . . . . . . 19
1.3 Theoretical Model of Payment Behavior . . . . . . . . . . . . 27
1.4 Research Topics and Objectives . . . . . . . . . . . . . . . . . 31
1.4.1 Contactless Payment and Transaction Frequency . . . 32
1.4.2 Contactless Payment and Cash Usage . . . . . . . . . 33
1.4.3 Mobile Payment and Payment Choice . . . . . . . . . 34
1.5 Research Contribution . . . . . . . . . . . . . . . . . . . . . . 35
1.5.1 Chapter 2: The Impact of Contactless Payment on
Transaction Frequency . . . . . . . . . . . . . . . . . . 36
1.5.2 Chapter 3: The Impact of Contactless Payment on
Cash Usage . . . . . . . . . . . . . . . . . . . . . . . . 37
i
1.5.3 Chapter 4: The Impact of Mobile Payment on Pay-
ment Choice . . . . . . . . . . . . . . . . . . . . . . . 37
1.6 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2 The Impact of Contactless Payment on Transaction Frequency 39
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2 Theoretical Considerations . . . . . . . . . . . . . . . . . . . . 44
2.2.1 Technology Acceptance Model . . . . . . . . . . . . . 44
2.2.2 Innovation Diffusion Theory . . . . . . . . . . . . . . . 45
2.2.3 Unified Theory of Acceptance and Use of Technology . 45
2.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3.1 Source . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3.2 Description . . . . . . . . . . . . . . . . . . . . . . . . 48
2.4 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.4.1 Identifying Assumptions . . . . . . . . . . . . . . . . . 57
2.4.2 Estimation Strategy . . . . . . . . . . . . . . . . . . . 59
2.4.3 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . 62
2.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.5.1 Estimation Results . . . . . . . . . . . . . . . . . . . . 62
2.5.2 Results of the Sensitivity Analysis . . . . . . . . . . . 72
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3 The Impact of Contactless Payment on Cash Usage 77
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . 81
3.3 Theoretical Background . . . . . . . . . . . . . . . . . . . . . 84
3.4 Estimation Strategy . . . . . . . . . . . . . . . . . . . . . . . 85
3.5 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.5.1 Source . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.5.2 Description . . . . . . . . . . . . . . . . . . . . . . . . 89
3.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.6.1 Estimation Results for Cash Value . . . . . . . . . . . 108
3.6.2 Estimation Results for Cash Volume . . . . . . . . . . 109
3.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
ii
3.A Appendix: Descriptives and Regression Tables . . . . . . . . 117
4 The Impact of Mobile Payment on Payment Choice 135
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.2 Related Literature . . . . . . . . . . . . . . . . . . . . . . . . 140
4.3 Model Specification of Payment Choice . . . . . . . . . . . . . 143
4.4 Identification Strategy . . . . . . . . . . . . . . . . . . . . . . 146
4.4.1 Identifying Assumptions . . . . . . . . . . . . . . . . . 147
4.4.2 Estimating the Adoption of Payment Instruments . . 149
4.4.3 Estimating the Usage of Payment Instruments . . . . 151
4.5 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.5.1 Source . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.5.2 Description . . . . . . . . . . . . . . . . . . . . . . . . 153
4.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.6.1 Estimation Results of the Adoption Stage . . . . . . . 162
4.6.2 Estimation Results of the Usage Stage . . . . . . . . . 167
4.7 Plausibility Check . . . . . . . . . . . . . . . . . . . . . . . . 175
4.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
4.A Appendix: Assessment of Payment Method Characteristics . 180
5 Concluding Remarks 183
5.1 Synopsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
5.2 Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
5.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
5.4 Directions for Future Research . . . . . . . . . . . . . . . . . 192
References 197
iii
List of Figures
1.1 Innovation Diffusion Theory . . . . . . . . . . . . . . . . . . . 7
1.2 Technology Acceptance Model . . . . . . . . . . . . . . . . . . 9
1.3 Unified Theory of Acceptance and Use of Technology . . . . . 10
1.4 Theoretical Model of Payment Behavior . . . . . . . . . . . . 29
2.1 Share of Credit Card Payments per Month at the POS . . . . 53
2.2 Share of Debit Card Payments per Month at the POS . . . . 53
2.3 Transaction Frequency vs. Propensity Score of Contactless
Credit Cards . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.4 Transaction Frequency vs. Propensity Score of Contactless
Debit Cards . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.5 Common Support for Contactless Credit Cards . . . . . . . . 68
2.6 Common Support for Contactless Debit Cards . . . . . . . . 69
3.1 Adoption Rate of Contactless Payment across Years . . . . . 91
3.2 Cash Measures (in USD) across Years . . . . . . . . . . . . . 94
3.3 Cash Measures across Years . . . . . . . . . . . . . . . . . . . 95
3.4 Interest Rates across Years . . . . . . . . . . . . . . . . . . . 104
3.5 Distribution of Interest Rates . . . . . . . . . . . . . . . . . . 105
3.6 Primary Cash Withdrawal Methods . . . . . . . . . . . . . . 107
v
List of Tables
2.1 Adoption and Usage of Payment Cards . . . . . . . . . . . . 48
2.2 Number of Payment Types by Contactless Credit Card Adopters
per Month . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.3 Number of Payment Types by Contactless Debit Card Adopters
per Month . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.4 Sample Summary Statistics of Credit Card Adopters . . . . . 54
2.5 Sample Summary Statistics of Debit Card Adopters . . . . . 55
2.6 Statistics of Perceived Characteristics . . . . . . . . . . . . . 56
2.7 Logit Propensity Score Marginal Effects . . . . . . . . . . . . 65
2.8 Matching Quality . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.9 Impact of Contactless Payment Cards on the Transaction Ratio 71
2.10 Rosenbaum Bounds Sensitivity Analysis and Significance Test 73
3.1 Adoption and Usage Rate of Payment Cards . . . . . . . . . 90
3.2 Adoption Patterns of Contactless Payment in the Entire Sam-
ple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.3 Cash Measures of Contactless Credit Card Innovators and
Non-Innovators . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.4 Cash Measures of Contactless Debit Card Innovators and Non-
Innovators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.5 Means of Cash Measures for Types of Contactless Credit Card
Adopters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.6 Means of Cash Measures for Types of Contactless Debit Card
Adopters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
vii
3.7 Sample Summary Statistics of Credit Card Innovators and
Non-Innovators . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.8 Sample Summary Statistics of Debit Card Innovators and
Non-Innovators . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.9 Regression Results of Contactless Credit Cards on Cash Value 110
3.10 Regression Results of Contactless Debit Cards on Cash Value 111
3.11 Regression Results of Contactless Payment on Cash Share Vol-
ume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
A1 Means of Demographics for Types of Contactless Credit Card
Adopters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
A2 Means of Demographics for Types of Contactless Debit Card
Adopters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
A3 OLS Regression Results of Contactless Credit on Usual Cash
Withdrawn . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
A4 OLS Regression Results of Contactless Credit on Number of
Withdrawals . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
A5 OLS Regression Results of Contactless Credit on Cash in Wal-
let . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
A6 FE Regression Results of Contactless Credit on Usual Cash
Withdrawn . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
A7 FE Regression Results of Contactless Credit on Number of
Withdrawals . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
A8 FE Regression Results of Contactless Credit on Cash in Wallet 124
A9 OLS Regression Results of Contactless Debit on Usual Cash
Withdrawn . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
A10 OLS Regression Results of Contactless Debit on Number of
Withdrawals . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
A11 OLS Regression Results of Contactless Debit on Cash in Wal-
let . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
A12 FE Regression Results of Contactless Debit on Usual Cash
Withdrawn . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
A13 FE Regression Results of Contactless Debit on Number of
Withdrawals . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
viii
A14 FE Regression Results of Contactless Debit on Cash in Wallet 130
A15 OLS Regression Results of Contactless Credit on Cash Share
Volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
A16 OLS Regression Results of Contactless Debit on Cash Share
Volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
A17 FE Regression Results of Contactless Credit on Cash Share
Volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
A18 FE Regression Results of Contactless Debit on Cash Share
Volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.1 Usage of Mobile Payment on an Annual Basis . . . . . . . . 154
4.2 Adoption Rates of POS Payment Instruments . . . . . . . . 154
4.3 Differences in Adoption Rates of POS Payment Instruments 155
4.4 Adoption Rates of Payment Portfolios . . . . . . . . . . . . . 155
4.5 Differences in Adoption Rates of Payment Portfolios . . . . . 156
4.6 Number of Transactions per Month by Payment Instrument
and Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
4.7 Differences of Transactions per Month by Payment Instrument
and Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.8 Payment Choice Frequencies in the Sample . . . . . . . . . . 159
4.9 Sample Summary Statistics . . . . . . . . . . . . . . . . . . . 160
4.10 Conditional Logit Estimates: Adoption Stage . . . . . . . . . 164
4.11 Conditional Logit Estimates: Adoption Stage (Cont.) . . . . 165
4.12 Adoption Stage: Average Marginal Effect of Mobile Payment 167
4.13 Nested Logit Estimates Usage Stage: POS Payments . . . . 169
4.14 Nested Logit Estimates Usage Stage: Retail Payments . . . . 170
4.15 Nested Logit Estimates Usage Stage: Services Payments . . 171
4.16 Usage Stage: AME of Mobile Payment for POS payments . . 173
4.17 Usage Stage: AME of Mobile Payment for Retail Payments . 174
4.18 Usage Stage: AME of Mobile Payment for Services Payments 174
A1 Assessment of Payment Instrument Characteristics . . . . . 181
ix
Abstract
This thesis examines the impact of payment innovations such as contactless
and mobile payment on individual payment behavior. It consists of three
empirical studies in the field of payment economics that fill the gap in this
relatively new field. Unique data sets on consumer payment choice provided
by the Federal Reserve Bank of Boston are analyzed. In addition, the intro-
ductory chapter offers a comprehensive overview of the various strands of
research on the factors driving consumer payment behavior. It also extends
literature by providing a theoretical model of individual payment behavior
that draws together the reviewed literature.
In the first study, the effect of contactless payment on the transaction
ratio for different transaction types at the stationary point-of-sale (POS)
is estimated. The specific devices that are investigated are debit and credit
cards, to which the feature is embedded. Using propensity score matching to
control for selection, the estimation shows that the contactless feature yields
a statistically significant increase in the transaction ratio at the POS for both
payment methods. The average treatment effect on the treated for credit and
debit cards is roughly 8% and 10%, respectively. These findings indicate
that the private industry can benefit from the innovation with respect to
additional revenue streams.
The second study explores the impact of contactless payment on cash
usage in terms of value spending and transaction frequency at the POS. The
specific devices that are investigated are debit and credit cards, to which
the feature is embedded. Employing cross-sectional estimation methods, the
estimation shows that contactless payment leads to a statistically significant
reduction in average cash usage at the POS in terms of value and volume.
The negative effect of contactless credit and debit cards on cash volume is 5%
and 6%, respectively. The negative impact of contactless credit cards on cash
value is estimated between 12% and 16%, but no effect is found for contactless
debit cards. Using the fixed-effects model, however, results in a negative
effect of 3% for contactless debit cards and a statistically insignificant effect
for contactless credit cards on cash volume. The results obtained on cash
value are unaffected.
xi
The third study focuses on the effect of mobile payment on the adoption
and usage patterns of traditional payment instruments such as cash, checks,
credit, debit, and prepaid cards used at the POS. Using discrete-choice ran-
dom utility models to simulate consumer behavior, the estimation provides
two major findings. First, pertaining to the adoption stage, mobile payment
does not replace physical payment cards, but is likely to substitute paper-
based payment methods such as cash and checks. Second, mobile payment
does not statistically significantly influence the choice of payment means at
the POS in terms of usage. However, there is suggestive evidence that it
is complementary to card payments and a substitute for paper-based pay-
ment instruments. The findings highlight the potential social welfare gains
of mobile payment.
xii
Zusammenfassung (Summary in German)
Diese Dissertation untersucht den Einfluss von Innovationen im Zahlungsver-
kehr wie kontaktloses und mobiles Bezahlen auf das individuelle Zahlungsver-
halten. Sie beinhaltet drei empirische Untersuchungen im Bereich der Zah-
lungsokonomie, welche die Lucke in diesem relativ neuen Forschungsfeld
schliessen. Es werden neuartige Datensatze uber die Wahl der Zahlungsmeth-
ode von Konsumenten analysiert, welche von der Bundesbank von Boston zur
Verfugung gestellt werden. Zusatzlich bietet das einfuhrende Kapitel einen
umfassenden Uberblick uber die verschiedenen Forschungsgebiete derjenigen
Faktoren, welche das Zahlungsverhalten der Konsumenten beeinflussen. Es
erweitert ausserdem die Literatur dahingehend, dass es ein theoretisches
Modell uber das individuelle Zahlungsverhalten zeigt, welches aus der rezen-
sierten Literatur hergeleitet wurde.
In der ersten Studie wird der Effekt des kontaktlosen Bezahlens auf
die Anzahl Zahlungstransaktionen verschiedener Transaktionsarten am sta-
tionaren Verkaufspunkt (VP) geschatzt. Dabei werden Debit- und Kred-
itkarten untersucht, in welche die Kontaktlos-Funktion eingebaut ist. Es
wird die Methodik des“Propensity Score Matching”angewendet, um die Selb-
stselektion zu kontrollieren. Die Schatzresultate zeigen, dass die Kontaktlos-
Funktion zu einer statistisch signifikanten Erhohung der Zahlungstransaktio-
nen am VP fur beide Zahlungskarten fuhrt. Der durchschnittliche Treatment-
Effekt ist ungefahr 8% fur Kredit- und 10% fur Debitkarten. Diese Resultate
zeigen, dass die Privatwirtschaft von der Innovation hinsichtlich zusatzlicher
Einnahmequellen profitiert.
Die zweite Studie exploriert die Auswirkungen des kontaktlosen Bezahlens
auf die Bargeldausgaben und -transaktionen am VP. Dabei werden Debit-
und Kreditkarten untersucht, in welche die Kontaktlos-Funktion eingebaut
ist. Die Schatzung ergibt bei Verwendung von Querschnittsdaten, dass die
Kontaktlos-Funktion zu einer statistisch signifikanten Reduktion des Bargeld-
gebrauchs hinsichtlich Ausgaben und Anzahl Transaktionen am VP fuhrt.
Der negative Effekt von kontaktlosen Kredit- und Debitkarten auf die An-
zahl Bargeldtransaktionen ist 5% bzw. 6%. Der negative Einfluss von kontak-
xiii
tlosen Kreditkarten auf Bargeldausgaben wird auf zwischen 12% und 16%
geschatzt, wahrend kein Effekt fur Debitkarten gefunden wird. Bei Ver-
wendung des “fixed-effects” Modells resultiert ein negativer Effekt von 3%
fur kontaktlose Debit- und kein statistisch signifikanter Effekt von kontak-
tlosen Kreditkarten auf die Anzahl Bargeldtransaktionen. Der Einfluss auf
die Bargeldausgaben bleibt unverandert.
Die dritte Studie fokussiert den Effekt des mobilen Bezahlens auf die
Adoption und den Gebrauch von herkommlichen Zahlungsmittel, welche am
VP eingesetzt werden. Es wird das“Discrete-Choice Random Utility”Modell
angewendet, um das Konsumentenverhalten zu simulieren. Die Schatzung of-
fenbart dabei zwei wesentliche Erkenntnisse: Erstens, bezogen auf die Stufe
Adoption, ersetzt mobiles Bezahlen nicht die physischen Zahlungskarten,
aber es substituiert die auf Papier basierten Zahlungsmethoden wie Bargeld
und Checks. Zweitens beeinflusst mobiles Bezahlen auf Stufe Einsatz nicht
statistisch signifikant die Wahl der Zahlungsmittel am VP. Es gibt jedoch
suggestive Hinweise, dass es komplementar zu Kartenzahlungen ist und auf
Papier basierte Zahlungsmittel substituiert.
xiv
Chapter 1
Introduction
“Cash is dirty. . .Cash is heavy. . .Cash is inequitable. . .Cash is quaint,
technologically speaking. . .Cash is expensive. . .Cash is obsolete.”
— James Gleick (1951 – present)
1.1 Introduction
Payments are an inherent element of economic activity. However, the evolu-
tion of payment instruments and the way individuals and businesses make
daily payments has undergone a tremendous change in human history, par-
ticulary in last decades due to major innovations in payment systems. In
the early beginning, payments were settled based on barter of goods that
individuals perceived as valuables like pearls or pelts, but soon commod-
ity money in the fashion of coins made out of precious metal such as gold
and silver served as a medium of exchange. During the 18th century, paper
notes gradually emerged as a uniform way of exchange encouraging foreign
trade, which finally led to the system of representative (fiat) money. This is
paper currency as well as non-precious coins that do not have an intrinsic
value. However, until the end of Bretton Woods in 1971, its value was still
redeemable for a specific amount of precious metal such as gold, also known
as the “Gold Standard”.
1
Meanwhile, fiat money is free-floating and allows central banks to print
and mint currency in excess of the actual amount of precious metal on de-
posit. By the 19th century, checks were the first cashless payment mode
that was introduced followed by payment cards (credit and debit cards) in
the mid 20th century. These methods enable an intangible and time-delayed
exchange of money.1 Payment cards are now the most prominent cashless
payment methods in developed countries, which, for instance, accounted for
more than a half of total consumer payments in the U.S. in 2013 and around
a third in France in 2011, respectively (see Schuh and Stavins (2015b) for
the U.S. and Bagnall et al. (2016) for France as well as for an international
overview).
The transfer of money between payers and payees when organizing eco-
nomic activity does not come without costs. The institutional framework
of payment systems hereby ascertains a significant amount of costs of pay-
ments. These are referred to as transaction costs, the concept coined by
Coase (1937) and Williamson (1985) in the context of their transaction cost
theory. The historical evolution of payment instruments has documented
that each innovation in payment markets has emerged in purpose of facili-
tating the transfer of value and making economic interactions more efficient,
thus minimizing operation and transaction costs of purchases, which overall
determine the social welfare costs of payments. Meanwhile, the latest gen-
eration of innovative payment instruments such as contactless and mobile
payment has been launched taking advantage of technological enhancements
in data communication, which tend to further improve payment efficiency
by reducing transactions costs and simultaneously foster electronic ways of
paying.
Contactless payment is based on the near-field communication (NFC)
technology, which is a standard radio communication technology that allows
to connect devices within a four centimeter range by waving or tapping the
objects without providing a signature or PIN for verification. The feature
is usually embedded in conventional payment cards, but also in other de-
vices such as mobile phones and key fobs. Contactless payment cards allow
1See Judt (2006) for a historical summary of the evolution of payment modes.
2
making instantaneous payment transactions by just waving the card over
the card reader. Mobile payment is referred to as any payment that is au-
thorized, initiated or confirmed through a mobile device (Au and Kauffman,
2008). Usually, traditional payment methods such as debit and credit cards
determine the payment process and settlement of payment underneath. How-
ever, other forms such as bank account deduction or charge of phone bills is
commonly practicable. Both payment modes enable a new form of payment
initiation that transfers value electronically.
This thesis examines the impact of payment innovations such as contact-
less and mobile payment on individual payment behavior with respect to
transaction motives, cash balances, and payment choices. It centres on the
latest payment innovations in retail payment markets focusing on proximity
payments at the stationary point-of-sale (POS) that account for the vast ma-
jority of consumer transactions compared to remote payments (Schuh and
Stavins, 2015b). Retail payments are the main drivers behind economic and
financial activities. Hence, their smooth functioning as well as understanding
their transformation process is essential for overall financial stability.
The motivation for this thesis derives from both a practical and theo-
retical challenge and it may be relevant for the following important reasons.
Providing and promoting efficient payment methods and systems is a key
element to underpin the sound operation of the economy, which in turn af-
fects private and social costs of all stakeholders involved in the ecosystem of
payments. Thus, the smooth functioning of payments is particularly perti-
nent to central banks, financial intermediaries, retailers, and consumers alike.
There are a series of insightful studies that analyze the private and social
costs structures of retail payment methods (see van Hove (2008), Schmiedel
et al. (2013) and references therein). A recent study by Schmiedel et al.
(2013) estimates the societal costs of retail payment instruments close to one
percent of the gross domestic product (GDP) in the EU-27 countries, whilst
the number of cash transactions contributes to almost half of the total social
welfare costs of payments. Correspondingly, striving for higher social costs
efficiency by stimulating electronic ways of paying is high on the agenda
of stakeholders in the payments ecosystem which, by contrast, may create
economic surplus.
3
Central banks are specifically interested in the impact of innovative pay-
ment products that spur the immediacy and convenience of transfers since
payment innovations tend to negatively affect the interest elasticity of money
demand (Alvarez and Lippi, 2009). This could hamper the efficacy of mone-
tary policy and may challenge the optimal provision of money, which deter-
mines a distinctive amount of central banks’ seignorage revenue. In addition,
it may result in decreasing welfare costs of inflation.2
Comprehending the effect of payment innovations is further relevant for
policy makers and government agencies alike since payment innovations may
pose new regulatory, policy, and liability issues. The rationale is that an
increasing number of non-banks attempts to expand in the area of consumer
mobile payments and may thus challenge the prevailing regulatory frame-
work. Moreover, the optimal regulation of interchange fees is a necessary
price mechanism in the two-sided markets of retail payments to ensure the
balance between supply and demand (cf. Weiner and Wright, 2005; Rochet
and Wright, 2010), which is ascertained by the number of payment card
transactions and cash purchases at the POS, amongst others, since they
incur different costs on merchants.
Striving to understand the impact of contactless and mobile payment on
individual payment behavior is crucial because the findings may ease the
decision-making process in light of managerial, promotional, strategic, as
well as investment issues of retailers and financial intermediaries. They could
help to reap the full gains of payment innovations regarding revenue streams
and consumer satisfaction. The aim of this thesis is hence to provide answers
to fundamental questions arising from recent trends in payment markets in
order to better understand consumer attitudes and behavior in the context
of different stakeholders in the payments ecosystem.
The rest of the introduction chapter expands on a review of existing
literature on determinants influencing individual payment behavior followed
by the evolvement of a theoretical model of payment behavior based on
reviewed literature. Subsequently, the main research objectives, the central
2Welfare costs of inflation are the amount of less seigniorage revenue due to highernominal interest rates (real rate plus expected inflation) (Briglevics and Schuh, 2013).
4
research topics, and the contribution of this thesis to extant literature is
discussed. Finally, the structure of this thesis is outlined.
1.2 Determinants of Payment Behavior
There has been a nascent body of academic literature in payment economics
in recent years that examines the determinants of individual payment behav-
ior.3 The majority of studies are empirical using self-reported survey data
that aim to track daily payments (see Bagnall et al. (2016) and references
therein). This approach in payment research is most common due to the
nature of cash payments, which feature anonymity and untraceability. Few
exceptional empirical studies rely on transactional-level data provided by
stores or financial intermediaries (e.g. Cohen and Rysman, 2013; Agarwal
et al., 2010; Klee, 2008; Rysman, 2007). Others offer theoretical models and
guidance on payment behavior that predominately relate to cash manage-
ment (e.g. Alvarez and Lippi, 2015, 2009; Bouhdaoui and Bounie, 2012).4
Another interesting avenue of research infers consumer payment selection
from the field of behavioral economics, where individual payment behavior
is analyzed in experimental settings (e.g. Prelec and Semester, 2001; Soman,
2003; Raghubir and Srivastava, 2008; Thomas et al., 2011).
The aim of this section is to provide a literature review to draw together
the various strands of research on the factors driving consumer payment
behavior. The first subsection lays out the theoretical background of tech-
nology adoption and usage pertaining to payment innovations. In this sense,
it is worth noting that the sequential decisions at the extensive and intensive
margin of payment products may differ. Second, payment instrument char-
acteristics that affect payment behavior followed by physical determinants
of consumer payment choice are presented. Finally, psychological aspects of
payment behavior are discussed.
3See Kahn and Roberds (2009) for a survey of the growing literature in paymenteconomics.
4There is also an important strand of theoretical research in payment economics fo-cusing on optimal competition and pricing of payment card network interchange fees (e.g.Rochet and Wright, 2010; Evans and Schmalensee, 2005; Rochet, 2003).
5
1.2.1 Models of Technology Acceptance
The basic theoretical framework of this thesis hinges on models explaining
technology acceptance, which aim at determining the adoption and usage
conditions of innovations. In order to gain a better understanding of con-
tactless and mobile payment behavioral patterns, three models, which are
most tailored to the research question, are henceforth presented in more
detail.
Innovation Diffusion Theory
The theory of the diffusion of innovations (IDT) is a popular model elab-
orated by Rogers (2003) that explains the proliferation of innovations in
societies. It consists of a micro and macro process which both are closely
related. While the former describes the decision-making process of individu-
als to adopt or reject innovations dependent on prior conditions and stages
(see Figure 1.1), the latter illustrates the diffusion of innovations that spread
through societies over time. It departs from the notion that the adoption of
new technologies typically follows a ‘S’-curve, based on the initial individual
decision process. That is, when a new technology is launched, it is only
adopted by a tiny share of so-called “innovators” followed by an increasing
number of “early adopters” as the time proceeds. Subsequently, gradually
increasing and finally reaching the tipping point, its adoption accelerates
until the stage of maturity. Finally, the process slows down and converges
to the level of saturation – the period when the so-called “laggards” get on.
Taking the micro perspective, Rogers (2003) distinguishes five consecu-
tive stages in the individual decision-making process of adoption, namely
(1) Knowledge, (2) Persuasion, (3) Decision, (4) Implementation, and (5)
Confirmation (see Figure 1.1). First, individuals confront new technologies
in the Knowledge stage while being predetermined by prior conditions such
as previous practice, problems and needs, innovativeness, norms of the so-
cial system, and their own characteristics, i.e. socioeconomic characteristics,
personality variables, and communication behavior. Subsequently, based on
this background, they shape their opinions and attitudes towards the inno-
6
vation in the Persuasion stage, where five perceived characteristics of the
innovation affect the adoption or rejection of a technology in the Decision
stage. These include Relative Advantage, Complexity, Trialability, Compati-
bility, and Observability. Observability relates to the fact whether the usage
and output of an innovation is observable while compatibility refers to the
degree of perceived consistency of the innovation with existing values and
past experiences of adopters. The first three of these constructs are most
closely linked to the intention to adopt contactless and mobile payment (e.g.
Leong et al., 2013; Mallat, 2007; Kim et al., 2010). In the forth stage – the
Implementation stage –, individuals employ the innovation, gain experience
and form attitudes towards usefulness. Finally, they decide to continue or
discontinue the use of the innovation in the last stage of Confirmation.
Source: Rogers (2003)
Figure 1.1: Innovation Diffusion Theory
According to the percentage value of contactless and mobile payment
adopters in the U.S. (see data sets used in sections 2.3.2, 3.5.2 and 4.5.2),
it is likely that the rate of adoption is still in its infancy and thus the dif-
fusion process at the very early stage. However, high penetration rates of
electronic payment products and widespread payment mode acceptance do
not necessarily guarantee a high rate of usage, as cash still accounts for a
7
significant amount of transactions in developed countries (see Bagnall et al.,
2016). This suggests that individuals use a mixture of different payment
modes, which their rate of diffusion cannot only explain. Thus, a number
of other factors also play an important role in describing payment behavior,
which is discussed in the following.
Technology Acceptance Model
The first version of the Technology Acceptance Model (TAM) proposed by
Davis (1989) and later refined by Davis et al. (1989) is one of the first tech-
nology acceptance models (see Figure 1.2). It suggests that the Attitude
Toward Using a technology is influenced by two key measures, namely Per-
ceived Usefulness and Perceived Ease of Use, which both are a function of
external variables encompassing technology features and end-user involve-
ment (see Figure 1.2). The Perceived Usefulness captures the subjective
belief that using a particular technology would enhance the individual’s job
performance, whereas Perceived Ease of Use measures the “degree to which
a person believes that using a particular system would be free from effort”.
(Davis, 1989, p. 320). Both concepts are very similar or identical to the con-
structs of relative advantage, complexity, and trialability in the previous IDT.
Furthermore, the Behavioral Intention to Use depends on Perceived Useful-
ness and Attitude Toward Using, which finally forms the decision towards
the actual usage of the technology.
Multiple studies have attempted to link specific characteristics of contact-
less and mobile payment to the two core constructs of Perceived Usefulness
and Perceived Ease of Use. In general, it has been found that improved con-
venience of paying, enhanced efficiency by reducing transaction costs, and
better record keeping are subject to Perceived Usefulness (e.g. Wang, 2008;
Mallat, 2007).
Unified Theory of Acceptance and Use of Technology
The model of Unified Theory of Acceptance and Use of Technology (UTAUT)
is an extended version of the prevailing TAM and IDT (among others). Tests
have shown that it is superior to these models in predicting the intention and
8
Source: Davis et al. (1989)
Figure 1.2: Technology Acceptance Model
use behavior of innovations (Venkatesh et al., 2003). The model, developed
by Venkatesh et al. (2003), entails four core constructs and four moderating
factors (see Figure 1.3). Performance Expectancy (PE), Effort Expectancy
(EE), and Social Influence (SI) directly influence the behavioral intention,
whereas Facilitating Conditions (FC) has a direct impact on use behavior.
These factors are thereby moderated by Gender, Age, Experience, and Vol-
untariness of Use.
PE is very similar to the concepts of Perceived Usefulness in the TAM
and Relative Advantage in the IDT and was found to be the best predictor
in behavioral intention (Venkatesh et al., 2003). According to Yu (2012), PE
can best be operationalized by key measures such as anticipated efficiency,
usefulness, convenience, and time savings. Studies conclude that these are
positively related to contactless and mobile payment (Liebana-Cabanillas
et al., 2014; Kim et al., 2010). Therefore, contactless and mobile payment
tend to foster electronic ways of paying. PE is particularly salient for men
because they are more task-oriented (Venkatesh et al., 2003).
The constructs of Perceived Ease of Use and Complexity of the previous
models are identical to EE and are directly measured by questions about
the difficulty of learning, interacting and becoming more skillful with new
technologies (Yu, 2012). Increasing age thereby moderates the processing
of complex stimuli concluding that younger cohorts are more attracted to
payment innovations (Venkatesh et al., 2003).
9
SI influences the intention to use innovations in a complex manner and is
associated with the way in which individuals believe other people view them
as a result of having used the technology (Venkatesh et al., 2003). That
is, an individual’s belief structure is formed in response to social conditions
such as social pressure and social status gains. The reliance on others’ opin-
ions is thereby most important in the early stage of technology adoption.
Venkatesh et al. (2003) find that SI is more salient for women and elderly
since they are more sensitive to others’ opinions and have higher affiliation
needs, respectively.
FC captures aspects of the technological and organizational environment
that supports the use of a technology. However, the effect becomes statis-
tically insignificant when incorporating EE because it largely captures the
ease of technology appliance, which is related to the FC construct (Venkatesh
et al., 2003).
Source: Venkatesh et al. (2003)
Figure 1.3: Unified Theory of Acceptance and Use of Technology
There are numerous studies that have tested the presented models above
as well as extensions and modifications in the context of innovative payment
methods, particularly with respect to mobile payment (e.g. Chen, 2008; Yang
et al., 2012). In sum, they find that these models predict the adoption and
use of payment innovations rather precisely to the extent that especially ease
10
of use, trust and security, usefulness, costs, and compatibility are among the
most important factors (see Dahlberg et al. (2008, p. 174) and Dahlberg
et al. (2015, p. 274) for a synopsis of constructs used to study consumer
payment instrument adoption including links to respective references).
1.2.2 Payment Instrument Attributes
As the technology acceptance models all have highlighted, perceived char-
acteristics of innovations are the main drivers for their actual adoption and
use. This has been empirically tested in the context of payment instruments
in various studies. In fact, perceived payment instrument attributes are one
of the crucial factors that explain consumer payment behavior and impor-
tantly augment the sociodemographic determinants (e.g. Ching and Hayashi,
2010; Schuh and Stavins, 2010; Klee, 2008). For instance, consumers with
higher preferences for time savings and convenience are more prone to use
electronic payment modes compared to paper-based counterparts (e.g. Klee,
2006; Borzekowski and Kiser, 2008), whereas the desire for anonymity and
budget control stimulate the use of cash (e.g. Hernandez et al., 2016; Arango
et al., 2015a; von Kalckreuth et al., 2014). In the following, a number of im-
portant payment instrument characteristics that influence payment behavior
are discussed more in-depth, while the focus lies on a broader context than
just the consumer’s point of view.
Acceptance
Obviously, it is necessary that consumers first have to accept and adopt, re-
spectively, new payment instruments following the mechanisms described in
the technology acceptance models in section 1.2.1 before they can actually
employ new payment technologies. However, due to the nature of retail pay-
ment markets, which feature network externalities and two-sidedness, con-
sumers will only adopt and use a new payment instrument if a minimum
number of others using it because its value increases with the number of
other users.5 Yet this threshold, referred to as “critical mass”, depends on
5This is clearly not subject to cash since it is ubiquitously accepted owing to legalrequirements.
11
the number of acceptance points of payment instruments and is difficult to
achieve by virtue of the two-sidedness of retail payment markets inherent
between consumers and merchants. This is, consumers will only adopt a spe-
cific payment instrument if it is accepted by a sufficient number of retailers.
Conversely, merchants will only accept it if the share of consumers exceeds
the critical level. This dilemma is often referred to as the “chicken-and-
egg problem”.6 To overcome the respective dilemma, a common approach
departs from the fact of pricing both sides differently while the least price-
sensitive side usually subsidizes the other one to reach a critical mass (Bolt
and Chakravorti, 2012). This aligns with the fact that payment card is-
suers offer rewards to consumers funded by the interchange fee passed on
merchants for the purpose of promoting the adoption and usage of cards
(Arango et al., 2015b). In this sense, merchants typically incur higher costs
in terms of transaction fees than consumers who only face a fixed fee for
adopting a payment instrument.7
Additionally, the decision to launch and accept payment innovations is
typically based on the economic concepts of economies of scale and scope.
They play an important role for the supply-side sector because they deter-
mine costs structures and hence revenue. Providing a dense payment network
incurs high fixed investment costs, but the marginal costs of processing an
additional transaction is distinctively small, meaning that average costs grad-
ually converge to zero as the number of transactions increase, referring to as
economies of scale. Moreover, regarding economies of scope, average costs
per transaction decrease as the number of different payment instruments in-
creases using the same payment network and infrastructure, respectively. As
a consequence, the decision among retailers and financial intermediaries to
invest in new payment methods and new infrastructure strongly hinges on
the possibility to process a significant share of transactions to realize strong
scale and scope economies (cf. Bolt and Chakravorti, 2012). Contactless and
partly mobile payment employs the same payment network of payment cards
6For more information on two-sided markets, see, for instance, Hoppli et al. (2011).7For a theoretical analysis of optimal payment network pricing, see, for instance, Ro-
chet and Wright (2010) and Rochet (2003) as well as Chakravorti (2010) for a review ofthe respective literature.
12
at the POS, but requires the latest payment terminals that allow initiating
payments wirelessly. Therefore, it is likely that payment providers and re-
tailers benefit from increasing economies of scale and scope when relying on
contactless and mobile payment.
There are several empirical studies that portray the importance of net-
work externalities and two-sidedness in terms of payment behavior (e.g. Cam-
era et al., 2016; Arango et al., 2015b; Rysman, 2007). For example, Jonker
(2007) reports fewer card payments with a decreasing degree of urbanization
of individuals’ living environment where card acceptance is less widespread.
In a similar vein, Rysman (2007) finds evidence that a denser payment card
network among merchants positively affects consumers’ payment behavior.
Other studies focusing on payment infrastructure find a strong effect of the
number of automated teller machines (ATMs), POS terminals, and bank
branches on the use of payment cards and cash (e.g. Drehmann et al., 2004;
Amromin and Chakravorti, 2009; Humphrey et al., 1996; Huynh et al., 2014;
Lippi and Secchi, 2009). However, Camera et al. (2016) provide experimen-
tal evidence in a laboratory setting that network externalities are not highly
prevalent in retail transactions.
Costs
Several different costs components of payment instruments largely determine
the decisions and behavior of stakeholders involved in the payment ecosys-
tem, whereby each single payment instrument incurs different amounts of
costs due to its specific characteristics. Enabling to transfer and receive re-
tail payments involves private costs for market participants in the payment
chain such as financial intermediaries, retailers, central banks, and consumers
alike. The private costs of each participant for a single payment instrument
equals the sum of external and internal costs, where the former consists of
fees and tariffs paid to other market participants and the latter includes re-
sources used and costs incurred by themselves, respectively (Schmiedel et al.,
2013).
Central banks, for instance, face costs for the physical production of
cash, its storage and delivery, as well as preventing counterfeiting. Banks
13
encounter private costs for operating ATMs, processing electronic payments,
preventing fraud and losses of payment cards and cash handling, whereas
retailers’ costs components entail costs for POS terminals, fraud and secu-
rity, as well as transaction fees for payment cards, and transaction costs
for handling cash, amongst others. Finally, consumers’ costs elements, for
instance, include the risk of losing and holding payment instruments, transac-
tion costs for undertaking payments and withdrawals (so-called“shoe-leather
costs”) and learning, search, and switching costs in seeking to change pay-
ment products. Consumers also pay fees for the adoption of payment cards,
interest rates on revolving debt, and penalty fees, amongst others.8
Each payment instrument – due to its features – further differs in its costs
structures for the society as a whole. The social costs of a single payment
instrument equal the sum of all internal costs incurred by the relevant partic-
ipants. They basically reflect the pure production costs to the economy, as
any payments between participants such as fees and tariffs are not included
because they represent costs for the one and revenue for the other (Schmiedel
et al., 2013). A number of research studies provides insights into the social
welfare costs of retail payments and concludes that they can amount up
to one percent of GDP (e.g. Schmiedel et al. (2013) and many references
therein).
Yet the total social costs of a payment instrument not only depend on its
characteristics, but also on features of the country-specific payment market,
i.e. the number of POS terminals, and the maturity and size of non-cash
payments (economies of scale). Therefore, the unit social costs of a payment
instrument – the social costs per transaction in terms of volume and value
– provide a better measure to compare the costs of payment modes. It has
been found that in some European countries cash incurs the lowest unit
social costs in terms of volume while in others debit cards are the cheapest.
The unit social costs in terms of value (costs per Euro of sales) are, however,
very similar for cash and payment cards (Schmiedel et al., 2013). This does
not necessarily imply that cash is most cost-efficient since its unit social costs
depend on the number as well as value of transactions, where the former is
usually very high.
8See Scholnick et al. (2008) and references therein for an explanation of (optimal)credit card pricing schemes.
14
For this reason, in order to assess the most cost-efficient payment instru-
ment for society, Jonker (2013) proposes to consider the costs of one single
additional transaction at a specific transaction size. The computation of
these variable costs enables the determination of the so-called “break-even”
transaction amount; the level where the costs of two payment instruments
are equal. In other words, one payment instrument may be cheaper for values
lower than the break-even point while the other may be more economical for
larger transaction amounts. Various studies find evidence that debit cards
are socially more cost-efficient than cash except for very low transaction
values (e.g. Jonker, 2013).9
From the individual perspective, however, market participants prefer to
employ those payment instruments that incur the lowest private costs for
them, which are not necessarily most cost-efficient for society. The discrep-
ancy between private and social costs may finally result in an overuse of
socially inefficient payment modes. As consumers are usually not confronted
with any transaction fees and charges for every transaction they make, they
are typically unaware of the social costs of each payment instrument. This
costs intransparency, which is primarily evoked by the strong interconnect-
edness of all payment market participants, has led to an overuse of cash by
consumers, which, from the social perspective, is the most cost-inefficient
payment mode (cf. Schmiedel et al., 2013).
In light of this, several attempts have been made to steer consumers away
from using socially cost-inefficient cash towards more cost-efficient electronic
payments, for example by launching electronic payment innovations such as
contactless and mobile payment and by introducing and removing transac-
tion fees for ATM withdrawals and debit card surcharges, respectively. There
is a series of influential empirical studies that examine the price and financial
incentive responses to consumer payment choice. In sum, consumers are very
price sensitive in the sense that they shift to electronic payments in disfavor
of paper-based transactions if every single transaction is directly priced or if
payment card transactions are made less expensive (Camera et al., 2016; Bolt
9The break-even point of cash and debit card payments in the Netherlands in 2009,for instance, is estimated at 3.06 EUR (Jonker, 2013).
15
et al., 2008; Jonker, 2007). They also respond elastically to debit card trans-
action fees that lead to a decline in debit card usage and to the substitution
for alternative payment instruments such as cash, checks and credit cards
(Camera et al., 2016; Borzekowski et al., 2008; Bolt et al., 2010; Koulayev
et al., 2012; Stavins, 2011).
Other studies underline the importance of loyalty programs and other fi-
nancial incentives such as card discounts and interest-free periods, amongst
others, that positively affect the use of payment cards at the expense of cash
(Arango et al., 2015b; Carbo-Valverde and Linares-Zegarra, 2011; Ching and
Hayashi, 2010; Simon et al., 2010). They also lead to increased spending be-
havior (Agarwal et al., 2010). Because payment transactions are not charged
differently according to the form of payment in order to recoup the costs of
merchant fees and card rewards, cash users (non-card users) implicitly sub-
sidize individuals who use credit cards (Schuh et al., 2010).10 Overall, it is
generally acknowledged that a complete transformation to a cashless society
yields social costs savings, which can amount to up to 0.6 percent of GDP
(Humphrey et al., 2001).
Security
Another important aspect of consumer payment choice refers to perceived
security and safety incidents of payment instruments, which has become an
important topic on the agenda of central banks and financial intermediaries
in previous years due to an increasing number of electronic payments pro-
cessing. The rationale is that new types of safety risks such as fraud and
identity theft associated with electronic payment instruments have emerged –
besides already existing security concerns like the risk of loss, theft, and coun-
terfeiting regarding paper-based payment methods. Thus, because payment
fraud may render individuals to use less cost-efficient forms of payments and
payment innovations, understanding the security risks of payment instru-
ments and their impact on payment choice is crucial to improve social costs
efficiency (Kosse, 2013b).
10Schuh et al. (2010) argue that cash users generate an implicit monetary transferto credit card users that amounts to 149 USD for cash-using payers and 1133 USD forcard-using payees per year.
16
The research on consumers’ perceptions of payment instrument security
and safety incidents on the impact of payment choices is very limited and
does not reach an unanimous conclusion (Kosse, 2013b). A few theoreti-
cal studies incorporate safety issues related to the risk of theft and to a
safe-keeping role for banks into monetary theory (He et al., 2008; Alvarez
and Lippi, 2009). He et al. (2008) implicitly assume that the risk of cash
theft is beneficial for the safety of payment cards while Alvarez and Lippi
(2009) conclude that individuals decrease money holdings and increase cash
withdrawals as the probability of cash theft increases.
A number of studies explores the effect of security concerns and safety
risks on the adoption and usage of payment means from the empirical point
of view. There is evidence that the perception of safety is a predominant
determinant of consumer payment choice (e.g. Arango et al., 2015b; Arango
and Taylor, 2009; Jonker, 2007; Borzekowski et al., 2008; Kosse, 2013b). For
instance, Arango and Taylor (2009) show that payment cards are employed
more frequently if consumers view them as less risky in terms of fraud, theft,
or counterfeiting compared to cash. Similar findings are provided by Arango
et al. (2015b). Kosse (2013b) finds that cash is less likely used at the POS if
consumers perceive it as unsafe and rather switch to debit cards. Conversely,
they use debit cards less frequently if they are dissatisfied about its safety.
Thereby, the perception of safety is largely influenced by views on the like-
lihood of possible safety incidents that may occur when using or holding a
payment instrument, which in turn is ascertained by personal characteristics
and past experiences. However, Stavins (2013) argues that once a payment
mode is adopted, the effect of the rating of security on actual use of payment
instruments such as debit cards becomes statistically insignificant, whereas
the reverse is found for more established forms of payment such as cash,
check and credit cards.
Moreover, the risk of fraud and identity theft negatively affects the usage
of payment instruments (e.g. Kosse, 2013a; Kahn and Linares-Zegarra, 2015;
Humphrey et al., 1996). For example, Humphrey et al. (1996) exhibit a
negative correlation between the rate of crime at country level and cash and
debit card usage. Kahn and Linares-Zegarra (2015) point out that certain
17
types of identity theft incidents have a positive impact on the adoption of
credit cards and stored-value cards, among others, and positively affect the
use of credit cards and payment modes not directly linked to bank accounts
(e.g. cash, money orders). By contrast, identity theft victims decrease the
usage of payment instruments related to bank account information such as
checks. Consumers also decline debit card usage after being confronted with
skimming fraud newspaper articles (Kosse, 2013a).
However, other studies such as Ching and Hayashi (2010), Schuh and
Stavins (2010) and Schuh and Stavins (2015a) find no compelling evidence
of security perceptions playing an essential role in consumer payment choice
both in terms of adoption and use.
1.2.3 Physical Factors
According to the theoretical framework of technology acceptance in section
1.2.1, consumer payment choice also hinges on environmental conditions as
well as socioeconomic characteristics, apart from payment instrument at-
tributes. There is a bulk of empirical studies dealing with these physical
determinants. Overall, the studies conclude that the choice of payment meth-
ods is a function of personal, transactional, and situational factors.
First, the payment literature reveals that sociodemographic characteris-
tics are pertinent indicators in explaining payment choice (e.g. Hernandez
et al., 2016; Schuh and Stavins, 2010; Borzekowski et al., 2008; Stavins, 2001;
Connolly and Stavins, 2015). Correspondingly, younger, more educated co-
horts with higher incomes use electronic payment modes more frequently
whereas elderly, less educated individuals with lower incomes are more likely
to use cash and other paper-based payment methods. One rationale is that
more educated, high income individuals face higher opportunity costs when
settling paper-based payments, which generally take a greater amount of
time (Humphrey et al., 2001; Polasik et al., 2013). In addition, race and
foreign background matters. For instance, Kosse and Jansen (2013) show
that foreign backgrounds affect payment choice after migration, whereby
cash is still most preferred by migrants from cash-oriented countries. Race
has been found to be strongly correlated with payment instrument use while
18
individual payment behavior only slightly evolves over time with respect to
demographics (Connolly and Stavins, 2015).
Second, transaction characteristics such as transaction size and the type
of good purchased are major factors in predicting payment choice at the POS.
In particular, the higher the transaction amount, the more likely consumers
pay by payment cards whilst cash is dominated for small-value purchases
(e.g. Arango et al., 2015b; von Kalckreuth et al., 2014; Klee, 2008; Bounie
and Francois, 2006). Also, the individual share of cash in transactions is
influenced by the type of purchase (von Kalckreuth et al., 2009). As op-
posed to these findings, Bouhdaoui and Bounie (2012) and in a similar vein
Eschelbach and Schmidt (2013) argue that payment choice is predominately
driven by the outstanding cash balance hold in wallets than by transaction
size.
Third, situational aspects and the physical location are relevant deter-
minants for payment selection. For example, the type of merchant as well
as the physical attributes of the POS such as the absence of a cashier influ-
ence payment selection, whereby the likelihood of cash payments increases
if payment venues are unattended (Jonker, 2007; Bounie and Francois, 2006;
Hayashi and Klee, 2003).
1.2.4 Psychological Aspects
Another approach infers consumer payment behavior from the psychological
and behavioral point of view, where payment behavior is viewed as a function
of payment modes, meaning that behavioral payment patterns follow an
intrinsic stimulus. This contrasts the perspective of the previous theoretical
and empirical setting above where the choice of payment means, i.e. payment
behavior, is a function of payment instrument characteristics and physical
determinants, implying that payment attitudes are influenced by extrinsic
determinants. However, as individuals only slightly change their payment
patterns over time suggesting a predominant inertia in payment composition
although a wide array of payment instruments with all their merits and
drawbacks is available (cf. Connolly and Stavins, 2015; Humphrey et al.,
1996), alternative explanations for payment method selection not directly
19
linked to rational decision-making and in light of behavioral economics have
been proposed.
The literature on this stream of research in the context of payment eco-
nomics is scarce and mainly derives from consumer and marketing research.
It builds on the notion that individuals have cognitive and emotional asso-
ciations with payment modes that in turn influence payment method selec-
tion, which hence impacts payment behavior (e.g. Khan, 2011; van der Horst
and Matthijsen, 2013). Overall, there is evidence that consumers’ emotions
related to payment instruments translate into automatic behavior, whereby
almost half of the payment modes selected is habitual (Leenheer et al., 2012).
This aligns with findings in other fields of research, where it is acknowledged
that half of consumers’ behavior is a product of habits (see Wood and Neal
(2009) and references therein). In the following, the most salient theoretical
concepts in explaining behavioral patterns of payment modes selection are
first discussed.11 Second, the results of some important empirical studies in
the context of irrational payment behavior are presented.
Theoretical Concepts
An important theoretical explanation of payment behavior derives from the
concept of “mental accounting”, which has evolved from the more general
“framing paradigm”. A frame serves as a point of reference that is formed
by mental filters influenced by biological and cultural characteristics (Khan,
2011). Individuals thereby rely on these frames to comprehend and react to
situations whilst it can create a cognitive bias that limits behavior depending
on how the situation is presented (e.g. as a loss or a gain).12 Elaborating on
this paradigm, Thaler (1985) established the concept of mental accounting
to describe consumer behavior, which basically represents “the set of cog-
nitive operations used by individuals and households to organize, evaluate,
and keep track of financial activities.” (Thaler, 1999, p. 183). These cog-
11Note that the presented theories, though distinctly presented, are interrelated andthus have a common explanatory base.
12For instance, a transaction is perceived more positively when an identical amount ofmoney is framed as “pennies a day” (one Dollar a day) compared to the aggregate amount(365 Dollar a year) (Gourville, 1998).
20
nitive operations are affected by a preset frame of references and anchors,
respectively (Thaler, 1999).13
According to Thaler (1999), mental accounting consists of three com-
ponents: a mental cost-benefit analysis of expenditures, the frequency of
account balancing, and the assignment of activities to specific mental ac-
counts. The first component describes how outcomes are perceived and how
decisions are made and evaluated. It basically reflects the awareness of the
benefits associated with the purchase and the price paid.14 The second com-
ponent relates to the frequency with which mental accounts are evaluated,
for instance, to gain an idea on how much it was spent and left in savings
in the specific mental account, because individuals are aware of restricted
monetary resources. The third component comprises the assignment of ac-
tivities to specific mental accounts such that expenditures are categorized
(e.g. necessities, housing, holiday) while being constrained by explicit and
implicit budgets (Thaler, 1999). In other words, each payment will be as-
signed to a specific mental account that correlates with financial accounts
linked to different payment methods. Consumers therefore place payments
to their preferable financial account based on the expense’s associated men-
tal account and the cost-benefit relationship. For instance, small purchases
may be preferably paid in cash that immediately gets the payment out of
the way, based on the assumption that it is placed in the perishable goods
account that offers only temporary utility. Mental accounting thereby helps
to reduce cognitive efforts to assess financial decision-making.
Based on this notion, Heath and Soll (1996) document that consumers
indeed mentally allocate their money for specific spending purposes, which
13The concept of“anchoring”lies within the framing paradigm and is termed by Tverskyand Kahnemann (1974). It implies that individuals’ decision-making is biased towards theadjustment and interpretation of other information around the anchor, which reflects theinitial point mostly based on the first piece of information offered in order to ease decision-making. It is thus completely feasible that anchoring is related to payment modes choicegiven the evolution of money representation that could have elicited different informationand responses once individuals were confronted with these money tokens at an early age.
14For instance, assume there are items of three different sizes with prices increasingwith size. Now they are all on sale for a price lower than the initial prices. An individualbuys the biggest item, though not perfectly suitable, and is quite pleased with the purchase(see Thaler, 1999).
