Omni-Channel Retailing: An Analysis of Channel Interdependencies, Integration Services and Specific Marketing Instruments (Handel und Internationales Marketing Retailing and International Marketing) 3658347066, 9783658347062

Amelie Winters investigates omni-channel strategies in retail and provides new insights and important implications for r

139 17 2MB

English Pages 332 [326] Year 2021

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Foreword
Acknowledgements
Contents
Abbreviations
List of Figures
List of Tables
1 Introduction
1.1 Focus and Relevance
1.2 Research Gaps and Literature Review
1.2.1 Overview
1.2.2 Channel Interdependencies
1.2.3 Channel Integration Services
1.2.4 Online- and Omni-channel-specific Marketing Instruments
1.2.5 General Research Objectives
1.3 Structure and Contribution of the Studies
1.3.1 Reciprocity within Major Retail Purchase Channels and their Effects on Overall, Offline and Online Loyalty
1.3.2 Effects of Perceived Offline-Online and Online-Offline Channel Integration Services
1.3.3 Importance of Marketing Instruments for Repurchase Intentions in Omni-channel Retailing
1.4 Further Remarks
2 Study 1: Reciprocity within Major Retail Purchase Channels and their Effects on Overall, Offline and Online Loyalty
2.1 Introduction
2.2 Conceptual Framework and Hypothesis Development
2.2.1 Overview
2.2.2 Theory
2.2.3 Hypothesis on Reciprocity and its Effect
2.2.4 Hypothesis on the Role of Prior Experience
2.3 Empirical Studies
2.3.1 Overview
2.3.2 Study 1: Longitudinal Study
2.3.3 Study 2: Cross-sectional Study
2.4 Discussion and Implications
2.4.1 Overview
2.4.2 Theoretical Implications
2.4.3 Managerial Implications
2.5 Limitations and Directions for Further Research
3 Study 2: Effects of Perceived Offline-Online and Online-Offline Channel Integration Services
3.1 Introduction
3.2 Conceptual Framework and Hypothesis Development
3.2.1 Overview
3.2.2 Theory
3.2.3 Mediation Paths of the Integration Services
3.2.4 Role of Consumers’ Online Shopping Experience
3.2.5 Role of Channel Congruence
3.3 Empirical Study
3.3.1 Sample Design
3.3.2 Measurement
3.3.3 Method
3.3.4 Results
3.4 Discussion and Conclusion
3.4.1 Overview
3.4.2 Theoretical Implications
3.4.3 Managerial Implications
3.5 Limitations and Further Research
4 Study 3: Importance of Marketing Instruments for Repurchase Intentions in Omni-Channel Retailing
4.1 Introduction
4.2 Conceptual Framework and Hypothesis Development
4.2.1 Overview
4.2.2 Theory
4.2.3 Hypotheses on the Effects of Marketing Instruments
4.2.4 Hypotheses on Reciprocal Effects
4.3 Empirical Studies
4.3.1 Overview
4.3.2 Study 1: Sequential Mediation Design
4.3.3 Study 2: Cross-lagged Panel Design
4.3.4 Results
4.3.5 Stability Tests and Alternative Models
4.4 Discussion and Implications
4.4.1 Overview
4.4.2 Theoretical Implications
4.4.3 Managerial Implications
4.5 Limitations and Future Research
5 Final Remarks
5.1 Discussion and Conclusions
5.1.1 Core Results
5.1.2 Theoretical Implications
5.1.3 Managerial Implications
5.2 Further Research
Appendix
2. Study 1: Reciprocity within Major Retail Purchase Channels and their Effects on Overall, Offline and Online Loyalty
2.1. Sample Selection and Manipulation Check
2.2. Common-method Bias-testing
2.3. Unobserved Heterogeneity
2.4. Measurement Invariance Tests
2.5. Description of the Cross-lagged Panel Model
2.6. Alternative Cross-lagged Panel Models
2.7. Alternative Cross-sectional Models
3. Study 2: Effects of Perceived Offline-Online and Online-Offline Channel Integration Services
3.1. Common Method Variance
3.2. Endogeneity Test
3.3. Test for Measurement Invariance
3.4. Description of the Latent Moderated Structural Equation Method
3.5. Further Tests and Models
4. Study 3: Importance of Marketing Instruments for Repurchase Intentions in Omni-channel Retailing
4.1. Common Method Variance
4.2. Endogeneity Test
4.3. Test for Measurement Invariance
4.4. Description of the Cross-lagged Panel Model
References
Recommend Papers

Omni-Channel Retailing: An Analysis of Channel Interdependencies, Integration Services and Specific Marketing Instruments (Handel und Internationales Marketing Retailing and International Marketing)
 3658347066, 9783658347062

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Handel und Internationales Marketing Retailing and International Marketing Bernhard Swoboda · Thomas Foscht Hanna Schramm-Klein Hrsg.

Amelie Winters

Omni-Channel Retailing An Analysis of Channel Interdependencies, Integration Services and Specific Marketing Instruments

Handel und Internationales Marketing Retailing and International Marketing Series Editors Bernhard Swoboda, Universität Trier, Trier, Germany Thomas Foscht, Karl-Franzens-Universität Graz, Graz, Austria Hanna Schramm-Klein, Lehrstuhl für Marketing, Universität Siegen, Siegen, Germany

Die Schriftenreihe fördert die Themengebiete Handel und Internationales Marketing. Diese charakterisieren – jedes für sich, aber auch in inhaltlicher Kombination – die Forschungsschwerpunkte der Herausgeber. Beide Themengebiete werden grundsätzlich breit aufgefasst; die Reihe bietet sowohl Dissertationen und Habilitationen als auch Tagungs- und Sammelbänden mit unterschiedlicher inhaltlicher und methodischer Ausrichtung ein Forum. Die inhaltliche Breite ist sowohl im Sinne eines konsumentenorientierten Marketings wie auch einer marktorientierten Unternehmensführung zu verstehen. Neben den Arbeiten, die von den Herausgebern für die Schriftenreihe vorgeschlagen werden, steht die Reihe auch externen wissenschaftlichen Arbeiten offen. Diese können bei den Herausgebern eingereicht und nach einer positiven Begutachtung publiziert werden. The book series focuses on the fields of Retailing and International Marketing. These two areas represent the research fields of the editors—each of them as a single research area, but also in combination. Both of these research areas are widely understood. Consequently, the series provides a platform for the publication of doctoral theses and habilitations, conference proceedings and edited books, as well as related methodological issues that encompass the focus of the series. The series is broad in the sense that it covers academic works in the area of consumer-oriented marketing as well as the area of marketoriented management. In addition to academic works recommended by the editors, the book series also welcomes other academic contributions. These may be submitted to the editors and will be published in the book series after a positive assessment.

More information about this series at http://www.springer.com/series/12697

Amelie Winters

Omni-Channel Retailing An Analysis of Channel Interdependencies, Integration Services and Specific Marketing Instruments

Amelie Winters Fachbereich IV Professur für Marketing und Handel Universität Trier Trier, Germany Dissertation, Fachbereich IV, Trier University, 2021

ISSN 2626-3327 ISSN 2626-3335 (electronic) Handel und Internationales Marketing Retailing and International Marketing ISBN 978-3-658-34706-2 ISBN 978-3-658-34707-9 (eBook) https://doi.org/10.1007/978-3-658-34707-9 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Responsible Editor: Marija Kojic This Springer Gabler imprint is published by the registered company Springer Fachmedien Wiesbaden GmbH part of Springer Nature. The registered company address is: Abraham-Lincoln-Str. 46, 65189 Wiesbaden, Germany

Foreword

The effort to unify sales channels has a long history across consumer goods producers, while the importance of omni-channel retailing has increased in the last years. However, while research intensively discusses this phenomenon, retailers, esp. grocery retailers (i.e., most of the 30 biggest retailers in the world, still do not offer omni-channel experiences and may still learn from other retail sectors). Especially consumers’ use, perceptions, or evaluations of various options to integrate channels on the way to an omni-channel firm is valuable. The objective of Dr. Amelie Winters’ thesis is to gain a deeper knowledge on such issues in three studies: – Reciprocity within Major Retail Purchase Channels: Although online stores drive offline stores and vice versa, knowledge of these reciprocal effects is still scarce. This study analyzes whether reciprocal effects between images exist and how they affect consumers’ loyalty across channels. Study 1 analyses longitudinal data from 573 consumer evaluations of leading omni-channel fashion retailers using cross-lagged panel models. The results indicate that a positive offline channel image enhances the online channel image and vice versa. The reciprocal effects of offline (vs. online) images are stronger on overall retailer, offline channel and, surprisingly, online channel loyalty. Moreover, the effects change for retailers with which consumers have more or less favorable prior experiences. Study 2 tests identical relationships based on 380 consumer evaluations and reveals possible shortcomings of a typical cross-sectional design. The findings have direct implications for managers.

v

vi

Foreword

– Effects of Perceived Offline-Online (OF-ON) and Online-Offline (ON-OF) Channel Integration Services: Channel integration services such as click-and-collect provide consumers with a seamless experience across channels. However, managers still find it challenging to decide on those OF-ON and ON-OF services that increase channel quality. This study applies latent moderated structural equations to analyze mediation paths of most useful OF-ON and ON-OF services for consumers. Based on data from 722 consumer evaluations of fashion firms, the results show indirect-only effects. OF-ON services provide knowledge about and ease of access to both offline and online channels; ON-OF services show no links to offline channels. Higher levels of consumers’ online shopping experience reduce the mediation paths, and higher levels of perceived channel congruence positively and negatively moderate them. – The study “Importance of Marketing Instruments in Omni-channel Retailing” examines the relative importance of online-specific instruments like aesthetic appeal or customer service and omni-channel-specific instruments like channel consistency or integration. Although various instruments have been considered promising for omni-channel retailers, their relative effects for consumer behavior across channels remain unclear. This study conceptualizes a model in which perceived online trust and online brand equity translate these instruments into repurchase intentions, and tests for these indirect effects in a sequential mediation study and for reciprocal effects of online trust and brand equity in a cross-lagged panel study. The results provide new evidence of a different importance of the instruments and a stronger role of online brand equity than of online trust. These findings have direct implications for managers interested in understanding which particular instruments affect consumer outcomes most. With her work Dr. Amelie Winters makes a significant contribution to modern retailing research. She significantly disentangles the various interrelations of offline and online channels and indicated that consumers do not use, perceive or evaluate them identically. Her work impresses with the extent of attention paid to the conceptualization but also with the combination of different studies and in particular methodologically. I’m very happy with her work, as Dr. Amelie Winters presents the eighteens dissertation at my chair for Marketing & Retailing at the University of Trier. Besides a very good teaching, she was moreover involved in three book projects and has organized practical projects. I therefore thank Dr. Amelie Winters for these years in which she was working as a research assistant

Foreword

vii

at my chair. I got to know her as a very honourable, happy, kind and open minded person and I wish her warmly all the best for her career as well as for her private life in the future. Trier April 2021

Professor Dr. Prof. h.c. Bernhard Swoboda

Acknowledgements

This thesis has been developed during my time as a research assistant at the Chair of Marketing and Retailing (University of Trier). At this point, I would like to express my great gratitude to all the people who have supported me in the processing of my dissertation. This includes my supervisor, my colleagues, and last but not least my friends and dear family. At the beginning, I would like to thank my supervisor Prof. Dr. Prof. h.c. Bernhard Swoboda. He offered me the opportunity to receive a doctoral degree and work at his chair in fall 2016, for which I am very grateful. I would like to express my special thanks to him for his supervision and ongoing support throughout the entire execution of the thesis. He influenced my work through constructive feedback and productive discussions, and I thank him for his efforts and thoughts on my work. Here, I would also like to take the opportunity to thank my second supervisor Prof. Dr. Katrin Muehlfeld (University of Trier) for evaluating my work and Prof. Dr. Jörn Block (University of Trier) for agreeing to chair the defense committee. Many productive discussions and conversations at various conferences, doctoral colloquia, and workshops have advanced this dissertation. My participation in a statistics course in Melbourne, Australia in 2017 provided the methodological foundation, while presentations of my research at international conferences in the U.S. (Boston, Chicago), the U.K. (Glasgow), and Germany (Hamburg) further advanced this work. I am very grateful for the numerous interesting meetings and experiences I have had here. At this occasion, I would like to express my gratitude to my colleagues Dr. Lukas Morbe, Dr. Nadine Batton, Magdalena Klar, Carolina Sinning, Nils Fränzel and Lukas Zimmer that accompanied me on my way with advice, productive conversations and lovely words. A special thanks also goes to our secretary Ursula

ix

x

Acknowledgements

Fassbender, who offered me moral support and always had an open ear. Moreover, I would like to thank our student assistants Judith Feilen, Rebecca Rieger, Julia Coen, and Angelina Klink for their support throughout the years. I would like to express my deepest gratitude to my friends and family. My brothers and their partners, my niece, my partner’s parents, his brothers and partners, my godparents, my parents and my partner who believed in my work and in my abilities. Thank you for your patience, encouragement, and assurances while I worked on this dissertation. You have had my back for many hours, weeks, and years. Amelie Winters

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Focus and Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Research Gaps and Literature Review . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.2 Channel Interdependencies . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.2.1 Effects of Unidirectional Channel Relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.2.2 Effects of Bidirectional Channel Relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.3 Channel Integration Services . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.3.1 Effects of Joint Channel Integration Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.3.2 Effects of Offline-Online Channel Integration Services . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.3.3 Effects of Online-Offline Channel Integration Services . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.4 Online- and Omni-channel-specific Marketing Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.4.1 Effects of One or Two Marketing Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.4.2 Effects of More than Two Marketing Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.5 General Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Structure and Contribution of the Studies . . . . . . . . . . . . . . . . . . . .

1 1 6 6 7 7 11 14 14 15 18 21 21 26 30 33

xi

xii

Contents

1.3.1 Reciprocity within Major Retail Purchase Channels and their Effects on Overall, Offline and Online Loyalty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.2 Effects of Perceived Offline-Online and Online-Offline Channel Integration Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.3 Importance of Marketing Instruments for Repurchase Intentions in Omni-channel Retailing . . . . . . . . . . . . . . . . . 1.4 Further Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33

35 37 39

2 Study 1: Reciprocity within Major Retail Purchase Channels and their Effects on Overall, Offline and Online Loyalty . . . . . . . . . . 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Conceptual Framework and Hypothesis Development . . . . . . . . . . 2.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3 Hypothesis on Reciprocity and its Effect . . . . . . . . . . . . . . 2.2.4 Hypothesis on the Role of Prior Experience . . . . . . . . . . . . 2.3 Empirical Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2 Study 1: Longitudinal Study . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2.1 Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2.2 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2.5 Alternative Models . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.3 Study 2: Cross-sectional Study . . . . . . . . . . . . . . . . . . . . . . . 2.3.3.1 Sample, Measurement and Method . . . . . . . . . . . 2.3.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Discussion and Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.2 Theoretical Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.3 Managerial Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Limitations and Directions for Further Research . . . . . . . . . . . . . .

41 41 44 44 44 45 47 50 50 50 50 52 58 59 68 74 74 75 79 79 79 84 85

3 Study 2: Effects of Perceived Offline-Online and Online-Offline Channel Integration Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Conceptual Framework and Hypothesis Development . . . . . . . . . . 3.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

87 87 90 90

Contents

xiii

3.2.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3 Mediation Paths of the Integration Services . . . . . . . . . . . . 3.2.4 Role of Consumers’ Online Shopping Experience . . . . . . . 3.2.5 Role of Channel Congruence . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Empirical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Sample Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.4.1 Results of the Path Coefficients . . . . . . . . . . . . . . 3.3.4.2 Results of the Moderators . . . . . . . . . . . . . . . . . . . 3.4 Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.2 Theoretical Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.3 Managerial Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Limitations and Further Research . . . . . . . . . . . . . . . . . . . . . . . . . . .

91 92 94 95 96 96 97 107 107 107 120 123 123 124 128 129

4 Study 3: Importance of Marketing Instruments for Repurchase Intentions in Omni-Channel Retailing . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Conceptual Framework and Hypothesis Development . . . . . . . . . . 4.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.3 Hypotheses on the Effects of Marketing Instruments . . . . 4.2.4 Hypotheses on Reciprocal Effects . . . . . . . . . . . . . . . . . . . . 4.3 Empirical Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Study 1: Sequential Mediation Design . . . . . . . . . . . . . . . . . 4.3.2.1 Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2.2 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3 Study 2: Cross-lagged Panel Design . . . . . . . . . . . . . . . . . . 4.3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.5 Stability Tests and Alternative Models . . . . . . . . . . . . . . . . 4.4 Discussion and Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Theoretical Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3 Managerial Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Limitations and Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . .

131 131 134 134 135 137 140 141 141 141 141 143 149 150 150 160 172 172 174 178 179

xiv

Contents

5 Final Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Discussion and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 Core Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.2 Theoretical Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.3 Managerial Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Further Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

181 181 181 186 190 194

Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

199

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

291

Abbreviations

ANOVA AVE b bn. CA CEO CFA CFI CI CMV CR df e.g. et al. etc. EUR FL fmRI H i.a. i.e. ItTC IV JIS KMO LIL

Analysis of variance Average variance extracted Unstandardized coefficient Billion Channel attributes Chief executive officer Confirmatory factor analysis Comparative fit index Channel image Common method variance Composite reliability Degrees of freedom Exempli gratia/for example Et alia/and others Et cetera/and so on Euro Factor loadings Functional magnetic resonance imaging Hypothesis Inter alia/among others Id est/that is Item-to-total-correlation Instrumental variable Joint perceived integration services Kaiser-Meyer-Olkin criterion Lower interval limit

xv

xvi

LMS LOY MANOVA MLM MS MSEM MV N ns OF OFB OFLOY OFO OF-ON OFP OFPI OFPRI ON ONB ONLOY ONO ON-OF ONP ONPI ONRPI OS ONT p p. PLS R2 RBE RMSEA RPI SCF SE SELF SEM SRMR

Abbreviations

Latent moderated structural equation method Loyalty Multivariate analysis of variance Maximum likelihood estimator with robust standard errors Mean square Multilevel structural equation modeling Mean value Sample size not significant Offline channel image Offline retail brand equity Offline channel loyalty Perceived quality of offline offerings Perceived offline-to-online services Offline experience Offline purchase intention Offline repurchase intention Online channel image Online retail brand equity Online channel loyalty Perceived quality of online offerings Perceived online-to-offline services Online experience Online purchase intention Online repurchase intention Perceived quality of offerings Online trust p-value Page Partial least squares R-square Retail brand equity Root mean square error of approximation Repurchase intention Scaling correction factor Standard error Self-efficacy Structural equation modeling Standardized root mean square residual

Abbreviations

Std. t TLI TPI U.S. UIL vs. α β λ % χ2 

xvii

Standard deviation t-value Tucker-Lewis index Total purchase intention United States of America Upper interval limit Versus Cronbach’s alpha Standardized coefficient standardized factor loadings (CFA) Per cent Chi-square Difference

List of Figures

Figure 1.1 Figure 1.2 Figure Figure Figure Figure

1.3 1.4 1.5 1.6

Figure 1.7 Figure 1.8 Figure 1.9 Figure 2.1 Figure Figure Figure Figure

2.2 3.1 3.2 3.3

Figure 3.4 Figure 4.1 Figure 4.2

Online sales across different business models in Germany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Online sales share of total sales in the food, electronics and fashion sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Studies on unidirectional channel relationships . . . . . . . . . . . Studies on bidirectional channel relationships . . . . . . . . . . . . Studies on effects of joint channel integration services . . . . Studies on effects of offline-online channel integration services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Studies on effects of online-offline channel integration services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Studies on one or two examined marketing instruments . . . . Studies on more than two examined marketing instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Literature review on channel relationships in omni-channel retailing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conceptual framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Literature review on integration services . . . . . . . . . . . . . . . . Conceptual framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Plots of the conditional total effects: consumers’ online shopping experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Plots of the conditional indirect effects: perceived channel congruence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Literature review on marketing instruments . . . . . . . . . . . . . . Conceptual framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 4 9 13 16 19 20 22 28 42 44 88 91 123 124 132 135

xix

xx

Figure 5.1 Figure 5.2 Figure A.1 Figure A.2 Figure A.3

List of Figures

Overview of analysis steps and resulting effects on behavioral outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Possible interdependencies between further touchpoints . . . Cross-lagged panel model . . . . . . . . . . . . . . . . . . . . . . . . . . . . Plots of the conditional indirect effects: consumers’ online shopping experience . . . . . . . . . . . . . . . . . . . . . . . . . . . Cross-lagged panel model . . . . . . . . . . . . . . . . . . . . . . . . . . . .

193 195 219 275 289

List of Tables

Table Table Table Table Table Table Table

2.1 2.2 2.3 2.4 2.5 2.6 2.7

Table 2.8 Table Table Table Table Table Table Table Table Table Table Table

2.9 2.10 2.11 2.12 3.1 3.2 3.3 3.4 3.5 3.6 3.7

Table 3.8 Table 3.9

Sample characteristics of cross-lagged design . . . . . . . . . . . . Reliability and validity of cross-lagged panel models . . . . . Reliability and validity of cross-lagged panel models II . . . Discriminant validity of cross-lagged panel models . . . . . . . Results of general cross-lagged panel models . . . . . . . . . . . . Results of cross-lagged panel moderator models . . . . . . . . . Results of alternative general cross-lagged panel models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of general cross-lagged panel models with perceived levels of channel integration . . . . . . . . . . . . . Sample characteristics of cross-sectional design . . . . . . . . . . Reliability and validity of cross-sectional models . . . . . . . . . Discriminant validity of cross-sectional models . . . . . . . . . . Results of proposed and rival cross-sectional models . . . . . . Sample characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Selection of perceived integration services (pretest) . . . . . . . Reliability and validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discriminant validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of path coefficients: offline purchase intention . . . . . Results of path coefficients: online purchase intention . . . . . Test for direct and indirect effect using bootstrapping: offline purchase intention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Test for direct and indirect effect using bootstrapping: online purchase intention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of alternative models: ON-OF services only and online offerings only . . . . . . . . . . . . . . . . . . . . . . . .

52 53 55 57 60 62 69 72 74 76 78 81 97 99 103 106 108 110 113 115 117

xxi

xxii

List of Tables

Table 3.10 Table 3.11 Table 3.12 Table Table Table Table Table Table Table Table Table Table

3.13 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9

Table 4.10 Table 4.11 Table 4.12 Table Table Table Table Table

4.13 4.14 4.15 A.1 A.2

Table Table Table Table

A.3 A.4 A.5 A.6

Table Table Table Table

A.7 A.8 A.9 A.10

Results of alternative models: offline purchase intention with one offering only . . . . . . . . . . . . . . . . . . . . . . . Results of alternative models: online purchase intention with one offering only . . . . . . . . . . . . . . . . . . . . . . . Results of alternative models: joint perspective of integration services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of alternative models: total purchase intention . . . . . Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: reliability and validity . . . . . . . . . . . . . . . . . . . . . . . . Study 1: discriminant validity . . . . . . . . . . . . . . . . . . . . . . . . . Study 2: reliability and validity . . . . . . . . . . . . . . . . . . . . . . . . Study 2: reliability and validity II . . . . . . . . . . . . . . . . . . . . . . Study 2: discriminant validity . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the sequential mediation models . . . . . . . Study 2: results of the cross-lagged models . . . . . . . . . . . . . . Study 1: results of the alternative model (online brand equity only) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the alternative model (online trust only) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the alternative model (total repurchase intention) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 2: results of the alternative model (total repurchase intention) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the alternative model (brand equity) . . . . Study 2: results of the alternative model (brand equity) . . . . Results overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sales growth 2015–2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Independent sample t-tests: prior offline and online experiences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ANOVA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the single-factor test . . . . . . . . . . . . . . . . . Study 1: results of the model comparisons (phase I) . . . . . . Study 1: results of the reliability decomposition (phase II) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the sensitivity analyses (phase III) . . . . . Study 1: F-test of strong instrumental variables . . . . . . . . . . Study 1: results of the efficient and consistent models . . . . . Study 1: measurement invariance across time points . . . . . .

118 119 121 122 143 144 148 151 152 154 155 159 161 163 166 168 170 173 175 200 200 201 202 203 207 209 212 213 218

List of Tables

Table A.11 Table A.12 Table A.13 Table A.14 Table A.15 Table A.16 Table A.17 Table A.18 Table A.19 Table A.20 Table Table Table Table Table Table Table Table Table

A.21 A.22 A.23 A.24 A.25 A.26 A.27 A.28 A.29

Table A.30 Table A.31 Table A.32 Table A.33 Table A.34 Table A.35

xxiii

Study 1: reliability and validity of alternative general models I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: reliability and validity of alternative general models II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: discriminant validity of alternative general models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the single-factor test of alternative general models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the model comparisons (phase I) for alternative general models . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the reliability and decomposition (phase II) for alternative general models . . . . . . . . . . . . . . . . Study 1: results of the sensitivity analyses (phase III) for alternative general models . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: F-test of strong instrumental variables for alternative general models . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the efficient and consistent alternative general models . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: measurement invariance across time points of alternative general models . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of alternative moderator models . . . . . . . . . Study 2: F-test of strong instrumental variables . . . . . . . . . . Study 2: results of the efficient and consistent models . . . . . Results of the single-factor tests . . . . . . . . . . . . . . . . . . . . . . . Results of the model comparisons (phase I) . . . . . . . . . . . . . Results of the reliability decomposition (phase II) . . . . . . . . Results of the sensitivity analyses (phase III) . . . . . . . . . . . . F-test of strong instrumental variables . . . . . . . . . . . . . . . . . . Results of the efficient and consistent models: offline purchase intention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of the efficient and consistent models: online purchase intention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Measurement invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of the multigroup models: offline purchase intention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of the multigroup models: online purchase intention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Study 1: results of the single-factor tests . . . . . . . . . . . . . . . . Study 2: results of the single-factor tests . . . . . . . . . . . . . . . .

219 222 224 226 228 234 236 239 240 247 248 256 257 259 260 261 262 264 265 267 269 271 273 276 276

xxiv

Table Table Table Table Table

List of Tables

A.36 A.37 A.38 A.39 A.40

Study Study Study Study Study

1: 1: 2: 2: 2:

F-test of strong instrumental variable . . . . . . . . . . . results of the efficient and consistent models . . . . . F-test of strong instrumental variables . . . . . . . . . . results of the efficient and consistent models . . . . . measurement invariance across time points . . . . . .

278 279 284 285 288

1

Introduction

1.1

Focus and Relevance

The retail landscape has changed over the last decade. New channels have been added to traditional ones, transforming formerly brick-and-mortar retailers into multi- and omni-channel retailers. By definition, multi-channel retailers offer more than one channel to attract consumers. In contrast, omni-channel retailers offer a seamless experience through integrated channels and touchpoints (Valentini et al. 2020; Verhoef et al. 2015). Today, consumers switch between channels and expect an integrated and consistent cross-channel shopping experience, especially between retailers’ major offline and online sales channels (Hult et al. 2019; Lemon and Verhoef 2016). If those criteria of a cross-channel shopping experience are met, retailers can benefit from enhanced customer loyalty, repurchase behavior, and ultimately greater profitability (e.g., Homburg et al. 2017; Kumar et al. 2017). Thus, in an increasingly competitive retail environment, retailers must identify stimuli and cues that influence and shape consumers’ shopping experiences and, consequently, consumer responses and behavior (Becker and Jaakkola 2020). Retailers are already placing particular emphasis on the coordination of physical and online stores due to assumed interdependencies instead of relying on channel-specific decisions (Bell et al. 2018). For example, retailers such as Walmart or Zara retain loyal consumers by encouraging image transfers between sales channels (Forbes 2019) or by offering technology-based integration services that help consumers engage in different interactive activities across channels (e.g., click-and-collect, Banerjee 2014; Sousa and Amorim 2018). In addition, retailers

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 A. Winters, Omni-Channel Retailing, Handel und Internationales Marketing Retailing and International Marketing, https://doi.org/10.1007/978-3-658-34707-9_1

1

2

1

Introduction

increasingly compete for consumers online and use online- and omni-channelspecific marketing instruments in online stores to benefit from cross-channel effects. However, retailers still face challenges in how to target consumers from different channels and to adapt their practices more effectively according to these challenges (e.g., Acquila-Natale and Chaparro-Peláez 2020). Therefore, this doctoral thesis focuses on interdependencies between retailers’ major purchase channels and emphasizes the need to identify supportive stimuli and cues, such as integration services and online- and omni-channel-specific marketing instruments, for such transfers. At the start of this doctoral thesis project, multi-channel retailing was already considered an important topic, as documented in a special issue of the Journal of Retailing in 2015. The influence of multiple channels on performance, consumer behavior, and initial channel integration had been analyzed (e.g., Cao and Li 2015; Melis et al. 2015; Herhausen et al. 2015). The ongoing importance of this topic remains evident, as underlined by a special issue of the same journal in 2020. It highlights customer journey management and, in particular, loop aspects of consumer experiences across channels, including consumers’ cognitive and behavioral responses, which are important in omni-channel retailing (Grewal and Roggeveen 2020). With special regard to the dynamic and practical development, questions arise whether interdependencies between channels exist and how offered integration services as well as online- and omni-channel-specific marketing instruments affect consumers and their behaviors today. In recent years, as mentioned above, an increasing number of retailers has made efforts to develop omni-channel strategies. Former brick-and-mortar retailers, such as Macy’s or Walmart, have added and integrated online capabilities, while online pure players are increasingly opening physical stores to benefit from cross-channel effects (e.g., Jindal et al. 2021; Valentini, Neslin and Montaguti 2020). The practical relevance of this strategy is confirmed when looking at online sales across different business models in Germany (see Figure 1.1). Omni-channel retailers increased their sales to a total of EUR 25.7 billion in 2019, compared to online pure players with EUR 10.8 billion. While online marketplaces with a different business model are still the strongest (dominated by Amazon’s high sales), omni-channel retailers have more than doubled their online sales in the last six years (BEVH/Beyonddata 2019, 2020; BEVH/GIM 2015, 2016, 2017, 2018). According to a survey by the German Trade Association, a further increase in sales for omni-channel retailers could have been expected (HDE 2020), but now the spread of the novel coronavirus poses challenges, especially for retailers of non-essential goods. These retailers have faced significant drops in sales and

1.1 Focus and Relevance

3

will have to adopt even more to new methods when trying to reach consumers (Roggeveen and Sethuraman 2020).

Online sales across different business models in Germany (in bn. EUR) 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0

25.0

23.0

11.0

20.1 14.0 6.3

33.9

30.6

27.9

26.7

25.7 22.7

16.7 7.6

8.6

9.8

10.8

2.0

2014

2015 Online pure player

2016 2017 Omni-channel retailer

2018 Online marketplaces

2019

Figure 1.1 Online sales across different business models in Germany. (Source: BEVH/Beyonddata (2019, 2020), BEVH/GIM (2015, 2016, 2017, 2018))

In light of the heavy investments and resources retailers need to build crosschannel capabilities, there is an urgent need to better understand whether and how such strategies reach the consumer (Luo et al. 2015; Tagashira and Minami 2019). Consumer perceptions of retail brands, like H&M or C&A, drive consumer behavior (e.g., Lee et al. 2019; Li et al. 2018), and therefore, this doctoral thesis focuses on the consumer perspective. Importantly, consumers not only decide on a distinct retailer but on one or more channels from that retailer, which requires a monitoring of retailer-level and channel-level outcomes (Herhausen et al. 2015). Regarding the retail sectors with the highest market shares in Germany, the sales distribution shows that online retailing is growing fastest, while offline retailing in Germany is still the anchor of the sectors (see Figure 1.2). Compared with the largest food sector, at only 0.7 % of EUR 215 billion, and the electronics sector, at 4.2 % of EUR 41 billion, the fashion sector has the largest share of online sales out of total sales at 4.9 % of EUR 37 billion in 2018 (Destatis 2020). This sector represents the primary driver of online sales growth. Moreover, the fashion sector is less concentrated compared to the larger food and electronics sectors, with over 40 firms accounting for two-thirds of total sales. Fashion retailers offer multiple channels and integration services and consumers take advantage of omni-channel retailing (e.g., Patten et al. 2020). With 71 fashion retailers, compared to 41 in the electronics sector, among the 312 omni-channel retailers in Germany, the fashion sector is the most dominant (EHI 2019). This sector offers a profound selection of several well-known omni-channel retailers, which helps to avoid single firm-specific results in empirical studies. Therefore,

4

1

Introduction

Online sales share of total sales in the sector (in % and in bn. EUR) 6.0

5.3

5.0 4.0

3.1

3.0

3.7

4.9

4.5

4.4

3.9

3.6

4.2

3.6

2.9 2.3

2.0 1.0 0.0

0.1

2005 2005

Total sales (in bn. EUR) 138 25 25

0.2

2010 2010 171 42 30

0.4

0.5

2015 2016 2015 2016 Food Electronics Fashion 193 41 37

202 40 36

0.7

0.7

2017 2017

2018 2018

210 41 37

215 41 37

Figure 1.2 Online sales share of total sales in the food, electronics and fashion sectors. (Source: Destatis (2020))

this doctoral thesis consistently focuses on leading and in pretests selected omnichannel retailers for the empirical studies. In the academic literature, interdependencies are considered to be critical in today’s retail environment. This is because transfers occur between channels (e.g., Herhausen et al. 2020; Verhagen et al. 2019). Channel images are witnessed as important antecedents in retail studies. They reflect the ways in which stores are defined in consumers’ mind, including attitudes based on subjective perceptions of various attributes (Martineau 1958; Kwon and Lennon 2009a). The image of one channel can be transferred to another channel and influence consumer behavior (for a detailed literature review on channel interdependencies, see Section 1.2.2). Extant research, however, has only studied either unidirectional, i.e., one-way (e.g., Chang and Tseng 2013) or bidirectional, i.e., cross-channel relationships (e.g., Badrinarayanan et al. 2012). Although, cross-channel effects are considered, it is surprising that important reciprocal relationships are neglected, since one channel can affect a second channel and the second channel in turn can affect the first. Conceptualizing such loop-like relationships could change extant results and implications and requires further investigation. In addition, in omni-channel retailing, consumer evaluations depend on prior experiences (e.g., Grewal and Roggeveen 2020), but their important role was seldom considered. To support such interdependencies with stimuli and cues, retailers are integrating multiple business functions across channels with integration services (Gallino et al. 2017; Tagashira and Minami 2019). These technology-based services provide consumers with knowledge about and ease of access to a channels’ offerings (e.g., Sousa and Amorim 2018). Research has shown that integration services affect various outcomes (e.g., sales growth, Cao and Li 2015; repurchase

1.1 Focus and Relevance

5

intention, Lee et al. 2019; for a detailed literature review on channel integration services, see Section 1.2.3). However, retailers still find it challenging to determine the appropriate services that increase channel quality and behavioral outcomes (Banerjee 2014; Bolton et al. 2018). Most often, a joint perspective of integration services has been examined (e.g., Hamouda 2019). Fewer studies have addressed offline-to-online (OF-ON) or online-to-offline (ON-OF) services separately (e.g., Bhargave et al. 2016; Jara et al. 2018). These studies fail to show cross-channel effects of simultaneously offered, but distinctly perceived OF-ON and ON-OF services. In a similar vein, consumers’ online shopping experiences are known to be an important context factor for ON-OF services (e.g. Herhausen et al. 2015), but this has not been conceptualized for the disentangled services yet. In addition, although retailers today exert considerable efforts in achieving channel congruence (e.g., Van Baal 2014), the effects for integration services remain unclear. Moreover, omni-channel retailers increasingly compete for consumers online and have to orchestrate online- and omni-channel-specific marketing instruments to remain competitive (e.g., aesthetic appeal, online-offline integration, Toufaily and Pons 2017; Yang et al. 2020). Scholars have acknowledged the importance of such stimuli for different outcomes across sectors (for a detailed literature review on online- and omni-channel-specific marketing instruments, see Section 1.2.4). An effective orchestration depends on the relative importance of instruments and the mechanisms that translate them into behavior (e.g., Bleier et al. 2019). Although numerous studies exist, usually only one or two instruments are examined, which is inadequate for effective management. A few studies on more than two instruments present indirect effects on consumer behavior, but do not consider online- and omni-channel-specific instruments simultaneously. Online trust, which is defined as a website’s delivery of confidence or reliability (Kim and Peterson 2017; Ye et al. 2020), is a known mechanism that translates instruments into behavior. In retailing, firms like H&M or Zara see consumer perceptions of their brand equity, defined as associations of a retailer’s website as a strong, attractive, unique, and favorable brand (Keller 2010), as a core competence (e.g., Swoboda et al. 2013). However, the important role of retail brands and assumed interrelationships between online trust and brand associations (e.g., Hollebeek and Macky 2019) have not been satisfactorily addressed. In summary, retail managers face complex decisions regarding their omnichannel strategies. These decisions concern (1) reciprocal relationships between major sales channels and the role of prior experience, (2) effects of OF-ON and ON-OF services and respective moderations of consumers’ online shopping experience and channel congruence, and (3) the importance of online- and

6

1

Introduction

omni-channel-specific instruments in the growing online competition (including reciprocal effects of online trust and brand equity). In view of the growing competition in offline and, especially, online retail markets due to increasing sales volumes, the following central questions were identified for the present doctoral thesis: (1) Do reciprocal relationships between offline and online channel images of major sales channels exist, and if so, how do they affect retailer-level and channel-level loyalty and whether and how these relationships are moderated by prior offline and online experiences? (2) Whether and how do perceived OF-ON vs. ON-OF services influence offline and online channel purchase intentions and how do consumers’ online shopping experience and channel congruence moderate these effects? (3) How important are online- and omni-channel-specific instruments in affecting online and offline repurchase intentions as mediated by online trust and brand equity, and do reciprocal relationships exist between online trust and brand equity? To answer these three central research questions, three specific studies have been conducted that are presented in this doctoral thesis. As a basis for the studies, Section 1.2 presents the specific research gaps and a literature review for each of the three topics, differentiated by studies on channel interdependencies (Section 1.2.2), channel integration services (Section 1.2.3), and online- and omni-channel-specific marketing instruments (Section 1.2.4).

1.2

Research Gaps and Literature Review

1.2.1

Overview

The following literature review is intended to provide a comprehensive overview of the relevant literature on the decisions of omni-channel retailers and the resulting effects on different behavioral channel-level and retailer-level outcomes. Literature was selected from journals which are ranked at least with level C and higher according to the VHB-JOURQUAL 3 journal ranking of the German Academic Association for Business Research (VHB 2020). This was decided because of the large amount of current literature on this topic and on the advice of my supervisor. Relevant and current literature for each of the three topics was selected. Literature reviews that include further work can be found

1.2 Research Gaps and Literature Review

7

in doctoral theses of Berg (2013) and Weindel (2016). The chapter is structured as follows. Section 1.2.2 summarizes the literature on channel interdependencies. Section 1.2.3 addresses former research on channel integration services. Section 1.2.4. compiles the literature on online- and omni-channel specific marketing instruments. Studies that can be assigned to Sections 1.2.2 to 1.2.4 are summarized in the first relevant section, but are mentioned repeatedly in the other sections if relevant. Finally, in Section 1.2.5, the general research objectives of this doctoral thesis are presented. These objectives form the basis of the subsequent three studies.

1.2.2

Channel Interdependencies

1.2.2.1 Effects of Unidirectional Channel Relationships Studies on unidirectional effects are more often examined on single channel outcomes than on various important outcomes in omni-channel retailing. The studies show effects on offline and online channel outcomes and towards the retailer in general, often in cross-sectional designs (see Figure 1.3). First, regarding offline channel outcomes, Fuentes-Blasco et al. (2017) analyzed whether marketing and technological innovation affect satisfaction and word-of-mouth mediated through store image, consumer value, and store brand equity. The authors found direct effects of the innovations on image, consumer value, and satisfaction. Store image and store brand equity enhance satisfaction, which in turn positively affects word-of-mouth. Hsu et al. (2010) analyzed the effects of offline store image, satisfaction and travel distance and found positive direct and indirect effects on behavioral intentions. Moliner-Velázquez et al. (2019) confirmed that store image positively affects the value dimensions excellence, efficiency, entertainment, and aesthetics, while retail innovation affects excellence, entertainment, and aesthetics only. Loyalty is affected by excellence, efficiency, entertainment, and aesthetics mediated by satisfaction. Second, regarding online channel outcomes, Aghekyan-Simonian et al. (2012) investigated whether product brand image and online channel image impacted consumer perceptions of specific types of risks and online purchase intentions. Their results showed that product brand image positively affects online purchase intentions and negatively affects product performance risk. Online channel image positively affects online purchase intentions, mediated by perceived product risk. Badrinarayanan et al. (2014) discovered the role congruity between offline and online stores and between self-image and online store image plays in building online trust and attitude. The authors pointed to direct and indirect effects of

8

1

Introduction

online trust on online attitude, while online attitude affected purchase intentions. Congruity directly increases online trust and affects online attitude, while self-image congruity is shown to enhance online trust and attitude. In a crosssectional study, online store image was moreover shown to be strongly linked to online purchase intention mediated by online shopping utilitarian value and online risk (Chang and Tseng 2013). The study of Kwon and Lennon (2009b) discussed the role of offline brand image for online brand image and online perceived risk, and how these constructs enhance online loyalty. The authors found positive direct effects of offline brand image on online brand image and on online loyalty. The way that navigation performance and incongruity between online and offline brand images affect consumer attitudes towards a firm’s website, perceived flow and revisit intentions was examined by Landers et al. (2015). The results confirmed that navigation positively influences perceived flow and attitude toward the website, while incongruity and the interaction between navigation and incongruity negatively impact flow. Perceived flow positively affects revisit intentions by affecting consumers’ attitudes toward the website. The moderations show that the relationship between navigation performance and perceived flow is weaker when incongruity is higher and that the indirect effect of navigation performance on revisit intentions is weaker at higher levels of incongruity. Third, regarding effects on overall retailer outcomes, Anselmsson et al. (2017) showed that image dimensions affect loyalty mediated by trust. Bezes (2013) hypothesized that perceived congruence of channels improves the website and store image of a retailer for different consumer buying groups. The author highlighted that website image positively affects a retailer’s overall image for offline and multi-channel buyers. Store image positively affects a retailer’s overall image for offline, multi-channel and online buyers. Both store and website image exert positive impacts on the perceived congruence between channels, while congruence between channels positively affects retailer image for multi-channel and online buyers. Hunneman et al. (2017) examined how different types of shopping trips change the effects of image on satisfaction. Specifically, they found that for major shoppers, treat service factors are less important to their satisfaction. Convenience is the most important driver of satisfaction for regular fill-in shoppers, and price most highly impacts satisfaction for fill-in trips related to special occasions. The only study that examined unidirectional effects on various important outcomes for retailers was Herhausen et al. (2015), who conceptualized perceived service quality and perceived risk as mediators between online-offline integration and search intention, purchase intention, and willingness to pay toward the retailer, at the offline and online stores. Their analysis reveals that the perceived service quality of the online store serves as a mediator between online-offline integration

1.2 Research Gaps and Literature Review

Author(s) and year

Research question

Fuentes- How do marketing and Blasco et al. technological innova(2017) tion affect satisfaction and word-of-mouth through store image, consumer value and store brand equity? Grosso et al. - What are the main driv(2018) ers of store loyalty and how does the possible range of levers available to retail companies look like to enter the Indian market? Hsu, Huang - Whether do interrelaand tionships among groSwanson cery store image, travel (2010) distance, customer satisfaction, and behavioral intentions exist?

Moliner- How do store image Velázquez et and retail innovation imal. (2019) pact value, satisfaction and loyalty?

Van Nierop - Does the introduction et al. (2011) and use of informational websites affect offline shopping trips and money spent offline? Aghekyan- - How do product brand Simonian et and online store image al. (2012) impact types of risks and online purchase intentions for fashion apparel products? Badrinara- What role does congruyanan, ity between offline and Becerra and online stores and beMadhatween self-image and varam online store image (2014) plays in building online trust and attitude? Bezes - What dimensions de(2014) fine store and website image? - How does image affect satisfaction and purchase attitude? Carlson et - How do perceived al. (2015) online channel values and customer perceptions affect online channel satisfaction and loyalty?

9

Theory/ Sample and Core findings framework method Effects on offline channel outcomes - Expectation - n = 820 cus- Marketing and technological innovation directly affect and discontomers tore image, whereas technological innovation also firmation - SEM impacts consumer value and satisfaction. theory - Consumer value directly affects store brand equity. - Signaling - Satisfaction is affected by store image and store theory brand equity and positively impacts word-of-mouth (referral and activity). - None - n = 1,651 cus- - Customer satisfaction is a main driver of store loyalty. tomers - Store environment and perceived value by customers af- SEM fect customer satisfaction. - Perceived value is influenced by the retailers’ product assortment decisions. - Promotions do not have an impact on the perceived value, while the perceived value has only a small and negative impact on store loyalty. - Central - n = 258 cus- Grocery store image is identified as a second-order conplace theory tomers struct reflected by the three key components of mer- SEM chandise attributes, store ambience and service, and marketing attractiveness. - Store image affects behavioral intentions, while its indirect effect through customer satisfaction is greater than its direct effect on behavioral intentions. - Travel distance directly influences customer satisfaction and indirectly impacts behavioral intentions. - None - n = 820 cus- Store image is positively related with the value ditomers mensions: excellence, efficiency, entertainment and - SEM aesthetics. - Retail innovation affects excellence, entertainment and aesthetics, but does not affect efficiency. - Excellence, efficiency, entertainment and aesthetics affect satisfaction, while satisfaction affects loyalty. - Theory of - n = 436 custom- - The use of a newly introduced informational website replanned beers sults in fewer shopping trips. havior - Tobit model - The use of a newly introduced informational website leads to fewer purchases and less money spent in the retailer’s offline store. Effects on online channel outcomes - Theory of - n = 875 re- Product brand image positively influences purchase perceived spondents intentions and negatively product performance risk. risk - SEM - Online store image does not affect purchase intentions, but positively affects perceived product risk and negatively influences perceived financial/time risk. - Financial/time risk and product performance risk reduce purchase intentions. - Theory of - n = 316 re- Online trust directly influences online attitude and has reasoned ac- spondents an indirect and a total effect through online attitude tion - SEM on purchase intentions. - Categoriza- Online attitude directly affects purchase intentions. tion theory - Congruity directly increases online trust and has an indirect and a total impact on online attitude. - Self-image congruity enhances online trust and attitude. - None - n = 1,478 con- - Store and website image dimensions consist of offersumers ings, price, layout, accessibility, promotions, cus- SEM tomer service, advice, reputation, institution and connections with other channels. - Store and website image influence satisfaction and purchase attitude. - Customer - n = 448 respond-- Service performance value, emotional value, moneperceived ents tary value, brand integration value and channel convalue theory - SEM venience value are formative first-order components of perceived online channel value. - Value affects online channel satisfaction and online channel loyalty intention, while the two are related.

Figure 1.3 Studies on unidirectional channel relationships. (Source: Own creation)

10

Chang and Tseng (2013)

Chu et al. (2017)

Kwon and Lennon (2009b) Landers et al. (2015)

Wang et al. (2009)

Wu et al. (2013)

1

- Are e-store image, perceived value and purchase intention related? - Does perceived risk moderate links of perceived value and purchase intention? - What are the antecedents of online attitude and offline attitude? - How do both attitudes influence offline purchase intention and ultimately online purchase intention? - How do online risk, online image, and offline image determine online loyalty? - Do navigation performance and incongruity between online and offline retailer brand images influence consumer attitudes toward a firm’s website and revisit intentions through perceived flow? - Does congruity between offline and online stores affect the relationship between website performance, offline and online attitudes? - Are store layout design, atmosphere, emotional arousal and online shopping intention related?

Anselmsson, - What constitutes cusBurt and tomer-based brand Tunca equity and image in (2017) retailing and how do they affect loyalty? Bezes - Does congruence of (2013) channels improve the website and store image of a retailer?

Hunneman, - How do different types Verhoef and of shopping trips Sloot (2017) change the effect of image on satisfaction?

Figure 1.3 (continued)

- None

- n = 332 respondents - SEM

Introduction

- E-store image and online shopping utilitarian value of an online retailer’s offers are positively related. - E-store image affects a consumer’s evaluation of hedonic value offered by an online retailer. - Higher online shopping utilitarian value and online shopping hedonic value enhance purchase intention. - Financial risk does not moderate the links between online shopping utilitarian and hedonic value. - Theory of - n = 520 re- Basic attributes and perceived risk influence online planned be- spondents attitude, whereas firm reputation and perceived risk havior - SEM positively affect offline attitude. - Technology - Offline attitude and online attitudes are positively reacceptance lated and offline purchase intention is positively afmodel fected by offline attitude. - Online purchase intention is enhanced by offline purchase intention and online attitude. - Theory of - n = 671 students - Offline brand image positively influences online brand cognitive dis- - (M)ANOVA image, online perceived risk, and online loyalty. sonance - SEM - Online brand image is positively related to online risk and online loyalty, but does not affect online loyalty. - Theory of op- - n = 290 re- Navigation performance positively impacts flow and timal flow spondents attitude toward the website. - SEM - Incongruity and the interaction between navigation performance and incongruity negatively affect flow. - Flow is positively associated with attitude toward the website and revisit intentions. - Attitude toward the website enhances revisit intentions. - The relationship between navigation performance and flow is weaker when incongruity is higher. - Indirect effect is weaker at higher levels of incongruity - Categoriza- - n = 290 respond-- Offline and online attitudes are positively related. tion theory ents - Website performance directly affects offline attitudes. - SEM - Offline attitudes relate more strongly with online attitudes, and website performance relates more weakly with online attitudes at higher versus lower levels of perceived retailer-website congruity. - Stimulus-or- - n = 626 respond-- Store layout design influences emotional arousal and ganism-reents attitude toward the website. sponse - SEM - Atmosphere affects emotional arousal. framework - Emotional arousal enhances attitude toward the website and purchase intention - Attitude directly affects purchase intention. Effects on overall retailer outcomes - None - n = 1,056 re- Awareness positively affects customer service, prodspondents (582 uct quality, pricing policy and physical store layout. grocery and 474 - Customer service enhances retailer trust. interior design) - Product quality, pricing policy, physical store layout - SEM and retailer trust influence loyalty. - Categoriza- - n = 1,478 con- - Website image positively influences a retailer’s overtion theory sumers all image for offline and multi-channel buyers. - SEM - Store image positively affects a retailer’s overall image for offline, multi-channel and online buyers. - Store and website image positively impact the perceived congruence between channels. - Congruence between channels influences retailer image for multi-channel and online buyers. - None - n = 220 con- - Service, price and convenience are important drivers sumers Dutch of satisfaction. grocery sector - For major shoppers, treat service factors are less imfrom 2009 to portant to their satisfaction. 2014 - Convenience is the most important driver of satisfac- Regression tion for regular fill-in shoppers. analysis - Price most importantly impacts satisfaction for fill-in trips related to special occasions.

1.2 Research Gaps and Literature Review

Jinfeng and - How does store image - None Zhilong influence retail brand (2009) equity and retailer loyalty?

Murray et al. (2017)

Herhausen et al. (2015)

-

-

11

- n = 6,530 respondents - SEM

- Retailer associations, awareness and perceived quality positively affect loyalty. - Convenience influences associations, awareness and perceived quality. - Institutional factors affect both associations and awareness. - Physical facilities, perceived price and employee services affect associations and perceived quality. - Each of the five image dimensions enhances loyalty. Are there differences - Information - n = 453 respond-- Store novelty positively affects both store complexity in loyalty building for processing ents (228 new and store design pleasure. new and established theory store design and - Store complexity impacts store design pleasure. store designs? - Perceptual 225 older store - Store design pleasure affects retail brand loyalty. fluency thedesign) - Retail price enhances retail brand loyalty. ory - SEM Effects on offline channel, online channel and overall retailer outcomes How does online-of- Technology - n = 107 re- Online-offline channel integration directly increases fline channel integraadaption the- spondents the perceived service quality of the online store. tion affect multi-chanory (Study 1) - Perceived service quality of the online store influnel outcomes (search - Diffusion the- - n = 129 reences overall and online outcomes. intention, purchase in- ory spondents - Online-offline channel integration has an indirect eftention and willingness (Study 2) fect on overall and online outcomes via the perceived to pay)? - n = 138 reservice quality of the online store. Do perceived online spondents - No significant mediation effects of perceived online service quality and (Study 3) risk on multi-channel outcomes emerge. perceived online risk - Regression - The direct and indirect effects of online-offline chanserve as mediators? analysis nel integration are negatively moderated by the cusIs Internet store expetomers’ Internet shopping experience. rience a moderator? - Online-offline channel integration does not negatively affect the physical store, i.e., no cannibalization effects between the channels occur.

Figure 1.3 (continued)

and exerts effects on the retailer and online outcomes. Online-offline integration does not negatively affect the physical store and no cannibalization effects between channels occur. Additionally, the authors provided empirical evidence for a negative moderation of the direct and indirect effects of the online-offline integration by consumers’ Internet shopping experiences.

1.2.2.2 Effects of Bidirectional Channel Relationships Only few studies have analyzed bidirectional effects that consider the crosschannel effects of two channels on different outcomes in omni-channel retailing. Figure 1.4 illustrates that effects of bidirectional relationships are only shown for online and overall retailer outcomes and that few studies have analyzed bidirectional effects on offline and online channel outcomes. First, regarding online channel outcomes, Badrinarayanan et al. (2012) analyzed whether attitude and trust transfer from offline to online stores, while considering effects of channel congruence between offline and online channel images on attitude, trust and online purchase intentions. They confirmed positive transfers from offline trust to online trust. Image congruence strengthens attitudes towards online stores and online trust, while both attitudes towards online stores

12

1

Introduction

and online trust enhance online purchase intentions. Moreover, positive effects of offline image dimensions on online image dimensions were shown for multichannel retailers by Verhagen and van Dolen (2009). In a recent study, Verhagen, van Dolen and Merikivi (2019) reported positive transfer effects of consumers’ offline store personnel perceptions that transferred to consumers’ evaluations of online store usefulness and enjoyment, which ultimately increased online store valuation. Second, regarding overall retailer outcomes, a study by Swoboda et al. (2016) is the only study that has explicitly focused on reciprocal effects. Their results showed cross-channel effects between consumers’ offline (online) brand beliefs and online (offline) retail brand equity. The total effects of offline (vs. online) brand beliefs on loyalty are stronger, whereas they are generally stronger for the fashion (vs. grocery) sector. Stronger total and indirect effects were found depending on strong prior offline and online performance of retailers. This study moreover reported positive reciprocal relationships between offline and online retail brand equity, whereas the total effects of offline retail brand equity affected loyalty more strongly, and were generally more stronger in the fashion (vs. grocery) sector. Third, regarding offline and online channel outcomes, Breugelmans and Campo (2016) analyzed the effects of offline and online channel outcomes and found cross-channel effects, that differed between more and less loyal consumers. Kwon and Lennon (2009a) analyzed the effects of online and offline brand beliefs, brand attitudes, and purchase intentions depending on retailers’ prior offline brand images. The authors suggested reciprocal relationships, but empirically only showed cross-channel effects of online (offline) brand beliefs on offline (online) brand attitudes. Offline and online brand attitudes determine online purchase intentions. The effects, moreover, depend on retailer’s prior offline brand image. In summary, the majority of studies have supported positive transfers of consumer perceptions from offline to online stores and vice versa, but have predominantly focused on unidirectional or bidirectional channel relationships and neglected prior experiences. Therefore, the above reviewed studies leave open questions regarding effects of reciprocity in omni-channel retailing. These research gaps will therefore be presented in Section 1.2.5.

1.2 Research Gaps and Literature Review

Author(s) and year

Research question

Badrinarayanan et al. (2012)

-

Verhagen and van Dolen (2009)

Verhagen, van Dolen and Merikivi (2019)

Yang et al. (2013)

-

Swoboda, Weindel and SchrammKlein (2016)

-

Breugelmans and Campo (2016)

-

Kwon and Lennon (2009a)

-

Verhoef, Neslin, et al. (2007)

-

-

13

Theory/ Sample and Core findings framework method Effects on online channel outcomes Do transfers between - Schema the- - n = 533 young - Offline trust of a multi-channel retailer transfers to online channels exist? ory adults students trust, while no transfer effects for attitude are found. Does congruence be- - Categorizafrom South Ko- - Image congruity strengthens the attitudes toward the tween offline and online tion theory rea and the US online store and the online trust. image affect online trust - Theory of - SEM - Attitude toward the online store and online trust enhance and attitude? reasoned aconline purchase intentions. tion How does multi-channel - None - n = 630 re- Offline service, offline merchandise, offline atmosphere, store image influence spondents and store layout have a positive impact on their online online purchase inten- SEM counterparts. tions? - Offline merchandise, online merchandise, online store atmosphere, and online store navigation positively influence online purchase intentions. How do perceptions of - Analogical - n = 269 re- Offline store personnel competence positively affects offline store personnel knowledge spondents customer perceptions of online store usefulness. influence online store transfer - PLS - Offline store personnel friendliness positively influences usefulness, online store customer perceptions of online store enjoyment and enjoyment and online online store usefulness. store valuation? - Usefulness and enjoyment improve value for customers. What are the effects of - Brand exten- - n = 308 cus- Effects of synergies and dissynergies across channels cross-channel synersion theory tomers exist simultaneously during the process of channel exgies and dissynergies - Expectation- - SEM tension. on evaluation and cusconfirmation - Both, the offline channel service quality and the confirtomer’s online channel theory mation of the offline channel service performance affect adaption behavior? online channel extension decisions. Effects on overall retailer outcomes Whether and how do - Theory of - n = 271 re- Consumers’ offline (online) brand beliefs positively affect crosswise and reciproreasoned ac- spondents online (offline) brand equity. cal relationships among tion fashion sector - The total effects of offline brand beliefs on loyalty are offline and online brand - Theory of - n = 274 restronger than those of online brand beliefs in fashion and beliefs, retail brand eqcognitive dis- spondents gro- grocery, while the total effects are stronger in fashion. uity and loyalty exist? sonance cery sector - For retailers with a strong prior offline and online perforHow do prior channel - SEM mance, the total effects of offline and online brand beperformance and the liefs on loyalty and the indirect effect of online brand beretail context (fashion liefs on loyalty via offline brand equity are stronger. vs. grocery) affect these - Offline and online brand equity are reciprocal related, relationships? whereas the total effects on loyalty are stronger for offline than for online brand equity. - The total effects of offline and online brand equity on loyalty and the reciprocal effects between offline and online brand equity are stronger in fashion retailing. Effects on offline channel and online channel outcomes How do the cross- None - n = 9,251 - Offline (online) price promotions affect category purchannel effects of households chase decisions in the offline (online) channel. price promotions affect from 2006 to - A high frequency of offline (online) price promotions negcategory purchase de2017 atively impacts promotion effectiveness in the offline cisions? - Regression (online) channel in subsequent periods. analysis - Online (offline) price promotions negatively affect category purchase decisions in the offline (online) channel. - A similar price promotion in the online channel has a stronger negative effect on category purchase behavior in the offline channel than vice versa. - A high frequency of online (offline) price promotions negatively influences promotion effectiveness in the offline (online) channel in subsequent periods. - Own- and cross-channel promotion effects differ between loyal and non-loyal customers of the chain. Do related offline and - Summative - n = 630 stu- Online brand beliefs are more favorable when the reonline brand images af- model of atti- dents tailer’s prior offline brand image is strong (vs. weak). fect offline and online tude (study 1) - Consumers’ online and offline brand beliefs affect online purchase intentions, de- - Theory of n = 650 stuand offline brand attitudes toward multi-channel retailers. pending on prior offline reasoned ac- dents (study 2) - Consumers’ online and offline brand attitudes positively brand image and online tion - (M)ANOVA affect online purchase intentions. performance? - Theory of - SEM cognitive dissonance How do consumers use - Theory of - n = 396 Dutch - Internet to store research shopping is the most common channels for search and reasoned consumers form of research shopping. purchase? action - Regression - Internet to store research shopping can be reduced eiWhat drives research analysis ther by an attribute approach, by managing lock-in, or by shopping? - Choice models managing cross-channel synergy.

Figure 1.4 Studies on bidirectional channel relationships. (Source: Own creation)

14

1.2.3

1

Introduction

Channel Integration Services

1.2.3.1 Effects of Joint Channel Integration Services Firms use channel integration services. These are defined as technology-based solutions that allow customers to execute different interactive activities across physical and online channels, for example, regarding information, transactions or fulfillment (Banerjee 2014; Sousa and Amorim 2018). Most studies have focused on the joint effects of integration services, which reflects a holistic view of channel integration (e.g., Hamouda 2019; Shen et al. 2018). As seen in Figure 1.5, a broad range of scholars has analyzed direct and, more often, indirect effects of joint channel integration services on different outcomes. First, regarding direct effects of joint integration services, Cao and Li (2015), for example, showed positive effects on firm sales growth. These effects were negatively moderated by firms’ online experiences and a larger physical store presence. Seck and Philippe (2013) indicated that physical and virtual service quality as well as multi-channel integration affect the overall satisfaction of the multi-channel customer. The authors provided positive evidence for the proposed relationships. Tagashira and Minami (2019) report direct effects of joint integration services on cost efficiency as a proxy for firm performance. The extent of a retailer’s experience with e-commerce and the level of face-to-face services reduces the effectiveness of the services on firm performance. Second, concerning the indirect effects of joint integration services, various mediators that translate integration services into consumer behavior have been studied. Bendoly et al. (2005) showed effects on the availability of channel choice and, as a consequence, less consumer switching to channels from competing firms. Frasquet and Miquel (2017) and Hamouda (2019) provided evidence for positive effects of joint integration services on offline and online loyalty mediated by consumer satisfaction. Li et al. (2018) examined the processes by which uncertainty, identity attractiveness, and switching costs of omni-channel retailers influenced consumer reactions to cross-channel integration. The authors pointed to positive effects of cross-channel integration on retailer uncertainty, identity attractiveness and switching costs. Retailer uncertainty negatively affects consumer retention, but exerts a positive influence on interest in alternatives. Identity attractiveness, and switching costs positively affect consumer retention, while switching costs negatively affects interest in alternatives. Moderator showrooming decreases the link between integration and retailer uncertainty.

1.2 Research Gaps and Literature Review

15

According to Lee et al. (2019), joint integration services enhance repurchase intentions and word-of-mouth through customer engagement. Their joint perspective included breadth of channel-service choice, transparency of channel-service configuration, content consistency, and process consistency. Comparing purchases for high-involvement vs. low-involvement products, the authors confirmed that integrated interactions have a stronger effect on customer engagement compared with channel-service configurations. For low-involvement products, the breadth of channel-service choice is more important to consumers. Oh et al. (2012) discussed how joint integration services affect firms’ exploitative and explorative competences. They found that these competences serve as mediators and enhance firm performance. Moreover, they found that cross-channel human resource capability positively moderates the relationship between channel integration capability and exploitative as well as explorative competence. Environmental dynamism also positively moderates the effects of innovation ability on performance. Shen et al. (2018) studied antecedences of omni-channel usage. The authors reported that consumers’ usage of omni-channel systems is indirectly affected by joint integration services through perceived fluency. They also investigated moderations and revealed that internal usage experience weakens the effect of perceived fluency on omni-channel service usage, whereas external usage experience enhances the effect of perceived fluency on omni-channel service usage. Zhang et al. (2018) proposed mediating effects of consumer empowerment between consumers’ positive responses to joint channel integration services and perceived trust as well as satisfaction. Joint channel integration services exert positive impacts on purchase intention towards a retailer indirectly through trust and satisfaction.

1.2.3.2 Effects of Offline-Online Channel Integration Services Few studies have focused on the effects of OF-ON services that support the ability of consumers in offline venues to interact with an online channel. The studies in Figure 1.6 showed direct effects of different services (e.g., support services for ordering articles that are physically not available or information integration). Regarding the direct effects of OF-ON services, Bhargave, Mantonakis and White (2016) analyzed whether consumers react to a presented reminder of the availability of online information in offline purchasing situations. They support a cue-of-the-cloud effect that enhances purchase intentions and consumers’ choice behaviors in offline stores. The amount of information available in the environment moderates the relationship, such that in low-information environments, decreased effects on offline purchases take place. Collier and Kimes (2013) provided evidence for the positive effects of using a self-service technology on trust and satisfaction, whereas Mosquera et al. (2018) showed positive effects of different interactive in-store technologies on offline purchase intentions.

16

1

Author(s) and year

Research question

Cao and Li (2015)

- Does cross-channel integration enhance firm sales growth? - Do firm-level factors moderate the effects of cross-channel integration on sales growth? - Do IT assets directly and indirectly affect cross-channel capabilities and managerial actions? - Which moderating effect do financial resources have on these relationships?

Luo, Fan and Zhang (2015)

Melis et al. (2015)

- Does the integration of the multi-channel retail mix impact online grocery store choice? - Whether do these drivers change when shoppers gain online experience?

Seck and Philippe (2013)

- How do physical and virtual service quality and multi-channel integration quality affect overall satisfaction? - What are the effects of cross-channel integration on firm performance (i.e., cost efficiency)?

Tagashira and Minami (2019)

Bendoly et al. (2005)

- What is the role that perceptions of channel integration have on risks beliefs and on purchase decisions?

Chiu et al. (2011)

- Whether do online and vicarious experience, switching costs, multichannel integration, service quality of competitors offline store and perceived risk of purchasing online deter consumers’ cross-channel free-riding?

Introduction

Theory/ Sample and Core findings framework method Direct effects on various outcomes - Grounded - n = 91 retailers, - Cross-channel integration positively affects firm sales theory longitudinal growth. data from 2008 - Firms’ online experience negatively moderates the effect to 2011 of cross-channel integration on firm sales growth. - Regression - Larger physical store presence negatively moderates analysis the effect of cross-channel integration on firm sales growth. - Resource- n = 49 U.S. ap- - Quantity and scope of investments in enterprise IT applibased view parel retailers, cations positively affect cross-channel capabilities. longitudinal - Theory of IT - Financial resources positively moderate the relationship data from 1995 between enterprise IT applications and cross-channel business to 2007 value capabilities. - Regression - Theory of - Enterprise IT applications increase the frequency and analysis competitive broadens the types of managerial actions. dynamics - Market-oriented capabilities such as e-commerce and multi-channel cross-selling capabilities broaden managerial actions and operation-oriented capabilities. - Cost-benefit - n = 7,907 - A strong integration of online and offline store leads conframework households sumers, who start to buy groceries online to choose the (4,069 multionline store that belongs to the same chain as their prechannel and ferred offline store. 3,838 single- With an increase of online shopping experience, multichannel) channel shoppers’ focus moves from comparing within a - Multinomial chain across channels to comparing stores across logit analysis chains within the online channel. - High online experience leads to an increase of the importance of online assortments, especially in terms of assortment attractiveness, and online loyalty - None - n = 445 con- Perceived physical service quality and perceived virtual sumers service quality have a positive influence on the overall - SEM satisfaction of the multi-channel customer. - Multi-channel integration quality has a positive effect on the overall multi-channel customer satisfaction. - None - Stochastic fron- - Cross-channel integration impacts cost efficiency. tier analysis - The extent of a retailer's experience with e-commerce - Dynamic panel and the level of face-to-face services negatively modermodel ates the relationship between cross-channel integration and cost efficiency. Indirect effects on various outcomes - Mental ac- n = 1,598 re- The greater options of channel integration are, the counting the- spondents smaller the impact of availability on channel selection ory - ANOVA with regard to those options. - Logistic regres- - Given an availability failure, the greater the options of sion analysis channel integration are, the less likely customers will seek out channel options provided by competing firms - Push-pull- n = 322 respond-- Multi-channel self-efficacy enhances cross-channel freemooring par- ents riding intentions. adigm - SEM - Service quality of competitors’ offline store and reduced risk in the offline channel affect the attractiveness and increase cross-channel free-riding intentions. - Perceived risk of purchasing online increases the attractiveness of competitors’ offline retail stores. - Attractiveness of competing offline retail stores increases cross-channel free-riding intentions. - Firms can avoid cross-channel free-riding through the development of within-firm lock-ins. - Channel integration has no significant effect on channel lock-in and does not prevent free-riding intentions.

Figure 1.5 Studies on effects of joint channel integration services. (Source: Own creation)

1.2 Research Gaps and Literature Review

Frasquet and Miquel (2017)

Hamouda (2019)

Hossain et al. (2020)

- How is multi-channel in- - None tegration be measured? - Does multi-channel integration impact offline and online loyalty, both directly and by mediation of customer satisfaction? - How are omni-channel - None integration quality, omni-channel perceived value, customer satisfaction and loyalty related?

- What are the dimensions and subdimensions of omni-channel integration quality and its effect on perceived value?

- Dynamic capabilities theory

Li et al. (2018)

- How do uncertainty, - Push-pullidentity attractiveness, mooring parand switching costs me- adigm diate customer reactions to cross-channel integration? - Does the moderator showrooming interact with these processes? Lee et al. - What are the effects of - Social ex(2019) channel integration change thequality on customer en- ory gagement, repurchase intention and word-ofmouth? - Are there differences in effects between purchasing contexts of products? Oh and Teo - How do aspects of - Customer (2010) physical and virtual re- contact tail channels that can model be integrated create value for consumer?

Oh, Teo and - What is the impact of - Resource Sambaretail channel integrabased view murthy tion capability on firm’s - Learning the(2012) explorative and exploiory tative competences and firm performance? - Do cross-channel human resource capability and environmental dynamism moderate these relationships?

Figure 1.5 (continued)

17

- n = 761 multi- - A multi-channel integration scale is developed, whereas channel fashion channel integration has two dimensions: reciprocity and shoppers coordination. - SEM - Multi-channel integration equally affects both offline and online loyalty directly. - Customer Satisfaction serves as a partial mediator for the relationship between multi-channel integration and offline and online loyalty. - n = 395 banking - Omni-channel integration quality increases the percustomers ceived value of the omni-channel by the customer. - SEM - Omni-channel perceived value in turn positively affects customer satisfaction, as well as customer loyalty. - Omni-channel integration quality and customer satisfaction have a positive relationship. - Customer loyalty is positively linked to customer satisfaction within omni-channel banking. - n = 18 in-depth - Omni-channel integration quality consists of four primary interviews dimensions (channel-service configuration, content con- n = 301 banking sistency, process consistency and assurance quality) customers and ten sub-dimensions. - PLS - Cross-buying intentions are a behavioral outcome of omni-channel integration quality. - Cross-buying intentions is a partial mediator between omni-channel integration quality and perceived value. - n = 259 respond-- Cross-channel integration positively impacts retailer unents certainty, identity attractiveness and switching costs. - SEM - Retailer uncertainty negatively affects customer retention, but positively the interest in alternatives. - Identity attractiveness and switching costs have positive effects on customer retention, while switching costs negatively affect interest in alternatives. - Showrooming negatively moderates the relationship between cross-channel integration and retailer uncertainty. - n = 269 respond-- Channel integration quality dimensions (including ents high-involve breadth of channel-service choice, transparency of ment products channel-service configuration, content consistency, and (Study 1) process consistency) influence customer engagement. - n = 221 respond-- Customer engagement leads to positive word-of-mouth ents low-involve- and enhances repurchase intention. ment products - Integrated interactions compared with channel-ser(Study 2) vice configuration exert a stronger influence on cus- SEM tomer engagement for high-involvement products, than vice versa for low-involvement products. - n = 300 consum-- Integration of promotion, product and pricing inforers mation, as well as transaction information enhances - PLS information quality. - Integration of information access, order fulfillment, and customer service increases service convenience. - Information quality and service convenience influence consumer value. - n = 125 multi- - Channel integration capability positively influences firm channel retailers exploitative competences (i.e., ability to improve the effi- PLS ciency of current operations) and explorative competences (i.e., the ability to provide new services). - Firm’s exploitative and explorative competences positively influence firm performance. - Cross-channel human resource capability positively moderates the relationship between channel integration capability and exploitative as well as explorative competence. - Environmental dynamism positively moderates the effects of innovation ability on performance.