21
also determines their purchase behavior. However, because they are not able
to perfectly anticipate consumption and money in one mental account is not
fungible, false budgeting leads them either to over- or underconsumption
of goods. By contrast, Soman (2001) proposes an alternative explanation
for purchase behavior in the domain of mental accounting that is different
from the budgeting model in the sense that past expenses are retrospectively
evaluated to guide current spending rather than prospectively and later re-
trieved at the time of purchase. The theory suggests that consumers know
about their budget balance based on the information of past expenses, which
serves as a reference point for current spending. The recall of past payments
is thereby facilitated by the transparency of payment modes – e.g. embod-
ied by cash (see below) – which is in turn additionally enhanced by the
requirement of writing down the amount paid, and when consumers’ wealth
is depleted immediately rather than with a delay (Soman, 2001). Corre-
spondingly, the intention to purchase is largely reduced by the probability
of accurate recall. For instance, credit cards are least transparent since they
decouple payment from purchase in terms of time and tight path tracking
and hence cause misinterpretation of past expenditures. The concept of
“coupling”, the direct mental linkage of consumption and payment, is an-
other important theoretical aspect of payment behavior coined by Prelec
and Loewenstein (1998).
Similar to the theory of mental accounting is the concept of “prospective
accounting” advanced by Prelec and Loewenstein (1998). They argue that
consumption that has been paid in advance (prepaid) lessens the intensity
of the act of payment, which in turn positively affects consumers’ evaluation
of the net benefit of the purchase. This is, consumption can be enjoyed
as if it were free while the pain associated with the act of paying prior to
consumption is attenuated by thinking about prospective benefits of the
purchase. In a more general context, Prelec (2009) refers to this as a “moral
tax”reflecting the psychological intrusion of payments into consumption that
reduces consumers’ hedonic value. For instance, consumers may have more
pleasure in enjoying a free meal than the identical meal for a cost, apart
from the pecuniary costs. Because of the present moral tax, Prelec (2009)
22
proposes different strategies that structure payments in a way to minimize
hedonic loss: prepayment, buffer currencies (token currencies such as casino
chips), frequency programs (e.g. frequent flyer miles), financial decoupling,
and fixed payment plans and subscriptions. It is worth noting that Prelec
(2009) interprets financial decoupling as similar to the experience of credit
card payments, but he argues that it differs in the way that it also accounts
for the strength and clarity of the causal link between the purchase and
the payment (apart from the decoupling effect in form of the time interval).
Therefore, decoupling impacts the degree of payment acuteness.15
Another behavioral aspect in payment choice infers from the “pain of
paying” – a concept introduced by Zellermayer (1996). It refers to the emo-
tions individuals experience in parting with money and complies with the
degree of annoyance from the act of making a payment (Zellermayer, 1996).
The pain of paying undermines the pleasure of consumption and its severity
is thereby influenced by the payment methods used. For instance, Soman
(2001) exhibits that consumers experience less pain in paying by credit cards
rather than checks and are therefore more willing to spend. Conversely,
Ariely and Silva (2002) demonstrate that the pain of paying substantially
affects payment and consumption. However, as Zellermayer (1996) argues,
payment mode selection is likely to be highly routine and driven by habits
rather then the immediacy of pain and pleasure associated with payments.16
The pain of paying highly and positively correlates with the transparency
of payment instruments that is defined as the “relative salience of the pay-
ment, both in terms of physical form and the amount, relative to paying by
cash”. (Soman, 2003, p. 175). Cash provides the most transparent form of
payment because it is tangible and visible enabling to instantly record what
amount is being transferred. In contrast, card-based payment products offer
15From this point of view, all-inclusive pricing is preferable to a la carte pricing (Prelec,2009).
16It is well noticeable that the pain of paying enhances self-regulation, but is hedonicallycostly meaning that there is a trade-off between hedonic and decision efficiency (Prelecand Loewenstein, 1998). The former relates to the ideal notion that payments ought tobe tightly linked to consumption, i.e. the act of payment is attenuated by the financedbenefits, while consumption in turn needs to be decoupled from payments, i.e. having thefeeling of “free” consumption. The latter refers to the fact that consumers actually wantto know what they are paying for (Prelec and Loewenstein, 1998).
23
low transparency whereas it is even lower with other electronic and mobile
payment formats (Soman, 2003). According to Soman (2003), the lower
the degree of transparency, the more the payer is willing to spend. This
is because the intransparency of payment products hampers the tallying of
expenditures.
In essence, payment methods are associated with different perceptions,
which are shaped by physical characteristics of the payment instrument re-
lated to visual representation, accessibility to money and historical associ-
ations (Khan, 2011). Perceptions affect payment mode selection and thus
purchase behavior. According to van der Horst and Matthijsen (2013), for
instance, cash triggers more positive emotions relative to debit cards, from
which both payment means provoke automatic payment behavior, whereas
the effect of cash is stronger. Since payment choices are highly habitual-
ized decisions, behavioral concepts such as “nudging”, termed by Thaler and
Sunstein (2008), have been proposed to alter habitualized payment behavior.
“Nudges” are soft interventions that subtly change the choice context and
harness non-intrusive influences (Thaler and Sunstein, 2008). They are not
related to pecuniary incentives and are governed by choice architecture. In
doing this, van Hove (2009) suggests altering the default option in payment
systems and to increase the“hassle factor”by making certain payments more
inconvenient to steer consumers towards cost-efficient payment means. He
proposes, for instance, to change the denomination of cash (e.g. eliminate
high denomination, remove small coins, issue less pocket friendly coins), re-
duce the number of ATMs, change the standard withdrawal amounts at
ATMs, and improve the visibility of POS terminals, amongst others. In a
field experiment, Aydogan and van Hove (2015) show that consumers indeed
increase the usage of payment cards due to the exposure of pro-card posters
at the cashier, but the intervention has not proved to be sustainable in the
sense that consumers soon revert to old habits.
Empirical Findings
In light of the theoretical explanations of payment behavior discussed above,
a number of influential empirical studies examine the nexus of payment
24
modes and purchase behavior, which are mostly based on field or laboratory
experiments. Apart from mere economic assumptions, there is compelling
evidence that the format of money itself, i.e. the physicality of payment
instruments, as well as the decoupling of payment and consumption signifi-
cantly impacts purchase behavior (e.g. Runnemark et al., 2015; Chatterjee
and Rose, 2012; Prelec and Semester, 2001; Soman, 2001; Feinberg, 1986).
The bulk of research in this field predominantly focuses on the difference be-
tween credit cards and cash (e.g. Prelec and Semester, 2001; Incekara-Hafalir
and Loewenstein, 2009; Raghubir and Srivastava, 2008; Thomas et al., 2011),
whereas other studies compare debit cards and cash (e.g. Runnemark et al.,
2015; Khan, 2011). Overall, they find that actual expenses and the willing-
ness to pay are higher when using credit and debit cards relative to cash,
referring to as the“credit card premium” in the former case. Research in this
field has also been conducted related to gift certificates and prepaid cards
(e.g. Raghubir and Srivastava, 2008; Soman, 2001, 2003), suggesting that
these money formats elicit an increase in spending compared to cash.17
The first study analyzing the relationship between payment instruments
and spending has been conducted by Hirschman (1979) based on consumer
shopping survey data, revealing that the possession of a credit card is posi-
tively related to higher expenditures. Later, Feinberg (1986) observes that
credit card paying consumers leave larger tips than cash payers. In a sub-
sequent experimental study, he shows that consumers are willing to spend
50–200 percent more for products by merely exposing them to credit card
paraphernalia, i.e. confronting them with a credit card logo on the desk un-
related to the task. This is in line with Prelec and Semester (2001) who find
a credit card premium of roughly 100 percent, improving upon methodologi-
cal issues of previous studies. In contrast, Incekara-Hafalir and Loewenstein
(2009) suggest that the effect of the credit card premium is only salient for
convenience users, whereas revolvers spend less. The effect of the credit card
premium is, however, not solely attributable to the decoupling from the pain
17There are also studies in the field of psychology that explain payment behavior in thecontext of pricing mechanisms (e.g. fixed and variable pricing) and social behavior (e.g.Gneezy et al., 2010; Jung et al., 2014; Xu et al., 2015).
25
of payment and the physical form, but also to the salience of product bene-
fits enhanced by positive perceptions and evaluations evoked by credit cards
(Chatterjee and Rose, 2012). Accordingly, consumers primed to use credit
cards focus more on product benefits, whereas cash primed consumers tend
to consider the costs of a product.
Consumers paying by debit and credit cards are more impulsive and
likely to buy unhealthy food products than when they pay by cash (Thomas
et al., 2011). Thomas et al. (2011) argue that the same effect both for debit
and credit cards is separately observable due to the salience of the payment
form, i.e. the degree of the pain of paying, in lieu of temporal decoupling
of payment settlement. Additionally, durable goods are more inclined to be
paid by credit cards than perishable items because they provide a continuing
flow of utility enabling to better match payment and consumption streams
(Prelec and Loewenstein, 1998).
Furthermore, denomination of cash matters. A large denomination (e.g.
one 20 Dollar note) is likely to reduce spending intention relative to many
small denominations such as 20 one Dollar bills (Mishra et al., 2006; Raghubir
and Srivastava, 2009). This “denomination effect” occurs since large denom-
inations are mentally less fungible than smaller ones, which in turn helps to
control spending more precisely (Raghubir and Srivastava, 2009). Another
rationale provides Mishra et al. (2006) for the “bias for the whole” – a simi-
lar concept to the denomination effect – arguing that a large denomination
is perceived as higher value due to greater fluency experienced in payment
processing compared to many small denominations. There is also evidence
that consumers tend to make more purchases of small value when carrying
coins relative to absent coins (Vandoros, 2013). Smaller amounts of cash in
banknotes are further preferred to higher values in coins, conditional on a
small amount of money (Vandoros, 2013).
26
1.3 Theoretical Model of Payment Behavior
This section aims at establishing a theoretical model of individual payment
behavior using an integrated framework that draws on the main empirical
findings and theoretical concepts discussed in the previous section 1.2. The
presented theoretical model does not seek to replicate the conceptual models
of technology adoption from section 1.2.1 by adding other important factors
(e.g. habit) to the well-established adoption and diffusion theories. In the
first step, it rather pursues to incorporate the relevant factors of payment
behavior in a utility function. In the second step, an integrated framework
is developed where consumers’ selection of payment methods in relation to
habitualized payment behavior is analyzed.
The model departs from the assumption of a utility maximizing individ-
ual. It abstracts from the theoretical underpinning that goods and services
purchased provide utility while the act of paying creates disutility for the
payer. Rather, it focuses on the level of utility a payment alternative pro-
vides under given conditions. Formally, there is an individual i who faces a
choice among J payment alternatives for transaction s while each alternative
provides a certain level of utility. Uijs, j = 1, . . . , J is the perceived utility
level which an individual i experiences when choosing payment method j for
transaction s. The utility function Uijs is denoted as
Uijs = U(Xi, CHi, RCj , TCj , TSs), (1.1)
where Xi are demographic characteristics as a set of proxies for opportunity
costs and CHi refers to the outstanding amount of cash holdings in wallet for
all i. RCj denotes relative perceived characteristics of payment instrument
j such as security, setup, acceptance, records, convenience (e.g. ease of use),
as well as financial costs (e.g. fees, surcharges, loyalty programs), amongst
others. TCj represents the transactional costs of payment instrument j
related to efficiency (e.g. speed). TSs refers to the size of a transaction s.
In line with utility maximizing behavior, individual i selects payment
alternative j at the POS yielding the highest utility for transaction s such
that
27
Uijs(·) > Uiks(·) (1.2)
for all k 6= j holds. The difference between the utility levels of payment
alternative j and k is referred to as the net utility NUijks with positive
values such that
NUijks = Uijs(·)− Uiks(·) > 0 (1.3)
must hold. Apart from the mere utility driven decision, the habitualization
of payment behavior for consumer payment choice is essential. The degree
of habitualization hi, i.e. the unit of payment habit, can be formulated as
hi = h(ais, pj , dj) (1.4)
where ais refers to the “mental account” of individual i where payment s is
placed. pj and dj denotes the degree of the “pain of paying” and the degree
of “decoupling” payment and consumption related to payment instrument
j, respectively. There is also a predetermined degree of individual payment
automatism reflected by a constant c, which leads to the stylized function of
habitualization H(hi) formulated as
H(hi) = c+ h2i . (1.5)
Figure 1.4 combines these two approaches of consumer payment choice,
illustrating the relationship between preference-based utility and habitual-
ization. It displays the function of habitualization H(hi), where complete
habitualization of payment behavior is at the point hi = H on the horizontal
axis. The vertical axis represents the perceived net utility NUijks, where the
dashed line marks the reference point in which the individual is indifferent
between two payment alternatives, i.e. Uijs(·) = Uiks(·). Above the dashed
line, the individual chooses payment alternative j over k. However, this may
not be tenable if the consumer’s degree of payment habitualization hi is too
high, meaning that he remains with his habitual payment patterns. As a
consequence, individual i changes payment behavior towards the usage of an
28
alternative payment instrument j relative to k for transaction s, represented
by the indicator variable I, such that:
I =
{1 if NUijks > H(hi)
0 otherwise.(1.6)
In other words, H(hi) indicates the behavioral change frontier meaning that
with decreasing payment habits, switching to an alternative payment method
becomes more likely, holding net utility constant. At the point H, payment
behavior is completely habitual.
Source: Own Figure based on Blume et al. (2015)
Figure 1.4: Theoretical Model of Payment Behavior
The presented theoretical framework of payment behavior serves as a
starting point for the thesis’ main research question to what extent payment
innovations impact individual payment behavior. The analysis departs from
29
the underlying assumption that the shift towards new payment behavior
primarily emerges from the change in perceived utility of payment methods
while payment habitualization is time invariant and only slightly changes
over time. This comes at the advantage that payment habits are generally
cumbersome to detect and hence hard to capture and measure. Therefore,
the analysis of the thesis’ main research question focuses on the concept of
perceived utility of payment instruments.
It is assumed that consumers have preferences for specific payment instru-
ments related to attributes that they value differently. Improved incentives
of payment instruments such as lower financial and transaction costs, better
security and ease of use, amongst others, tend to foster perceived added value.
Payment innovations such as contactless and mobile payment are inclined
to particularly lower transaction costs that may enhance perceived utility.
Thus, if consumers perceive the value-added of contactless and mobile pay-
ment with respect to transaction costs as sufficient, utility increases, holding
all else constant. Consequently, individual i switches to an alternative, more
efficient payment method if
Uijs(TCj) > Uiks(TCk). (1.7)
Accordingly, if the perceived utility of payment instrument j with respect to
transaction costs TCj is greater than the corresponding utility of payment
method k, individual i selects the more efficient payment instrument j, hold-
ing all else constant. This is basically the theoretical concept underneath the
empirical analysis that is implicitly tested in the presented thesis. It also
implies that net utility NUijks increases. However, in case of taking into
account the aspect of habitualized payment behavior, individual i changes
to an alternative payment instrument j only if
Uijs(TCj)− Uiks(TCk) > c+ h2i . (1.8)
30
Rearranging the inequation leads to
hi → 0 : Uijs(TCj) > c+ Uiks(TCk), (1.9)
hi → H : Uijs(TCj) > c+ h2i + Uiks(TCk). (1.10)
This implies that under these conditions, the perceived utility of payment
innovation with respect to transaction efficiency must be more pronounced
compared to the case of absent payment habitualization in order to induce
consumers to switch payment instruments, even in the case when automa-
tism is very low (h → 0). However, if behavioral payment patterns are
dominating (h → H), the perceived utility of payment innovation with re-
spect to transaction efficiency has to be even higher to the prior case inducing
individuals to change payment patterns.
To conclude, the model shows that consumers must encounter clear incen-
tives of payment instruments, for instance, such as lower transaction costs
that elicit a substantial increase in perceived utility and hence net utility
in order to change payment behavior relative to their old habits. Accord-
ing to the model, they do not select new payment modes if the positive
change in net utility, particulary with respect to perceived characteristics,
is insufficient, given a certain degree of payment habitualization. This is
the principal theoretical framework that serves as the reference point in the
empirical studies presented within this thesis.
1.4 Research Topics and Objectives
Despite the wide array of research in the field of payment economics (see, for
instance, section 1.2), there are still some under-exposed areas in payments
literature that await further investigation. In particular, it exists only limited
knowledge and unsettled empirical evidence on the effects of the latest pay-
ment innovations in retail payment markets such as contactless and mobile
payment on individual payment patterns. The main objective of this thesis
is therefore to enhance the understanding of the impact of contactless and
mobile payment on individual payment behavior at the stationary POS. By
31
empirically exploring these under-researched topics analyzing unique data
sets on consumer payment choice, this thesis fills the gap in payment eco-
nomics providing novel knowledge and new insights on whether there is room
for a further shift towards the use of electronic payments. Comprehending
these mechanisms is a relevant topic since stimulating electronic ways of
paying fosters social costs efficiency of overall payment systems.
Specifically, the thesis centres on the research topics including the role
of contactless payment on transaction motives and cash usage as well as
the role of mobile payment on payment instrument choice. These payment
innovations tend to decrease the transaction costs of paying, which in turn
may influence existing payment behavior by increasing the perceived utility
that induces individuals to switch payment methods. The thesis examines
whether contactless payment affects the number of payment card transac-
tions and cash usage in terms of volume and value of transactions. It also
analyzes the impact of mobile payment on payment mode selection hypoth-
esizing a change in the usage of traditional payment instruments such as
cash, checks, credit, debit, and prepaid cards. In the following, the role of
contactless and mobile payment in payment economics is outlined providing
a clear picture of where payments literature in this context is lacking.
1.4.1 Contactless Payment and Transaction Frequency
The payments literature has so far paid little attention to the relationship
between contactless payment and individual transaction behavior, which re-
sults in a poor understanding of these dynamics. The major rationale is
that contactless payment is an innovative way of paying having launched
and established only recently in the payment landscape. It is therefore a
neglected topic that has not yet been approached in payments research with
respect to transaction motives. More insights are therefore desired to gain
a better understanding of how transaction behavior is influenced by new
forms of paying. This may be especially of great interest for retailers, finan-
cial intermediaries, central banks, and policy makers alike, of which a few
may benefit from increased electronic payment processing raising additional
revenue streams.
32
There is an influential stream of research that highlights the importance
of the representation of money on purchase behavior, suggesting that actual
expenses and the willingness to pay are higher when using credit and debit
cards relative to cash (e.g. Runnemark et al., 2015; Raghubir and Srivastava,
2008; Soman, 2003; Prelec and Semester, 2001). It is argued that psycho-
logical factors related to the format of money influence spending motives
(see section 1.2.4). However, as contactless payment represent a feature em-
bedded in conventional payment cards – it can also be embedded in other
devices –, it is suggested that the effect on payment behavior may arguably
be attributable to higher transaction costs efficiency inferred from the possi-
bility to pay instantly. This is because checkout time has been found to be
an important determinant for the choice of payment means (cf. Schuh and
Stavins, 2010; Jonker, 2007; Klee, 2006; Borzekowski and Kiser, 2008). Ana-
lyzing transactions data of Mastercard customers using Mastercard contact-
less payment cards has revealed an increased usage of payment cards both
in terms of value spending and transaction frequency (Mastercard, 2013).
However, this research tends to be biased since it is restricted to Mastercard
customers only. It is also of a purely descriptive nature and incapable of es-
tablishing causality. For this reason, it is essential to provide more objective
research to gain key insights into transaction motives of contactless payment
adopters.
1.4.2 Contactless Payment and Cash Usage
Similar to the status quo in payments literature regarding contactless pay-
ment and transaction behavior, the literature on contactless payment and
cash usage is scant, albeit increasing in recent years. First, this is similarly
attributed to the relatively new existence of contactless payment in retail
payments markets. Second, there are only limited empirical data sets that
provide information both on the adoption and use of contactless payment and
cash, which exactly report on the number and expenses of cash payments.
This is because cash payments are anonymous and not directly traceable.
Therefore, they have to be studied by self-reported surveys that are usually
costs- and time-consuming to conduct.
33
Nevertheless, there are a few studies dealing with the effect of contact-
less payment on cash demand. Using household survey data in Japan, Fujiki
and Tanaka (2014) find no reduction in average cash balances due to con-
tactless payment. By contrast, Fung et al. (2014) observe a statistically sig-
nificant negative effect of contactless payment on cash usage both in terms
of value and volume analyzing individual-level survey data in Canada. How-
ever, these empirical studies lack an appropriate identification strategy due
to data restrictions and also fail to consider long-term effects on cash usage.
Chen et al. (2017) enhance previous identification strategies by using house-
hold panel data in Canada, but they unfortunately confront a high attrition
rate, which leave the results questionable. They find no statistically signifi-
cant impact of contactless credit cards on cash usage both in terms of value
and volume.
It is therefore still unclear and inconclusive what effect contactless pay-
ment exerts on cash usage and deserves further investigation and clarification.
To this end, this thesis serves as an effort to unravel the mechanisms be-
tween contactless payment and cash usage by advancing identification using
a unique balanced panel data set. Comprehending the relationship between
contactless payment and cash usage is relevant for having answers to the
question whether payment innovation may steer consumers away from using
socially cost-inefficient cash, which is still the most prominent payment in-
strument at the POS in numerous countries (e.g. von Kalckreuth et al., 2014;
Bouhdaoui and Bounie, 2012; Arango et al., 2015a; Bagnall et al., 2016).
1.4.3 Mobile Payment and Payment Choice
Analyzing the effect of mobile payment on consumer payment choice has not
attracted much academic interests yet, thus leaving it an entirely uninves-
tigated issue. This has already been pointed out in an extended literature
review on mobile payment research in Dahlberg et al. (2008) and most re-
cently in an updated version in Dahlberg et al. (2015). One major reason lies
within the unavailability of comprehensive real-world data incurred by the
dynamically evolving mobile payment technology. For this reason, it invites
first-hand information to shed light on this unexplored area of research prov-
34
ing empirical knowledge on the dynamics between mobile and traditional
payment methods. Examining this question is particularly important in or-
der to find ways to further stimulate electronic forms of paying, whereby it is
more interesting, unlike in developing countries, to focus on the effects in de-
veloped countries, as mobile payment is suggested to hardly penetrate these
markets due to their advanced financial markets and sophisticated payment
infrastructure (Dahlberg et al., 2015).
However, a few disruption analysis studies attempt to prospect the prolif-
eration of mobile payment adoption and use, suggesting that card payments
are preferable to mobile payment while the latter becomes a complement
rather than a substitute to traditional payment instruments (cf. Ondrus and
Pigneur, 2005, 2006a, 2006b). Conversely, Garcia-Swartz et al. (2002) es-
timate that mobile payment will negatively affect the use of cash and the
number of credit and debit card payments. Polasik et al. (2013) argue for a
breakthrough of mobile payment due to its superior time efficiency relative
to cash. In sum, scholars reach an unanimous conclusion on the impact of
mobile payment on existing payment methods from a prospective point of
view. Hence, this thesis provides improved knowledge and novel empirical
evidence on this issue.
1.5 Research Contribution
This section outlines the innovative aspect of this thesis by providing an
overview of the remaining chapters, which answer the thesis’ research ques-
tion to what extend payment innovation affects individual payment behavior
on the basis of three self-contained studies. Chapter 2 elaborates on the work
by Trutsch (2014) that was published in a peer-reviewed journal. It has been
modified and updated for the purpose of this thesis. Chapter 3 provides un-
published work. The study in chapter 4 was published as Trutsch (2016)
in a slightly revised version. This thesis is among the first endeavors to
investigate the role of innovative payment instruments such as contactless
and mobile payment on transaction behavior, cash usage, and payment in-
strument choice. It is complementary to the various strands of literature in
35
payment economics and contributes to empirical payments research with re-
gards to latest payment innovations. Hereinafter, the research contribution
of each study is presented in more detail including their methodologies and
data sets used.
1.5.1 Chapter 2: The Impact of Contactless Paymenton Transaction Frequency
In chapter 2, the effect of contactless payment on the transaction ratio for
different transaction types at the POS is estimated. The specific devices that
are investigated are debit and credit cards, to which the feature is embedded
and which are the most popular cashless payment means (cf. Schuh and
Stavins, 2015b). The focus of the analysis specifically lies on retail and
services payments made at the POS, which account for the vast majority of
individual payments (Schuh and Stavins, 2015b).
It is implicitly tested whether contactless payment is more efficient to pay
than conventional debit and credit cards, which in turn leads to the explicit
estimation of whether contactless payment positively affects card usage and
hence transaction frequency. To this end, a comprehensive, representative
individual survey on consumer payment behavior in the U.S. in 2010 is used
to econometrically estimate by means of propensity score matching the effect
of contactless payment on the number of transactions. In doing so, it allows
to purge selection bias inherent in the empirical setting. This research makes
major contributions in the context of financial innovation (e.g. Alvarez and
Lippi, 2009; Amromin and Chakravorti, 2009; von Kalckreuth et al., 2009)
and is related to earlier work that highlights the very importance of payment
efficiency in payment processing (e.g. Jonker, 2007; Klee, 2006; Borzekowski
and Kiser, 2008; Polasik et al., 2013). It provides new insights into the nexus
of one of the latest payment innovations and transaction patterns, which fills
an important gap in literature.
36
1.5.2 Chapter 3: The Impact of Contactless Paymenton Cash Usage
Chapter 3 explores the impact of contactless payment on consumer demand
for cash in terms of value and volume at the POS. The focus lies on debit and
credit cards, to which the feature is embedded and which counts among the
most popular cashless payment means (cf. Schuh and Stavins, 2015b). Since
it has been found that contactless payment is competitive to and under
certain conditions even outperforms cash payments with respect to time
efficiency (Polasik et al., 2013), the study in chapter 3 explicitly estimates
whether contactless payment substitutes cash payments in terms of value
spending and transaction frequency and provides key insights into whether
it fosters electronic ways of paying at the expense of cash.
The novelty of this research is manifold. First, although contactless credit
cards have been subject to a couple of studies regarding cash usage, the
effect of contactless debit cards has remained an unexplored area. However,
debit cards are among the most popular and widespread cashless payment
instruments (cf. Bagnall et al., 2016; Schuh and Stavins, 2015b). This study
therefore provides insights into the relationship between contactless debit
cards and cash usage. Second, the availability of a unique microeconomic
balanced panel data set that is merged from national representative surveys
on consumer payment behavior in the U.S. from 2009 to 2013 is novel in
payments literature. It enables to control for possible layers of endogeneity in
the econometric specifications, which fills an important gap in literature. The
study contributes to existing literature on cash inventory models providing
novel empirical evidence and advances knowledge in payment economics with
regards to latest payment innovation.
1.5.3 Chapter 4: The Impact of Mobile Payment onPayment Choice
The third study, presented in chapter 4, analyzes the impact of mobile pay-
ment on the adoption and usage patterns of traditional payment instruments
such as cash, checks, credit, debit, and prepaid cards used at the POS. Data
37
are drawn from a representative survey on consumer payment choice in the
U.S. in 2012. The effect is econometrically analyzed employing discrete-
choice random utility models on the probability of using conventional pay-
ment instruments to simulate changes in consumer behavior with respect
to the composition of payment instrument portfolios and the instruments’
employment.
The contribution of this study to existing literature comes from the fol-
lowing aspects. First and foremost, it is the first analysis that gauges the
effect of mobile payment on the array of traditional payment instruments
used at the POS. Second, the quality of the data set enables the investiga-
tion of the impact of mobile payment both at the extensive and intensive
margin of conventional payment methods, which advances the examination
strategy. Third, the consumer-level data encourage to employ discrete-choice
models that offer novel evidence on the substitution and complementation
patterns of existing payment means and may hence shed light on potential
market disruptions. The study contributes to the various strands of litera-
ture on consumer payment choice focusing on the most up-to-date payment
device.
1.6 Outline
The remainder of this thesis is as follows. In chapter 2, the impact of con-
tactless payment on transaction frequency is analyzed. Chapter 3 focuses on
the empirical analysis of the impact of contactless payment on cash usage.
The effect of mobile payment on payment choice is investigated in chapter
4. Chapter 5 summarizes and concludes with respect to the thesis’ research
question. Policy implications and the direction for further research are addi-
tionally presented.
38
Chapter 2
The Impact of ContactlessPayment on TransactionFrequency1
Abstract
This paper estimates the effect of contactless payment on the trans-action ratio for different transaction types at the point-of-sale. Thespecific devices that are investigated are debit and credit cards, towhich the feature is embedded. Data are drawn from a national rep-resentative survey on consumer payment behavior in the U.S. in 2010.Using propensity score matching to control for selection, the estima-tion shows that the contactless feature yields a statistically significantincrease in the transaction ratio at the point-of-sale for both paymentmethods. The average treatment effect on the treated for credit anddebit cards is roughly 8 and 10 percent, respectively. These findingsindicate that the private industry can benefit from the innovation withrespect to additional revenue streams. This paper contributes to theexisting literature in payment economics by analyzing one of the mostrecent payment products.
JEL-Classification: C21, D12, D14, O33
Keywords: contactless payment, payment innovation, transaction behavior, creditcards, debit cards
1This article has been published in a modified version as Trutsch, T. (2014), TheImpact of Contactless Payment on Spending, International Journal of Economic Sciences,III, 70–98.
39
2.1 Introduction
The way consumers make daily payments has changed significantly in recent
years due to innovations such as debit, credit, and prepaid cards, online
banking, and mobile payments, to mention a few. By 2010, consumers in
the U.S. have undertaken within a month on average 50 percent of their
transactions by payment cards, 40 percent by paper instruments such as
cash and 9.2 percent by electronic and other instruments (Foster et al., 2013).
Meanwhile, new forms of retail payment innovations have come up among
which contactless payment.2
This paper investigates the impact of contactless payment on individ-
ual transaction frequency for different transaction types at the point-of-sale
(POS). This new form of payment device has mainly been developed by
the private industry sector for revenue purposes. The specific technology
is embedded in the most prominent payment cards and mobile phones. Its
convenience, safety and efficiency, which are expected to be perceived as su-
perior to cash, should support the proliferation of electronic payments and
substitution away from cash, which still accounts for a significant share of
transactions.
Understanding the effect of contactless payment on individual transaction
habits is crucial for three main reasons. First and foremost, there is limited
knowledge on the adoption and usage behavior of the contactless payment
innovation due to its very recent emergence and establishment. Retailers
can use the information for evaluating whether to invest in the most up-to-
date payment terminals in order to have full gains of the newest payment
technologies because an efficient payment process is one of the most crucial
conditions to reduce waiting lines at the counter and consequently a decline
in sales inferring from negative shopping experiences.
2Contactless payment is based on the near-field communication (NFC) technology,which is a standard radio communication technology that allows to connect devices withina four centimeter range by waving or tapping the objects without providing a signature orPIN for verification. The feature is usually embedded in conventional payment cards, butalso in other devices such as mobile phones and key fobs. For instance, contactless creditcards allow making instantaneous payment transactions by just waving the card over thecard reader. The terms “NFC” and “contactless” are used interchangeably in this study.
40
Second, the findings provide information on specific usage and adoption
patterns among cashless payment means, which may be relevant for finan-
cial intermediaries with respect to managerial, promotional, and revenue
purposes. In general, increasing card transactions that they might process
will result in raising revenue streams generated through their fees.
Third, the paper provides information for policy makers with regards to
evaluating and implementing interchange fee regulation for payment cards
(cf. Weiner and Wright, 2005), which is an ongoing issue in several countries
such as the U.S. (Johnson, 2014), Switzerland (Brouzos, 2014), and the Eu-
ropean Union (EP, 2014).3 For instance, more card transactions incur higher
costs on shop owners due to the current interchange fee structure, as it is
demonstrated in Wakamori and Welte (2012). Additionally, Wiechert (2009)
concludes for Swiss retailers that contactless payment increases the payment
costs for retail shops even more dramatically since it would mean the transfer
of low-cost cash payments to cards implying a higher burden of interchange
fees. The cost increase is more accentuated for micro than macro payments.4
However, the provision of an efficient and cheap payment service is crucial
to underpin the sound operation of the economy. This is also highlighted in
the new strategic focus for financial services announced by the president of
the Federal Reserve Bank of Cleveland (Pianalto, 2012), which specifically
considers payment preferences of end consumers when making future deci-
sions about the payment system. Providing such information in this paper
contributes to support the decision-making process.
This paper can be seen as complementary to the strands of literature
in payment economics and makes a contribution in the context of financial
innovation (e.g. Alvarez and Lippi, 2009; Amromin and Chakravorti, 2009;
Drehmann et al., 2004; Humphrey et al., 2001; von Kalckreuth et al., 2009;
Schuh and Stavins, 2010) and may be relevant for the literature in the two-
sided markets as well (e.g. Rysman, 2007; Rochet and Tirole, 2002; Rochet
3I refer to Rochet and Wright (2010), Evans and Schmalensee (2005), Rochet andTirole (2002), and Rochet (2003), among others, for a theoretical consideration of theinterchange fee regulation and to Hoppli et al. (2011) with special focus on Switzerland.
4Avoiding the cost increase for retailers entails growth in sales or reduction in opera-tion costs. If both are not sufficient, an overall card fees reduction or a discount for micropayment transactions is more appropriate (Wiechert, 2009).
41
and Wright, 2010). Although the model in this paper does not account for
price sensitivity and the two-sidedness in terms of merchant decisions, the
study gives insights in the individual adoption and usage of contactless pay-
ment cards under the interchange fee regulation in 2010 from a consumer’s
point of view.5
The topic is also relevant in the context of efficient payment methods.
Checkout time is an important determinant for the choice of payment means
(cf. Schuh and Stavins, 2010; Jonker, 2007). This is highlighted in Klee
(2006) who finds evidence that debit cards are preferred over checks to save
time. Contactless payment allows to pay efficiently and may therefore lead
to higher transaction frequency. For instance, Polasik et al. (2013) find that
contactless payment cards and NFC mobile payments are competitive to
cash payments with respect to time efficiency and under certain conditions
even outperform cash. Borzekowski and Kiser (2008) quantify the effect of
contactless debit cards in the U.S. applying rank-ordered logit models and
prospect an increase in market share of contactless debit cards compared to
cash, checks, and credit cards because merchants can save up to 0.03 USD
per transaction by accepting contactless debit cards, which is exclusively
driven by faster checkout.6
There is substantial literature on the relationship between reward pro-
grams, interest free periods, and use of credit cards, which this paper is
related to since time savings at the checkout are associated with pecuniary
incentives. Participation in loyalty programs and access to interest free pe-
riods are likely to increase credit card use at the expense of alternative
payment methods such as debit cards and cash (Simon et al., 2010; Agarwal
et al., 2010; Ching and Hayashi, 2010; Carbo-Valverde and Linares-Zegarra,
2011; Arango et al., 2011). Similarly, Agarwal et al. (2010) find that rewards
are associated with elevated spending. This paper is also linked to the lit-
erature on the relationship between payment modes and purchase behavior,
5In July 2010, the Dodd-Frank Wall Street Reform was enacted capping interchangefees of debit cards at 0.12 USD per transaction compared to 0.44 USD before the reform(Board of Governors of the Federal Reserve System, 2011). The interchange fee of creditcards was roughly around 3 percent of the transaction amount in 2010 (Visa, 2010).
6With average cost of 0.70 USD per debit card transaction.
42
suggesting that actual expenses and the willingness to pay are higher when
using credit and debit cards relative to cash (e.g. Runnemark et al., 2015;
Raghubir and Srivastava, 2008; Soman, 2003; Prelec and Semester, 2001).
There are also some consumer-side studies conducted by the private in-
dustry sector focusing on contactless payment. For example, Mastercard
(2013) observes an increased usage of Mastercard PayPass-payment cards
both in terms of value spending and transaction frequency.7 This research,
however, tends to be biased because it might serve as a sales argument for
merchants and the data are restricted to Mastercard customers only. This
paper aims to provide more objective research to gain insights in individual
payment habits in the context of retail payment innovations.
The novelty of this study is twofold. On the one hand, due to the very
recent emergence of contactless payment, it exists only limited knowledge
of its effect on individual payment habits. This paper fills the gap in this
relatively new field. On the other hand, using unique, detailed, and represen-
tative individual survey data from the U.S. in 2010 allows to investigate the
effect of contactless payment on the transaction frequency of the most promi-
nent payment cards (credit and debit cards) for different transaction types
(POS payments distinguished by retail and services payments) by applying
propensity score matching to control for selection bias, which is inherent in
this setting. Since the data set encompasses the rating of perceived charac-
teristics such as ease of use, security, speed, setup, and costs of numerous
payment instruments, I also can control for unobserved heterogeneity (cf.
Jonker, 2007; Kim et al., 2006; Ching and Hayashi, 2010).
My empirical analysis yields the following important results. Using the
2010 Survey of Consumer Payment Choice (SCPC) I estimate the impact of
contactless payment on the transaction ratio at the individual level. First,
I find that the average treatment effect on the treated of contactless credit
cards leads to an increase in the transaction ratio of 8.3 percent at the POS
while the effect for retail and services purchases is 4.8 and 3.5 percent, re-
spectively. Second, the average treatment effect on the treated of contactless
debit cards on the transaction ratio at the POS is 10 percent. In terms of re-
7The Mastercard PayPass-payment card is NFC-enabled.
43
tail and services payments the impact both results in 4.5 percent. Sensitivity
analysis shows that the results are robust to unobserved heterogeneity.
The structure of the paper is as follows. Section 2.2 derives the theoretical
framework and section 2.3 describes the data. In section 2.4, I elaborate my
estimation strategy and present the econometric model. Section 2.5 includes
the results of the empirical analysis and section 2.6 concludes.
2.2 Theoretical Considerations
The theoretical background for this study is drawn from technology accep-
tance models, which aim at explaining the adoption and usage conditions of
innovations. There are numerous models that explain technology adoption
and use from different points of view, from which I choose the most tailored
to the research question.8
2.2.1 Technology Acceptance Model
The Technology Acceptance Model (TAM) explains when individuals will
accept and make use of a technology and has originally been applied to pre-
dict end-user acceptance of information systems within organizations. The
model consists of two main technology acceptance measures: Perceived Use-
fulness and Perceived Ease of Use. Davis (1989, p. 320) defines the former
as “the degree to which a person believes that using a particular system
would enhance his or her job performance”. Enhanced efficiency, time sav-
ings, and convenience are subject to Perceived Usefulness, which pertain to
contactless payment (Wang, 2008), and therefore should foster its deploy-
ment. Perceived Ease of Use is specified as “the degree to which a person
believes that using a particular system would be free from effort”. (Davis,
1989, p. 320). Accordingly, contactless payment is more likely to be used if
it is easy to handle.
8Clearly, the theoretical framework of this study is also linked to the transaction coststheory coined by Coase (1937) and Williamson (1985).
44
2.2.2 Innovation Diffusion Theory
The Innovation Diffusion Theory (IDT), developed by Rogers (2003), ex-
plains how and why innovations spread through societies. It basically con-
sists of two interrelated processes, namely the diffusion and adoption process.
The former can be described as a macro process that explains how innova-
tions spread through societies whereas the latter is a micro process focusing
on the individual’s decision-making process of adopting innovations.
The innovation-decision process consists of five consecutive stages: (1)
Knowledge, (2) Persuasion, (3) Decision, (4) Implementation, and (5) Con-
firmation (Rogers, 2003). In the Knowledge stage, the individual learns
about the emergence of an innovation influenced by prior conditions (pre-
vious practice, problems and needs, innovativeness, and norms of the so-
cial system) and by his own characteristics (socioeconomic characteristics,
personality variables, and communication behavior). Thus, some adoption
mechanisms are predetermined. Subsequently, opinions are formed about the
innovation in the Persuasion stage where five innovation characteristics affect
the adoption of innovations: Relative Advantage, Complexity, Trialability,
Compatibility, and Observability (Rogers, 2003). The first three concepts
are similar to the ones in the previous TAM.
Out of these constructs, the first three of them have provided the most
accurate prediction for the intention to use NFC-enabled mobile credit cards
(Leong et al., 2013). With respect to complexity, (mobile) contactless pay-
ment is expected to increase the convenience of payments and therefore usage
by reducing the need for coins and cash in small transactions (Mallat et al.,
2004). In the third stage, the Decision stage, the individual finally chooses
to adopt or reject the innovation based on the former stages.
2.2.3 Unified Theory of Acceptance and Use of Tech-nology
The model of Unified Theory of Acceptance and Use of Technology (UTAUT)
represents an extension of the previous TAM and IDT model – among others
– and explains user intentions and subsequent usage behavior (Venkatesh
45
et al., 2003). The model consists of four key effects and four moderating
factors. While the first three core constructs – Performance Expectancy
(PE), Effort Expectancy (EE), and Social Influence (SI) – directly influence
the behavioral intention, the forth construct – Facilitating Conditions (FC)
– has a direct impact on use behavior. The four remaining factors Gender,
Age, Experience, and Voluntariness of Use thereby moderate the initial key
effects.
Empirical testing has shown that PE, which is similar to Perceived Use-
fulness in the TAM model, is the strongest predictor of intention in the con-
text of the UTAUT. Time savings, usefulness, and convenience are concepts
which measure performance expectancy and are positively related to con-
tactless payment (Yu, 2012). These characteristics should therefore advance
the usage of contactless payment. Gender studies have revealed that PE is
especially salient for men since they tend to be more task-oriented. Also,
age differences determine technology adoption (Venkatesh et al., 2003).
EE is evaluated by questions about the difficulty of learning, interacting
and becoming skillful in applying new technologies (Yu, 2012). Venkatesh
et al. (2003) show that this construct is only significant for users with a
non-existing or low experience level, becoming non-significant over periods
of extended and sustained usage. EE is more salient for women than for men
whereas increasing age is associated with difficulties in processing complex
stimuli (Venkatesh et al., 2003). This implies younger cohorts to be more
prone to contactless payment.
SI suggests that individuals’ behavior is affected by the way in which they
believe other people will view them as a result of having used the technol-
ogy (Venkatesh et al., 2003). Its role in technology acceptance decisions is
complex and influences individuals through three mechanisms: compliance,
internalization, and identification. The latter two intend to alter an individ-
ual’s belief structure and/or to cause an individual to respond to potential
social status gains. The former mechanism causes an individual to alter his
intention in response to social pressure. Positively attributed characteristics
of contactless payment such as transaction speed and convenience positively
alters the individual’s belief structure and hence can positively influence
46
usage. However, the reliance on others’ opinions, i.e. manifested itself in
social pressure, is particularly significant in the early stages of the technol-
ogy experience when individuals are uninformed. This in turn will attenuate
over time since a more instrumental (rather than social) basis will affect the
technology usage due to increased experience (Venkatesh et al., 2003). For
instance, Stavins (2001) finds that the share of other people in a region using
a payment instrument positively correlates with consumers’ usage behavior.
Social Influence is more salient for women regarding the technology accep-
tance decision process since they tend to be more sensitive to others’ opinions.
Moreover, elderly people are more likely to place increased salience on social
influences since they possess higher affiliation needs (Venkatesh et al., 2003).
In conclusion, the adoption and usage of contactless payment is influenced
by various factors that are partly predetermined and therefore it follows a
non-random pattern.
2.3 Data
2.3.1 Source
Data are drawn from the Federal Reserve Bank of Boston that supports the
Consumer Payments Research Center (CPRC), which regularly conducts
the Survey of Consumer Payment Choice (SCPC).9 It is a rich nationally-
representative and publicly-available data set on consumer payment behav-
ior in the U.S. The survey focuses on the adoption and use of nine common
payment instruments including cash.10 Also, the perceptions on method
of payment attributes are questioned and information on demographics is
provided. The data date back to 2010 and were administrated online by
the RAND Corporation, using RAND’s American Life Panel, to a random
sample of 2102 U.S. consumers primarily in October during fall 2010, whose
responses were weighted to represent all U.S. consumers aged 18 years and
older. The reporting unit of the SCPC is an individual consumer in the U.S.
9See Foster et al. (2013) for a comprehensive description of the data.10These include checks, bank account number payments, online banking bill payments,
money order, traveler’s checks, debit, credit, and stored-value cards.
47
The reason to monitor individuals rather than households stems from the
fact that it is unlikely that the head of the household can track the payment
behavior of all household members in detail. However, some information
about each reporting consumer’s household is collected in the survey such
as income. It is worth noting that the estimates are not adjusted for sea-
sonal variation, inflation, or item non-response (missing values). Also, the
tumultuous years after the financial crisis in 2008 accompanied by a severe
recession could have led to unusual reporting of the number of payments.
2.3.2 Description
The survey specifically asks respondents if one of their credit and debit cards
is equipped with the contactless feature, but unfortunately does not provide
exact information on the usage of the technology. Instead, detailed statistics
on the usage of conventional credit and debit cards are available as well
as their adoption rates. Table 2.1 shows the market shares of contactless
and conventional credit and debit cards as well as the corresponding use of
the latter. It reveals that about 9 percent (187 individuals) of the entire
sample of 2084 respondents reported that their credit card is equipped with
the contactless feature, whereas approximately 12 percent (258 individuals)
stated to possess a contactless debit card. In contrast, more than 70 percent
have a conventional credit card and around 78 percent a debit card. Credit
and debit cards are used at least once within a month by 56 and 63 percent
of the people in the sample.
Table 2.1: Adoption and Usage of Payment Cards
Variable Mean NContactless Credit 0.092 2084Contactless Debit 0.124 2084Credit 0.703 2088Debit 0.784 2090Credit Usage 0.568 2059Debit Usage 0.631 2056
Note: Usage describes the fact that respondents make the corresponding type of payment atleast once in a typical month. Survey weights used.
48
To estimate the impact on transaction frequency, I refer to the exact num-
ber of specific card transactions (credit and debit cards) that an individual
has conducted within a typical month distinguished by types of payment at
the POS, i.e. retail goods11 and services.12 Accuracy of reporting was ame-
liorated by asking respondents about the number of payments for a typical
period rather than a specific calendar period. Typical periods shall repre-
sent an implicit average of their perceived regular or trend behavior and have
the advantage of eliminating unusual events that might affect high-frequency
payments and veil longer-run trends (Foster et al., 2013). Also, respondents
are allowed to choose the frequency (week, month or year) that best suits
their recollection of payments for each type of transaction. On the basis of
the responses, the number of payments was calculated for a typical month
and then corrected for invalid data entries. Tables 2.2 and 2.3 provide sum-
mary statistics on the number of transactions of different payment types per
month distinguished by contactless card adopters. Additionally, a simple
mean comparison test (t-test) between non-innovators and innovators is re-
ported showing (significant) differences in the average transaction frequency.
As shown in Table 2.2, contactless credit card adopters undertake around
9 credit card payments more at the POS within a month than non-adopters
(17 vs. 8 transactions) with approximately 5 and 4 transactions more for
retail goods and services, respectively (10 vs. 5 and 7 vs. 3 payments). These
means are significantly different from each other indicating enhancement in
payment frequency for innovators. This holds true also for overall payment
card statistics at the POS. Innovators on average pay 31 times by payment
cards at the POS per month (18 retail and 13 services payments), while non-
innovators conduct around 23 payments (14 retail and 19 services payments).
These mean differences are highly significant. On the contrary, contactless
credit card adopters pay significantly less frequently by cash for services
(roughly 2 payments) than non-innovators.
Table 2.3 distinguishes the number of transactions by contactless debit
11These include items purchased in food and grocery stores, superstores, warehouses,club stores, drug or convenience stores, gas stations, department stores, electronics, hard-ware and appliances stores.
12These include services paid for restaurants, bars, fast food and beverage, transporta-tion and tolls, medical, dental, and fitness, education and child care, personal care (e.g.
49
Table 2.2: Number of Payment Types by Contactless Credit Card Adopters per Month
Non-Innovator Innovator t-Test
Variable Mean Std. Dev. Max. N Mean Std. Dev. Max. N Mean Diff.CC POS 8.36 16.61 117.40 1869 17.10 25.44 108.71 188 −8.67 ***
CC Retail 4.99 10.72 100 1851 9.81 14.83 65.22 188 −4.67 ***CC Services 3.45 7.61 95.66 1849 7.39 12.34 86.96 186 −3.99 ***
DC POS 15.09 23.03 139.14 1868 14.52 24.14 130.45 185 0.67DC Retail 9.22 15.13 108.71 1857 8.59 15.37 86.97 184 0.67DC Services 6.06 10.49 100 1834 5.95 10.58 43.48 185 0.00
SVC POS 0.39 1.81 20 1849 0.21 1.10 12 183 0.18SVC Retail 0.24 1.26 20 1843 0.15 0.87 10 182 0.08SVC Services 0.15 0.77 8.69 1839 0.06 0.31 2 181 0.09*
Overall Card POS 23.61 26.72 165.22 1886 30.77 35.48 173.93 190 −7.82 **Overall Card Retail 14.21 17.57 109.71 1884 17.90 21.27 108.71 190 −3.92 *Overall Card Services 9.41 12.4 105 1884 13.01 16.91 86.96 189 −3.90 **
Cash POS 16.56 19.26 130.45 1841 14.05 17.87 108.71 187 3.01Cash Retail 9.52 12.72 100 1822 8.38 11.65 65.22 185 1.07Cash Services 7.27 9.91 86.96 1813 5.75 8.66 43.48 185 1.95**
Total POS 42.84 38.24 245.5 1893 47.05 42.87 217.41 191 −4.62Total Retail 24.91 24.19 153.19 1893 27.25 26.31 148.84 191 −2.75Total Services 17.93 19.32 158 1893 19.80 20.41 91.73 191 −1.87
Note: Survey weights used. Subcategories do not sum to main category due to rounding and weighting. For brevity, the minimum is droppedbut equals zero for every type of payment. T-test of mean differences of innovator and non-innovator. They can differ from true values due torounding and weighting. Significance levels 1% ***, 5% ** and 10% *. CC represents credit cards, DC debit cards and SVC stored-valuecards. Overall card payments are the sum of CC, DC and SVC payments. Total POS payments are the sum of overall card POS payments,cash POS payments plus check and money order payments.
50
card adopters and non-adopters. Mean comparison tests between adopters
and non-adopters reveal that statistically significant differences in the trans-
action frequency exist. Innovators buy goods and services at the POS by
debit cards more frequently than non-innovators, namely 4 and 6 transac-
tions more within a month (13 vs. 9 and 11 vs. 5 payments, respectively).
Also, their overall card and total POS payments for services exceed those of
non-adopters by 4 and 6 transactions, respectively. In contrast, they trans-
act 5 payments fewer by credit cards at the POS than non-innovators (4 vs.
10 transactions).
To sum up, contactless credit and debit card adopters undertake statis-
tically significantly more transactions by their corresponding payment cards
compared to non-adopters while this also holds true for overall card services
payments.
For the purpose of the analysis, I computed the ratio of credit and debit
card transactions separately to total payments at the POS, which is a more
robust measurement towards outliers.13 The majority of individuals exhibits
a very small transaction ratio both for credit and debit cards (see Figures 2.1
and 2.2) because roughly 31 percent and 35 percent of individuals stated to
have conducted zero credit and debit card payments per month, respectively
(only restricted to those who possess a credit and debit card). This may stem
either from those who did not make any purchases during a typical month or
from those who forgot or refused to report any payments. Thus, the reasons
for reporting zero payments may differ from the determinants of the actual
non-negative integer number of card payments recorded.
The data set also provides rich information about consumer demographic
characteristics and financial status. Tables 2.4 and 2.5 give insights in de-
mographic characteristics and financial status of contactless credit and debit
card holders separately. Obviously, referring to Table 2.4, the sample of
contactless credit card adopters is more skewed towards higher income and
education brackets as well as higher asset shares. For instance, 14 percent
hair), recreation, entertainment and travel, maintenance and repairs, other professionalservices (business, legal etc.), and charitable donations.
13The total number of POS payments encompasses cash, checks, money order, debit,credit, and stored-value card payments.
51
Table 2.3: Number of Payment Types by Contactless Debit Card Adopters per Month
Non-Innovator Innovator t-Test
Variable Mean Std. Dev. Max. N Mean Std. Dev. Max. N Mean Diff.CC POS 9.83 18 117.4 1875 4.37 15.26 108.71 181 5.56***
CC Retail 5.88 11.5 100 1858 2.35 8.59 65.22 180 3.51***CC Services 4.06 8.34 95.66 1854 2.03 7.18 43.48 180 2.06***
DC POS 13.76 22.53 139.14 1876 24.13 25.13 130.45 179 −10.33 ***DC Retail 8.64 15.12 86.96 1865 12.85 14.8 108.71 178 −4.20 ***DC Services 5.29 9.85 100 1843 11.33 13.02 65.22 178 −6.14 ***
SVC POS 0.3 1.51 17.39 1853 0.88 2.94 20 180 −0.48SVC Retail 0.17 0.92 12 1848 0.66 2.37 20 178 −0.40SVC Services 0.13 0.72 8.70 1843 0.22 0.85 4.35 178 −0.08
Overall Card POS 23.6 27.15 173.93 1893 29.04 31.02 173.93 183 −5.25Overall Card Retail 14.4 17.98 108.71 1891 15.65 17.88 109.71 183 −1.08Overall Card Services 9.23 12.38 105 1890 13.39 15.75 86.96 183 −4.16 **
Cash POS 15.96 18.92 130.45 1850 18.98 20.54 108.71 179 −3.11Cash Retail 9.30 12.63 100 1831 10.22 12.57 65.22 177 −0.91Cash Services 6.88 9.64 86.96 1821 8.89 10.78 43.48 178 −2.20
Total POS 42.24 37.96 245.5 1900 50.34 42.95 196.84 184 −8.17Total Retail 24.88 24.43 153.19 1900 26.91 24.14 148.84 184 −1.93Total Services 17.36 18.82 158 1900 23.43 22.61 117.4 184 −6.24 **
Note: Survey weights used. Subcategories do not sum to main category due to rounding and weighting. For brevity, the minimum is droppedbut equals zero for every type of payment. T-test of mean differences of innovator and non-innovator. They can differ from true values due torounding and weighting. Significance levels 1% ***, 5% ** and 10% *. CC represents credit cards, DC debit cards and SVC stored-valuecards. Overall card payments are the sum of CC, DC and SVC payments. Total POS payments are the sum of overall card POS payments,cash POS payments plus check and money order payments.
52
010
20
30
40
Perc
ent
0 .2 .4 .6 .8 1
Credit Card Share
Figure 2.1: Share of Credit Card Payments per Month at the POS
010
20
30
40
Perc
ent
0 .2 .4 .6 .8 1
Debit Card Share
Figure 2.2: Share of Debit Card Payments per Month at the POS
53
of individuals earning 125,000 USD above possess a contactless credit card,
which is also observable for 25 percent of individuals who have completed
some post graduate studies. On average, innovators also withdraw money
less frequently than non-innovators and are mostly male, working, and mar-
ried compared to non-innovators.
Table 2.4: Sample Summary Statistics of Credit Card Adopters
Non-Innovator Innovator
Variable Mean SD Min. Max. N Mean SD Min. Max. NIncome (in 1000)
<25 0.26 0 1 1890 0.13 0 1 19025–49 0.28 0 1 1890 0.22 0 1 19050–74 0.21 0 1 1890 0.26 0 1 19075–99 0.11 0 1 1890 0.19 0 1 190100–124 0.08 0 1 1890 0.07 0 1 190>125 0.07 0 1 1890 0.14 0 1 190
Education<High School 0.05 0 1 1893 0.08 0 1 191High School 0.4 0 1 1893 0.28 0 1 191Some College 0.29 0 1 1893 0.23 0 1 191College 0.15 0 1 1893 0.17 0 1 191Post Graduate 0.11 0 1 1893 0.25 0 1 191
EmploymentWorking 0.62 0 1 1893 0.70 0 1 191Retired 0.19 0 1 1893 0.18 0 1 191Unemployed 0.1 0 1 1893 0.06 0 1 191
Marital StatusSingle 0.2 0 1 1893 0.08 0 1 191Married 0.62 0 1 1893 0.77 0 1 191
OthersMale 0.48 0 1 1893 0.57 0 1 191Age 46.59 16.82 18 109 1893 45.24 15.7 21 88 191HH Members 1.4 1.56 0 9 1893 1.08 1.22 0 5 191Assets 1.31 8.21 0 100 1807 1.54 8.19 0 78 183Cash WD 6.15 12.31 0 434.82 1885 3.74 3.67 0 30.44 191
Note: Survey weights used. Subcategories do not sum to main category due to rounding andweighting. Cash withdrawals (WD) per month. Assets (in 1000) do not include primary home.
Contrarily, the sample of contactless debit card adopters is more skewed
towards the lower income and education brackets as well as lower wealth
status, as highlighted in Table 2.5. Approximately 32 percent of innovators
earn less than 25000 USD and around 40 percent graduated from high school.
54
Furthermore, they are mostly male, working, younger, and single compared
to non-innovators. Also, they withdraw cash around twice as much as non-
innovators (10 vs. 5 withdrawals). This reflects higher preferences for out-
of-the-way rather than credit payments, which cash and debit cards can
provide. Contactless debit card holders do not seem to adopt contactless
payment for the purpose of reducing cash transactions, which could indicate
complementarity of cash and debit cards.
Table 2.5: Sample Summary Statistics of Debit Card Adopters
Non-Innovator Innovator
Variable Mean SD Min. Max. N Mean SD Min. Max. NIncome (in 1000)
<25 0.23 0 1 1895 0.32 0 1 18425–49 0.27 0 1 1895 0.28 0 1 18450–74 0.21 0 1 1895 0.23 0 1 18475–99 0.13 0 1 1895 0.07 0 1 184100–124 0.08 0 1 1895 0.04 0 1 184>125 0.08 0 1 1895 0.06 0 1 184
Education<High School 0.05 0 1 1900 0.1 0 1 184High School 0.38 0 1 1900 0.44 0 1 184Some College 0.29 0 1 1900 0.28 0 1 184College 0.16 0 1 1900 0.11 0 1 184Post Graduate 0.13 0 1 1900 0.07 0 1 184
EmploymentWorking 0.61 0 1 1900 0.73 0 1 184Retired 0.2 0 1 1900 0.11 0 1 184Unemployed 0.09 0 1 1900 0.11 0 1 184
Marital StatusSingle 0.18 0 1 1900 0.23 0 1 184Married 0.64 0 1 1900 0.59 0 1 184
OthersMale 0.47 0 1 1900 0.56 0 1 184Age 47.22 16.87 18 109 1900 41.06 14.49 19 77 184HH Members 1.31 1.5 0 9 1900 1.79 1.72 0 8 184Assets 1.34 8.27 0 100 1818 1.28 7.76 0 80 172Cash WD 5.35 9.98 0 434.82 1893 10.07 20.05 0 130.45 184
Note: Survey weights used. Subcategories do not sum to main category due to rounding andweighting. Cash withdrawals (WD) per month. Assets (in 1000) do not include primary home.
Previous studies have found significant evidence that perceptions about
payment attributes such as costs, safety, and convenience improve the ex-
55
planation of consumer payment decisions since they largely account for un-
observable preferences (e.g. Jonker, 2007; Schuh and Stavins, 2013). The
SCPC explicitly asks respondents to evaluate their perceptions about debit
and credit cards in terms of security, setup, acceptance, cost, records, and
convenience on a categorical scale from one to five, where the latter implies
the strongest view. Innovators in general rate the six characteristics listed as
higher than non-innovators implying that contactless payment might have
subtly and positively altered the perception and affinity towards these cards
(see Table 2.6). It is noteworthy that especially convenience is highly at-
tributed to contactless payment. Costs for debit cards are perceived as lower
by innovators than non-innovators in contrast to credit cards, which costs
are rated higher by contactless credit card adopters.
Table 2.6: Statistics of Perceived Characteristics
Credit Cards Debit Cards
NI I NI I
Variable Mean SD N Mean SD N Mean SD N Mean SD NSecurity 3.09 1.26 1886 3.29 1.27 191 3.04 1.18 1893 3.44 1.29 182Setup 3.69 1.14 1889 3.95 0.95 191 3.97 0.93 1894 4.16 0.89 184Acceptance 4.44 0.81 1889 4.5 0.69 190 4.32 0.82 1893 4.51 0.75 184Cost 2.85 1.35 1886 2.93 1.36 190 3.96 0.98 1890 3.73 1.08 183Records 4.3 0.85 1881 4.43 0.76 190 4.1 0.93 1888 4.36 0.68 184Convenience 4.25 1.02 1884 4.49 0.79 191 4.27 0.97 1891 4.49 0.93 184
Note: Survey weights used. The perceived characteristics are measured with a Likert scaleranging from one to five with five representing the strongest view.
The perceived characteristics of credit and debit cards are constructed
for the purpose of this paper as the average of each respondent’s perception
relative to all other payment methods at the POS similar to the procedure
in Arango et al. (2011). It is calculated as
RCHARkij ≡CHARkij∑J
j=1 CHARkij′
(2.1)
where k describes the six characteristics such as security, setup, acceptance,
cost, records, and convenience, i indexes the consumer, j relates to the
56
payment instrument debit or credit card and j’ is every other payment in-
strument besides j that is commonly used at the POS.14 The construction is
applied to every consumer regardless of the adoption stage of the payment
methods. This allows normalizing the perception of a particular attribute
by the individual’s overall absolute perceived levels of satisfaction across
payments at the POS (Arango et al., 2011).
To conclude, the descriptive statistics distinguished by innovators and
non-innovators, defined by the adoption of contactless payment either for
credit or debit cards, have offered some suggestive evidence that contactless
payment leads to increased transactions at the POS. Also, there is strong
evidence that individuals do not randomly adopt the contactless payment
innovation because some distinct adoption patterns between innovators and
non-innovators are observable. Lastly, the perception of attributed character-
istics towards credit and debit cards analyzed separately for innovators and
non-innovators raises issues about endogeneity since positively attributed
experiences of contactless payment may have affected its usage. The next
section outlines the empirical strategy to estimate the relationship of con-
tactless payment on transaction frequency.
2.4 Methodology
2.4.1 Identifying Assumptions
To estimate the relationship between contactless payment and the transac-
tion ratio one can use standard OLS regression:
TRANSRij = αIij + βXi + ǫi (2.2)
where TRANSRij is the share of transactions of individual i for payment
method j, where j relates to debit or credit cards, relative to every other
payment instrument j’ besides j that is commonly used at the POS, Iij takes
the value of one if the individual is an innovator, i.e. a contactless payment
14Such as cash, stored-value cards, and checks.
57
adopter for payment method j, Xi are the observed characteristics for indi-
vidual i and ǫi is the error term.15 It is necessary that the variable Iij is
strictly exogenous to obtain an unbiased estimate of the parameter α. How-
ever, as the descriptives have shown, it is most likely that the adoption of
the contactless feature (Iij) is non-randomly assigned and thus the estimate
may be biased and inconsistent (selection bias).16 There is great concern
that some unobserved variables cause individuals to select into treatment
and simultaneously to make more card payments. For instance, individu-
als could deliberately adopt contactless payment because they pay generally
more by payment cards resulting in higher preferences towards the contact-
less technology. The utility of contactless payment might be much greater
for these individuals than for others.
Moreover, it might be the case that Iij is correlated with some other
variables that could also have an impact on the number of payments and
cannot be measured directly (omitted variable bias). For instance, individ-
uals that frequently use payment cards are specifically addressed by card
issuers promoting the use of the contactless feature. Another important un-
observed factor that might determine the adoption and usage of contactless
payment could be an individual’s affinity for new technologies, labeled per-
sonal innovativeness, that influences preferences for electronic payments and
the likelihood of adopting payment innovations.17
Further, it is most likely that contactless payment and transaction fre-
quency suffers from reverse causality since contactless payment may induce
individuals to make more transactions or individuals could adopt contactless
15Other payment methods used at the POS are cash, stored-value cards, checks, andmoney order.
16See also Fung et al. (2014).17One might also consider the fact that the payment market is inherently character-
ized by a special market structure, i.e. the two-sided market, where network effects arepredominant. Put differently, the value of contactless payment for a consumer dependson the number of others using it. If the critical level of users had not been exceeded, themerchants would not invest in payment terminals and offer this payment method due tosmall economies of scale. This is typically referred to as the “chicken-and-egg problem”.Hence, the adoption and usage of contactless payment may face feedback effects, implyingthat consumers will actually choose contactless payment conditional on the number ofterminals available that allow deploying this technology. However, this issue cannot beaddressed adequately in the estimation due to data restrictions.
58
payment to meet their personal preferences for frequent usage of payment
cards. It is thus not evident if innovation drives the number of transactions
or vice-versa.
These biases all stem from endogeneity, i.e. the regressor Iij is correlated
with the error term ǫi. In these circumstances, OLS provides biased estimates
of the effect of the treatment Iij .18
A common and reliable methodology to control for endogeneity is the
instrumental variable (IV) research design providing high order of internal
validity. In this sense, the IV (or alternatively the excluded instrument)
must be highly correlated with the endogenous explanatory variable – the
treatment Iij – und must not be correlated with the error term ǫi. However,
the IV estimates are only as good as the excluded instruments used. It has
been cumbersome to find plausible instruments in this context.
A significant amount of unobserved heterogeneity can be captured by
the inclusion of individuals’ perceptions on payment cards characteristics
(Jonker, 2007; Kim et al., 2006; Ching and Hayashi, 2010).19 Also, some
proxy variables that account for personal innovativeness help to control for
unobservables. Therefore, the issue of endogeneous treatment is largely mit-
igated. However, the problem of non-random assignment into treatment has
to be eliminated.
2.4.2 Estimation Strategy
To cope with the problem of selection into treatment, I apply propensity
score matching (PSM) that generally provides a high order of internal validity
(Nichols, 2007). Regarding the measurement of the difference in transactions
between innovators and non-innovators at the POS, I define the potential
outcome TRANSRij(Iij) as the ratio of transactions for individual i and
18Only with strong distributional assumptions on Iij and βi, i.e. both parameters arenormally distributed implying the effect of the treatment Iij does not vary across individ-uals, the causal effect may be consistently estimated by OLS (Nichols, 2007). However,one can hardly think of such a homogeneous effect in reality.
19Some other endogeneity issues may arise since it is far from clear-cut whether theperceived characteristics of contactless payment lead to more transactions or is it thatthe gained positive experiences of transactions by contactless cards induce the perceivedcharacteristics to rise.
59
payment method j, where j relates to debit or credit cards, relativ to every
other payment instrument j′ besides j that is commonly used at the POS,
and where Iij equals one if individual i receives treatment (Iij = 1) of
payment method j and zero otherwise (Iij = 0).20
According to Caliendo and Kopeinig (2005), the treatment effect for an
individual i and payment method j can be written as
τij = TRANSRij(1)− TRANSRij(0). (2.3)
However, the problem arises that only one of the potential outcomes is
observed for each individual i, where i = 1, ..., N andN denotes the total pop-
ulation. Therefore, the individual treatment effect τij cannot be estimated
and one has to focus on (population) average treatment effects, which can
be measured by invoking some identifying assumptions. Under the assump-
tion that the selection into treatment solely depends on the observables Xi
and the potential outcome is independent on the treatment assignment, the
PSM gives consistent and efficient estimates of the average treatment effects.
This is a strong assumption known as unconfoundedness or conditional in-
dependence assumption. It implies that the decision to adopt contactless
payment is random and exogenous to other variables such as the number of
payment card transactions. Given this assumption, the average difference in
the transaction ratio is thus defined as the expectation of the difference in
the transaction ratio of adopters and non-adopters. The parameters to be
estimated are
τATE|X = E[TRANSRij(1)− TRANSRij(0)|Xi], (2.4)
τATT |X = E[TRANSRij(1)− TRANSRij(0)|Xi, Iij = 1], (2.5)
where the ATE (τATE|X) represents the average treatment effect and the
ATT (τATT |X) the average treatment effect on the treated that measures
the mean effect of the treatment for the sample of innovators. The ATT is
20Other payment methods used at the POS are cash, stored-value cards, checks, andmoney order.
60
more relevant in this context since individuals tend to become more and more
contactless payment adopters due to the diffusion process of the innovation.
Since conditioning on all relevant covariatesXi is restricted in case of high
dimensions, Rosenbaum and Rubin (1983) suggest using balancing scores
such as the propensity score. It requires that all variables relevant to the
probability of being selected into treatment may be observed and included
in Xi. In the first step, the PSM estimates each individual’s probability
of receiving the treatment p(Iij = 1|Xi), i.e. the probability of adopting
contactless payment for payment method j, conditional on the observables
Xi, and matches individuals with similar predicted propensities p(Xi) in
the second step. There are actually various matching algorithms to choose
from (e.g. nearest neighbor, caliper and radius, kernel and local linear). The
matching allows the untreated units to be used to construct an unbiased
counterfactual for the treatment group. Based on the propensities provided
by Logit or Probit estimation, the ratio of transactions of seemingly similar
individuals is then compared and averaged. The PSM estimators for τATE|X
and τATT |X then result in
ATE = τATE|X =1
N
N∑
i=1
[Iij − p(Xi)]TRANSRij(Iij)
p(Xi)[1− p(Xi)], (2.6)
ATT = τATT |X =1
N1
N1∑
i=1
[Iij − p(Xi)]TRANSRij(Iij)
1− p(Xi), (2.7)
whereas N1 equals the number of innovators. The estimators are the mean
differences in outcomes weighted by the propensity score.
Another requirement besides the conditional independence assumption
is the overlap assumption ensuring that individuals with the same Xi have
positive probability of both adopting and non-adopting contactless payment,
such that 0 < p(Iij = 1|Xi) < 1. This ensures having a comparison group
in the sample.
61
2.4.3 Sensitivity Analysis
If selection is not exclusively on observables, the estimator will be both bi-
ased and inefficient. In order to check if the estimates are robust and to
calculate how sensitive the estimates are to unobserved variables, I estimate
the Rosenbaum bounds (RB), which provide evidence on the degree to which
significant results hinge on the unconfoundedness assumption. However, it
cannot directly be tested because this would mean explicitly observing vari-
ables that affect selection into treatment (Rosenbaum, 2002). The participa-
tion probability of payment innovation is given by
p(Iij |Xi) = F (Xiβ + γui), (2.8)
where Iij equals one if individual i receives treatment of payment method j
and zero otherwise, Xi are the observed characteristics for individual i, F
is the cumulative density function, ui is the unobserved variable and γ is
the effect of ui on the participation decision into treatment. The log-odds
ratios p(Iij |Xi)/p(Ikj |Xk) = 1 for matched individuals with the same char-
acteristics Xi = Xk if there is no hidden bias, γ = 0, implying that the
participation probability is exclusively determined by Xi and there is no
unobserved variable that simultaneously affect the probability of receiving
treatment and the outcome variable. However, two individuals with identi-
cal X will have different chances of treatment if there is hidden bias, γ > 0,
so that the log-odds will be p(Iij |Xi)/p(Ikj |Xk) 6= 1. In fact, the sensitivity
analysis evaluates how changing the values of γ affects inference of the treat-
ment effect while the RB are the bounds on the odds ratio that either of the
two matched individuals will receive treatment (Rosenbaum, 2002).
2.5 Results
2.5.1 Estimation Results
First, to estimate the effect of contactless payment on the ratio of transac-
tions, I obtain the propensity score of adopting contactless credit or debit
cards separately for each individual, where contactless payment adopters rep-
62
resent the treatment and non-adopters the control group. Second, I compare
the share of credit and debit card transactions to the total POS transactions
of individuals in the treatment and control group with similar propensity
scores based on the kernel matching algorithm and average it over the whole
sample N and subsample N1 resulting in the ATE and ATT. The focus of
the results lies on the expected effect of contactless payment for a randomly
chosen innovator (ATT). I thereby apply the Stata module psmatch2 to
implement PSM, which is provided by Leuven and Sianesi (2003).
Regarding the inclusion of optimal covariates in the propensity score
model, only those that are unaffected by participation should be considered,
i.e. they should be time invariant or measured in advance of the treatment
(Caliendo and Kopeinig, 2005). According to theory (e.g. Venkatesh et al.,
2003; Rogers, 2003) and previous research on contactless payment (Fujiki and
Tanaka, 2014; Lee and Kwon, 2002; Wang, 2008), I estimate two Logit models
separately for contactless credit and debit cards that control for demograph-
ics, financial status, perceptions on card attributes, personal innovativeness,
the number of cash withdrawals and residential states.21 The corresponding
link tests indicate that the Logit models are properly specified.22
The marginal effects of the Logit estimations both for contactless credit
and debit cards are displayed in Table 2.7. It is observable that the num-
ber of cash withdrawals, education, some income and age brackets, as well
as certain perceptions and whether being single and having adopted mobile
banking are statistically significant effects in describing the adoption of con-
tactless credit cards, holding all else constant. The probability of adopting
contactless credit cards for individuals earning between 75000 and 99000
USD is 7.2 percent higher than for those earning 100,000–125,000 USD and
11 percent higher for people aging 25–34 compared to people younger than
25 years. Singles and college graduates are less likely to adopt contactless
payment compared to the widowed (–9.3 percent) and also less likely than
high school graduates (–16.4 percent), respectively. A one percent increase
in the number of cash withdrawals lowers the probability of adopting con-
21For more details on the theoretical background, see section 2.2.22Test statistics are not provided.
63
tactless credit cards by 2.2 percent, whereas the adoption of mobile banking
rises the probability by 4.4 percent. This may give evidence that personal
innovativeness has a crucial effect on the adoption behavior of innovations.
As convenience of credit cards in relation to all other payment methods in-
creases, individuals are more likely to adopt contactless credit (21 percent).
This is a strong indicator that contactless credit cards may meet this require-
ment.
I find evidence that education, younger cohorts, low income individuals,
certain perceived attributes, the number of cash withdrawals and whether
to revolve on credit cards or not are, ceteris paribus, statistically significant
factors that predict the adoption of contactless debit cards. For instance,
people that attended college are 23 percent less likely to adopt contactless
debit cards compared to lower than high school attendants. As costs of debit
cards decrease and acceptance increase, the probability to adopt contactless
debit rises by around 30 and 42 percent, respectively, implying the impor-
tance of supply-side factors. Credit card revolvers are 4.6 percent less likely
to adopt contactless debit, which may suggest that these heavily rely on the
provision of credit, which debit cards cannot provide. Also, a one percent in-
crease in cash withdrawals rises the probability to adopt contactless debit by
4.5 percent indicating some complementarity between cash and debit cards.
As opposed to theory, gender does not have any influence on the adoption
patterns of contactless payment.
The relationship between the transaction ratio and the propensity score
for innovators and non-innovators both for credit and debit cards is depicted
in Figures 2.3 and 2.4. It can be inferred that as the propensity score in-
creases, adopters have a higher ratio of transactions. This relationship is
slightly stronger for contactless credit adopters than non-adopters (0.7 vs.
0.67) while for contactless debit adopters, the correlation is less pronounced
(0.13 vs. 0.79).
64
Table 2.7: Logit Propensity Score Marginal Effects
Contactless Credit Contactless Debit
Mfx Std. Err. Mfx Std. Err.Income (in 1000)
<25 0.075* (0.043) 0.094* (0.052)25–49 −0.002 (0.038) 0.017 (0.049)50–74 0.043 (0.032) 0.030 (0.047)75–99 0.072** (0.034) 0.015 (0.050)>125 0.058 (0.037) 0.036 (0.051)
EducationHigh School −0.167*** (0.052) −0.133** (0.052)Some College −0.151*** (0.053) −0.173*** (0.054)College −0.164*** (0.056) −0.232*** (0.060)Post Graduate −0.110* (0.056) −0.252*** (0.061)
Age25–34 0.110* (0.063) 0.165*** (0.058)35–44 0.071 (0.067) 0.166*** (0.060)45–54 0.035 (0.066) 0.136** (0.061)55–64 0.014 (0.069) 0.095 (0.065)>65 −0.059 (0.077) 0.035 (0.075)
EmploymentWorking −0.016 (0.036) 0.036 (0.039)Retired 0.036 (0.045) 0.049 (0.044)Others −0.033 (0.036) −0.004 (0.038)
Marital StatusMarried −0.008 (0.039) 0.000 (0.052)Separated −0.030 (0.042) 0.038 (0.061)Single −0.093* (0.051) 0.005 (0.062)
PerceptionSecurity −0.021 (0.076) −0.122 (0.096)Setup 0.008 (0.143) 0.182 (0.197)Acceptance −0.118 (0.234) 0.426* (0.230)Cost 0.031 (0.118) 0.307** (0.156)Records −0.012 (0.134) −0.192 (0.156)Convenience 0.210* (0.125) −0.029 (0.139)
OthersMale 0.013 (0.020) 0.005 (0.023)log(Assets) −0.002 (0.005) 0.007 (0.006)CC Revolver 0.012 (0.020) −0.046** (0.022)HH Members −0.007 (0.008) 0.009 (0.008)Mobile Banking 0.044* (0.027) 0.107*** (0.028)log(Cash WD) −0.022** (0.010) 0.045*** (0.011)
Observations 1565 1466Pseudo-R2 0.219 0.302log(likelihood) -18377 -18470
Note: Marginal effects are displayed. Survey weights used. Significance levels 1% ***, 5% **,and 10% *. Base category for income is between 100,000–125,000 USD, for education is lowerthan high school, for age under 25, for employment unemployed and for marital statuswidowed. For brevity, coefficients of residential state dummies are not displayed.
65
0.5
1
0 .2 .4 .6 0 .2 .4 .6
Non�Adopter Adopter
Fitted values Credit Card Share POS
psmatch2: Propensity Score
Graph� b� Conta�t�e�� CC
Note: Logit propensity score, share of credit card payments at the POS
Figure 2.3: Transaction Frequency vs. Propensity Score of Contactless CreditCards
66
0.5
1
0 .2 .4 .� 0 .2 .4 .�
�on��dopter �dopter
itted a��e Debit Card Share ��S
p �at�h2� �ropen it� S�ore
�raph� b� Conta�t�e�� DC
Note: Logit propensity score, share of debit card payments at the POS
Figure 2.4: Transaction Frequency vs. Propensity Score of Contactless DebitCards
67
Common Support
Figures 2.5 and 2.6 exhibit the distribution of the propensity scores of con-
tactless payment adopters and non-adopters both for credit and debit cards.
They visually show that both distributions overlap and thus the common
support assumption is fulfilled. It is also worth noting that the identified
heterogeneity between these two groups, which is discussed in section 2.3.2,
is recognizable. The majority of cases within the control group concentrates
on the interval from 0 to 0.1, where those of the treatment group mostly lie
above 0.1. Consequently, individuals differ from the covariates being used in
the analysis.
02
4
�
8
Den
�
ity
0 .2 .4 .�Propen�ity Score
kdensity Contactless_CC_Adopters kdensity Non_Adopters
Note: Logit propensity score
Figure 2.5: Common Support for Contactless Credit Cards
Matching Quality
The principle of the PSM estimator is to compare the outcome of a treated
individual with the outcomes of counterfactual group members. There are
various matching algorithms to choose from, which all define the neighbor-
68
02
46
810
D
ensity
0 .2 .4 .6 .8�ropensity Score
kdensity Contactless_�C_Adopters kdensity Non_Adopters
Note: Logit propensity score
Figure 2.6: Common Support for Contactless Debit Cards
hood of the treated individual and the weights assigned to these neighbors
differently (cf. Caliendo and Kopeinig, 2005). In this study, I choose kernel
matching using the Epanechnikov kernel function with default bandwidth
0.06. It uses weighted averages of all individuals in the control group to con-
struct the counterfactual outcomes, where weights depend on the distance
between each individual from the control group and the participant obser-
vation for which the counterfactual is estimated (Caliendo and Kopeinig,
2005). This comes at the advantage of lower variance due to more available
information, but also observations could be included that are bad matches.
Other matching algorithms use the nearest neighbor as the control in
terms of being closest to the propensity score of the treated individual. How-
ever, bad matches are possible if the closest neighbor is far away. Caliper and
radius matching algorithms avoid this problem by imposing a tolerance level
on the maximum propensity score distance to select the nearest match. Since
these algorithms use only one observation from the control group to construct
69
the counterfactual outcome of the treated, it makes sense to choose more
than one closest neighbor in this study by using kernel matching to improve
precision in estimates (lower variance), because there are many comparable
untreated individuals in the sample (cf. Caliendo and Kopeinig, 2005).23
After matching, significant differences between the control and treatment
group should not be observable anymore. To test whether unequally dis-
tributed covariates between the groups are in sum well balanced by the
propensity score, I here present test statistics of the matching quality in
Table 2.8. The test statistics show that the Pseudo-R2 is close to zero and
statistically insignificant in all cases, implying that none of the covariates
is suitable to predict participation anymore. Furthermore, the mean bias
before and after matching indicates strong matching quality since the bias is
reduced below 3 percent in all cases, which is the threshold for appropriate
matching.24 Therefore, kernel matching has proved to be a good match-
ing algorithm to balance the distribution of variables X in the control and
treatment group.
Table 2.8: Matching Quality
Pseudo-R2 Mean BiasCC POS 0.002 1.7
(0.067***) (12.0)
CC Retail 0.004 1.4(0.110***) (9.3)
CC Services 0.002 1.7(0.066***) (11.7)
DC POS 0.004 2.4(0.119***) (17.2)
DC Retail 0.004 2.5(0.118***) (17.5)
DC Services 0.003 2.1(0.120***) (17.3)
Note: Significance levels 1% ***, 5% **, and 10% *. After matching, the likelihood-ratio testis not significant indicating that the regressors cannot predict participation into treatmentanymore, i.e. good matching quality. Figures before matching are in parentheses.
23In fact, only around 9 and 12 percent of respondents, respectively, possess a contact-less credit and debit card (see Table 2.1).
24A bias reduction below 3 or 5 percent is considered to be sufficient (Caliendo andKopeinig, 2005).
70
Results
The results of the treatment effects of contactless payment on the transaction
ratio of different transaction types are presented in Table 2.9. As a reference
point – besides PSM estimation – the ATE and ATT are additionally calcu-
lated using Tobit estimation that accounts for data censoring at zero, but
does not consider non-random assignment into treatment. These parameters
are obtained by the basic regression equation 2.2. The statistical significance
of the ATT in the PSM estimation, where the focus of the effect lies on, is
calculated with the bootstrapping method as proposed in Lechner (2002),
because also the variance due to the propensity score and the imputation of
the common support, besides the variance of the treatment effect, has to be
considered to estimate standard errors (see Table 2.10).25
Table 2.9: Impact of Contactless Payment Cards on the Transaction Ratio
ATETobit ATTTobit ATEPSM ATTPSM
CC POS 0.096 0.131 0.080 0.083(0.076) (0.032) (-) (0.027)
CC Retail 0.074 0.094 0.047 0.048(0.065) (0.032) (-) (0.019)
CC Services 0.020 0.049 0.031 0.035(0.050) (0.024) (-) (0.014)
DC POS 0.239 0.158 0.144 0.100(0.086) (0.010) (-) (0.029)
DC Retail 0.120 0.099 0.085 0.045(0.079) (0.028) (-) (0.023)
DC Services 0.142 0.105 0.052 0.045(0.050) (0.012) (-) (0.015)
Note: Tobit and PSM-kernel matching estimates are provided. Standard errors in parentheses,but they are not available for the ATE. Survey weights are used for Tobit estimation.
Overall, I find that contactless payment has a positive impact on the
transaction ratio of credit and debit card POS payments, both of retail and
services payments (see Table 2.9). Comparing the results of the OLS and
PSM estimation leads to the conclusion that self-selection into contactless
payment is evident since the effects throughout are higher in the Tobit estima-
tion (with the exception of the ATTTobit for credit card services payments).
25The standard errors are not available for the ATE.
71
In the following, the discussion of the estimation results focuses on the ATT
because the interest is in the effect of contactless payment for a randomly
chosen innovator.
The results of the ATT are statistically significant except for debit card
retail transactions (see Table 2.10 for significance tests). The ATT of con-
tactless credit cards on the transaction ratio is associated with an increase
of 8.3 percent, of which 4.8 percent stem from retail and 3.5 percent from
services payments, respectively. The ATT of contactless debit cards is 10
percent while the effect is similar for retail and services payments (4.5 per-
cent). The results imply that an average contactless credit card adopter,
who makes roughly 17 credit card transactions at the POS within a month
and with a transaction ratio of 36 percent, increases the number of credit
card transactions to approximately 21 payments under the assumption of
constant overall POS payments. An average contactless debit card adopter
with a transaction ratio of around 48 percent and 24 monthly debit card
payments rises the corresponding transaction volume by 5 transactions to 29
payments, holding total POS payments constant. Consequently, an average
debit card innovator increases fee turnover of debit card issuers by roughly
7 USD per year.26
2.5.2 Results of the Sensitivity Analysis
Table 2.10 displays the results of the sensitivity analysis, which are provided
by the Rosenbaum bounds for average treatment effects on the treated. Since
potential overestimation of the true treatment effects is suspected due to
positive selection, the upper bound significance levels are reported (p-values).
The test statistics show – under the assumption of no hidden bias (γ = 0 or
Γ = 1, respectively) – that the ATTs are statistically significant indicating
that no selection bias occurs, i.e. those who have a contactless feature do
not have higher transaction ratios even without adopting with the exception
of debit card retail payments.27 Further, the results reveal that the ATTs
for credit and debit card POS payments are still statistically significant even
26Assuming an interchange fee of 0.12 USD per transaction.27Note that Γ = eγ .
72
if a confounding factor would alter the odds of the adoption of contactless
credit cards (Γ = 1.25) and debit cards (Γ = 1.5). The upper bound Hodges-
Lehman point estimates (in parentheses) indicate that in case of Γ = 1.25,
the treatment effect on the treated for credit and debit card POS payments
is still 4.7 and 7.4 percent, respectively.
Table 2.10: Rosenbaum Bounds Sensitivity Analysis and Significance Test
Γ 1 1.25 1.5 1.75 2 Std. Err.ATTPSM
CC POS 0.001 0.038 0.206 0.494 0.752 0.045**(0.078) (0.047) (0.023) (0.000) (-0.021)
CC Retail 0.026 0.229 0.591 0.854 0.962 0.019***(0.035) (0.014) (-0.005) (-0.021) (-0.033)
CC Services 0.059 0.356 0.726 0.923 0.984 0.014**(0.021) (0.005) (-0.009) (-0.018) (-0.027)
DC POS 0.000 0.008 0.061 0.206 0.423 0.032***(0.106) (0.074) (0.048) (0.024) (0.005)
DC Retail 0.124 0.483 0.807 0.949 0.990 0.036(0.025) (0.001) (-0.017) (-0.033) (-0.046)
DC Services 0.005 0.074 0.288 0.575 0.798 0.078***(0.037) (0.020) (0.007) (-0.003) (-0.011)
Note: Upper bound significance levels are displayed (p-values). Upper bound Hodges-Lehmanpoint estimates are in parentheses. Standard errors for the PSM estimation of the ATT arecalculated using 100 bootstrap replications taking into account the propensity score while forthe ATE it is not applicable. Γ = eγ .
2.6 Conclusion
The aim of this paper was to investigate the effect of contactless payment on
the number of transactions for different transaction types at the point-of-sale
using a comprehensive U.S. data set. Controlling for selection into treatment
by propensity score matching, my analysis reveals that recent retail payment
innovations such as contactless credit and debit cards lead to an increase in
the transaction ratio by roughly 8 and 10 percent for credit and debit cards,
respectively. The results are insensitive to any hidden bias.
The results provide evidence that faster and more convenient payment
products that can be deployed at the POS such as contactless payment in-
duce individuals to undertake more frequent transactions. These findings
73
give advice for contactless card issuers to actively promote the payment
product and thus accelerate the diffusion process, which finally is expected
to lead to increasing revenue streams. Also, they show that policy makers
should pay attention to regular market monitoring to ensure balanced fee
structures in the payment market, as more frequent transactions put higher
burdens on shop owners. Under the current interchange fee structure, for
instance, incremented costs for merchants due to more frequent debit card
usage cannot be compensated by the reduction in costs due to faster check-
outs.28
The analysis faces several limitations. First, the major downside of the
data set entails the absence of information on the exact spending in terms
of volume and value of contactless devices. It only reports their adoption
rate. In fact, there may exist two different and independent processes de-
termining the adoption in the first and the usage of contactless payment in
the second stage. For instance, contactless payment adopters could never
use the technology, but instead pay more frequently by conventional pay-
ment cards than those who do not possess a contactless card, resulting in
a possible overestimation of the corresponding effect. Payment diaries that
report each transaction in detail would help to obtain more accurate results.
Additionally, the effect on value spending could then be investigated.
Secondly, the data set does not obtain supply-side factors that obviously
play a crucial role in the context of individual payment preferences. In this
sense, the question raises how generalizable the setting of the empirical study
and the results are. There are major cultural and institutional differences
between the U.S. and European payment composition at present stemming
from history. High actual payment card usage in the U.S. can be traced back
to the historical reliance on check use in conjunction with an undeveloped
giro system, whereas the importance of credit transfers and debit cards in
Europe originated from the historical establishment of the postal giro system.
There seems to be a predominant inertia in payment instrument use and the
current patterns depend strongly on the past composition (Humphrey et al.,
28Given the fee of 0.12 USD and the reduction of costs of 0.03 USD per transaction (cf.Board of Governors of the Federal Reserve System, 2011; Borzekowski and Kiser, 2008).
74
1996). Therefore, specific payment patterns in the two payment areas may
have a significant impact on the strength of the effects. Also, the U.S. may
experience greater network effects since the diffusion of contactless payment
terminals is already at an advanced stage.
75
Chapter 3
The Impact of ContactlessPayment on Cash Usage
AbstractThis paper explores the impact of contactless payment on the con-sumer demand for cash in terms of value and volume. The specificdevices that are investigated are debit and credit cards, to which thefeature is embedded. A novel microeconomic panel data set that isdrawn from national representative surveys on consumer payment be-havior in the U.S. from 2009 to 2013 is exploited to control for unob-served heterogeneity. Employing cross-sectional estimation methods,the estimation shows that contactless payment leads to a statisticallysignificant reduction in average cash usage at the point-of-sale in termsof value and volume. The negative effect of contactless credit and debitcards on cash volume is 5 and 6 percent, respectively. The negativeimpact of contactless credit cards on cash value is estimated between12 and 16 percent, but no effect is found for contactless debit cards.Using the fixed-effects model, however, results in a statistically in-significant effect for contactless credit cards and an effect of 3 percentfor contactless debit cards on cash volume. The results obtained oncash value are unaffected.
JEL-Classification: C33, D12, E41, E42
Keywords: contactless payment, money demand, cash usage, credit cards, debitcards
77
3.1 Introduction
Cash is still the most prominent payment method at the point-of-sale (POS)
in numerous developed countries, especially at low transaction values, as mul-
tiple transactional survey data have revealed (see, for example, von Kalck-
reuth et al. (2014) for Germany, Bouhdaoui and Bounie (2012) for France,
Arango et al. (2015a) for Canada, and Bagnall et al. (2016) for the U.S. as
well as for an international comparison). However, the promotion of various
technological innovations in retail payment markets such as credit, debit,
and prepaid cards have led to an actual decline in cash usage in past years
(e.g. Lippi and Secchi, 2009; Amromin and Chakravorti, 2009; Stix, 2003).
Recent innovative payment means, such as contactless payment, attempt to
mimic the desirable features of cash promising efficient and convenient pay-
ment services and tend to negatively affect the transaction costs of payment.1
Contactless payment is therefore becoming a more competitive payment al-
ternative to traditional cash payments compared to conventional payment
cards. Thus, discussing the dismal prospects of cash usage is high on the
agenda of central banks, which organize cash management.
In this paper, I estimate the effect of contactless payment on the consumer
demand for cash. More precisely, I analyze the impact of the latest payment
technology on the use of cash in terms of value and volume. Thereby, I exploit
a unique microeconomic balanced panel data set that allows controlling for
sources of confoundedness.
Analyzing the effect of contactless payment on the usage of cash is rele-
vant for the following reasons. First, one of the major goals of central banks
is to provide efficient payment services that guarantee the sound operation
of the economy. The efficiency of payment systems – as expressed in social
welfare costs – significantly relies on the number and transaction size of cash
1Contactless payment is based on the near-field communication (NFC) technology,which is a standard radio communication technology that allows to connect devices withina four centimeter range by waving or tapping the objects without providing a signatureor PIN for verification. The feature is usually embedded in conventional payment cards,but also in other devices such as mobile phones and key fobs. Contactless payment cardsallow making instantaneous payment transactions by just waving the card over the cardreader.
78
payments. For example, van Hove (2008) measures social welfare costs of
cash usage in the Netherlands in the amount of 0.48 percent of GDP while
Humphrey et al. (2001) estimate that Norway would save up 0.6 percent of
its GDP if it changed to a cashless society. Similar findings are provided
by Schmiedel et al. (2013) for the EU-27 member states showing substan-
tial social welfare costs of cash that amount one-half percent of GDP. Thus,
measuring the demand for cash is crucial to determine the costs of payment
systems.
Second, since central banks are sole institutions that are entitled to issue
legal tender money, assessing future trends in cash demand is a key mone-
tary policy issue. Greater usage of cash substitutes such as debit and credit
cards, which may be augmented by the contactless payment feature, affects
how much cash central banks should supply. Provisioning the right amount
of money is important with respect to achieving inflation targets that, in con-
trast, minimizes the welfare costs of inflation. Similarly, decreasing money
holdings due to a reduction in “shoe-leather” costs through technological
improvements tend to increase the velocity of money circulation (assuming
constant nominal GDP) and to decrease the interest elasticity of the demand
for money, resulting in lower welfare costs of inflation (cf. Alvarez and Lippi,
2009).2
Third, the amount of money outstanding, among others, determines an
important share of central banks’ revenue. This so-called seignorage revenue
is a prominent way to raise interest-free income.3 As a consequence, central
banks are interested in prospecting future demands of cash to evaluate and
forecast a significant share of their profits.
Three papers have focused on the impact of contactless media on cash
demand so far. While Fujiki and Tanaka (2014) find, using household-level
survey data in Japan, that average cash balances do not decrease with the
adoption of electronic money and under some specifications even increase,
Fung et al. (2014) exhibit a reduction in average cash usage for transactions
2Welfare costs of inflation are the amount of less seigniorage revenue due to highernominal interest rates (real rate plus expected inflation) (Briglevics and Schuh, 2013).
3Seignorage is basically the difference between the profit of printing money and theactual costs of its physical production.
79
both in terms of value and volume due to contactless credit and stored-value
cards.4 They analyze consumer-level survey data in Canada. Apart from
the fact that the results of the former study are at odds with theoretical
predictions, it also fails to accurately correct for endogeneity issues with ap-
propriate instrumental variables. Similar problems face Fung et al. (2014),
who cannot adequately purge unobserved heterogeneity due to data restric-
tions. Chen et al. (2017) improve upon this issue by using household panel
data in Canada, but they unfortunately confront a high attrition rate of
about 50 percent, which in turn aggravates the proper analysis of the nexus
of contactless credit cards and cash usage. They find no statistically signifi-
cant impact of contactless credit cards on cash usage both in terms of value
and volume.
The contribution of my paper to existing literature comes from three
major aspects. First, the availability of a unique micro balanced panel data
set allows estimating econometric specifications that expunge endogenous
components stemming from various sources. Using rich balanced panel data
for these purposes in this setting is novel. Second, the data set enables to
investigate the effect of contactless debit cards on cash demand, which fills
an important gap in literature, since debit cards are the most popular cash-
less payment method. Third, there is still limited knowledge and unsettled
empirical evidence about the impact of contactless payment, as one of the
latest payment innovations, on cash usage. This paper therefore contributes
to existing literature of cash inventory models.
By merging five Surveys of Consumer Payment Choice (SCPC), which
form a valuable balanced panel in the years 2009–2013, I estimate the impact
of contactless payment on cash usage with respect to value and volume. The
results yield support for a statistically significant reduction in average cash
usage at the POS in terms of value and volume. Employing cross-sectional
4Fung et al. (2014) estimate a decline in cash value due to contactless credit and stored-value cards by roughly –14 and –12 percent and a reduction in cash volume by around –13and –15 percent, respectively. Electronic money (“e-money”) in the setting of Fujiki andTanaka (2014) is equivalent to money provided by contactless payment instruments. AsFujiki and Tanaka (2014, p. 1) put: “Electronic money is a payment medium that allowsbuyers and sellers to make secure and instantaneous monetary transactions with a slighttouch of the card on a terminal”.
80
estimation methods, I find a negative effect of contactless credit and debit
cards on cash volume by roughly 5 and 6 percent, respectively. The negative
impact of contactless credit cards on cash value is estimated between 12 and
16 percent, but no effect is found for contactless debit cards. After controlling
for unobserved heterogeneity using the fixed-effects model, however, I find
no evidence that contactless credit cards have an impact on cash volume
and the effect of contactless debit cards of 3 percent is half the size. The
results obtained on cash value remain unaffected. The findings align with
the theoretical concept of the transaction demand for money and highlight
the importance of transaction costs in cash management and of controlling
for unobserved heterogeneity.
I proceed as follows. Section 3.2 gives an overview of previous research
on inventory cash management models accounting for new transaction tech-
nologies. Section 3.3 lays out the theoretical framework. The empirical
specification is presented in section 3.4. Section 3.5 describes the data and
section 3.6 discusses the results. Section 3.7 concludes.
3.2 Literature Review
There is a vast number of academic work focusing on the estimation of money
demand functions and the future use of cash with regard to technological
change. In the following, I will briefly discuss the most relevant and latest
studies most closely related to this paper with special attention to technolog-
ical innovations. The theoretical background of this paper is related to the
seminal work of Baumol (1952) and Tobin (1956), which has experienced
various extensions in model applications. I refer to Duca and van Hoose
(2004), who give a profound survey on the large theoretical literature on
money demand. The model specification in this paper is undertaken in the
spirit of Attanasio et al. (2002) and Lippi and Secchi (2009), who use panel
micro data for Italian households to estimate the parameters of the demand
for currency by considering the effects of technological improvements. They
both find that the interest rate elasticity of cash holdings depends on the
withdrawal technology and is significantly lower for ATM card users than
81
non-users. The estimation of precise parameters of the money demand func-
tion is an important body of literature, which basically allows to evaluate
the welfare costs of inflation. Similar findings are provided by Alvarez and
Lippi (2009), who capture technological improvements such as the diffusion
of ATM terminals and bank branches in a structural model that accounts
for free withdrawals at random times. They show with Italian household
data that money demand and the interest rate elasticity decreases as the
frequency of free withdrawal opportunities increases.
While these studies focus on infrastructural innovations in payment mar-
kets provided by the supply-side sector, others approach the effect on cash
demand from the consumer’s point of view. For instance, Briglevics and
Schuh (2013) analyze the effect on cash demand allowing for credit card pay-
ments and revolving debt. Thereby, credit card convenience users (without
revolving debt) exhibit a negative interest elasticity, as have been found in
other studies, whereas revolvers (with revolving debt) are interest inelastic.
These face higher costs when substituting cash for credit since they imme-
diately accrue interest charges after payment. In contrast, Briglevics and
Schuh (2013) point out that bank density and ATM diffusion does not affect
consumer demand for cash in the U.S., which may be explained by the com-
pleted proliferation process of bank branches and ATMs. Several objections
against the use of the identifying variables in the exclusion restrictions of
the various empirical models in all of these studies above can be raised. The
presented paper advances identification by panel data models.
Another strand of literature deals with the approximation of the share
of cash transactions at the POS and its future usage in light of payment en-
hancement. The analysis is predominately based on aggregate data sources
such as money balances and the stock of money outstanding. From an em-
pirical point of view, the effect of payment innovations on aggregate cash
demand is mixed. Taking advantage of the natural experiment represented
by the introduction of the Euro that makes use of the heterogeneity in the
distribution of cash across Italian provinces, Columba (2009) studies the ef-
fect of ATM and POS terminals on the demand for currency and narrow
money M1. He shows that the impact on cash in circulation is negative
82
and thus in line with previous studies, whereas it has a positive effect on
narrow money. Other scholars in earlier work find that modern payment
technologies have little effect on currency usage mainly due to its superior
characteristic of anonymity. For instance, Amromin and Chakravorti (2009)
claim that the demand for low denomination notes and coins decreases as
debit card usage increases because merchants need to make less change for
customer purchases, while the demand for high denomination notes is less af-
fected, inferring from non-transactional purposes such as hoarding and illegal
activities. This is highlighted in Drehmann et al. (2004), who point out that
POS terminals negatively and ATMs positively affect the demand for small
banknotes. Snellman et al. (2001) purport, studying ten European countries
from 1987 to 1996, that the diffusion of debit and credit cards has been the
main driver of substituting away from cash, while the effect of ATMs remains
ambiguous.5 The common drawback of these studies is the use of aggregated
data sources that does not allow to perfectly disentangle cash from serving
as payment medium and store of value. Consequently, multiple-time usage
of bank notes and coins cannot be taken into consideration, which may lead
to an underestimation of cash demand.
To overcome the aforementioned difficulties, household survey data of
individual payment activities aim at studying money demand more precisely.
Thereby, Stix (2003) elucidates for Austria that the demand for purse cash
is negatively affected by debit cards both for POS payments and ATMs
withdrawals, while the effect varies with usage frequency. On the other hand,
von Kalckreuth et al. (2009) argue that the possession of credit cards does
not have any impact on the number of cash transactions, analyzing German
survey data. In contrast, Huynh et al. (2014) find both for Austria and
Canada that the acceptance of payment cards by merchants has a substantial
negative impact on the demand for cash.
The presented paper fits into the context of survey data analysis and fills
the gap in money demand literature by embracing one of the newest payment
technologies.
5Similar evidence provide Humphrey (2004) that estimates cash use in the era ofchecks, debit, and credit cards in the U.S.
83
3.3 Theoretical Background
The theoretical background for the estimation strategy is derived from the
McCallum and Goodfriend (1987) framework, which is presented as an ex-
tension of the traditional Baumol-Tobin model (Baumol, 1952; Tobin, 1956)
in Attanasio et al. (2002), who take into account innovations in transaction
technologies. Accordingly, individuals adopt payment innovations such as
contactless payment if the benefits exceed the costs of adopting the technol-
ogy. Benefits of payment innovations increase with efficiency improvements
in transacting and thus make the adoption of contactless payment more
likely since it allows for an efficient payment process. Benefits also tend
to rise with more consumption expenditures and higher transaction values,
respectively, because more spending is subject to higher time spent trans-
acting. Consequently, the rate of adoption of contactless payment varies
between consumers by their demographic characteristics (e.g. income, age,
education etc.) that determine their opportunity costs of paying.6 It is there-
fore more likely that high income individuals adopt contactless payment to
reduce transactional costs of paying since their opportunity costs of paying
tend to be higher than for others. For the same reason, cash demand tends to
be lower for contactless adopters than non-adopters because cash payments
take more time to settle than contactless payments.
In general, individuals need time to undertake transactions, which can
be reduced by money as a form of exchange and even more by financial
innovations (Attanasio et al., 2002). In the traditional Baumol-Tobin set-
ting, individuals face a trade-off between holding liquidity in the form of
money to carry out transactions and the forgone interest paid on deposited
assets. In the extended version of the model in Attanasio et al. (2002), how-
ever, consumers choose optimal money holdings to trade off time costs of
transactions against the costs of holding cash. Time costs of transactions
originate from the shadow value of time and from the “shoe-leather” costs
of withdrawing cash. Consumers therefore demand optimal money holdings
6Other costs encompass one-time operational learning costs and annual fees, amongstothers.
84
by minimizing the costs of transaction time and the forgone interest paid on
deposited assets subject to their consumption expenditures. Improvements
in transaction technology such as contactless payment and lower transaction
costs therefore lessen the demand for cash. Contactless payment also enables
to instantly access liquid assets in accounts for making payments that further
reduces the demand for cash and maximizes the return of interest paid on
deposited assets. Thereby, higher interest rates on deposited accounts create
more incentives to park money holdings that in turn reduce the demand for
cash. Conversely, higher consumption expenditures increase the demand for
cash.
3.4 Estimation Strategy
According to the theoretical background, one can use OLS to estimate the
relationship between contactless payment and cash demand similar to the
standard panel data model in equation 3.1:
Mit = αIijt + βXit + γYit + δRit + ηi + ǫit, (3.1)
where Mit denotes the measurement of cash usage, Iijt takes the value of
one if the individual is an innovator, i.e. a contactless payment adopter for
payment method j, where j relates to debit and credit cards, respectively,
Xit are the observed individual characteristics and a vector of proxies for
transaction costs, as evidenced by Connolly and Stavins (2015), Yit is the
household income to proxy for consumption expenditures, Rit is the interest
rate for primary checking accounts and the alternative cost of holding cash,
ηi is referred to as unobserved individual fixed-effects, and ǫit is the error
term for all i and t. α is the parameter of interest, which measures the effect
of contactless payment on cash usage in terms of value and volume.