18

SchrammKlein et al. (2011)

- How do channel evaluations and channel integration affect customer loyalty and usage of a multi-channel system mediated by trust and image? Shen et al. - Is there a link between (2018) channel integration quality, perceived fluency and omni-channel usage? - Do internal and external usage experience moderate this link? Yang et al. - What factors influence (2017) consumer repurchase intention in an onlinecum-mobile retail context? - What role does channel integration play in this context? Zhang et al. - How does channel inte(2018) gration affect consumers’ intention to purchase through perceived trust and satisfaction about their shopping experiences?

1

Introduction

- None

- n = 981 custom- - The evaluation of channel integration positively impacts ers trust, and stronger image. - SEM - Retailer’s image strongly affects customer loyalty and the differentiated use of the multi-channel system. - A relatively small, but still significant, impact of customer’s trust toward the retailer on customer loyalty and differentiated use of the multi-channel system is found. - Wixom & - n = 401 re- Channel integration quality significantly affects perceived Todd framespondents fluency across different channels. work - SEM - Omni-channel usage is indirectly affected by channel integration quality through perceived fluency. - Internal usage experience weakens the effect of perceived fluency on omni-channel service usage. - External usage experience enhances the effect of perceived fluency on omni-channel service usage - None - n = 317 consum-- Channel integration has a strong and positive effect on ers service quality perceptions in both online and mobile en- SEM vironments - The quality perceptions influence transaction-specific satisfaction and cumulative satisfaction. - Transaction-specific satisfaction has a positive influence on cumulative satisfaction, and both of them in turn positively affect repurchase intention. - Stimulus-or- - n = 173 respond-- Significant mediating effects of consumer empowerment ganism-reents between consumers’ positive responses to channel intesponse - SEM gration and perceived trust as well as satisfaction framework emerge. - Perceptions of channel integration have an influence on purchase intention towards a retailer through perceived trust and satisfaction.

Figure 1.5 (continued)

1.2.3.3 Effects of Online-Offline Channel Integration Services ON-OF services support the ability of consumers in online venues to interact with an offline channel. The studies in Figure 1.7 addressed, for example, services that support the pickup or return of articles purchased online at offline stores. Both direct and indirect effects of ON-OF services were shown on different outcomes. First, with regard to the direct effects of ON-OF services, Akturk et al. (2018) examined the consequences for retailers that introduced ship-to-store services. The authors demonstrated that online sales decreased after the introduction of ship-to-store, although offline store sales increased due to increased cross-channel consumer retention. Both, Gallino and Moreno (2014) and Gao and Su (2017) studied the consequences of services that allow consumers to buy online and pick up in-store. Their results showed lower online sales, higher store sales, and higher store traffic due to these click-and-collect services. Retailers can reach new customers with click-and-collect services, but for existing customers, the shift from online fulfillment to store fulfillment may decrease profit margins. Second, with respect to indirect effects, Herhausen et al. (2015) revealed that ON-OF services affect multi-channel outcomes through perceived online service quality. Moreover, consumers’ Internet shopping experiences negatively moderate

1.2 Research Gaps and Literature Review

Author(s) and year

Research question

Bhargave, Mantonakis and White (2016)

- How does a reminder of the availability of online information for consumers in offline purchasing situations impacts offline purchase intention? - Does the amount of information in the environment acts as a moderator?

Collier and Kimes (2013)

- How does convenience of using a self-service technology influence satisfaction and trust customers place in the technology via exploration, accuracy, and speed of transaction?

Mosquera et - What is the effect of the al. (2018) intention to use different interactive technologies in a fashion store on purchase intention? - What is the moderating role of gender in this relationship?

19

Theory/ Sample and Core findings framework method Direct effects on various outcomes - Metacogni- - n = 438 con- When consumers are presented with a reminder of the tive theory sumers availability of online information in offline purchasing situ(Study 1) ations, a cue-of-the-cloud can enhance purchase inten- n = 164 contions and choice behaviors. sumers - The cue increases consumers’ confidence in being able (Study 2) to retain and access the information seen in-store, which - n = 133 conengenders positive feelings about the decision to pursumers chase. (Study 3) - The cue-of-the-cloud decreases offline purchases in a - n = 251 conlow vs. high information environment sumers (Study 4) - ANOVA - Regression analysis - Resource - n= 214 re- Convenience influences the perceived accuracy, speed, matching spondents and exploration intentions of a self-service technology. theory (Study 1) - Self-service technology users have a lower need for in- n = 228 reteraction when they are satisfied with a technology. spondents - Nonusers’ trust perceptions has the greatest influence (Study 2) on the need for human interaction. - n =242 re- Satisfaction can be enhanced by focusing on the speed spondents and accuracy of a self-service technology, whereas non(Study 3) users’ perceptions of accuracy and exploration in- SEM creased the trust placed in a self-service technology. - None - n = 628 omni- - The intention to use in-store technology, the intention to channel cususe fitting-room technology and the intention to use the tomers customer’s own technology positively affects offline pur- PLS chase intention. - In-store technology is the strongest predictor of offline purchase intention, followed by own technology use. - No significant differences were found between female and male consumer groups.

Figure 1.6 Studies on effects of offline-online channel integration services. (Source: Own creation)

the direct and indirect effects of the ON-OF services. Murfield et al. (2017) analyzed the effects of click-and-collect and buy-in-store-ship-direct on consumer satisfaction and loyalty. Consumer satisfaction partially mediates the relationship between click-and-collect and loyalty and fully mediates the relationship between the services’ timeliness and loyalty. For buy-in-store-ship-direct, consumer satisfaction partially mediates the relationship between the services’ timeliness and consumer loyalty. Finally, Yang et al. (2020) analyzed how informational onlineto-offline integration affects consumer use intention of online to offline commerce. The authors reported significant impacts of informational online-to-offline integration on consumer use intention of online to offline commerce. This relationship is mediated by perceived benefit, perceived usefulness, and perceived value. As a result, studies focused on joint effects of channel integration services or dealt with OF-ON or ON-OF services only. More general channel evaluations that translate the services into consumer behavior were studied, which failed to reveal cross-channel effects. Moreover, only initial results on context factors

20

1

Author(s) and year

Research question

Akturk, Ketzenberg and Heim (2018)

- Does the introducing of ship-to-store impact a retailer’s operating performance, in terms of sales and customer returns?

Gallino and Moreno (2014)

- What are the effects of buy-online-pickup-instore on online sales, store sales and store traffic? Gallino, - How do cross-channel Moreno and functionalities influence Stamatosales dispersion? poulos - Which implications (2017) emerge for inventory management? Gao and Su - For what types of prod(2017) uct will buy-onlinepickup-in-store be profitable? - How does buy-onlinepickup-in-store impact the customer base? - How should the integration services’ revenue be allocated between offline and online channels? Jara et al. - What are the key fac(2018) tors in explaining the long-term value creation for click-and-collect systems depending on consumer profiles? Vyt et al. - Does a grocery pickup (2017) system create value for consumers? - Can spatial development of grocery pickup increase retailer’s territory coverage?

Theory/ Sample and Core findings framework method Direct effects on various outcomes - Transaction - n = 6,500 con- - Online sales decreased after ship-to-store was introcost theory sumer purduced, although offline store sales increased. chase/return - After ship-to-store implementation, some customers transactions at switched from the online channel to the offline channel. jewelry retail- - This switch occurred mainly for high-value purchases. er's online and - The customers who actually remained with and fully offline stores, completed a sale using the ship-to-store service typically from August 1, were those that bought low-value items. 2009 to July - Introducing ship-to-store increased cross-channel cus31, 2013 tomer returns of online purchases to physical stores. - Logit regres- - These new ship-to-store returns generated additional ofsion analysis fline store sales. - None - n = 210 DMAs - Buy-online-pick-up-in-store results in lower online - n = 83 stores sales, higher store sales, and higher store traffic. - Regression - The influence occur due to cross-selling effects from analysis customers who go to the store to pick up items they purchased using this channel integration service. - None - n = 158,663 - Increasing sale dispersions emerged (the contribution of Units from the 90 % lowest-selling products to total sales increased 2011 to 2012 by 0.75 percentage points). - Regression - To respond optimally to the observed increase in disperanalysis sion, retailers need to increase its cycle and safety inventories by approximately 2.7 %. - None - Stylized model - Buy-online-pickup-in-store affects customer choice by providing real-time information about inventory availability and by reducing the hassle cost of shopping. - Not all products are well suited for in-store pickup (specifically not for products that sell well in-stores). - Buy-online-pickup-in-store enables retailers to reach new customers, but for existing customers, the shift from online fulfillment to store fulfillment may decrease profit margins when the latter is less cost effective. - In decentralized retail systems, buy-online-and-pick-upin-store revenue can be shared across channels to alleviate incentive conflicts. - None - n = 479 re- Customers’ relations, website and pickup station are the spondents most important factors creating value for customers - SEM whatever the click-and-collect system (drive-out, drive-in or in-store picking model). - Customers’ benefits vary regarding the kind of click-andcollect model and the age of customers. - None - n = 1,576 hy- Long term value exist through the satisfaction of funcpermarkets with tional, experiential and relational benefits responsible for 1,473 grocery the uniqueness of the retailer’s positioning. pickups - Grocery pickup locations lead to closeness in retailing. - Relative en- Grocery pickups are not deployed in strategic locatropy model tions to improve retail chain global territory coverage but are mostly developed in overstored areas. Indirect effects on various outcomes -

Herhausen - See Figure A—3 et al. (2015) Murfield et - What are the effect of - None al. (2017) buy-online-pickup-instore, and buy-instore-ship-direct on consumer satisfaction and loyalty? - How do the effects differ across both contexts? Yang et al. (2020)

Introduction

- n = 323 respondents (Study 1) - n = 184 respondents (Study 2) - PLS

- Do physical experi- Perceived - n = 416 reence and integration value theory spondents of online and offline in- - Technology - SEM formation affect the acceptance use of online to offline model commerce?

- The three dimensions of each integration service (condition, availability, and timeliness) are distinct in their impact on satisfaction and loyalty. - In the buy-online-pickup-in-store context, consumer satisfaction partially mediates the relationship between condition and loyalty and fully mediates the relationship between timeliness and loyalty. - In the buy-in-store-ship-direct context, consumer satisfaction partially mediates the relationship between timeliness and consumer loyalty. - Both physical experience and integration of online and offline information have significant impacts on consumer use intention of online to offline commerce. - The effects are mediated by perceived benefit, perceived usefulness, and perceived value.

Figure 1.7 Studies on effects of online-offline channel integration services. (Source: Own creation)

1.2 Research Gaps and Literature Review

21

have been provided. Therefore, these studies have provided limited findings for the mediation paths of the offered but distinctly perceived integration services on offline and online outcomes and important context factors. The questions arising from synthesizing these gaps in the literature will therefore be presented in Section 1.2.5.

1.2.4

Online- and Omni-channel-specific Marketing Instruments

1.2.4.1 Effects of One or Two Marketing Instruments Marketing instruments are strategically applied by retailers in online stores to attract consumers (e.g., Bressolles et al. 2015; Toufaily and Pons 2017). Especially for omni-channel retailers, which offer seamless experiences across various channels, these instruments are of primary concern for online and cross-channel for offline consumer behavior. As seen in Figure 1.8, most research has focused on one or two marketing instruments, while different mediators, most often online trust and more seldom brand associations, have been conceptualized between effects of marketing instruments on consumer behavior. First, regarding different mediators, Al-Qeisi et al. (2014), for example, sought to determine which elements of website design impact quality. They explored how different elements affect usage behavior, as mediated by performance expectancy and experience. The authors found that website design quality consists of technical, general, and appearance quality. Considering the instrument of website design quality, the authors provided evidence for a positive effect on usage behavior through performance expectancy. Performance expectancy mediates the effect of effort expectancy on usage behavior. Experience was moreover shown to have positive effects on effort expectancy, performance expectancy, and website design quality perceptions. Emrich et al. (2015) showed that multi-channel assortment integration affects patronage intentions, mediated through perceived variety, risk, and convenience. Fang et al. (2016) questioned whether e-service quality, sacrifice and product quality affect repurchase intention mediated through perceived value. The authors demonstrated that e-service quality and product quality positively affect perceived value and repurchase intentions. Customer value in turn enhances repurchase intentions. The effects change depending on age and gender, and are different for task-focused vs. experiential shoppers. Moreover, the digital presence of service employees was found to affect website quality, which in turn enhances customer loyalty (Herhausen et al. 2020). Wen and Lurie (2019) considered the effects of visual boundaries on perceived variety under high and

22

Author(s) and year

1

Introduction

Research question

Theory/ Sample and Core findings framework method - Effects with different mediators Al-Qeisi et - Which elements of web- - Unified the- - n = 216 re- Technical, general and appearance quality build website al. (2014) site design are imory of acspondents design quality, which positively affects usage behavior. portant for quality? ceptance - SEM - Performance expectancy has a mediating role on the ef- How do these elements and use of fect of effort expectancy on usage behavior. affect usage behavior technology - Website design quality has an indirect effect on usage through performance behavior through performance expectancy. expectancy and experi- Experience affects effort expectancy, performance exence? pectancy and website design quality perceptions. Barrutia and - How do consumer re- - Resource- n = 472 re- Process quality represented by efficiency, system Gilsanz sources and firm readvantage spondents availability, design and information as well as out(2013) sources enhance theory - SEM come quality are antecedents of electronic service value? quality. - Does the interaction of - Social expertise and consumer innovativeness have consumer expertise and a significant positive effect on consumer expertise. electronic service qual- Consumer expertise affects value perception. ity have a moderating - Electronic service quality affects value perception. effect on value? - The positive effect of consumer expertise on value is weaker as electronic service quality is enhanced. Bertrandie - Whether and how does - None - n = 594 re- For external assortment integration, choice difficulty and Zielke a retailer’s assortment spondents is the highest for no integration. (2017) integration affects con- SEM - For internal assortment integration, asymmetrical (vs. sumer overload confufull) integration is a superior assortment perception. sion, assortment per- Choice overload affects purchase abandonment, attiception and consumer tude towards the retailer and patronage intentions. behavior? - Effects of choice overload on purchase abandonment, attitude towards the retailer and patronage intentions are mediated by assortment perception. Emrich, Paul - How does multi-channel - Negativity - n = 959 re- Full integration dominates no integration, but asymmetand Rudolph assortment integration bias theory spondents rical integration has impacts for substitutive relations (2015) affect patronage inten(study 1) compared with no integration. tions through shopping - n = 2,005 re- For independent relations, asymmetrical integration can benefits? spondents be more beneficial than full integration, while customer (study 2) outcomes differ less for complementary relations. - SEM Fang et al. - How does e-service - Meands-End - n = 651 re- E-service quality and product quality influence perceived (2016) quality, sacrifice and Chain Thespondents value and repurchase intention. product quality affect re- ory - SEM - Customer value positively affects repurchase intention. purchase intention via - The effects change depending on age and gender, and perceived value? for task-focused vs. experiential shoppers. Herhausen - See Figure A—3 et al. (2015) Herhausen - Does digital presence of - Social pres- - n = 11,152 re- - Digital presence of service employees positively affects et al. (2020) service employees afence theory spondents perceived website service quality and perceived emfect website service - MSEM ployee service quality, which in turn affect loyalty. quality, employee ser- The relationship become stronger with more signals of vice quality and cusemployee accessibility and for firms with a stronger customer loyalty? tomer orientation. Landers et - See Figure A—3 al. (2015) Lee et al. - See Figure A—5 (2019) Mallapra- How do website charac- - Signaling - n = 773,262 - Contextual factors are associated with online browsing. gada et al. teristics influence an theory browsing ses- - Website scope in terms of product variety is associated (2016) online transaction’s sions from 385 with visit durations and basket values but negatively with basket value? websites page views. - Mixed-effects - Communication functionality is associated with basket type II tobit value for hedonic products. model Murfield et - See Figure A—7 al. (2017)

Figure 1.8 Studies on one or two examined marketing instruments. (Source: Own creation)

1.2 Research Gaps and Literature Review

Wen and - Do visual boundaries - None Lurie (2019) affect perceived variety under high and low cognitive load conditions?

Yang, Lu and Chau (2013)

Awad and Ragowsky (2008)

- n = 311 respondents (Study 1) - n = 900 respondents (Study 2) - n = 576 respondents (Study 3) - ANOVA - What factors affect a - Brand exten- - n = 308 reconsumer’s decision on sion theory spondents moving or extending - Expectation- - SEM from an offline channel confirmation to an online channel? theory

23

- Cognitive load moderates the impact of visual boundaries on perceived variety such that when cognitive load is high the effect of by-attribute (by-alternative) boundaries on increasing (decreasing) perceived variety is stronger than when cognitive load is low. - Consumer perceptions of retailer intent moderate the interaction between visual boundaries and cognitive load on perceived variety.

- Offline service quality affects online service quality, while the latter affects the intention to use the online channel. - Offline service quality impacts offline performance. - Offline channel performance negatively affects relative benefits of the online channel, whereas relative benefits of online channel affects intention to use online channel. - Online service quality influences the relative benefits of the online channel. - Effects with trust as mediator - How are online trust - Technology - n = 1,561 re- - The ability to post an opinion online affects word-of(and its antecedents), acceptance spondents mouth quality for men. word-of-mouth quality model - SEM - Responsive participation of others has an effect on (and its antecedents) word-of-mouth quality for both men and women, while and intention to shop the effect on is higher for women than men. online related? - Word-of-mouth quality has an effect on online trust for - Are this relationships men and women, but stronger for men. moderated by gender? - Ease of use affects online trust for women more. - Perceived usefulness affects online trust. - Online trust has a stronger effect on intention to shop online for women than for men. - See Figure A—4

Badrinarayanan et al. (2012) Badrinara- See Figure A—3 yanan, Becerra and Madhavaram (2014) Bansal et al. - How do previous online (2016) privacy invasion affect Internet privacy concern, and in turn trust in the website and intention to disclose information? - What role do personality traits play? Bashir et al. - How do perceived fi(2018) nancial risk and online trust affect consumer’s online buying intention? Benedicktus - Whether do consensus et al. (2010) information and brand familiarity exert independent or interactive effects on consumer perceptions across retailers that possess, or lack, a physical presence? - Is generalized suspicion a potential boundary condition for the effects of consensus information and brand familiarity?

Figure 1.8 (continued)

- Theory of - n = 367 rereasoned ac- spondents tion - SEM - Prospect theory

- None

-

- Categorization theory

-

-

- Trust affects the intention to disclose information. - Privacy concern negatively affects the intention to disclose private information. - Privacy concern negatively affects trust. - Different personality traits affect the relationship between online privacy concern, trust and the intention to disclose information. - The effects change depending on more vs. less sensitive contexts. n = 400 re- Perceived financial risk negatively influences consumspondents ers’ online buying intention and online trust, while online SEM trust enhances consumer’s online buying intention. - Online trust mediates this negative relationship. n = 118 re- The reliability and benevolence dimensions of trust and spondents purchase intentions are higher when consensus infor(Study 1) mation indicates satisfactory performance is high. n = 261 respond-- Benevolence and purchase intentions are higher for hyents (Study 2) brid firms than pure online retailers. ANOVA - Both dimensions of trust and purchase intentions are SEM higher for familiar brands than unfamiliar brands. - Consensus information and store presence interact on both dimensions of trust and purchase intentions. - Both dimensions of trust mediate the effects of consensus, physical presence, brand familiarity and their interactions on purchase intentions. - Both dimensions of trust and purchase intentions are lower when consumers have been previously deceived by another company.

24

1

Bock et al. (2012)

- What are the anteced- - Transferents of online trust and ence-based what are their effects on trust theory online purchase intentions? - How does the product type moderate the relationships of the antecedences on online trust? Casado- Whether and how do - None Aranda, seals of approval affect Dimoka and willingness to pay and Sánchezpurchase intention, meFernández diated by trust and risk? (2019) Fazal-e- Does consumer hope - Affect theory Hasan et al. affect perceived brand of social ex(2018) value, overall satisfacchange tion, trust in a brand offered by an online retailer and commitment? - How does customer goal attainment moderates these relationships? Javed and - Whether and how do af- - Social exWu (2020) ter delivery services af- change thefect customer percepory tion of satisfaction, trust, and repurchase intention? Sullivan and - How do perceived value - Mental acKim (2018) and online trust influcounting theence consumers’ online ory repurchase intention? - Equity theory

Rahimnia and Hassanzadeh (2013)

- How do the website in- formation and the design dimension impact e-marketing effectiveness mediated by online trust? Toufaily, - How do security/privacy Souiden and and social presence afLadhari fect online trust dimen(2013) sions, attitudes toward websites and word-ofmouth? Ye et al. (2020)

- Do effects of website social perception affect behavioral intentions mediated by cognitive and affective trust? - Do the effects differ for repeat vs. first-time visitors?

Baek et al. (2020)

- How do sensory, behavioral, emotional and intellectual perceptions of a virtual store brand experience impact brand equity and visit intentions online?

Figure 1.8 (continued)

- n = 853 respondents - PLS

Introduction

- Offline trust, efficacy of sanctions and website quality affect online trust. - Perceived structural assurance affects online trust more at the before-interaction phase. - Word-of-mouth affects online more at the before-interaction phase than at the initial-interaction phase. - Offline trust influences online trust at the initial-interaction phase than at the before-interaction phase. - Word-of-mouth, offline trust, and perceived efficacy have stronger effects for experience products. - n = 29 respond- - Brain areas related to reward and value expectations are ents strongly elicited when online trust signals are perceived - fMRI scans as more trustworthy. - Conjoint analy- - Brain areas related to risk, ambiguity and negative feelsis ings are more strongly activated in response to riskier online trust signals. - n = 418 re- Perceived brand value dimensions quality, price and sospondents cial influences consumer hope. - SEM - Consumer hope affects overall customer satisfaction, trust in a brand offered by an online retailer and customer affective commitment. - Customer goal attainment moderates the relationship between consumer hope and overall customer satisfaction in a purchase from an online retailer and between consumer hope and customer trust in a purchase from an online retailer. - n = 262 re- Satisfaction affects customer trust and repurchase intenspondents tion; trust is a predictor of repurchase intention. - SEM - Customer satisfaction and trust mediate the effects of after delivery services on repurchase intention, whereby customer satisfaction also mediates the link of after delivery services and trust in a retailer. - n = 312 re- Perceived risk is negatively associated with online trust spondents and repurchase intention, while online trust affects the - SEM website’s perceived usefulness and repurchase intention. - Website reputation affects quality, online trust and perceived value, the latter is affected by competitive price. - Product quality is associated with perceived value, online trust and repurchase intention. None - n = 100 re- Website content, informational and design dimensions spondents affect online trust and all affect e-marketing effective- PLS ness. - Content dimensions’ indirect impact on e-marketing effectiveness via online trust is greater than their direct impact on the e-marketing effectiveness. Theory of - n = 989 re- Perceived website social presence and perceived secuplanned bespondents rity/privacy exert strong impacts on website credibility havior - SEM and benevolence (online trust). Theory of - Website credibility and benevolence directly affect webreasoned acsite attitudes and indirectly word-of-mouth. tion - Role of social presence in developing online benevolence is more important in the case of online pure player. Social re- n = 226 re- Repeat (vs. initial) visitors exhibit a higher level of cognisponse thespondents tive, affective trust, and behavioral intentions. ory - SEM - A website with higher social perception generates a higher level of cognitive and affective trust, and these mediate behavioral intentions. - The relationship between perceived website social perception and affective trust is stronger (weaker) for repeat (first-time) visitors. - Effects with brand as mediator None - n = 250 re- Emotional and intellectual experience affect brand eqspondents uity, while sensory and behavioral experience do not af- SEM fect brand equity - Sensory and behavioral experience increase visit intention, while emotional and intellectual experience do not increase visit intention. - Brand equity and visit intention are positively related.

1.2 Research Gaps and Literature Review

Khan et al. (2019)

- How do brand trust and - Social ex- n = 414 recommitment mediate change thespondents the relationship of brand ory - SEM engagement and brand - Commitment experience with brand trust theory loyalty in the online service context?

Wu et al. (2013)

- What are the relation- - Stimulus-or- - n = 626 reships between store ganism-respondents layout design, atmossponse - SEM phere, emotional framework arousal and attitude toward the website and do they impact purchase intention?

25

- Brand engagement directly affects brand trust and brand commitment and indirectly brand commitment via brand trust, brand loyalty via brand commitment and via brand trust/commitment. - Brand experience directly affects brand trust and brand commitment; indirectly brand commitment via brand trust, brand loyalty via brand commitment and via brand trust/commitment in online service. - Store layout design impacts emotional arousal and attitude toward the website. - Atmosphere has a stronger impact on emotional arousal than store layout design. - Emotional arousal affects attitude toward the website and purchase intention. - Attitude toward the website and purchase intention are positively related.

Figure 1.8 (continued)

low cognitive load conditions. The authors provided evidence for a moderation of cognitive load on the link between visual boundaries and perceived variety and a moderation of consumer perceptions of retailer intent on the links between visual boundaries and cognitive load on perceived variety. Moreover, Yang, Lu and Chau (2013) examined the factors that affect channel switching from an offline to an online channel. Their results confirmed that offline service quality affects online service quality, while online service quality affects the intention to use the online channel. Offline service quality has a direct effect on offline channel performance, while the latter negatively affects the relative benefits of the online channel. The relative benefits of online channels are affected by online service quality and directly impact the intention to use online channels. Second, regarding trust as a mediator, Awad and Ragowsky (2008) analyzed how perceived ease of use and perceived usefulness affect the intention to shop online, while considering gender as a moderator. The authors confirmed the effects for both instruments and found that online trust has a stronger effect on intention to shop online for women than for men. Further positive effects on consumer behavior were found by Casado-Aranda et al. (2019). The authors analyzed whether and how seals of approval affect willingness to pay and purchase intentions, as mediated by trust and risk. Using fmRI scans, the authors demonstrated that brain areas related to reward and value expectations elicit strong responses when online trust signals are perceived as more trustworthy. Brain areas related to risk, ambiguity, and negative feelings are more strongly activated in response to riskier online trust signals. Moreover, customer satisfaction and trust have been shown to mediate the effects of after-delivery services on repurchase intention, whereby customer satisfaction also mediates the link between after-delivery services and trust in a retailer (Javed and Wu 2020). Analyzing how perceived value

26

1

Introduction

and online trust influence consumers’ online repurchase intentions, Sullivan and Kim (2018) highlighted that perceived risk negatively affects online trust and repurchase intentions, while online trust influences a website’s perceived usefulness and repurchase intentions. Website reputation affects perceived quality, online trust, and perceived value, and competitive price has an effect on perceived value. Moreover, Toufaily et al. (2013) investigated how security/privacy and social presence affect online trust dimensions, attitudes toward websites, and word-of-mouth. The authors confirmed that perceived website social presence and perceived security/privacy exert strong impacts on two dimensions of online trust: namely, credibility and benevolence. Website credibility and benevolence directly affect attitudes towards websites and indirectly affect word-of-mouth. The role of social presence in developing online benevolence is shown to be more important in the case of online pure players. Finally, online trust also mediates the impact of social perceptions on behavioral intentions as demonstrated by Ye et al. (2020) using cross-sectional data and a Solomon-like design that captured longitudinal effects. Third, regarding brand associations as a mediator, Baek et al. (2020) analyzed antecedents of brand equity and found that emotional and intellectual experience affect brand equity, which in turn enhances visit intentions. Khan et al. (2019) showed that brand engagement directly affects brand trust and brand commitment, which in turn enhance loyalty. Wu et al. (2013) discovered the relationships between store layout design, atmosphere, emotional arousal, and attitude toward a website and analyzed whether these constructs affect purchase intentions. Positive effects for store layout design on emotional arousal and attitude toward a website are found. Atmosphere more stronger influences emotional arousal than store layout design. Emotional arousal affects attitude toward the website and purchase intentions, while attitude and purchase intention are positively related.

1.2.4.2 Effects of More than Two Marketing Instruments Only few studies have analyzed the effects of more than two marketing instruments on different outcomes, and in the following, this literature will be presented in Figure 1.9. Different mediators, online trust, and more seldom, brand associations have been shown to mediate the impacts of the marketing instruments on different outcomes. First, regarding different mediators, effects of aesthetic appeal, ease of use, security/privacy, information, and content have been shown to influence behavior through perceived value (Bressolles, Durrieu and Deans 2015) or enjoyment (Floh and Madlberger 2013). Hsieh et al. (2014) explored how certain website dimensions affected purchase intentions, mediated through arousal, pleasure, and

1.2 Research Gaps and Literature Review

27

dominance. The authors confirmed that informativeness, navigational cues, and perceived organization positively affect perceived dominance, while entertainment enhances arousal. Arousal and perceived dominance enhance purchase intentions via perceptions of pleasure. Moreover, system, information, and service quality are reported to influence utilitarian and hedonic shopping values, which in turn increase consumer repurchase intentions (Kim et al. 2012). Second, regarding online trust as a mediator, Chen and Dibb (2010) found that website usability, security/privacy, and the quality of a website’s product information affect online trust, while website usability enhances shoppers’ website attitudes. Online trust influences attitudes toward the website and consumers’ website approach intentions. Familiarity with the website is found to moderate these relationships. Loureiro et al. (2018) studied different factors that drive website performance. The authors validated that perceived fashion website quality and technology quality are related to performance. Furthermore, social influence and past experience are shown to be related to performance expectancy, which in turn affects satisfaction and word-of-mouth behavior. Online trust is found to mediate these relationships. Moreover, Toufaily and Pons (2017) supported the positive effects of security/privacy, quality of support, personalization, and social presence on online trust, which in turn is positively associated with online loyalty. Differences between online pure players and multi-channel retailers were found, such that interactivity and personalization more strongly influence online trust for multi-channel retailers (vs. online pure players). Support quality and virtual community affect online trust more for online pure players. Third, regarding brand associations as mediators, Al-Hawari (2011) studied brand awareness and image as mediators for effects of online service quality factors on brand loyalty. The results of their study showed that e-escape and security directly affect brand awareness and brand image, while e-responsiveness affects brand image only. Both brand image and awareness enhance loyalty. Focusing on different verbal and visual design elements (e.g., linguistic style, photo functions, customer star ratings), Bleier, Harmeling and Palmatier (2019) analyzed how each design element elicits purchases through perceptions of online customer experience. They found that cognitive, affective, social and sensory experiences mediate the impact of design elements on customer purchases. Product (search vs. experience) and brand trustworthiness characteristics increase or decrease the uncertainty inherent in online shopping such that they moderate the influence of each experience dimension on purchases. In conclusion, the above-listed research has contributed substantially to the understanding of different online- and omni-channel-specific instruments. However, these studies are limited in that they cannot predict a successful orchestration

28

Author(s) and year

1

Introduction

Research question

Theory/ Sample and Core findings framework method - Effects with different mediators Bressolles, - Do online service qual- - None - n = 2,813 Inter- - Information, aesthetics, ease of use, security/privacy, Durrieu and ity dimensions impact net users and reliability relate to online value dimensions. Deans online customer value - SEM - Social, functional, economic, and emotional value have (2015) dimensions? an effect on online satisfaction, while - Do the online customer - Online satisfaction influences online loyalty. - Social value mediates the effect of information, secuvalue dimensions medirity/privacy and ease of use, functional value the effect of ate the effects of the information, aesthetics, ease of use, security/privacy and service quality dimenreliability and economic value the effect of information, sions on online satisfacaesthetics and security/privacy on satisfaction. tion and online loyalty? Demangeot - How do retail websites - Stimulus-or- - n = 301 re- Interaction engagement with a website influences beand engage customers dur- ganism-respondents havioral engagement. Broderick ing the course of a web- sponse - SEM - Activity engagement with the website affects communi(2016) site navigation? framework cation engagement and behavioral engagement. - Communication engagement affects behavioral engagement with the website. - Experiential exploration potential and informational exploration potential influence activity engagement with the website and activates activity engagement with the site navigation more than it does interaction engagement. - Sense-making potential and experiential exploration potential affect interaction engagement. - Informational exploration potential influences interaction engagement and activates interaction engagement with the site more than it does activity engagement. Dickinger - How do certain website - Classical test n = 455 respond- - Website performance is most strongly determined by conand Stangl dimensions relate to theory ents tent quality and usefulness, followed by ease of use and (2013) website performance? n = 316 respond- website design. - Does website perforents - Trust and system availability are less important for website mance enhance satis- SEM performance determination. faction, perceived value - Website performance affects satisfaction and perceived and loyalty? value, while satisfaction has a direct effect on loyalty. Floh and - How do online content, - Stimulus-or- - n = 555 re- A pleasing online design affects consumer’s shopping Madlberger design and navigation ganism-respondents enjoyment in the online store. (2013) influence shopping ensponse - SEM - Perceived online navigation has an impact on consumjoyment, impulsiveness, framework ers’ shopping enjoyment, while the latter has an effect browsing and impulse on impulsiveness and browsing. buying behavior as well - Impulsiveness and browsing are associated with online as expenditure? impulse-buying behavior. Hsieh et al. - How do website dimen- - Theory of - n = 417 re- Perceived dominance positively affects online purchase (2014) sions affect purchase planned bespondents intentions and consumers’ pleasure. intention, through havior - PLS - Situational involvement moderates the relationships bearousal, pleasure and - Reactance tween perceived dominance and pleasure. perceived dominance? theory - Informativeness, navigational cues and perceived organ- Cognitive ization positively affect perceived dominance. theory of - Entertainment positively relates to arousal. emotions Kim et al. - What are the effects of - None - n = 293 re- Utilitarian and hedonic shopping value are positively re(2012) system, (security and spondents lated with customer satisfaction and customer repuraccessibility) infor- SEM chase intention, whereas customer satisfaction enmation (variety and curhances customer repurchase intention. rency) and service (vari- Security, accessibility, service quickness and service reety and currency) qualceptiveness are related with utilitarian shopping value. ity on utilitarian and he- Information variety, service quickness and service redonic values, customer ceptiveness are related to hedonic shopping value. satisfaction and online - The impact of information and service quality on utilitarrepurchase intention? ian and hedonic shopping value differs for income levels. - Do the links vary be- The impact of hedonic shopping values on repurchase tween income levels? intention differs across user income levels. Liu et al. - How do website cues - Stimulus-or- - n = 318 re- Impulsiveness is related to normative evaluation and the (2013) affect personality traits ganism-respondents urge to buy impulsively. (instant gratification, sponse - SEM - Normative evaluation affects instant gratification and the normative evaluation framework urge to buy impulsively. and impulsiveness) to - Instant gratification relates to the urge to buy impulsively. urge the impulse pur- Product availability affects visual appeal. chase online? - Visual appeal impacts normative evaluation and instant gratification, while it is also related to ease of use.