In this study, the left-hand side variable M as a parameter for transac-
tional cash demand is separately measured by the usual withdrawal amount,
the cash kept in wallet, and the number of withdrawals in a typical month
to analyze the effect on cash value.7 To analyze the impact of contactless
7In the classic model of cash demand, money holdings M conceptually represent cash
85
payment on cash volume, M represents the cash share at the POS defined
as the ratio of the total number of cash transactions in a typical month at
the POS to the total number of all purchases in a typical month at the POS,
which is a robust measure towards outliers (see section 3.5).
Equation 3.1 represents the baseline specification to be estimated using
pooled OLS, which uses both between and within variation. As an additional
set of controls, I include the perceived characteristics k = security, setup, ac-
ceptance, cost, records, and convenience of cash (RCHARkit) into the second
specification (see section 3.5 for variable definition). The third specification
further controls for individuals’ primary cash withdrawal methodWMit such
as ATM, bank teller, check casher, cashback, employer, family, and others,
similar to the procedure in Briglevics and Schuh (2013).
To obtain an unbiased estimate of the parameter α, it is necessary that
the variable Iijt is strictly exogenous. However, endogeneity issues may arise
since it is most likely that the adoption of the contactless feature (Iijt) is
non-randomly assigned and thus the estimate may be biased and inconsis-
tent (selection bias). In fact, there is great concern that some unobserved
variables cause individuals to select into innovation Iijt and simultaneously
use less cash (cf. Fung et al., 2014). For instance, individuals who have
an affinity for new technologies may be more prone to hold less cash and
have a higher probability to use contactless payment. Also, since payment
behavior is suggested to be to a large extent habitual (van der Horst and
Matthijsen, 2013), omitted variables with respect to payment automatism
may confound the estimation results (omitted variable bias). Unobserved
individual-specific fixed-effects ηi may thus be correlated with the explana-
tory variable Iijt, which introduce a bias in the pooled OLS estimation.8
for transactional purposes. Individuals, however, do not only hold money for spendingmotives, but also for hoarding and precautionary reasons. Large cash holdings may bealso motivated by anticipating large purchases and could be related not only to retailpayments, but also to in-person payments beyond POS payments. In this context, themeasures of actual cash holdings may thus differ from balances consumers held for actualcash transactions. Without accurate transactional-level data, it is unfeasible to measurecash demand accurately. However, according to Briglevics and Schuh (2013), the reportedamount of cash usually withdrawn, the cash kept in wallet, and the number of withdrawalsare closely related to transactional cash balances.
8One might also consider the fact that the payment market is inherently character-
86
Furthermore, it is perfectly conceivable that contactless payment and
cash usage suffers from reverse causality. Individuals who rely less on cash
may also adopt contactless payment to meet their personal preferences for
frequent usage of payment cards, as it offers instantaneous payment. It is
thus not evident if innovation drives cash demand or vice-versa (simultaneity
bias). The inclusion of perceived characteristics of cash relative to payment
cards (RCHARkit) in the estimation equation is suggested to address this
issue.
One popular method to overcome the potential problem of unobserved
individual fixed-effects is through the use of the within-groups estimator
(mean-difference model) by exploiting the panel dimension of the data to
yield:
(Mit−Mi) = α(Iijt− Iij)+β(Xit−Xi)+γ(Yit− Yi)+δ(Rit−Ri)+(ǫit− ǫi).
(3.2)
In this way, unobserved individual fixed-effects ηi are eliminated. Therefore,
in the second step and as a test of robustness, the baseline specification in
equation 3.1 as well as the second and third specification using additional
controls are separately estimated using the fixed-effects model according to
equation 3.2.9
Estimating the models requires the implicit assumption that consumers
have an interest-bearing bank account. Because this does not hold for all
individuals in the sample used, I estimate the effect of contactless payment
on a subsample of checking account holders. This comes at the advantage of
ized by a special market structure, i.e. the two-sided market, where network effects arepredominant. Put differently, the value of contactless payment for a consumer dependson the number of others using it. If the critical level of users had not been exceeded,the merchants would not invest in payment terminals and offer this payment method dueto small economies of scale. This is typically referred to as the chicken-and-egg problem.Hence, the adoption and usage of contactless payment may face feedback effects, implyingthat consumers will actually choose contactless payment conditional on the number ofterminals available that allow deploying this technology.
9Note that one can assume that individual effects are random and uncorrelated withthe variable Iijt, leading to the random-effects model. However, employing the Hausman-test on the balanced panel for all specifications rejects the null hypothesis of the random-effects model (test statistics are not provided). Therefore, the random-effects model doesnot provide consistent estimates compared to the fixed-effects model.
87
eliminating self-selection bias since it is likely that individuals open up bank
accounts to reduce transaction costs.
In the next section, I turn to the description of the data set which is used
to estimate the model specifications.
3.5 Data
3.5.1 Source
Data are drawn from the Federal Reserve Bank of Boston, which has annually
conducted the Survey of Consumer Payment Choice (SCPC) since 2008. The
surveys are implemented in fall (fourth quarter) – primarily in October –
by the RAND Corporation as online surveys using RAND’s American Life
Panel (ALP). They are unique, comprehensive and representative surveys
that provide detailed payment information of individuals with respect to
nine payment instruments in the U.S. including cash.10 The sampling unit
is an individual consumer in the U.S. older than 18 years, whose responses
of each survey were weighted to represent all U.S. consumers aged 18 years
and older. Due to major revisions in the questionnaire and methodology
across years, the 2008 responses are not comparable. However, by merging
the 2009–2013 surveys, they form a valuable longitudinal balanced panel,
as 1132 respondents have completed all five surveys, which include similar
and identical questions.11 Hence, the SCPC provides a unique panel data
structure with respect to payment choice.
The SCPC asks consumers what payment instruments they have and how
often they use these instruments by employing a flexible reporting strategy
to enhance recall and optimize the accuracy of the number of payments.12
It also collects comprehensive data on consumer cash holdings and cash
10These include cash, checks, money orders, traveler’s checks, debit, credit, and prepaidcards, online banking bill payments, and bank account number payments.
11I refer to Foster et al. (2013), Schuh and Stavins (2014), and Schuh and Stavins(2015b) for a comprehensive description of each data set, a synopsis of its results, anddetailed information about the collection process.
12Typical periods used to measure the number of payments were during a week, amonth, or a year. They are quite consistent with the implicit average that representsconsumers’ trend behavior (Schuh and Stavins, 2015b).
88
withdrawal behavior. Since low value payments are mostly paid in cash, they
tend to get more easily forgotten due to their high frequency and low budget
impact, which may lead to underreporting. Thus, cleaning procedures were
applied to identify and edit invalid data entries for the number of monthly
payments of all payment instruments, the number of transactions, and the
typical value of cash withdrawals. The data set also provides rich information
about consumer demographic characteristics, financial status, and the rating
of payment instrument attributes.
It is noticeable that the 2009–2013 estimates are not throughout adjusted
for seasonal variation, inflation, or item non-response (missing values). Also,
the calendar time period of the 2009 survey conduction differs slightly from
the one of the 2010–2013 surveys, while the latest are very similar in terms
of size, composition, and timing of the sample. The comparability of the
surveys across years may suffer from this different survey timing, if there
were crucial monthly seasonal differences in individual payment behavior.
Also, consumers may have underreported the number of payments and with-
drawals in the tumultuous years 2009–2010 during the financial crisis and
the corresponding severe recession. Additionally, there are no longitudinal
sample weights available.
3.5.2 Description
With respect to the research question, the survey specifically asks respon-
dents if one of their credit and debit cards is equipped with the contactless
feature. However, it does not provide information on its specific usage pat-
terns. Instead, detailed statistics on the adoption and usage of basic credit
and debit cards are available.
Table 3.1 indicates that 9.6 percent of consumers of the entire 2009–2013
balanced panel reported that one of their credit cards has embedded the con-
tactless feature, whereas approximately 10 percent stated that they possess
a contactless debit card. In contrast, roughly 76 percent of respondents own
a conventional credit card and 78 percent a debit card. Credit and debit
cards are at least once used within a month by around 61 and 63 percent of
consumers, respectively.
89
Table 3.1: Adoption and Usage Rate of Payment Cards
Variable Mean Obs.Contactless Credit Cards 0.096 5624Contactless Debit Cards 0.102 5612Credit Cards 0.759 5628Debit Cards 0.78 5620Credit Cards Usage 0.613 5625Debit Cards Usage 0.63 5619
Note: Usage describes the fact that respondents make the corresponding typ of payment atleast once in a typical month. Numbers correspond to the entire 2009–2013 year balancedpanel. Survey weights used.
The adoption rate of contactless credit cards remained relatively stable at
around 10 percent over the years 2009–2013, but ranging between lowest 8.2
percent in 2010 and highest 10.5 percent in 2011 (see Figure 3.1). Contrarily,
the adoption rate of contactless debit cards annually decreased from 11.6
percent in 2011 to 8.1 percent in 2013. Contactless debit cards therefore
became less popular over the years, whereas the popularity of contactless
credit cards more or less remained unchanged.13
To gain insights in transition patterns of contactless payment in all five
consecutive years in the entire 2009–2013 balanced panel, five different types
of contactless payment adopters are defined:
13One might assume that the rate of adoption typically follows a ‘S’-curve over theyears, as proposed by Rogers (2003). However, due to the very fragmented banking andmerchants sector in the U.S. the adoption of contactless payment failed to take off (SPA,2016). On the one hand, retailers have never been eager to install contactless-enabled POSterminals. On the other hand and as a result, individuals’ contactless payment adoptionlagged behind, leading banks to slow down issuing contactless payment cards (SPA, 2016).Instead, they started to focus on the EMV (Europay International, MasterCard and VISA)standard that enables to store data in chips in addition to magnetic stripes.
90
10.3%
8.2%
10.5%10.0%
9.1%
9.9%
11.6%11.3%
10.0%
8.1%
0%
2%
4%
6%
8%
10%
12%
14%
2009 2010 2011 2012 2013
Contactless Credit Contactless Debit
Note: Survey weights used.
Figure 3.1: Adoption Rate of Contactless Payment across Years
1. Never-innovators (non-innovator; non-innovator)
2. Stayers (start-adopters and one-time switchers, respectively), who start
without having contactless payment and finally adopt it within the five
year period holding it to the end (non-innovator; innovator)
3. Leavers (stop-adopters and one-time switchers, respectively), who start
having contactless payment and then return it within the five year
period (innovator; non-innovator)
4. Permanent innovators (innovator; innovator)
5. Multiple switchers (the rest), who switch between adoption and non-
adoption of contactless payment one or several times within the five
year period
Table 3.2 provides adoption patterns of contactless payment for these five
types of adopters in the years 2009–2013 separately for contactless credit and
debit cards. The matrix both displays the total number of each adoption
type as well as all possible combinations of them. In this way, it is possible to
reveal the proportions of consumers that, for instance, simultaneously have
91
contactless debit and credit cards. Overall, penetration rates of contactless
credit and debit cards are very similar in the observational period 2009–2013.
The presence of contactless payment is quite modest in the sample meaning
that around 72 and 71 percent of respondents never adopted neither con-
tactless credit nor debit cards, respectively (non-innovator; non-innovator).
Roughly one percent of consumers are permanent innovators of both payment
cards, around 6 (credit) and 5 (debit) percent are stayers (non-innovator; in-
novator), roughly 6 (credit) and 4.5 (debit) percent are leavers (innovator;
non-innovator), and approximately 15 (credit) and 19 (debit) percent are
multiple switchers.14
In Table 3.2, there is also information about multiple payment innovation
adopters. Around 57 percent of respondents have never adopted any contact-
less payment card in the entire period. 41.5 percent of consumers who have
once adopted contactless credit cards also once adopted contactless debit
cards within the years 2009–2013, whereas 47.5 percent of one-time contact-
less debit card adopters once had contactless credit cards.15 Furthermore,
around 43 percent of all consumers in the sample once had one of the two
innovations, but only around 14 percent possessed both innovations at the
same time.16
The survey also collects data on consumer cash holdings and cash with-
drawal behavior. In the questionnaire, consumers report on two distinct
measurements of cash holdings, namely the amount of cash kept in wallet
and cash on property, which in sum is the total amount of cash holdings. The
former portrays the actual Dollar amount of cash in the consumer’s pocket,
purse, or wallet while the latter asks for the Dollar amount of cash stored
elsewhere for safe keeping in homes, cars, and offices. Cash on property is
likely to be used for hoarding, but may also serve for (unexpected) trans-
actions. For this reason, cash in wallet may effectively underestimate the
actual cash amount used for transactional purposes. However, this rather
14The relatively sizable number of multiple switchers could rely on the fact that U.S.consumers possess on average 6.5 debit and credit cards issued by different banks.
15The proportions are calculated using the sum of debit card innovators among creditcard innovators (11.56 percent) in relation to the sum of credit card innovators (27.89percent) and vice versa (13.79 vs. 29.02 percent).
16This information cannot be drawn from Table 3.2.
92
Table 3.2: Adoption Patterns of Contactless Payment in the Entire Sample
Contactless Credit Contactless Debit Cards for t;TCards for t;T
N-I; N-I; I; I; Multiple TotalN-I I N-I I Switcher
Non-Innovator; 56.81 3.03 2.03 0.62 9.62 72.11Non-InnovatorNon-Innovator; 2.83 0.94 0.22 0 1.86 5.85InnovatorInnovator; 3.21 0 1.59 0 1.24 6.04Non-InnovatorInnovator; 0.34 0.24 0 0 0.37 0.95InnovatorMultiple 7.71 1.05 0.58 0.1 5.6 15.04Switcher
Total 70.92 5.26 4.42 0.72 18.69 100
Note: Numbers are in proportions, correspond to the 2009–2013 year balanced panel, surveyweights used. N-I and I denote Non-Innovators and Innovators, respectively. Missings arecoded according to the value of their previous year. t = 2009, 2010, 2011, 2012, 2013.
objective measurement of located-specific cash holdings tend to be more pre-
cise than the amount of cash holdings a consumer usually keeps for everyday
purchases, which may be harder to recall (cf. Briglevics and Schuh, 2013).
For cash withdrawals, consumers are questioned the amount of cash they
most often withdraw and the number of withdrawals they usually make in
a typical period (week, month, or year). Both questions are asked for two
separate withdrawal locations: For the primary one, where consumers most
often get cash, and all other sources.17 In this study, I focus on the numbers
relating to the primary location as in Briglevics and Schuh (2013), because
these estimates tend to be more precise. Reporting the usual rather than
actual withdrawal amount reduces the mental burden to compute averages
of potentially diverse cash withdrawals (cf. Briglevics and Schuh, 2013). The
SCPC also declares the number of cash payments and the total number of all
purchases made in a typical month at the POS, of which its ratio measures
the cash share in terms of volume.17Withdrawal locations of cash include ATMs, bank tellers, check cashing stores, cash
back at retail stores, family or friends and others, as well as the possibility to be paid incash.
93
The evolution of these cash measure variables between 2009 and 2013
is described in Figures 3.2 and 3.3. The usual withdrawal amount and the
amount of cash in wallet do not vary considerably over time meaning that
they remained relatively stable around the average of 129 and 73 USD, re-
spectively (see Figure 3.2). Conversely, cash on property incrementally in-
creased from 253 USD in 2009 to 495 USD in 2013. This indicates an elevated
need for cash hoarding, whereas the demand for cash used for transactional
purposes remained unchanged.
122136 126 129 130
253
324
274
509 495
73 74 68 74 74
0
100
200
300
400
500
600
2009 2010 2011 2012 2013
in U
SD
Usual Withdrawal Amount Cash on Property Cash in Wallet
Note: Means are displayed. Numbers correspond to the entire 2009–2013 year balanced panel.Survey weights used.
Figure 3.2: Cash Measures (in USD) across Years
The average number of withdrawals at the primary location per month
was quite consistent across years ranging between 3.3 in 2009 and 4.0 in
2013 (see Figure 3.3). The cash ratio in volume, however, has gradually
decreased since 2009 from 40 to 33 percent in 2013. This reflects the fact
that individuals have paid less by cash at the POS from year to year.
Tables 3.3 and 3.4 report detailed statistics of the relevant cash measures
for the entire balanced panel sample distinguished by contactless credit and
debit card innovators and non-innovators, respectively. Mean comparison
94
3.3
4.0
3.63.7
4.040% 40%
35% 35%
33%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
2009 2010 2011 2012 2013
Number of Withdrawals Cash Share (right scale)
Note: Means of the number of withdrawals per month are displayed. Numbers correspond tothe entire 2009–2013 year balanced panel. Survey weights used.
Figure 3.3: Cash Measures across Years
tests between innovators and non-innovators are provided to detect statisti-
cally significant differences.
According to Table 3.3, contactless credit card adopters statistically sig-
nificantly undertake fewer cash withdrawals within a month than non-adopters
(roughly 3.0 vs. 3.8). They also have a 9.5 percentage points lower cash ratio
in volume than non-adopters, which is highly statistically different. Further-
more, they tend to have less Dollar value on property (around 89 USD) and
usually withdraw less cash than non-innovators (around 7 USD), but carry
slightly more cash in wallet (1.4 USD). What contactless credit card innova-
tors and non-innovators have in common is the fact that their average cash
value in wallet amounts little more than half of their usual cash withdrawn
and roughly a fifth of their cash on property. Contrarily, median values of
cash on property and cash in wallet are very similar. This indicates that a
small number of people hoards an extensive amount of money. In general,
some people heavily rely on cash since average cash measures are distinctively
higher than the median values resulting in high standard errors.
95
Table 3.3: Cash Measures of Contactless Credit Card Innovators and Non-Innovators
Non-Innovator Innovator t-Test
Variable Mean SD Max. Med. Obs. Mean SD Max. Med. Obs. Mean Diff.Usual Withdrawal 129.676 175.301 5000 80 5007 122.205 147.188 1300 60 550 7.471Nr. of Withdrawals 3.789 6.834 434.821 2 5019 3.029 3.885 79.904 2 549 0.698***Cash on Property 378.727 2395.259 100000 20 4858 290.246 925.621 25000 35 528 88.542Cash in Wallet 72.434 135.753 3500 35 5016 73.831 124.241 1500 40 554 −1.397Cash Share .377 .285 1 .332 4842 .281 .233 1 .239 535 0.095***
Table 3.4: Cash Measures of Contactless Debit Card Innovators and Non-Innovators
Non-Innovator Innovator t-Test
Variable Mean SD Max. Med. Obs. Mean SD Max. Med. Obs. Mean Diff.Usual Withdrawal 129.307 174.081 5000 80 5129 127.028 163.032 1000 60 419 2.476Nr. of Withdrawals 3.634 6.704 434.821 2 5139 4.114 4.867 79.904 3 421 −0.601*Cash on Property 363.539 2138.131 100000 20 4972 440.159 3412.949 55000 20 406 −76.62Cash in Wallet 72.856 126.975 2000 40 5140 70.935 191.193 3500 30 422 1.921Cash Share .369 .282 1 .323 4973 .345 .272 1 .332 398 0.026
Note: Cash management measures are reported in USD except the number of withdrawals and cash share. The usual cash withdrawalamount and the number of withdrawals relate to the primary location. Cash share is the ratio of the total number of cash transactions in atypical month at the POS to the total number of all purchases in a typical month at the POS. The minimum numbers equal zero for everyvariable. Numbers correspond to the entire 2009–2013 year balanced panel. Survey weights used. T-tests of mean differences of innovatorsand non-innovators are displayed. Differences can differ from true values due to rounding and weighting. Significance levels are denoted as*** p<0.01, ** p<0.05, * p<0.1.
96
There are also statistically significant differences between contactless
debit card innovators and non-innovators with respect to the average number
of withdrawals, meaning that the former make on average roughly 0.6 more
cash withdrawals a month than their counterparts (see Table 3.4). Contact-
less debit card adopters also tend hoard higher cash holdings on property (77
USD). Conversely, they hold around 2 USD less in wallet, withdraw lower
amounts of money (roughly 2.5 USD), and have a 2.6 percentage points
lower cash ratio compared to non-innovators. Overall, this indicates that
debit card innovators generally less rely on cash for transactional purposes
than non-innovators, but more frequently withdraw money and hoard more
cash. The differences are smaller than in case of contactless credit cards.
Similarities are observable between contactless debit card adopters and
non-adopters with respect to every single average cash measure that is con-
siderably larger than its median value. This relates to the fact that a small
proportion of respondents withdraw and hold significant amounts of money
on property and in wallet. For instance, the average Dollar value on prop-
erty is roughly five times higher than in wallet for non-innovators (364 vs.
73 USD), but the median value is around half the size (20 vs. 40 USD).
To sum up, there is descriptive evidence that contactless payment reduces
both the value and volume of cash transactions, but the effect tends to be
more pronounced for contactless credit card adopters.
Apart from the mere differences of cash usage between contactless pay-
ment adopters and non-adopters displayed in Tables 3.3 and 3.4, it is inter-
esting to analyze the differences of cash usage across the five different types
of adopters. Tables 3.5 and 3.6 compare the means of cash measure vari-
ables for the five types of contactless payment adopters. Permanent credit
card innovators statistically significantly withdraw less cash (–47 USD), un-
dertake fewer cash withdrawals (–1.7), have a lower cash share in terms of
the number of transactions (–13 percentage points), but hold more cash on
property (+384 USD) than never-innovators (see Table 3.5). Also, leavers
(stop-adopters) hold 145 USD less cash at home and 16 USD less cash in
wallet while both leavers and multiple switchers have an approximate 8 per-
centage points lower cash ratio in volume than never-innovators, which are
highly statistically significant differences (see Table 3.5).
97
According to Table 3.6, highly statistically significant mean differences
of cash usage across different types of contactless debit card adopters are
observable. Permanent debit card innovators make lower usual cash with-
drawals (–97 USD), hold 40 USD less cash in wallet, and have a 12 percent-
age points lower cash ratio in volume than never-innovators. Stayers and
leavers have lower cash holdings in wallet (–23 USD and –25 USD, respec-
tively) and on property (–275 USD and –208 USD, respectively) compared to
never-innovators. Additionally, stayers withdraw 52 USD less cash, whereas
leavers more frequently withdraw cash (+1.4) than never-innovators. In con-
trast, multiple contactless debit card switchers have a higher cash share in
volume (+4.3 percentage points) and more often withdraw cash per month
(+0.6) compared to the reference type “never-innovator”.
To conclude, contactless payment seems to be clearly correlated with
lower cash usage both in terms of volume and value. The relationship tends
to be more accentuated for always-adopters compared to start-adopters (stay-
ers) and stop-adopters (leavers), whereby it suggests to be ambiguous for
multiple switchers.
98
Table 3.5: Means of Cash Measures for Types of Contactless Credit CardAdopters
Usual Number of Cash on Cash in CashWithdrawal Withdrawals Property Wallet Share
Non-Innovator; 129.304 3.729 338.804 72.312 0.388Non-Innovator (1)Non-Innovator; 128.325 3.944 343.468 99.801 0.337Innovator (2)Innovator; 143.973 4.343 193.647 56.377 0.308Non-Innovator (3)Innovator; 82.118 2.048 723.138 60.086 0.259Innovator (4)Multiple 123.632 3.419 578.635 70.673 0.313Switcher (5)
Difference (1)-(2) 0.979 −0.215 −4.664 −27.489 0.051Difference (1)-(3) −14.669 −0.614 145.157*** 15.935*** 0.08***Difference (1)-(4) 47.186*** 1.681*** −384.334 ** 12.226 0.129***Difference (1)-(5) 5.672 0.31 −239.831 1.639 0.075***
Note: Means correspond to the 2009–2013 year balanced panel, survey weights used. T-testsof mean differences are displayed. Significance levels are denoted as *** p<0.01, ** p<0.05, *p<0.1.
Table 3.6: Means of Cash Measures for Types of Contactless Debit CardAdopters
Usual Number of Cash on Cash in CashWithdrawal Withdrawals Property Wallet Share
Non-Innovator; 131.107 3.542 367.235 75.001 0.359Non-Innovator (1)Non-Innovator; 79.154 3.572 92.546 52.552 0.351Innovator (2)Innovator; 173.051 4.879 159.353 50.026 0.408Non-Innovator (3)Innovator; 33.888 3.823 319.881 35.020 0.240Innovator (4)Multiple 127.307 4.143 505.945 75.768 0.402Switcher (5)
Difference (1)-(2) 51.953*** −0.03 274.689*** 22.449*** 0.008Difference (1)-(3) −41.944 −1.337* 207.882*** 24.975** −0.049Difference (1)-(4) 97.219*** −0.281 47.354 39.981*** 0.119**Difference (1)-(5) 3.8 −0.601** −138.71 −0.767 −0.043**
Note: Means correspond to the 2009–2013 year balanced panel, survey weights used. T-testsof mean differences are displayed. Significance levels are denoted as *** p<0.01, ** p<0.05, *p<0.1.
99
The panel data set additionally provides rich information on demographic
and financial characteristics, which I have tabulated separately for contact-
less credit and debit card holders in Tables 3.7 and 3.8 including the results
of the mean comparison tests. Referring to Table 3.7, the sample of contact-
less credit card adopters is statistically significantly more skewed towards
higher income and education brackets. In other words, more than 7 percent
of individuals earning 200,000 USD above possess a contactless credit card
compared to 2.5 percent non-adopters. Credit card innovators are more fre-
quently between 35 and 44 years old, working, widowed, Asian, revolvers,
house owners, and live in smaller households compared to non-innovators.
These mean differences are all statistically significant.
As opposed to credit card innovators, the sample of contactless debit card
adopters is statistically significantly more skewed towards lower income and
education brackets (see Table 3.8). For instance, roughly 30 percent of debit
card innovators earn less than 25000 USD and around 46 percent graduated
from high school. They are mostly younger, working, Black, Asian, Latino,
or other ethnicity, and live in larger households compared to non-innovators.
Whether and to what extent the various types of contactless payment
adopters differ from never-innovators with respect to demographics and fi-
nancial characteristics shows Table A1 for credit cards and Table A2 for
debit cards in the Appendix, including the statistics of the mean comparison
tests. Obviously, there are numerous statistically significant differences be-
tween never-innovators and the remaining types of adopters with respect to
income, education, employment, marital status, age, and ethnicity. Accord-
ing to Table A1 in the Appendix, permanent credit card innovators are likely
to be very high income and education individuals, working, married, 35–44
years old, and Asian. Stayers are more skewed towards higher income and
education brackets, working, separated, lower than 25 years old, and male.
Stop-adopters, in contrast, are poorer, less educated, have special employ-
ment status, are 35–44 years old, Latino or other races, and revolvers com-
pared to never-innovators. Multiple switchers are likely to have an income
between 50,000–74,000 USD and 125,000–199,000 USD, are well educated,
widowed, older than 65, Asian, revolvers, and home owners.
100
Table 3.7: Sample Summary Statistics of Credit Card Innovators and Non-Innovators
Non-Innovator Innovator t-Test
Variable Mean SD Max. Obs. Mean SD Max. Obs. Mean Diff.Income (in 1000)
<25 0.206 1 5055 0.141 1 555 0.059***25–49 0.285 1 5055 0.25 1 555 0.02950–74 0.199 1 5055 0.202 1 555 −0.00375–99 0.129 1 5055 0.134 1 555 0.000100–124 0.084 1 5055 0.09 1 555 −0.004125–199 0.072 1 5055 0.111 1 555 −0.037**>200 0.025 1 5055 0.072 1 555 −0.044***
Education<High School 0.039 1 5068 0.058 1 556 −0.025High School 0.391 1 5068 0.248 1 556 0.132***Some College 0.279 1 5068 0.278 1 556 0.010College 0.169 1 5068 0.199 1 556 −0.022Post Graduate 0.122 1 5068 0.217 1 556 −0.095***
EmploymentWorking 0.641 1 4956 0.712 1 554 −0.070***Retired 0.21 1 4956 0.193 1 554 0.015Unemployed 0.064 1 4956 0.045 1 554 0.018Other 0.177 1 5068 0.141 1 556 0.041**
Marital StatusSingle 0.143 1 5068 0.11 1 556 0.030*Married 0.66 1 5068 0.676 1 556 −0.010Separated 0.148 1 5068 0.145 1 556 0.059Widowed 0.049 1 5068 0.068 1 556 −0.024*
Age<25 0.044 1 5068 0.052 1 556 −0.00625–34 0.155 1 5068 0.119 1 556 0.03035–44 0.163 1 5068 0.229 1 556 −0.066***45–54 0.246 1 5068 0.243 1 556 0.01055–64 0.185 1 5068 0.142 1 556 0.043**>65 0.207 1 5068 0.216 1 556 −0.012
EthnicityWhite 0.77 1 5068 0.739 1 556 0.029Black 0.147 1 5068 0.076 1 556 0.070***Asian 0.024 1 5068 0.104 1 556 −0.074***Latino 0.091 1 5068 0.113 1 556 −0.023Other 0.06 1 5068 0.081 1 556 −0.025
OthersMale 0.456 1 5068 0.431 1 556 0.026HH Members 1.304 1.586 10 5068 0.978 1.303 8 556 0.354***Revolver 0.41 1 5045 0.499 1 553 −0.086***Home owner 0.707 1 5046 0.752 1 555 −0.055*
Note: HH refers to household. The minimum numbers equal zero for every variable. Numberscorrespond to the 2009–2013 year balanced panel. Survey weights used. T-tests of meandifferences of innovators and non-innovators are displayed. Differences can differ from truevalues due to rounding and weighting. Significance levels are denoted as *** p<0.01, **p<0.05, * p<0.1.
101
Table 3.8: Sample Summary Statistics of Debit Card Innovators and Non-Innovators
Non-Innovator Innovator t-Test
Variable Mean SD Max. Obs. Mean SD Max. Obs. Mean Diff.Income (in 1000)
<25 0.186 1 5171 0.3 1 426 −0.118***25–49 0.281 1 5171 0.283 1 426 0.01050–74 0.204 1 5171 0.168 1 426 0.03475–99 0.133 1 5171 0.098 1 426 0.038**100–124 0.088 1 5171 0.059 1 426 0.026*125–199 0.079 1 5171 0.054 1 426 0.023*>200 0.029 1 5171 0.038 1 426 −0.013
Education<High School 0.039 1 5186 0.06 1 426 −0.020High School 0.367 1 5186 0.457 1 426 −0.094**Some College 0.28 1 5186 0.274 1 426 0.013College 0.175 1 5186 0.147 1 426 0.026Post Graduate 0.139 1 5186 0.062 1 426 0.076***
EmploymentWorking 0.637 1 5076 0.727 1 422 −0.097***Retired 0.218 1 5076 0.129 1 422 0.087***Unemployed 0.061 1 5076 0.082 1 422 −0.021Other 0.176 1 5186 0.161 1 426 0.021
Marital StatusSingle 0.133 1 5186 0.206 1 426 −0.049Married 0.666 1 5186 0.609 1 426 0.031Separated 0.146 1 5186 0.17 1 426 −0.018Widowed 0.055 1 5186 0.015 1 426 0.036***
Age<25 0.04 1 5186 0.087 1 426 −0.03225–34 0.139 1 5186 0.242 1 426 −0.097**35–44 0.171 1 5186 0.157 1 426 0.01245–54 0.243 1 5186 0.276 1 426 −0.04255–64 0.189 1 5186 0.115 1 426 0.070***>65 0.218 1 5186 0.123 1 426 0.091***
EthnicityWhite 0.788 1 5186 0.598 1 426 0.197***Black 0.131 1 5186 0.199 1 426 −0.067**Asian 0.029 1 5186 0.059 1 426 −0.039**Latino 0.081 1 5186 0.169 1 426 −0.101**Other 0.052 1 5186 0.144 1 426 −0.092**
OthersMale 0.447 1 5186 0.51 1 426 −0.054HH Members 1.238 1.537 10 5186 1.57 1.772 9 426 −0.328**Revolver 0.429 1 5159 0.342 1 424 0.076**Home owner 0.736 1 5164 0.494 1 424 0.228***
Note: HH refers to household. The minimum numbers equal zero for every variable. Numberscorrespond to the 2009–2013 year balanced panel. Survey weights used. T-tests of meandifferences of innovators and non-innovators are displayed. Differences can differ from truevalues due to rounding and weighting. Significance levels are denoted as *** p<0.01, **p<0.05, * p<0.1.
102
The distribution of characteristics among contactless debit card adopters
shows – as opposed to credit card innovators – a slightly different picture, as
evidenced in Table A2 in the Appendix. Permanent debit card innovators
are likely to earn between 50,000–74,000 USD, are working, married, 45–54
years old, male, and either Asian, Latino, or other races, whereas start-
adopters (stayers) have an income between 25,000–49,000 USD and greater
than 200,000 USD, special employment status, are between 25–34 years old,
and Asian compared to never-innovators. Stop-adopters (leavers) are low
educated, working, under 25 years, black, Latino, and have more household
members than never-innovators. Multiple switchers are skewed towards the
lowest income, education, and age brackets, are working, unemployed, single,
black, Latino, other races, male, and have more household members than the
adoption type of never-innovators.
To sum up, the assorted characteristics across the five types of adopters
are well associated with the theoretical background of the adoption of inno-
vations (see section 1.2.1).
Annual household income, as proxy for consumption, is surveyed as cate-
gorical variable with 17 categories. Also, interest rates of checking accounts
are reported as categories in the SCPC. For the purpose of the analysis,
however, I computed the average of each category’s bounds to convert it
into a continuous variable, which makes the interpretation of the coefficient
straightforward. Data for the median household income earning more than
200,000 USD are drawn from the 2013 Survey of Consumer Finances as
proxy for the top income category (see SCF, 2014). Since test statistics con-
clude that the distribution of household income is skewed, the variable is
transformed for estimation taking the logarithm.18
Average checking account interest rates for the entire balanced panel are
displayed across years in Figure 3.4 including the trend line. They vary con-
siderably over time without any specific pattern ranging from 0.06 percent
in 2011 to 0.22 percent in 2010. The era between 2009–2013 is characterized
by very low interest rates that decline as the years proceed.19 In line with
18Test statistics are not provided.19Low checking account interest rates are clearly correlated with the federal funds rate
103
the theoretical prediction, the relationship between checking account inter-
est rates and cash measure variables shows a negative correlation, which is,
however, not statistically significant.20
Figure 3.5 shows the distribution of checking account interest rates of
the entire sample. The bulk of checking account holders bear zero interest
(almost one-half) or very low rates, meaning an interest rate of less than 0.25
percent. The rest of respondents – a minority of 7.3 percent – bear interest
greater than 0.25 percent, of which 3.3 percent of respondents receive 1.75
percent interest. Approximately 97 percent of individuals in the sample have
a checking account.
0.19%
0.22%
0.06%
0.12%0.13%
0.00%
0.05%
0.10%
0.15%
0.20%
0.25%
2009 2010 2011 2012 2013
Interest Rate Linear (Interest Rate)
Note: Means are displayed. Survey weights used.
Figure 3.4: Interest Rates across Years
whose target has been historically low since December 2008 ranging between zero to 0.25percent.
20Test statistics are not provided.
104
Note: The percentage numbers correspond to the share of respondents in the entire sampleyielding the interest rate on their checking account.
Figure 3.5: Distribution of Interest Rates
105
The absolute ratings of the perceived characteristics of cash are trans-
formed into relative ones – as in Schuh and Stavins (2013) – by using the
following transformation:
RCHARki(j, h) ≡ log
(CHARkij∑H
h=1 CHARkih
), (3.3)
where k describes the six characteristics such as security, setup, acceptance,
cost, records, and convenience, i indexes the consumer, j relates to cash, and
h is payment instrument h = 1, . . . , H such as credit and debit cards. The
construction is applied to every consumer regardless of the adoption stage of
the payment methods. The higher the value of variable RCHARki, the more
favorable is cash than debit and credit cards with respect to the particular
characteristic k.21
Figure 3.6 shows the distribution of individuals’ primary cash withdrawal
methods WMit in the entire balanced panel. The most popular primary cash
withdrawal method is ATMs (48 percent) followed by bank tellers (28 per-
cent), cash back at retail stores (13 percent), being paid in cash by employers
(3.8 percent), family or friends (3.7 percent), by others (3.5 percent), and
check cashing stores (0.5 percent).
To conclude, the descriptive statistics have offered some suggestive ev-
idence that contactless payment leads to reduced cash usage in terms of
value and volume. They have also revealed that the assignment into the
contactless feature is likely to be non-random due to obvious demographic
discrepancies between innovators and non-innovators. Finally, a negative
relationship between checking account interest rates and cash measure vari-
ables is observable across years.
3.6 Results
This section presents the estimation results of the model specification in
equation 3.1 using pooled OLS and in equation 3.2 using fixed-effects (FE).
21Note that the perceived characteristics such as records and setup are not surveyed inthe year 2009.
106
47.8%
27.6%
0.5%
13.2%
3.8% 3.7% 3.5%
0%
10%
20%
30%
40%
50%
60%
ATM Bank teller Check casher Cashback Employer Family Others
Note: Means are displayed. Survey weights used.
Figure 3.6: Primary Cash Withdrawal Methods
Fixed-effects and pooled OLS estimates are compared to comprehend the
importance of controlling for unobserved heterogeneity. Overall, the results
suggest that individual-specific fixed-effects – for instance, individual pay-
ment habitualization – are predominant and contactless payment adopters
positively selected, since FE estimates of contactless payment are mostly
smaller than pooled OLS estimates. Also, the Hausman-test rejects the use
of pooled OLS in favor of FE estimates in almost all specifications. The
striking differences in the goodness-of-fit (R2) between the pooled OLS and
the FE models indicate that cash usage differs more between individuals
than over time (within).22
First, I will discuss the effect of contactless payment on cash value. Sec-
ond, the impact of contactless payment on cash volume is analyzed.
22This can also be inferred from comparing the evolution of cash measures in Figures3.2 and 3.3 with these in Tables 3.5 and 3.6, especially with special attention to standarddeviations.
107
3.6.1 Estimation Results for Cash Value
Three specifications using different controls for three types of cash measures
such as usual withdrawal amount, the number of withdrawals, and cash
in wallet – serving as the outcome variables – are estimated to separately
obtain the parameter for contactless credit and debit cards on cash value.23
The main results of contactless credit cards on cash value are first discussed
according to Table 3.9. In the second step, Table 3.10 presents the main
results of the effect of contactless debit cards. The results of the full set of
covariates for each regression are reported in the Appendix.
Effects of Contactless Credit Cards
Table 3.9 shows results of the pooled OLS and FE regressions of contactless
credit cards on cash value. The validity of each specification can be checked
by two statistical tests, the F- and Hausman-test, which are provided at the
bottom of each regression.24 I find evidence that contactless credit cards have
statistically significant negative effects on the number of cash withdrawals
across all specifications. The results are rather robust against the inclusion
of additional controls. The Hausman-test clearly supports the use of the
pooled OLS regression. Accordingly, the estimated negative impact has a
modest magnitude ranging between –0.6 and –0.45, holding all else constant
(see column [1] and [3]). In other words, contactless credit cards induce
individuals to make 0.45 to 0.6 fewer cash withdrawals per month, which is
compared to the average number of withdrawals per month in the SCPC of
3.7, however, a sizeable reduction of 12 to 16 percent. These effects on cash
value are comparable to the ones in Fung et al. (2014) and Chen et al. (2017)
obtained on cross-sectional data.25
23Note that the number of observations decreases in the estimations using controlset one and two because withdrawal methods and some perceived characteristics are notsurveyed in 2009.
24The F-test analyzes the joint significance of all variables included in the regression.The Hausman-test compares the consistent but inefficient estimator (fixed-effects) to thepotentially inconsistent but efficient estimator (pooled OLS).
25Fung et al. (2014) and Chen et al. (2017) find a significant negative impact of con-tactless credit cards on cash value of roughly –14 and –10 percent, respectively, usingcross-sectional estimation methods.
108
The estimated coefficients of contactless credits cards on the usual amount
withdrawn have the expected negative sign, but I find no evidence that they
have a statistically significant effect in all specifications. Similarly, the point
estimates on cash in wallet in all regressions are not statistically relevant
and, against expectation, exhibit a positive sign. In line with theory, income
is a statistically significant factor meaning, for instance, that a one percent
increase in household income positively affects the value of cash in wallet by
around 0.16 USD (see column [6]). This is, however, a rather modest impact.
The estimated coefficients of the interest rate have theoretically predicted
negative signs, but the model fails to identify a significant negative impact.
This could reflect the fact that individuals hold sufficient cash on person for
precautionary and unexpected reasons irrespective of its opportunity costs.
Effects of Contactless Debit Cards
The hypothesis that contactless debit cards reduce the demand for cash
in value is rejected. According to Table 3.10, the estimated coefficients of
contactless debit cards in all model specifications employed are not statisti-
cally significant and show a positive relationship with cash measure variables
(except one coefficient), which is in contrast to expectation. Therefore, con-
tactless debit card adopters basically tend to hold more cash in value. The
magnitude of the contactless payment coefficients are overall more sizeable
compared to the contactless credit cards point estimates. Moreover, the
remaining coefficients of income and interest rate show analogous signs, sig-
nificance levels, and magnitude as in the contactless credit cards regressions
in Table 3.9.
3.6.2 Estimation Results for Cash Volume
The main estimates of the regressions that analyze the impact of contactless
payment on cash volume are displayed in Table 3.11, whereby the full set
of estimates is provided in the Appendix. All coefficients of the contactless
feature (except one) have predicted negative signs and are insensitive to the
inclusion of additional control variables. The estimates obtained on pooled
109
Table 3.9: Regression Results of Contactless Credit Cards on Cash Value
Pooled OLS FEBaseline Controls 1 Controls 2 Baseline Controls 1 Controls 2
Variable (1) (2) (3) (4) (5) (6)
Usual Cash Amount Withdrawn
Contactless Credit −2.051 −3.671 −2.725 −4.067 −1.065 −1.773(9.852 ) (9.804 ) (9.037 ) (5.912 ) (5.564 ) (5.429 )
log(Income) 27.989*** 29.230*** 34.260*** 3.298 −1.472 −2.342(7.620 ) (7.844 ) (6.772 ) (11.210 ) (17.331 ) (16.851 )
Interest Rate 16.764 18.420 10.115 −4.047 −5.239 −6.238(14.906 ) (15.561 ) (11.851 ) (4.324 ) (5.436 ) (5.295 )
R2 0.080 0.092 0.208 0.006 0.011 0.055
Observations 4734 3768 3768 4734 3768 3768Individuals 1119 1110 1110 1119 1110 1110F-test (p-value) 0.000 0.000 0.000 0.363 0.619 0.000Hausman-test (p-value) 0.001 0.000 0.000
Number of Withdrawals
Contactless Credit −0.598** −0.554* −0.456* −0.649** −0.862** −0.866**(0.233 ) (0.282 ) (0.272 ) (0.260 ) (0.364 ) (0.366 )
log(Income) 0.056 0.104 0.138 0.528* 0.455 0.483(0.147 ) (0.177 ) (0.179 ) (0.280 ) (0.452 ) (0.453 )
Interest Rate 0.118 0.088 −0.009 0.121 0.213 0.171(0.237 ) (0.293 ) (0.298 ) (0.245 ) (0.401 ) (0.405 )
R2 0.020 0.026 0.036 0.003 0.008 0.011
Observations 4745 3778 3777 4745 3778 3777Individuals 1119 1112 1111 1119 1112 1111F-test (p-value) 0.000 0.000 0.000 0.148 0.354 0.152Hausman-test (p-value) 0.950 0.874 0.173
Cash in Wallet
Contactless Credit 6.950 12.569 13.257 5.206 5.417 5.446(8.089 ) (9.823 ) (9.842 ) (6.154 ) (7.191 ) (7.231 )
log(Income) 17.370*** 17.972*** 18.995*** 25.438*** 15.825** 16.041**(5.674 ) (6.361 ) (6.301 ) (9.428 ) (6.425 ) (6.400 )
Interest Rate −1.368 −5.217 −4.021 −4.809 −7.218 −6.368(5.748 ) (6.178 ) (6.094 ) (4.020 ) (5.570 ) (5.457 )
R2 0.062 0.076 0.108 0.013 0.012 0.021
Observations 4741 3777 3777 4741 3777 3777Individuals 1117 1109 1109 1117 1109 1109F-test (p-value) 0.000 0.000 0.000 0.050 0.047 0.017Hausman-test (p-value) 0.034 0.025 0.000
Demographics Yes Yes Yes Yes Yes YesRelative Characteristics No Yes Yes No Yes YesWithdrawal Method No No Yes No No Yes
Note: OLS is the pooled OLS estimator and FE is the fixed-effects estimator obtained on thebalanced panel. Time-invariant variables such as gender and race are not included in the FEmodel. Cluster-robust standard errors are used and in parentheses. Survey weights are usedfor OLS. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1. TheHausman-test tests the null hypothesis that the preferred model is pooled OLS vs. thealternative the fixed-effects model. The full set of OLS estimates are reported in the Appendixin Tables A3, A4, and A5. The full set of FE estimates are displayed in the Appendix inTables A6, A7, and A8.
110
Table 3.10: Regression Results of Contactless Debit Cards on Cash Value
Pooled OLS FEBaseline Controls 1 Controls 2 Baseline Controls 1 Controls 2
Variable (1) (2) (3) (4) (5) (6)
Usual Cash Amount Withdrawn
Contactless Debit 2.196 3.139 17.018 2.449 3.067 5.453(18.450 ) (16.517 ) (12.329 ) (5.747 ) (6.380 ) (6.279 )
log(Income) 27.946*** 29.161*** 34.432*** 3.255 −1.529 −2.451(7.590 ) (7.830 ) (6.776 ) (11.209 ) (17.353 ) (16.878 )
Interest Rate 16.997 18.315 9.387 −3.993 −5.242 −6.245(14.388 ) (15.158 ) (11.615 ) (4.348 ) (5.436 ) (5.295 )
R2 0.079 0.092 0.208 0.006 0.011 0.055
Observations 4732 3767 3767 4732 3767 3767Individuals 1119 1110 1110 1119 1110 1110F-test (p-value) 0.000 0.000 0.000 0.364 0.622 0.000Hausman-test (p-value) 0.001 0.000 0.000
Number of Withdrawals
Contactless Debit 0.191 0.060 0.023 −0.224 0.090 0.073(0.320 ) (0.367 ) (0.361 ) (0.328 ) (0.347 ) (0.345 )
log(Income) 0.045 0.087 0.125 0.510* 0.438 0.466(0.143 ) (0.172 ) (0.174 ) (0.280 ) (0.453 ) (0.454 )
Interest Rate 0.110 0.092 −0.005 0.122 0.215 0.174(0.234 ) (0.291 ) (0.295 ) (0.244 ) (0.401 ) (0.405 )
R2 0.019 0.026 0.036 0.003 0.007 0.010
Observations 4744 3777 3776 4744 3777 3776Individuals 1119 1112 1111 1119 1112 1111F-test (p-value) 0.000 0.000 0.000 0.182 0.482 0.179Hausman-test (p-value) 0.938 0.898 0.172
Cash in Wallet
Contactless Debit 19.375 15.329 24.001 21.241 22.860 22.755(15.483 ) (15.050 ) (14.954 ) (18.481 ) (21.153 ) (20.622 )
log(Income) 17.849*** 18.594*** 19.774*** 25.488*** 15.632** 15.854**(5.611 ) (6.295 ) (6.229 ) (9.460 ) (6.471 ) (6.446 )
Interest Rate −2.273 −6.069 −5.202 −4.883 −7.277 −6.437(5.370 ) (5.959 ) (5.850 ) (4.028 ) (5.551 ) (5.434 )
R2 0.063 0.077 0.109 0.015 0.014 0.023
Observations 4740 3777 3777 4740 3777 3777Individuals 1117 1109 1109 1117 1109 1109F-test (p-value) 0.000 0.000 0.000 0.053 0.048 0.011Hausman-test (p-value) 0.039 0.023 0.000
Demographics Yes Yes Yes Yes Yes YesRelative Characteristics No Yes Yes No Yes YesWithdrawal Method No No Yes No No Yes
Note: OLS is the pooled OLS estimator and FE is the fixed-effects estimator obtained on thebalanced panel. Time-invariant variables such as gender and race are not included in the FEmodel. Cluster-robust standard errors are used and in parentheses. Survey weights are usedfor OLS. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1. TheHausman-test tests the null hypothesis that the preferred model is pooled OLS vs. thealternative the fixed-effects model. The full set of OLS estimates are reported in the Appendixin Tables A9, A10, and A11. The full set of FE estimates are displayed in the Appendix inTables A12, A13, and A14.
111
data are all highly statistically significant. The effect of contactless payment
has a sensible magnitude. It is associated with a decline in cash share volume
by approximately 5 percent for credit and 6 percent for debit cards, holding
all else constant. Innovators are therefore likely to reduce the number of
cash transactions due to contactless payment.
However, controlling for fixed-effects, the results show that the effect of
contactless credit cards is not statistically relevant anymore, indicating that
unobserved heterogeneity drives the results using cross-sectional data. Simi-
lar arguments can be put forward regarding the impact of contactless debit
cards, where the magnitude is half the size compared to pooled OLS, but the
coefficients are still statistically significant. The estimated negative effect of
contactless debit cards is approximately 3 percent. Ignoring unobserved het-
erogeneity will therefore lead to overestimation of the effect of contactless
payment on cash usage in terms of volume. Overall, tests statistics show that
fixed-effects models perform better than pooled OLS, revealing that income
and interest rates are not statistically significant factors that predict cash
share at the POS.
The main findings partly contrast the prior, but may highlight the su-
perior characteristics of contactless payment being more efficient and more
convenient compared to cash payments that affects cash volume. The impact
of contactless credit cards on cash volume in the amount of 5 percent using
pooled data is around half the size as estimated by Fung et al. (2014) and
comparable to the one in Chen et al. (2017).26 However, the results indicate
that ignoring unobserved heterogeneity leads to spurious results, which is in
line with the arguments proposed in Chen et al. (2017). They find no sta-
tistically significant impact of contactless credit cards on cash volume after
controlling for unobserved heterogeneity
26Fung et al. (2014) and Chen et al. (2017) estimate a negative effect of contactlesscredit cards on cash volume of roughly 13 and 8 percent using cross-sectional and pooleddata, respectively. They do not analyze contactless debit cards.
112
Table 3.11: Regression Results of Contactless Payment on Cash Share Vol-ume
Pooled OLS FEBaseline Controls 1 Controls 2 Baseline Controls 1 Controls 2
Variable (1) (2) (3) (4) (5) (6)
Contactless Credit −0.054*** −0.045*** −0.046*** 0.002 −0.003 −0.003(0.015 ) (0.017 ) (0.016 ) (0.010 ) (0.012 ) (0.012 )
log(Income) −0.031*** −0.015 −0.013 0.009 0.015 0.014(0.010 ) (0.010 ) (0.010 ) (0.011 ) (0.013 ) (0.013 )
Interest Rate 0.022 0.021 0.011 0.003 0.010 0.011(0.019 ) (0.017 ) (0.017 ) (0.008 ) (0.009 ) (0.009 )
R2 0.093 0.154 0.177 0.011 0.028 0.030
Observations 4595 3683 3681 4595 3683 3681Individuals 1115 1109 1108 1115 1109 1108F-test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000Hausman-test (p-value) 0.000 0.000 0.000
Contactless Debit −0.057*** −0.065*** −0.062*** −0.029** −0.032** −0.031**(0.021 ) (0.021 ) (0.020 ) (0.014 ) (0.015 ) (0.015 )
log(Income) −0.033*** −0.017* −0.016 0.010 0.015 0.015(0.010 ) (0.010 ) (0.010 ) (0.011 ) (0.013 ) (0.013 )
Interest Rate 0.026 0.025 0.015 0.003 0.010 0.011(0.019 ) (0.017 ) (0.017 ) (0.008 ) (0.009 ) (0.009 )
R2 0.092 0.157 0.179 0.013 0.030 0.032
Observations 4594 3682 3680 4594 3682 3680Individuals 1116 1109 1108 1116 1109 1108F-test (p-value) 0.000 0.000 0.000 0.000 0.000 0.000Hausman-test (p-value) 0.006 0.000 0.000
Demographics Yes Yes Yes Yes Yes YesRelative Characteristics No Yes Yes No Yes YesWithdrawal Method No No Yes No No Yes
Note: OLS is the pooled OLS estimator and FE is the fixed-effects estimator obtained on thebalanced panel. Time-invariant variables such as gender and race are not included in the FEmodel. Cluster-robust standard errors are used and in parentheses. Survey weights are usedfor OLS. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1. The Hausman-testtests the null hypothesis that the preferred model is pooled OLS vs. the alternative thefixed-effects model. The full set of OLS estimates are reported in the Appendix in Tables A15and A16. The full set of FE estimates are displayed in the Appendix in Tables A17 and A18.