Figure 1.9 Studies on more than two examined marketing instruments. (Source: Own creation)

1.2 Research Gaps and Literature Review

McDowell et - What are the empirical al. (2016) relationships between website functions and online conversion rates? Chen and Dibb (2010)

Loureiro, Cavallero and Miranda (2018)

Rose et al. (2012)

Shin et al. (2013)

Toufaily and Pons (2017)

Zhou et al. (2009)

Al-Hawari (2011)

Bleier, Harmeling and Palmatier (2019)

- n = 114 con- Website features that enhance purchase intention are sumers from associated with levels of conversion rate in the visitor firms greeting, catalog and in the shopping cart stage. - Regression - Website features that enhance purchase intention are analysis related with conversion rate levels in the checkout stage. - Effects with trust as mediator - Do website quality di- Theory of - n = 452 re- Website usability, security/privacy, quality of the webmensions affect online reasoned ac- spondents site’s product information affect online trust. trust, attitudes toward tion - SEM - Website usability affects shoppers’ website attitudes. the website and ulti- Theory of - Online trust in the website affect attitudes toward the site mately website applanned beand site approach intentions. proach intentions? havior - Website attitudes affect site approach intentions. - Does familiarity with the - Familiarity with the website moderates the relationship site moderate these rebetween usability, security/privacy, speed of download, lationships? product information quality, service information quality, aesthetics and online trust, respectively. - What are the factors - Unified the- - n = 312 re- Perceived fashion website quality and technology quality driving website perforory of acspondents are related to performance. mance and trust? ceptance - PLS - Social influence by other fashion consumers and past - How does trust mediate and use of experience are related to performance expectancy. the relationship betechnology - Performance expectancy is related to satisfaction and to tween satisfaction and word-of-mouth behavior. word-of-mouth? - Online trust mediates the relationship between customer satisfaction and word-of-mouth behavior. - Customer satisfaction is related with online trust - Online trust and satisfaction affect word-of-mouth. - How do website dimen- - Stimulus-or- - n = 220 re- Telepresence and challenge enhance the cognitive exsions affect consumers’ ganism-respondents periential state. cognitive and affective sponse - PLS - Ease of use, the opportunity for customization and the experiential state? framework connectedness enhance the level of perceived control. - Do the cognitive and af- Perceived control affects affective experiential state. fective components af- Perceived benefits affect the affective experiential state. fect online shopping - The affective experiential state of the online shopper insatisfaction, online trust fluences the cognitive experiential state. and ultimately online re- Cognitive and affective experiential state affect online purchase intention? shopping satisfaction and online trust, which affect online repurchase intention. - How does site quality - None - n = 230 re- Site quality has six dimensions: shopping convenience, affect online repurchase spondents site design, information usefulness, transaction security, intention through satis- SEM payment system, and customer communication. faction, trust and con- Site quality has an effect on repurchase intention medisumer commitment? ated by satisfaction, trust, and commitment. - Do functional and rela- - None - n = 326 re- Website design, ease of use, website interactivity and tional attributes of webspondents virtual community do not affect online trust, while websites impact online trust - SEM site security/privacy, quality of support, personalization and online loyalty? and social presence do. - Are there differences in - Online trust and online loyalty are related. the relationships be- Interactivity and personalization more strongly influence tween multi-channel online trust for multi-channel vs. pure players. and pure player firms? - Support quality and virtual community affect online trust more for pure players vs. multi-channel firms. - How do website design - Expectation - n = 184 re- Website design quality does not directly impact online quality and service confirmation spondents trust, but affects online satisfaction. quality affect consumtheory - SEM - Service quality affects online trust and satisfaction. ers’ repurchase inten- Compared with website design quality, service quality tion, mediated by online has stronger affect online trust and satisfaction. trust and online satis- Online satisfaction affects online trust faction? - Online satisfaction and online trust affect the repurchase intention on the website. - Effects with brand as mediator - What are the relation- - None - n = 232 re- E-escape and security have a direct relationship with ships between online spondents brand awareness and brand image. service quality factors - SEM - E-responsiveness affects brand image. and brand value in a - None of the online service quality factors has a direct rebanking context? lationship with brand loyalty. - Image and awareness enhance loyalty. - How do design ele- None - n = 10,470 - The conceptualization of online customer experience in ments shape dimenworkers informativeness, entertainment, social presence, and sions of online cus- SEM sensory appeal reveals why the effectiveness of any tomer experience and given design element varies. ultimately influence pur- Search vs. experience negatively and brand trustworthichases? ness positively moderate the relationships.

Figure 1.9 (continued)

- None

29

30

1

Introduction

of online- and omni-channel-specific instruments, which is important for omnichannel managers. Although, indirect effects of the instruments on consumer behavior have been frequently shown, the mechanism that transforms these instruments into behavior is not fully understood yet. Open questions that arise from combining these gaps in the literature will therefore be presented in Section 1.2.5.

1.2.5

General Research Objectives

This comprehensive review of the literature shows that imperative decisions of omni-channel retailers have been investigated in previous research. Particular emphasis has been placed on interdependencies between channels and on integration services and instruments that support channel transfers. However, reciprocal effects between channels, and stimuli and cues have not been adequately addressed, due in part to insufficiently applied empirical methods. As a result, there are open questions that need to be further studied. Accordingly, the focus of each of the three studies that make up this doctoral thesis is derived from the gaps identified in the literature review above. The first gap in the literature addressed by the present doctoral thesis relates to the effects of important reciprocal relationships between major offline and online purchase channels of omni-channel retailers. The existing literature has primarily analyzed unidirectional relationships, which provide different insights for channel image effects (e.g., Kwon and Lennon 2009b; Chang and Tseng 2013). Less frequently bidirectional channel relationships have also been addressed (e.g., Badrinarayanan et al. 2012; Breugelmans and Campo 2016). These have considered cross-channel effects, but neglected reciprocity, leading to inconclusive generalizations. Reciprocity in non-recursive models goes beyond cross-channel relationships, since one channel can affect a second channel and the second channel in turn can affect the first in a loop-like relationship (e.g., Nagase and Kano 2017). The only study that analyzed reciprocal relationships was by Swoboda, Weindel and Schramm-Klein (2016) and showed reinforcing effects. The authors, however, analyzed effects on retailer-level loyalty only, and did not consider further cross-channel effects. Furthermore, only initial evidence on the role of reciprocal effects of channel images over time exists. Previous studies have mostly adopted cross-sectional designs, thereby limiting the understanding of interdependencies between channel images over time. Scholars have called for research to uncover all relationships in multi- and omni-channel retailing while considering reciprocal relationships (Kwon and Lennon 2009a; Loupiac and Goudey 2019;

1.2 Research Gaps and Literature Review

31

Wiener et al. 2018). In addition, in omni-channel retailing, consumers draw inferences based not only on current information but also on their prior experiences (e.g., Lemon and Verhoef 2016; Grewal and Roggeveen 2020), and reciprocal effects might vary for retailer with which consumers have more (vs. less) favorable experiences. Unfortunately, previous work has failed to cover these prior experiences. Thus, the aim of Study 1 is to move the research field forward by investigating reciprocal effects on channel-level and retailer-level loyalty while considering consumers’ prior experiences. The second gap refers to the disentangled effects of OF-ON and ON-OF services offered by omni-channel retailers to provide consumers seamless experiences across channels. Previous studies have not considered differences in the results of simultaneously perceived integration services. Studies have included the joint effects of integration services (e.g., Cao and Li 2015; Hossain et al. 2020), OF-ON services (e.g., Bhargave, Mantonakis and White 2016; Mosquera et al. 2018), and ON-OF services individually on behavioral and performance outcomes. However, consumers perceive and value integration services differently (Akturk, Ketzenberg and Heim 2018; Li et al. 2018; Pan et al. 2017). Analyzing ramifications of integration services can therefore change the results of studies that analyze only joint effects or one direction only (e.g., Hamouda 2019; Herhausen et al. 2015). In addition, omni-channel retailers must design their channel offerings in such a way that they do not disrupt integration service outcomes (Banerjee 2014). Although studies have mostly shown indirect effects of perceived integration services on consumer behavior, they have frequently examined general channel evaluations such as brand image or quality perceptions that translate integration services into consumer behavior (e.g., Schramm-Klein et al. 2011; Herhausen et al. 2015). Cross-channel effects of mediators on offline or online consumer behavior are evident in omni-channel systems (e.g., Ravula et al. 2020) but have not been conceptualized. Research calls exist for this sort of research (e.g., Shen et al. 2018; Zhang et al. 2018). Moreover, knowledge of important context factors for the integration services’ mediation paths is scarce. Consumers’ online shopping experience is known to reduce the effects of ON-OF services (e.g., Herhausen et al. 2015) and should be therefore analyzed in more detail. In addition, today, firms expend considerable effort on synchronizing their channels to achieve channel congruence (e.g., Van Baal 2014), and it might increase the probability that integration services enhance purchase intentions if consumers obtain similar offerings across channels. Previous work has failed to address these important context factors. Therefore, in Study 2, insights into the respective mediation paths for offline and online purchase intentions depending on important context factors are provided.

32

1

Introduction

The third gap in the literature concerns the relative importance of online- and omni-channel-specific marketing instruments for online and offline repurchase intentions over time. Scholars have called for such research (e.g., Bleier, Harmeling and Palmatier 2019; Blut et al. 2018) because it is unclear whether existent findings can be generalized to omni-channel retailers. Few studies have addressed one or two instruments (e.g., Al-Qeisi et al. 2014; Sullivan and Kim 2018), and only inconclusive conclusions can be drawn from these studies for successful management of instruments in omni-channel systems. Studies that focus on more than two instruments have shown their relative importance, but do not focus on online- and omni-channel specific instruments simultaneously (e.g., Loureiro, Cavallero and Miranda 2018; McDowell, Wilson and Kile Jr. 2016). Frequently, indirect effects of instruments show that online trust is vital in e-commerce research. Nevertheless, for omni-channel retailers especially, retail brand equity is known to be of paramount importance, and thus should be considered as a further mechanism that translates instruments into behavior. Furthermore, previous research has relied on cross-sectional data, which does not allow rigorous inferences to be drawn about the causal relationships over time. Only initial evidence on the role of online trust over time exists and scholars have called for studies to verify its effects over time (Kim and Peterson 2017; Ye et al. 2020). Such analysis can contribute substantially to the debate over whether trust is a result of brand associations or whether brand associations are affected by trust (Hollebeek and Macky 2019; Rajavi et al. 2019). Finally, although experiences from website interactions transfer cognitively to additional channels (e.g., Grewal and Roggeveen 2020), analyses of such cross-channel effects are still missing. Study 3 addresses these gaps. This lack of knowledge needs to be discussed in further detail. The following research objectives form the basis for the conceptualization and implementation of the three studies in this doctoral thesis. (1) The first research objective is to analyze the reciprocal relationships between omni-channel retailers’ images of major purchase channels and their effects on channel-level and retailer-level loyalty. These interrelationships will be examined in the light of prior experiences with retailers. (2) The second research objective seeks to investigate the mediation paths of perceived OF-ON and ON-OF channel integration services for purchase intentions in offline and online channels as moderated by consumers’ online shopping experiences and channel congruence. (3) The third research objective is to examine the relative importance of onlineand omni-channel-specific marketing instruments for online and offline

1.3 Structure and Contribution of the Studies

33

repurchase intentions as mediated by online trust and brand equity and taking into account reciprocal effects of online trust and brand equity. Each of the three main research objectives is dealt with in a separate study in chapters B to D. Each study raises more detailed research questions that relate to the respective objectives mentioned above and requires specific conceptual frameworks and analyses. While Study 1 directly deals with reciprocity between channel images, Studies 2 and 3 address stimuli and cues that support seamless channel transfers of consumers in omni-channel retailing. Cross-channel effects are examined in all three studies. Based on the results of the studies, Chapter E summarizes the findings and discusses them in general terms to point out the empirical and theoretical implications. Subsequently, the structure and contributions of the three studies are presented in detail.

1.3

Structure and Contribution of the Studies

1.3.1

Reciprocity within Major Retail Purchase Channels and their Effects on Overall, Offline and Online Loyalty

Multi-channel and omni-channel retailers attract consumers through a variety of channels, and the latter additionally offer seamless experiences through further touchpoints (e.g., Verhoef, Kannan and Inman 2015). However, physical stores and online stores are still the two dominant sales channels for both forms of retailers (e.g., Bell, Gallino and Moreno 2018). These retailers retain loyal consumers by supporting channel image transfers. However, they still find it challenging to build strong consumer loyalty through a combination of sales channels (Hult et al. 2019). In this study, reciprocal effects of major purchase channel images were examined because consumer perceptions and attitudes formed for each channel are stored in memory and transfer to further channels to affect purchase behavior (Bezes 2014; Hunneman, Verhoef and Sloot 2017). The motivation for Study 1 stems from the lack of knowledge about the interdependencies between channels and reciprocal effects of major sales channels. This study therefore focuses on reciprocal effects, i.e., consumers’ attitudes influence one another’s attitudes in a loop (Nagase and Kano 2017). The respective knowledge is important for managers to coordinate major sales channels owing to their interdependencies. Reciprocity may change extant findings, for example, results concerning online channels only (e.g., Wu et al. 2013). Additionally, the role of prior experiences, which are generally important in omni-channel studies

34

1

Introduction

(e.g., Grewal and Roggeveen 2020), are analyzed for reciprocal relationships. This is a new way of looking at the interdependencies between the channels that multiand omni-channel retailer own and directly control. Study 1 therefore analyzes the following research questions: • Does a reciprocal relationship between offline and online images of major purchase channels exist and how do these reciprocal relationships affect overall retailer, offline channel and online channel loyalty? • Whether and how do reciprocal effects differ for retailer with which consumers have more vs. less favorable prior offline and online experiences? The conceptual framework of Study 1 is based on categorization theory and the hypothesis development follows cognitive rationales and empirical studies. Categorization theory provides a stringent cognitive explanation for the research aims because consumers in multi- and omni-channel retailing tend to categorize channels and transfer the respective knowledge for evaluation (Shen et al. 2018; Sohn 2017). Categories represent omni-channel retailers, and category members represent their channels. Consumers faced with a channel to evaluate will categorize the channel as related to a retailer (Wang, Beatty and Mothersbaugh 2009). In a seamless retail environment, consumers easily switch from one channel to another (Acquila-Natale and Chaparro-Peláez 2020) and reciprocally transmit the images of one channel to another (Bezes 2013; Lemon and Verhoef 2016). Consumers are assumed to draw inferences about channels based on the most knowledgeable purchase channel (i.e., offline in the case of former brick-and-mortar retailers, Benedicktus et al. 2010) and use category-based inferences for decision-making, for example, decisions to be loyal to a retailer and related channels (Loken 2006). Prior experiences are informative in their own right and may change the categorization and transmission process (Grewal and Roggeveen 2020; Lemon and Verhoef 2016). The empirical analysis comprises two studies. In the first study, longitudinal data from 573 consumer evaluations of leading omni-channel fashion retailers were analyzed using cross-lagged panel models. Multigroup models are used for the experience effects. Study 2 tests identical relationships based on 380 consumer evaluations and reveals possible shortcomings of cross-sectional designs. Alternative models strengthen the observations and provide further insights. The results of the first study indicate that a positive offline channel image enhances the online channel image and vice versa. The reciprocal effects of offline (vs. online) channel image are stronger on overall retailer, offline channel,

1.3 Structure and Contribution of the Studies

35

and, surprisingly, online channel loyalty. Moreover, the effects change for retailer with which consumers have more or less favorable prior offline and online experiences. Thus, the results support the theoretical assumption that both offline and online channel images affect loyalty, i.e., consumers’ category knowledge causes responses. Individuals rely on the most knowledgeable (i.e., representative) category member to draw inferences and use these category inferences for decision-making (Loken 2006; Sohn 2017). For retailer, with which consumers have more favorable prior experiences, the respective effects increase, which underlines the theoretical rationales. The second study points to shortcomings of cross-sectional studies, differentiated for reciprocal models and mediation models between offline and online image and with no relations between both constructs. This study design delivers different results and implications than the reciprocal study. For example, cross-sectional studies tend to overestimate or underestimate online and offline channel image effects. Models without reciprocal effects do not consider the directionality of effects. The findings have direct implications for research and managers interested in understanding how reciprocity and prior experience affect consumer behavior.

1.3.2

Effects of Perceived Offline-Online and Online-Offline Channel Integration Services

Omni-channel retailers apply integration services to provide consumers with a seamless experience. These services allow customers to execute different interactive activities across channels (Banerjee 2014; Sousa and Amorim 2018). Consumers benefit from increased knowledge and access to channels in decisions, both of which increase their purchase intentions (Herhausen et al. 2015; Lee et al. 2019). In particular, OF-ON services support consumers in offline venues to interact with an online channel, whereas ON-OF services support interaction with offline channels online (e.g., Hamouda 2019; Hossain et al. 2020). Consumers value the services differently, and the services’ important mediation paths to offline and online purchase intentions via the perceived quality of offline and online offerings are studied here. Two important context factors for the mediation paths are conceptualized: consumers’ online shopping experience and channel congruence (e.g., Herhausen et al. 2015; Vyt, Jara and Cliquet 2017). The motivation for Study 2 emerges from a substantial gap in the extant literature. Previous research on the effects of simultaneously offered but differently perceived OF-ON and ON-OF services observed joint effects or one direction

36

1

Introduction

of services only (e.g., Li et al. 2018; Yang et al. 2020). Often, general channel evaluations are conceptualized as mediators for integration services’ effects (e.g., image, Schramm-Klein et al. 2011), although the aim of integration services is to increase the salience of both offline and online channels. Cross-channel effects are not considered. Moreover, managers find it challenging to offer OF-ON and/or ON-OF services that increase channel quality and behavioral outcomes. Therefore, this study analyzes ramifications of integration services, important cross-channel links, and moderators for the mediation paths. Therefore, Study 2 aims to answer the following research questions: • Whether and how are perceived OF-ON vs. ON-OF services transformed into offline and online channel purchase intentions through perceived quality of offline and online channel offerings? • How do two important context factors—consumers’ online shopping experience and channel congruence—moderate these mediation paths? The conceptual framework of Study 2 is based on accessibility-diagnosticity theory and on studies on the effects of integration services on consumer behavior. The theory suggests that the likelihood of using a cue or input for decision-making depends on the specific input’s accessibility and its relative diagnosticity (Feldman and Lynch 1988). Accessibility refers to the ease and extent of retrieving a cue. Relative diagnosticity refers to the extent to which the inferences based on this input can be used to make a decision (Lynch et al. 1988). Consumers use all available cues in decisions, although their relevance for evaluation varies as the most accessible and diagnostic input will be more frequently used (Menon and Raghubir 2003). The probability that the integration services will be used for evaluation is a function of their accessibility and relative diagnosticity. Integration services only become relatively diagnostic by affecting the perceived quality of offerings, thus rendering this information relevant for purchases. Context factors are known to be further diagnostic cues. Such higher- or lower-order cues can change the relevance of integration services (Feldman and Lynch 1988) and are therefore considered in this study. The empirical analysis of Study 2 is based on a survey with 722 consumer evaluations of leading omni-channel fashion firms. Insights on conditional effects are provided by proving that the respective paths of the mediation model are conditional on a continuous latent moderator by using the latent moderated structural equation method (LMS; Cheung and Lau 2017). The moderated mediation is further probed using the Johnson-Neyman floodlight test, which illustrates the significance of the conditional effects (Spiller et al. 2013).

1.3 Structure and Contribution of the Studies

37

The results of Study 2 show that OF-ON services provide knowledge about and ease access to both offline and online channels and enhance purchase intentions. ON-OF services show no links to offline channels, and purchase intentions depend on the perceived quality of online offerings only. Surprisingly, OF-ON services increase purchase intentions the most. Additionally, higher levels of consumers’ online shopping experience as a diagnostic cue reduce the mediation paths, and higher levels of perceived channel congruence positively and negatively moderate them. Knowing the cross-channel effects of OF-ON and ON-OF services that are most important to consumers enables effective management. In general, this study adds to research by applying the accessibility-diagnosticity theory to the mediation paths across major sales channels. Moreover, this study enhances views of joint perspectives by disentangling ramifications of integration services and by considering cross-channel effects. For managers this study identifies OFON services as the major lever for offline and online purchase intentions and in alternative models, at the retailer level.

1.3.3

Importance of Marketing Instruments for Repurchase Intentions in Omni-channel Retailing

Omni-channel retailers increasingly compete for consumers online. They have to orchestrate online- and omni-channel-specific instruments to benefit from crosschannel effects (Toufaily and Pons 2017; Yang et al. 2020). Therefore, in this study, the relative importance of marketing instruments—online-specific like aesthetic appeal, ease of use, security/privacy, consumer service, and omni-channelspecific like online-offline integration and channel consistency—are examined. Effective orchestration depends on the relative importance of instruments and on mechanisms that translate them into behavior (Bleier, Harmeling and Palmatier 2019). Online trust is a known mechanism. In retail, brands are also important to consumers. Thus, this study examines the effects of the instruments on consumer behavior mediated by online trust and brand equity. The motivation for Study 3 is twofold. Although various instruments have been considered promising for omni-channel retailers, their relative effects on consumer behavior across channels remain unclear. Most studies have addressed single instruments (e.g., Al-Qeisi et al. 2014). Fewer studies have analyzed more instruments, while online trust is most often conceptualized as a mediator (e.g., Loureiro, Cavallero and Miranda 2018). These studies predict the relative importance of instruments but not of online- and omni-channel-specific ones

38

1

Introduction

simultaneously. Importantly, only a few studies have addressed retail brand associations (e.g., Al-Hawari 2011; Bleier, Harmeling and Palmatier 2019). However, brand equity and trust are assumed to be interlinked (e.g., Hollebeek and Macky 2019), as online trust increases brand equity and brand equity leads to higher trustworthiness evaluations of websites (e.g., Benedicktus et al. 2010; Fazale-Hasan et al. 2018; Khan et al. 2019). This reciprocity has not yet been analyzed (i.e., interdependencies of both in a loop, Nagase and Kano 2017). Thus, research has not sufficiently considered the relative effects of online- and omnichannel-specific instruments on consumer behavior. Therefore, this doctoral thesis analyzes effects of online- and omni-channel-specific instruments on online and offline repurchase behavior and considers reciprocal effects between online trust and online brand equity. Therefore, the following research questions are addressed in Study 3. • How do important online- and omni-channel-specific instruments affect repurchase intentions in retailer’s online and offline channels, as mediated by online trust and brand equity? • Do reciprocal relationships between online trust and brand equity exist, and how do they affect online and offline repurchase intentions? In the conceptual framework of Study 3, perceived online trust and brand equity mediate the impact of the studied instruments on repurchase intentions. A systematic process based on a literature review guides the selection of online- and omni-channel-specific instruments. Scholars often conceptualize marketing instruments as stimuli perceived by consumers as information cues to form attitudes or cognitions and to facilitate decisions. However, in repurchase decisions, consumers refer to learned and stored information in memory (Grewal and Roggeveen 2020). Thus, this study, refers to schema and associative network theory, which provide organizing mechanisms for such knowledge (e.g., Fiske and Taylor 1991, p. 98; Krishnan 1996). Moreover, online trust and brand equity are conceptualized as nodes in consumers’ memory (e.g., Keller 2010; Bock et al. 2012). Crosschannel effects on online and offline repurchase intentions and toward the retailer in general in alternative models are determined. The empirical analysis of Study 3 is based on two methods. Using a sample of 377 respondents that evaluated fashion retailers, sequential mediation modeling over time was applied in a first study. The dependent, independent, and mediating variables were measured at different time points, which avoided the limitations of cross-sectional studies that may bias the estimation of mediation parameters. In

1.4 Further Remarks

39

a second study, cross-lagged structural equation modeling is applied by using the same data to evaluate the reciprocal effects over time. The results of Study 3 provide evidence for different indirect effects of the instruments and a stronger role of online brand equity than of online trust. Four online- and omni-channel-specific instruments are identified as the most influential, leading to important theoretical implications. Reciprocal relationships reveal that online brand equity (general retail brand equity in alternative models) stronger impacts on offline and online repurchase intentions than trust. These findings have direct implications for managers interested in understanding which particular marketing instruments and mechanisms affect consumer outcomes.

1.4

Further Remarks

In order to extend the aforementioned research questions, this thesis consists of three studies that focus on cross-channel effects in omni-channel retailing. Study 1, Study 2, and Study 3 are organized around the following structure: • • • • •

Introduction Conceptual framework and hypothesis development Empirical study, including sample design, measurement, method, and results Discussion with theoretical and managerial implications Limitations and directions for further research

This structure is set irrespective of the applied theory. Regarding theory, Study 1 is based on categorization theory, which is established in research on channel interdependencies (e.g., Keaveney and Hunt 1992; Sohn 2017). Study 2 follows an accessibility-diagnosticity theoretical rational to explain effects of accessible channel integration service cues (Feldman and Lynch 1988). Study 3 is based on cognitive theoretical approaches rooted in schema and associative network theory (Fiske and Taylor 1991, p. 98; Anderson 1983; Keller 1993). Study 1 and Study 3 deviate slightly from the proposed structure. Study 1 consists of two empirical studies. The empirical studies, including sample design, measurement, method, and results are therefore provided separately for both studies. In Study 3, two different methods are used based on the same sample and measurement. The methods and results for each study are presented in separate chapters. The remainder of this doctoral thesis is organized as follows. After having illustrated the research gaps and the current state of the literature, the three studies

40

1

Introduction

are addressed in Chapters B, C, and D. Chapter E summarizes the conclusions of the three studies with regard to the general research objectives of this doctoral thesis. The author finally outlines further research avenues.

2

Study 1: Reciprocity within Major Retail Purchase Channels and their Effects on Overall, Offline and Online Loyalty

2.1

Introduction

This study addresses reciprocal effects, i.e., consumers’ attitudes towards major offline and online purchase channels influence one another in a loop (Kwon and Lennon 2009a; Nagase and Kano 2017). These transfers occur in today’s retail environment as associations with one channel transfer to other channels and affect purchase behavior (Shen et al. 2018). The effects are relevant for multi-channel retailers that offer more than one channel and for omni-channel retailers that additionally offer seamless experiences across integrated channels and further touchpoints, which are not studied subsequently (Valentini, Neslin and Montaguti 2020; Verhoef, Kannan and Inman 2015). Physical and online stores are the two currently dominant sales channels for omni-channel retailers (Bell, Gallino and Moreno 2018; Hult et al. 2019). Leading retailers such as H&M, Walmart, and Sephora retain loyal consumers by supporting image transfers between sales channels because online stores are known to drive physical stores and vice versa (Forbes 2019; Grewal et al. 2020). However, retailers find it challenging to build strong consumer loyalty through a combination of sales channels (Hult et al. 2019), e.g., they close physical stores, possibly ignoring their role for online stores (Ailawadi and Farris 2017; Valentini, Neslin and Montaguti 2020). Therefore, this study analyzes reciprocal effects of major purchase channel images (i.e., consumers’ perceptions and attitudes formed for each channel) that are stored in memory and affect purchase behavior (Bezes 2014; Hunneman, Verhoef and Sloot 2017). The effects of categorized offline and online channels in memory (Benedicktus et al. 2010) on loyalty to the overall retailer, the offline, and the online

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 A. Winters, Omni-Channel Retailing, Handel und Internationales Marketing Retailing and International Marketing, https://doi.org/10.1007/978-3-658-34707-9_2

41

42

2

Study 1: Reciprocity within Major Retail Purchase Channels …

channel are focused and additionally the role of prior experiences, important in omni-channel studies. Scholars mostly study unidirectional relationships (i.e., one-way effects of one channel on either online or offline outcomes, see Figure 2.1). For example, Fuentes-Blasco et al. (2017) examine the effects of offline store images on offline outcomes, Aghekyan-Simonian et al. (2012) analyze the effects of online channel images on online intentions, and Bezes (2013) shows unidirectional effects of offline and online channel images on a retailer. Fewer scholars address bidirectional relationships (i.e., cross-channel effects), such as offline and online outcomes depending on offline and online price promotions (e.g., Breugelmans and Campo 2016), or offline and online channel images (e.g., Kwon and Lennon 2009a), but not reciprocally. However, reciprocity in non-recursive models goes beyond cross-channel relationships because one channel has an effect on a second channel, and the second channel in turn has an effect on the first (Zyphur et al. 2019). The only study on reciprocity indicates reinforcing effects (Swoboda, Weindel and Schramm-Klein 2016). Reciprocity may change extant findings (e.g., regarding the roles of exclusively observed online channels in omni-channel studies, Wu et al. 2013). Scholars call for research on reciprocity to reveal all relationships in multi- and omni-channel retailing (Kwon and Lennon 2009a; Loupiac and Goudey 2019; Wiener, Hoßbach and Saunders 2018). Knowledge of reciprocity is important for managers to coordinate major sales channels due to their interdependencies and because a theory-based conceptualization and empirical consideration are valuable.

Unidirectional

Bidirectional

Offline Channel Fuentes-Blasco et al. (2017); Grosso, Castaldo and Grewal (2018); Moliner-Velázquez et al. (2019); Van Nierop et al. (2011); etc. none

Omni-channel Outcomes Online Channel Aghekyan-Simonian et al. (2012); Bezes (2014); Chang and Tseng (2013); Chu et al. (2017); Kwon and Lennon (2009b); Landers et al. (2015); Wang, Beatty and Mothersbaugh (2009); Wu et al. (2013); etc.

Overall Retailer Anselmsson, Burt and Tunca (2017); Bezes (2013); Hunneman, Verhoef and Sloot (2017); Murray, Elms and Teller (2017); etc.

Herhausen et al. (2015) Badrinarayanan et al. (2012); Verhagen and van Dolen (2009); Yang, Lu and Chau (2013)

Swoboda, Weindel and Schramm-Klein (2016)

Breugelmans and Campo (2016); Kwon and Lennon (2009a); Verhoef, Neslin and Vroomen (2007) This Study Note: Italic = unidirectional interdependent relationships, i.e., from online (offline) channels to offline/overall retailer (online) outcomes; Underlined = reciprocal relationships of channels.

Figure 2.1 Literature review on channel relationships in omni-channel retailing. (Source: Own creation)

2.1 Introduction

43

This study aims to answer the research question of whether reciprocal relationships between offline and online images of major purchase channels exist and how they affect conative loyalty (i.e., intention and readiness to purchase at or to recommend a retailer or its channels, Oliver 1999). Additionally, this study asks whether and how reciprocal effects differ for retailers with which consumers have more vs. less favorable prior offline and online experiences (Grewal and Roggeveen 2020). Accordingly, this study offers two contributions. First, it contributes to the literature by studying the effects of reciprocity on overall retailer, offline, and online channel loyalty (typical aims to retain consumers, Ailawadi and Farris 2017). Moreover, this study contributes to the application of categorization theory in a multi- and omni-channel context because consumers group objects into categories to draw inferences. Consumers categorize an offline or online channel within the same brand and reciprocally transfer knowledge to evaluate retailers’ channels (Benedicktus et al. 2010; Grewal and Roggeveen 2020). Furthermore, this study provides empirical contributions to extant knowledge. Longitudinal study 1 provides new insights into channel images’ reciprocity in today’s retail environment (channel associations in alternative models). Study 2 tests identical relationships and implies possible consequences of the exclusion of reciprocity from traditional cross-sectional designs. Practical implications are provided for managers on the challenges of retaining consumers through channel images. Second, this study contributes to the literature by considering prior experiences, i.e., accumulated attitudes formed based on prior exposure to retailers’ channels, incorporating consumers’ responses to their interactions with a retailer (Lemon and Verhoef 2016; Grewal and Roggeveen 2020).1 Theoretically, consumers draw inferences based not only on current information but also on their prior experience (Badrinarayanan et al. 2012; Loken 2006). However, because no common agreement on a measure exists (e.g., Wang and Goldfarb 2017), in this study, retailers will be preselected. By conceptualizing the role of experiences in the reciprocal transfers of images in loyalty formation, we contribute to the application of categorization theory (e.g., Badrinarayanan et al. 2012). The findings extend studies that analyze prior experiences but not reciprocity (e.g., Wang, Beatty and Mothersbaugh 2009). The remainder of this study proceeds as follows. By drawing from theory, the authors derive hypotheses and test them based on consumer evaluations of 1

Consumers’ prior experience is distinct from consumers’ internet shopping experience (i.e., extent of searching for or purchasing products online, e.g., Herhausen et al. 2015) and from customer experience (i.e., cognitive, emotional, behavioral, sensorial, and social responses to a firm’s offerings in customers’ purchase journey, e.g., Lemon and Verhoef 2016).

44

2

Study 1: Reciprocity within Major Retail Purchase Channels …

leading omni-channel fashion firms in two studies. After presenting the results, implications and directions for research are provided.

2.2

Conceptual Framework and Hypothesis Development

2.2.1

Overview

To address the research aims, this study first refers to categorization theory and then derives hypotheses based on theory and empirical studies. This study assumes that offline and online channel image reciprocally affect overall retailer, offline channel, and online channel loyalty and that the reciprocal effects differ for retailers with which consumers have more (vs. less) favorable prior offline and online experiences (see Figure 2.2).

More vs. less favorable prior offline experiences

Offline channel image

1) Overall retailer loyalty 2) Offline channel loyalty 3) Online channel loyalty Online channel image

More vs. less favorable prior online experiences

Figure 2.2 Conceptual framework. (Source: Own creation)

2.2.2

Theory

According to categorization theory, individuals structure their knowledge by categorizing objects for efficient information processing (e.g., Mervis and Rosch 1981; Loken 2006). Individuals who perceive an object compare it with existing categories and information stored in their memory (Keaveney and Hunt 1992). If they find a match, they categorize the object into the corresponding category and transfer category knowledge to the category member to draw inferences (Wang, Beatty and Mothersbaugh 2009). This transmission process is used to evaluate the category member. Moreover, individuals rely on the most knowledgeable (i.e., most representative) category member to draw the strongest inferences (Mervis and Rosch 1981). Category knowledge and inferences are used for judgments in

2.2 Conceptual Framework and Hypothesis Development

45

decisions (Loken 2006), i.e., category knowledge affects individuals’ behavioral responses (e.g., loyalty, Keaveney and Hunt 1992; Morales et al. 2005). Because consumers in multi- and omni-channel retailing categorize channels and transfer knowledge between them (e.g., Shen et al. 2018; Sohn 2017), we argue according to categorization theory.2 In our context, categories represent retailers, and category members represent their channels (e.g., catalog, stores, Badrinarayanan et al. 2012; Saghiri et al. 2017). Theory holds that consumers faced with a channel categorize the channel as related to a retailer (Wang, Beatty and Mothersbaugh 2009). Especially in today’s retailing environment, consumers easily switch from one channel to other channels (Acquila-Natale and ChaparroPeláez 2020) and reciprocally transmit the images of one channel to another (e.g., Bezes 2013; Lemon and Verhoef 2016). Moreover, consumers draw inferences about channels based on the most knowledgeable purchase channel (i.e., offline or online in the case of former brick-and-mortar or purely online retailers, Benedicktus et al. 2010). They use category-based inferences for decision-making (Loken 2006). The reciprocity of offline and online channel images affects decisions to be loyal to a retailer and related channels. Previous experiences change the categorization and transmission process (Keaveney and Hunt 1992; Grewal and Roggeveen 2020). Experiences are informative in their own right, and consumers draw inferences based on them (Lemon and Verhoef 2016). Prior experiences reduce efforts in cognitive processing, and induce categorization (Alba and Hutchinson 1987). Favorable experiences with a retailer’s channel strengthen reciprocal image transfers as they facilitate the use of knowledge and elicit stronger responses; less favorable experiences weaken the effects. Experience leads to an accumulation of information, allowing consumers to retrieve and use more and similar information for decisions (Alba and Hutchinson 1987). We expect stronger reciprocal effects for more (vs. less) favorable prior experiences of offline and online channels.