113
3.7 Conclusion
The goal of this paper was to investigate the impact of contactless payment
on cash usage in terms of value and volume. It is suggested that multiple
layers of endogeneity are inherent in the presented setting, which asks for an
appropriate estimation strategy to obtain unbiased and consistent estimates.
Therefore, apart from cross-sectional estimations and as a test of robustness,
I employ the within estimator using a comprehensive U.S. balanced panel
data set from 2009–2013 that allows eliminating individual-specific fixed-
effects.
The findings suggest that contactless payment leads to a decreasing de-
mand for money regarding cash value and volume. Contactless credit and
debit cards economically and statistically significantly reduce cash usage in
volume by roughly 5 and 6 percent, respectively. Controlling for unobserved
heterogeneity, however, shows that the effect of contactless credit cards is
statistically insignificant, whereas the impact of contactless debit cards of 3
percent is half the size. In addition, I find a statistically significant negative
effect of contactless credit cards on cash usage in terms of value by roughly
12 to 16 percent and no effect of contactless debit cards, irrespective of the
estimation method.27 The results are in line with theoretical predictions
of the inventory cash management model proposed by Baumol (1952) and
Tobin (1956). This study makes major contributions to the literature on
money demand and advances knowledge in payment economics with regards
to latest payment innovations.
The results portray important implications. First, contactless payment
increases the efficiency of payment systems and reduces social welfare costs.
On the one hand, this is achieved by the decline in the number of cash
transactions and withdrawals. For instance, merchants may benefit from
faster electronic checkout and less cash handling costs, particularly at low
transaction values, where efficiency gains due to the vast number of cash
27In fact, I estimate a negative impact of contactless credit cards on the number ofwithdrawals. However, as the number of withdrawals are defined as the ratio of totalcash spending to the usual withdrawal amounts, cash spending has to decrease giventhe estimated negative effect on the number of withdrawals and the estimated inexistentimpact on usual withdrawal amounts.
114
transactions are greatest. On the other hand, banks and financial intermedi-
aries alike may gain from reduced withdrawal behavior with respect to cash
management that negatively affects social welfare costs. Central banks are
thus asked to enhance the use of contactless payment for the purpose of an
overall efficiency gain in payment processing.
Second, the findings may highlight the entrenched use of cash as a pay-
ment medium since consumers do not opt for smaller cash holdings on person
due to contactless payment. In principle, they may refrain from parting with
the entire cash amount on person, that is in wallet or pocket, because it of-
fers ubiquitous payment and precautionary reasons when facing uncertainty
about future purchases may hamper decreasing cash usage behavior, as in-
dicated in Eschelbach and Schmidt (2013). Consequently, the postulation
of phasing out paper currency may turn out not to be a simple task, which,
however, prominent scholars have recently proclaimed, particulary in the
era of low interest rates (e.g. Rogoff, 2016). The results may therefore offer
important information for central banks.
Third, comparing pooled OLS and fixed-effects estimates have suggested
that contactless payment adopters are positively selected, meaning that they
self-select into payment innovation. Therefore, ignoring unobserved hetero-
geneity will lead to overestimation of the effect of contactless payment on
cash demand.
The study may have several caveats. First, the data set gives no insights
in transactional purposes and henceforth, the true values of cash purchases
have to be proxied by the typical amount of cash per withdrawal, kept in
wallet, and the frequency of withdrawals. However, these measures are not
perfectly equivalent to the theoretical concept in the Baumol-Tobin frame-
work. The respective unobserved discrepancy between cash hoarding and
cash usage for transactional purposes is likely to result in measurement errors
that may bias the estimation results of the cash value estimation. Similarly,
the absence of information on the exact amount of checking account interest
rates and household income – instead of averages of each category’s bounds –
may also lead to measurement errors. Additionally, the measurement of cash
usage in terms of volume may suffer from recall effects. Payment diaries that
115
report each transaction in detail in conjunction with information on exact
interest rates and income would help to obtain more accurate results.
Second, external validity of the results could be limited since the pay-
ment composition and payment infrastructure between the U.S. and Europe
differ significantly due to their culturally and institutionally differentiated
evolution of payment systems. Whilst Europeans heavily rely on cash as
a payment mean, Americans pay more by payment cards (Bagnall et al.,
2016). Therefore, specific payment patterns in the two payment areas may
affect the magnitude of the estimated effects. Also, since the diffusion of
contactless payment terminals and contactless payment cards is already at
an advanced stage in the U.S. compared to Europe, the U.S. may experience
greater network effects that could influence the estimates.
116
3.A Appendix: Descriptives and RegressionTables
Table A1: Means of Demographics for Types of Contactless Credit CardAdopters
Variable N-I; N-I; I; I; Multiple t-TestN-I I N-I I Switcher Mean Diff.(1) (2) (3) (4) (5) (1)-(2) (1)-(3) (1)-(4) (1)-(5)
Income (in 1000)<25 0.216 0.255 0.157 0.033 0.141 -0.039 0.059* 0.183*** 0.075***25–49 0.286 0.226 0.361 0.284 0.240 0.06*** -0.075** 0.002 0.046**50–74 0.192 0.127 0.217 0.188 0.257 0.065** -0.025 0.004 -0.065***75–99 0.133 0.125 0.121 0.097 0.121 0.008 0.012 0.036 0.012100–124 0.079 0.112 0.096 0.059 0.095 -0.033 -0.017 0.02 -0.016125–199 0.070 0.107 0.017 0.168 0.107 -0.037** 0.053*** -0.098 -0.037***>200 0.024 0.047 0.032 0.171 0.039 -0.023* -0.008 -0.147* -0.015
Education<High School 0.038 0.000 0.115 0.000 0.048 0.038*** -0.077*** 0.038*** -0.01High School 0.409 0.497 0.330 0.114 0.230 -0.088 0.079** 0.295*** 0.179***Some College 0.279 0.145 0.252 0.066 0.350 0.134*** 0.027 0.213*** -0.071***College 0.159 0.158 0.190 0.379 0.212 0.001 -0.031 -0.22*** -0.053***Post Graduate 0.115 0.200 0.112 0.440 0.161 -0.085*** 0.003 -0.325*** -0.046***
EmploymentWorking 0.634 0.829 0.567 0.857 0.667 -0.195*** 0.067 -0.223*** -0.033Retired 0.212 0.076 0.221 0.158 0.236 0.136*** -0.009 0.054 -0.024Unemployed 0.069 0.058 0.055 0.000 0.041 0.011 0.014 0.069*** 0.028***Other 0.184 0.101 0.246 0.075 0.135 0.083*** -0.062* 0.109*** 0.049***
Marital StatusSingle 0.148 0.272 0.077 0.184 0.085 -0.124 0.071*** -0.036 0.063***Married 0.659 0.519 0.708 0.770 0.687 0.14* -0.049 -0.111* -0.028Separated 0.144 0.202 0.143 0.046 0.149 -0.058* 0.001 0.098*** -0.005Widowed 0.049 0.006 0.072 0.000 0.078 0.043*** -0.023 0.049*** -0.029***
Age<25 0.038 0.132 0.067 0.068 0.029 -0.094** -0.029 -0.03 0.00925–34 0.163 0.215 0.087 0.158 0.100 -0.052 0.076*** 0.005 0.063***35–44 0.163 0.129 0.223 0.439 0.176 0.034 -0.06* -0.276*** -0.01345–54 0.246 0.212 0.267 0.085 0.263 0.034 -0.021 0.161*** -0.01755–64 0.185 0.179 0.150 0.091 0.186 0.006 0.035 0.094* -0.001>65 0.206 0.133 0.206 0.159 0.247 0.073* 0 0.047 -0.041**
EthnicityWhite 0.768 0.779 0.688 0.612 0.787 -0.011 0.08** 0.156* -0.019Black 0.160 0.051 0.169 0.000 0.085 0.109*** -0.009 0.16*** 0.075***Asian 0.019 0.027 0.018 0.295 0.079 -0.008 0.001 -0.276*** -0.06***Latino 0.081 0.155 0.192 0.000 0.095 -0.074 -0.111*** 0.081*** -0.014Other 0.052 0.143 0.125 0.093 0.049 -0.091 -0.073** -0.041 0.003
OthersMale 0.448 0.580 0.442 0.465 0.426 -0.132** 0.006 -0.017 0.022HH Members 1.358 1.280 1.327 0.853 0.883 0.078 0.031 0.505*** 0.475***Revolver 0.402 0.285 0.518 0.393 0.514 0.117*** -0.116*** 0.009 -0.112***Home owner 0.692 0.693 0.738 0.736 0.790 -0.001 -0.046 -0.044 -0.098***
Note: N-I and I denote Non-Innovators and Innovators, respectively. HH refers to household.Means correspond to the 2009–2013 year balanced panel. Survey weights used. T-tests ofmean differences of innovators and non-innovators are displayed. Differences can differ fromtrue values due to rounding and weighting. Significance levels are denoted as *** p<0.01, **p<0.05, * p<0.1.
117
Table A2: Means of Demographics for Types of Contactless Debit CardAdopters
Variable N-I; N-I; I; I; Multiple t-TestN-I I N-I I Switcher Mean Diff.(1) (2) (3) (4) (5) (1)-(2) (1)-(3) (1)-(4) (1)-(5)
Income (in 1000)<25 0.174 0.215 0.189 0 0.315 -0.041 -0.015 0.174*** -0.141***25–49 0.265 0.383 0.369 0.142 0.292 -0.118** -0.104 0.123 -0.02750–74 0.214 0.127 0.122 0.611 0.167 0.087*** 0.092*** -0.397*** 0.047**75–99 0.144 0.144 0.063 0.072 0.09 0 0.081*** 0.072 0.054***100–124 0.091 0.036 0.121 0.13 0.06 0.055*** -0.03 -0.039 0.031**125–199 0.084 0.033 0.083 0.045 0.053 0.051*** 0.001 0.039 0.031***>200 0.028 0.061 0.053 0 0.023 -0.033* -0.025* 0.028 0.005
Education<High School 0.036 0.061 0.143 0 0.036 -0.025 -0.107** 0.036*** 0High School 0.365 0.422 0.339 0.598 0.422 -0.057 0.026 -0.233 -0.057**Some College 0.268 0.274 0.274 0.137 0.322 -0.006 -0.006 0.131* -0.054**College 0.179 0.137 0.186 0.265 0.142 0.042 -0.007 -0.086 0.037***Post Graduate 0.151 0.107 0.058 0 0.078 0.044 0.093*** 0.151*** 0.073***
EmploymentWorking 0.623 0.63 0.727 0.954 0.719 -0.007 -0.104*** -0.331*** -0.096***Retired 0.237 0.195 0.164 0 0.117 0.042 0.073** 0.237*** 0.12***Unemployed 0.054 0.05 0.039 0.046 0.102 0.004 0.015 0.008 -0.048***Other 0.183 0.233 0.158 0 0.138 -0.05* 0.025 0.183*** 0.045***
Marital StatusSingle 0.115 0.151 0.205 0.137 0.227 -0.036 -0.09 -0.022 -0.112***Married 0.677 0.633 0.593 0.863 0.604 0.044 0.084 -0.186** 0.073**Separated 0.143 0.167 0.179 0 0.155 -0.024 -0.036 0.143*** -0.012Widowed 0.064 0.05 0.023 0 0.015 0.014 0.041*** 0.064*** 0.049***
Age<25 0.025 0.113 0.171 0 0.07 -0.088 -0.146*** 0.025*** -0.045***25–34 0.114 0.209 0.212 0.265 0.26 -0.095** -0.098** -0.151 -0.146***35–44 0.17 0.15 0.085 0 0.194 0.02 0.085*** 0.17*** -0.02445–54 0.243 0.176 0.281 0.598 0.256 0.067 -0.038 -0.355** -0.01355–64 0.206 0.16 0.114 0.137 0.113 0.046 0.092*** 0.069 0.093***>65 0.242 0.192 0.138 0 0.107 0.05 0.104*** 0.242*** 0.135***
EthnicityWhite 0.822 0.761 0.662 0 0.605 0.061 0.16*** 0.822*** 0.217***Black 0.111 0.079 0.284 0.137 0.241 0.032 -0.173*** -0.026 -0.13***Asian 0.022 0.12 0.04 0.265 0.033 -0.098** -0.018 -0.243** -0.011Latino 0.066 0.053 0.112 0.598 0.182 0.013 -0.046* -0.532*** -0.116***Other 0.046 0.04 0.013 0.598 0.122 0.006 0.033*** -0.552*** -0.076**
OthersMale 0.441 0.298 0.36 0.735 0.548 0.143*** 0.081* -0.294** -0.107***HH Members 1.146 1.088 1.557 0.863 1.768 0.058* -0.411** 0.283*** -0.622***Revolver 0.442 0.403 0.404 0.402 0.34 0.039 0.038 0.04 0.102***Home owner 0.78 0.637 0.415 0.153 0.559 0.143*** 0.365*** 0.627*** 0.221***
Note: N-I and I denote Non-Innovators and Innovators, respectively. HH refers to household.Means correspond to the 2009–2013 year balanced panel. Survey weights used. T-tests ofmean differences of innovators and non-innovators are displayed. Differences can differ fromtrue values due to rounding and weighting. Significance levels are denoted as *** p<0.01, **p<0.05, * p<0.1.
118
Table A3: OLS Regression Results of Contactless Credit on Usual CashWithdrawn
(1) (2) (3)
Variable b se b se b se
Contactless Credit −2.051 (9.852 ) −3.671 (9.804 ) −2.725 (9.037 )log(Income) 27.989*** (7.620 ) 29.230*** (7.844 ) 34.260*** (6.772 )Interest Rate 16.764 (14.906 ) 18.420 (15.561 ) 10.115 (11.851 )Education
High School −80.447 (49.537 ) −10.565 (25.896 ) 18.129 (26.192 )Some College −92.320 * (49.992 ) −19.752 (25.510 ) 16.727 (25.487 )College −101.321 ** (49.932 ) −27.025 (26.260 ) 14.866 (26.242 )Post Graduate −86.791 * (50.073 ) −10.980 (27.548 ) 31.580 (27.147 )
EmploymentWorking −32.013 ** (14.387 ) −32.964 ** (13.799 ) −33.698 *** (12.919 )Retired 12.718 (15.003 ) 14.459 (15.417 ) 16.923 (15.036 )Other 13.138 (13.781 ) 12.103 (13.360 ) 12.492 (12.723 )
Marital StatusSingle 22.259 (27.799 ) 26.889 (26.640 ) 27.128 (21.291 )Married −34.788 * (20.993 ) −28.117 (19.616 ) −16.867 (15.658 )Separated −21.060 (21.972 ) −23.293 (20.402 ) −1.323 (16.527 )
Age25–34 77.567*** (29.342 ) 58.188** (27.339 ) 63.518*** (24.502 )35–44 81.156*** (22.157 ) 63.788*** (22.716 ) 72.647*** (22.019 )45–54 84.508*** (21.908 ) 68.826*** (22.346 ) 75.851*** (21.902 )55–64 100.442*** (24.459 ) 81.699*** (23.983 ) 80.444*** (23.053 )>65 87.765*** (27.770 ) 66.516** (28.031 ) 67.132** (26.642 )
EthnicityWhite 0.599 (25.990 ) −10.569 (28.341 ) −11.162 (24.907 )Black 21.347 (30.835 ) −12.969 (31.957 ) −12.829 (28.993 )Latino 5.441 (13.454 ) 13.392 (16.413 ) 13.657 (13.777 )Other −0.220 (31.667 ) −13.243 (34.102 ) −5.569 (30.846 )
OtherMale 32.816*** (9.640 ) 34.889*** (10.075 ) 30.154*** (9.080 )HH Members 4.105 (4.086 ) 3.534 (3.755 ) 1.943 (3.335 )CC Revolver −45.418 *** (8.038 ) −37.644 *** (7.594 ) −29.303 *** (7.009 )Home owner 3.156 (11.324 ) 6.875 (11.468 ) −5.403 (9.579 )
Rel. CharacteristicsSecurity 7.176 (5.349 ) 2.119 (4.920 )Setup 16.603 (10.429 ) 2.588 (8.552 )Acceptance −0.405 (11.519 ) −10.125 (10.486 )Cost −7.292 (10.009 ) −13.207 (9.162 )Records 20.068*** (6.863 ) 13.649** (6.082 )Convenience 29.927*** (9.259 ) 25.866*** (8.369 )
Withdrawal MethodBank teller 85.927*** (11.478 )Check casher 169.979** (70.484 )Cashback −53.647 *** (5.355 )Employer 211.834*** (48.504 )Family −12.027 (23.349 )Other 128.990*** (37.643 )
Constant −145.009 (99.291 ) −141.429 (92.973 ) −298.318 *** (83.571 )
R2 0.080 0.092 0.208Observations 4734 3768 3768Individuals 1119 1110 1110
Note: OLS is the pooled OLS estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors and survey weights areused. HH and CC refers to household and credit card, respectively. Base category ofcategorical variables is lower than high school, unemployed, widowed, lower than 25 years,Asian, and ATM. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
119
Table A4: OLS Regression Results of Contactless Credit on Number of With-drawals
(1) (2) (3)
Variable b se b se b se
Contactless Credit −0.598** (0.233 ) −0.554* (0.282 ) −0.456* (0.272 )log(Income) 0.056 (0.147 ) 0.104 (0.177 ) 0.138 (0.179 )Interest Rate 0.118 (0.237 ) 0.088 (0.293 ) −0.009 (0.298 )Education
High School 0.415 (0.663 ) 0.286 (0.872 ) 0.178 (0.826 )Some College 0.190 (0.649 ) −0.103 (0.837 ) −0.218 (0.788 )College −0.075 (0.664 ) −0.254 (0.869 ) −0.408 (0.824 )Post Graduate −0.258 (0.692 ) −0.498 (0.894 ) −0.737 (0.859 )
EmploymentWorking 0.302 (0.478 ) 0.433 (0.483 ) 0.433 (0.462 )Retired −0.605 (0.487 ) −0.673 (0.553 ) −0.589 (0.541 )Other −0.564 (0.375 ) −0.561 (0.389 ) −0.458 (0.393 )
Marital StatusSingle 0.030 (0.841 ) 0.046 (1.046 ) 0.025 (1.052 )Married −0.609 (0.712 ) −0.702 (0.883 ) −0.633 (0.892 )Separated −0.810 (0.763 ) −0.980 (0.936 ) −0.940 (0.963 )
Age25–34 0.729 (0.775 ) −0.553 (1.456 ) −0.651 (1.432 )35–44 1.140 (0.784 ) −0.009 (1.463 ) −0.229 (1.422 )45–54 2.271*** (0.797 ) 1.201 (1.469 ) 1.124 (1.421 )55–64 1.922** (0.810 ) 0.778 (1.466 ) 0.801 (1.444 )>65 2.292** (0.945 ) 1.406 (1.599 ) 1.453 (1.595 )
EthnicityWhite 0.395 (0.452 ) 0.256 (0.435 ) 0.298 (0.439 )Black 1.246* (0.651 ) 1.277* (0.737 ) 1.323* (0.727 )Latino 0.037 (0.431 ) −0.121 (0.451 ) −0.139 (0.446 )Other 1.098 (0.847 ) 0.921 (0.809 ) 0.869 (0.790 )
OtherMale 0.029 (0.260 ) 0.086 (0.296 ) 0.029 (0.288 )HH Members 0.169* (0.101 ) 0.173 (0.122 ) 0.142 (0.123 )CC Revolver 0.358 (0.228 ) 0.211 (0.273 ) 0.214 (0.271 )Home owner −0.866** (0.384 ) −0.919** (0.457 ) −0.855* (0.452 )
Rel. CharacteristicsSecurity 0.071 (0.179 ) 0.041 (0.189 )Setup 0.086 (0.346 ) 0.035 (0.334 )Acceptance −0.620* (0.368 ) −0.501 (0.366 )Cost 0.803** (0.407 ) 0.921** (0.424 )Records 0.127 (0.196 ) 0.147 (0.201 )Convenience 0.766*** (0.247 ) 0.832*** (0.250 )
Withdrawal MethodBank teller −1.038*** (0.280 )Check casher −1.968*** (0.490 )Cashback 0.117 (0.322 )Employer 1.292 (0.952 )Family −0.186 (1.295 )Other 2.767 (1.981 )
Constant 1.506 (1.903 ) 3.424 (2.519 ) 3.436 (2.472 )
R2 0.020 0.026 0.036Observations 4745 3778 3777Individuals 1119 1112 1111
Note: OLS is the pooled OLS estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors and survey weights areused. HH and CC refers to household and credit card, respectively. Base category ofcategorical variables is lower than high school, unemployed, widowed, lower than 25 years,Asian, and ATM. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
120
Table A5: OLS Regression Results of Contactless Credit on Cash in Wallet
(1) (2) (3)
Variable b se b se b se
Contactless Credit 6.950 (8.089 ) 12.569 (9.823 ) 13.257 (9.842 )log(Income) 17.370*** (5.674 ) 17.972*** (6.361 ) 18.995*** (6.301 )Interest Rate −1.368 (5.748 ) −5.217 (6.178 ) −4.021 (6.094 )Education
High School 9.171 (19.222 ) 25.972 (25.557 ) 33.368 (24.701 )Some College −0.567 (17.783 ) 15.962 (23.492 ) 24.393 (22.809 )College 4.701 (17.937 ) 23.566 (23.713 ) 35.401 (23.278 )Post Graduate −2.764 (18.954 ) 17.628 (25.160 ) 29.751 (24.846 )
EmploymentWorking −6.657 (17.594 ) −7.828 (17.795 ) −3.581 (17.440 )Retired 3.588 (17.667 ) 2.523 (18.414 ) 5.950 (17.741 )Other 23.925 (19.947 ) 27.452 (20.626 ) 28.374 (19.890 )
Marital StatusSingle 27.066* (14.715 ) 17.302 (15.069 ) 19.284 (14.544 )Married −1.527 (12.452 ) −0.415 (14.057 ) 3.620 (13.465 )Separated 10.193 (13.631 ) 10.771 (15.108 ) 17.962 (14.999 )
Age25–34 5.128 (16.027 ) −9.965 (16.108 ) −12.254 (14.802 )35–44 23.745 (17.518 ) 6.918 (17.831 ) 3.166 (16.457 )45–54 25.842 (16.128 ) 1.436 (17.062 ) −2.175 (15.898 )55–64 39.256** (16.860 ) 19.238 (18.706 ) 11.869 (17.165 )>65 44.129** (18.504 ) 21.724 (20.216 ) 13.699 (18.908 )
EthnicityWhite 13.895 (11.152 ) 14.019 (11.735 ) 13.869 (11.217 )Black 16.915 (13.139 ) 17.943 (13.579 ) 16.543 (13.235 )Latino −1.430 (8.244 ) 3.257 (9.171 ) 5.873 (9.435 )Other 11.722 (15.613 ) 8.123 (16.828 ) 15.068 (16.201 )
OtherMale 40.426*** (7.813 ) 45.841*** (8.516 ) 44.979*** (8.262 )HH Members −0.245 (2.378 ) −1.436 (2.602 ) −1.922 (2.492 )CC Revolver −22.370 *** (5.647 ) −21.728 *** (5.838 ) −18.528 *** (5.512 )Home owner 22.654*** (6.647 ) 24.399*** (7.431 ) 19.198*** (7.083 )
Rel. CharacteristicsSecurity −2.807 (3.605 ) −4.645 (3.710 )Setup 1.541 (5.656 ) −2.708 (5.683 )Acceptance −9.453 (8.452 ) −14.324 * (8.528 )Cost −19.480 ** (9.459 ) −20.301 ** (9.170 )Records 12.747** (5.213 ) 10.287** (4.989 )Convenience 21.919*** (7.325 ) 20.945*** (7.201 )
Withdrawal MethodBank teller 41.173*** (8.027 )Check casher 13.049 (23.448 )Cashback −12.438 *** (4.807 )Employer 11.316 (10.836 )Family 28.640 (18.869 )Other 109.867** (47.782 )
Constant −192.266 *** (59.847 ) −177.431 *** (67.039 ) −224.449 *** (67.110 )
R2 0.062 0.076 0.108Observations 4741 3777 3777Individuals 1117 1109 1109
Note: OLS is the pooled OLS estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors and survey weights areused. HH and CC refers to household and credit card, respectively. Base category ofcategorical variables is lower than high school, unemployed, widowed, lower than 25 years,Asian, and ATM. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
121
Table A6: FE Regression Results of Contactless Credit on Usual Cash With-drawn
(1) (2) (3)
Variable b se b se b se
Contactless Credit −4.067 (5.912 ) −1.065 (5.564 ) −1.773 (5.429 )log(Income) 3.298 (11.210 ) −1.472 (17.331 ) −2.342 (16.851 )Interest Rate −4.047 (4.324 ) −5.239 (5.436 ) −6.238 (5.295 )Education
High School 37.105 (36.530 ) 59.014 (42.931 ) 1.215 (54.225 )Some College 37.649 (43.229 ) 33.532 (55.509 ) −22.264 (60.018 )College 17.834 (47.059 ) 6.583 (59.358 ) −42.088 (62.621 )Post Graduate 99.053 (97.167 ) 125.525 ( 143.207 ) 72.915 ( 134.893 )
EmploymentWorking 5.116 (7.851 ) 3.808 (9.010 ) −2.756 (8.848 )Retired 30.262** (13.355 ) 14.954 (17.680 ) 15.468 (16.047 )Other 12.373 (9.544 ) 2.269 (13.204 ) 2.517 (12.285 )
Marital StatusSingle −16.936 (30.942 ) 13.113 (20.666 ) 10.381 (18.139 )Married −24.834 (28.549 ) 3.411 (12.845 ) 8.225 (11.983 )Separated −44.904 (33.635 ) −19.885 (21.564 ) −12.709 (21.260 )
Age25–34 0.043 (18.657 ) −27.787 (33.448 ) −21.711 (22.661 )35–44 11.337 (24.225 ) −21.557 (40.207 ) −14.741 (30.167 )45–54 27.280 (27.773 ) −4.907 (42.413 ) 1.614 (32.023 )55–64 35.397 (30.561 ) −7.577 (44.864 ) 2.490 (35.210 )>65 −0.912 (57.039 ) −76.900 (78.575 ) −68.593 (74.079 )
OtherHH Members 1.030 (4.223 ) −0.666 (4.469 ) −0.516 (4.429 )CC Revolver −8.604 (5.234 ) −4.434 (6.186 ) −2.335 (5.662 )Home owner −8.797 (10.550 ) −18.817 (13.905 ) −18.082 (13.807 )
Rel. CharacteristicsSecurity −6.775 (6.142 ) −6.293 (5.939 )Setup 8.329* (4.575 ) 3.632 (4.515 )Acceptance 13.180 (8.831 ) 15.850* (8.307 )Cost 5.229 (9.071 ) 3.090 (8.573 )Records −3.691 (5.058 ) −1.962 (4.784 )Convenience 1.221 (7.413 ) 1.818 (7.341 )
Withdrawal MethodBank teller 75.429*** (17.369 )Check casher 140.405** (58.746 )Cashback −32.260 *** (9.240 )Employer 109.707** (48.246 )Family −2.515 (22.229 )Other 111.719*** (32.173 )
Constant 49.167 ( 133.445 ) 132.954 ( 203.999 ) 165.169 ( 206.235 )
R2 0.006 0.011 0.055Observations 4734 3768 3768Individuals 1119 1110 1110
Note: FE is the fixed-effects estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors are used. HH and CCrefers to household and credit card, respectively. Base category of categorical variables islower than high school, unemployed, widowed, lower than 25 years, Asian, and ATM.Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
122
Table A7: FE Regression Results of Contactless Credit on Number of With-drawals
(1) (2) (3)
Variable b se b se b se
Contactless Credit −0.649** (0.260 ) −0.862** (0.364 ) −0.866** (0.366 )log(Income) 0.528* (0.280 ) 0.455 (0.452 ) 0.483 (0.453 )Interest Rate 0.121 (0.245 ) 0.213 (0.401 ) 0.171 (0.405 )Education
High School 1.107** (0.514 ) 0.958 (1.084 ) 1.935* (1.112 )Some College 0.432 (1.125 ) −0.295 (2.653 ) 0.744 (2.733 )College 1.081 (1.428 ) 0.187 (3.082 ) 1.124 (3.127 )Post Graduate 1.770 (1.375 ) 0.554 (3.163 ) 1.578 (3.206 )
EmploymentWorking −0.598 (0.432 ) −0.406 (0.591 ) −0.345 (0.583 )Retired 0.116 (0.805 ) −0.004 (2.058 ) 0.203 (2.070 )Other −0.244 (0.367 ) −0.110 (0.582 ) −0.146 (0.572 )
Marital StatusSingle 0.761 (1.599 ) 1.842 (2.324 ) 2.010 (2.358 )Married −0.771 (0.618 ) −0.681 (0.719 ) −0.631 (0.691 )Separated −2.176 (1.664 ) −2.746 (2.135 ) −2.660 (2.135 )
Age25–34 1.108 (0.863 ) 0.985 (1.406 ) 0.881 (1.194 )35–44 1.502 (0.951 ) 1.983 (1.467 ) 2.022 (1.267 )45–54 1.781* (0.988 ) 2.211 (1.563 ) 2.196 (1.362 )55–64 2.483* (1.297 ) 2.828 (1.988 ) 2.710 (1.842 )>65 2.834* (1.667 ) 3.423 (2.479 ) 3.222 (2.363 )
OtherHH Members 0.024 (0.108 ) 0.113 (0.155 ) 0.094 (0.152 )CC Revolver 0.772 (0.937 ) 1.517 (1.700 ) 1.478 (1.706 )Home owner −0.238 (0.702 ) −0.091 (0.893 ) −0.133 (0.892 )
Rel. CharacteristicsSecurity −0.401 (0.537 ) −0.423 (0.539 )Setup 0.541 (0.555 ) 0.535 (0.557 )Acceptance −0.806 (0.542 ) −0.769 (0.539 )Cost 1.166 (1.155 ) 1.184 (1.158 )Records 0.227 (0.280 ) 0.178 (0.277 )Convenience −0.344 (0.528 ) −0.333 (0.529 )
Withdrawal MethodBank teller −1.229*** (0.418 )Check casher −2.236 (1.528 )Cashback 0.289 (0.389 )Employer 0.024 (1.215 )Family −0.570 (0.833 )Other 1.990* (1.088 )
Constant −4.132 (3.364 ) −3.722 (5.285 ) −4.817 (5.291 )
R2 0.003 0.008 0.011Observations 4745 3778 3777Individuals 1119 1112 1111
Note: FE is the fixed-effects estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors are used. HH and CCrefers to household and credit card, respectively. Base category of categorical variables islower than high school, unemployed, widowed, lower than 25 years, Asian, and ATM.Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
123
Table A8: FE Regression Results of Contactless Credit on Cash in Wallet
(1) (2) (3)
Variable b se b se b se
Contactless Credit 5.206 (6.154 ) 5.417 (7.191 ) 5.446 (7.231 )log(Income) 25.438*** (9.428 ) 15.825** (6.425 ) 16.041** (6.400 )Interest Rate −4.809 (4.020 ) −7.218 (5.570 ) −6.368 (5.457 )Education
High School 18.731** (8.551 ) 44.677** (19.910 ) 23.463 (18.377 )Some College 25.473 (17.441 ) 55.661 (34.395 ) 28.902 (34.335 )College 50.114** (23.168 ) 71.950* (39.839 ) 46.776 (40.666 )Post Graduate 71.879** (34.577 ) 81.254* (48.095 ) 52.418 (47.432 )
EmploymentWorking −6.268 (9.623 ) −3.211 (7.962 ) −2.401 (7.794 )Retired −23.436 (21.822 ) 8.581 (15.329 ) 7.886 (15.903 )Other 7.369 (11.538 ) −3.316 (9.037 ) −2.506 (9.144 )
Marital StatusSingle −0.082 (23.291 ) 26.317 (33.088 ) 29.650 (33.189 )Married −3.901 (21.936 ) 14.218 (31.499 ) 18.548 (31.521 )Separated −33.440 (30.041 ) −24.166 (40.373 ) −20.479 (39.968 )
Age25–34 −16.353 (11.413 ) −35.569 ** (16.415 ) −35.543 ** (17.095 )35–44 −14.611 (14.556 ) −26.662 (19.930 ) −25.613 (20.404 )45–54 −13.204 (16.171 ) −27.679 (21.768 ) −25.098 (22.160 )55–64 2.504 (20.642 ) −1.646 (28.982 ) 0.728 (29.263 )>65 49.614 (40.069 ) 50.195 (51.144 ) 54.060 (52.464 )
OtherHH Members −0.045 (2.826 ) −1.572 (3.087 ) −1.873 (3.093 )CC Revolver −2.617 (5.113 ) −0.994 (5.920 ) −1.793 (6.055 )Home owner −0.385 (7.936 ) −7.489 (9.163 ) −8.860 (9.328 )
Rel. CharacteristicsSecurity −2.638 (2.508 ) −1.962 (2.503 )Setup −2.664 (6.130 ) −3.042 (5.931 )Acceptance −1.630 (6.894 ) −1.262 (7.141 )Cost −1.996 (5.983 ) −2.336 (5.994 )Records 2.236 (4.688 ) 2.622 (4.682 )Convenience −0.881 (5.144 ) −0.817 (5.209 )
Withdrawal MethodBank teller 26.523*** (9.892 )Check casher 25.003 (20.471 )Cashback −2.933 (6.255 )Employer −46.294 (42.550 )Family 23.599 (27.059 )Other 29.353 (31.258 )
Constant −229.622 ** ( 100.832 ) −159.911 * (83.000 ) −148.203 * (82.714 )
R2 0.013 0.012 0.021Observations 4741 3777 3777Individuals 1117 1109 1109
Note: FE is the fixed-effects estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors are used. HH and CCrefers to household and credit card, respectively. Base category of categorical variables islower than high school, unemployed, widowed, lower than 25 years, Asian, and ATM.Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
124
Table A9: OLS Regression Results of Contactless Debit on Usual Cash With-drawn
(1) (2) (3)
Variable b se b se b se
Contactless Debit 2.196 (18.450 ) 3.139 (16.517 ) 17.018 (12.329 )log(Income) 27.946*** (7.590 ) 29.161*** (7.830 ) 34.432*** (6.776 )Interest Rate 16.997 (14.388 ) 18.315 (15.158 ) 9.387 (11.615 )Education
High School −79.280 (49.397 ) −9.667 (26.154 ) 20.363 (26.010 )Some College −91.560 * (50.046 ) −18.802 (25.595 ) 19.551 (25.371 )College −100.621 ** (49.959 ) −25.946 (26.263 ) 17.944 (26.047 )Post Graduate −85.938 * (50.557 ) −9.962 (27.954 ) 35.056 (27.293 )
EmploymentWorking −32.044 ** (14.249 ) −33.169 ** (13.754 ) −34.295 *** (12.940 )Retired 12.684 (15.003 ) 14.426 (15.436 ) 16.940 (15.069 )Other 13.081 (13.847 ) 12.073 (13.388 ) 12.307 (12.776 )
Marital StatusSingle 21.628 (27.531 ) 27.043 (26.450 ) 27.011 (21.018 )Married −34.602 * (20.889 ) −27.957 (19.594 ) −17.140 (15.673 )Separated −20.828 (22.068 ) −23.308 (20.403 ) −1.773 (16.534 )
Age25–34 77.802*** (29.219 ) 57.679** (26.688 ) 61.933*** (23.748 )35–44 80.801*** (22.210 ) 63.458*** (22.781 ) 72.388*** (21.813 )45–54 84.139*** (21.595 ) 68.543*** (22.232 ) 74.931*** (21.534 )55–64 100.256*** (24.324 ) 81.629*** (23.963 ) 80.073*** (22.717 )>65 87.507*** (27.638 ) 66.396** (28.063 ) 66.832** (26.362 )
EthnicityWhite 1.289 (25.947 ) −9.251 (28.136 ) −7.442 (25.144 )Black 22.184 (31.331 ) −11.794 (31.941 ) −10.432 (29.186 )Latino 6.797 (14.010 ) 13.022 (16.163 ) 12.177 (13.827 )Other −1.130 (32.093 ) −12.347 (34.603 ) −3.093 (31.388 )
OtherMale 32.759*** (9.673 ) 34.793*** (10.128 ) 29.635*** (9.144 )HH Members 4.127 (4.040 ) 3.622 (3.726 ) 1.988 (3.285 )CC Revolver −45.574 *** (8.105 ) −37.616 *** (7.526 ) −28.642 *** (6.984 )Home owner 3.101 (11.593 ) 7.163 (11.768 ) −4.097 (9.499 )
Rel. CharacteristicsSecurity 7.114 (5.287 ) 1.765 (4.856 )Setup 16.670 (10.536 ) 2.830 (8.523 )Acceptance −0.459 (11.552 ) −10.194 (10.495 )Cost −7.355 (9.995 ) −13.568 (9.193 )Records 20.010*** (6.896 ) 13.112** (6.101 )Convenience 30.089*** (9.153 ) 26.701*** (8.335 )
Withdrawal MethodBank teller 86.805*** (11.481 )Check casher 169.717** (71.232 )Cashback −53.956 *** (5.440 )Employer 212.158*** (47.451 )Family −10.214 (23.271 )Other 130.638*** (37.571 )
Constant −146.223 ( 100.944 ) −143.599 (92.512 ) −308.820 *** (84.522 )
R2 0.079 0.092 0.208Observations 4732 3767 3767Individuals 1119 1110 1110
Note: OLS is the pooled OLS estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors and survey weights areused. HH and CC refers to household and credit card, respectively. Base category ofcategorical variables is lower than high school, unemployed, widowed, lower than 25 years,Asian, and ATM. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
125
Table A10: OLS Regression Results of Contactless Debit on Number ofWithdrawals
(1) (2) (3)
Variable b se b se b se
Contactless Debit 0.191 (0.320 ) 0.060 (0.367 ) 0.023 (0.361 )log(Income) 0.045 (0.143 ) 0.087 (0.172 ) 0.125 (0.174 )Interest Rate 0.110 (0.234 ) 0.092 (0.291 ) −0.005 (0.295 )Education
High School 0.529 (0.677 ) 0.384 (0.875 ) 0.257 (0.828 )Some College 0.297 (0.665 ) −0.012 (0.848 ) −0.146 (0.797 )College 0.038 (0.678 ) −0.155 (0.876 ) −0.330 (0.831 )Post Graduate −0.157 (0.707 ) −0.410 (0.906 ) −0.669 (0.871 )
EmploymentWorking 0.284 (0.478 ) 0.415 (0.481 ) 0.421 (0.460 )Retired −0.608 (0.486 ) −0.678 (0.552 ) −0.591 (0.541 )Other −0.561 (0.376 ) −0.560 (0.389 ) −0.455 (0.393 )
Marital StatusSingle 0.082 (0.844 ) 0.075 (1.046 ) 0.048 (1.052 )Married −0.572 (0.712 ) −0.667 (0.880 ) −0.604 (0.889 )Separated −0.785 (0.767 ) −0.967 (0.937 ) −0.928 (0.962 )
Age25–34 0.727 (0.780 ) −0.593 (1.452 ) −0.680 (1.429 )35–44 1.138 (0.790 ) −0.054 (1.462 ) −0.265 (1.419 )45–54 2.283*** (0.801 ) 1.171 (1.468 ) 1.101 (1.420 )55–64 1.959** (0.816 ) 0.772 (1.468 ) 0.800 (1.446 )>65 2.319** (0.951 ) 1.389 (1.600 ) 1.440 (1.596 )
EthnicityWhite 0.554 (0.442 ) 0.370 (0.434 ) 0.388 (0.436 )Black 1.405** (0.654 ) 1.404* (0.733 ) 1.425** (0.723 )Latino 0.009 (0.451 ) −0.138 (0.450 ) −0.149 (0.447 )Other 1.219 (0.848 ) 1.002 (0.796 ) 0.930 (0.779 )
OtherMale 0.031 (0.261 ) 0.089 (0.297 ) 0.033 (0.289 )HH Members 0.180* (0.099 ) 0.185 (0.119 ) 0.151 (0.121 )CC Revolver 0.345 (0.227 ) 0.205 (0.271 ) 0.209 (0.269 )Home owner −0.850** (0.387 ) −0.910** (0.459 ) −0.850* (0.455 )
Rel. CharacteristicsSecurity 0.072 (0.177 ) 0.042 (0.188 )Setup 0.086 (0.346 ) 0.033 (0.334 )Acceptance −0.636* (0.364 ) −0.515 (0.363 )Cost 0.804** (0.410 ) 0.924** (0.426 )Records 0.133 (0.197 ) 0.153 (0.201 )Convenience 0.780*** (0.244 ) 0.844*** (0.248 )
Withdrawal MethodBank teller −1.038*** (0.281 )Check casher −1.922*** (0.493 )Cashback 0.129 (0.321 )Employer 1.316 (0.954 )Family −0.156 (1.289 )Other 2.795 (1.985 )
Constant 1.223 (1.903 ) 3.337 (2.507 ) 3.365 (2.444 )
R2 0.019 0.026 0.036Observations 4744 3777 3776Individuals 1119 1112 1111
Note: Note: OLS is the pooled OLS estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors and survey weights areused. HH and CC refers to household and credit card, respectively. Base category ofcategorical variables is lower than high school, unemployed, widowed, lower than 25 years,Asian, and ATM. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
126
Table A11: OLS Regression Results of Contactless Debit on Cash in Wallet
(1) (2) (3)
Variable b se b se b se
Contactless Debit 19.375 (15.483 ) 15.329 (15.050 ) 24.001 (14.954 )log(Income) 17.849*** (5.611 ) 18.594*** (6.295 ) 19.774*** (6.229 )Interest Rate −2.273 (5.370 ) −6.069 (5.959 ) −5.202 (5.850 )Education
High School 9.798 (18.993 ) 25.289 (25.451 ) 33.643 (24.105 )Some College 0.474 (17.379 ) 16.054 (23.500 ) 25.772 (22.376 )College 5.729 (17.510 ) 23.452 (23.653 ) 36.682 (22.810 )Post Graduate −0.765 (18.458 ) 18.468 (25.001 ) 32.327 (24.350 )
EmploymentWorking −7.221 (17.639 ) −8.095 (17.820 ) −4.051 (17.429 )Retired 3.421 (17.733 ) 2.527 (18.447 ) 6.005 (17.707 )Other 23.729 (20.020 ) 27.269 (20.678 ) 28.063 (19.935 )
Marital StatusSingle 25.728* (14.539 ) 16.343 (15.065 ) 18.140 (14.485 )Married −2.757 (12.498 ) −1.717 (14.068 ) 2.060 (13.420 )Separated 8.992 (13.313 ) 9.832 (14.854 ) 16.823 (14.701 )
Age25–34 4.373 (15.896 ) −10.487 (16.493 ) −13.545 (15.444 )35–44 23.964 (17.685 ) 7.889 (18.514 ) 4.153 (17.381 )45–54 24.644 (16.248 ) 1.162 (17.695 ) −2.926 (16.874 )55–64 38.585** (16.951 ) 19.003 (19.192 ) 11.267 (17.949 )>65 43.532** (18.579 ) 21.921 (20.718 ) 13.617 (19.630 )
EthnicityWhite 15.626 (11.133 ) 14.168 (11.715 ) 15.487 (11.263 )Black 17.288 (13.253 ) 16.506 (13.619 ) 15.804 (13.335 )Latino −1.440 (9.186 ) 2.177 (9.467 ) 4.141 (9.766 )Other 11.338 (16.021 ) 7.996 (16.993 ) 16.081 (16.531 )
OtherMale 40.102*** (7.726 ) 45.335*** (8.391 ) 44.206*** (8.108 )HH Members −0.491 (2.316 ) −1.729 (2.527 ) −2.243 (2.406 )CC Revolver −21.680 *** (5.607 ) −20.873 *** (5.760 ) −17.223 *** (5.436 )Home owner 24.335*** (7.149 ) 25.579*** (7.812 ) 20.913*** (7.402 )
Rel. CharacteristicsSecurity −3.109 (3.587 ) −5.137 (3.717 )Setup 1.872 (5.604 ) −2.276 (5.641 )Acceptance −9.059 (8.488 ) −14.061 (8.571 )Cost −19.773 ** (9.566 ) −20.794 ** (9.293 )Records 12.147** (5.191 ) 9.346* (4.952 )Convenience 22.503*** (7.301 ) 21.845*** (7.209 )
Withdrawal MethodBank teller 42.501*** (8.103 )Check casher 11.144 (23.821 )Cashback −13.230 *** (5.004 )Employer 10.984 (10.201 )Family 30.180 (18.915 )Other 111.282** (47.714 )
Constant −199.862 *** (59.357 ) −183.977 *** (66.836 ) −236.660 *** (67.218 )
R2 0.063 0.077 0.109Observations 4740 3777 3777Individuals 1117 1109 1109
Note: OLS is the pooled OLS estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors and survey weights areused. HH and CC refers to household and credit card, respectively. Base category ofcategorical variables is lower than high school, unemployed, widowed, lower than 25 years,Asian, and ATM. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
127
Table A12: FE Regression Results of Contactless Debit on Usual Cash With-drawn
(1) (2) (3)
Variable b se b se b se
Contactless Debit 2.449 (5.747 ) 3.067 (6.380 ) 5.453 (6.279 )log(Income) 3.255 (11.209 ) −1.529 (17.353 ) −2.451 (16.878 )Interest Rate −3.993 (4.348 ) −5.242 (5.436 ) −6.245 (5.295 )Education
High School 36.495 (36.271 ) 58.996 (42.945 ) 1.132 (54.260 )Some College 36.829 (43.009 ) 34.063 (55.548 ) −21.337 (59.995 )College 16.446 (46.849 ) 6.908 (59.372 ) −41.514 (62.608 )Post Graduate 97.879 (97.044 ) 125.897 ( 143.199 ) 73.578 ( 134.836 )
EmploymentWorking 5.140 (7.862 ) 3.843 (9.013 ) −2.712 (8.849 )Retired 30.427** (13.403 ) 15.027 (17.678 ) 15.595 (16.042 )Other 12.305 (9.555 ) 2.339 (13.199 ) 2.623 (12.270 )
Marital StatusSingle −16.675 (31.021 ) 13.251 (20.699 ) 10.613 (18.181 )Married −24.809 (28.572 ) 3.394 (12.822 ) 8.195 (11.933 )Separated −45.046 (33.629 ) −19.826 (21.546 ) −12.611 (21.220 )
Age25–34 0.127 (18.687 ) −27.819 (33.606 ) −21.791 (22.855 )35–44 11.587 (24.221 ) −21.634 (40.321 ) −14.924 (30.301 )45–54 27.577 (27.743 ) −5.054 (42.508 ) 1.301 (32.123 )55–64 35.760 (30.534 ) −7.621 (44.982 ) 2.381 (35.325 )>65 −0.686 (57.095 ) −76.989 (78.669 ) −68.786 (74.179 )
OtherHH Members 0.431 (4.256 ) −0.631 (4.470 ) −0.451 (4.428 )CC Revolver −8.850* (5.224 ) −4.479 (6.175 ) −2.405 (5.661 )Home owner −8.840 (10.621 ) −18.570 (13.904 ) −17.640 (13.822 )
Rel. CharacteristicsSecurity −6.799 (6.160 ) −6.335 (5.957 )Setup 8.329* (4.583 ) 3.631 (4.525 )Acceptance 13.132 (8.832 ) 15.763* (8.306 )Cost 5.262 (9.083 ) 3.144 (8.583 )Records −3.731 (5.063 ) −2.030 (4.790 )Convenience 1.273 (7.436 ) 1.929 (7.371 )
Withdrawal MethodBank teller 75.497*** (17.383 )Check casher 139.873** (59.037 )Cashback −32.341 *** (9.228 )Employer 109.973** (48.289 )Family −2.428 (22.246 )Other 111.755*** (32.176 )
Constant 50.388 ( 133.431 ) 132.644 ( 203.886 ) 164.759 ( 206.090 )
R2 0.006 0.011 0.055Observations 4732 3767 3767Individuals 1119 1110 1110
Note: FE is the fixed-effects estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors are used. HH and CCrefers to household and credit card, respectively. Base category of categorical variables islower than high school, unemployed, widowed, lower than 25 years, Asian, and ATM.Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
128
Table A13: FE Regression Results of Contactless Debit on Number of With-drawals
(1) (2) (3)
Variable b se b se b se
Contactless Debit −0.224 (0.328 ) 0.090 (0.347 ) 0.073 (0.345 )log(Income) 0.510* (0.280 ) 0.438 (0.453 ) 0.466 (0.454 )Interest Rate 0.122 (0.244 ) 0.215 (0.401 ) 0.174 (0.405 )Education
High School 1.176** (0.513 ) 0.945 (1.095 ) 1.926* (1.124 )Some College 0.467 (1.128 ) −0.295 (2.656 ) 0.750 (2.736 )College 1.096 (1.434 ) 0.169 (3.087 ) 1.112 (3.131 )Post Graduate 1.840 (1.395 ) 0.534 (3.170 ) 1.566 (3.212 )
EmploymentWorking −0.592 (0.432 ) −0.400 (0.590 ) −0.338 (0.581 )Retired 0.124 (0.803 ) 0.018 (2.057 ) 0.225 (2.069 )Other −0.262 (0.367 ) −0.115 (0.580 ) −0.152 (0.570 )
Marital StatusSingle 0.773 (1.620 ) 1.875 (2.370 ) 2.043 (2.405 )Married −0.759 (0.613 ) −0.665 (0.715 ) −0.616 (0.687 )Separated −2.185 (1.667 ) −2.734 (2.133 ) −2.648 (2.134 )
Age25–34 1.093 (0.861 ) 0.903 (1.409 ) 0.799 (1.200 )35–44 1.513 (0.951 ) 1.921 (1.469 ) 1.957 (1.269 )45–54 1.836* (0.991 ) 2.180 (1.568 ) 2.166 (1.368 )55–64 2.703** (1.351 ) 2.848 (1.994 ) 2.730 (1.848 )>65 3.064* (1.717 ) 3.447 (2.485 ) 3.246 (2.369 )
OtherHH Members 0.031 (0.111 ) 0.117 (0.154 ) 0.098 (0.151 )CC Revolver 0.763 (0.928 ) 1.462 (1.692 ) 1.423 (1.698 )Home owner −0.227 (0.702 ) −0.045 (0.894 ) −0.088 (0.893 )
Rel. CharacteristicsSecurity −0.391 (0.537 ) −0.414 (0.539 )Setup 0.520 (0.553 ) 0.514 (0.556 )Acceptance −0.818 (0.543 ) −0.781 (0.541 )Cost 1.178 (1.158 ) 1.196 (1.161 )Records 0.223 (0.280 ) 0.174 (0.277 )Convenience −0.316 (0.526 ) −0.304 (0.527 )
Withdrawal MethodBank teller −1.234*** (0.419 )Check casher −2.245 (1.526 )Cashback 0.277 (0.391 )Employer 0.021 (1.218 )Family −0.546 (0.825 )Other 1.981* (1.087 )
Constant −4.174 (3.383 ) −3.634 (5.304 ) −4.733 (5.314 )
R2 0.003 0.007 0.010Observations 4744 3777 3776Individuals 1119 1112 1111
Note: FE is the fixed-effects estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors are used. HH and CCrefers to household and credit card, respectively. Base category of categorical variables islower than high school, unemployed, widowed, lower than 25 years, Asian, and ATM.Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
129
Table A14: FE Regression Results of Contactless Debit on Cash in Wallet
(1) (2) (3)
Variable b se b se b se
Contactless Debit 21.241 (18.481 ) 22.860 (21.153 ) 22.755 (20.622 )log(Income) 25.488*** (9.460 ) 15.632** (6.471 ) 15.854** (6.446 )Interest Rate −4.883 (4.028 ) −7.277 (5.551 ) −6.437 (5.434 )Education
High School 14.547 (9.250 ) 44.707** (19.607 ) 23.370 (18.127 )Some College 22.228 (18.336 ) 59.827* (34.902 ) 33.024 (34.104 )College 44.898* (24.283 ) 74.760* (40.235 ) 49.627 (40.727 )Post Graduate 66.063* (35.422 ) 84.504* (48.301 ) 55.763 (47.275 )
EmploymentWorking −6.307 (9.623 ) −3.169 (7.958 ) −2.457 (7.797 )Retired −23.222 (21.550 ) 8.700 (15.378 ) 8.001 (15.923 )Other 7.443 (11.516 ) −2.749 (8.955 ) −2.005 (9.081 )
Marital StatusSingle −0.460 (23.026 ) 26.768 (32.602 ) 30.071 (32.741 )Married −4.146 (21.694 ) 13.841 (31.034 ) 18.160 (31.041 )Separated −33.637 (29.867 ) −23.867 (40.161 ) −20.206 (39.748 )
Age25–34 −15.219 (10.895 ) −34.454 ** (16.131 ) −34.598 ** (16.393 )35–44 −14.193 (14.242 ) −26.213 (19.706 ) −25.353 (19.798 )45–54 −13.743 (15.900 ) −28.265 (21.592 ) −25.904 (21.611 )55–64 1.990 (20.470 ) −2.047 (28.749 ) 0.136 (28.763 )>65 48.371 (39.401 ) 49.384 (50.298 ) 53.038 (51.343 )
OtherHH Members −0.484 (2.860 ) −1.352 (3.106 ) −1.642 (3.101 )CC Revolver −2.032 (5.109 ) −0.465 (5.998 ) −1.253 (6.122 )Home owner 0.429 (8.118 ) −6.269 (9.409 ) −7.628 (9.470 )
Rel. CharacteristicsSecurity −2.934 (2.546 ) −2.230 (2.508 )Setup −2.353 (5.999 ) −2.714 (5.819 )Acceptance −1.840 (7.042 ) −1.463 (7.279 )Cost −1.905 (5.911 ) −2.244 (5.923 )Records 2.040 (4.543 ) 2.416 (4.537 )Convenience −0.872 (5.191 ) −0.901 (5.243 )
Withdrawal MethodBank teller 26.656*** (10.002 )Check casher 22.786 (20.264 )Cashback −3.212 (6.272 )Employer −45.220 (41.458 )Family 23.455 (27.067 )Other 29.490 (31.133 )
Constant −226.729 ** ( 100.982 ) −163.580 ** (83.319 ) −151.728 * (82.749 )
R2 0.015 0.014 0.023Observations 4740 3777 3777Individuals 1117 1109 1109
Note: FE is the fixed-effects estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors are used. HH and CCrefers to household and credit card, respectively. Base category of categorical variables islower than high school, unemployed, widowed, lower than 25 years, Asian, and ATM.Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
130
Table A15: OLS Regression Results of Contactless Credit on Cash ShareVolume
(1) (2) (3)
Variable b se b se b se
Contactless Credit −0.054*** (0.015 ) −0.045*** (0.017 ) −0.046*** (0.016 )log(Income) −0.031*** (0.010 ) −0.015 (0.010 ) −0.013 (0.010 )Interest Rate 0.022 (0.019 ) 0.021 (0.017 ) 0.011 (0.017 )Education
High School −0.065 (0.056 ) −0.114* (0.059 ) −0.095* (0.058 )Some College −0.091* (0.055 ) −0.130** (0.059 ) −0.106* (0.057 )College −0.128** (0.056 ) −0.157*** (0.060 ) −0.135** (0.058 )Post Graduate −0.126** (0.057 ) −0.165*** (0.061 ) −0.145** (0.060 )
EmploymentWorking 0.012 (0.021 ) −0.002 (0.021 ) −0.004 (0.020 )Retired −0.009 (0.023 ) −0.009 (0.023 ) −0.006 (0.023 )Other 0.003 (0.022 ) 0.003 (0.022 ) 0.002 (0.021 )
Marital StatusSingle 0.037 (0.047 ) 0.013 (0.044 ) 0.009 (0.043 )Married −0.033 (0.036 ) −0.032 (0.035 ) −0.027 (0.034 )Separated −0.072* (0.039 ) −0.058 (0.038 ) −0.044 (0.036 )
Age25–34 0.005 (0.076 ) −0.062 (0.067 ) −0.050 (0.067 )35–44 0.030 (0.075 ) −0.036 (0.067 ) −0.018 (0.066 )45–54 0.084 (0.075 ) 0.008 (0.066 ) 0.026 (0.065 )55–64 0.079 (0.075 ) 0.002 (0.067 ) 0.014 (0.066 )>65 0.077 (0.077 ) −0.018 (0.069 ) 0.000 (0.069 )
EthnicityWhite 0.018 (0.030 ) −0.008 (0.032 ) 0.000 (0.032 )Black 0.054 (0.037 ) 0.024 (0.038 ) 0.031 (0.038 )Latino −0.001 (0.037 ) 0.000 (0.034 ) −0.009 (0.034 )Other 0.036 (0.058 ) 0.026 (0.057 ) 0.038 (0.056 )
OtherMale 0.052*** (0.016 ) 0.052*** (0.015 ) 0.046*** (0.014 )HH Members 0.005 (0.006 ) −0.001 (0.006 ) −0.002 (0.006 )CC Revolver −0.041*** (0.013 ) −0.035*** (0.012 ) −0.030** (0.012 )Home owner −0.034* (0.020 ) −0.035* (0.019 ) −0.039** (0.019 )
Rel. CharacteristicsSecurity 0.015* (0.008 ) 0.011 (0.008 )Setup 0.041*** (0.016 ) 0.033** (0.015 )Acceptance −0.020 (0.018 ) −0.023 (0.019 )Cost 0.054*** (0.019 ) 0.050*** (0.019 )Records 0.036*** (0.011 ) 0.033*** (0.011 )Convenience 0.096*** (0.016 ) 0.094*** (0.016 )
Withdrawal MethodBank teller 0.008 (0.016 )Check casher 0.067 (0.098 )Cashback −0.083*** (0.017 )Employer 0.149** (0.066 )Family 0.005 (0.047 )Other −0.010 (0.033 )
Constant 0.729*** (0.138 ) 0.889*** (0.135 ) 0.814*** (0.135 )
R2 0.093 0.154 0.177Observations 4595 3683 3681Individuals 1115 1109 1108
Note: OLS is the pooled OLS estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors and survey weights areused. HH and CC refers to household and credit card, respectively. Base category ofcategorical variables is lower than high school, unemployed, widowed, lower than 25 years,Asian, and ATM. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
131
Table A16: OLS Regression Results of Contactless Debit on Cash ShareVolume
(1) (2) (3)
Variable b se b se b se
Contactless Debit −0.057*** (0.021 ) −0.065*** (0.021 ) −0.062*** (0.020 )log(Income) −0.033*** (0.010 ) −0.017* (0.010 ) −0.016 (0.010 )Interest Rate 0.026 (0.019 ) 0.025 (0.017 ) 0.015 (0.017 )Education
High School −0.064 (0.054 ) −0.112* (0.057 ) −0.093* (0.056 )Some College −0.091* (0.054 ) −0.131** (0.057 ) −0.107* (0.055 )College −0.127** (0.055 ) −0.157*** (0.058 ) −0.136** (0.056 )Post Graduate −0.129** (0.057 ) −0.170*** (0.060 ) −0.149** (0.058 )
EmploymentWorking 0.012 (0.021 ) 0.000 (0.021 ) −0.003 (0.020 )Retired −0.009 (0.023 ) −0.009 (0.023 ) −0.006 (0.023 )Other 0.003 (0.021 ) 0.003 (0.021 ) 0.002 (0.021 )
Marital StatusSingle 0.047 (0.047 ) 0.018 (0.044 ) 0.013 (0.043 )Married −0.026 (0.036 ) −0.027 (0.035 ) −0.022 (0.034 )Separated −0.065* (0.039 ) −0.055 (0.038 ) −0.042 (0.036 )
Age25–34 0.008 (0.076 ) −0.053 (0.067 ) −0.042 (0.067 )35–44 0.033 (0.075 ) −0.034 (0.067 ) −0.017 (0.066 )45–54 0.093 (0.074 ) 0.015 (0.066 ) 0.032 (0.065 )55–64 0.087 (0.074 ) 0.009 (0.066 ) 0.021 (0.066 )>65 0.084 (0.077 ) −0.012 (0.069 ) 0.006 (0.069 )
EthnicityWhite 0.020 (0.030 ) −0.010 (0.032 ) −0.002 (0.032 )Black 0.056 (0.036 ) 0.028 (0.038 ) 0.035 (0.038 )Latino −0.012 (0.034 ) 0.005 (0.034 ) −0.005 (0.034 )Other 0.052 (0.056 ) 0.025 (0.056 ) 0.037 (0.055 )
OtherMale 0.052*** (0.016 ) 0.054*** (0.015 ) 0.048*** (0.014 )HH Members 0.007 (0.006 ) 0.001 (0.006 ) −0.001 (0.006 )CC Revolver −0.043*** (0.013 ) −0.038*** (0.012 ) −0.034*** (0.012 )Home owner −0.038** (0.019 ) −0.040** (0.018 ) −0.043** (0.019 )
Rel. CharacteristicsSecurity 0.017** (0.008 ) 0.012* (0.008 )Setup 0.040** (0.016 ) 0.032** (0.015 )Acceptance −0.021 (0.019 ) −0.023 (0.019 )Cost 0.055*** (0.019 ) 0.051*** (0.019 )Records 0.038*** (0.011 ) 0.035*** (0.011 )Convenience 0.093*** (0.016 ) 0.092*** (0.015 )
Withdrawal MethodBank teller 0.005 (0.016 )Check casher 0.073 (0.100 )Cashback −0.081*** (0.017 )Employer 0.150** (0.063 )Family 0.002 (0.047 )Other −0.013 (0.033 )
Constant 0.735*** (0.139 ) 0.911*** (0.136 ) 0.837*** (0.135 )
R2 0.092 0.157 0.179Observations 4594 3682 3680Individuals 1116 1109 1108
Note: OLS is the pooled OLS estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors and survey weights areused. HH and CC refers to household and credit card, respectively. Base category ofcategorical variables is lower than high school, unemployed, widowed, lower than 25 years,Asian, and ATM. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
132
Table A17: FE Regression Results of Contactless Credit on Cash ShareVolume
(1) (2) (3)
Variable b se b se b se
Contactless Credit 0.002 (0.010 ) −0.003 (0.012 ) −0.003 (0.012 )log(Income) 0.009 (0.011 ) 0.015 (0.013 ) 0.014 (0.013 )Interest Rate 0.003 (0.008 ) 0.010 (0.009 ) 0.011 (0.009 )Education
High School 0.250*** (0.048 ) 0.319*** (0.012 ) 0.305*** (0.017 )Some College 0.279** (0.118 ) 0.470*** (0.103 ) 0.455*** (0.102 )College 0.139 (0.130 ) 0.197* (0.119 ) 0.186 (0.118 )Post Graduate 0.099 (0.149 ) 0.190 (0.135 ) 0.180 (0.132 )
EmploymentWorking 0.003 (0.014 ) 0.020 (0.017 ) 0.019 (0.017 )Retired 0.010 (0.016 ) −0.014 (0.024 ) −0.016 (0.024 )Other 0.020 (0.015 ) 0.012 (0.019 ) 0.012 (0.019 )
Marital StatusSingle −0.017 (0.057 ) −0.018 (0.062 ) −0.016 (0.062 )Married −0.001 (0.044 ) 0.034 (0.048 ) 0.035 (0.048 )Separated 0.032 (0.047 ) 0.093* (0.050 ) 0.094* (0.050 )
Age25–34 −0.082* (0.044 ) −0.145* (0.080 ) −0.146* (0.080 )35–44 −0.085* (0.048 ) −0.144* (0.083 ) −0.147* (0.083 )45–54 −0.097* (0.052 ) −0.144* (0.086 ) −0.146* (0.086 )55–64 −0.113** (0.055 ) −0.166* (0.089 ) −0.167* (0.088 )>65 −0.088 (0.058 ) −0.173* (0.091 ) −0.172* (0.091 )
OtherHH Members −0.005 (0.006 ) 0.001 (0.007 ) 0.002 (0.007 )CC Revolver −0.022** (0.011 ) −0.031** (0.013 ) −0.031** (0.013 )Home owner −0.030* (0.018 ) −0.021 (0.021 ) −0.022 (0.021 )
Rel. CharacteristicsSecurity −0.001 (0.005 ) −0.001 (0.005 )Setup 0.017* (0.010 ) 0.017* (0.010 )Acceptance −0.007 (0.013 ) −0.007 (0.013 )Cost 0.026** (0.012 ) 0.026** (0.012 )Records 0.005 (0.008 ) 0.005 (0.008 )Convenience 0.000 (0.010 ) 0.001 (0.010 )
Withdrawal MethodBank teller 0.017 (0.014 )Check casher 0.005 (0.045 )Cashback −0.020 (0.014 )Employer −0.008 (0.034 )Family 0.004 (0.025 )Other −0.013 (0.022 )
Constant 0.155 (0.161 ) 0.021 (0.181 ) 0.041 (0.181 )
R2 0.011 0.028 0.030Observations 4595 3683 3681Individuals 1115 1109 1108
Note: FE is the fixed-effects estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors are used. HH and CCrefers to household and credit card, respectively. Base category of categorical variables islower than high school, unemployed, widowed, lower than 25 years, Asian, and ATM.Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
133
Table A18: FE Regression Results of Contactless Debit on Cash Share Vol-ume
(1) (2) (3)
Variable b se b se b se
Contactless Debit −0.029** (0.014 ) −0.032** (0.015 ) −0.031** (0.015 )log(Income) 0.010 (0.011 ) 0.015 (0.013 ) 0.015 (0.013 )Interest Rate 0.003 (0.008 ) 0.010 (0.009 ) 0.011 (0.009 )Education
High School 0.256*** (0.047 ) 0.319*** (0.012 ) 0.306*** (0.017 )Some College 0.282** (0.117 ) 0.463*** (0.104 ) 0.449*** (0.103 )College 0.146 (0.129 ) 0.192 (0.120 ) 0.182 (0.119 )Post Graduate 0.106 (0.149 ) 0.185 (0.136 ) 0.176 (0.133 )
EmploymentWorking 0.003 (0.014 ) 0.019 (0.017 ) 0.018 (0.017 )Retired 0.009 (0.016 ) −0.014 (0.024 ) −0.016 (0.024 )Other 0.020 (0.015 ) 0.012 (0.018 ) 0.012 (0.019 )
Marital StatusSingle −0.016 (0.057 ) −0.018 (0.062 ) −0.017 (0.062 )Married −0.001 (0.044 ) 0.035 (0.048 ) 0.036 (0.048 )Separated 0.032 (0.047 ) 0.092* (0.050 ) 0.093* (0.050 )
Age25–34 −0.084* (0.043 ) −0.148* (0.079 ) −0.148* (0.079 )35–44 −0.085* (0.048 ) −0.145* (0.082 ) −0.147* (0.082 )45–54 −0.097* (0.052 ) −0.144* (0.085 ) −0.145* (0.085 )55–64 −0.113** (0.055 ) −0.166* (0.088 ) −0.166* (0.088 )>65 −0.088 (0.058 ) −0.173* (0.090 ) −0.171* (0.090 )
OtherHH Members −0.005 (0.006 ) 0.001 (0.007 ) 0.001 (0.007 )CC Revolver −0.022** (0.011 ) −0.032** (0.013 ) −0.031** (0.013 )Home owner −0.031* (0.018 ) −0.023 (0.021 ) −0.023 (0.021 )
Rel. CharacteristicsSecurity −0.001 (0.005 ) 0.000 (0.005 )Setup 0.017* (0.010 ) 0.017* (0.010 )Acceptance −0.006 (0.013 ) −0.007 (0.013 )Cost 0.026** (0.012 ) 0.026** (0.012 )Records 0.006 (0.008 ) 0.006 (0.008 )Convenience 0.000 (0.010 ) 0.001 (0.010 )
Withdrawal MethodBank teller 0.016 (0.014 )Check casher 0.008 (0.044 )Cashback −0.020 (0.014 )Employer −0.010 (0.034 )Family 0.004 (0.025 )Other −0.013 (0.022 )
Constant 0.150 (0.160 ) 0.028 (0.180 ) 0.047 (0.180 )
R2 0.013 0.030 0.032Observations 4594 3682 3680Individuals 1116 1109 1108
Note: FE is the fixed-effects estimator obtained on the balanced panel. b are the pointestimates and se the standard errors. Cluster-robust standard errors are used. HH and CCrefers to household and credit card, respectively. Base category of categorical variables islower than high school, unemployed, widowed, lower than 25 years, Asian, and ATM.Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
134
Chapter 4
The Impact of MobilePayment on PaymentChoice1
AbstractThis paper focuses on the effect of mobile payment on the adoptionand usage patterns of traditional payment instruments such as cash,checks, credit, debit, and prepaid cards used at the point-of-sale. Dataare drawn from a representative survey on consumer payment choicein the U.S. in 2012. Using discrete-choice random utility models tosimulate consumer behavior, the estimation provides two major find-ings. First, pertaining to the adoption stage, mobile payment doesnot replace physical payment cards, but is likely to substitute paper-based payment methods such as cash and checks. Second, mobilepayment does not statistically significantly influence the choice of pay-ment means at the POS in terms of usage. However, there is suggestiveevidence that it is complementary to card payments and a substitutefor paper-based payment instruments. The findings highlight the po-tential social welfare gains of mobile payment and provide key insightsinto challenging issues for the private industry sector. This paper fur-thermore offers novel evidence on the impact of mobile payment onthe use and adoption of existing payment instruments and aims atcontributing to the literature on consumer payment choice.
JEL-Classification: D12, G21, O33
Keywords: mobile payment, payment innovation, retail payments, payment cards
1This article has been published in a slightly revised version as Trutsch, T. (2016),The Impact of Mobile Payment on Payment Choice, Financial Markets and Portfolio
Management, 30(3), 299–336.
135
4.1 Introduction
The availability of an increasing number of different payment instruments
and of new online payment opportunities offers individuals various payment
alternatives from which to choose. For instance, consumers can nowadays
select from among at least nine payment instruments in the U.S. that – apart
from cash – either authorize the transfer of money or can access funds in
checking and other deposit accounts to initiate payments (Schuh and Stavins,
2014). Traditional banking payment services are increasingly facing fierce
competition by novel and established non-financial companies such as Paypal,
Google, Apple and Square, amongst others. These attempt to gain market
share in retail payment markets to meet actual consumer preferences by
launching innovative payment products such as mobile payment that offers
technological advances in payment processing and more convenient payment
initiation.2 Therefore, the effects on consumer payment choice in the context
of mobile payment has been attracting growing attention by researchers.
This paper studies the effect of mobile payment on the adoption and
usage patterns of traditional payment instruments used at the point-of-sale
(POS) and provides empirical evidence of actual changes in the composition
of payment instrument portfolios as well as the instruments’ deployment. On
the one hand, mobile payment comprises payments for purchases and there-
fore may compete or complement current POS payment means (e.g. cash,
checks, credit, debit, and prepaid cards). On the other hand, mobile pay-
ment offers a new access channel to account-based payment services such as
online banking payments and bank account number payments (Kim et al.,
2010). For the purpose of this study, however, I exclusively focus on POS
payments since they account for the majority of consumer payments in the
2Mobile payment is here defined as any payment that is authorized, initiated or con-firmed through a mobile device (Au and Kauffman, 2008). Mobile payment activitiesentail sending a text message, tapping or waving the phone to make a contactless pay-ment, scanning a barcode or QR code using the mobile phone, using the mobile phone’sweb browser, using a downloadable app, and swiping a credit, debit or prepaid card in adevice attached to the mobile phone. It is usually debit and credit cards that determinethe underlying payment process and settlement of payment. However, other forms suchas bank account deduction or charge of phone bills is commonly practicable. I refer to Liuet al. (2015) for a survey of recent changes in the mobile payment technology ecosystem.
136
U.S. (Schuh and Stavins, 2014).3 The multiple fields of application and the
high market penetration of mobile phones suggest that mobile payment is
experiencing a promising prospect as a means of payment. The paper pre-
sented therefore aims to give insights into the complementing or substituting
patterns of mobile payment and its potential welfare gains.
This paper is motivated by recent developments in payment markets,
as innovative mobile payment products have been frequently launched, and
where the interrelation between payment alternatives is still unclear. Un-
derstanding these effects is important for the following reasons. First, con-
sumers’ payment instrument choices for transactional purposes significantly
affect the efficiency and effectiveness of an overall payment system, which
in turn determines financial stability. Since the payment system in the U.S.
predominantly relies on paper-based payment methods (Schuh and Stavins,
2014), which cause high inefficiencies and high operation costs – for instance,
due to handling and distribution of cash – a shift to more electronic pay-
ments would confer overall economic surplus (cf. Schmiedel et al., 2013). In
general, substituting cash by electronic payment instruments is found to be
associated with decreasing social costs (e.g. van Hove, 2008; Humphrey et al.,
2001). Also, a transfer of conventional card payments to mobile payments
may additionally create economic value due to expected lower fee structures.
Investigating these effects is supported by the strategic plan for 2012–2016
issued by the Federal Reserve Financial Services in the U.S., which high-
lights to enhance understanding end-user needs for and barriers to payments
system efficiency and faster payment (Pianalto, 2012).
Second, comprehending and anticipating the impact of mobile payment
and its implications on traditional payment services is elementary since it
could create market disruptions and could threaten financial stability due to
its new features relevant to convenience and security. For instance, mobile
payment – especially such applications that are operated by non-bank market
players – may raise new policy issues and may challenge the existing regula-
tory framework since these market participants might impact the payment
3According to Schuh and Stavins (2014), 68.1 percent of total average monthly pay-ments a consumer made in 2012 were non-bill payments.
137
landscape to a large extent posing new liability issues. Providing detailed in-
formation for policy makers and regulatory authorities could facilitate their
decision-making process.
Third, assessing the impact of mobile payment is interesting for vari-
ous market participants especially of the private industry sector. On the one
hand, banks and other financial intermediaries may be confronted by eroding
revenue streams incurred by new market players of mobile payment appli-
cations. On the other hand, mobile network operators (MNO) may benefit
from new revenue streams stemming from charging conventional payment ser-
vice providers for the right to undertake payments using their system. The
findings therefore provide information for strategic and investment decision-
making purposes for mobile payment stakeholders of the private industry.
This paper aims at contributing to the literature on payment economics
with respect to consumer payment choice and fills the gap in understand-
ing the dynamics between mobile and traditional payment methods. To the
best of my knowledge, there has not been any literature targeting solely
the effect of mobile payment on conventional payment means, which is also
highlighted by Dahlberg et al. (2008) and most recently reemphasized by
Dahlberg et al. (2015). However, a few disruption analysis studies conclude
that card payments would still be preferred to mobile payments from an
industry point of view and that the latter tend to become rather a comple-
ment than substitute for traditional payment methods in Switzerland (cf.
Ondrus and Pigneur, 2005, 2006a, 2006b). In contrast, others prospect that
mobile payment reduces the use of central bank cash as well as credit and
debit card payments (e.g. Garcia-Swartz et al., 2002). Polasik et al. (2013)
argue that contactless mobile payment, applied for proximity payments at
the POS, will lead to a breakthrough in payment markets due to its superior
time efficiency compared to cash. It is thus ambiguous what effect mobile
payment exerts. Consequently, this paper presents improved knowledge and
empirical evidence on payment habits and their changes in the context of
mobile payment from the consumer’s point of view.
The novelty of this paper is manifold. First and foremost, this paper
is the first study that gauges the impact of mobile payment on the array
138
of traditional payment methods used at the POS. Second, the richness and
quality of a unique data set allows for a detailed assessment of mobile pay-
ment technologies on conventional payment instruments, as well as in terms
of adoption rates and usage behavior. Third, the individual consumer-level
data set enables estimating random utility models that quantify the effect
of mobile payment for varying control variables. From this vantage point,
I eventually can evaluate substitution patterns across the current payment
instruments and can shed light on potential market disruptions.
To this end, I apply discrete-choice econometric models on the probabil-
ity of adopting and using conventional payment instruments such as cash,
checks, debit, credit, and prepaid cards with regards to mobile payment at
the POS. Drawing data from the 2012 Survey of Consumer Payment Choice
(SCPC) in the U.S., my analysis yields the following important findings.
First, mobile payment statistically significantly increases the probability of
possessing all available payment instruments at the POS by roughly two
percentage points and reduces the likelihood of adopting payment portfolios
comprising checks and only cash (the extensive margin). This implies that
mobile payment does not emerge as a substitute for physical payment cards,
but does so for paper-based payment products. Second, mobile payment does
not statistically significantly impact consumer payment choice at the POS
in terms of usage (the intensive margin) with the exception of prepaid card
payments, which are positively affected by mobile payment. However, there
is an indication that mobile payment can be regarded as complementary to
traditional card payments and as a substitute for paper-based payment in-
struments such as cash and checks. The results may reflect the fact that
the usage of payment instruments strongly depends on other factors, such
as perceived characteristics of payment methods, individual habits, and au-
tomatism, among others.
The remainder of this paper is as follows. Section 4.2 shows related
payment literature focusing on the choice of payment methods. Thereby,
literature on mobile payment is also discussed. In section 4.3, I introduce the
framework of the random utility model and provide a theoretical background
of its properties. I proceed by the identification strategy and the specification
139
of the models that are used for estimation in section 4.4. The description
of the data is presented in section 4.5. Section 4.6 discusses the estimation
results and compares the model specifications of their overall fit of the data.
Furthermore, I run a plausibility check to validate the results in section 4.7.
Finally, section 4.8 concludes.
4.2 Related Literature
This paper is related to a growing amount of academic literature in payment
economics in recent years dealing with the determinants of consumer pay-
ment choice.4 The bulk of empirical studies rely on individual-level survey
data due to restricted accurate transactional-level data sources.5 In sum,
the payment literature concludes that the adoption and usage of electronic
payment instruments is primarily determined by personal, transactional and
situational characteristics as well as payment instrument attributes. More-
over, price characteristics and financial incentives are strong predictors for
adoption and deployment.
A few studies reveal that socioeconomic and financial attributes of con-
sumers are relevant indicators for payment choice (Schuh and Stavins, 2010;
Borzekowski et al., 2008; Stavins, 2001). Accordingly, younger, more edu-
cated cohorts with higher incomes are more likely to use electronic payment
instruments than elderly, less educated people with lower incomes, who tend
to preferably use paper-based payment methods such as cash. One reason is
that the former face higher opportunity costs for undertaking paper-based
transactions, which generally take more time to settle (Polasik et al., 2013).
Further research find influence of region and foreign background. For in-
stance, consumers’ usage patterns of payment instruments are highly affected
4I refer to Scholnick et al. (2008) and Humphrey (2010) for a synopsis of empiricalfindings and recent developments in payment literature in the context of various subdisci-plines.
5There are a few exceptional studies based on transactional-level data provided bystores (scanner data), banks or credit card companies (e.g. Cohen and Rysman, 2013;Agarwal et al., 2010; Klee, 2008; Rysman, 2007). Other papers focus on consumer pay-ment choice over time using aggregate data sources provided by central banks or paymentsystems (e.g. Humphrey et al., 2001, 1996; Snellman et al., 2001; Amromin and Chakra-vorti, 2009).
140
by the fraction of other people in the region using the same type of payment
method (Stavins, 2001). In addition, Kosse and Jansen (2013) find that for-
eign backgrounds still influence payment instrument choice after migration
using Dutch data. To put differently, migrants from cash-oriented countries
are more prone to use cash after migration.
A plethora of literature deals with the effect of transaction size, the type
of good purchased and the spending location on payment choice. The size
of a transaction is a leading indicator for the choice of payment methods at
the POS (von Kalckreuth et al., 2014; Cohen and Rysman, 2013; Klee, 2008;
Bounie and Francois, 2006).6 The higher the size of the transaction, the
more likely people pay by payment cards. Conversely, cash is dominated for
small-value purchases. As opposed to these findings, Bouhdaoui and Bounie
(2012) and in a similar vein Eschelbach and Schmidt (2013) argue that pay-
ment choice is predominately driven by the outstanding cash balance held in
wallets than by transaction size. This is in line with the theoretical model
established by Alvarez and Lippi (2015) where cash is used whenever cash
holdings are sufficient allowing for an optimal dynamic cash management
and the choice of credit cards.
Another approach infers consumer choice from situational aspects. Rys-
man (2007) states that the amount of local merchant acceptance of the spe-
cific payment card network favored by the consumer positively affects his
payment behavior. This finding indicates the underlying network effects of
retail payments. Consumer payment choice also differs regarding location,
where, for instance, the physical characteristics of the POS such as the ab-
sence of a cashier are relevant (Jonker, 2007; Bounie and Francois, 2006;
Hayashi and Klee, 2003).
There is a number of influential empirical papers that investigates price
and financial incentive responses to payment choice. Overall, consumers are
very price sensitive. For instance, many scholars exhibit a very elastic re-
sponse to fees and surcharges imposed on debit card transactions, inferring
consumers to substitute for alternative payment methods (Koulayev et al.,
2012; Stavins, 2011; Borzekowski et al., 2008; Bolt et al., 2010). In contrast,
6See Arango et al. (2013) for an international comparison.
141
Bolt et al. (2008) argue that direct pricing of payment transactions may not
necessarily ensure the substitution of electronic payments for paper-based
transactions.7 From a seller’s perspective, however, it is not optimal to use
price discounts to steer consumers to pay by cheaper payment instruments.
On the one hand, rewards on credit card transactions may exceed the price
discounts merchants can provide and their cost reduction may be insufficient
to offset the cost increase of administering the varying price schemes (Brigle-
vics and Shy, 2014). On the other hand, merchants fear intensified price
competition that is almost never optimal for two vertically differentiated
merchants (Choudhary and Tyagi, 2009).
There is separate literature that highlights the significant and positive
effects of loyalty programs and other financial incentives (card discounts,
points, cash-back, interest-free periods) on the use of payment cards at the
expense of cash (Arango et al., 2015b; Carbo-Valverde and Linares-Zegarra,
2011; Ching and Hayashi, 2010; Simon et al., 2010; Agarwal et al., 2010). In
turn, credit card charges hinder revolvers to pay by credit cards and instead
they prefer to pay by debit cards (Zinman, 2009).8
With regards to mobile payment, this paper hinges on empirical work that
points out the very importance of payment instrument attributes such as con-
venience, ease of use, speed, record keeping, and security, amongst others,
for the choice of payment instruments (e.g. Arango et al., 2015a; Ching and
Hayashi, 2010; Schuh and Stavins, 2010; Klee, 2008). Overall, these determi-
nants are found to be more important than demographic variables. Mobile
payment is considered to both increase the convenience of payments, reduce
the transaction costs, and provide better record keeping (Mallat, 2007). It
can therefore improve the attractiveness of electronic card payments, as it
enhances their convenience by technological modifications (Jonker, 2007).
Conversely, personal experience with mobile payment is found to negatively
influence cost-conscious payment choice behavior such as choosing cash at
the POS (EC, 2013).9 My analysis aims to fill the gap in understanding
these mechanisms.7See also Ten Raa and Shestalova (2004) for an optimal transaction pricing.8See also Massoud et al. (2011) who find that card interest rates are direct substitutes
of card penalty fees, which are increasing in customer risk.9Cost-consciousness in this context represents transparency of payment charges.
142
In another vein, mobile payment has been extensively studied from the
behavioral decision and intention perspective towards innovations (see for a
synopsis Dahlberg et al. (2008), Dahlberg et al. (2015), and many references
therein).10 The adoption and usage mechanisms of mobile payment are influ-
enced by the relative importance of certain factors such as trust, usefulness,
ease of use, external influences, and personal traits, amongst others (Xin
et al., 2015; Liebana-Cabanillas et al., 2014; Yang et al., 2012; Kim et al.,
2010). Thereby, perceived ease of use, perceived usefulness, as well as trust
are found to be the most important factors (Dahlberg et al., 2015). Contrar-
ily, Mallat (2007) illustrates that the relative advantages of mobile payment
differ from the traditional adoption theories, as it offers ubiquity of payment,
independence of time, queue avoidance, and the ability to complement cash
payments.
4.3 Model Specification of Payment Choice
The behavioral model of payment choice is derived from the random utility
framework, which explains decision behavior (cf. Train, 2009). The descrip-
tion of payment decisions in discrete choice models is based on the utility
maximizing choice between discrete alternatives and allows to estimate con-
sumer preferences over the choices. Formally, there is a decision maker i
who faces a choice among J payment alternatives while each alternative pro-
vides a certain level of utility. Uij , j = 1, . . . , J is the utility level which a
decision maker faces when choosing payment method j. If Uij > Uik for all
k 6= j holds, then the alternative j is selected. In other words, the payment
alternative yielding the highest utility is the one that is chosen such that
Vij = maxUij . (4.1)
The decision maker’s utility can be decomposed as
Uij = Vij + ǫij (4.2)
10I refer to Au and Kauffman (2008) for a survey of stakeholder issues in the field ofmobile payment.
143
where Vij is a function which relates observed factors to the decision maker’s
utility. This function is denoted Vij = V (Xi, Sj , βj , γ)∀j where the factors
are attributes of the decision maker Xi and of the payment alternatives Sj .
It is called representative utility. Vij also depends on unknown parameters βj
and γ, which have to be estimated. βj relates the attributes of the decision
maker Xi to his utility for choice j. γ shows the relationship between the
decision maker and the alternatives to his utility for alternative j. Since
some factors cannot be observed and therefore are not included in Vij but
affect utility, they are captured by ǫij , which is assumed to be randomly
distributed. It can be seen as the error made in evaluating alternative j.
Since ǫij is simply the difference between Uij and Vij , this decomposition is
completely general.
The logit model is obtained by assuming that each ǫij is independently
and identically distributed extreme value (i.i.d.) implying homogeneous error
variances. This distribution is referred to as extreme value type-I or also as
Gumbel distribution leading to the assumption of independence of irrelevant
alternatives (IIA). Put differently, the consumer’s preference for payment
method j over method k is independent of the availability of other choice
alternatives.11 The density and cumulative distribution of ǫij is, respectively,
f(ǫij) = exp(−ǫij) exp(− exp(−ǫij)) (4.3)
and
F (ǫij) = exp(− exp(−ǫij)). (4.4)
Under this assumption, the probability that decision maker i chooses pay-
ment alternative j over k is given by
11This problem is not very likely to occur so starkly in the context of payment in-strument adoption since the choice set for payment methods at the POS is more finite.According to Borzekowski and Kiser (2008), the IIA assumption can be relaxed (on anaggregate level) if interaction terms between individual- and alternative-specific attributesare included.
144
Pij = Prob(Uij > Uik, ∀k 6= j) (4.5)
= Prob(Vij + ǫij > Vik + ǫik, ∀k 6= j) (4.6)
=exp(Vij)∑J
k=1 exp(Vik). (4.7)
As can be inferred, the expression in 4.7 requires that the probabilities
lie between zero and one and that they must sum to one. The fact that
the above choice probabilities lead to i.i.d. extreme value distributed errors
was proved by McFadden (1973). The basic setup in 4.7 is referred to as the
multinomial logit model (MNL), where the utility of all payment alternatives
depends on the same factors such as personal characteristics Xi. It leads to
the binary logit model for J = 1. In the conditional logit model (CL), the
utility of each payment alternative solely depends on attributes Sj of that
alternative, which require to vary across alternatives. The conjunction of
these models is the mixed conditional logit model where both alternative- and
case-specific variables are included. These models allow to infer structural
behavioral responses of consumers by using the estimates of the model to
perform counterfactual experiments in the consumer choice set. Because the
logit probabilities take a closed form and are fairly easily computed, the
traditional maximum likelihood procedure can be applied. For more details
on the estimation procedure with maximum likelihood, see Train (2009).
These models, however, can only explain taste variation to the extent
that tastes vary with the observed characteristics of individual i. In the
presence of unobserved heterogeneity and to relax the IIA assumption, how-
ever, random coefficient models are necessary to avoid biased estimates. The
nested logit model (NL) is regarded as the most tractable of these more gen-
eral multinomial logit models, where the vector of the payment instrument
error terms ǫij exhibits the generalized value distribution (GEV) with the
following cumulative distribution function (see Train, 2009)12:
12In addition, the mixed logit model allows for random parameters by varying theelements of β over the decision makers (see Train, 2009).
145
F (ǫij) = exp
−
L∑
l=1
∑
j∈Bl
exp(−ǫij/λl)
λl
. (4.8)
In this model, the alternatives J are partitioned into groups L with each
alternative j belonging to an upper nest B. It is thought of a decision tree
where the consumer first decides which nest to choose and then within the
nest the alternative is selected. The error terms are allowed to be corre-
lated within nests, but are uncorrelated across nests following an univariate
extreme value distribution. Consequently, the probability that individual i
chooses payment instrument h is:
Pih =exp(Vih/λl)
(∑j∈Bl
exp(Vij/λl))λl−1
∑L
d=1
(∑j∈Bd
exp(Vij/λd))λd
. (4.9)
The parameter λl measures the correlation between alternatives within
the different nests, i.e. the degree of independence in unobserved utility. The
higher λl, the less correlation within nests is observable. According to Train
(2009), the model is consistent with utility maximization for all values of the
explanatory variables if λl is between zero and one, but only for some range
of variables if λl is greater than one.
4.4 Identification Strategy
In this section, I describe the econometric models of consumer payment
choice, which are directly derived from the random utility model laid out
above. Consumer payment behavior can be thought of as a two-step deci-
sion process. First, individuals decide whether to adopt a specific payment
instrument leading to the possession of different payment portfolios (the ex-
tensive margin). Second, they choose how much to use each instrument
adopted in different contexts (the intensive margin). In this study, the two
different processes are modeled independently and sequentially, respectively.
This approach is selected since individuals usually have to apply first for a
146
payment card before they can use it. Also, this paper aims to provide evi-
dence on the causal effect of mobile payment on either the adoption or the
usage stage of traditional payment instruments, rather than both simulta-
neously. On the one hand, mobile payment might replace physical payment
cards due to electronic storing of payment card information, which nega-
tively affects the adoption stage. On the other hand, mobile payment might
promote electronic transactions, which positively affects the usage stage.13
In this way, it is plausible to consider these issues separately.
Hereafter, in section 4.4.1, I will first deal with potential endogeneity
issues related to my empirical strategy allowing to correctly estimate the
impact of mobile payment on payment choice. Second, the model specifi-
cation that identifies the impact of mobile payment in the adoption stage
is discussed in section 4.4.2. Third, section 4.4.3 lays out the econometric
specification associated with the impact of mobile payment in the use stage.
4.4.1 Identifying Assumptions
Inferring from the random utility model above, I specify a stylized utility
function to estimate the effect of mobile payment on the choice of payment
method j as follows:
Uij = V (MPi, Xi, Sj) + ǫij . (4.10)
The observed utility V (·) of individual i can be described as a function of
mobile payment MPi, consumer characteristics Xi, and payment instrument
attributes Sj . ǫij captures the measurement error. It is important to note
that the variable MPi has to be strictly exogenous to estimate an unbiased
effect in the adoption and usage stage. However, several endogeneity issues
are likely to appear in this context. First, selection bias may be prevalent
due to unobservable factors causing consumers to adopt mobile payment and
simultaneously to decrease the number of paper-based payment methods.
For example, consumers with preferences for new technologies may generally
13I refer to Koulayev et al. (2012) who develop a structural model of simultaneousadoption and use of payment instruments accounting for the amount of usage at the timeof adoption while at the same time identifying the effect of use on adoption.
147
possess less paper-based payment means and are likely to adopt innovations
such as mobile payment more extensively. Similarly, this holds for the usage
of mobile payment. For these people, the utility of mobile payment is greater
than for others.
Second, the direction of causation of MPi is not obvious and could also
be reverse. For instance, consumers who tend to pay less by cash are more
prone to use innovative payment alternatives such as mobile payment.14
To obtain an unbiased causal parameter of MPi, the aforementioned en-
dogeneity issues can largely be circumvented by the inclusion of individuals’
perceptions towards payment characteristics of traditional payment instru-
ments. These capture consumer preferences what otherwise would have been
unobserved heterogeneity (cf. Ching and Hayashi, 2010; Schuh and Stavins,
2010; Jonker, 2007). Moreover, the assessment of attitudinal characteristics
of mobile payment allows to control for heterogeneous effects of mobile pay-
ment across consumers and for other unobservables such as personal affinity
towards innovations.15 Therefore, consumer preferences for payment alter-
natives can comprehensively be explained by their perceived characteristics,
which conversely purge unobserved heterogeneity.
Nevertheless, it is perfectly conceivable that the error terms could be cor-
related across payment alternatives since different groups of payment meth-
ods have similar unobservable characteristics. In this way, the assumption
of IIA is violated. For instance, paper-based payment methods such as cash
share the distinct feature of anonymity and transparency (“pain of paying”)
whereas payment cards are subject to nearly unlimited liquidity and credit
functions and leave personal data tracks. Additionally, payment decision-
14Another important endogeneity issue stems from the two-sided structure of the pay-ment market where network effects are dominating. In other words, the interdependenceof supply and demand of payment methods by merchants and consumers, respectively,results in feedback effects. These could result in higher utility for consumers who haveadopted payment alternatives that are ubiquitously accepted by merchants. The perceivedrating of acceptance for payment methods by consumers can help to largely explain theirvariation in payment choice, which the Survey of Consumer Payment Choice (SCPC) con-tends (see below). However, due to data restrictions, an adequate control for this issuefrom the seller’s point of view is not possible.
15It is conceivable that consumers who rate characteristics of mobile payment morepositively than others have higher coefficients than consumers who do not.
148
making is considered to be largely habitual and unconscious (van der Horst
and Matthijsen, 2013). This implies that consumers often have preferences
for either one of the payment instruments, particularly subjected to the set
of paper or plastic payment products. While the IIA assumption is likely
to be violated in the usage stage due to the above mentioned arguments, it
is assumed to hold true in the adoption stage. The rationale is that it is
unlikely that choosing a specific payment instrument portfolio is dependent
of whether to adopt another bundle is an option.
Another methodological issue pertains to sample selection in the usage
stage, as the adoption of payment instruments is a prerequisite for their
usage. However, since the penetration of available POS payment instruments
is relatively advanced (see Table 4.2) and assuming that the adoption and use
decisions are made sequentially and not simultaneously, sample selection bias
in the usage stage is considered to be negligible. Evidence may provide Schuh
and Stavins (2013) in their study where the Mills Ratios of the first-stage
probit models in the usage stage of POS payment means are insignificant
in the majority of cases. Similar caveats can put forward regarding the
ownership of mobile phones that are necessary to perform mobile payment
tasks. Because the rate of diffusion, however, appears to be relatively high
(around 95 percent), the sample is not restricted for estimation accordingly.
4.4.2 Estimating the Adoption of Payment Instruments
Crucial to identification of the effect of mobile payment in the adoption stage
is the specification of the utility function. Thereby, I draw on previous work
by Schuh and Stavins (2013). I estimate the econometric model using the
(mixed) conditional logit method. Individual i obtains the following utility
from choosing payment instrument j:
Uij = αMPi + βXi + γYi + δZi + λRCj + ǫij , (4.11)
where MPi takes the value of one if consumer i has used any form of mobile
payment, Xi is a vector of consumer demographics, Yi is a set of financial
variables related to individual i, Zi are additional control variables including
149
the attitudinal data of the consumer’s valuation of mobile payment, and RCj
is a vector of relative attributes of payment method j perceived by individual
i. ǫij represents the unobserved preference component that is related to the
particular payment choice and is assumed to be i.i.d.
Since consumers are very heterogeneous in their adoption patterns of
payment instruments (Schuh and Stavins, 2014), i.e. they generally adopt
different payment portfolios instead of a single instrument, it makes sense
to proceed by determining the observed individual payment portfolios, of
which each instrument is solely applicable at the POS. This comes at the
advantage of identifying an exhaustive, mutually exclusive, and finite num-
ber of discrete alternatives, which is a prerequisite of the conditional logit
model. Available payment instruments j = 1, . . . , J at the POS are cash,
checks, credit, debit, and prepaid cards, from which every combination can
be adopted.16 Following the approach in the spirit of Koulayev et al. (2012),
the consumer selects bundle b ∈ B, where b is a subset of all possible sets B
of payment instruments. Assuming that every consumer adopts cash, there
are four payment choices remaining, leading to the maximum number of dif-
ferent payment portfolios B = 16 (24). Thereby, the direct utility function
in 4.11 of choosing bundle b converts to
Uib = αMPi + βXi + γYi + δZi + λRCb + ǫib, (4.12)
where individual i derives utility from choosing payment portfolio b. MPi
takes the value of one if consumer i has used any form of mobile payment.
The set of demographic variables Xi include age, gender, education, and
household size. The vector Yi encompasses employment status and income.
Zi consists of a dummy indicating whether having ever been bankrupt in
the last 12 months prior to the survey. Also, the perceived assessment of
mobile phones, mobile phones with internet access, voice calling, and texting
in terms of security are included. RCb is a set of relative measures of per-
ceived security, setup, acceptance, cost, records, and convenience of payment
16Note that an additional payment method such as money order is applicable at thePOS. However, since its share of adoption and usage is negligible, I exclude it from theanalysis.
150
instrument j belonging to bundle b (see section 4.5.2 for variable definition).
It varies across payment portfolios (alternative-specific regressors). The case-
specific regressors, which are constant over alternatives, comprise individual
characteristics MPi, Xi, Yi, and Zi.
4.4.3 Estimating the Usage of Payment Instruments
The specification of the utility function in the usage stage hinges on the study
by Ching and Hayashi (2010) and Schuh and Stavins (2013). The model
is estimated using the nested logit method that explains which payment
instrument j is most frequently selected by consumer i for each transaction
type h. As for the analysis, I constructed two nests (L = 2), where the
paper-based payment methods cash and checks share one nest (Bpaper) and
the remaining card-based payment alternatives (debit, credit, and prepaid
cards) share another nest (Bcard). The utility function to be estimated has
the form
Uijh = αMPi + βXi + γYi + δZi + λRCj + ǫijh, (4.13)
where MPi is a dummy variable for having used any form of mobile pay-
ment in the past 12 months, Xi is a vector of consumer demographics, Yi
is a set of financial variables related to individual i, Zi are additional con-
trol variables including the attitudinal data about the consumer’s valuation
of mobile payment, and RCj is a vector of relative attributes of payment
method j perceived by individual i. The specific variables incorporated in
the utility function are similar to those in the adoption stage (see equation
4.12).
Accordingly, I can account for observed heterogeneity across individuals
in the model. Put more simply, it means that the marginal utility of pay-
ment method j in context h is different across consumers. Consumer i can
choose among five payment instruments j = 1, . . . , J such as cash, checks,
credit, debit, and prepaid cards to pay for three transaction types h such
as total POS payments, which are further distinguished by retail payments
151
and services payments.17 It is noteworthy that I abstract from the adoption
decision of available payment methods and hence focus on every consumer
irrespective of the number of adopted instruments.
4.5 Data
4.5.1 Source
I draw data from the Federal Reserve Bank of Boston that supports the Con-
sumer Payments Research Center (CPRC), which regularly conducts the
Survey of Consumer Payment Choice (SCPC).18 The cross-sectional data
set conducted in October 2012 consists of 2065 participants whose responses
were weighted to represent all U.S. consumers aged 18 years and older. The
survey is implemented by the RAND Corporation as an online survey using
RAND’s American Life Panel (ALP). It is a unique, comprehensive, publicly
available, and representative survey that provides detailed payment informa-
tion of individual consumers with respect to nine common payment methods
in the U.S.19
The survey primarily measures the adoption and use of these payment
instruments by employing a flexible reporting strategy to enhance recall and
optimize the accuracy of the number of payments.20 However, low value pay-
ments, which are mostly paid in cash, tend to get more easily forgotten due
17The SCPC partitions POS payments into a third type of payment such as person-to-person payments. However, only a minor share of transactions is made from person-to-person why the effect on this type of payment is not analyzed (see section 4.5.2). Fur-thermore, the focus of the analysis lies on payments made through a retail establishment.Retail payments comprise purchases of goods at stores such as grocery stores, superstores,department stores and drug stores. Services payments include purchases of services suchas restaurants, bars, fast food and beverage, transportation and tolls, doctor’s visits,child care, haircuts, education, recreation and entertainment. Person-to-person paymentsare payments to people not made through a retail establishment such as payments forallowances, paying back a friend, or gifts to other people (cf. Schuh and Stavins, 2014).
18I refer to Schuh and Stavins (2014) for a comprehensive description of the data, asynopsis of the results and detailed information about the collection process.
19These include cash, check, money order, traveler’s check, debit card, credit card,prepaid card, online banking bill payments (OBBP), and bank account number payments(BANP).
20Typical periods used to measure the number of payments were during a week, amonth or a year.
152
to their high frequency and low budget impact. Cleaning procedures were im-
plemented for the number of monthly payments for each payment instrument
by defining upper limits based on the number of adopted instruments and an
extreme limit on the total number of monthly payments (300 total payments)
(Schuh and Stavins, 2014). Also, the SCPC asks consumers to evaluate six
payment instrument characteristics for each payment method. These ratings
may be vulnerable to incomplete information, memory loss, estimation, or
even subjective perceptions because consumers base their ratings on their
own objective knowledge. The data set also provides rich information about
consumer demographic characteristics, financial status, and residential state.
It it worth noting that the estimates are not adjusted for seasonal variation,
inflation or item non-response (missing values).
4.5.2 Description
The survey addresses the specific question whether the respondent has made
any form of mobile payment in the past 12 months prior to the survey using
a mobile phone. The incidence of use of mobile payment has been split
into specific activities to enhance recall. Table 4.1 presents the share of
consumers having used mobile payment on an annual basis distinguished by
different mobile payment concepts. As can be inferred, 18 percent of the
respondents have used mobile payment within the past year. Respondents
mostly undertake mobile payments by using a web browser (12 percent),
an application (7 percent), or a device attached to the mobile phone (6
percent) followed by sending a text message (3 percent), scanning a barcode
(2 percent), or using the contactless feature (1 percent). All these concepts
enable to purchase goods and services at a stationary POS. However, the
survey does not question the exact number of mobile payment transactions
made as well as how mobile payments are generally funded.
The data set also gives insights in the adoption rates of available payment
instruments at the POS (see Table 4.2). Not surprisingly, every respondent
in the sample adopts cash, as it offers ubiquitous payment. The adoption
of checks is also widespread (85 percent) as is the one of debit cards (78
percent) and of credit cards (72 percent). Prepaid cards are somewhat less
153
Table 4.1: Usage of Mobile Payment on an Annual Basis
Variable Mean NTotal 0.18 2032
Text/SMS 0.03 2032Contactless 0.01 2032Scanned a barcode 0.02 2031Mobile phone’s web browser 0.12 2032Mobile application 0.07 2031Device attached to mobile phone 0.06 2031
Note: Usage describes the fact that respondents make the corresponding type of payment atleast once in a typical year. Survey weights used.
preferred (52 percent). Table 4.3 provides in addition to summary statis-
tics a simple mean comparison test (t-test) between mobile payment users
(innovator) and non-users (non-innovator). Interestingly, innovators signif-
icantly possess more debit and prepaid cards than non-innovators (15 and
13 percentage points, respectively), providing ad hoc evidence for preferred
electronic payment methods.
Table 4.2: Adoption Rates of POS Payment Instruments
Variable Mean NCash 1 2032Check 0.85 2031Debit Card 0.78 2031Credit Card 0.72 2030Prepaid Card 0.52 2029
Note: Survey weights used.