2.2.3

Hypothesis on Reciprocity and its Effect

Theoretically, consumers compare a retailer’s channels against categories stored in their memory. Consumers categorize the channel to which they switch into the 2

Multi- and omni-channel retailers differ. For example, the latter offer integrated channels (e.g., Verhoef, Kannan and Inman 2015, the effect of which is controlled). Theoretical categorization effects occur for the dominant sales channels of both retailer types, but empirically, this study focuses on omni-channel retailers (e.g., with advantages due to forced categorization by redirecting consumers); therefore, they are usually used as the theoretical reference.

46

2

Study 1: Reciprocity within Major Retail Purchase Channels …

corresponding retailer category (e.g., Loken 2006; Wang, Beatty and Mothersbaugh 2009). Moreover, they transmit category inferences. i.e., they convey offline channel images to an online channel to evaluate the latter. The same process applies to consumers who switch to an offline channel: they categorize the channel and convey online images to offline images. Information on channels with a high category fit can be more easily processed (e.g., Sohn 2017). Successful categorization stimulates positive feelings, resulting in a positive evaluation (e.g., Badrinarayanan et al. 2012; Loken 2006; for an example, Verhagen, van Dolen and Merikivi 2019). Consumers faced with a disparate category member (e.g., an unappealing online store) rely more on typical category members (e.g., offline stores, Kwon and Lennon 2009a; Loken 2006). Thus, we assume a positive relationship. Unidirectional empirical studies have demonstrated a transmission from offline to online channels or vice versa. Kwon and Lennon (2009b) and Verhagen and van Dolen (2009) show that offline channel image positively affects online channel image; Bezes (2014) observes this relationship in the opposite direction. Badrinarayanan et al. (2012) identify reciprocal reasoning when they show that high congruence between retailers’ online and offline channel images positively affects behavior, but they do not consider reciprocity between channels. Finally, studies on cannibalization between channels suggest positive effects through seamless channel transfers (e.g., Cao and Li 2015; Pauwels and Neslin 2015). We propose the following: H1:

Offline and online channel images have a positive reciprocal relationship.

Theoretical rationales explain how the total effects, i.e., the sum of the direct and indirect effects of offline and online channel image, affect loyalty. If a categorization is successful and reciprocal effects occur between the channels, consumers transfer category knowledge and use it for evaluation (Keaveney and Hunt 1992). The total effects of offline and online channel image affect loyalty because consumers regularly use category knowledge and inferences in making judgments about category members (Loken 2006). Category knowledge and inferences affect consumers’ behavioral intentions (Loken 2006; Morales et al. 2005). In multi- or omni-channel retailing, consumers draw the strongest inferences from the most knowledgeable channel. Loyalty towards the overall retailer and the offline (vs. online) channel of former brick-and-mortar retailers is likely to be more strongly affected by the offline channel image (Benedicktus et al. 2010; Hult et al. 2019). The online channel is the most knowledgeable for former purely online players (Saghiri et al. 2017).

2.2 Conceptual Framework and Hypothesis Development

47

However, studies have observed mixed unidirectional and bidirectional effects of either offline or online channels on overall retailer, offline channel and online channel loyalty. Scholars have shown significant effects of online and offline image on overall retailer loyalty (e.g., Hunneman, Verhoef and Sloot 2017; Murray, Elms and Teller 2017) and positive effects of offline (online) channel image on offline (online) loyalty (e.g., Chang and Tseng 2013; Verhagen and van Dolen 2009). Swoboda, Weindel and Schramm-Klein (2016) show stronger reciprocal effects of offline (vs. online) channels on overall retailer loyalty. The following hypothesis is proposed: H2:

2.2.4

The total effect of offline channel image is stronger than the total effect of online channel image on (a) overall retailer loyalty and (b) offline channel loyalty but weaker than that on (c) online channel loyalty.

Hypothesis on the Role of Prior Experience

Prior experience with a retailer’s channels is important because consumers draw inferences not only from external information but also from their prior experience (Shen et al. 2018). Experienced consumers accumulate knowledge that they retrieve more easily from memory, which leads to stronger categorization effects (Alba and Hutchinson 1987). They draw inferences and make decisions based on accumulated information (Morales et al. 2005). Favorable experiences facilitate the use of specific knowledge and elicit strong behavioral responses, whereas less favorable experiences do not. Experience affects information processing and behavior and is important in multi- and, especially, omni-channel retailing due to easier switching between channels through integration (Grewal and Roggeveen 2020). Next, offline and online experiences are addressed in detail. Prior Offline Experiences: Retailers with which consumers have more (vs. less) favorable prior offline experiences might expect stronger total offline channel image effects because consumers have increased available cognitive resources, leading to less effort in cognitive processing (Yang et al. 2020). However, whether the same effect occurs for the online channel is unclear. First, considering more (vs. less) favorable prior offline experiences, we assume that there are strengthened reciprocal offline channel image effects on overall retailer and offline channel loyalty. More experienced consumers accumulate offline channel knowledge and retrieve it more easily. This similar information drives their decisions or related behavior (Verhoef, Neslin and Vroomen 2007; Yang

48

2

Study 1: Reciprocity within Major Retail Purchase Channels …

et al. 2020). However, we also expect stronger effects on online channel loyalty for consumers with more experience. Stronger retrieval of offline channel images affects online loyalty, as shown in cross-channel studies (e.g., Kwon and Lennon 2009a). The predominant online channel image effect is reciprocally more weakly affected (Swoboda, Weindel and Schramm-Klein 2016). Second, regarding online channel image effects, we assume similar relationships as experiences formed through interactions with retailers’ physical stores are transmitted to further channels, such as online stores (De Kerviler et al. 2016). More experienced consumers more easily retrieve knowledge and transfer it from one channel to another (Yang et al. 2020). For them, cognitive efforts in decision-making are reduced, and stronger responses are likely. A stronger retrieval of online channel images has a cross-channel and reciprocal effect on overall retailer and offline channel loyalty (Swoboda, Weindel and Schramm-Klein 2016; Van Nierop et al. 2011). Scholars assume that favorable prior offline experiences increase online attitudes, confidence in online information and online behavior (Gensler et al. 2012; Melis et al. 2015). Prior offline experiences affect the image effects on online channel loyalty. We propose the following: H3:

H4:

For retailers with which consumers have more (vs. less) favorable prior offline experiences, the total effects of offline channel image on (a) overall retailer loyalty, (b) offline channel loyalty, and (c) online channel loyalty are stronger. For retailers with which consumers have more (vs. less) favorable prior offline experiences, the total effects of online channel image on (a) overall retailer loyalty, (b) offline channel loyalty, and (c) online channel loyalty are stronger.

Prior Online Experiences: Reciprocal effects of prior online experiences are important because former brick-and-mortar retailers may overcome less favorable online experiences through an attractive offline channel (Wang and Goldfarb 2017). A more favorable prior online experience leads to stronger categorization effects of the online channel as it facilitates cognitive processing (Emrich and Verhoef 2015). However, whether the same occurs for the offline channel is questionable. First, for retailers with which consumers have more favorable prior online experiences, strengthened offline channel image effects on overall retailer and offline channel loyalty are assumed. Favorable prior online experiences facilitate reciprocal offline information processing and elicit stronger offline behavior

2.2 Conceptual Framework and Hypothesis Development

49

(Yang, Lu and Chau 2013). Scholars show that consumers rely on accumulated online experiences for offline information processing (Gensler, Verhoef and Böhm 2012). Stronger reciprocal effects on online channel loyalty are also likely. Experiences accumulated through prior exposure to online stores are transmitted to other categories, e.g., offline stores (Loupiac and Goudey 2019; Yang, Lu and Chau 2013). More experienced consumers use these stronger retrievals and transfer them more easily from one channel to another. This results in a reduction of efforts for related behavior and more elaborate evaluations, which lead to stronger responses (e.g., Melis et al. 2015). Thus, the stronger retrieval of offline channel images has a cross-channel and reciprocal effect on online channel loyalty (Kwon and Lennon 2009b). Second, regarding online channel image, strengthened effects on online channel loyalty are obvious. Favorable experiences support categorization and the use of similar online information, which affect online attitude transfers and online channel loyalty (Dickinger and Stangl 2013; Emrich and Verhoef 2015). Stronger reciprocal effects on overall retailer and offline channel loyalty are likely. Experienced consumers accumulate and more easily retrieve offline channel knowledge. Favorable prior online experiences lead to a stronger retrieval of online channel images. Stronger retrieved online channel images are known to have cross-channel and reciprocal effects on the decision to be loyal towards retailers and offline channels (Kwon and Lennon 2009a; Swoboda, Weindel and Schramm-Klein 2016). We hypothesize the following: H5:

H6:

For retailers with which consumers have more (vs. less) favorable prior online experiences, the total effects of offline channel image on (a) overall retailer loyalty, (b) offline channel loyalty, and (c) online channel loyalty are stronger. For retailers with which consumers have more (vs. less) favorable prior online experiences, the total effects of online channel image on (a) overall retailer loyalty, (b) offline channel loyalty, and (c) online channel loyalty are stronger.

50

2

Study 1: Reciprocity within Major Retail Purchase Channels …

2.3

Empirical Studies

2.3.1

Overview

Two empirical studies were conducted. Study 1 analyzed the reciprocal effects of images between channels and the role of prior experiences in a longitudinal design. Study 2 analyzed reciprocity and its effects with a typical cross-sectional design.

2.3.2

Study 1: Longitudinal Study

2.3.2.1 Sample The German fashion sector was chosen for both studies for several reasons. This sector offers a profound selection of several well-known omni-channel retailers, which helps to avoid single firm-specific results (Landers et al. 2015). Fashion is the third largest retail sector and has the highest online sales share in Germany (24.9 % according to the national retail association, compared to 24.3 % within the larger electronics sector and 8.4 % within the largest food sector; HDE/IFH 2019). The fashion sector is less concentrated, with over 40 firms accounting for two-thirds of total sales (while one electronic firm dominates with 40 % of total sales), and images/brands are important for consumers, who have considerable offline and online experiences. Finally, the largest fashion retailers offer various communication channels and integration services (Acquila-Natale and ChaparroPeláez 2020), and this study focuses on major offline vs. online purchase channels (dominant in multi- and omni-channel decisions, e.g., Bell, Gallino and Moreno 2018; Hult et al. 2019). Pretests were conducted to select retailers with which consumers had more (vs. less) favorable prior offline and online experiences. First, a list of the 12 largest fashion retailers in Germany was compiled. Based on awareness data from a first pretest (N = 130, quota sample), the eight best-known retailers were chosen. In a second pretest, a convenience sample of 20 consumers was used to pre-evaluate whether prior experiences were more or less favorable using a seven-point, fouritem offline and online channel image scale and mean values (ranging from 1 = strongly disagree to 7 = strongly agree): I like the offline/online stores of [retailer]; The offline/online stores of [retailer] are pleasant; I like to shop at the offline/online stores of [retailer]; The offline/online stores of [retailer] are favorable (e.g., Bezes 2014). On this basis, two retailers offering broad assortments (not only fashion products) were eliminated, and the remaining six retailers were

2.3 Empirical Studies

51

pre-categorized into four groups in a matrix (axes: more vs. less favorable prior offline vs. online experiences). These retailers are omni-channel (assured according to Oh, Teo and Sambamurthy 2012). In a final pretest (N = 475, quota sample), respondents again rated the six retailers on the scales. According to the offline channel image, three retailers each were chosen to represent more and less favorable prior offline experience based on the most positive vs. negative mean values relative to the neutral point, 4.0 (p < .05, Mm = 4.6–4.1, Ml = 3.4– 3.0, see Appendix 2.1). The same was done for favorable prior online experience (p < .05 for H0 : μ = 4, Mm = 5.0–4.1, Ml = 2.9–2.5). These procedures guided the choice of the four most heterogeneous retailers that best fit the matrix (three verticals, one discounter with competing store locations). The categorization was verified by objective performance data; two of the retailers grew and two declined in sales (years 2015–2017). To develop the sample in study 1, quota sampling for 600 consumers was employed (i.e., national distribution of the population according to age and gender), who were recruited from a consumer panel and regularly made offline and online fashion purchases. The survey was conducted over a ten-month period in 2017–2018 in three waves four to five months apart and with the same respondents. This period is adequate because inter-purchase times for fashion are known to be short (40 days, HDE/IFH 2019). Trained interviewers conducted face-toface in-home interviews using standardized questionnaires because of a higher data quality and lower non-response bias than web surveys (e.g., Bennink et al. 2013). Lottery low-prize vouchers were used as incentives for participants to complete all waves of the survey (e.g., Beal 2015). In a screening phase prior to the first wave, respondents were asked to list fashion retailers and then to name four retailers at which they often shopped offline and online. The first, second, or third mentioned retailer that fit the matrix was randomly chosen for each respondent to be evaluated in all subsequent waves. Respondents who participated in all three waves and made online and offline fashion purchases were included in the analysis. Eight incomplete cases occurred. The Mahalanobis distance was used to identify outliers, and 19 cases were eliminated. A total of 573 observations per wave remained. Compared to the planned quotas, the 40–65 age group was slightly underrepresented (see Table 2.1). Tests revealed deviation from multivariate normality. A mean-adjusted maximum likelihood estimator (MLM) was chosen to test the hypotheses because it provides robust chi-square tests and handles potential threats within the data structure (Maydeu-Olivares 2017). Chi-square difference tests were conducted using scaling corrections (Satorra and Bentler 2010).

52

2

Study 1: Reciprocity within Major Retail Purchase Channels …

Table 2.1 Sample characteristics of cross-lagged design Realized quota sample (in %)

Planned quota sample (in %)

Male

Male

Female

Female

Total

Total

Fashion sector (N = 573) Age 15-29

20.2

18.9

39.2

19.1

18.3

37.4

Age 30-39

13.8

13.1

26.8

13.2

12.9

26.1

Age 40-65

18.1

15.9

34.0

18.6

17.9

36.5

Total

52.1

47.9

100.0

50.9

49.1

100.0

Source: Own creation

2.3.2.2 Measurement We used scales from previous studies, which were quantitatively pretested in the first pretest with minor changes. The image scales were again used in the main study. Conative loyalty towards the overall retailer and the channel was measured with four items: I am certain that I will shop at [retailer/offline/online channel] again; In the future, I will purchase more at [retailer/offline/online channel] than at any other retailer; I would recommend [retailer/offline/online channel] to friends and others; When I need to make a purchase, [retailer/offline/online channel] is my first choice (e.g., Sirohi et al. 1998; Srinivasan et al. 2002). Covariates were used because loyalty is likely to be affected by gender (0/1 = male/female) and age (Hult et al. 2019). Internet expertise was controlled for because highly experienced consumers are more confident in using online shops (How would you characterize your level of expertise with the Internet?, MontoyaWeiss et al. 2003). Consumers’ familiarity with the retailer was also controlled for (I have often shopped at [retailer], Kent and Allen 1994). The reliability of measurements was ensured across the three time points (e.g., corrected item-to-total correlations and Cronbach’s alpha, see Table 2.2). The values for construct and convergent validity were above the common thresholds. Average variance extracted values exceeded the squared correlations of the constructs and supported discriminant validity (see Table 2.3 and Table 2.4; Fornell and Larcker 1981). The fit values were satisfactory (Hair et al. 2018, p. 93). Common-method variance (CMV) was addressed a priori by an appropriate questionnaire design. A single-factor test was performed. Models with all items loading on a single factor provided significantly worse fit values than the proposed models (overall: χ2 (3) = 2,133.773, p < .001, offline: χ2 (3) = 2,116.100, p < .001, online: χ2 (3) = 2,512.496, p < .001, see Appendix 2.2). The marker variable technique (Lindell and Whitney 2001) was applied using self-efficacy

Offline channel loyalty

Overall retailer loyalty

Online channel image

Offline channel image

.916 .858 .806

OFLOY3 3.6/1.8

OFLOY4 2.4/1.6

.815

OFLOY2 2.9/1.7

2.4/1.6

LOY4

.854

.907

.761 .781

3.6/1.8

LOY3

.763 .791

.893

.947

.949

.941 .862

.924

.899

.947

.953 .817

3.9/1.6

3.9/1.7

4.0/1.6

4.2/1.5

4.2/1.5

4.3/1.4

2.6/1.6

3.6/1.8

3.0/1.6

.742

.810

.839 2.7/1.7

3.6/1.8

3.1/1.6

.723 .897 4.6/2.0

.749

.807

.829

.725 .896 4.5/2.0

.876

.923

.923

.916 .963 4.4/1.4

.904

.880

.919

.841

.880

.929

.753 .792

.840

.886

.907

.874 .790

.920

.941

.944

.946 .856

.924

.940

.961

Time point three

3.9/1.6

3.9/1.7

4.0/1.7

4.2/1.5

4.2/1.5

4.3/1.4

2.7/1.6

3.5/1.8

3.0/1.6

.779

.834

.859

2.8/1.7

3.5/1.8

3.1/1.6

.721 .908 4.5/2.0

.775

.837

.837

.715 .904 4.5/2.0

.903

.921

.921

.923 .966 4.3/1.4

.910

.923

.940

.844

.914

.938

.757 .810

.850

.909

.909

.751 .804

.932

.946

.958

.940 .861

.939

.931

.960

.916

.910

.970

.972

α

(continued)

.789

.865

.876

.729

.789

.858

.846

.722

.916

.928

.937

.921

.924

.915

.940

.937

KMO ItTC

.956 .840

MV/Std. FL

.935 .971 4.0/1.7

KMO ItTC α

.955 .843

MV/Std. FL

Time point two .926 .962 4.0/1.6

KMO ItTC α

OFLOY1 4.7/2.1

2.9/1.6

LOY2

4.2/1.5

ON4

4.7/2.1

4.2/1.5

ON3

LOY1

4.4./1.4

3.8/1.6

OF4

ON2

3.8/1.7

OF3

4.4/1.4

4.0/1.6

ON1

4.0/1.6

OF2

MV/Std. FL

Time point one

OF1

Construct Item

Table 2.2 Reliability and validity of cross-lagged panel models

2.3 Empirical Studies 53

.715 .810 .829 .758 .687

ONLOY2 2.5/1.5

ONLOY3 2.9/1.7

ONLOY4 2.2/1.5

.782

.830

.854 2.3/1.4

2.9/1.6

2.5/1.5 .824

.895

.896

Time point three

.778

.847

.843 2.3/1.5

2.7/1.6

2.5/1.4 .850

.932

.932

.8081

.887

.883

.789

KMO ItTC

.817 .835

MV/Std. FL

.792 .917 2.9/1.8

KMO ItTC α

.831 .828

MV/Std. FL

Time point two .807 .918 3.0/1.8

KMO ItTC α

ONLOY1 3.0/1.8

MV/Std. FL

Time point one

.929

α

Source: Own creation

2

Notes: OF = Offline channel image, ON = Online retailer image, LOY = Overall retailer loyalty, OFLOY = Offline channel loyalty, ONLOY = Online channel loyalty, MV/Std. = Mean values and standard deviations, FL = Factor loadings (exploratory), KMO = Kaiser-Meyer-Olkin Criterion (≥ .5), ItTC = Item-to-Total Correlation (≥ .3), α = Cronbach’s alpha (≥. 7).

Online channel loyalty

Construct Item

Table 2.2 (continued)

54 Study 1: Reciprocity within Major Retail Purchase Channels …

Offline channel loyalty

Overall retailer loyalty

Online channel image

Offline channel image

CR

CR

λ CR

λ CR

λ CR

Time point three λ CR

λ

.885

.902

OF4

.903

.885

.961 .897

.875

.965 .916

.831

.966 .915

.931

.966 .912

.926

.969 .933

.924

.965

.932

.923

.966

.929

.917

.968

.876 .869 .859

OFLOY2

OFLOY3

OFLOY4

.807

LOY4 .903 .773

.872

LOY3

.890

.940

.951

OFLOY1

.871

LOY2

.889

ON4

.900 .777

.940

ON3

LOY1

.950

ON2 .890

.941

.950

.843

.888

.885

.904 .757

.917

.937

.946

.859

.877

.910

.913 .752

.918

.938

.947 .917

.938

.946

.855

.906

.893

.916 .762

.930

.941

.958

.867

.901

.923

.921 .753

.930

.941

.958

(continued)

.930

.941

.958

.964 .948 .964 .947 .964 .947 .967 .951 .967 .950 .967 .950 .970 .945 .970 .945 .970 .944

.961

OF3

ON1

λ

.962 .967 .962 .967 .961 .969 .969 .963 .970 .964 .970 .964 .972 .963 .972 .963 .972 .964

CR

λ

λ

Time point two

CR

λ

CR

Time point one

OF2

OF1

Construct Item

Table 2.3 Reliability and validity of cross-lagged panel models II

2.3 Empirical Studies 55

CR

λ

.907 .838 .902 .865 .842

ONLOY1

ONLOY2

ONLOY3

ONLOY4

CR

CR

λ

λ

Time point two

CR

λ

CR

Time point one λ

λ

.844

.888

.886

.915 .822

CR

CR

λ CR

Time point three λ

Source: Own creation

λ

.862

.919

.926

.924 .816

CR

2

Notes: OF = Offline channel image, ON = Online channel image, LOY = Overall retailer loyalty, OFLOY = Offline channel loyalty, ONLOY = Online channel loyalty, CR = Composite reliability (≥.6), λ = Standardized factor loadings (confirmatory) (≥.5), SCF = Scaling correction factor for MLM.

Online channel loyalty

Construct Item

Table 2.3 (continued)

56 Study 1: Reciprocity within Major Retail Purchase Channels …

2.3 Empirical Studies

57

Table 2.4 Discriminant validity of cross-lagged panel models Constructs

Time point one

Time point two

Time point three

1

1

1

2

1

Offline channel image

.866

2

Online channel image

.364

.870

3

Overall retailer loyalty

.666

.305

1

Offline channel image

.866

2

Online channel image

.364

.870

3

Offline channel loyalty

.659

.293

1

Offline channel image

.862

2

Online channel image

.358

.870

3

Online channel loyalty

.387

.348

3

2

3

.894

.699

.407

.879

.709

.392

.719

.430

.891

.682

.342

.737

.897

.407

.880

.701

.334

.731

.892

.748

3

.897

.894

.701

2

.430

.892

.646

.310

.753

.895

.403

.880

.392

.375

.744

.428

.891

.399

.333

.783

Confirmatory model fits: Time point one (Overall retailer loyalty): CFI .936, TLI .917, RMSEA .130, SRMR .047, χ2 (51) = 546.749, SCF = 1.15. Time point two (Overall retailer loyalty): CFI .948, TLI .933, RMSEA .118, SRMR .042, χ2 (51) = 459.111, SCF = 1.20. Time point three (Overall retailer loyalty): CFI .957, TLI .944, RMSEA .111, SRMR .041, χ2 (51) = 414.077, SCF = 1.17. Time point one (Offline channel loyalty): CFI .934, TLI .915, RMSEA .131, SRMR.046, χ2 (51) = 554.436, SCF = 1.17. Time point two (Offline channel loyalty): CFI .949, TLI .934, RMSEA .117, SRMR. 043, χ2 (51) = 453.417, SCF = 1.21. Time point three (Offline channel loyalty): CFI .955, TLI .942, RMSEA .114, SRMR .043, χ2 (51) = 432.814, SCF = 1.41. Time point one (Online channel loyalty): CFI .961, TLI .950, RMSEA .096, SRMR .033, χ2 (51) = 321.629, SCF = 1.26. Time point two (Online channel loyalty): CFI .970, TLI .961, RMSEA .087, SRMR .026, χ2 (51) = 273.642, SCF = 1.23. Time point three (Online channel loyalty): CFI .975, TLI .968, RMSEA .082, SRMR .025, χ2 (51) = 249.878, SCF = 1.19. Notes: AVE = Average Variance Extracted (≥ 5), SCF = Scaling correction factor for MLM. Values in italics represent squared correlations between constructs, values in bold represent the AVE of the construct. Source: Own creation

58

2

Study 1: Reciprocity within Major Retail Purchase Channels …

(measured with three items: I am able to achieve the goals I have set for myself; Compared to other people, I can do most tasks very well; Even when things are tough, I can perform well, Chen et al. 2001). Self-efficacy is theoretically unrelated to our constructs and is equivalently vulnerable to the same causes of CMV because it is similar to our constructs in content and format (Simmering et al. 2015). The tests indicated no significant changes in the coefficients or correlations (method variance < 14.4 %, Williams and McGonagle 2016). The probability of CMV is reduced.

2.3.2.3 Method To reduce the possibility of omitted variables, endogeneity was tested with the instrumental variable (IV) method (Antonakis et al. 2014). Perceived offline and online channel attributes were used as IVs for offline and online channel images (offline attributes: The assortment at the [retailer’s] store is very good; I think the prices at the [retailer’s] store are always reasonable; I like the store layout of the [retailer’s] offline store very much; The [retailer’s] store has excellent advertising, adapted from Chowdhury et al. 1998; online attributes: The website of [retailer] makes the products look very appealing; The website of [retailer] is well organized; The website of [retailer] is easy to use; The website of [retailer] is informative, Kwon and Lennon 2009a). These attributes are suitable IVs because of their antecedent role in offline and online store images (Bezes 2014). F-tests showed that the IVs were strong predictors (see Appendix 2.3). In addition to the efficient models, consistent models including the IVs were calculated and did not significantly differ (Hausman 1978, z-values < 1.96). Endogeneity was not a serious problem. Measurement invariance was determined to ensure the comparability of the results across the time points (Van de Schoot et al. 2012). The results indicated a good fit for all models (overall: χ2 (540) = 1,527.592, p > .05, offline: χ2 (540) = 1,506.523, p > .05, online: χ2 (540) = 1,155.085, p > .05, see Appendix 2.4). Cross-lagged structural equation modeling using Mplus was applied. This method facilitates the analysis of reciprocal effects and is based on two assumptions: a variable Xt can be predicted by Xt−1 , and Xt can be influenced in a cross-lagged manner by other variables Yt−1 (Zyphur et al. 2019, see Appendix 2.5). All fit values were satisfactory. Next, the results using standardized coefficients are presented.

2.3 Empirical Studies

59

2.3.2.4 Results Manipulation Check. T-tests based on data from the screening phase prior to the first wave supported the experience categorization. The results showed higher mean values for more (vs. less) favorable categorization regarding prior offline channel image (Mm = 4.7–4.5 vs. Ml = 3.9–3.3, above/below the neutral point, p < .05 for H0: μ = 4.0) and online channel image (Mm = 4.9–4.6 and Ml = 3.6–3.2, p < .05 for H0: μ = 4.0). A 2 × 2 MANOVA underscored the differences (p < .001) in main effects for more (vs. less) favorable prior experiences (Wilks’ λ = .965/.744, F = 2.5/24.4). Finally, ANOVAs showed significant main and interaction effects (see Appendix 2.1). Hypotheses Tests. Models 1–3 support H1 (see Table 2.5): Offline channel image positively affects online channel image (overall: β1–2 = .103, p < .001, β2–3 = .115, p < .001, offline: β1–2 = .097, p < .001, β2–3 = .108, p < .001, online: β1–2 = .098, p < .001, β2–3 = .108, p < .001). Simultaneously, the reverse effect exists (overall: β1–2 = .032, p < .05, β2–3 = .035, p < .01, offline: β1–2 = .034, p < .05, β2–3 = .038, p < .05, online: β1–2 = .035, p < .01, β2–3 = .039, p < .01). Models 1–3 support H2a/b: The total effect of offline vs. online channel image on overall and offline channel loyalty is stronger over time (overall: β = .209, p < .01 vs. β = .068, p < .05, offline: β = .206, p < .001 vs. β = .064, p < .01). T-tests confirmed these differences. Surprisingly, offline channel image affects online channel loyalty more strongly (online: β = .131, p < .001 vs. β = .048, p < .05), rejecting H2c. The offline channel remains the most knowledgeable for multi- and omni-channel retailers, and consumers draw stronger inferences to be loyal to online channels. Models 1–3 in Table 2.6 support H3a-c. The total effects of offline channel image on loyalty are stronger for retailers with which consumers have more (vs. less) favorable prior offline experiences (overall: βm = .277, p < .001, βl = .189, p < .001, difference test p < .01, offline: βm = .258, p < .001, βl = .199, p < .001, p < .01, online: βm = .170, p < .001, βl = .107, p < .01, p < .01). H4a-b are supported due to the significantly stronger total effects of online channel image depending on more (vs. less) favorable prior offline experiences (overall: βm = .100, p < .001, βl = .073, p < .01, difference test p < .05, offline: βm = .102, p < .001, βl = .038, p > .10, p < .01). H4c is rejected. The total effects on online channel loyalty are only marginally significant and do not differ significantly (βm = .063, p < .10, βl = .061, p < .10, p > .05). Consumers draw inferences based on online channel images, whereas prior offline experiences do not affect the decision to be loyal to online channels.

60

2

Study 1: Reciprocity within Major Retail Purchase Channels …

Table 2.5 Results of general cross-lagged panel models Model 1: Overall retailer loyalty

Model 2: Offline loyalty

Model 3: Online loyalty

β

p

β

β

p

Offline CI (1) → Online CI (2)

.103

***

.097

***

.098

***

Online CI (1) → Offline CI (2)

.032

*

.034

*

.035

**

Offline CI (1) → LOY (2)

.121

***

.120

***

.072

***

Online CI (1) → LOY (2)

.040

**

.038

**

.027

*

Offline CI (1) → Offline CI (2)

.852

***

.850

***

.848

***

Online CI (1) → Online CI (2)

.734

***

.737

***

.737

***

LOY (1) → LOY (2)

.708

***

.702

***

.738

***

Offline CI (2) → Online CI (3)

.115

***

.108

***

.108

***

Online CI (2) → Offline CI (3)

.035

**

.038

*

.039

**

Offline CI (2) → LOY (3)

.133

***

.131

***

.081

***

Online CI (2) → LOY (3)

.045

*

.043

**

.032

*

Offline CI (2) → Offline CI (3)

.913

***

.910

***

.909

***

Online CI (2) → Online CI (3)

.841

***

.846

***

.845

***

LOY (2) → LOY (3)

.758

***

.754

***

.812

***

R2

.879

***

.867

***

.788

***

Offline CI (1) → LOY (3)

.209

**

.206

***

.131

***

Online CI (1) → LOY (3)

.068

*

.064

**

.048

*

Diff. in total effects

t = 5.036**

t = 5.217**

t = 3.795**

Gender (1) → LOY (1)

.003

ns

−.010

ns

.012

ns

Gender (2) → LOY (2)

.003

ns

−.010

ns

.013

ns

Gender (3) → LOY (3)

.003

ns

−.010

ns

.012

ns

Age (1) → LOY (1)

.032

**

.011

ns

.004

ns

Age (2) → LOY (2)

.031

**

.011

ns

.004

ns

Age (3) → LOY (3)

.032

**

.011

ns

.004

p

Direct effects

LOY (3)

Total effects

Covariates

ns (continued)

2.3 Empirical Studies

61

Table 2.5 (continued) Model 1: Overall retailer loyalty

Model 2: Offline loyalty

Model 3: Online loyalty

β

β

β

p

p

p

Internet expertise (1) → LOY (1)

.024

*

.007

ns

.040

**

Internet expertise (2) → LOY (2)

.023

*

.006

ns

.037

**

Internet expertise (3) → LOY (3)

.024

*

.006

ns

.040

**

Familiarity (1) → LOY (1)

.157

***

.170

***

.039

**

Familiarity (2) → LOY (2)

.144

***

.154

***

.035

**

Familiarity (3) → LOY (3)

.154

***

.163

***

.039

** χ2 (1009)

Structural model fits: Model 1: CFI .935, TLI .931, RMSEA .066, SRMR .079, = 3506.910, SCF = .87. Model 2: CFI .933, TLI .930; RMSEA .066, SRMR .078, χ2 (1009) = 3547.310, SCF = .89. Model 3: CFI .949, TLI .946, RMSEA .056, SRMR .063, χ2 (1009) = 2835.315, SCF = .88. Notes: CI = Channel image, LOY = Loyalty, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 573. Standardized coefficients are shown. Differences between total effects have been tested using t-tests. ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001. Source: Own creation

Models 4–6 support H5a-c. The total offline channel image effects on loyalty are stronger for retailers with which consumers have more (vs. less) favorable prior online experiences (overall: βm = .314, p < .001, βl = .090, p < .10, difference test p < .01, offline: βm = .291, p < .001, βl = .118, p < .001, p < .01, online: βm = .240, p < .001, βl = .069, p < .01, p < .01). H6a and H6c are supported due to the stronger effects of online channel image on loyalty for more (vs. less) favorable prior online experiences (overall: βm = .066, p < .05, βl = .042, p > .10, difference test p < .05, online: βm = .065, p < .05, βl = .043, p > .10, p < .01). H6b is rejected. The total online channel image effects on offline channel loyalty differ but are minor significant and are stronger for less favorable experiences (βm = .036, p > .10, βl = .051, p < .10, p < .05). Retailers with which consumers have more favorable prior online experiences do not benefit from online effects when trying to retain consumers offline. Consumers draw inferences based on offline images to be loyal to offline channels.

**

**

Online CI .033 (1) → Offline CI (2)

.099

.045

Offline CI (1) → LOY (2)

Online CI (1) → LOY (2)

**

***

ns

Online CI .036 (2) → Offline CI (3)

.121

−.002

Offline CI (2) → LOY (3)

Online CI (2) → LOY (3)

***

***

Offline CI .104 (2) → Online CI (3)

Online CI .845 (2) → Online CI (3)

***

.649

LOY (1) → LOY (2)

***

***

Online CI .720 (1) → Online CI (2)

Offline CI .902 (2) → Offline CI (3)

***

Offline CI .804 (1) → Offline CI (2)

.834

.930

.056

.176

.036

.134

.727

.750

.903

.091

.173

.033

.134

***

***

**

***

**

***

***

***

***

***

***

**

***

ns

ns

**

**

ns

ns

**

ns

**

*

*

ns

ns

.847

.901

− .022

.087

.036

.099

.634

.722

.806

.043

.125

.032

.078

***

***

ns

***

**

***

***

***

***

†(.061)

***

**

**

p

.836

.929

.081

.162

.035

.132

.731

.749

.903

.069

.166

.032

.131

β

***

***

***

***

**

***

***

***

***

**

***

**

***

p

More favorable

ns

ns

***

***

ns

**

**

ns

**

ns

ns

ns

ns

p

Diff.

.846

.900

.012

.080

.037

.101

.708

.721

.805

.037

.050

.033

.079

β

***

***

ns

**

**

***

***

***

***

ns

ns

**

**

p

Less favorable

.836

.929

.044

.096

.036

.132

.761

.749

.901

.050

.108

.033

.131

β

***

***

*

***

**

***

***

***

***

*

***

**

***

p

More favorable

Model 3: Online channel loyalty

(continued)

ns

ns

ns

ns

ns

**

**

ns

**

ns

ns

ns

ns

p

Diff.

2

*

**

β

Less favorable

p

Diff.