As individuals, however, usually adopt different payment portfolios, I
constructed 16 possible payment bundles assuming every individual adopts
cash (see Table 4.4). As shown in Table 4.4, the majority of consumers in the
sample have all five payment instruments available (around 30 percent).21
Roughly 28 percent of individuals adopt the portfolio of four instruments
without prepaid cards and around seven percent without credit cards. The
21Available payment instruments at the POS include cash, check, debit, credit andprepaid cards.
154
Table 4.3: Differences in Adoption Rates of POS Payment Instruments
Non-Innovator Innovator t-Test
Variable Mean N Mean N Mean Diff.Cash 1 1699 1 328 0.00Check 0.85 1697 0.87 327 −0.02Debit Card 0.76 1698 0.91 327 −0.15 ***Credit Card 0.71 1700 0.75 328 −0.04Prepaid Card 0.5 1701 0.63 328 −0.13 ***
Note: Survey weights used. T-test of mean differences of innovator and non-innovator. Theycan differ from true values due to rounding and weighting. Significance levels 1% ***, 5% **and 10% *.
payment instrument portfolios “cash, check, debit” and “cash, check, credit”
are held by around seven percent of individuals. Cash held together with pre-
paid cards are widespread among five percent of consumers whereas around
two percent solely rely on cash.
Table 4.4: Adoption Rates of Payment Portfolios
Bundle Mean(1) all five instruments 0.303(2) cash, debit, credit, prepaid 0.012(3) cash, check, credit, prepaid 0.047(4) cash, check, debit, prepaid 0.074(5) cash, check, debit, credit 0.275(6) cash, check, debit 0.075(7) cash, check, credit 0.070(8) cash, check, prepaid 0.007(9) cash, debit, credit 0.009(10) cash, debit, prepaid 0.024(11) cash, credit, prepaid 0.005(12) cash, check 0.004(13) cash, debit 0.015(14) cash, credit 0.003(15) cash, prepaid 0.053(16) only cash 0.024
Note: Survey weights used. N = 2065. Available payment instruments at the POS are cash,check, debit card, credit card and prepaid card.
Additionally, a simple mean comparison test between innovators and non-
innovators indicates significant differences in the adoption rates of payment
155
bundles (see Table 4.5). First, mobile payment users have statistically sig-
nificantly higher adoption rates of the payment portfolio encompassing all
payment methods than non-users (around 18 percentage points). Second,
they adopt the portfolios “cash, check, credit, prepaid”, “cash, check, credit”
and “cash, check” significantly less often. What these bundles have in com-
mon is the absence of debit cards, which seem to be linked to mobile payment
users.
Table 4.5: Differences in Adoption Rates of Payment Portfolios
Non-Innovator Innovator t-Test
Bundle Mean Mean Mean Diff.(1) all five instruments 0.271 0.446 −0.175***(2) cash, debit, credit, prepaid 0.012 0.008 0.004(3) cash, check, credit, prepaid 0.055 0.011 0.044***(4) cash, check, debit, prepaid 0.075 0.066 0.009(5) cash, check, debit, credit 0.277 0.265 0.012(6) cash, check, debit 0.079 0.06 0.019(7) cash, check, credit 0.084 0.006 0.078***(8) cash, check, prepaid 0.006 0.012 −0.007(9) cash, debit, credit 0.008 0.015 −0.007(10) cash, debit, prepaid 0.021 0.037 −0.016(11) cash, credit, prepaid 0.006 0.001 0.005(12) cash, check 0.005 0 0.005**(13) cash, debit 0.015 0.013 0.002(14) cash, credit 0.004 0 0.004(15) cash, prepaid 0.055 0.049 0.006(16) only cash 0.027 0.01 0.017
N 1688 327
Note: T-test of mean differences of innovator and non-innovator. They can differ from truevalues due to rounding and weighting. Significance levels 1% ***, 5% ** and 10% *. Surveyweights used.
To estimate the impact of mobile payment on the usage of traditional
payment methods, I refer to Table 4.6 that shows descriptive statistics re-
garding the number of transactions made by different payment methods for
various transactions types within a month. On average, consumers under-
take roughly 43 POS payments a month, from which 24 are retail and 15
services payments.22 These are usually paid in cash (around 16 POS trans-
22Note that person-to-person payments account for around 3 transactions out of thetotal POS payments.
156
actions, nine and six retail and services transactions, respectively). Debit
cards are used the second most often in these contexts, namely around 13
POS payments a month including eight and five retail and services transac-
tions, respectively. Credit cards are somewhat less frequently deployed with
an average of 9 POS transactions entailing six retail and four services pay-
ments. Checks and most prominently prepaid cards are not very popular in
usage at the POS (roughly three transactions).
Table 4.6: Number of Transactions per Month by Payment Instrument andType
Variable Mean SD Min. Max.Total POS Payments 42.83 37.27 0 302.56
Cash 16.18 19.86 0 126.1Check 2.84 5.7 0 41.67Debit Card 13.14 19.8 0 111.89Credit Card 9.42 18.1 0 130.45Prepaid Card 0.62 2.74 0 35
Retail Payments 24.12 23.69 0 154.79Cash 8.66 12.26 0 65.22Check 1.23 3.3 0 30Debit Card 8.16 13.24 0 86.96Credit Card 5.63 11.47 0 78.27Prepaid Card 0.35 1.77 0 21.74
Services Payments 15.37 16.48 0 130.45Cash 5.61 8.93 0 86.96Check 1.11 2.59 0 21.74Debit Card 4.66 8.4 0 80Credit Card 3.66 8 0 65.22Prepaid Card 0.27 1.53 0 26.09
Note: Survey weights used. N = 2041. Subcategories do not exactly sum to main categorydue to rounding, weighting and dropping money orders and person-to-person payments.
Comparing the means of innovators and non-innovators by a t-test reveals
that the former undertake significantly more POS payments than the latter
(approximately six transactions), referring to Table 4.7. This holds also true
for debit card POS payments (around five transactions) while, in contrast,
innovators roughly make one check POS transaction less than non-innovators.
Moreover, significant differences are observed in retail payments made by
checks and prepaid cards for mobile payment users (approximately 0.5 and
0.2 fewer transactions, respectively) as well as by debit cards (around three
157
payments more). Innovators overall pay more frequently for services than
non-innovators, namely around two transactions. They significantly use their
debit cards more and checks less often for this transaction type, respectively
(roughly two vs. 0.5 transactions). It is important to note that the survey
does not report which and how many card payments are initiated, authorized
or confirmed through a mobile phone.
Table 4.7: Differences of Transactions per Month by Payment Instrumentand Type
Non-Innovator Innovator t-Test
Variable Mean SD Min. Max. Mean SD Min. Max. Mean Diff.Total POS Payments 42.1 36.46 0 302.56 47.84 40.29 0 236.67 −5.74 **
Cash 16.42 19.84 0 117.4 15.72 20.1 0 126.1 0.7Check 3.09 5.95 0 41.67 1.82 4.33 0 41 1.27***Debit Card 12.28 18.89 0 108.71 17.55 23.25 0 111.89 −5.27 ***Credit Card 9.08 17.45 0 130.45 11.34 20.96 0 109.04 −2.26Prepaid Card 0.61 2.69 0 35 0.65 3 0 26.09 −0.04
Retail Payments 23.74 23.04 0 138.46 26.8 26.37 0 154.79 −3.05Cash 8.85 12.3 0 65.22 8.14 12.19 0 65.22 0.71Check 1.33 3.41 0 30 0.82 2.82 0 23 0.50**Debit Card 7.67 12.73 0 86.96 10.74 15.26 0 86.96 −3.08 ***Credit Card 5.44 11 0 78.27 6.73 13.52 0 65.22 −1.28Prepaid Card 0.38 1.89 0 21.74 0.22 1.08 0 13.04 0.16*
Services Payments 15.04 16.18 0 130.45 17.48 17.73 0 87.05 −2.44 *Cash 5.7 8.99 0 86.96 5.43 8.77 0 52.18 0.27Check 1.23 2.76 0 21.74 0.59 1.58 0 15 0.65***Debit Card 4.29 7.84 0 80 6.55 10.48 0 52.18 −2.26 ***Credit Card 3.52 7.71 0 65.22 4.46 9.27 0 43.48 −0.93Prepaid Card 0.23 1.21 0 26.09 0.43 2.51 0 21.74 −0.20
N 1704 328
Note: Subcategories do not exactly sum to main category due to rounding, weighting anddropping money orders and person-to-person payments. N = 2041. T-test of mean differencesof innovator and non-innovator. They can differ from true values due to rounding andweighting. Significance levels 1% ***, 5% ** and 10% *. Survey weights used.
For the purpose of analyzing the effect of mobile payment on the usage
of payment instruments, I constructed an individual dummy variable indi-
cating the most frequently used payment method for every transaction type
at the POS. Table 4.8 presents the payment choice frequencies in the sam-
ple revealing that debit cards are the most preferred payment instrument
followed by cash (around 37 vs. 35 percent). 21 percent of consumers most
158
frequently choose credit cards while checks and prepaid cards are less fre-
quently selected as primary payment choice (4 vs. 1 percent). Interestingly,
consumers do not significantly vary their primary payment instrument for
different transaction types such as retail and services payments.
Table 4.8: Payment Choice Frequencies in the Sample
Variable Total POS Retail ServicesPayments Payments Payments
Cash 0.352 0.355 0.353Check 0.040 0.044 0.069Debit Card 0.368 0.376 0.353Credit Card 0.210 0.216 0.214Prepaid Card 0.012 0.015 0.014
Note: Survey weights used. The share of payment instruments most frequently used isdisplayed. Total POS payments do not include person-to-person payments.
The survey also provides rich information about financial and demo-
graphic consumer characteristics. Table 4.9 compares individual attributes
of innovators with those of the entire sample. Mobile payment users are gen-
erally younger, more educated, and richer than an average consumer. They
are mostly male, working, and more likely to already have been bankrupt
once. This is in line with previous studies (see section 4.2).
To sum up, the descriptives have shown that significant differences in the
adoption and usage patterns of mobile payment users and non-users exist.
There is suggestive evidence that mobile payment influences the adoption of
payment portfolios – especially towards holding all payment instruments –
and leads to increased usage of debit cards and decreased usage of checks
for overall POS payments. Also, consumers who are younger, richer, more
educated, and male are more likely to use mobile payment technology. As
a whole, individuals in the sample most frequently pay by debit cards and
cash at the POS.
As pointed out in section 4.2, perceived characteristics of payment instru-
ments explain a significant amount of taste variation for payment methods.
A major advantage of the SCPC is that respondents – both adopters and
non-adopters of payment instruments – assessed the attributes such as se-
159
Table 4.9: Sample Summary Statistics
All Innovator
Variable Mean SD Min. Max. N Mean SD Min. Max. NAge
<25 0.06 0 1 2065 0.09 0 1 32825–34 0.24 0 1 2065 0.4 0 1 32835–44 0.16 0 1 2065 0.24 0 1 32845–54 0.19 0 1 2065 0.15 0 1 32855–64 0.16 0 1 2065 0.08 0 1 328>65 0.18 0 1 2065 0.03 0 1 328
Education<High School 0.07 0 1 2065 0.05 0 1 328High School 0.35 0 1 2065 0.26 0 1 328Some College 0.29 0 1 2065 0.31 0 1 328College 0.17 0 1 2065 0.24 0 1 328Post Graduate 0.12 0 1 2065 0.14 0 1 328
Income (in 1000)<25 0.23 0 1 2062 0.16 0 1 32825–49 0.25 0 1 2062 0.23 0 1 32850–74 0.19 0 1 2062 0.21 0 1 32875–99 0.13 0 1 2062 0.18 0 1 328100–124 0.09 0 1 2062 0.07 0 1 328>125 0.11 0 1 2062 0.14 0 1 328
EmploymentWorking 0.61 0 1 2065 0.78 0 1 328Retired 0.19 0 1 2065 0.04 0 1 328Unemployed 0.09 0 1 2065 0.1 0 1 328
OthersMale 0.48 0 1 2065 0.54 0 1 328Female 0.52 0 1 2065 0.46 0 1 328Household Size 3 1.63 1 11 2065 3.34 1.71 1 11 328Bankruptcy 0.01 0 1 2020 0.03 0 1 326
Note: Survey weights used. Bankruptcy refers to having been bankrupt in the past 12 months.
160
curity, setup, acceptance, cost, records, and convenience for every existing
payment instrument on an absolute scale from one to five, where the latter
implies the strongest view. The assessment of these payment instrument
characteristics is described in Appendix 4.A.
For the purpose of this paper, the perceived characteristics of every pay-
ment method applicable at the POS were constructed as the average of each
respondent’s perception of each payment method relative to every other pay-
ment method at the POS similar to the approach in Schuh and Stavins
(2013). They were calculated as
RCHARki(j, j′) ≡ log
(CHARkij
CHARkij′
), (4.14)
where k describes the six characteristics such as security, setup, acceptance,
cost, records, and convenience, i indexes the consumer, j relates to the in-
teresting payment instrument applicable at the POS, and j′ is every other
payment instrument besides j that is commonly used at the POS. As for the
baseline specification and to account for the number of available payment
instruments, the log relative characteristics were transformed as
RCki(j) ≡1
Ji
∑
j′ 6=j
RCHARki(j, j′) (4.15)
over all available POS payment instruments Ji = 5 for consumer i resulting
in the average relative characteristics for each payment attribute k. For
instance, RCcosti(credit card) represents the average of the log ratios of the
perceived credit card cost for consumer i to the cost of each of the other
payment alternatives for consumer i. The construction is applied to every
consumer regardless of the adoption stage of payment methods, i.e. relative
to all payment instruments. However, to explicitly deal with the different
bundles b ∈ B in the adoption stage, I constructed the relative perceived
characteristics RCkib of bundle b as
RCki(b) ≡1
Ji
∑
j∈b
RCki(j), (4.16)
where Ji is the number of payment instruments adopted by consumer i.
161
4.6 Results
In this section, I first discuss the estimation results of the random utility
model according to the adoption stage (see section 4.6.1). Thereby, I evaluate
the effect of mobile payment on the adoption patterns of payment portfolios.
Second, I proceed by presenting the estimation results of the usage stage
(see section 4.6.2). I will follow by examining the effect of mobile payment
on the usage of payment instruments in different contexts.
4.6.1 Estimation Results of the Adoption Stage
First, this section presents the estimation results of the adoption stage re-
gression. Second, the effect of mobile payment on the adoption of payment
portfolios is discussed.
Results of the Conditional Logit Model
The estimation results of the conditional logit model are reported in Tables
4.10 and 4.11. Testing for IIA by the Hausman-McFadden test implies that
the model is well specified, i.e. dropped alternatives are irrelevant in the
majority of cases.23 Additionally, I employ a Wald test that reveals that
mobile payment creates a statistically significant improvement in the fit of
the model.24 This test also demonstrates that the effects are not statistically
different from each other in predicting the adoption of different portfolios
compared to the base outcome.
Overall, I find corroborating evidence that mobile payment basically ex-
erts a negatively, statistically significant effect on the probability of choosing
all other payment portfolios relative to the probability of having adopted all
five payment methods available, holding all else constant. For instance, it
is less likely that individuals who have used mobile payment adopt payment
23Test statistics are not provided. The Hausman-McFadden test compares two estima-tors of the same parameter, from which one is consistent and efficient (IIA holds) while theother is consistent, but inefficient. The first estimator is obtained by a correctly specifiedmodel while the second is obtained by estimating the model on a restricted number ofpayment bundles (cf. Hausman and McFadden, 1984).
24Test statistics are not shown.
162
portfolios including checks compared to the base outcome. Most prominently,
they are significantly less likely to jointly adopt cash and checks, cash and
credit cards as well as to solely rely on cash.
Coefficients of the perceived attributes of payment instruments have in-
tuitive signs and are highly statistically significant indicating that a more
positive rating increases the demand for one payment bundle while decreases
the demand for the remaining portfolios. In other words, utility of payment
portfolios is increasing in perceived characteristics. For instance, consumers
who rate the basket of adopted payment methods as relatively more secure
and convenient are more likely to adopt it, which aligns with previous stud-
ies (e.g. Schuh and Stavins, 2015a; Arango et al., 2015a). Education, age,
whether being male and having ever been bankrupt is statistically signifi-
cant in most of the cases across bundles. The probability of adopting only
cash, cash and debit cards as well as cash and prepaid cards, for example,
is decreasing if the level of education increases. Furthermore, the results
illustrate that the inclusion of attitudinal characteristics of mobile payment
can significantly explain heterogenous effects of consumers across payment
portfolios.
However, coefficients are cumbersome to interpret in nonlinear models
such as the conditional logit model. For this reason, I will display average
marginal effects in the next section.
Average Marginal Effects
To quantitatively appraise changes in payment portfolio choice subsequent
to mobile payment, I compute the average predicted probabilities of choosing
each of the payment bundles with and without having used mobile payment
in the model as
AMEMP =1
N
N∑
i=1
{(Pij |MPi = 1)− (Pij |MPi = 0)} (4.17)
resulting in the average marginal effect (AMEMP ) of mobile payment for
a typical person. The effect of mobile payment on the portfolio choice is
evaluated as the difference in the predicted probabilities of whether having
163
Table 4.10: Conditional Logit Estimates: Adoption Stage
Variables DC/CC/SVC CHK/CC/SVC CHK/DC/SVC CHK/DC/CC CHK/DC CHK/CC CHK/SVC
MP −0.327 −1.331*** −0.589** −0.464*** −0.621** −1.502*** −0.533(0.620 ) (0.435 ) (0.294 ) (0.170 ) (0.314 ) (0.473 ) (1.002 )
Age −0.180 0.044 −0.019 −0.064** −0.148*** −0.047 −0.242**(0.142 ) (0.049 ) (0.047 ) (0.027 ) (0.045 ) (0.048 ) (0.108 )
Age2 0.001 0.000 0.000 0.001** 0.001** 0.001* 0.002**(0.002 ) (0.000 ) (0.000 ) (0.000 ) (0.000 ) (0.000 ) (0.001 )
Education −0.721*** 0.123 −0.570*** −0.209*** −0.524*** −0.183* −1.535***(0.170 ) (0.096 ) (0.097 ) (0.061 ) (0.119 ) (0.105 ) (0.521 )
Working −0.228 0.224 −0.519 −0.084 −0.134 0.198 1.537(1.073 ) (0.320 ) (0.321 ) (0.194 ) (0.329 ) (0.304 ) (1.039 )
Retired 2.213* 0.359 −0.610 −0.006 −0.364 0.410 −17.312 ***(1.249 ) (0.337 ) (0.425 ) (0.226 ) (0.433 ) (0.326 ) (0.789 )
Other employment 0.828 0.330 0.466 −0.021 0.110 −0.044 1.542(0.790 ) (0.299 ) (0.324 ) (0.202 ) (0.355 ) (0.295 ) (1.124 )
Male 0.791 0.482** 0.380* 0.347*** 0.436** 0.335* −0.253(0.697 ) (0.195 ) (0.215 ) (0.125 ) (0.219 ) (0.200 ) (1.033 )
Household size 0.082 −0.026 0.079 −0.068 −0.013 −0.137 0.177(0.130 ) (0.069 ) (0.073 ) (0.045 ) (0.086 ) (0.107 ) (0.257 )
Bankruptcy −16.515 *** −16.571 *** 2.516*** −1.217 2.378*** −16.446 *** −16.454 ***(1.058 ) (0.546 ) (0.736 ) (1.091 ) (0.766 ) (0.514 ) (1.520 )
MP internet security −0.287 0.133 0.034 0.080 −0.087 0.079 −0.738**(0.478 ) (0.127 ) (0.134 ) (0.079 ) (0.140 ) (0.133 ) (0.366 )
MP text security 0.367 −0.120 −0.107 0.191** 0.268* 0.282** 0.607(0.422 ) (0.134 ) (0.151 ) (0.080 ) (0.149 ) (0.125 ) (0.544 )
MP voice security 0.014 −0.065 0.100 −0.139* −0.107 −0.235* 0.448(0.452 ) (0.124 ) (0.131 ) (0.075 ) (0.129 ) (0.121 ) (0.476 )
Security 1.205***(0.266 )
Setup 2.891***(0.434 )
Acceptance 1.382**(0.566 )
Cost 1.019***(0.330 )
Records 1.835***(0.306 )
Convenience 2.762***(0.406 )
Constant 3.547 −4.165*** 1.463 1.444* 3.795*** −1.150 3.473(2.314 ) (1.495 ) (1.206 ) (0.745 ) (1.197 ) (1.530 ) (3.255 )
Note: Adopting cash (CSH), check (CHK), debit (DC), credit (CC) and prepaid card (SVC) as elements of the payment portfolio is the base outcome. Cash iselement of all payment portfolios. Base category for employment is unemployed. Bankruptcy refers to having been bankrupt in the past 12 months. MP is mobilepayment. Robust standard errors are in parentheses. The number of cases is 1987, the number of observations is 31’792 and the log(likelihood) is -3076. Income hasbeen dropped due to collinearity. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
164
Table 4.11: Conditional Logit Estimates: Adoption Stage (Cont.)
Variables DC/CC DC/SVC CC/SVC CHK DC CC SVC CSH
MP −1.209* −0.208 −0.286 −16.980 *** −0.887 −16.435 *** 0.209 −2.169**(0.703 ) (0.424 ) (1.282 ) (0.501 ) (0.600 ) (0.797 ) (0.478 ) (1.063 )
Age −0.142 0.177 0.027 −0.274 −0.134 −0.289 −0.094 0.064(0.138 ) (0.130 ) (0.190 ) (0.172 ) (0.104 ) (0.194 ) (0.099 ) (0.252 )
Age2 0.000 −0.003* 0.000 0.003 0.001 0.003 0.000 −0.002(0.002 ) (0.002 ) (0.002 ) (0.002 ) (0.001 ) (0.002 ) (0.001 ) (0.003 )
Education −0.383 −0.630*** −1.304** −0.192 −0.764*** −0.824 −1.306*** −1.021***(0.278 ) (0.143 ) (0.594 ) (0.449 ) (0.223 ) (0.659 ) (0.267 ) (0.318 )
Working 0.295 −0.406 1.893 0.956** −0.486 −1.310 −1.425*** −1.005(0.930 ) (0.459 ) (1.505 ) (0.421 ) (0.524 ) (0.812 ) (0.516 ) (0.737 )
Retired 1.927 0.666 1.948 −1.476 −0.967 −17.143 *** 0.286 0.827(1.898 ) (1.401 ) (1.614 ) (2.328 ) (1.421 ) (1.769 ) (0.910 ) (2.164 )
Other employment −0.700 0.511 1.412** 1.261 1.537*** −15.851 *** 0.906* 0.956(1.070 ) (0.467 ) (0.710 ) (1.134 ) (0.575 ) (1.624 ) (0.502 ) (0.747 )
Male 0.764 −0.207 0.696 2.131*** 0.878* 1.338 0.152 1.840***(0.524 ) (0.406 ) (0.686 ) (0.827 ) (0.498 ) (1.232 ) (0.402 ) (0.569 )
Household size −0.184 0.156 0.244 0.393 0.002 −0.133 0.009 0.084(0.234 ) (0.103 ) (0.223 ) (0.244 ) (0.153 ) (0.415 ) (0.107 ) (0.171 )
Bankruptcy −16.608 *** −16.216 *** −16.185 *** −15.324 *** 2.272* −14.792 *** −16.336 *** −16.321 ***(0.906 ) (0.740 ) (1.015 ) (1.509 ) (1.232 ) (1.068 ) (1.119 ) (0.991 )
MP internet security −0.362 −0.070 0.241 −0.073 0.129 0.587 0.078 0.451(0.286 ) (0.217 ) (0.450 ) (0.299 ) (0.276 ) (0.469 ) (0.332 ) (0.290 )
MP text security 0.435 0.014 −0.092 0.431 0.396 0.441 0.459 −0.056(0.270 ) (0.213 ) (0.571 ) (0.291 ) (0.266 ) (0.416 ) (0.286 ) (0.308 )
MP voice security −0.296 0.026 0.167 −0.928*** −0.271 −0.824*** −0.205 −0.023(0.351 ) (0.168 ) (0.424 ) (0.268 ) (0.321 ) (0.289 ) (0.285 ) (0.281 )
Security 1.205***(0.266 )
Setup 2.891***(0.434 )
Acceptance 1.382**(0.566 )
Cost 1.019***(0.330 )
Records 1.835***(0.306 )
Convenience 2.762***(0.406 )
Constant 2.754 −2.627 −5.455 0.314 1.468 4.182 4.800** −1.719(3.133 ) (2.502 ) (3.405 ) (4.028 ) (2.990 ) (5.493 ) (2.022 ) (4.890 )
Note: Adopting cash (CSH), check (CHK), debit (DC), credit (CC) and prepaid card (SVC) as elements of the payment portfolio is the base outcome. Cash iselement of all payment portfolios. Base category for employment is unemployed. Bankruptcy refers to having been bankrupt in the past 12 months. MP is mobilepayment. Robust standard errors are in parentheses. The number of cases is 1987, the number of observations is 31’792 and the log(likelihood) is -3076. Income hasbeen dropped due to collinearity. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
165
used mobile payment, while holding all other factors constant. In this way,
substitution patterns across payment portfolios can be computed.
The predicted probabilities are set out in the first column in Table 4.12,
which further can be compared to the actual frequencies of payment portfo-
lios in the sample (see Table 4.4). It is well noticeable that the overall fit
of the model is fairly satisfied since the predicted choice probabilities corre-
spond closely to the actual frequencies in the entire sample. In other words,
the explanatory variables in the model can predict the choice of payment
portfolios rather precisely. This implies that every change in the observed
variables, which lead to a change in the predicted probabilities, is actually
closely related to the observed frequencies. However, the average predicted
probabilities are somewhat more accurate for specific payment bundles, while
they are less precise for other choices. The reason may be that some alterna-
tives have considerably higher shares in the sample and are therefore better
approximated than less preferred alternatives.
In addition, average marginal effects (AME) of mobile payment for every
choice alternative are depicted in the third column in Table 4.12. As can be
inferred, the signs of the AME vary across payment bundles and the mag-
nitude of the effects are very modest compared to the initial probabilities,
meaning that the probability of adopting payment portfolios is not highly
dependent on mobile payment.25 There are two main findings: First, mobile
payment increases the probability of adopting the payment bundle includ-
ing all POS payment instruments by around 2.1 percentage points at the
expense of reducing the probability of choosing portfolios primarily entailing
checks (with fewer than five instruments) and cash as a sole instrument. The
reductions range from –0.05 to –0.42 percentage points. Second, consumers
are generally more likely to adopt payment bundles that encompass more
than one payment card if they have mobile payment (maximum increase of
+0.12 percentage points).
To conclude, mobile payment positively influences the probability of hav-
ing all POS payment methods. This indicates that it does not replace phys-
25Attitudinal characteristics of payment methods are expected to have greater impacton the adoption choice (cf. Schuh and Stavins, 2013).
166
ical payment cards. Conversely, it reduces the probability of adopting only
cash and payment portfolios with checks, which include fewer than five in-
struments. This suggests that mobile payment principally does not emerge
as a substitute for payment cards while it does so for paper-based payment
options.
Table 4.12: Adoption Stage: Average Marginal Effect of Mobile Payment
Bundle Predicted Predicted AverageProb. Prob. w/o MP Marginal Effect
all 29.325 27.188 2.137DC/CC/SVC 1.124 1.104 0.019CHK/CC/SVC 5.076 5.458 -0.382CHK/DC/SVC 8.340 8.618 -0.279CHK/DC/CC 27.336 27.680 -0.342CHK/DC 7.504 7.776 -0.273CHK/CC 6.552 6.975 -0.423CHK/SVC 0.715 0.726 -0.011DC/CC 1.068 1.307 -0.240DC/SVC 2.614 2.499 0.115CC/SVC 0.472 0.463 0.008CHK 0.452 0.504 -0.052DC 1.835 1.980 -0.145CC 0.264 0.352 -0.088SVC 4.902 4.538 0.363CSH 2.422 2.831 -0.409
Note: Survey weights used. Numbers are in percentages and percentage points, respectively.CSH refers to cash, CHK to check, DC to debit, CC to credit and SVC to prepaid card.
4.6.2 Estimation Results of the Usage Stage
This section proceeds by first discussing the estimation results of the us-
age stage regression. Second, the impact of mobile payment on the use of
traditional payment instruments in different payment contexts is presented.
Results of the Nested Logit Model
The estimation results of the usage stage regarding the transaction type of
overall POS payments is displayed in Table 4.13. More specifically, Tables
4.14 and 4.15 separately show the estimation results of the usage stage in
167
terms of retail and services payments. At the bottom of all models estimated,
the dissimilarity parameters λ of both nests are presented, revealing highly
significant results of the Likelihood Ratio Test that leads to a strong rejection
of the conditional logit model in favor of the nested logit model, i.e. there is
strong evidence for correlated errors. The parameter λ of the nest comprising
payment cards is throughout smaller than one indicating that payment cards
are closer substitutes than across the other group of paper-based methods.
In contrast, this is not true for the nest comprising paper-based payment
methods since λ is slightly greater than one. Overall, there is compelling
evidence that the model is appropriately specified and consistent with theory.
According to Table 4.13, the coefficients of mobile payment all show the
expected sign meaning that mobile payment positively affects the use of
payment cards and negatively affects the use of checks for overall POS pay-
ments compared to cash. It exerts a statistically significant effect on the
usage of prepaid cards relative to cash while the effect is insignificant for the
remaining payment instruments such as check, debit, and credit cards, hold-
ing all else constant. The corresponding joint significant test (Wald test)
affirms that mobile payment has no significant impact on the use of pay-
ment instruments and thus does not statistically significantly improve the
fit of the model.26 Similar findings are provided for retail and services pay-
ments (see Tables 4.14 and 4.15). This may reflect the fact that consumers
are generally not eager to change their payment habits at the POS due to
mobile payment technologies. It probably results both from force of habit
and general resistance to new technologies, as Humphrey et al. (1996) argue
that individual payment patterns strongly depend on the past compositions
concluding inertia in payment instrument use.27
Estimates of the perceived characteristics of payment methods all have
expected signs and are highly statistically significant. The higher payment
attitudes of one payment instrument are evaluated, the more likely it is that
the corresponding instrument is more heavily used while the probability of
26Test statistics are not provided.27Another rationale could be the fact that the possibility to use mobile payment at the
POS is not widely applicable. Also, mobile payment as a new way of payment may notbe able to overwhelm the benefits of traditional payment instruments.
168
Table 4.13: Nested Logit Estimates Usage Stage: POS Payments
Variables Check Debit Credit PrepaidMP −1.465 0.108 0.034 0.493*
(1.021 ) (0.161 ) (0.166 ) (0.281 )Age 0.046 −0.019 −0.040 −0.040
(0.079 ) (0.025 ) (0.025 ) (0.055 )Age2 0.000 0.000 0.000 0.000
(0.001 ) (0.000 ) (0.000 ) (0.001 )Education 0.063 0.168*** 0.243*** 0.181
(0.166 ) (0.059 ) (0.060 ) (0.127 )Income −0.103 0.076* 0.148*** −0.027
(0.125 ) (0.041 ) (0.041 ) (0.110 )Working 0.656 0.798*** 0.736*** 0.614*
(0.510 ) (0.180 ) (0.184 ) (0.330 )Retired 1.719** 0.632*** 0.693*** 1.291**
(0.667 ) (0.226 ) (0.230 ) (0.561 )Other employment 0.452 0.176 0.271 0.212
(0.452 ) (0.181 ) (0.186 ) (0.321 )Male −0.668* −0.586*** −0.520*** −1.062***
(0.364 ) (0.116 ) (0.118 ) (0.354 )Household size −0.086 −0.060 −0.092** 0.050
(0.137 ) (0.041 ) (0.043 ) (0.063 )MP internet security 0.106 0.180** 0.168** 0.317**
(0.213 ) (0.074 ) (0.075 ) (0.148 )MP txt security −0.414* −0.048 −0.067 −0.205
(0.240 ) (0.077 ) (0.079 ) (0.140 )MP voice security 0.099 −0.001 0.019 −0.039
(0.200 ) (0.071 ) (0.072 ) (0.141 )Security 0.151***
(0.047 )Setup 0.629***
(0.114 )Acceptance 0.466***
(0.125 )Cost 0.341***
(0.064 )Records 0.430***
(0.075 )Convenience 1.063***
(0.134 )Constant −3.906 −0.740 −0.749 −0.688
(2.597 ) (0.683 ) (0.698 ) (1.268 )Observations 9230N 1846log(likelihood) -1783Paper λ 1.287***
(0.306)Card λ 0.257***
(0.033)
Note: Base outcome is cash. MP is mobile payment. Base category for employment is unemployed. Robuststandard errors are in parentheses. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
169
Table 4.14: Nested Logit Estimates Usage Stage: Retail Payments
Variables Check Debit Credit PrepaidMP −0.509 0.115 0.109 −0.004
(0.911 ) (0.182 ) (0.186 ) (0.431 )Age 0.281* −0.028 −0.048* 0.037
(0.157 ) (0.028 ) (0.028 ) (0.055 )Age2 −0.002* 0.000 0.000 0.000
(0.001 ) (0.000 ) (0.000 ) (0.001 )Education −0.078 0.191*** 0.286*** 0.210
(0.198 ) (0.065 ) (0.066 ) (0.129 )Income 0.037 0.108** 0.163*** −0.350**
(0.144 ) (0.046 ) (0.046 ) (0.176 )Working 0.490 0.744*** 0.656*** 0.062
(0.595 ) (0.197 ) (0.201 ) (0.332 )Retired 1.304* 0.632** 0.646** 0.195
(0.730 ) (0.250 ) (0.253 ) (0.464 )Other employment 0.589 0.169 0.217 −0.166
(0.542 ) (0.200 ) (0.204 ) (0.337 )Male −0.736* −0.642*** −0.542*** −0.223
(0.433 ) (0.129 ) (0.130 ) (0.250 )Household size −0.046 −0.012 −0.041 0.150**
(0.166 ) (0.046 ) (0.048 ) (0.075 )MP internet security −0.180 0.208** 0.182** 0.010
(0.279 ) (0.084 ) (0.085 ) (0.189 )MP text security −0.348 −0.017 −0.041 0.026
(0.290 ) (0.087 ) (0.088 ) (0.164 )MP voice security 0.230 −0.100 −0.054 0.067
(0.242 ) (0.079 ) (0.080 ) (0.150 )Security 0.134***
(0.049 )Setup 0.580***
(0.112 )Acceptance 0.460***
(0.131 )Cost 0.373***
(0.066 )Records 0.449***
(0.081 )Convenience 1.212***
(0.155 )Constant −10.599 ** −0.508 −0.620 −2.465*
(5.224 ) (0.766 ) (0.783 ) (1.445 )Observations 8430N 1686log(likelihood) -1603Paper λ 1.488***
(0.421)Card λ 0.266***
(0.035)
Note: Base outcome is cash. MP is mobile payment. Base category for employment is unemployed. Robuststandard errors are in parentheses. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
170
Table 4.15: Nested Logit Estimates Usage Stage: Services Payments
Variables Check Debit Credit PrepaidMP −0.685 0.048 0.046 0.358
(0.838 ) (0.207 ) (0.210 ) (0.282 )Age 0.099 0.005 −0.005 0.015
(0.116 ) (0.031 ) (0.031 ) (0.061 )Age2 −0.001 0.000 0.000 0.000
(0.001 ) (0.000 ) (0.000 ) (0.001 )Education 0.366 0.114 0.230*** 0.116
(0.315 ) (0.087 ) (0.089 ) (0.130 )Income 0.036 0.106** 0.188*** 0.105
(0.165 ) (0.053 ) (0.053 ) (0.093 )Working 0.623 0.591*** 0.385* −0.163
(0.724 ) (0.227 ) (0.229 ) (0.337 )Retired 1.558* 0.749** 0.809*** 0.631
(0.947 ) (0.305 ) (0.308 ) (0.601 )Other employment 0.006 0.153 0.156 −0.306
(0.645 ) (0.223 ) (0.226 ) (0.364 )Male −1.148* −0.527*** −0.484*** −0.589**
(0.606 ) (0.168 ) (0.170 ) (0.262 )Household size −0.075 −0.009 −0.051 0.069
(0.185 ) (0.053 ) (0.055 ) (0.071 )MP internet security 0.741* 0.357*** 0.344*** 0.435***
(0.397 ) (0.114 ) (0.115 ) (0.164 )MP text security −0.571 −0.072 −0.063 −0.294*
(0.400 ) (0.111 ) (0.112 ) (0.163 )MP voice security −0.112 −0.117 −0.104 −0.154
(0.261 ) (0.087 ) (0.089 ) (0.148 )Security 0.115**
(0.048 )Setup 0.496***
(0.113 )Acceptance 0.533***
(0.133 )Cost 0.323***
(0.066 )Records 0.492***
(0.085 )Convenience 0.910***
(0.143 )Constant −7.725 −1.095 −1.401 −0.830
(5.165 ) (0.875 ) (0.892 ) (1.390 )Observations 8015N 1603log(likelihood) -1666Paper λ 1.993**
(0.922)Card λ 0.252***
(0.037)
Note: Base outcome is cash. MP is mobile payment. Base category for employment is unemployed. Robuststandard errors are in parentheses. Significance levels are denoted as *** p<0.01, ** p<0.05, * p<0.1.
171
usage of the remaining instruments decreases. Put more simply, individuals
who see a specific payment mean as relatively cheaper, more secure, more ac-
cepted, more convenient, more easily to set up, and more supportive to track
payments are more likely to use it. Furthermore, I find overall evidence that
education, income, whether being male, retired or working – compared to
being unemployed – are statistically significant factors that predict the em-
ployment of payment methods. For instance, more educated, higher income
and working individuals are more likely to pay by debit and credit cards
for all types of POS payments compared to cash. Also, the likelihood of
paying by payment cards rises if consumers rate internet security of mobile
payments higher.
Notwithstanding the fact of mostly insignificant effects of mobile payment
in the regression analysis, I will focus on average marginal effects in the next
section.
Average Marginal Effects
Analogous to the previous section 4.6.1, I compute the average marginal ef-
fect of mobile payment for a typical person in different payment contexts
according to equation 4.17. The AMEMP basically represents the differ-
ence in the predicted probabilities of paying by instrument j for transaction
type h with and without mobile payment, holding all else constant. How-
ever, according to the model, the effects are not statistically relevant with
the exception of the impact on the choice of prepaid cards for overall POS
payments.
Table 4.16 exhibits the predicted probabilities for overall POS payments
in the first column. They are quite close to the actual frequencies in the
sample with the exception of cash payments (compare Tables 4.8 and 4.4).28
However, I conclude the overall fit of the model to be well satisfied, as the
choice probabilities of checks, debit, credit, and prepaid cards are rather
precise.
28The predicted probabilities of cash primarily differ from the sample frequencies dueto sample weights which tend to account for underrepresented cash payments.
172
The AMEMP of POS payments is shown in the third column of Table
4.16. It is noticeable that all effects have expected signs except credit cards
and their magnitude is very moderate ranging from –0.36 to +0.65 percent-
age points. This implies that the decision to pay by instrument j at the POS
is not considerably influenced by mobile payment. While mobile payment
typically decreases the probability to choose cash and check as a payment
mean at the POS (around –0.27 and –0.24 percentage points, respectively),
it increases the likelihood of paying by debit and prepaid card (+0.65 and
+0.23 percentage points, respectively). Ironically, the effect of mobile pay-
ment on the probability of paying by credit card is negative (–0.36 percentage
points). The reason may be that mobile payment users preferably rely on
debit and prepaid cards to which the amount of payment is charged. It is
also very likely that it negatively affects the use of paper-based methods in
favor of card-based methods, particularly debit card usage. This is more
pronounced in the context of retail payments, as the magnitude of the ef-
fect on cash and checks nearly offsets the one on debit card use. Thus, it
may suggests that mobile payment emerges as a substitute for paper-based
instruments, especially cash usage.
Table 4.16: Usage Stage: AME of Mobile Payment for POS payments
Instrument Predicted Predicted AverageProb. Prob. w/o MP Marginal Effect
Cash 39.155 39.428 -0.270Check 3.874 4.109 -0.240Debit 36.605 35.975 0.648Credit 19.320 19.673 -0.364Prepaid 1.046 0.815 0.227
Note: Survey weights used. Numbers are in percentages and percentage points, respectively.
Similar findings are separately reported both for retail and services pay-
ments, which are presented in Tables 4.17 and 4.18. Overall, the effects are
not very sizeable compared to the choice probabilities. As for retail payments,
mobile payment positively affects debit and credit card usage by around 0.37
and 0.07 percentage points, respectively, while it negatively influences cash,
173
checks, and prepaid card use (roughly –0.32, –0.09, and –0.03 percentage
points). Moreover, mobile payment has a negative impact on the proba-
bility to both use cash and check for services payments by approximately
–0.24 percentage points. In contrast, it fosters the usage of payment cards
for services payments ranging from +0.07 to +0.32 percentage points. This
could indicate that mobile payment especially facilitates electronic payment
processing at counters where consumers typically ask for faster checkout.29
Table 4.17: Usage Stage: AME of Mobile Payment for Retail Payments
Instrument Predicted Predicted AverageProb. Prob. w/o MP Marginal Effect
Cash 34.229 34.550 -0.317Check 3.728 3.816 -0.092Debit 39.925 39.557 0.374Credit 20.903 20.823 0.069Prepaid 1.215 1.254 -0.034
Note: Survey weights used. Numbers are in percentages and percentage points, respectively.
Table 4.18: Usage Stage: AME of Mobile Payment for Services Payments
Instrument Predicted Predicted AverageProb. Prob. w/o MP Marginal Effect
Cash 34.349 34.572 -0.243Check 6.058 6.288 -0.236Debit 36.950 36.879 0.095Credit 20.727 20.659 0.066Prepaid 1.915 1.602 0.317
Note: Survey weights used. Numbers are in percentages and percentage points, respectively.
To conclude, there is suggestive evidence in the usage stage that mo-
bile payment generally tends to be substitutional to paper-based payment
methods and complementary to card-based instruments with regards to POS
payments, especially to debit cards. However, the magnitude of the impact
29For instance, services payments comprise purchases at bars, fast food restaurants aswell as for transportation and tolls, amongst others.
174
is not very sizeable. With respect to the transaction context of retail and
services payments, these effects principally remain stable, but vary in terms
of magnitude and sign. Since the estimates of mobile payment, however, are
not throughout statistically significant, it suggests that current individual
payment compositions are not influenced by mobile payment technologies
and are thus determined by other factors. The usage of payment instru-
ments seems to be strongly habitual and unconscious, which in turn may
impede the change towards the use of innovative payment products.
4.7 Plausibility Check
At this stage, the question arises as to how reasonable and robust the results
are. I do not intend to focus on alternative estimation strategies here since
the presented test statistics of the estimated models all demonstrate that the
models are well specified. Rather, I aim at critically scrutinizing the data set
used with respect to the question targeting the usage of mobile payment. To
this end, I draw on the latest report by Brown et al. (2015) – published by the
Board of Governors of the Federal Reserve System (FRS) – who exclusively
examine the adoption and use of mobile banking and mobile payment as
well as individual interaction with financial institutions facilitated by mobile
phones and other technologies.30 The survey has been conducted online
every year in December since 2011, using a representative sample of the U.S.
population aged 18 and older. The latest survey dates back to 2014.
Comparing the responses of the survey question of mobile payment usage
in 2012 between the SCPC and FRS illustrates that the number of users
does not vary in a pronounced way (18 vs. 15 percent). In 2014, this figure
gradually increased to 22 percent. The FRS also shows that the median
reported frequency of mobile payment was two times in the month prior to
the 2014 survey, whereby roughly 18 percent of respondents stated to had
used it more than five times. Further, around 27 percent of mobile payment
users had applied it in the past year, but not in the month prior to the survey.
30Unfortunately, the survey does not provide any information on the adoption and useof other payment instruments.
175
This rather sporadic and low rate of usage may explain why mobile payment
does not causally impact the use of conventional payment instruments in
the empirical analysis above. It appears that this technology is still in its
infancy (cf. Rogers, 2003).
The most paramount reasons why individuals had never used mobile pay-
ment is the fact that it is easier to pay with other methods (75 percent), they
do not see a clear benefit from it (59 percent), and due to security concerns
(59 percent) (cf. Brown et al., 2015). Yet for a successful breakthrough of
mobile payment, it is hence essential that the payment industry attempts to
ameliorate the perception of mobile payment pertaining to these issues. It is
interesting to note that the facts and figures expressed above are similar to
the ones of the recent survey in Germany conducted by the Deutsche Bun-
desbank (2015): A tiny share of roughly two to four percent regularly uses
mobile payment. Thereby, security concerns as well as absent needs are the
main rationales for non-usage.
Another striking finding relates to the funding principle of mobile pay-
ment. According to Brown et al. (2015), debit cards (55 percent), credit
cards (51 percent), and bank account deductions (41 percent) are the most
prominent channels to fund mobile purchases. It is therefore likely that mo-
bile payment will affect card payments more extensively in near future the
more intensively it will be used.
4.8 Conclusion
This paper studied for the first time the impact of mobile payment on the
adoption and usage of traditional retail payment instruments at the POS us-
ing a comprehensive U.S. data set on individual payment patterns. Applying
the random utility framework for estimation yields the following important
results: First, mobile payment increases the probability of possessing all
available payment instruments at the POS by roughly two percentage points
and reduces the likelihood of adopting payment portfolios comprising checks
and only cash. This implies that it does not replace physical payment cards,
but substitutes paper-based payment methods. Second, no causal relation-
176
ship between mobile payment and the use of traditional payment means
except prepaid cards has been found. In other words, mobile payment does
not statistically significantly impact the usage of payment instruments at
the POS, but positively does so for prepaid cards. The estimation provides
supportive evidence that mobile payment principally serves as a complement
to card payments and as a substitute for paper-based payment methods such
as cash and checks, as it could allow for a more efficient checkout. Overall,
the results of the usage stage may reflect the fact that payment instrument
use seems to strongly depend on other factors such as perceived characteris-
tics of payment methods, individual habits, and automatism, amongst others.
The presented study fits into existing literature on consumer payment choice
and fills the gap in understanding the role of mobile payment in the retail
payment landscape.
The findings may have several important implications. First, the results
show that the payment industry should actively promote mobile payment
products, as it enhances the shift from adopting paper-based products to
electronic payment cards, which in turn is a necessary prerequisite for a
faster proliferation of electronic payment transactions. They also highlight
the important difference between the extensive and intensive margin of pay-
ment products and thus give advice for the private industry to make more
efforts to incentivize mobile payment usage, as more frequent use also tends
to foster electronic (card) payments and consequently to increase profits.
Second, the findings imply that the overall payment system can benefit from
decreasing social costs due to the shift from paper- to card-based payment
methods. Third, although mobile payment does not throughout exerts causal
effects on the use of payment means, policy makers should be aware of dif-
ferent consumer regulations and regulatory agencies that cover the payment
method used to fund mobile payment (cf. Martin, 2012).
The study is subject to a number of limitations. The information on
mobile payment usage in the survey may not be sufficiently extensive to an-
alyze its impact in detail. For instance, it is unclear how many times mobile
payment is deployed at the POS prior to the survey, which conversely could
affect the use of traditional payment instruments more significantly. Also,
177
the data set may suffer from recall effects since respondents may have forgot-
ten or could conceptually be uncertain about reporting the number of card
payments that have been undertaken through a mobile device. Consequently,
this would lead to a possible underestimation of the corresponding effects.
In addition, although the multiple forms of mobile payment enable to pay at
the POS, the survey does not provide sufficient information whether mobile
payments have been solely applied for POS payments. In this sense, the chan-
nels through which mobile payment affects traditional payment instrument
use may be confounded. Therefore, payment surveys that conduct detailed
information on the usage of different mobile payment concepts would help
to obtain more accurate results.
Furthermore, there are caveats that missing information about supply-
side factors in the data set could lead to correlated error terms. Because
mobile payment represents a relatively new form of payment, it could be
less accepted by merchants, for instance, due to inadequate infrastructure
or personal reservations. As a consequence, lacking payment options at the
POS results in negative feedback effects for consumers and thus in less mobile
payment usage. A potential hint is given by Rysman (2007), who finds
evidence that people first choose a specific card brand if a large number of
merchants accept this brand. In these circumstances, data on merchants’
mobile payment acceptance would help to mitigate this issue.
With respect to external validity, it is doubtable to what extent the find-
ings can be generalized to other countries, as major cultural and institutional
differences in payment markets across countries are prevalent. For instance,
the heavy and earlier reliance on payment cards rather than credit trans-
fer and cash in the U.S. compared to Europe could have induced American
consumers to be more open-minded towards innovative payment products in
terms of adoption and usage. Therefore, the magnitude and significance of
the effect of mobile payment may differ across payment areas. Also, there
are many different concepts of mobile payment technology with varying in-
cidence across countries. Whilst mobile payment may experience greater
proliferation in developing countries regarding person-to-person payments
due to the vast number of unbanked and underbanked persons, it may be
178
more successful in developed countries regarding POS payments. Thus, the
impact of mobile payment across countries may vary in terms of the under-
lying technological concept.
After all, it remains unclear how mobile payment affects the use of tradi-
tional payment instruments in detail. Possible are improved efficiency and
convenience compared to traditional payment instruments that drives its de-
ployment, as several studies presume (see section 4.2). However, to obtain
a clear picture of the specific channels through which mobile payment has
an effect on other payment methods and to provide detailed results of its
multiple forms of applications and features, it is necessary to collect qualita-
tively improved data on mobile payment usage in all its facets. I leave this
for future research.
179
4.A Appendix: Assessment of Payment MethodCharacteristics
In the SCPC, every respondent rated the payment characteristics security,
acceptance, cost, convenience, setup, and records of all payment instruments
available regardless whether he had adopted or used all of them. Available
payment methods include cash, checks, money order, debit, credit, and pre-
paid cards, bank account number, and online banking bill pay. The survey
questions are presented in Table A1.
180
Table A1: Assessment of Payment Instrument Characteristics
SECURITYSuppose a payment method has been stolen, misused, or accessed withoutthe owner’s permission. Please rate the SECURITY of each method againstpermanent financial loss or unwanted disclosure of personal information.
Very Risky RiskyNeither riskynor secure
Secure Very Secure
ACCEPTANCE FOR PAYMENTPlease rate how likely each payment method is to be ACCEPTED for paymentby stores, companies, online merchants, and other people or organizations.
Rarelyaccepted
Occasionallyaccepted
Oftenaccepted
Usuallyaccepted
Almostalwaysaccepted
COSTPlease rate the COST of using each payment method. Examples: Fees, penal-ties, postage, interest paid or lost, subscriptions, or materials can raise thecost of a payment method. Cash discounts and rewards (like frequent flyermiles) can lower the cost of a payment method. Consider the cost of usingor owning the payment method, not the cost of an item purchased. Pleasechoose one answer in each row for all payment methods.Very high
costHigh cost
Neither highnor low cost
Low costVery low
cost
CONVENIENCEPlease rate the CONVENIENCE of each payment method. Examples: speed,control over payment timing, ease of use, effort to carry, ability to keep orstore.
Veryinconvenient
Inconvenient
Neitherinconvenient
norconvenient
ConvenientVery
convenient
GETTING and SETTING UPRate the task of getting or setting up each payment method before you canuse it. Examples: getting cash at the ATM, length of time to get or set up,paperwork, learning to use or install it, or travel.Very hard toget or set up
Hard to getor set up
Neither hardnor easy
Easy to getor set up
Very easy toget or set up
PAYMENT RECORDSRate the quality of payment records offered by each payment method. Con-sider both paper and electronic records. Examples: proof of purchase, accountbalances, spending history, usefulness in correcting errors or dispute resolu-tion, or ease of storage.Very poorrecords
Poor recordsNeither good
nor poorGoodrecords
Very goodrecords
181
Chapter 5
Concluding Remarks
5.1 Synopsis
A large variety of different payment instruments is nowadays available that
allow to make transfers of money electronically or in paper form in purpose of
daily economic activities. However, within past decades, the way individuals
made payments has changed considerably by virtue of new payment products
and major revisions in payments processing. Meanwhile, innovative forms of
paying such as contactless and mobile payment have emerged making use of
technological advancements in data communication and changing consumer
needs. Yet, cash still accounts for a significant share of retail payments
that is overall more costly for society than to pay electronically. Thus, it is
essential to know whether there is room for a further digitization of retail
payments to improve overall costs efficiency and raising new revenue streams.