β

β p

More favorable

Less favorable

p

Model 2: Offline channel loyalty

Model 1: Overall retailer loyalty

Offline CI .079 (1) → Online CI (2)

Direct effects

Prior Offline Experiences

Table 2.6 Results of cross-lagged panel moderator models

62 Study 1: Reciprocity within Major Retail Purchase Channels …

.842

.073

Online CI (1) → LOY (3)

−.017

−.015

−.017

.013

.011

.012

.000

.000

.000

.177

.150

.160

Gender (1) → LOY (1)

Gender (2) → LOY (2)

Gender (3) → LOY (3)

Age (1) → LOY (1)

Age (2) → LOY (2)

Age (3) → LOY (3)

Internet expertise (1) → LOY (1)

Internet expertise (2) → LOY (2)

Internet expertise (3) → LOY (3)

Familiarity (1) → LOY (1)

Familiarity (2) → LOY (2)

Familiarity (3) → LOY (3)

Covariates

.189

Offline CI (1) → LOY (3)

Total effects

R2 LOY (3)

**

***

***

***

ns

ns

ns

†(.075)

†(.075)

†(.075)

**

**

**

***

***

.125

.117

.120

.000

.000

.000

.012

.012

.011

−.015

−.014

−.014

.100

.277

.915

.749

.739

LOY (2) → LOY (3)

***

Model 1: Overall retailer loyalty

Prior Offline Experiences

Table 2.6 (continued)

***

***

***

ns

ns

ns

†(.077)

†(.076)

†(.076)

**

**

**

***

***

***

***

*

**

ns

.168

.155

.191

.007

.006

.008

.011

.011

.012

−.015

−.015

−.017

.038

.199

.827

.761

***

***

***

ns

ns

ns

†(.095)

†(.095)

†(.095)

*

*

*

ns

***

***

***

.132

.123

.128

.007

.006

.006

.012

.011

.011

−.015

−.014

−.014

.102

.258

.913

.760

***

***

***

ns

ns

ns

†(.096)

†(.097)

†(.096)

*

*

*

***

***

***

***

Model 2: Offline channel loyalty

**

**

ns

.034

.031

.036

.010

.010

.011

−.019

−.018

−.019

−.012

−.011

−.012

.061

.107

.780

.810

***

***

***

ns

ns

ns

*

*

*

ns

ns

ns

†(.064)

**

***

***

.028

.026

.027

.011

.010

.010

−.020

−.019

−.019

−.011

−.011

−.011

.063

.170

.796

.806

*

*

*

ns

ns

ns

*

*

*

ns

ns

ns

†(.061)

***

***

***

Model 3: Online channel loyalty

(continued)

ns

**

ns

2.3 Empirical Studies 63

.047

Online CI (1) → LOY (2)

***

.660

.762

.169

Online CI (1) → Online CI (2)

LOY (1) → LOY (2)

Offline CI (2) → Online CI (3)

**

***

***

Offline CI .811 (1) → Offline CI (2)

*

*

.078

Offline CI (1) → LOY (2)

***

*

.158

.058

.580

.686

.874

.061

.246

.029

.159

β

**

***

***

***

*

***

*

***

p

More favorable

β

p

Less favorable

Model 4: Overall retailer loyalty

**

**

ns

ns

ns

*

ns

ns

p

.168

.711

.660

.813

.063

.127

.026

.156

β

***

***

***

***

*

***

*

***

p

Diff. Less favorable

.054

.585

.689

.875

.041

.258

.029

.154

β

**

***

***

***

ns

***

*

***

p

More favorable

Model 5: Offline channel loyalty

**

**

ns

ns

ns

ns

ns

ns

p

.163

.758

.660

.807

.061

.062

.026

.154

β

***

***

***

***

*

*

**

***

p

Diff. Less favorable

.056

.562

.690

.875

.082

.206

.029

.156

β

*

***

***

***

*

***

*

***

p

More favorable

Model 6: Online channel loyalty

(continued)

**

**

ns

**

ns

†(.073)

ns

ns

p

Diff.

2

Online CI .027 (1) → Offline CI (2)

Offline CI (1) → Online CI (2)

Direct effects

Prior Online Experiences

Table 2.6 (continued)

64 Study 1: Reciprocity within Major Retail Purchase Channels …

***

.083

.020

Offline CI (2) → LOY (3)

Online CI (2) → LOY (3)

.827

.849

LOY (2) → LOY (3)

R2 LOY (3)

Gender (1) → LOY (1)

−.015

.042

Online CI (1) → LOY (3)

Covariates

.090

Offline CI (1) → LOY (3)

Total effects

***

.763

Online CI (2) → Online CI (3)

ns

†(.056)

***

***

***

Offline CI .900 (2) → Offline CI (3)

ns

*

*

−.014

.066

.314

.915

.786

.894

.922

.020

.135

.029

Model 4: Overall retailer loyalty

Online CI .029 (2) → Offline CI (3)

Prior Online Experiences

Table 2.6 (continued)

*

*

***

***

***

***

***

ns

***

*

*

**

*

**

ns

ns

*

ns

−.015

.051

.118

.855

.835

.764

.900

.291

.871

.804

.896

.920

†(.053) −.013

†(.098) .036

***

***

***

***

***

.018

ns

−.011

.029

†(.099) .093

*

.049

.028

Model 5: Offline channel loyalty

†(.052)

ns

***

***

***

***

***

ns

*

*

*

**

ns

ns

ns

ns

ns

ns

−.007

.043

.069

.720

.827

.765

.898

−.026

.029

.029

ns

ns

**

***

***

***

***

ns

ns

*

−.007

.065

.240

.856

.851

.894

.920

.021

.072

.029

Model 6: Online channel loyalty

ns

*

***

***

***

***

***

ns

**

*

(continued)

**

**

ns

ns

ns

ns

ns

ns

2.3 Empirical Studies 65

ns

ns

ns

.012

.011

.012

−.003

−.003

−.003

.163

Age (2) → LOY (2)

Age (3) → LOY (3)

Internet expertise (1) → LOY (1)

Internet expertise (2) → LOY (2)

Internet expertise (3) → LOY (3)

Familiarity (1) → LOY (1)

***

.140

−.003

−.003

−.004

†(.087) .011

†(.086) .011

***

ns

ns

ns

†(.086)

†(.086)

†(.086)

.183

−.002

−.002

−.002

.011

.010

.011

***

ns

ns

ns

ns

ns

ns

.141

−.002

−.002

−.002

.010

.010

.009

†(.053) −.013

***

ns

ns

ns

ns

ns

ns

†(.053)

†(.052)

ns

ns

−.011

−.012

.035

.014

.013

−.012

−.012

−.012

−.007

−.007

**

.033

†(.057) .016

†(.057) .016

†(.057) .016

ns

−.011

.013

ns

ns

−.007

−.006

Model 6: Online channel loyalty

**

†(.058)

†(.058)

†(.061)

ns

ns

ns

ns

ns

(continued)

2

†(.085) .011

−.015

Age (1) → LOY (1)

−.014 *

*

−.015

Gender (3) → LOY (3)

†(.053) −.014

−.013

*

*

−.014

Gender (2) → LOY (2)

−.014

Model 5: Offline channel loyalty

Model 4: Overall retailer loyalty

Prior Online Experiences

Table 2.6 (continued)

66 Study 1: Reciprocity within Major Retail Purchase Channels …

.157

Familiarity (3) → LOY (3)

.130

***

***

.172

.146

***

***

.150

.153

Model 5: Offline channel loyalty

***

***

.035

.030

**

**

.034

.035

Model 6: Online channel loyalty

**

**

Source: Own creation

Notes: CI = Channel image, LOY = Loyalty, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, Standardized coefficients are shown. N-Offline (286 less favorable, 285 more favorable), N-Online (284 less favorable, 287 more favorable). Differences in effects between groups tested using χ2 tests of difference. ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001.

Structural model fit: Model 1: CFI .922, TLI .919, RMSEA .079, SRMR .100, χ2 (1900) = 5309.364, SCF = .78. Model 2: CFI .923, TLI .919, RMSEA .078, SRMR .100, χ2 (1900) = 4086.386, SCF = .79. Model 3: CFI .938, TLI .936, RMSEA .068, SRMR .084, χ2 (1900) = 4421.349, SCF = .79. Model 4: CFI .909, TLI .905, RMSEA .081, SRMR .170, χ2 (1894) = 5482.639, SCF = .78. Model 5: CFI .910, TLI .906, RMSEA .080, SRMR .169, χ2 (1894) = 5387.256, SCF = .79. Model 6: CFI .926, TLI .923, RMSEA .071, SRMR .163, χ2 (1894) = 4622.412, SCF = .79.

***

.141

.137

Familiarity (2) → LOY (2)

***

Model 4: Overall retailer loyalty

Prior Online Experiences

Table 2.6 (continued)

2.3 Empirical Studies 67

68

2

Study 1: Reciprocity within Major Retail Purchase Channels …

2.3.2.5 Alternative Models Alternative models strengthen the observations. First, offline and online retail brand equity (RBE) were used as consumer associations (of retailers’ stores as a strong, attractive, unique, and favorable brand, Keller 2010): (Offline/online) store of [retailer] is a strong brand; (Offline/online) store of [retailer] is a well-known brand; (Offline/online) store of [retailer] is an attractive brand; (Offline/online) store of [retailer] is a unique brand (Verhoef, Langerak, et al. 2007). The same tests for reliability, validity, measurement invariance, CMV, and endogeneity were conducted (see Appendix 2.6). Similar results are found for H1–2 (see Table 2.7). Offline RBE positively affects online RBE (overall: β1–2 = .149, p < .001, β2–3 = .158, p < .001, offline: β1–2 = .150, p < .001, β2–3 = .160, p < .001, online: β1–2 = .148, p < .001, β2–3 = .157, p < .001), and vice versa (overall: β1–2 = .062, p < .001, β2–3 = .066, p < .001, offline: β1–2 = .059, p < .01, β2–3 = .063, p < .01, online: β1–2 = .064, p < .001, β2–3 = .068, p < .01). The image results are confirmed (similar to those for overall retailer loyalty by Swoboda, Weindel and Schramm-Klein 2016). Offline vs. online RBE has a stronger total effect on overall and offline channel loyalty (β = .316, p < .001 vs. β = .095, p < .01, β = .337, p < .001 vs. β = .056, p < .10, significant differences). The effect on online channel loyalty is weaker (β = .097, p < .05 vs. β = .107, p > .05). Identical results emerge for H3a-c (see Appendix 2.6). The total effects of offline RBE are stronger for retailers with which consumers have more (vs. less) favorable prior offline experiences (overall: βm = .649, p < .001, βl = .532, p < .001, difference test p < .01, offline: βm = .621, p < .001, βl = .590, p < .001, p < .05, online: βm = .107, p < .05, βl = .098, p < .01, p > .05). Stronger total effects of online RBE on overall retailer and offline channel loyalty depend on more (vs. less) favorable prior offline experiences (βm = .193, p < .001, βl = .185, p > .10, p < .05 and βm = .205, p < .001, βl = .104, p < .01, p < .01). The effect on online channel loyalty is similar for more (vs. less) favorable prior offline experiences (βm = .404, p < .10, βl = .406, p < .10, p > .05). Finally, H5a-c are supported. The total effects of offline RBE on loyalty are stronger for retailers depending on more (vs. less) favorable prior online experiences (overall: βm = .531, p < .001, βl = .499, p < .001, difference test p < .01, offline: βm = .566, p < .001, βl = .504, p < .001, p < .01, online: βm = .204, p < .001, βl = .138, p < .001, p < .01). Stronger effects of online RBE on overall retailer and online loyalty occur due to more (vs. less) favorable prior online experiences (βm = .206, p < .001, βl = .175, p > .001, p < .01 and βm = .327, p < .001, βl = .246, p > .001, p < .01) but not for offline loyalty (βm = .153, p > .10, βl = .143, p > .10, p > .05). Therefore, H6a/c are supported, while H6b is not.

2.3 Empirical Studies

69

Table 2.7 Results of alternative general cross-lagged panel models Model 1: Overall retailer loyalty

Model 2: Offline channel loyalty

Model 3: Online channel loyalty

β

p

β

p

β

p

Offline RBE (1) → Online RBE (2)

.149

***

.150

***

.148

***

Online RBE (1) → Offline RBE (2)

.062

***

.059

**

.064

**

Offline RBE (1) → LOY (2)

.205

***

.220

***

.051

*

Online RBE (1) → LOY (2)

.058

**

.030

ns

.065

*

Offline RBE (1) → Offline RBE (2)

.839

***

.842

***

.836

***

Online RBE (1) → Online RBE (2)

.725

***

.723

***

.726

***

LOY (1) → LOY .578 (2)

***

.584

***

.722

***

Offline RBE (2) → Online RBE (3)

.158

***

.160

***

.157

***

Online RBE (2) → Offline RBE (3)

.066

***

.063

**

.068

**

Offline RBE (2) → LOY (3)

.215

***

.230

***

.055

*

Online RBE (2) → LOY (3)

.062

**

.032

ns

.071

*

Direct effects

(continued)

70

2

Study 1: Reciprocity within Major Retail Purchase Channels …

Table 2.7 (continued) Model 1: Overall retailer loyalty

Model 2: Offline channel loyalty

Model 3: Online channel loyalty

β

p

β

p

β

p

Offline RBE (2) → Offline RBE (3)

.873

***

.876

***

.870

***

Online RBE (2) → Online RBE (3)

.784

***

.780

***

.782

***

LOY (2) → LOY .617 (3)

***

.631

***

.797

***

R2 LOY (3)

.874

***

.872

***

.792

***

Offline RBE (1) → LOY (3)

.316

***

.337

***

.097

*

Online RBE (1) → LOY (3)

.095

**

.056

† (.064)

.107

**

Diff. in total effects

t = 7.379**

t = 9.560**

Gender (1) → LOY (1)

.001

ns

−.015

ns

.013

ns

Gender (2) → LOY (2)

.001

ns

−.016

ns

.012

ns

Gender (3) → LOY (3)

.001

ns

−.016

ns

.014

ns

Age (1) → LOY (1)

.034

*

.012

ns

−.002

ns

Age (2) → LOY (2)

.033

*

.012

ns

−.002

ns

Age (3) → LOY (3)

.033

*

.012

ns

−.002

ns

Total effects

t = 1.125 ns

Covariates

Internet expertise .017 (1) → LOY (1)

ns

−.002

ns

.037

*

Internet expertise .016 (2) → LOY (2)

ns

−.002

ns

.034

* (continued)

2.3 Empirical Studies

71

Table 2.7 (continued) Model 1: Overall retailer loyalty

Model 2: Offline channel loyalty

Model 3: Online channel loyalty

β

p

β

p

β

Internet expertise .016 (3) → LOY (3)

ns

−.002

ns

.037

*

Familiarity (1) → LOY (1)

.256

***

.260

***

.072

***

Familiarity (2) → LOY (2)

.231

***

.231

***

.065

***

Familiarity (3) → LOY (3)

.245

***

.243

***

.073

***

p

Structural model fits: Model 1: CFI .918, TLI .914, RMSEA .070, SRMR .062, χ2 (1009) = 3847.824, SCF = .90. Model 2: CFI .916, TLI .912, RMSEA .061, SRMR .070, χ2 (1009) = 3856.525, SCF = .91. Model 3: CFI .930, TLI .926, RMSEA .063, SRMR .052, χ2 (1009) = 3270.553, sSCF = .91. Notes: RBE = Retail brand equity, LOY = Loyalty, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, Standardized coefficients are shown. N = 573, Differences between total effects have been tested using t-tests. ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001. Source: Own creation

Second, we control for perceived levels of channel integration, which may affect consumer loyalty in omni-channel retailing (e.g., Zhang et al. 2018). Absolute difference scores for perceived assortment, promotion and ease of transaction service were built and measured as follows: The variety of products at the [retailer’s] offline/online store is sufficient; [Retailer’s] offline/online store has informative promotion; It looks easy to find what I am looking for at [retailer’s] offline/online store (e.g., Bezes 2014). The covariates are almost all insignificant, and the results and all coefficients in the reciprocal models are stable (see Table 2.8). Loyalties are not affected.

72

2

Study 1: Reciprocity within Major Retail Purchase Channels …

Table 2.8 Results of general cross-lagged panel models with perceived levels of channel integrationn Model 1: Overall retailer loyalty

Model 2: Offline loyalty

Model 3: Online loyalty

β

β

β

p

p

p

Direct effects Offline CI (1) → Online CI (2)

.104

***

.098

***

.098

***

Online CI (1) → Offline CI (2)

.032

**

.035

**

.036

**

Offline CI (1) → LOY (2)

.115

***

.108

***

.065

***

Online CI (1) → LOY (2)

.043

**

.043

**

.029

*

Offline CI (1) → Offline CI (2)

.852

***

.849

***

.848

***

Online CI (1) → Online CI (2)

.734

***

.738

***

.738

***

LOY (1) → LOY (2)

.710

***

.708

***

.740

***

Offline CI (2) → Online CI (3)

.115

***

.108

***

.109

***

Online CI (2) → Offline CI (3)

.035

**

.038

**

.039

**

Offline CI (2) → LOY (3)

.126

***

.118

***

.074

***

Online CI (2) → LOY (3)

.049

**

.048

**

.034

*

Offline CI (2) → Offline CI (3)

.913

***

.909

***

.908

***

Online CI (2) → Online CI (3)

.841

***

.846

***

.845

***

LOY (2) → LOY (3)

.759

***

.759

***

.813

***

R2

.877

***

.866

***

.786

***

.199

***

.187

***

.119

***

LOY (3)

Total effects Offline CI (1) → LOY (3)

(continued)

2.3 Empirical Studies

73

Table 2.8 (continued)

Online CI (1) → LOY (3) Diff. in total effects

Model 1: Overall retailer loyalty

Model 2: Offline loyalty

Model 3: Online loyalty

β

β

β

p .073

**

t = 4.309**

p .072

**

t = 4.225**

p .051

*

t = 4.258**

Covariates Gender (1) → LOY (1)

.002

ns

−.011

ns

.013

ns

Gender (2) → LOY (2)

.002

ns

−.011

ns

.012

ns

Gender (3) → LOY (3)

.003

ns

−.012

ns

.014

ns

Age (1) → LOY (1)

.031

**

.011

ns

.006

ns

Age (2) → LOY (2)

.031

**

.011

ns

.006

ns

Age (3) → LOY (3)

.032

**

.011

ns

.006

ns

Internet expertise (1) → LOY (1)

.025

*

.009

ns

.041

**

Internet expertise (2) → LOY (2)

.024

*

.008

ns

.039

**

Internet expertise (3) → LOY (3)

.025

*

.009

ns

.042

**

Familiarity (1) → LOY (1)

.158

***

.169

***

.040

**

Familiarity (2) → LOY (2)

.145

***

.154

***

.037

**

Familiarty (3) → LOY (3)

.156

***

.164

***

.040

**

Perceived integration level of assortment → LOY (3)

−.015

ns

−.020

ns

−.023

ns

Perceived integration level of promotion → LOY (3)

−.001

ns

−.011

ns

.032

**

(continued)

74

2

Study 1: Reciprocity within Major Retail Purchase Channels …

Table 2.8 (continued) Model 1: Overall retailer loyalty β Perceived integration level of ease of transaction service → LOY (3)

β

p

−.015

Model 2: Offline loyalty

β

p

−.024

ns

Model 3: Online loyalty −.013

ns

p ns

Structural model fits: Model 1: CFI .926, TLI .923, RMSEA .071, SRMR .097, χ2 (1114) = 4331.062, SCF = .77. Model 2: CFI .925, TLI .921; RMSEA .072, SRMR .097, χ2 (1114) = 4370.487, SCF = .78. Model 3: CFI .939, TLI .936, RMSEA .063, SRMR .082, χ2 (1114) = 3609.702, SCF = .77. Notes: CI = Channel image, LOY = Loyalty, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 573. Standardized coefficients are shown. Differences between total effects have been tested using t-tests. Ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001. Source: Own creation

2.3.3

Study 2: Cross-sectional Study

2.3.3.1 Sample, Measurement and Method To develop the sample, interviews were conducted at one point in time using the same procedure as in study 1 for 400 different consumers. Twenty cases were omitted in the outlier analysis, leading to 380 observations. With respect to the intended quotas, the 15–29 age group was slightly underrepresented (see Table 2.9). The data deviated from normality, and the MLM estimator was chosen. Table 2.9 Sample characteristics of cross-sectional design Realized quota sample (in %)

Planned quota sample (in %)

Male

Male

Female

Total 26.1

Female

Total

Fashion sector (N = 380) Age 15–29

11.8

11.8

23.7

13.3

12.8

Age 30–39

8.4

31.6

17.6

9.2

9.0

18.2

Age 40–65

27.1

31.6

58.7

28.0

27.7

55.7

Total

47.4

52.6

100.0

50.5

49.5

100.0

Source: Own creation

2.3 Empirical Studies

75

The same scales as in study 1 were used to measure offline and online channel image and overall retailer, offline channel, and online channel loyalty (with satisfactory values for reliability, validity and discriminant validity; see Table 2.10 and Table 2.11). The same covariates were used. To test the hypothesis, non-recursive structural equation modeling using Mplus was applied. Non-recursive models need to be identified by IVs (offline and online channel attributes of study 1) and require a disturbance correlation between the constructs (which include reciprocity, Nagase and Kano 2017). To account for endogeneity, offline and online channel trust were used as antecedents of the attributes ([Retailer’s] offline/online store is trustworthy, Dickinger and Stangl 2013, see Appendix 2.7). The proposed models with reciprocity showed satisfactory fit values (see Table 2.12). Rival models 1 and 2 included each channel image as an antecedent of the other image (mediation); rival model 3 did not include any offline-online-image relationship.

2.3.3.2 Results First, Table 2.12 visualizes that in the proposed reciprocal models, offline channel image positively affects online channel image (overall: β = .214, p < .05, offline: β = .215, p < .01, online: β = .226, p < .01), while online does not affect offline channel image (overall: β = .065, p > .05, offline: β = .064, p > .05, online: β = .068, p > .05). H1 is not supported. However, offline (vs. online) channel image more strongly affects overall retailer and offline channel loyalty (β = .612, p < .001 vs. β = .200, p < .01, difference test p < .01, β = .628, p < .001 vs. β = .186, p < .05, p < .01) and equally affects online channel loyalty (β = .330, p < .001 vs. β = .359, p < .001, p > .05), an effect opposite that observed in study 1 and proposed in H2c. Second, rival model 1 shows significant unidirectional relationships of offline channel image on online channel image, while rival model 2 shows the opposite relationship. However, the results change in at least three ways. • The unidirectional relationships of offline and online channel image are equally strong (rival models 1–2), while those of online image are significantly weaker in the reciprocal models (study 1) and are insignificant in the proposed model (study 2). All models/studies lead to different conclusions regarding the image link between offline and online channels. • The total effects of online and offline channel image are stronger than those in study 1, and offline channel image affects loyalty even more strongly. This occurs in the proposed models and in revised models 1–3 for overall retailer loyalty and offline channel loyalty. The effect of online channel image is weakened (and often insignificant).

Overall retailer loyalty

Online channel attributes

Offline channel attributes

4.2/1.6

ON4

5.1/1.3 4.5/1.4

IV23

IV24 4.2/2.1

5.0/1.3

LOY1

4.4/1.5

IV14

IV22

3.5/1.5

IV13

IV21

4.7/1.5 3.8/1.6

IV12a

3.9/1.5

4.2/1.7

ON3

IV11

4.5/1.6

ON2

3.8/1.8

OF4 4.6/1.6

3.7/1.8

OF3

ON1

4.1/1.8

OF2

.779

.689

.597

.759

.712

.739

.731



.755

.903

.929

.952

.937

.930

.897

.942

.943

FL

.778

.719

.706

.859

.827

KMO

.747

.595

.510

.644

.606

.623

.619



.633

.884

.907

.925

.912

.908

.878

.914

.915

ItTC

.907

.781

.785

.962

.961

α

.924

.781

.785

.963

.961

CR



.956

.764

.575

.613

.815

.717

.761

.747

.901

.926

.952

.943

.918

.892

.952

.949

λ

Overall retailer loyalty model

.768

.750

.963

.961

CR

.764

.575

.613

.815

.715

.758



.751

.901

.926

.952

.943

.917

.891

.952

.950

λ

Offline channel loyalty model

.782

.785

.963

.961

CR

(continued)

.760

.580

.617

.816

.714

.773



.735

.901

.926

.951

.943

.913

.884

.956

.953

λ

Online channel loyalty model

2

Online channel image

4.1/1.7

OF1

Offline channel image

MV/Std.

Item

Construct

Table 2.10 Reliability and validity of cross-sectional models

76 Study 1: Reciprocity within Major Retail Purchase Channels …

2.4/1.6 3.1/1.8 2.1/1.5

ONLOY2

ONLOY3

ONLOY4

2.4/1.6

OFLOY4 3.2/2.0

3.1/1.9

OFLOY3

ONLOY1

2.3/1.6

2.3/1.6

LOY4 3.2/1.9

3.5/1.9

LOY3

OFLOY2

2.7/1.7

LOY2

OFLOY1

MV/Std.

Item

.845

.816

.929

.859

.837

.875

.897

.784

.835

.889

.899

FL

.791

.779

KMO

.793

.786

.864

.820

.774

.832

.826

.751

.774

.843

.831

ItTC

.916

.905

α CR

.725

.991

.773

λ

Overall retailer loyalty model

.920

CR

.724

.970

.788

.948

λ

Offline channel loyalty model

.912

CR

.673

.993

.774

.890

λ

Online channel loyalty model

Source: Own creation

Notes: MV/Std. = Mean values and Standard deviations, FL = Factor loadings (exploratory), KMO = Kaiser-Meyer-Olkin Criterion (≥ .5), ItTC = Item-to-Total Correlation (≥.3), α = Cronbach’s alpha (≥.7), CR = Composite reliability (≥.6), λ = Standardized factor loadings (confirmatory) (≥.5). a Item deleted due to low factor loading.

Online channel loyalty

Offline channel loyalty

Construct

Table 2.10 (continued)

2.3 Empirical Studies 77

78

2

Study 1: Reciprocity within Major Retail Purchase Channels …

Table 2.11 Discriminant validity of cross-sectional models Constructs

1

2

3

4

5

Overall retailer loyalty model 1

Offline channel image

.861

2

Online channel image

.579

.866

3

Offline channel attributes

.780a

.497

.550

4

Online channel attributes

.593a

.745a

.632a

.576

5

Overall retailer loyalty

.563

.361

.607 a

.420

.755

Offline channel loyalty model 1

Offline channel image

.861

2

Online channel image

.579

.866

3

Offline channel attributes

.778a

.496

.500

4

Online channel attributes

.596 a

.745a

.632a

.562

5

Offline channel loyalty

.581

.360

.646 a

.432

.746

Online channel loyalty model 1

Offline channel image

.859

2

Online channel image

.579

.866

3

Offline channel attributes

.780a

.497

.634

.577

.446

.412

4

Online channel attributes

.596

.745a

5

Online channel loyalty

.379

.353

.549 .726 χ2 (137)

Overall retailer loyalty model: CFI .951, TLI .939, RMSEA .071, SRMR .040, = 458.882, SCF = 1.01. Offline channel loyalty model: CFI .953, TLI .942, RMSEA .069, SRMR .040, χ2 (137) = 444.852, SCF = 1.01. Online channel loyalty model: CFI .957, TLI .947, RMSEA .065, SRMR .036, χ2 (137) = 415.479, SCF = 1.01. Notes: AVE = Average variance extracted (≥ .5), values in italics represent squared correlations between constructs, values in bold represent the AVE of the construct, SCF = Scaling correction factor for MLM a For situations in which the criterion of Fornell and Larcker (1981) was violated, the test of Anderson and Gerbing (1988) assured discriminant validity. This procedure yielded satisfactory results as the nested model (the more restrictive model with more degrees of freedom due to successively fixed correlations at value one) fits significantly more poorly (p < .001) than the comparison model. Source: Own creation

2.4 Discussion and Implications

79

• For (III) online channel loyalty, the results change. The proposed model and rival model 3 show an equal (insignificant) total effect of offline and online channel image (no reciprocity), while in study 1, the effect is significantly stronger for offline channel image. In rival models 1–2, the importance of both images changes again depending on the mediation; e.g., the offline-onlineimage (online-offline) link increases the total effect of online (offline) image. The cross-sectional design may bias reciprocity (due to equilibrium or stationarity, Mitchell and Maxwell 2013), but the results point to different conclusions, which are discussed next.

2.4

Discussion and Implications

2.4.1

Overview

This study contributes to understanding of the reciprocity between major purchase channel images (brand equity in alternative models) and their effects on overall retailer, offline and online loyalty. Additionally, prior offline and online experiences change the reciprocal effects. While this study examined only one industry and only four leading omni-channel retailers, the results have important theoretical and managerial implications.

2.4.2

Theoretical Implications

Regarding the first research question, the longitudinal study shows that reciprocal relationships of major offline and online purchase channel images exist and reciprocally affect all loyalties. The results extend the dominant unidirectional and bidirectional studies and provide theoretical implications. First, the results regarding reciprocity indicate that consumers categorize major purchase channels into a corresponding retailer category and reciprocally transfer images between them (e.g., Badrinarayanan et al. 2012). They do so positively for attitudinal channel images and for associative brand equity, which is theoretically notable and promising (e.g., compared to other cognitive theories, Swoboda, Weindel and Schramm-Klein 2016, or categorization rationales in experiments, Kwon and Lennon 2009a). The results are stable in all models: for overall retailer, offline and online channel loyalty, offline image more strongly affects online image than vice versa (extending Swoboda, Weindel and Schramm-Klein 2016).

80

2

Study 1: Reciprocity within Major Retail Purchase Channels …

Unidirectional studies have indicated categorization effects (e.g., Chu et al. 2017; Grosso, Castaldo and Grewal 2018). However, reciprocal effects are often not revealed (Mitchell and Maxwell 2013) but provide different insights; in study 2, online image does not reciprocally affect offline image. Cross-lagged panel models have causal inference capability (e.g., Wiedermann and von Eye 2015; Zyphur et al. 2019). We therefore believe that this study significantly contributes to extant research by revealing the relationships of major purchase channels— which are relevant for multi- and omni-channel retailers—and answering related calls. However, the role of reciprocity beyond the studied (attitudinal/associative) constructs should be questioned. Moreover, it is questionable whether measures of perceived channel integration or congruence (e.g., Shen et al. 2018; Lee et al. 2019) adequately capture reciprocity and theoretically cover the respective cognitive mechanisms in consumers’ minds. Asking consumers whether channels are congruent is unlikely to arouse reciprocal cognitive processes (Emrich and Verhoef 2015). Second, the results regarding the reciprocal effects support our theoretical assumption that both offline and online purchase channel images affect loyalty, i.e., consumers’ category knowledge causes responses. Individuals rely on the most knowledgeable (i.e., representative) category member to draw inferences and use these category inferences for decision-making (e.g., Loken 2006; Sohn 2017). The results support this rationale for offline (vs. online) channel image effects on offline channel and overall retailer loyalty (i.e., consumers mostly rely on offline channel image to guide their loyalty intentions) and, surprisingly, on online loyalty. The observations are stable for brand equity. We may conclude that in today’s retail environment, offline purchase channels are still more knowledgeable and predominant, although loyalty to physical stores and retailers also depends on online channels. However, we also conclude that the results are to be expected when analyzing former brick-and-mortar retailers (Swoboda, Weindel and Schramm-Klein 2016), while for former purely online players (e.g., Apple), opposite insights will occur. Despite the origin, managers must understand the role of reciprocity between channels from the consumer perspective. Of course, consumers switch channels and touchpoints flexibly, but knowledge of the reciprocity of major purchase channels is still predominant (e.g., Wiener, Hoßbach and Saunders 2018). As mentioned, there is a research gap because reciprocity changes the extant results/implications. Based on cross-sectional study 2, which does not support reciprocity, three examples emerge (for dependent variables, model types, and further conceptualizations, see Table 2.12):

.599***

.068ns

.740***

.813***

→ LOY

→ LOY

→ Offline CI

→ Online CI

Offline CI

Online CI

IV1

IV2

.612***

.200**

→ LOY

→ LOY

Offline CI

Online CI

Total effects

.546***

.065ns

→ Offline CI

Online CI

R2 loyalty

.214*

→ Online CI

Offline CI

Direct effects

.236***

.599***

.540***

.869***

.690***

.069ns

.599***

– –

.279***

.222***

.615***

.540***

.869***

.691***

.051ns

.615***

.278**

– –

βp

.081*

.593***

.539***

.887***

.930***

.083*

.593***

– –

– –

βp

.186*

.628***

.546***

.813***

.740***

.051ns

.616***

.064ns

.215**

βp

βp

βp

.222***

.615***

.540***

.869***

.691***

.051ns

.615***

– –

.278***

βp

Rival 1

Offline channel loyalty Proposed

Rival 3

Rival 1

Proposed

Rival 2

Overall retailer loyalty

Table 2.12 Results of proposed and rival cross-sectional models

.059ns

.625***

.540***

.653***

.917***

.059ns

.610***

.262***

– –

βp

Rival 2

.064ns

.608***

.540***

.887***

.931***

.064ns

.608***

– –

– –

βp

Rival 3

.359***

.330***

.418***

.811***

.726***

.284***

.306***

.068ns

.226*

βp

.371***

.307***

.411***

.870***

.681***

.284***

.307***

– –

.284***

βp

Rival 1

Online channel loyalty Proposed

.289***

.379***

.411***

.654***

.912***

.289***

.303***

.263***

– –

βp

Rival 2

(continued)

.290***

.307***

.409***

.888***

.926***

.290***

.307***

– –

– –

βp

Rival 3

2.4 Discussion and Implications 81

−.070*

−.035ns

.326***

→ LOY

→ LOY

→ LOY

Age

Internet expertise

Familiarity

.326***

−.035ns

−.070*

.037ns

t=21.502**

.327***

−.034ns

−.073*

.017ns

t=23.279**

.325***

−.035ns

−.069*

.037ns

t=24.624**

βp

.327***

−.034ns

−.074*

.017ns

t=8.378**

βp

.327***

−.034ns

−.073*

.017ns

t=23.279**

βp

.327***

−.034ns

−.074*

.017ns

t=22.099**

βp

Rival 2

.326***

−.035ns

−.073*

.017ns

t=24.824**

βp

Rival 3

.266***

.012ns

−.116***

.075*

t=1.100ns

βp

.266***

.012ns

−.146***

.075*

t=1.970*

βp

Rival 1

Online channel loyalty Proposed

.266***

.011ns

−.146***

.075*

t=2.954**

βp

Rival 2

.265***

.011 ns

−.146***

.075*

t=1.628ns

βp

Rival 3

Source: Own creation

Notes: CI = Channel image, LOY = Loyalty, IV = Instrumental variable, Prop. Model = Proposed Model, Ri. Model = Rival Model, SCF = Scaling correction factor for MLM, N = 380. Standardized coefficients are shown. Differences between total effects tested using t-tests. ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001.

2

Structural model fits: Overall retailer loyalty: Prop. model 1: CFI .926, TLI .913, RMSEA .076, SRMR .115, χ2 (211) = 748.261, SCF = 1.04, Ri. model 1: CFI .926, TLI .914, RMSEA .081, SRMR .115, χ2 (213) = 749.388, SCF = 1.04. Ri. model 2: CFI .927, TLI .916, RMSEA .081, SRMR .115, χ2 (213) = 740.000, SCF = 1.03. Ri. model 3: CFI .923, TLI .911, RMSEA .083, SRMR .116, χ2 (214) = 772.680, SCF = 1.04. Offline channel loyalty: Prop. model: CFI .927, TLI .915, RMSEA .075, SRMR .115, χ2 (211) = 738.964, SCF = 1.03, Ri. model 1: CFI .927 TLI .916, RMSEA .081, SRMR .115, χ2 (213) = 740.000, SCF = 1.03. Ri. model 2: CFI .926, TLI .914, RMSEA .081, SRMR .116, χ2 (213) = 749.447, SCF = 1.03. Rival model 3: CFI .924, TLI .913, RMSEA .082, SRMR .116, χ2 (214) = 763.400, SCF = 1.03.. Online channel loyalty: Prop. model: CFI .932, TLI .920, RMSEA .072, SRMR .111, χ2 (211) = 698.554, SCF = 1.02, Ri. model 1: CFI .932, TLI .921, RMSEA .078, SRMR .111, χ2 (213) = 699.595, SCF=1.02. Rival model 2: CFI .931, TLI .920, RMSEA .078, SRMR .111, χ2 (213) = 708.553, SCF = 1.02. Ri. model 3: CFI .929, TLI .918, RMSEA .079, SRMR .112, χ2 (214) = 723.538, SCF = 1.02.