Understanding the transformation process of payment instruments is also
important for the smooth operation of the economy.
Against this background, the objective of this thesis was to explore the
impact of innovative payment instruments on individual payment behavior.
To this end, three studies in the field of payment economics were conducted,
each of which analyzed one of the latest payment innovations such as con-
tactless and mobile payment on either individual transaction motives (chap-
ter 2), cash usage (chapter 3), and payment instrument choice (chapter 4).
183
Prior to the these studies, a comprehensive overview of factors determining
individual payment behavior was provided followed by the establishment of
a theoretical model of individual payment behavior comprising aspects of
utility maximizing and habitual behavior based on the reviewed literature
(chapter 1). The theoretical model implicitly served as the reference point
and guidance for the subsequent empirical studies conducted in chapter 2, 3,
and 4.
Overall, the findings provide corroborating evidence that the latest pay-
ment innovations such as contactless and mobile payment influence individ-
ual payment behavior to the extent that individuals make more payment
card transactions and undertake fewer cash payments at the POS with re-
spect to value and volume due to the contactless feature embedded in debit
and credit cards. Also, mobile payment substitutes paper-based payment
methods (cash and checks) regarding adoption and suggests to complement
card payments and to substitute paper-based payments with respect to us-
age.
The first study, presented in chapter 2, estimated the effect of contactless
payment on the transaction ratio for different transaction types at the POS.
Credit and debit cards served as the object of investigation, to which the con-
tactless feature is embedded. The empirical analysis was undertaken using
data from a national representative survey on consumer payment behavior
in the U.S. in 2010. Since it is likely that unobserved heterogeneity is present
in the underlying econometric specification, propensity score matching was
employed to control for selection into contactless payment. The estimation
showed that contactless payment yields a statistically significant increase
in the transaction ratio at the POS. The average treatment effect on the
treated for credit and debit cards is estimated at roughly 8 and 10 percent,
respectively. In other words, the contactless feature induces individuals to
employ payment cards more frequently implying that, irrespective of the for-
mat of money, transaction costs efficiency is a relevant factor in payment
decisions. Higher debit and credit card usage results in higher fee turnovers
for payment card issuers, which can amount to approximately 7 USD per
184
person and year for increased debit card usage.1 Hence, the private industry
can highly benefit from this innovation regarding additional revenue streams
and efficiency gains.
The second study explored the impact of contactless payment on cash
usage in terms of value spending and transaction frequency at the POS (see
chapter 3). Debit and credit cards were investigated, to which the feature is
embedded. A novel microeconomic balanced panel data set that was drawn
from national representative surveys on consumer payment behavior in the
U.S. from 2009 to 2013 was exploited to econometrically analyze the effect
of contactless debit and credit cards on cash demand. Employing cross-
sectional estimation methods, the results showed that contactless payment
economically and statistically significantly reduces cash usage at the POS
in terms of value spending and transaction frequency. The negative effect
of contactless credit and debit cards on cash volume is 5 and 6 percent, re-
spectively. The negative impact of contactless credit cards on cash value is
estimated between 12 and 16 percent, but no effect is found for contactless
debit cards. However, after controlling for unobserved heterogeneity using
the fixed-effects model, contactless credit cards have no statistically signifi-
cant impact on cash volume, whereas the negative effect of contactless debit
cards of 3 percent is half the size. The results obtained on cash value are
unaffected. This implies that contactless payment is able to steer consumers
away from the use of cost-inefficient cash transactions.
In chapter 4, the third study was presented analyzing the effect of mo-
bile payment on the adoption and usage patterns of traditional payment
instruments such as cash, checks, credit, debit, and prepaid cards at the
POS. Discrete-choice random utility models were employed to simulate con-
sumer behavior using data from a representative survey on consumer pay-
ment choice in the U.S. in 2012. The estimation yielded the following im-
portant findings. First, mobile payment does not exert an effect on payment
cards at the adoption stage (extensive margin), but is likely to substitute
for paper-based payment methods such as cash and checks. Second, mo-
bile payment does not affect the selection of payment instruments at the
1Assuming an interchange fee of 0.12 USD per transaction.
185
usage stage (intensive margin). Yet, there is suggestive evidence that it is
complementary to card payments and a substitute for paper-based payment
instruments when choosing how to pay. The results hint at potential social
welfare gains of mobile payment induced by the reduction of cost-inefficient
paper-based payment methods. They also shed light on the challenging is-
sues for the private industry sector to enhance the usage of mobile payment
that in turn may increase profits.
The rest of this concluding chapter first points to implications derived
from the thesis’ main results. Second, the thesis limitations are discussed.
Third, directions and suggestions for future research are presented.
5.2 Implications
The thesis’ main findings offer valuable implications not only for policy mak-
ers, but also for other stakeholders in the payments ecosystem such as central
banks, financial intermediaries, merchants, and consumers. In sum, the re-
sults imply that the use of electronic payment instruments can further be
stimulated by means of contactless and mobile payment, meaning there is
still room for a further digitization of retail payments that improves the
overall social costs efficiency of retail payments.
The results in chapter 2 demonstrated that contactless payment has a
positive impact on the use of debit and credit cards, to which the feature is
embedded. Pertaining to the concept of utility maximizing behavior, which
served as the conceptual basic framework in this thesis to analyze the effect of
contactless and mobile payment (see section 1.3), the findings make clear that
the feature substantially improves the added value of payment instruments
and hence individual perceived utility, which is likely to be attributable to the
elevated incentive of greater transaction efficiency, as the representation of
money remains unchanged. They underline the relevance of transaction costs
efficiency in payment instrument choice. The respective positive increase in
perceived net utility is therefore as remarkably large that it makes individuals
to change their payment behavior. In other words, contactless payment shifts
a within the “no behavioral change” area randomly chosen starting point in
186
the presented theoretical model above the behavioral change frontier (see
Figure 1.4).
Correspondingly, the results give advice for the private industry to ac-
tively promote this innovative payment product, as greater usage of payment
cards leads to efficiency gains in payment processing and to increasing rev-
enue streams. For instance, more transactions translate onto more revenue
caused by the increased amount of fees processed by financial intermediaries.
Conversely, it implies that merchants face higher burdens given the actual
interchange fee policy. Thus, the results point to the importance of regular
monitoring of payment markets by policy makers to guarantee the smooth
operation of retail payments and an optimally balanced fee structure within
the framework of the four-party networks.2 From the consumer’s perspective,
the findings may additionally provide evidence that value spending increases
due to the contactless feature, assuming the average transaction amount of
card payments remains constant.
The results in chapter 3 documented that contactless payment decreases
the number and value of cash payments at the POS. This has influential
implications for the social welfare costs of payments systems, which tend to
decrease since contactless payment steers consumers away from using socially
costs-inefficient cash. The results also imply that central banks benefit from
lower welfare costs of inflation due to decreasing money demand. Therefore,
central banks should advance and promote the use of payment innovations
such as contactless payment to achieve gains in payment processing and to
minimize social welfare costs.
The results further hint at the entrenched usage of cash, which may be
highly routine since the amount of cash holdings in wallet is not affected by
contactless payment. Indeed, according to Keynes (1936), there are three mo-
tives for holding money that relate to the transaction motive, precautionary
motive, and speculative motive. This thesis focused on money that is used
to carry out planned expenditures. It is hence assigned to the transaction
2The four-party networks include consumers and their banks (issuers) as well as mer-chants and their banks (acquirers). Issuers and acquirers belong to a specific networkthat sets the rules for clearing and settling payment card transactions among its members(Bolt and Chakravorti, 2012).
187
motive. However, as the precautionary motive indicates, it is likely that indi-
viduals also hold money balances as a reserve to cover unexpected expenses.
Another rationale for holding cash in reserves is when costs are incurred for
converting invested money (deposits) into liquid money (cash) for making
payments (Keynes, 1936). Money balances held for speculative motives refer
to cash holdings that are used for profitable investment opportunities.
This implies that a further shift towards using less cash with respect to
value spending can occur when payment cards, including contactless pay-
ment, are ubiquitously accepted as a payment medium. In doing so, the
precautionary motive of cash becomes obsolete since payment cards provide
instantaneous and almost costless access to deposited money that offer liquid-
ity for unforeseen transactions. Therefore, payment card terminals should
be actively promoted by the private industry sector. Also, policy makers as
well as members in the four-party networks are requested to modify the in-
terchange fee structures to the extent that it is attractive for every merchant,
irrespective of its size and revenue, to accept electronic payment instruments.
The results further indicate that the role of cash as a medium of exchange is
still vast, meaning that central banks are asked to monitor the demand for
cash regularly to ensure the optimal provision of money.
The findings in chapter 4 suggest that mobile payment does not substitute
physical payment cards, but paper-based payment instruments such as cash
and checks. It has no effect on the usage of traditional payment instruments
except on prepaid cards. However, there is supportive evidence that it is
complementary to card payments and substitutional to paper-based payment
instruments. The results portray important implications for the relevant
stakeholders in the payment ecosystem because mobile payment strives for
further stimulation of socially costs-efficient payment selection. This leads
to decreasing social welfare costs in the overall payment system.
In particular, the findings point to different drivers and mechanisms un-
derneath at the extensive and intensive margin of payment instruments. This
is, it is crucial not only to foster the proliferation of mobile payment, as it
is a prerequisite for use, but more importantly to make more efforts to in-
centivize its usage, for instance, by means of marketing and pricing. This
188
comes at the advantage of processing an increasing number of electronic
card payments at the expense of paper-based payments that in turn yields
raising revenue streams for the participants in the retail payments chain.
The results also imply that consumer payment choice strongly depends on
perceived characteristics of payment methods, leading to the conclusion that
promoters specifically need to highlight mobile payment’s attributes that are
superior to traditional payment instruments in order to enhance its usage.
5.3 Limitations
The methodological approach used in this thesis to answer the main research
question whether payment innovation impacts individual payment behavior
relies on representative surveys asking consumers about their actual payment
behavior. This approach is most common primarily because of the nature
of cash payments, which are characterized by anonymity and are therefore
not registered in contrast to their electronic counterparts. However, as cash
payments account for the vast majority of retail payments, it is absolutely es-
sential to have accurate data on the use of all payment instruments available
to assess the effects of payment innovations properly.
The data sets used in this thesis are representative self-reported surveys
in the U.S. that are drawn from the Federal Reserve Bank of Boston. They
provide information on the adoption and use of nine common payment in-
struments including cash, but they shed no light on transaction values and
total expenditures. The sample unit is an individual consumer in the U.S.
older than 18 years. The surveys were conducted based on retrospective re-
call questionnaires. However, this survey method is sensitive to errors since
it may suffer from poor recall, particularly low value payments that have low
budget impact and are high in numbers and non-salient are vulnerable to
underestimation. This is exactly the case with cash payments. Consequently,
the data sets used in this thesis may not be sufficiently extensive and may
be subject to measurement errors, although flexible reporting strategies and
subsequent cleaning procedures were employed to optimize data accuracy.
189
In this sense, the best method to measure consumer payment choice
including error-sensitive cash payments is to record daily purchases in a
payment diary. According to Jonker and Kosse (2013), one-day diaries are
preferable to multiple-day diaries since the latter suffer from underreporting
of low value cash payments too, which is caused by diary fatigue and diary
despair. This is, individuals experience a gradual or immediate loss of com-
mitment and accuracy when participating several days (Jonker and Kosse,
2013). Diaries allow the recording of payments at such a detailed level that
they provide accurate insights into their frequency, value, time, location,
and instrument used. Having comprehensive payment information available
is key to fully understanding consumers’ payment patterns and motivation
for payment choice. In that case, it may facilitate the decision-making pro-
cess in purpose of stimulating the use of more socially costs-efficient payment
instruments.
However, according to van der Horst and Matthijsen (2013), self-reported
surveys are particularly useful in ascertaining individual characteristics re-
lated to specific patterns of payment behavior, but they give less insights on
individuals’ motives behind the choice of payment instruments, which is ex-
pected to be largely routine. For this reason, experimental research methods
such as the virtual-reality study design are preferable. It allows to directly
observe virtual payment behavior based on real-life simulated payment deci-
sions that can be manipulated by various stimuli, for instance, by modifying
scenarios and variables or giving specific instructions to participants. Com-
pared to surveys, the direct observation of payment behavior has the added
value of an inexistent time lag between actual behavior and the questions
asked as well as the absence of social desirability in surveying (cf. van der
Horst and Matthijsen, 2013).
In light of this, the thesis may have several limitations. First and fore-
most, the data sets used may be subject to incomplete recall resulting in mea-
surement errors that lead to inappropriate statistical inferences. Second, the
absence of information on exact transaction values of payment instruments
hinders to make final conclusions about the effect of contactless payment on
value spending and cash usage with respect to expenditures. In that case,
190
these measures either have to be proxied or approximated. Third, the sur-
veys report the adoption of contactless payment only, but not the frequency
of usage resulting in a possible over- or underestimation of its corresponding
effect. This is because adoption and use may be two sequentially different
and independent processes. Payment diaries would therefore obtain more
accurate results.
Similarly, accurate data on mobile payment may be insufficient in this
thesis. There are many different existing concepts referring to mobile pay-
ment, suggesting the surveys not to be exhaustive in the variety of respective
terminologies. In turn, this may have resulted in false reporting of mobile
payments since respondents may have been conceptually uncertain about
their exact declaration. Also, the surveys do not provide sufficient infor-
mation whether mobile payments have been made proximately, remotely or
between persons, which overall may confound the results. Data are also lack-
ing about the frequency of mobile payments. It is unclear how many times
prior to the survey mobile payment has exactly been employed. However,
having accurate usage data is key to estimate the impact on traditional pay-
ment instruments more thoroughly. Payment diaries along with real-world
data about actual mobile payment usage, partitioned into the various mobile
payment concepts available, would improve the accuracy of the estimation
results.
There are caveats in interpreting the exact motivation underneath using
innovative payment instruments when analyzing the present self-reported
payment surveys. The thesis suggests that lower transaction costs of pay-
ment innovations are one of the main drivers for their elevated perceived
utility, which in turn may affect payment behavior. Yet, it remains incon-
clusive through which channels payment innovations effectively impact indi-
vidual payment behavior. An experimental research approach would help to
enhance the understanding of the underlying mechanisms.
The thesis’ results may also suffer from feedback effects due to the nature
of network externalities inherent in the two-sidedness of the payment mar-
kets. This is, the perceived utility of payment innovations not only depends
on their perceived characteristics, but also on the number of terminals ac-
191
cepting these new forms of paying. If their points of acceptance are scarce,
individuals are limited to use payment innovations, meaning that the impact
of payment innovation on conventional payment instrument use may be af-
fected more extensively. However, data restrictions in this thesis impair an
appropriate analysis of the supply-side factors.
Using U.S. survey data raise issues about the external validity of the re-
sults, i.e. the relevance for the findings for extra-territorial payment areas.
It is widely acknowledged that the economics of payment are different from
country to country, as each country has unique retail payment markets and
features its own payment market characteristics with respect to costs of pay-
ment instruments, market development, and payment behavior (Schmiedel
et al., 2013). These cultural, institutional, and infrastructural differences
in payment compositions across countries might pose a serious threat to
the generalizability of the thesis’ findings. For instance, Kosse and Jansen
(2013) hint at cultural factors playing a potential role for payment choice.
They exhibit that foreign background individuals from cash-oriented coun-
tries still use cash extensively after migration. However, there are country
clusters with similar payment landscapes observable, whereby actual national
payment patterns chiefly hinge on the countries’ past payment market struc-
tures (Humphrey et al., 1996).3 It is thus deemed important to highlight that
while the U.S. payment composition may be close to Scandinavian countries,
it might remain challenging to compare U.S. payment markets with the rest
of European payment areas, which is likely to persist sooner or later.
5.4 Directions for Future Research
The thesis documented that contactless payment fosters the use of payment
cards while it reduces the usage of cash. It further showed that mobile
payment is substitutional to paper-based payment methods at the extensive
margin and likely to be complementary to card payments and substitutional
to paper-based payments at the intensive margin. Nevertheless, there are a
few issues that remain to be addressed regarding forthcoming studies in the
3See Schmiedel et al. (2013) for an overview of European retail payments clusters.
192
field of payment economics to enhance the quality of the results. In essence, it
would be useful for future payments research in general and in particular with
respect to exploring the impact of innovative payment instruments to make
use of qualitatively superior payment data based on payment diaries in order
to improve the estimates. Thereby, innovative data collection methods rather
than paper-based diary surveys and questionnaires may be more fruitful to
advance data quality and reliability by minimizing recall errors.
In this sense, new electronic devices such as smart phones and tablets
may serve as potential survey tools that could facilitate the records of daily
payments. For instance, special targeted tracking applications installed on
smart phones may help to instantly update one’s personal virtual payment
diary at the moment a payment is settled, whereby electronic payments that
are made remotely or alternatively stationary by mobile phones could hereby
automatically be registered by the respective application. These innovative
survey techniques could further provide easy access to an unprecedented
amount of individuals’ payments data that could open up new and enrich-
ing possibilities of data analysis in the field of “Big Data” that allows to
investigate individual payment behavior more thoroughly. There is a clear
trend towards more electronic payments observable due to growing famil-
iarity with electronic payment instruments on top of increasing e-commerce
activities. Also, new virtual money tokens based on cryptographic technolo-
gies (e.g. “Blockchain”) such as the cryptocurrency “Bitcoin” are on the rise.
Therefore, it would be interesting to deepen the understanding in payment
behavior particularly with regards to online payments, focusing on these new
forms of paying.
Collecting detailed payments data, however, does not provide a method-
ologically viable approach to effectively detect individuals’ psychological bar-
riers and motivations behind the usage of (innovative) payment methods.
Thus, future empirical endeavors in payment economics should depart from
the promising role of experiments to gain a clear picture of personal pay-
ment behavior, particularly with respect to payment innovation. Though
it is acknowledged that payment behavior is highly habitual, there are still
too many unanswered questions on how these habits are formed, developed,
193
and how they can be changed. Research questions in the context of payment
habitualization further should comprise the role of cultural, personal, and
environmental factors. Providing enlightening insights in these behavioral
patterns may help to encourage the shift to use more costs-efficient payment
means.
Another important area that deserves further investigation is the effect of
supply-side determinants in payment economics, for instance, by conducting
“multi-sided adoption research” in the payment ecosystem (Dahlberg et al.,
2015).4 This includes the analysis of simultaneous adoption and usage of
payment instruments by several groups within the payment ecosystem. It is
highly welcomed to explore the views of merchants and retailers on payment
instrument choice, as they play the other key role along with consumers in
payment markets. This is a highly neglected topic in payments research
so far. In a similar vein, future studies could center on the role of payment
infrastructure and payment institutions, for instance, by using merged aggre-
gate supply-side-level with consumer-level payments data. This would help
to enhance the understanding of payment decisions in a comprehensively
rigor way, as it may allow to control for network externalities.
Since every single country features its own payment market characteris-
tics, it leaves unanswered to what extent country-specific payment market
analyses are valid for other payment areas, which causes a serious threat
to external validity of payments research. For this reason, there is an obvi-
ous demand for future multinational comparative payment studies based on
internationally harmonized payments data that enable useful cross-country
comparisons to be made. This allows to gauge effects of payment innova-
tion and to assess determinants of payment behavior that are specific to a
particular country or are more general. Therefore, future research based on
internationally harmonized data is desired.
Lastly, a grounded theoretical foundation of payment behavior is lacking.
This thesis attempted to establish a basic theoretical framework encompass-
ing the relationship between payment habitualization and utility maximiza-
4See Dahlberg et al. (2015) for vivid recommendations in future mobile payment re-search.
194
tion that served as the starting point for the three studies conducted in this
thesis. However, more theoretical work needs to be done in the area of pay-
ment behavior to fully comprehend individuals’ payment patterns and habits.
Thereby, the theoretical fundament may come from the field of psychology
of payments, behavioral and mental action theories, economic theories com-
prising theories of money, the history of payment instrument use in different
countries, as well as the pricing of payment instruments, amongst others
(Dahlberg et al., 2015). Following these directions is likely to contribute to
an integrated theoretical framework of individual payment behavior.
195
196
References
Agarwal, S., Chakravorti, S. and Lunn, A. (2010) Why do banks rewardtheir customers to use their credit cards?, Working Paper 2010-19,Federal Reserve Bank of Chicago.
Alvarez, F. and Lippi, F. (2009) Financial innovation and the transactionsdemand for cash, Econometrica, 77(2), 363–402.
Alvarez, F. and Lippi, F. (2015) Cash burns: An inventory model with acash-credit choice, Working Paper 21110, National Bureau of EconomicResearch.
Amromin, G. and Chakravorti, S. (2009) Whither Loose Change? TheDiminishing Demand for Small-Denomination Currency, Journal ofMoney, Credit and Banking, 41(2-3), 315–335.
Arango, C., Bouhdaoui, Y., Bounie, D., Eschelbach, M. and Hernandez, L.(2013) Cash Management and Payment Choices: A Simulation Modelwith International Comparisons, Working Paper 2013-53, Bank ofCanada.
Arango, C., Hogg, D. and Lee, A. (2015a) Why Is Cash (Still) SoEntrenched? Insights from Canadian Shopping Diaries, ContemporaryEconomic Policy, 33(1), 141–158.
Arango, C., Huynh, K. P. and Sabetti, L. (2011) How Do You Pay? TheRole of Incentives at the Point-of-Sale, Working Paper 2011-23, Bank ofCanada.
Arango, C., Huynh, K. P. and Sabetti, L. (2015b) Consumer paymentchoice: Merchant card acceptance versus pricing incentives, Journal ofBanking and Finance, 55, 130–141.
197
Arango, C. and Taylor, V. (2009) The Role of Convenience and Risk inConsumers’ Means of Payment, Discussion Paper 2009-8, Bank ofCanada.
Ariely, D. and Silva, J. (2002) Payment Method Design: Psychological andEconomic Aspects of Payments, Center for eBusiness@MIT, Paper 196 .
Attanasio, O. P., Guiso, L. and Jappelli, T. (2002) The demand for money,financial innovation, and the welfare cost of inflation: An analysis withhousehold data, Journal of Political Economy, 110(21), 317–351.
Au, Y. A. and Kauffman, R. J. (2008) The economics of mobile payment:Understanding stakeholder issues for an emerging financial technologyapplication, Electronic Commerce Research and Applications, 7, 141–164.
Aydogan, S. and van Hove, L. (2015) Nudging consumers towards cardpayments: A field experiment, in: The Usage, Costs and Benefits of Cash– Revisited: Proceedings of the 2014 International Cash Conference, pp.589–630. Frankfurt: Deutsche Bundesbank.
Bagnall, J., Bounie, D., Huynh, K. P., Kosse, A., Schmidt, T., Schuh, S.and Stix, H. (2016) Consumer Cash Usage: A Cross-CountryComparison with Payment Diary Survey Data, International Journal ofCentral Banking, 12(4), 1–61.
Baumol, W. J. (1952) The transactions demand for cash: An inventorytheoretic approach, The Quarterly Journal of Economics, 66(4),545–556.
Blume, M., Graf, S. and Scherrer, C. (2015) Discount vs. reward: Whathas greater effect on altering payment behaviour?, ZHAW ZurichUniversity of Applied Sciences, Institute of Marketing Management.
Board of Governors of the Federal Reserve System (2011) 2009 InterchangeRevenue, Covered Issuer Cost, and Covered Issuer and Merchant FraudLoss Related to Debit Card Transactions, [Online].www.federalreserve.gov/paymentsystems (26.05.2014).
Bolt, W. and Chakravorti, S. (2012) Digitization of retail payments, in:Peitz, M. and Waldfogel, J. (eds), The Oxford Handbook of the DigitalEconomy, Oxford University Press.
Bolt, W., Humphrey, D. B. and Uittenbogaard, R. (2008) TransactionPricing and the Adoption of Electronic Payments: A Cross-CountryComparison, International Journal of Central Banking, pp. 89–123.
198
Bolt, W., Jonker, N. and van Renselaar, C. (2010) Incentives at thecounter: An empirical analysis of surcharching card payments andpayment behaviour in the Netherlands, Journal of Banking and Finance,34, 1738–1744.
Borzekowski, R. and Kiser, E. K. (2008) The choice at the checkout:Quantifying demand across payment instruments, International Journalof Industrial Organization, 26, 889–902.
Borzekowski, R., Kiser, E. K. and Ahmed, S. (2008) Consumers’ Use ofDebit Cards: Patterns, Preferences, and Price Response, Journal ofMoney, Credit and Banking, 40(1), 149–172.
Bouhdaoui, Y. and Bounie, D. (2012) Modeling the share of cash paymentsin the economy, International Journal of Central Banking, 8(4), 175–195.
Bounie, D. and Francois, A. (2006) Cash, Check or Bank Card? The effectsof transaction characteristics on the use of payment instruments,Presented at the FMG and CASS Business School Conference, Workshopon Financial Regulation and Payment Systems, London.
Briglevics, T. and Schuh, S. (2013) U.S. Consumer Demand for Cash in theEra of Low Interest Rates and Electronic Payments, Working PaperNo. 13-23, The Federal Reserve Bank of Boston.
Briglevics, T. and Shy, O. (2014) Why don’t most merchants use pricediscounts to steer consumer payment choice?, Review of IndustrialOrganization, 44(4), 367–392.
Brouzos, J. (2014) Kreditkartengebuhren: Weko schaltet sich ein,Handelszeitung, Ausgabe vom 12.03.2014, [Online].www.handelszeitung.ch.
Brown, A., Dodini, S., Gonzalez, A., Merry, E. and Thomas, L. (2015)Consumers and Mobile Financial Services 2015, Board of Govenors of theFederal Reserve System, [Online].www.federalreserve.gov/publications/default.htm (08.09.2015).
Caliendo, M. and Kopeinig, S. (2005) Some practical guidance for theimplementation of propensity score matching, IZA Discussion PaperNo. 1588.
Camera, G., Casari, M. and Bortolotti, S. (2016) An experiment on retailpayments systems, Journal of Money, Credit and Banking, 48, 363–392.
199
Carbo-Valverde, S. and Linares-Zegarra, J. M. (2011) How effective arerewards programs in promoting payment card usage? Empirical evidence,Journal of Banking and Finance, 35(12), 3275–3291.
Chakravorti, S. (2010) Externalities in Payment Card Networks: Theoryand Evidence, Review of Network Economics, 9(2).
Chatterjee, P. and Rose, R. L. (2012) Do Payment Mechanisms Change theWay Consumers Perceive Products?, Journal of Consumer Research, 38.
Chen, H., Felt, M.-H. and Huynh, K. P. (2017) Retail Payment Innovationsand Cash Usage: Accounting for Attrition Using Refreshment Samples,Journal of the Royal Statistical Society, Series A, Statistics in Society,180(2), 503–530.
Chen, L. (2008) A model of consumer acceptance of mobile payment,International Journal of Mobile Communications, 6(1), 32–52.
Ching, A. T. and Hayashi, F. (2010) Payment card rewards programs andconsumer payment choice, Journal of Banking and Finance, 34(8),1773–1787.
Choudhary, V. and Tyagi, R. K. (2009) Economic incentives to adoptelectronic payment schemes under competition, Decision SupportSystems, 46, 552–561.
Coase, R. H. (1937) The Nature of the Firm, Economica, 4, 386–405.
Cohen, M. and Rysman, M. (2013) Payment Choice with Consumer PanelData, Working Papers No. 13-06, Federal Reserve Bank of Boston.
Columba, F. (2009) Narrow money and transaction technology: Newdisaggregated evidence, Journal of Economics and Business, 61,312–325.
Connolly, J. and Stavins, J. (2015) Payment Instrument Adoption and Usein the United States 2009–2013 by Consumers’ DemographicCharacteristics, Research Data Reports No. 15-6, Federal Reserve Bankof Boston.
Dahlberg, T., Guo, J. and Ondrus, J. (2015) A critical review of mobilepayment research, Electronic Commerce Research and Applications, 14,265–284.
200
Dahlberg, T., Mallat, N., Ondrus, J. and Zmijewska, A. (2008) Past,present and future of mobile payments research: A literature review,Electronic Commerce Research and Applications, 7, 165–181.
Davis, F. D. (1989) Perceived usefulness, perceived ease of use, and useracceptance of information technology, MIS Quarterly, 13(3), 319–340.
Davis, F. D., Bagozzi, R. P. and Warshaw, P. R. (1989) User acceptance ofcomputer technology: A comparison of two theoretical models,Management Science, 35(8), 982–1003.
Deutsche Bundesbank (2015) Zahlungsverhalten in Deutschland 2014 –Dritte Studie uber die Verwendung von Bargeld und unbarenZahlungsinstrumenten, [Online]. www.bundesbank.de (22.08.15).
Drehmann, M., Goodhart, C. and Krueger, M. (2004) The Challengesfacing currency usage: Will the traditional transaction medium be ableto resist competition from new technologies?, Economic Policy, 56,195–227.
Duca, J. V. and van Hoose, D. D. (2004) Recent developments inunderstanding the demand for money, Journal of Economics andBusiness, 56(4), 247–272.
EC, European Commission (2013) Study on the effects of informationdisclosure on consumer choice of payment instruments, [Online].http://ec.europa.eu/competition/sectors/financial services (11.08.2015).
EP, European Parliament (2014) MEPs back cap on card payment fees,Committees Committee on Economic and Monetary Affairs, PressRelease 20.02.14, [Online]. www.europarl.europa.eu (20.03.14).
Eschelbach, M. and Schmidt, T. (2013) Precautionary Motives inShort-Term Cash Management – Evidence from German POSTransactions, Bundesbank Discussion Paper No. 38/2013, DeutscheBundesbank.
Evans, D. and Schmalensee, R. (2005) The Economics of Interchange Feesand Their Regulation: An Overview, Proceedings – Payments SystemResearch Conferences, Federal Reserve Bank of Kansas City, Issue May,73–120.
Feinberg, R. A. (1986) Credit Cards as Spending Facilitating Stimuli: AConditioning Interpretation, Journal of Consumer Research, 13.
201
Foster, K., Schuh, S. and Zhang, H. (2013) The 2010 Survey of ConsumerPayment Choice, Research Data Reports No. 13-2, The Federal ReserveBank of Boston.
Fujiki, H. and Tanaka, M. (2014) Currency demand, new technology, andthe adoption of electronic money: Micro evidence from Japan,Economics Letters, 125(1), 5–8.
Fung, B. S., Huynh, K. P. and Sabetti, L. (2014) The impact of retailpayment innovations on cash usage, Journal of Financial MarketInfrastructures, 3(1), 1–29.
Garcia-Swartz, D. D., Hahn, R. H. and Layne-Farrar, A. (2002) The movetowards a cashless society: Calculating the costs and the benefits, Reviewof Network Economics, 5(2), 199–228.
Gneezy, A., Gneezy, U., Nelson, L. D. and Brown, A. (2010) Shared socialresponsibility: A field experiment in pay-what-you-want pricing andcharitable giving, Science, 329(5989), 325–327.
Gourville, J. T. (1998) Pennies-a-Day: The Effect of Temporal Reframingon Transaction Evaluation, Journal of Consumer Research, 24(4),395–403.
Hausman, J. and McFadden, D. (1984) Specification tests for themultinomial logit model, Econometrica, 52(5), 1219–1240.
Hayashi, F. and Klee, E. (2003) Technology adoption and consumerpayments: Evidence from survey data, Review of Network Economics,2(2), 175–190.
He, P., Huang, L. and Wright, R. (2008) Money, banking, and monetarypolicy, Journal of Monetary Economics, 55(6), 1013–1024.
Heath, C. and Soll, J. B. (1996) Mental Budgeting and ConsumerDecisions, Journal of Consumer Research, 23(1), 40–52.
Hernandez, L., Jonker, N. and Kosse, A. (2016) Cash versus debit card:The role of budget control, The Journal of Consumer Affairs.
Hirschman, E. C. (1979) Differences in consumer purchase behavior bycredit card payment system, Journal of Consumer Research, 6, 58–66.
Hoppli, T., Jaeger, F. and Koller, J. (2011) Studie: SchweizerKreditkartenmarkt, Executive School of Management, Technology andLaw (ES-HSG), University of St. Gallen, [Online]. www.es.unisg.ch(15.09.13).
202
Humphrey, D. B. (2004) Replacement of cash by cards in U.S. consumerpayments, Journal of Economics and Business, 56, 211–225.
Humphrey, D. B. (2010) Retail payments: New contributions, empiricalresults, and unanswered questions, Journal of Banking and Finance,34(8), 1729–1737.
Humphrey, D. B., Kim, M. and Vale, B. (2001) Realizing the gains fromelectronic payments: Cost, pricing and payment choice, Journal ofMoney, Credit and Banking, 33(2), 216–234.
Humphrey, D. B., Pulley, L. B. and Vesala, J. (1996) Cash, Paper andElectronic Payments: A Cross-Country Analysis, Journal of Money,28(4), 914–939.
Huynh, K. P., Schmidt-Dengler, P. and Stix, H. (2014) The Role of CardAcceptance in the Transaction Demand for Money, Working PaperNo. 44, Bank of Canada.
Incekara-Hafalir, E. and Loewenstein, G. (2009) The Impact of CreditCards on Spending: A Field Experiment, Social Science ResearchNetwork (SSRN).
Johnson, J. J. (2014) Debit Card Interchange Fees and Routing: InterimFinal Rule, [Online]. www.federalreserve.gov (20.03.14).
Jonker, N. (2007) Payment instruments as perveived by consumers: Resultsfrom a household survey, De Economist, 155, 21–38.
Jonker, N. (2013) Social costs of POS payments in the Netherlands2002–2012: Efficiency gains from increased debit card usage, DNBOccasional Studies, 11(2).
Jonker, N. and Kosse, A. (2013) Estimating Cash Usage: The Impact ofSurvey Design on Research Outcomes, De Economist, 161, 19–44.
Judt, E. (2006) Zahlungsverkehrsinnovationen im Wandel der Zeit und ihreVermarktung, in: Lammer, T. (ed), Handbuch E-Money, E-Payment andM-Payment, Physica-Verlag Heidelberg.
Jung, M. H., Nelson, L. D., Gneezy, A. and Gneezy, U. (2014) Paying morewhen paying for others, Journal of Personality and Social Psychology,107(3), 414–431.
203
Kahn, C. M. and Linares-Zegarra, J. M. (2015) Identity Theft andConsumer Payment Choice: Does Security Really Matter?, Journal ofFinancial Services Research, pp. 1–39.
Kahn, C. M. and Roberds, W. (2009) Why pay? An introduction topayments economics, Journal of Financial Intermediation, 18, 1–23.
Keynes, J. M. (1936) The General Theory of Employment, Interest andMoney, New York.
Khan, J. (2011) Cash or Card: Consumer Perceptions of Payment Modes,Dissertation, Auckland University of Technology.
Kim, B.-M., Yilmazer, T. and Widdows, R. (2006) Adoption of internetbanking and consumers’ payment choices, Working Paper, PurdueUniversity.
Kim, C., Mirusmonov, M. and Lee, I. (2010) An empirial examination offactors influencing the intention to use mobile payment, Computers inHuman Behavior, 26, 310–322.
Klee, E. (2006) Paper or Plastic? The effect of time on check and debitcard use at grocery stores, Working Paper, Board of Governors of theFederal Reserve System.
Klee, E. (2008) How people pay: Evidence from grocery store data, Journalof Monetary Economics, 55, 526–541.
Kosse, A. (2013a) Do newspaper articles on card fraud affect debit cardusage?, Journal of Banking and Finance, 37(12), 5382–5391.
Kosse, A. (2013b) The safety of cash and debit cards: A study on theperception and behavior of Dutch consumers, International Journal ofCentral Banking, 9(4), 77–98.
Kosse, A. and Jansen, D.-J. (2013) Choosing how to pay: The influence offoreign backgrounds, Journal of Banking and Finance, 37, 989–998.
Koulayev, S., Rysman, M., Schuh, S. and Stavins, J. (2012) ExplainingAdoption and Use of Payment Instruments by U.S. Consumers?,Working Paper 2012-14, Federal Reserve Bank of Boston.
Lechner, M. (2002) Some practical issues in the evaluation of heterogenouslabour market programmes by matching methods, Journal of RoyalStatistical Society, 165, 59–82.
204
Lee, J. and Kwon, K.-N. (2002) Consumers’ Use of Credit Cards: StoreCredit Card Usage as an Alternative Payment and Financing Medium,The Journal of Consumer Affairs, 36(2), 239–262.
Leenheer, J., Elsen, M. and Pieters, R. (2012) Consumentenprikkels voorEfficient Betalen – Management Summary, CentERdata, Tilburg.
Leong, L.-Y., Hewb, T.-S., Tan, G. W.-H. and Ooi, K.-B. (2013) Predictingthe determinants of the NFC-enabled mobile credit card acceptance: Aneural networks approach, Expert Systems with Applications, 40,5604–5620.
Leuven, E. and Sianesi, B. (2003) PSMATCH2: Stata module to performfull Mahalanobis and propensity score matching, common supportgraphing, and covariate imbalance testing, Boston College Department ofEconomics, Statistical Software Components. Version 4.0.10 10feb2014.[Online.] http://ideas.repec.org/c/boc/bocode/s432001.html(11.06.2014).
Liebana-Cabanillas, F., Sanchez-Fernandez, J. and Munoz-Leiva, F. (2014)Antecedents of the adoption of the new mobile payment systems: Themoderating effect of age, Computers in Human Behavior, 35, 464–478.
Lippi, F. and Secchi, A. (2009) Technological change and the households’demand for currency, Journal of Monetary Economics, 56, 222–230.
Liu, J., Kauffman, R. J. and Ma, D. (2015) Competition, cooperation, andregulation: Understanding the evolution of the mobile paymentstechnology ecosystem, Electronic, 14, 372–391.
Mallat, N. (2007) Exploring consumer adoption of mobile payments – Aqualitative study, Journal of Strategic Information Systems, 16, 413–432.
Mallat, N., Rossi, M. and Tuunainen, V. K. (2004) Mobile banking services,Communications of the ACM, 47(5), 42–46.
Martin, S. (2012) Statement before the Committee on Financial ServicesSubcommittee on Financial Institutions and Consumer Credit U.S.House of Representatives, Board of Governors of the Federal ReserveSystem, Washington, D.C.
Massoud, N., Saunders, A. and Scholnick, B. (2011) The cost of being late?The case of credit card penalty fees, Journal of Financial Stability, 7,49–59.
205
Mastercard (2013) Mastercard paypass performance insights, [Online].http://mastercard-mobilepartner.com (04.07.2013).
McCallum, B. T. and Goodfriend, M. S. (1987) Demand for Money:Theoretical Studies, in: Eatwell, J., Milgate, M., and Newman, P. (eds),The New Palgrave: A Dictionary of Economics, London: Macmillan.
McFadden, D. (1973) Conditional Logit Analysis of Qualitative ChoiceAnalysis, in: Zarembka, P. (ed), Frontiers in Econometrics, AcademicPress, London/New York.
Mishra, H., Mishra, A. and Nayakankuppam, D. (2006) Money: A bias forthe whole, Journal of Consumer Research, 32, 541–549.
Nichols, A. (2007) Causal inference with observational data, The StataJournal, 7(4), 507–541.
Ondrus, J. and Pigneur, Y. (2005) A disruption analysis in the mobilepayment market, in: Sprague, R. (ed), Proceedings of the 38th HawaiiInternational Conference on Systems Science, Big Island, HI, USA,January 3–6, IEEE Computing Society Press, Los Alamitos, CA.
Ondrus, J. and Pigneur, Y. (2006a) A systematic approach to explain thedelayed deployment of mobile payments in Switzerland, in: Proceedingsof the Fifth International Conference on Mobile Business (ICMB),Copenhagen, Denmark, June 26–27.
Ondrus, J. and Pigneur, Y. (2006b) Towards a holistic analysis of mobilepayments: A multiple perspective approach, Electronic CommerceResearch and Applications, 5(3), 246–257.
Pianalto, S. (2012) Collaborating to Improve the U.S. Payments System,[Online]. www.clevelandfed.org (22.10.12).
Polasik, M., Gorka, J., Wilczewski, G., Kunkowski, J., Przenajkowska, K.and Tetkowska, N. (2013) Time efficiency of point-of-sale paymentmethods: Empirical results for cash, cards and mobile payments, in:Cordeiro, J., Maciaszek, L. A., Filipe, J. (eds), Enterprise InformationSystems, Lecture Notes in Business Information Processing, Vol. 141.Springer, Berlin Heidelberg, pp. 306–320.
Prelec, D. (2009) Consumer behavior and the future of consumer payments,in: Litan, R. E. and Baily, M. N. (eds) Moving Money: The Future ofConsumer Payments, pp. 77–101. Washington, D.C.: BrookingsInstitution Press.
206
Prelec, D. and Loewenstein, G. (1998) The Red and the Black: MentalAccounting of Savings and Debt, Marketing Science, 17(1), 4–28.
Prelec, D. and Semester, D. (2001) Always leave home without it. Afurther investigation of the credit card effect on willingness to pay,Marketing Letters, 12(1), 5–12.
Raghubir, P. and Srivastava, J. (2008) Monopoly money: The effect ofpayment coupling and form on spending behavior, Journal ofExperimental Psychology, 14(3), 213–225.
Raghubir, P. and Srivastava, J. (2009) The denomination effect, Journal ofConsumer Research, 36, 701–713.
Rochet, J. (2003) The theory of interchange fees: A synthesis of recentcontributions, Review of Network Economics, 2(2), 97–124.
Rochet, J. and Tirole, J. (2002) Cooperation among competitors: Someeconomics of payment card associations, RAND Journal of Economics,33(4), 549–570.
Rochet, J.-C. and Wright, J. (2010) Credit card interchange fees, Journalof Banking and Finance, 34(8), 1788–1797.
Rogers, E. M. (2003) Diffusion of Innovation, The Free Press, New York(5th ed.).
Rogoff, K. S. (2016) The Curse of Cash, Princeton: Princeton UniversityPress.
Rosenbaum, P. R. (2002) Observational Studies, New York: Springer (2nded.).
Rosenbaum, P. R. and Rubin, D. B. (1983) The Central Role of thePropensity Score in Observational Studies for Causal Effects, Biometrika,70(1), 41–55.
Runnemark, E., Hedman, J. and Xiao, X. (2015) Do consumers pay moreusing debit cards than cash?, Electronic Commerce Research andApplications, 14(5), 285–291.
Rysman, M. (2007) An empirical analysis of payment card usage, Journalof Industrial Economics, 55, 1–36.
SCF, Survey of Consumer Finances (2014) 2013 Survey of ConsumerFinances, [Online]. www.federalreserve.gov/econresdata/scf/scfindex.htm(10.10.15).
207
Schmiedel, H., Kostova, G. L. and Ruttenberg, W. (2013) The Social andPrivate Costs of Retail Payment Instruments: A European Perspective,Journal of Financial Market Infrastructures, 2(1).
Scholnick, B., Massoud, N., Saunders, A., Carbo-Valverde, S. andRodriguez-Fernandez, F. (2008) The economics of credit cards, debitcards and ATMs: A survey and some new evidence, Journal of Bankingand Finance, 32, 1468–1483.
Schuh, S., Shy, O. and Stavins, J. (2010) Who Gains and Who Loses fromCredit Card Payments? Theory and Calibrations, Public PolicyDiscussion Papers No. 10-03, The Federal Reserve Bank of Boston.
Schuh, S. and Stavins, J. (2010) Why are (some) consumers (finally)writing fewer checks? The role of payment characteristics, Journal ofBanking and Finance, 34(8), 1745–1758.
Schuh, S. and Stavins, J. (2013) How Consumers Pay: Adoption and Use ofPayments, Accounting and Finance Research, 2(2), 1–21.
Schuh, S. and Stavins, J. (2014) The 2011 and 2012 Survey of ConsumerPayment Choice, Research Data Report No. 14-1, The Federal ReserveBank of Boston.
Schuh, S. and Stavins, J. (2015a) How do speed and security influenceconsumers’ payment behavior?, Current Policy Perspectives No. 15-1,The Federal Reserve Bank of Boston.
Schuh, S. and Stavins, J. (2015b) The 2013 Survey of Consumer PaymentChoice: Summary Results, Research Data Report No. 15-4, The FederalReserve Bank of Boston.
Simon, J., Smith, K. and West, T. (2010) Price incentives and consumerpayment behaviour, Journal of Banking and Finance, 34(8), 1759–1772.
Snellman, J., Vesala, J. and Humphrey, D. B. (2001) Substitution ofNoncash Payment Instruments for Cash in Europe, Journal of FinancialServices Research, 19(2/3), 131–145.
Soman, D. (2001) Effects of Payment Mechanism on Spending Behavior:The Role of Rehearsal and Immediacy of Payments, Journal ofConsumer Research, 27, 460–474.
Soman, D. (2003) The Effect of Payment Transparency on Consumption:Quasi Experiments from the Field, Marketing Letters, 14(3), 173–183.
208
SPA, Smart Payment Association (2016) An Overview of ContactlessPayment Benefits and Worldwide Developments, [Online].www.smartpaymentassociation.com (20.12.16).
Stavins, J. (2001) Effect of consumer characteristics on the use of paymentinstruments, New England Economic Review, 3, 19–31.
Stavins, J. (2011) Potential Effects of an Increase in Debit Card Fees,Public Policy Briefs No. 11-3, Federal Reserve Bank of Boston.
Stavins, J. (2013) Security of Retail Payments: The New StrategicObjective, Public Policy Discussion Paper No. 13-9, Federal ReserveBank of Boston.
Stix, H. (2003) How do debit cards affect cash demand? Survey DataEvidence, Working Paper 82, Oesterreichische Nationalbank.
Ten Raa, T. and Shestalova, V. (2004) Empirical evidence on paymentmedia costs and switch points, Journal of Banking and Finance, 28,203–213.
Thaler, R. H. (1985) Mental Accounting and Consumer Choice, MarketingScience, 4(3), 199–214.
Thaler, R. H. (1999) Mental Accounting Matters, Journal of BehavioralDecision Making, 12(3), 183–206.
Thaler, R. H. and Sunstein, C. R. (2008) Nudge: Improving DecisionsAbout Health, Wealth, and Happiness, New Haven: Yale University Press.
Thomas, M., Desai, K. K. and Seenivasan, S. (2011) How Credit CardPayments Increase Unhealthy Food Purchases: Visceral Regulation ofVices, Journal of Consumer Research, 38, 126–139.
Tobin, J. (1956) The interest elasticity of the transactions demand for cash,Review of Economics and Statistics, 38(3), 241–247.
Train, K. (2009) Discrete Choice Methods with Simulation, CambridgeUniversity Press (2nd ed.).
Trutsch, T. (2014) The impact of contactless payment on spending,International Journal of Economic Sciences, III, 70–98.
Trutsch, T. (2016) The impact of mobile payment on payment choice,Financial Markets and Portfolio Management, 30(3), 299–336.
209
Tversky, A. and Kahnemann, D. (1974) Judgment under uncertainty:Heuristics and biases, Science, 185(4157), 1124–1131.
van der Horst, F. and Matthijsen, E. (2013) The irrationality of paymentbehaviour, DNB Occasional Studies, 11(4).
van Hove, L. (2008) On the war on cash and its spoils, InternationalJournal of Electronic Banking, 1(1), 36–45.
van Hove, L. (2009) Could ’Nudges’ Steer Us Towards a ’Less-CashSociety’?, Social Science Research Network (SSRN) 1317360 .
Vandoros, S. (2013) My five pounds are not as good as yours, so I willspend them, Experimental Economics, 16(4), 546–559.
Venkatesh, V., Morris, M. G., Davis, G. B. and Davis, F. D. (2003) Useracceptance of information technology: Toward a unified view, MISQuarterly, 27(2), 425–478.
Visa (2010) Visa U.S.A. Interchange Reimbursement Fees, [Online].www.usa.visa.com (26.05.2014).
von Kalckreuth, U., Schmidt, T. and Stix, H. (2009) Choosing and usingpayment instruments: Evidence from German micro-data, WorkingPaper Series No. 1144, European Central Bank.
von Kalckreuth, U., Schmidt, T. and Stix, H. (2014) Using cash to monitorliquidity – Implications for payments, currency demand and withdrawalbehavior, Journal of Money, Credit and Banking, 46(8), 1753–1785.
Wakamori, N. and Welte, A. (2012) Why Do Shoppers Use Cash? Evidencefrom Shopping Diary Data, Working Paper 2012-24, Bank of Canada.
Wang, Y. (2008) Determinants affecting consumer adoption of contactlesscredit card: An empirical study, Cyber Psychology and Behavior, 11(6),687–689.
Weiner, S. and Wright, J. (2005) Interchange fees in various countries:Developments and determinants, Review of Network Economics, 4,290–323.
Wiechert, T. J. P. (2009) The Economics of Contactless Payment – AnAnalysis of the Financial Impact of Near Field Communication onStationary Retailers, Dissertation, University of St. Gallen.
210
Williamson, O. E. (1985) The economic institutions of capitalism: firms,markets, relational contracting, New York: The Free Press.
Wood, W. and Neal, D. T. (2009) The habitual consumer, Journal ofConsumer Psychology, 19, 579–592.
Xin, H., Techatassanasoontorn, A. A. and Tan, F. B. (2015) Antecedents ofconsumer trust in mobile payment adoption, Journal of ComputerInformation Systems, 55(4), 1–10.
Xu, Q., Zhou, Y., Ye, M. and Zhou, X. (2015) Perceived social supportreduces the pain of spending money, Journal of Consumer Psychology,25(2), 219–230.
Yang, S., Lu, Y., Gupta, S., Cao, Y. and Zhang, R. (2012) Mobile paymentservices adoption across time: An empirical study of the effects ofbehavioral beliefs, social influence, and personal traits, Computers inHuman Behavior, 28, 219–142.
Yu, C.-S. (2012) Factors affecting individuals to adopt mobile banking:Empirical evidence from the UTAUT model, Journal of ElectronicCommerce Research, 13(2), 104–121.
Zellermayer, O. (1996) The Pain of Paying, Ph.D. Dissertation, CarnegieMellon University.
Zinman, J. (2009) Debit or credit?, Journal of Banking and Finance, 33,358–366.
211
Curriculum Vitae
Born December 19, 1985 in St. Gallen, Switzerland
Education
02/2012–05/2017 University of St. Gallen (HSG), Switzerland
Ph. D. Program in International Affairs and Political
Economy, Main Focus: Economics
01/2010–06/2010 University of Maastricht, the Netherlands
Exchange Semester
10/2005–4/2011 University of Berne, Switzerland
Master of Science in Economics
08/2000–09/2004 Cantonal High School St. Gallen, Switzerland
Maturity
Professional Experience
01/2017–present Executive School of Management, Technology and Law
(ES-HSG), University of St. Gallen, Switzerland
Head of the Economics Division, Program Head of
“CAS Weiterbildung fur Politik”
06/2010–12/2016 Executive School of Management, Technology and Law
(ES-HSG), University of St. Gallen, Switzerland
Research and Teaching Assistant, Team of
Prof. em. Dr. Franz Jaeger, Program Head of “CAS
Weiterbildung fur Politik”
213
09/2009–11/2009 Swisscom AG, Berne, Switzerland
Research Assistant at the User Observatory
09/2008–10/2008 UBS AG, Zurich, Switzerland
Research Assistant Global Wealth Management
Economic Research Switzerland
Teaching Experience
2011–present University of St. Gallen (HSG), Switzerland
Teaching Assistant of the Economics Module in the
Executive Education (e.g. EMBA, EMBE, Inhouse
Programs)
2014–present University of St. Gallen (HSG), Switzerland
Teaching Assistant of the Course “Introduction to
Economics” on the Assessment Level
2011–2013 University of St. Gallen (HSG), Switzerland
Teaching Assistant of the Course “Microeconomics II”
on the Bachelor Level
214