.037ns

→ LOY

t=7.746**

Gender

Covariates

Diff. in total effects

βp

βp

βp

Rival 1

Offline channel loyalty Proposed

Rival 3

Rival 1

Proposed

Rival 2

Overall retailer loyalty

Table 2.12 (continued)

82 Study 1: Reciprocity within Major Retail Purchase Channels …

2.4 Discussion and Implications

83

• Cross-sectional studies tend to overestimate offline (online) channel image effects in the models for overall and offline loyalty (online loyalty). While this study examines conative loyalty, the logic may hold for action loyalty (e.g., spending, Ailawadi and Farris 2017). • The conceptualization of mediations, i.e., offline-online-image links or onlineoffline-image links (rival models 1 or 2), tends to underestimate online effects for overall retailer and offline (not online) loyalty. Model 3 without interdependencies does not consider directionality. • We suggest that studying only online channels (and possibly further touchpoints) overestimates their effects because strong offline channels are not conceptualized (e.g., in multi- and omni-channels, Herhausen et al. 2015; Murray, Elms and Teller 2017). Therefore, a theory-based conceptualization of reciprocity is valuable; otherwise, unobserved effects occur (Chu et al. 2017). For example, bidirectional studies account for interdependent cross-channel effects but not for reciprocity (e.g., Breugelmans and Campo 2016). They theoretically imply piecemeal (i.e., isolated) information processing. Cross-sectional studies need to theoretically embed and empirically control such theoretical mechanisms (e.g., Saghiri et al. 2017). Options include IVs and tests for mutual interactions between channels; however, these do not completely capture the direction of causality. Thus, we call for the use of cross-lagged panel models that incorporate time lags and autoregressive relationships to improve the modeling of causal inferences of multi- and omni-channel reciprocal effects (e.g., Wiedermann and von Eye 2015). Regarding the second research question, study 1 shows that prior experiences affect the reciprocal effects of major offline and online purchase channel images on overall retailer, offline and online channel loyalty (extending Badrinarayanan et al. 2012). It appears that consumers consider experience and knowledge for efficient processing to support categorization and reciprocal links between images of channels in decisions (Lemon and Verhoef 2016; Yang et al. 2020). More (vs. less) favorable prior experiences strengthen (weaken) the reciprocal effects because they facilitate the use of favorable knowledge and elicit stronger responses (which are also stable for associative brand equities). The core implications are as follows.

84

2

Study 1: Reciprocity within Major Retail Purchase Channels …

• Prior offline experiences change offline channel image effects (the strongest lever for loyalty). For retailers with more (vs. less) favorable prior offline experiences, significantly stronger total offline channel image effects on overall retailer, offline and online channel loyalty occur. Respective online channel image effects occur (though not on online loyalty). • Prior online experiences affect offline channel image effects. For retailers with more favorable prior online experiences, significantly stronger total offline channel image effects on overall retailer, offline and online channel loyalty emerge. For less favorable prior online experiences, respective effects emerge (though not for offline loyalty). We conclude that consumers’ prior experiences provide competitive advantages to some retailers—especially via offline channel images, the strongest lever in our models—and limit other retailers that use reciprocity to retain consumers. These levers apply to reciprocity and prior offline and online experiences. However, opposite results for previous purely online players may emerge. Despite the origin, managers need to know the role of prior experience in the reciprocity of major purchase channels; otherwise, they ignore certain consumer inferences (e.g., Loupiac and Goudey 2019). Less favorable prior experiences, for example, weaken coordination mechanisms (due to the largely nonexistent or insignificant online channel image effects; see Table 2.5).

2.4.3

Managerial Implications

Known and carefully managed reciprocal channel images help retailers retain loyal consumers. Managers who are unaware of reciprocity between major purchase channels may make ineffective marketing investments, e.g., impacting the wrong channels or behaviors (Hult et al. 2019). We believe that the results are relevant for former brick-and-mortar and former purely online players (that increasingly open offline stores, e.g., Amazon, Valentini, Neslin and Montaguti 2020). Because retailers still often have siloed structures (i.e., online and physical store divisions operate independently; Yrjölä et al. 2018), managers should know how one channel affects the others. A triple strong offline-to-online vs. online-to-offline link emerges in this study, which is a strong lever for marketing activities (e.g., investments in offline-online integration, communication at respective touchpoints, Lee et al. 2019; Herhausen et al. 2019), and an alternative to pure channel harmonization perspectives (e.g., Emrich and Verhoef 2015).

2.5 Limitations and Directions for Further Research

85

The relationships were shown for attitudinal/associative constructs, which strongly affect consumers’ store choice. Collaborative strategic decision-making by the separate divisions may therefore rely on these constructs to better match online and physical stores (Acquila-Natale and Chaparro-Peláez 2020). Managers may aim to increase loyalty to the overall retailer, to the offline channel (as is currently the major lever), or to the online channel (as is likely to be the focus in the future). In all cases, knowledge of the reciprocal strength of offline and online purchase channels is useful. The stronger effect of offline channel images on all types of loyalty shows that physical stores continue to play a key role in multi- and omni-channel retailing (in line with other sectors, e.g., banking, Cambra-Fierro et al. 2020). Instead of closing offline stores that do not perform well according to offline metrics, managers may actively pursue strategies to foster the physical store image and consider its importance for online stores (Bell, Gallino and Moreno 2018). However, the latter might not be an opportunity for all managers because of the role of consumers’ prior experiences, which influence current evaluations and future experiences (Grewal and Roggeveen 2020). Advantages emerge for firms with favorable prior offline/online experiences (due to the reinforcing effects of reciprocity over time). Retailers with less favorable prior offline and online experience cannot rely on significant online-offline reciprocity in most cases and may overestimate it (Gensler, Verhoef and Böhm 2012). They need a strategic consideration of how to overcome these disadvantages, e.g., by increasing initial confidence by using further touchpoints and strengthening experience in the targeted channel (Grewal and Roggeveen 2020; Herhausen et al. 2015). Monitoring how consumers’ prior offline/online experiences account for their short-term and long-lasting effects (e.g., dashboards including experiences, Lemon and Verhoef 2016).

2.5

Limitations and Directions for Further Research

This study has certain limitations that point to future research directions. Although special attention was paid to the data collection, using quota sampling, reasonable choices of one sector and leading omni-channel retailers, broadening the database will allow further conclusions. Studying online-affine consumer groups (e.g., Herhausen et al. 2019), other sectors (e.g., Swoboda, Weindel and Schramm-Klein 2016) or former online pure players may change the results. We choose retailers to study categorization and reciprocity (unknown retailers may increase internal validity), but we could not consider consumers’ switching behavior between touchpoints. Broadening the channel scope to further

86

2

Study 1: Reciprocity within Major Retail Purchase Channels …

analyze reciprocity (e.g., of social media and online stores, Sohn 2017) will add to extant research by revealing inference mechanisms at the touchpoint level. Regarding the measures, enhancing conative loyalty by objective purchase data for channels is promising (but difficult to realize, Yrjölä, Spence and Saarijärvi 2018). Although channel image is carefully captured, there is no common agreement on a measure (except for brand equity). Future research may study the reciprocity of fine-grained constructs (e.g., marketing-mix, Grosso, Castaldo and Grewal 2018). This study preselected retailers according to the more (vs. less) favorable prior experiences of consumers. Future measures might allow tests for continuous moderations (e.g., Melis et al. 2015) or lagged counts of experiences (e.g., Grewal and Roggeveen 2020), which account for further dynamics between channels. The conceptual framework was based on cognitive reasoning, but it would be interesting to explain reciprocal effects affectively. Emotions or motivations may enhance online effects because affective behavior is strongly linked to online channels (e.g., Verhagen, van Dolen and Merikivi 2019). Studying reciprocity between further communication touchpoints or their antecedent role is promising in our framework but is challenging in cross-lagged models. However, studies may adjust channel effects, as mentioned for omni- (but not multi-) channel retailers (e.g., promotion increases online effects, Valentini, Neslin and Montaguti 2020). Similarly, studying perceived levels of integration (e.g., Zhang et al. 2018) as antecedents of reciprocal effects is challenging. Theoretically, an increasing level of integration will strengthen reciprocity between channels due to forced categorybased knowledge transmission (Lee et al. 2019).

3

Study 2: Effects of Perceived Offline-Online and Online-Offline Channel Integration Services

3.1

Introduction

Firms use channel integration services, i.e., technology-based solutions that allow customers to execute different interactive activities across physical and online channels (e.g., functional and informational integration, Banerjee 2014; Sousa and Amorim 2018), to provide consumers a seamless experience. These services provide consumers knowledge about and ease of access to a channel, i.e., they make a channel more transparent and salient which increases the purchase intention in a channel (Herhausen et al. 2015; Lee et al. 2019. OF-ON services support consumers in offline venues to interact with an online channel, e.g., see/order articles not physically available, whereas ON-OF services support consumers to, e.g., pick up/return online purchased articles in offline channels (e.g., Hamouda 2019; Hossain et al. 2020). However, consumers value integration services differently (Akturk, Ketzenberg and Heim 2018; Li et al. 2018; Pan, Wu and Olson 2017). For example, fashion shoppers prefer ON-OF services the most (clickand-collect/-return, EY Parthenon 2019). Moreover, omni-channel firms that offer a seamless experience by integrating all channels (Hosseini et al. 2018) still find it challenging to detect the services for an appropriate integration that increase channel quality and behavioral outcomes (Banerjee 2014; Bolton et al. 2018). While previous research stresses the importance of integration quality, we show how to reach it (namely through offering integration services). Therefore, we examine the indirect effects (i.e., mediation paths) of perceived OF-ON and ON-OF services on purchase intention through perceived quality of offerings in major sales channels.

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 A. Winters, Omni-Channel Retailing, Handel und Internationales Marketing Retailing and International Marketing, https://doi.org/10.1007/978-3-658-34707-9_3

87

88

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

In research, integration and management across channels have a high priority (Chen et al. 2018). Studies have mostly shown indirect effects of perceived integration services on consumer behavior (e.g., loyalty, purchase or repurchase intention, Bendoly et al. 2005; Lee et al. 2019; Murfield et al. 2017; see Figure 3.1). They have often studied general channel evaluations that translate integration services into consumer behavior, e.g., offline channel image or quality (Oh and Teo 2010; Schramm-Klein et al. 2011; Zhang et al. 2018) or online channel quality or usefulness (Herhausen et al. 2015; Lee et al. 2019; Shen et al. 2018; Yang et al. 2020). The mediators have cross-channel effects on offline or online consumer behavior in omni-channel firms (e.g., Melis et al. 2015; Sousa and Amorim 2018), which has not been simultaneously examined in former studies. More importantly, most scholars have focused on the joint effects of integration services (e.g., Hamouda 2019; Seck and Philippe 2013) or have addressed only OF-ON or ON-OF integration (e.g., Bhargave, Mantonakis and White 2016; Collier and Kimes 2013 or Jara et al. 2018; Kleinlercher et al. 2018). They do not yield insights into the effects of the offered but distinctly perceived OF-ON and ON-OF services. Effect on various outcomes Direct Indirect Cao and Li (2015); Luo, Fan and Zhang (2015); Bendoly et al. (2005); Chiu et al. (2011); Frasquet Melis et al. (2015); Seck and Philippe (2013); and Miquel (2017); Hamouda (2019); Hossain et al. Tagashira and Minami (2019) (2020); Li et al. (2018); Lee et al. (2019); Oh and Teo 2010; Oh, Teo and Sambamurthy (2012); SchrammKlein et al. (2011); Shen et al. (2018); Yang et al. (2017); Zhang et al. (2018) Offline-to-online Bhargave, Mantonakis and White (2016); Collier and Kimes (2013); Mosquera et al. (2018); Patrício et al. (2008) Online-to-offline Akturk, Ketzenberg and Heim (2018); Gallino Herhausen et al. (2015); Murfield et al. (2017); Yang and Moreno (2014); Gao and Su (2017); Jara et al. (2020) et al. (2018); Kleinlercher et al. (2018); Vyt, Jara and Cliquet (2017) Both This study Notes: Studies in italic analyze performance (not behavioral) outcomes. Directions of integration services

Joint perspective

Figure 3.1 Literature review on integration services. (Source: Own creation)

Scholars have called for respective research (e.g., Shen et al. 2018; Zhang et al. 2018). We simultaneously study the effects of the OF-ON and ON-OF services that are most important to consumers. Analyzing ramifications of integration services can change the results of the studies that analyze only the joint effects or ON-OF services (e.g., Hamouda 2019). Insights into the respective effects for offline and online purchase intentions are crucial for managers who aim to attract consumers in sales channels (Acquila-Natale and Iglesias-Pradas 2020; Bolton et al. 2018).

3.1 Introduction

89

Therefore, this study aims to answer the research question of whether and how perceived OF-ON vs. ON-OF services are transformed into offline and online channel purchase intentions. Additionally, we question how two important context factors moderate these effects. Consumers’ online shopping experience is an important moderator for the integration service effects (e.g., Herhausen et al. 2015; Shen et al. 2018). Channel congruence is also important, but its role in the perceived OF-ON and ON-OF services’ mediation paths is unexplored (e.g., Acquila-Natale and Iglesias-Pradas 2020; Vyt, Jara and Cliquet 2017). Accordingly, we offer three contributions. First, we contribute to the extant research by studying the effects of major OF-ON and ON-OF services from consumers’ view, as both are important in omni-channel firms (Hosseini et al. 2018). Both are offered to consumers but differently affect offline and online purchase intentions (Bhargave, Mantonakis and White 2016; Li et al. 2018). Moreover, we initially conceptualize the indirect effects of both integration services on purchase intention through perceived quality of offerings (also testing direct effects). Finally, we add to research by applying the accessibility-diagnosticity theory to the mediation paths across major sales channels (Feldman and Lynch 1988). For consumers, accessible integration service cues (inputs) become relative diagnostics for purchase intentions via the perceived quality of offerings in sales channels. In omni-channel firms, crosschannel links are fundamental. For example, OF-ON services provide consumers knowledge about and ease of access to the online channels’ quality, i.e., they make this channel more salient and increase purchase intentions there (e.g., Herhausen et al. 2015; Lee et al. 2019). Studying the cross-channel effects of both integration services on both purchase intentions provides finer-grained implications for managers. Second, consumers’ online shopping experience, i.e., the extent of their knowledge/experience of searching/purchasing products online (Melis et al. 2015), is a possible diagnostic cue for decision-making. Higher levels of experience are known to reduce the effects of ON-OF services (Herhausen et al. 2015). For OFON and ON-OF services, we do not know if this result holds (e.g., Shen et al. 2018). We contribute to our knowledge by analyzing if the original inputs, i.e., the integration services’ relative diagnosticity for purchase intentions, change when a possible superior cue, i.e., online shopping experience, is available. We provide insights into the conditional effects, i.e., continuous latent moderations on respective mediation paths, by using LMS (Cheung and Lau 2017). Third, firms aim to achieve congruent channels, i.e., synchronizing structures/offerings across channels (e.g., Bezes 2013). Perceived channel congruence results in holistic views of consumers, whereby a lower congruence leads to

90

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

piecemeal channel evaluations (Hammerschmidt et al. 2016; Wang, Beatty and Mothersbaugh 2009). We examine how perceived congruence as a possible superior cue moderates the services-purchase intentions links. Perceived congruence can reduce or increase the importance of the perceived quality of offerings and thus the effects of integration services on purchases intentions (Emrich, Paul and Rudolph 2015; Falk et al. 2007). The remainder of this study proceeds as follows. By drawing from theory, we derive hypotheses and test them based on 722 consumer evaluations of omnichannel fashion firms in a cross-sectional design. After presenting the results, we provide implications and directions for further research.

3.2

Conceptual Framework and Hypothesis Development

3.2.1

Overview

To address our aims, we build on accessibility-diagnosticity theory and studies on the effects of integration services on consumer behavior. We study the (total) indirect effects of perceived OF-ON and ON-OF services (see Figure 3.2) on offline and online purchase intention (i.e., the likelihood that consumers use firms’ channels to purchase, Hult et al. 2019). Integration services (sometimes called multi-channel services in service research) support activities across channels. They enhance transparency and point consumers to offerings available in other channels (Banerjee 2014; Sousa and Amorim 2018). As previous research stresses the importance of integration quality, we show how to reach an appropriate integration. We study the services that are most important to consumers (see examples in Table 3.2): OF-ON services including the perceived access to online stores, product information in online stores, and helpful employees when using its online store; ON-OF services including information about available products offline, pick up and return of products in offline stores. The perceived quality of offline and online offerings represents consumers’ channel-specific evaluation of quality, value and benefits (Acquila-Natale and Iglesias-Pradas 2020; Hammerschmidt, Falk and Weijters 2016). The mediation paths are moderated by consumers’ online shopping experiences and perceived channel congruence.

3.2 Conceptual Framework and Hypothesis Development

91

Consumers’ online shopping experience

Perceived offlineto-online services

Perceived quality of offline offerings

Perceived onlineto-offline services

Perceived quality of online offerings

Purchase intention Offline channel Online channel

Perceived channel congruence

Figure 3.2 Conceptual framework. (Source: Own creation)

3.2.2

Theory

Accessibility-diagnosticity theory provides a cognitive explanation for our research aims (Feldman and Lynch 1988). This theory suggests that the likelihood of using a cue or input for decision-making depends on the specific input’s accessibility and its relative diagnosticity. Accessibility refers to the ease and extent of retrieving a cue. This input is more likely to be used for evaluation when it is easily perceived as useful and easily recalled from the mind. Relative diagnosticity refers to the extent to which the inferences based on this input can be used to make a decision, i.e., the degree to which the input is decision relevant (Lynch, Marmorstein and Weigold 1988). In decisions, all available cues are utilized, although their relevance for evaluation differ. The most accessible, diagnostic input, i.e., the superior cue, will be more used (Menon and Raghubir 2003). Boundary conditions as further diagnostic cues can change the relevance. But, even if such a higher-order cue is available, the lower-order cue still operates and is attenuated with the availability of a higher-order cue (Feldman and Lynch 1988). In our omni-channel context, multiple cues exist. The probability that OF-ON and ON-OF services are used to evaluate firms’ channels in purchase decisions is a function of the accessibility and relative diagnosticity of these cues. Although integration services are accessible, they may not be useful in a given evaluation or not equally relevant for offline or online purchase intentions (i.e., one dominates, Alba and Hutchinson 1987; Lynch, Marmorstein and Weigold 1988). They become relatively diagnostic by affecting the perceived quality of offerings

92

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

(rendering this information relevant for purchases). The latter is known to have channel-specific relevance, e.g., perceived quality of offline offerings will contribute most to offline purchase intentions. Whether perceived OF-ON and ON-OF services enhance perceived qualities and purchase intentions is the subject of the next steps in our analysis. Boundary conditions will moderate the mediation paths. The relative diagnosticity of the original input (integration services) diminishes if a more diagnostic, superior cue is available (Feldman and Lynch 1988). For example, individuals tend to rely on their experience for evaluations and rate further accessible cues as less useful (Schwarz 2004). For consumers with higher levels of online shopping experience, perceived OF-ON services might lose their relative diagnosticity in purchase decisions and might just be seen as services added to existing online services (Falk et al. 2007). Higher levels of perceived channel congruence may affect the relative diagnosticity by promoting a more holistic evaluation of channels and their relevance (Bezes 2013; Hammerschmidt, Falk and Weijters 2016). The mediation paths of the integration services are likely to be reduced or increased, as higher levels of channel congruence as a superior cue affect the processing of the service inputs in mind (Alba and Hutchinson 1987; Wang, Beatty and Mothersbaugh 2009). Next, we provide hypotheses for the mediation paths of the integration services and the moderations. We are interested in the cross-channel indirect and total effects (of OF-ON vs. ON-OF services on purchase intentions via the quality of offerings) in offline and online purchase decisions.

3.2.3

Mediation Paths of the Integration Services

We believe that although integration services are easily accessible in consumers’ minds, they are not directly diagnostic for purchase intentions in omni-channel firms (e.g., Li et al. 2018). Thus, we do not hypothesize direct services-purchase intention links but test them. We hypothesize the indirect effects as such mediation paths depend on the integration services’ accessibility and diagnosticity. In offline purchase decisions, OF-ON and ON-OF services become theoretically useful when consumers link them to the quality perceptions of channel offerings. The integration services become a useful piece of information and a relative diagnostic input for decision-making, i.e., relevant to offline purchase intentions (Lynch, Marmorstein and Weigold 1988). However, differences in the mediation paths are likely. OF-ON and ON-OF services lead consumers to rely

3.2 Conceptual Framework and Hypothesis Development

93

on the channel-specific perceived quality of offline offerings (Campo and Breugelmans 2015), and OF-ON services cues dominate in usefulness and relevance. Reduced cross-channel effects from OF-ON and ON-OF services on the perceived quality of online offerings are likely (e.g., Swoboda, Weindel and Schramm-Klein 2016). Moreover, OF-ON services put consumers at ease and make a channel’s quality of offerings more salient (Bhargave, Mantonakis and White 2016). We therefore believe that OF-ON (vs. ON-OF) services more strongly enhance offline purchase intention via the perceived quality of offline (vs. online) offerings. ON-OF services also become more diagnostic via the perceived quality of offline offerings. In online purchase decisions, we also assume differences in the usefulness and relevance of the integration services (Lynch, Marmorstein and Weigold 1988). Here, consumers rely more on the channel-specific quality of online offerings, and ON-OF services are useful in expanding the salience of the quality of offerings (Montoya-Weiss, Voss and Grewal 2003). Cross-channel links through the quality of offline offerings are also likely, as ON-OF services provide consumers with knowledge about offline channel offerings. The mediation path of ON-OF services becomes relatively diagnostic given the channel-specific relevant quality of online (vs. offline) offerings; also, the OF-ON services will proceed more strongly via the perceived quality of online offerings. This theoretical rationale is novel. Empirically, Patrício, Fisk and Falcão e Cunha (2008) showed that OF-ON services are a valuable part of an offline service experience, whereas ON-OF services via online evaluations were weakly linked to offline outcomes (Herhausen et al. 2015). Strong links of ON-OF services to online outcomes are evident. The joint integration were found to affect consumers’ evaluation and offline and online behavior (e.g., Frasquet and Miquel 2017; Yang et al. 2017; Zhang et al. 2018). We propose the following: H1:

H2:

The indirect effects of (a) OF-ON services and (b) ON-OF services on offline purchase intention are stronger through the perceived quality of offline (vs. online) offerings. The indirect effects of (a) OF-ON services and (b) ON-OF services on online purchase intention are stronger through the perceived quality of online (vs. offline) offerings.

94

3.2.4

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

Role of Consumers’ Online Shopping Experience

Consumers’ online shopping experience will moderate the mediation paths. Consumers rely on their online shopping experience as a superior cue, while integration services become less useful. Total effects in parallel mediation models are hypothesized, as consumers’ online shopping experience interferes with the services’ accessibility and relative diagnosticity. In offline purchase decisions, consumers’ online shopping experience helps by providing additional knowledge in decisions, and consumers are less likely to rely on original accessible cues (Menon and Raghubir 2003). Consumers with higher levels of online shopping experience find accessible integration services less useful in assessing the quality of offline and online offerings (Shen et al. 2018). OF-ON and ON-OF services become relatively less diagnostic, i.e., their effects on offline purchase intentions will be reduced. In omni-channel systems, consumers rely on their online shopping experience as a useful and superior cue for their offline purchase intentions (Melis et al. 2015). Experience interferes with the retrieval of accessible cues, consumers rely on it when developing intentions (Schwarz 2004). In online purchase decisions, consumers’ online shopping experience also affects the responses to integration services. Highly experienced consumers are more confident in online shopping environments and less dependent on integration services (Melis et al. 2015). They may see integration as another service that has been added to existing online services (Falk et al. 2007). Theoretically, consumers rate accessible integration services as less useful. The effects of OF-ON and ON-OF services on online purchase intention are reduced. Moreover, when consumers rely more on online shopping experience as a superior cue, the experience itself becomes relatively diagnostic (e.g., Schwarz 2004). Studies partly support our reasoning. Herhausen et al. (2015) show for highly experienced consumers decreasing ON-OF service-online channel quality-channel outcomes-links. Shen et al. (2018) find a diminishing moderation of online usage experience and joint integration services on omni-channel service usage. We hypothesize the following: H3:

H4:

Higher levels of consumers’ online shopping rate the total effects of (a) OF-ON services offline purchase intention. Higher levels of consumers’ online shopping rate the total effects of (a) OF-ON services online purchase intention.

experience negatively modeand (b) ON-OF services on experience negatively modeand (b) ON-OF services on

3.2 Conceptual Framework and Hypothesis Development

3.2.5

95

Role of Channel Congruence

Perceived channel congruence will moderate the mediation paths. The relative diagnosticity mechanism is affected as congruence leads to a more holistic evaluation of channels (Bezes 2013). We hypothesize indirect effects as congruence affects consumers’ use of inputs and the links of the quality of offerings. In offline purchase decisions, a more holistic evaluation evoked by higher levels of perceived channel congruence reinforces the integration services’ relative diagnosticity (Alba and Hutchinson 1987). The perceived quality of online channel offerings gains salience; the former important quality of offline channel offerings loses relevance for evaluation and intentions (Badrinarayanan et al. 2012; Bezes 2013). The effects of OF-ON and ON-OF services through the perceived quality of offline offerings become less useful, i.e., the integration services’ diagnosticity is reduced. In contrast, the effects of the integration services through the perceived quality of online offerings become more useful and relevant. In online purchase decisions, higher levels of channel congruence may similarly reinforce integration services’ diagnosticity through the quality of perceived offline offerings, while a former channel-specific quality of online offerings loses relevance (Alba and Hutchinson 1987). However, in online decisions, channel congruence might lose its effect strength due to a more piecemeal (vs. holistic) channel evaluation (Bezes 2013; Wang, Beatty and Mothersbaugh 2009). We assume an increasing (decreasing) relative diagnosticity and relevance of integration services via the quality of offline (vs. online) offerings. The relative diagnosticity of the paths of OF-ON and ON-OF services via the quality of online offerings is reduced. In contrast, the paths of the services via the quality of offline offerings will increase. This moderator has seldom been studied in our context. Wang, Beatty and Mothersbaugh (2009) show that high channel congruence leads consumers to evaluate offline and online channels more holistically. Emrich, Paul and Rudolph (2015) show that congruent channels positively affect perceptions, which leads to higher levels of purchase intention. In contrast, Carlson and O’Cass (2011) support the decreasing effects on online evaluations-online behavior links. We hypothesize the following: H5:

With higher levels of perceived channel congruence, the indirect effects of OF-ON services and ON-OF services on offline purchase intention are moderated (a) negatively via the perceived quality of offline offerings and (b) positively via the perceived quality of online offerings.

96

H6:

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

With higher levels of perceived channel congruence, the indirect effects of OF-ON services and ON-OF services on online purchase intention are moderated (a) negatively via the perceived quality of online offerings and (b) positively via the perceived quality of offline offerings.

3.3

Empirical Study

3.3.1

Sample Design

For the empirical study, we carefully chose one retail sector, leading firms and face-to-face interviews. We chose fashion retailing for several reasons. It is one of the largest retail sectors in most Western economies with high shares of online or omni-channel sales (40 % of online sales shares are forecast in the U.S. or Germany for 2023, Planet Retail 2019). Consumers buy across channels and take advantage of integration services. For example, in Germany, 50 % of offline sales occur after ON-OF service usage, and 25 % of online sales occur after OF-ON service usage (Planet Retail 2019). Retail firms and verticals dominate here. We selected the leading four omni-channel pure fashion firms in Germany (due to sales in 2017). These firms—two former offline retailers and two verticals—have stores in suburban areas across the country and online stores. They are omni-channel, and their OFON and ON-OF services are likely to be well-known. We ensured the offer of services in the following areas: integrated information access, product/price information, promotions, transactions, order fulfillment (Oh, Teo and Sambamurthy 2012). This procedure is superior to choosing one firm or various firms selected by consumers. To improve data quality and reduce nonresponse bias, we conducted face-toface in-home interviews by using standardized questionnaires (Heerwegh 2009). To develop the sample in 2018, quota sampling was employed for 1,000 consumers (with a national distribution according to age and gender). Respondents were recruited from an existing consumer panel in a midsized city and had to have made offline and online purchases at the selected firms in the last year (Li et al. 2018). We asked individuals to participate by e-mail/phone by providing general information about the research (Zhang and Bloemer 2008). At least 1,550 individuals who followed the intended quotas were contacted until 1,000 agreed to participate. They were visited by 15 trained interviewers during one week (twice when a first attempt failed). About 100 respondents each were not available or

3.3 Empirical Study

97

canceled their appointments. The multi-stage response rate (originally invited) was 50.9 % and 79.0 % for the final-stage. The average interview length was 20 minutes. In a screening phase, respondents were first asked to name and identify the fashion firms they knew (awareness, Badrinarayanan et al. 2012). They next indicated the firms where they had shopped during the last year (based on a list that included the preselected firms and a 5-point Likert-type scale from 1 = I don’t purchase at the firm’s offline/online channel to 5 = I purchase at the firm’s offline/online channel very often). Respondents who had at least occasionally (= 2) shopped at the firms’ offline and online channels were interviewed. 762 respondents had purchased from both channels of at least two of the selected firms. We randomly chose one of both firms for evaluation: better/weaker known for a first/second respondent. Twenty-four incomplete cases emerge. Sixteen cases were omitted based on the Mahalanobis distance. This procedure led to 722 observations. With respect to the quotas, the 15–29 age group was slightly overrepresented; the 40–65 age group was slightly underrepresented (see Table 3.1). Table 3.1 Sample characteristics Realized quota sample (in %)

Planned quota sample (in %)

Male

Male

Female

12.3

11.5

Female

Total

Total

Fashion sector (N = 722) Age 15–29

14.0

13.9

27.9

23.8

Age 30–39

8.3

8.7

17.0

8.2

8.0

16.2

Age 40–69

25.1

30.0

55.1

28.4

31.6

60.0

Total

47.4

52.6

100.0

48.9

51.1

100.0

Source: Own creation

Prior to the analysis, we tested for univariate and multivariate normality. All values indicated that the data were normally distributed. We chose the maximum likelihood estimator.

3.3.2

Measurement

We used the scales from previous studies (all based on seven-point Likert-type scales from 1 = strongly disagree to 7 = strongly agree) and additionally provided pretests.

98

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

Purchase intention was measured for both channels with three items each: “if I find something I like, it is (1) likely, (2) probable, (3) possible, that I will shop at the [firm’s] offline/online store” (Schlosser et al. 2006). Pretests were used to identify the most important integration services. We selected 35 OF-ON and ON-OF services from literature (e.g., Bendoly et al. 2005; Chiu et al. 2011; Oh, Teo and Sambamurthy 2012). Twenty-three graduate students rated the nine best-known OF-ON and ON-OF services (also offered by our firms). These were then evaluated by undergraduate students regarding importance and usefulness (n = 106, see Table 3.2) as follows: “please rate how important (useful) are the following integration services provided by fashion firms in general (1 = not important (useful) at all to 7 = very important (useful)).” A clear preference for three OF-ON services emerged (over the neutral point, Mimportant = 5.4–5.1, p < .05, Museful = 5.8–5.5, for H0 : μ = 4): access to online stores, online products, employee help. These were ranked 1st -3rd place in 59–70 % of the cases (the fourth item, gift coupons for the next purchase, was ranked 1st -3rd in 27 % of the cases). Similarly, three ON-OF services were selected (Mimportant = 6.0–5.8, Museful = 6.2–5.9, p < .05, for H0 : μ = 4): product information, click-and-collect, and returns (ranked 1st -3rd place in 63–74 % of the cases). In the main survey, we asked, “please rate on a seven point scale to what extent [firm] usefully provides the following: when I purchase from [firm’s] offline store, (1) the firm provides access to its online store, (2) I can inform myself about available products in its online store, (3) employees are helpful when using its online store.” For ON-OF services, we asked, “when I purchase from [firm’s] online store (1), I can inform myself about the availability of products in its offline store, (2) I can pick up the product from its offline store, and (3) I can return the product to its offline store.” Perceived quality of offline and online offerings was measured by typical channel attributes in the fashion sector (Acquila-Natale and Iglesias-Pradas 2020; Hammerschmidt, Falk and Weijters 2016). We used one item per attribute for the offline and online channels respectively: “[firm’s] offline/online store has a good variety of products; I like the store layout of [firm’s] offline/online store very much” (Chowdhury, Reardon and Srivastava 1998) and “I think that the prices at [firm’s] offline/online store are always reasonable; [firm’s] offline/online advertising is very informative” (Yoo et al. 2000). Consumers’ online shopping experience was measured with the following three items: “I have often shopped at [firm’s] online store; I am very familiar with [firm’s] online store; I have a lot of experience with [firm’s] online store” (Diamantopoulos et al. 2005). Perceived channel congruence was measured with four items: “the offline and online stores of [firm] are similar; the service/functions in the offline and

3.6/1.8

5.1/1.5 5.2/1.6

2.6/1.3

5.4/1.4

4.5/1.9

The physical store provides Internet kiosks for customers to place orders for out-of-stock items.

The firm provides access to the website within its stores.

The employees at firm’s stores are knowledgeable and helpful regarding the use of the website.

The firm allows customers to access their prior integrated purchase history.

The local store allows checking for inventory status of products available in the online store.

The gift coupons or vouchers issued by the store can be redeemed either online or offline for the next purchase.

The firm advertises the website at 3.0/1.3 its local stores.

Perceived OF-ON services

.632 .534

4.7/1.9

5.8/1.4

2.9/1.3

−2.545 .018

5.789 .000

5.5/1.4

5.5/1.6

3.8/1.8

3.2/1.4

5.298 .000

6.952 .000

.554 .585

−3.219 .004

p

.295 .087

9.660 .000

−1.207 .240

5.058 .000

7.666 .000

1.374 .183

−2.238 .036

Usefulness MV/Std. t-test

MV/Std. t-test

p

Importance

Table 3.2 Selection of perceived integration services (pretest) Stage of purchase

X

X

X

X

X

X

X

(continued)

Pre-purchase Pur-chase Post-purchase

3.3 Empirical Study 99

2.9/1.3

The firm makes future purchase recommendations based on past consolidated online and offline purchases.

3.5/1.8

−5.378 .000

3.1/1.6

6.0/1.2

The website provides interactive access to the customer service assistant through a real-time chat program.

The website allows customers to search for products available in the physical store.

6.2/1.1

4.1/1.7

−3.418 .001

The firm advertises its local stores 3.4/1.6 through the website.

16.787 .000

3.9/1.9

−3.352 .001

3.4/1.3

−3.325 .003

You can receive nonproduct 3.4/1.8 information on firm’s stores (e.g., driving directions) via e-mail contact or other electronic communication made available through firm’s website.

3.8/1.9

.312 .758

19.720 .000

−7.811 .000

.507 .613

−.505 .615

−1.704 .102

.564 .579

p

Stage of purchase

X

X

X

X

X

X

(continued)

Pre-purchase Pur-chase Post-purchase

3

Perceived ON-OF services

3.8/1.9

Usefulness MV/Std. t-test

MV/Std. t-test

p

Importance

The website provides post-purchase services such as support for the products purchased at physical stores.

Table 3.2 (continued)

100 Study 2: Effects of Perceived Offline-Online and Online-Offline …

5.9/1.3

5.8/1.2

3.9/2.0

2.7/1.4

The firm allows consumers to purchase items through firm’s website and pick them up in physical stores.

The firm accept returns at its stores for purchases made through firm’s website.

The in-store customer service center accepts repair or exchange of products purchased online.

After each purchase, the website customizes webpages for customers based on past consolidated online and offline purchases.

6.0/1.2

4.1/2.0

3.0/1.7

−.186 .853

−9.650 .000

5.9/1.3

3.4/1.7

14.497 .000

15.175 .000

−4.163 .000

−5.772 .000

.462 .645

17.609 .000

14.954 .000

−3.546 .001

p

Stage of purchase

X

X

X

X

X

Pre-purchase Pur-chase Post-purchase

Source: Own creation

Notes: N = 106, MV/Std. = Mean value and standard deviation. Items selected for this analysis (in italic) show the most positive mean difference due to t-tests to the neutral point of 4.0. All items measured on 7-point Likert-type scales: 1 = not important (useful) at all, 7 = very important (useful). *** p < .001; ** p < .01; * p < .05; ns = not significant.

3.2/1.8

Usefulness MV/Std. t-test

MV/Std. t-test

p

Importance

The firm allows customers to make payment in the physical store for their online purchases.

Table 3.2 (continued)

3.3 Empirical Study 101

102

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

online stores of [firm] are consistent; the online store represents the offline store of [firm]; the offline and online stores are typical for [firm]” (Badrinarayanan et al. 2012). We additionally obtained an “objective” measure. Six fashion experts (CEOs/professors) evaluated the congruence of the offline and online channels of our firms with six items (Badrinarayanan et al. 2012; Bezes 2013). Higher/lower congruent groups emerge: Mhigh1 = 6.4, Mhigh2 = 5.7 vs. Mlow1 = 3.0, Mlow2 = 3.1. The higher congruent firms were also ranked in 1st and 2nd place in 67–100 % of the cases. We used covariates because offline and online purchase intentions are likely to be affected by gender (0/1 = male/female) and age (Hult et al. 2019). We also controlled for general Internet expertise as users with high expertise are more confident in using online shops (“how would you characterize your level of expertise with the Internet?” Montoya-Weiss, Voss and Grewal 2003). We tested for the nested data structure and found only small intraclass correlations (025/.002 for offline/online purchase intentions). No additional explanations of variance were found; hypotheses were not tested with multilevel modeling (Hox et al. 2017, p. 244). The reliability of the measurements was confirmed see Table 3.3. The values for construct and convergent validity were above the common thresholds. The average variance extracted values exceeded the squared correlations of the constructs and support discriminant validity (Fornell and Larcker 1981, see Table 3.4). The fit values for the confirmatory models were satisfactory. CMV was addressed by using an appropriate questionnaire design. A singlefactor test was performed (with lower fit values than the proposed model: offline and online purchase intention: χ2 (24) = 2,834.24, p < .001, χ2 (24) = 2,868.78, p < .001). We applied the marker variable technique by using selfefficacy. Self-efficacy is theoretically unrelated to our constructs but is similar in content/format and has some correlations with quality measures (Simmering et al. 2015, see Appendix 3.1). It was measured with three items (“I am able to achieve the goals that I have set for myself; compared to other people, I can do most tasks very well; even when things are tough, I can perform quite well,” Chen, Gully and Eden 2001). The highest amount of method variance was 16.80 % (lower than in former research, Williams and McGonagle 2016). The coefficients and correlations showed no significant changes, and CMV appeared not to be a problem in this study.

MV/Std.

FL

KMO

5.7/1.3 5.4/1.4

If I find something I like, it’s probable that I’ll shop at the [firm’s] offline store

If I find something I like, it’s possible, that I’ll shop at the [firm’s] offline store

.924

.919

.909

.772

4.6/1.7 4.3/1.8

If I found something I like, it’s probable that I’ll shop at [firm’s] online store.

If I found something I like, it’s possible, that I’ll shop at [firm’s] online store. 3.7/2.0 4.3/1.8 3.7/1.8

When I purchase from [firm’s] offline store, the firm provides access to its online store.

When I purchase from [firm’s] offline store, I can inform myself about available products in its online store.

When I purchase from [firm’s] offline store, employees are helpful when using its online store.

Perceived OF-ON services

4.0/1.8

If I found something I like, it’s likely that I’ll shop at [firm’s] online store.

.728

.717

.920

.954

.934

.914

.690

.772

Online purchase intention (adapted from Schlosser, Barnett White and Lloyd (2006))

5.3/1.5

If I find something I like, it’s likely that I’ll shop at the [firm’s] offline store.

Offline purchase intention (adapted from Schlosser, Barnett White and Lloyd (2006))

Table 3.3 Reliability and validity

.657

.650

.762

.916

.901

.887

.881

.878

.871

ItTC

.829

.953

.939

α

OFPI

.835



.897

CR



λ

.735

.743

.894

.924

.917

.910



.735

.744

.892

.953

.932

.918



λ

(continued)

.835

.954

ONPI CR

3.3 Empirical Study 103

.602

When I purchase from [firm’s] online store, I can return the 4.8/1.8 product to its offline store.

.663

KMO

.555

.695

.731

ItTC .804

α

OFPI

.821

CR

λ

.596

.819

.900

5.1/1.2 4.5/1.3

I like the store layout of [firm’s] offline store very much.

I think the prices at the [firm’s] offline store are always reasonable.a

[Firm’s] offline advertising is very informative.

.545

.489



.492



.657 .559

.612 .702

.799

.749

.520



.663

.680

5.6/1.1 5.4/1.2 5.4/1.2

[Firm’s] online store has a good variety of products.

I like the store layout of [firm’s] online store very much.

I think the prices at the [firm’s] online store are always reasonable.a





.554 .704

.680 .864

.624

.789

.753



.791

.658

Perceived quality of online offerings (e.g., Chowdhury, Reardon and Srivastava (1998); Yoo, Donthu and Lee (2000))

5.3/1.2 5.4/1.2

[Firm’s] offline store has a good variety of products.



.792

.652

.812



.791

.652

.596

.820

.898

λ

(continued)

.798

.756

.821

ONPI CR

3

Perceived quality of offline offerings (e.g., Chowdhury, Reardon and Srivastava (1998); Yoo, Donthu and Lee (2000))

.821

5.4/1.6

When I purchase from [firm’s] online store, I can pick up the product from its offline store.

.902

FL

5.1/1.8

MV/Std.

When I purchase from [firm’s] online store, I can inform myself about the availability of products in its offline store.

Perceived ON-OF services

Table 3.3 (continued)

104 Study 2: Effects of Perceived Offline-Online and Online-Offline …

5.4/1.3

MV/Std. .764

FL

KMO

3.5/1.9 3.3/1.9

I am very familiar with [firm’s] online store.

I have a lot of experience with [firm’s] online store.

4.8/1.3 4.9/1.4 5.3/1.2

The service/functions in offline and online store of [firm] are consistent.

The online store represents the offline store of [firm].

The offline and online store are typical for [firm].

.707

.897

.747

.784

.989

.955

.907

.813

.747

.653

.799

.687

.716

.953

.929

.894

.650

ItTC

.864

.965

α

OFPI

.867

.966

CR

λ

.712

.881

.758

.791

.987

.957

.908

.808 966

.867

ONPI CR

.712

.881

.758

.791

.986

.957

.910

.812

λ

Source: Own creation

Confirmatory model fits: Model 1: CFI .971, TLI .964, RMSEA .047, SRMR .043, χ2 (188) = 485.715. Model 2: CFI .969, TLI .962, RMSEA .049, SRMR .043, χ2 (188) = 519.490. Notes: OFPI = Offline purchase intention, ONPI = Online purchase intention, MV/Std. = Mean values and standard deviations; FL = Factor loadings (exploratory factor analysis); KMO = Kaiser-Meyer-Olkin criterion (≥ .5), ItTC = Item-to-total correlation (≥ .3), α = Cronbach’s alpha (≥ .7), CR = Composite reliability (≥ .6), λ = Standardized factor loadings (confirmatory factor analysis) (≥ .5). All items measured on 7-point Likert-type scales: 1 = strongly disagree, 7 = strongly agree. a Item deleted due to low factor loading.

4.8/1.2

The offline and online store of [firm] are similar.

Perceived channel congruence (Badrinarayanan et al. 2012)

3.6/1.9

I have often shopped at [firm’s] online store.

Consumers‘ online shopping experience (Diamantopoulos, Smith and Grime 2005)

[Firm’s] online advertising is very informative.

Table 3.3 (continued)

3.3 Empirical Study 105

106

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

Table 3.4 Discriminant validity Constructs

1

2

3

4

5

6

7

Model 1 1

Perceived OF-ON services

.630

2

Perceived ON-OF services

.222

.612

3

Perceived quality of offline offerings

.042

.020

.503

4

Perceived quality of online offerings

.065

.038

.487

.506

5

Offline purchase intention

.014

.017

.308

.160

.783

6

Consumers’ online shopping experience

.020

.030

.310

.222

.075

.905

7

Perceived channel congruence

.045

.065

.366

.442

.088

.124

.621

Model 2 1

Perceived OF-ON services

.630

2

Perceived ON-OF services

.223

.612

3

Perceived quality of offline offerings

.042

.018

.510

4

Perceived quality of online offerings

.065

.038

.489

.570

5

Online purchase intention

.031

.028

.110

.277

.873

6

Consumers’ online shopping experience

.020

.030

.109

.223

.404

.905

7

Perceived channel congruence

.045

.065

.368

.441

.133

.124

.621

Notes: AVE = Average variance extracted (≥ .5); values in italics represent squared correlations between constructs; values in bold represent the AVE of the construct. Correlations of the covariates (gender/age/general Internet expertise) to all other variables are below .135/-.089/.157. Source: Own creation

3.3 Empirical Study

3.3.3

107

Method

We tested for endogeneity with the IV method to control for omitted variables (Antonakis et al. 2014, see Appendix 3.2). Antecedents of perceived integration services were used as IVs. Offline store accessibility as an important antecedent of ON-OF services (e.g., Jin et al. 2018) was measured with one item (“I can easily reach the offline store of [firm],” Teller and Reutterer 2008). Perceived online service quality as an important antecedent of OF-ON services (e.g., a first step toward stimulating consumer interest in the channel, Shen et al. 2018) was measured with one item: “[firm] provides helpful service through its online store” (Montoya-Weiss, Voss and Grewal 2003). F-tests show that the IVs were strong predictors (Antonakis et al. 2014). In addition to the efficient models, consistent models with the IVs were calculated, while both did not significantly differ (the z-values were all < 1.96). The probability of endogeneity among the perceived integration services seems to be reduced. We additionally tested for measurement equivalence to ensure comparability across the objective congruence groups (we found full metric invariance for all constructs, offline model: χ2 (10 = 14.245, p > .05; online model: χ2 (10) = 16.981, p > .05, see Appendix 3.3). We used Mplus 8 and the LMS approach for the latent interaction effects, i.e., continuous latent moderators and conditional effects, to predict the moderated mediation (Cheung and Lau 2017). LMS provides unbiased estimators and standard errors that justify its complexity (Klein and Moosbrugger 2000). Two models are necessary: without latent interactions and with latent interactions (see Appendix 3.4, Cheung and Lau 2017). The conditional effects were probed by using the Johnson-Neyman floodlight test to identify the regions in the moderator-measures where the conditional effects are significant. This test is superior to other tests as the effects are interpreted by using all moderator values to limit the potentially arbitrary choice of values (Spiller et al. 2013).

3.3.4

Results

3.3.4.1 Results of the Path Coefficients The results show no direct effect of perceived OF-ON or ON-OF services on both purchase intentions in all models (see Table 3.5 and Table 3.6). Indirect-only mediation is established (Zhao et al. 2010). Offline purchase intention is significantly linked to OF-ON services via both the quality of offline and online offerings, but the difference is nonsignificant (model

.074 .118 .528 .375 −.076 −.001

→ ONO

→ OFO

→ ONO

→ OFPI

→ OFPI

→ OFPI

→ OFPI

OF-ON

ON-OF

ON-OF

OFO

ONO

OF-ON

ON-OF

.084

*

.135

−.123

−.129

→ ONO (H3a)

→ ONO (H3b)

→ OFPI (H5a)

→ OFPI (H5b)

× ONSE

× ONSE

→ OFPI

× CON

× CON

OF-ON

ON-OF

CON

OFO

ONO

.013

.006

−.065

.303

ns

**

***

ns

.106

.014

.434

→ ONO

ns

ONSE

−.004

−.006

→ OFO (H3b)

× ONSE

−.009

−.010

→ OFO (H3a)

× ONSE

.025

−.025

.214

.452

.108

.090

.173

ON-OF

***

ns

ns

**

***

ns

ns

**

OF-ON

.234

.017

−.058

.594

.773

.009

.015

.097

.288

.021

−.061

.313

.473

.021

.028

.192

.143

→ OFO

ns

ns

***

***

**

ns

***

**

β .119

.398

.707

.080

.080

.128

*

*

*

ns

ns

ns

**

***

ns

***

**

p

(continued)

.196

−.192

.018

.034

−.035

b

Model 1b

ONSE

.001

−.073

.708

.867

.053

.038

.095

.091

p

3

Interactions

.155 .186

→ OFO

OF-ON

Direct effects

b

β

Diff. test

Model 1a p

β b

Model 1

Table 3.5 Results of path coefficients: offline purchase intention

108 Study 2: Effects of Perceived Offline-Online and Online-Offline …

→ OFPI

→ OFPI

→ OFPI

→ ONO

→ OFO

→ ONO

OF-ON

ON-OF

ON-OF

.014

.013

.225 −.004

.085

.070

.146

.037

.033

.067

ns

ns

*

*

**

*

ns

**

*

t = 4.527**

t = .271 ns (H1b)

t = .519 ns (H1a)

−.003 −.009

−.010

.145

.0172

.1222

.006

.011

.058

.065

−.029

.054

.020

.140

.008

.017

.075

.085

ns

ns

*

ns

**

ns

ns

*

**

p

.077

.076

.117

.027

.048

.044

.073

.014

−.047

β

.200

.088

.135

.012

ns

ns

*

*

**

*

ns .0323

* .0563

*

p

.0513

.0843

−.004

b

Model 1b

Source: Own creation

Structural model fit Model 1: CFI .957, TLI .944, RMSEA .053, SRMR .049, χ2 (114) = 345.046. 1 Further differences tests: OF-ON vs. ON-OF → OFO → OFPI t = 3.612**, OF-ON vs. ON-OF → ONO → OFPI t = 2.524**. 2 Total effect in the mean value of moderator is shown (see Figure 3.3 for conditional results). 3 Indirect effect in the mean value of moderator is shown (see Figure 3.4 for conditional results). Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, ONSE = Consumers’ online shopping experience, CON = Perceived channel congruence, OFPI = Offline purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests. *** p < .001; ** p < .01; * p < .05; ns = not significant.

General Internet expertise

Age

Gender

.079

−.049

.110

→ OFPI

ON-OF

Covariates

.200

→ OFPI

.059

.051

.092

.108

b

β

Diff. test

Model 1a p

β b

Model 1

OF-ON

Total effects

→ OFPI

→ OFO

OF-ON

Indirect

effects1

Table 3.5 (continued)

3.3 Empirical Study 109

.194 .059 .107 .318 .495 −.021 .009

→ ONO

→ OFO

→ ONO

→ ONPI

→ ONPI

→ ONPI

→ ONPI

OF-ON

ON-OF

ON-OF

OFO

ONO

OF-ON

ON-OF

.081

−.005

−.121

→ ONO (H4a)

→ ONO (H4b)

→ ONPI (H6a)

→ ONPI (H6b)

× ONSE

× ONSE

→ ONPI

× CON

× CON

OF-ON

ON-OF

CON

OFO

ONO

.013

.006

−.064

.380

ns

**

***

−.051

.075

.524

→ ONO

ns

ns

ONSE

−.004

.000

→ OFO (H4b)

× ONSE

−.007

−.001

→ OFO (H4a)

× ONSE

.021

.015

.440

.064

.109

.068

.174

.142

ON-OF

***

ns

ns

***

***

ns

ns

**

*

OF-ON

.339

.036

−.079

1.236

.855

−.004

−.009

.098

.439

.035

−.066

.536

.395

−.009

−.018

.187

.146

→ OFO

ns

ns

***

***

*

ns

***

**

β .114

.134

.080

.055

.128

ns

ns

ns

ns

ns

*

ns

*

ns

***

**

p

(continued)

−.107

−.012

.119

.037

.025

1.015

b

Model 2b

ONSE

.010

−.025

1.169

.681

.047

.028

.097

.088

p

3

Interactions

.159

→ OFO

OF-ON

Direct effects

b

Model 2a Diff. test

β

p

β b

Model 2

Table 3.6 Results of path coefficients: online purchase intention

110 Study 2: Effects of Perceived Offline-Online and Online-Offline …

→ ONPI

→ ONPI

→ ONPI

→ ONO

→ OFO

→ ONO

OF-ON

ON-OF

ON-OF

.080

.087

.035 −.019

.011

.074

.174

.055

.019

.114

*

***

ns

*

**

*

ns

***

*

t = 4.233**

t = 2.300* (H2b)

t = 2.079* (H2a)

.034

−.143

−.050

−.015

.223

−.007

−.011

.153

.088

.038

−.016

−.168

−.0122

.1902

−.005

−.007

.121

.070

ns

***

ns

ns

**

ns

ns

**

*

p

.005

.061

.100

.056

.005

.089

.011

.084

−.168

β

.017

.089

.145

.089

*

***

ns

*

**

*

ns .0823

** .0073

ns

p

.1293

.0153

−.018

b

Model 2b

Source: Own creation

Structural model fit Model 2: CFI .960, TLI .947, RMSEA .053, SRMR .050, χ2 (114) = 345.222. 1 Further differences tests: OF-ON vs. ON-OF → OFO → ONPI t = 4.733**, OF-ON vs. ON-OF → ONO → ONPI t = 4.100**. 2 Total effect in the mean value of moderator is shown (see Figure 3.3 for conditional results). 3 Indirect effect in the mean value of moderator is shown (see Figure 3.4 for conditional results). Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, ONSE = Consumers’ online shopping experience, CON = Perceived channel congruence, ONPI = Online purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests. *** p < .001; ** p < .01; * p < .05; ns = not significant.

General Internet expertise

Age

Gender

.060

−.168

.093

→ ONPI

ON-OF

Covariates

.190

→ ONPI

.068

.024

.124

.065

b

Model 2a Diff. test

β

p

β b

Model 2

OF-ON

Total effects

→ ONPI

→ OFO

OF-ON

Indirect

effects1

Table 3.6 (continued)

3.3 Empirical Study 111

112

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

1: ß = .108, p < .05, ß = .092, p < .01, t = .519, p > .05). Offline purchase intention is differently linked to ON-OF services: nonsignificant through the quality of offline offerings (ß = .051, p > .05) and significant through the quality of online offerings (ß = .059, p < .05) but with a nonsignificant difference (t = .271, p > .05). We reject H1a-b but support H2a-b. Online purchase intention is linked to OFON services via the quality of offline and stronger via online offerings (model 2: ß = .065, p < .05, ß = 124, p < .001, significant difference, t = 2.079, p < .05). Online purchase intention is linked to ON-OF services via the quality of online but not offline offerings (significant difference, ß = .068, p < .05, ß = .024, p > .05, t = 2.300, p < .05). A reason for the rejection of H1a-b (OF-ON services’ equal effects via both offerings) is that the services put consumers at ease (Bhargave, Mantonakis and White 2016), i.e., both offerings are salient. The nonsignificant effect of ON-OF services is based on the lack of a direct link to quality of offline offerings, i.e., the stores’ salience. We consider stability tests and alternative models for a richer discussion (see Table 3.7 and Table 3.8). First, a bootstrap test confirms the missing direct effect of OF-ON and ON-OF services on purchase intention (5,000 samples, Model1OF-ON /ON-OF /Model2OF-ON /ON-OF : ß = -.076/-.001/-.021/.009, SE = .058/.058/ .052/.050, lower interval limit [LIL] = -.172/-.097/-.107/-.078, upper interval limit [UIL] = .019/.093/.065/.087). The mediation paths show the same results (Zhao, Lynch Jr. and Chen 2010). Second, ON-OF services (without OF-ON) show effects on online purchase intention via quality of online offerings only (ß = .116, p < .001) and support respective studies (e.g., Herhausen et al. 2015; see Table 3.9) but neglect the OF-ON effects. However, even models with only one of the perceived qualities (both OF-ON/ON-OF services) support the dominance of OF-ON services as their effects through the quality of online offerings lose significance (even for online purchase intention, ßOF-ON = .122, p < .001, ßON-OF = .058, p > .05, t = 5.657, p < .05; see Table 3.10 and Table 3.11). Thus, neglecting the mediation paths that make the perceived quality of offline/online offerings more salient disregards certain effects. Third, we applied alternative measures. A joint view of perceived integration services was measured as a second-order construct of both integration services (see Table 3.12). The results show (slightly) stronger paths via the quality of offline offerings to offline purchase intentions (ßOFO = .169, p < .001; ßONO = .157, p < .001, t = .424; p > .05) and online offerings to online purchase intentions (ßOFO = .092 p < .001; ßONO = .201; p < .001, t = 3.873; p < .01). These results are less insightful than our differentiated findings. Moreover, we measure the total purchase intention toward the firm: “if I find something that I like, it is

.044 .049 .058 .058

→ OFO

→ ONO

→ OFPI

→ OFPI

→ OFPI

→ OFPI

ON-OF

ON-OF

OFO

ONO

OF-ON

ON-OF

→ OFPI

→ OFPI

→ OFPI

→ ONO

→ OFO

→ ONO

OF-ON

ON-OF

ON-OF .073 .066

→ OFPI

→ OFPI

ON-OF

.031

.044

.036

OF-ON

Total effects

→ OFPI

→ OFO

OF-ON

.047

.059 .057

→ ONO

OF-ON

Indirect effects

.062 .060

→ OFO

OF-ON

Direct effects

S.E.

Model 1

.375

.528

.118

.074

.186

.155

.110

.200

.059

.051

.092

.108

−.001

−.076

β

.708

.867

.053

.038

.095

.091

.070

.146

.037

.033

.067

.079

.001

−.073

b

.009

.094

.014

−.018

.042

.035

−.097

−.172

.287

.450

.028

−.025

.092

.054

Lower

Table 3.7 Test for direct and indirect effect using bootstrapping: offline purchase intention

.225

.333

.117

.125

.157

.192

.093

.019

.450

.597

.215

.171

.287

.260

Upper

***

***

***

ns

***

***

ns

ns

***

***

***

ns

***

***

p

(continued)

t = 4.527**

t = .271 ns

t = .519 ns

Diff. test

3.3 Empirical Study 113

.039 .040

General Internet expertise

.014

.013

.225 −.004

.085

b

−.049

β

.020 .081

−.049

.141

Upper

−.110

.029

Lower

ns

ns

***

p

Diff. test

Source: Own creation

3

Structural model fit Model 1: CFI .957; TLI .944; RMSEA .053; SRMR .049; χ2 (114) = 345.046. Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, OFPI = Offline purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests; Lower: Lower limit of confidence interval; Upper: Upper limit of confidence interval; 95 % confidence interval limits are shown; bootstrap samples = 5,000. *** p < .001; ** p < .01; * p < .05; ns = not significant.

.034

Age

S.E.

Model 1

Gender

Covariates

Table 3.7 (continued)

114 Study 2: Effects of Perceived Offline-Online and Online-Offline …

.060 .058 .045 .041 .052 .050

→ OFO

→ ONO

→ ONPI

→ ONPI

→ ONPI

→ ONPI

ON-OF

ON-OF

OFO

ONO

OF-ON

ON-OF

→ ONPI

→ ONPI

→ ONPI

→ ONO

→ OFO

→ ONO

OF-ON

ON-OF

ON-OF .067 .057

→ ONPI

→ ONPI

ON-OF

.039

.026

.045

OF-ON

Total effects

→ ONPI

→ OFO

OF-ON

.031

.060

→ ONO

Indirect effects

.063

→ OFO

OF-ON

S.E.

OF-ON

Direct effects

Model 2

.495

.318

.107

.059

.194

.159

.093

.190

.068

.024

.124

.065

.009

−.021

β

.681

.047

.028

.097

.088

.074

.174

.055

.019

.114

.060

.010

−.025

1.169

b

.007

.093

.009

−.017

.059

.022

−.078

−.107

.423

.243

.016

−.042

.098

.056

Lower

Table 3.8 Test for direct and indirect effect using bootstrapping: online purchase intention

Upper

p

.193***

.310***

.138***

.069ns

.204***

.122***

.087ns

.065ns

.560***

.392***

.206***

.157 ns

.295***

.264***

(continued)

t=4.233**

t=2.300*

t=2.079*

Diff. test

3.3 Empirical Study 115

.039 .040

Age

General Internet expertise

.080

.080

.011 −.168

.011 −.168

Source: Own creation

χ2 (114)

.015

−.227

−.045 .067ns .142***

−.106*** 3

Structural model fit Model 2: CFI .960; TLI .947; RMSEA .053; SRMR .050; = 345.222. Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, OFPI = Offline purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests; Lower: Lower limit of confidence interval; Upper: Upper limit of confidence interval; 95 % confidence interval limits are shown; bootstrap samples = 5,000. *** p < .001; ** p < .01; * p < .05; ns = not significant.

.034

Gender

Covariates

Model 2

Table 3.8 (continued)

116 Study 2: Effects of Perceived Offline-Online and Online-Offline …

→ PI

ON-OF CI

.000 .014

−.005 .016

.228

.086

.060

.042

.700

.086

ns

ns

*

***

ns

***

***

.080

−.160

.008

.116

.045

.502

.192

.087

−.018

.025

.099

.047

1.186

.084

*

***

ns

***

ns

***

***

p

Diff. test

Source: Own creation

Structural model fits: Model 1: CFI .978; TLI .971; RMSEA .048; SRMR .046; χ2 (48) = 128.469. Model 2: CFI .982; TLI .976; RMSEA .046; SRMR .046; χ2 (48) = 121.004. Notes: ON-OF = Perceived online-to-offline services, ONO = Perceived quality of online offerings, PI = Purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests. *** p < .001; ** p < .01; * p < .05; ns = not significant.

General Internet expertise

Age

Gender

Covariates

.078

ON-OF CI

→ ONO

.373 .051

→ PI

→ PI

ONO

Indirect effect

.193

→ ONO

ON-OF

Direct effects

b

β

Diff. test

Model 2: Online purchase intention

p

β b

Model 1: Offline purchase intention

Table 3.9 Results of alternative models: ON-OF services only and online offerings only

3.3 Empirical Study 117

.552 −.014 .039

→ OFPI

→ OFPI

→ OFPI

OS

OF-ON

ON-OF

→ OFPI

→ OFPI

→ OS

→ OS

.023

.020

.256 −.005

.096

.035

−.052

.051

.086

.033

−.013

.891

.040

.096

ns

ns

**

ns

**

ns

ns

***

ns

**

t = 4.427**

.016

−.005

.086

.040

.083

.050

−.001

.375

.098

.203

.014

.000

.228

.031

.073

.042

−.001

.704

.043

.104

ns

ns

*

ns

***

ns

ns

***

ns

***

t = 3.701**

Diff. test

Source: Own creation

Structural model fits: Model 1: CFI .959; TLI .948; RMSEA .056; SRMR .050; χ2 (81) = 263.772. Model 2: CFI .966; TLI .957; RMSEA .052; SRMR .050; χ2 (81) = 240.943. Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OS = Perceived quality of offerings, OFPI = Offline purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests. *** p < .001; ** p < .01; * p < .05; ns = not significant.

General Internet expertise

Age

Gender

Covariates

ON-OF

OF-ON

.107

.076

→ OS

ON-OF

p

3

Indirect effects

.160

→ OS

OF-ON

Direct effects

b

β

Diff. test

Model 2: Online offerings

p

β b

Model 1: Offline offerings

Table 3.10 Results of alternative models: offline purchase intention with one offering only

118 Study 2: Effects of Perceived Offline-Online and Online-Offline …

.327 .069 .055

→ ONPI

→ ONPI

→ ONPI

OS

OF-ON

ON-OF

→ ONPI

→ ONPI

→ OS

→ OS

.107

−.021

−.188 .098

.155

.019

.047

.021

.067

.057

.082

.706

.027

.095

*

***

ns

ns

**

ns

ns

***

ns

**

t = 4.371**

.079

−.161

.008

.058

.122

.034

.024

.500

.097

.203

.086

−.018

.027

.050

.120

.035

.028

1.181

.042

.102

*

***

ns

ns

***

ns

ns

***

ns

***

p

t = 5.657**

Diff. test

Source: Own creation

Structural model fits: Model 1: CFI .957; TLI .946; RMSEA .058; SRMR .053; χ2 (81) = 280.605. Model 2: CFI .970; TLI .962; RMSEA .051; SRMR .051; χ2 (81) = 235.554. Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OS = Perceived quality of offerings, ONPI = Online purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests. *** p < .001; ** p < .01; * p < .05; ns = not significant.

General Internet expertise

Age

Gender

Covariates

ON-OF

OF-ON

.062

.057

→ OS

ON-OF

Indirect effects

.172

→ OS

OF-ON

Direct effects

b

β

Diff. test

Model 2: Online offerings

p

β b

Model 1: Offline offerings

Table 3.11 Results of alternative models: online purchase intention with one offering only

3.3 Empirical Study 119

120

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

(1) likely, (2) probable, (3) possible, that I will shop at the [firm]” (Schlosser, Barnett White and Lloyd 2006; see Table 3.13). This variable is linked to OF-ON services equally via the quality of offline and online offerings (ß = .096, p < .05, ß = .099, p < .01, t = .519, p > .05) and to ON-OF services via the quality of online offerings only (ß = .062, p < .05, ß = .044, p > .05, t = .519, p > .05). OFON services still dominate, which appear to reduce the idiosyncrasies of various purchase intentions.

3.3.4.2 Results of the Moderators Next, we present the results of the LMS approach by testing the conditional effects. Consumers’ online shopping experience moderates the total effects of OF-ON services on offline purchase intention (model 1a: ßOF-ONxONSE → ONO = −.129, p < .01). Figure 3.3a illustrates this negative interaction. With very high levels of experience, the lower confidence band cuts the x-axis, i.e., integration services are valuable again. H3a is supported, while H3b is not, as the nonsignificant interactions of ON-OF services via the quality of both offerings emerge (ßON-OFxONSE → OFO = -.009, p > .05; ßON-OFxONSE → ONO = .013, p > .05). Consumers’ online shopping experience moderates the total effects of OF-ON services on online purchase intention (model 2a: ßOF-ONxONSE → ONO = -.121, p < .01); Figure 3.3b illustrates this negative interaction. The effects of ON-OF services are again nonsignificant (ßON-OFxONSE → OFO = -.007, p > .05; ßON-OFxONSE → ONO = .013, p > .05). Therefore, H4a is supported, and H4b is not. A reason for the nonsignificant effect is that consumers’ experience itself becomes useful, while ON-OF services become less relevant to the evaluation. Perceived channel congruence moderates the indirect effects of perceived integration services. In the offline decisions for OF-ON services, reduced effects through the quality of offline offerings emerge (ßOFOxCON → PUI = −.123, p < .05). Figure 3.4a shows the interaction; Figure 3.4b shows the same marginally for ON-OF services. A reinforcing effect of both OF-ON and ON-OF services on offline purchase intention through the quality of online offerings occurs (ßONOxCON → PUI = .106, p < .05; see Figure 3.4c-d). H5a-b are supported. Regarding online decisions, H6a-b are not supported (model 2b). The interactions with the quality of online/offline offerings are nonsignificant (ßONOxCON→PUI = − .005, p > .05; ßOFOxCON→PUI = −.051, p > .05). Increasing levels of channel congruence do not affect the relative diagnosticity mechanism. In online decisions, the channels are evaluated in a more piecemeal way. The objective congruence measure and multi-group SEM provide further insights (see Appendix 3.5.1). In offline decisions, firms with a higher (vs. lower)

.524 .369 −.083

→ PI

→ PI

→ PI

OFO

ONO

JIS

JIS

.016

.014

.229 −.004

.086

.324

.156

.168

−.050

.326

.157

−.108

.697

.857

.224

.196

ns

ns

*

***

***

***

ns

***

***

***

***

t = .424 ns

.080

−.168

.011

.293

.201

.092

−.010

.487

.306

.321

.236

.088

−.019

.037

.355

.243

.112

−.015

1.152

.652

.211

.171

*

***

ns

***

***

**

ns

***

***

***

***

p

t = 3.873**

Diff. test

Source: Own creation

Structural model fits: Model 1: CFI .958; TLI .945; RMSEA .052; SRMR .050; χ2 (116) = 345.596. Model 2: CFI .960; TLI .949; RMSEA .052; SRMR .050; χ2 (116) = 345.458. Notes: JIS = Joint perceived integration services; OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, PI = Purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests. *** p < .001; ** p < .01; * p < .05; ns = not significant.

General Internet expertise

Age

Gender

Covariates

→ PI

→ ONO

JIS

Total effect

→ PI

→ PI

→ OFO

JIS

.169

.325

→ ONO

JIS

Indirect effects

.247

→ OFO

JIS

Direct effects

b

Model 2: Online purchase intention

Diff. test

β

p

β b

Model 1: Offline purchase intention

Table 3.12 Results of alternative models: joint perspective of integration services

3.3 Empirical Study 121

122

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

Table 3.13 Results of alternative models: total purchase intention Model 1 β

b

p

Diff. test

Direct effects OF-ON

→ OFO

.157

.092

**

OF-ON

→ ONO

.188

.096

***

ON-OF

→ OFO

.071

.037

ns

ON-OF

→ ONO

.117

.052

*

OFO

→ TPI

.473

.800

***

ONO

→ TPI

.407

.793

OF-ON

→ TPI

−.064

−.063

ns

ON-OF

→ TPI

−.025

−.022

ns

***

Indirect effects OF-ON

→ OFO

→ TPI

.096

.073

*

OF-ON

→ ONO

→ TPI

.099

.076

**

ON-OF

→ OFO

→ TPI

.044

.029

ns

ON-OF

→ ONO

→ TPI

.062

.041

*

t = .137 ns t = .795 ns

Total effects OF-ON

→ TPI

.196

.149

**

ON-OF

→ TPI

.106

.070

*

t = 3.721**

Covariates Gender Age General Internet expertise

.124

.339

**

−.069

−.006

ns

.025

.023

ns

Structural model fit Model 1: CFI .954; TLI .940; RMSEA .053; SRMR .049; χ2 (114) = 345.848. Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, ONSE = Consumers’ online shopping experience, CON = Perceived channel congruence, TPI = Total purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests. *** p < .001; ** p < .01; * p < .05; ns = not significant. Source: Own creation

3.4 Discussion and Conclusion

a) Perceived OF-ON services on offline purchase intention

123

b) Perceived OF-ON services on online purchase intention

Notes: Plots are based on unstandardized coefficients. The horizontal lines denote a total effect of zero. Dashed lines represent the confidence bands. For any values of the moderator for which the confidence bands do not contain zero, the conditional total effect is significantly different from zero. Dots represent the conditional total effects in the mean value of moderator (ßOF-ON→OFPI = .122, p < .01; ßOF-ON→ONPI = .190, p < .001; Spiller et al. 2013).

Figure 3.3 Plots of the conditional total effects: consumers’ online shopping experience. (Source: Own creation)

channel congruence achieve weaker (stronger) mediation paths of OF-ON services via the perceived quality of offline (vs. online) offerings (ßOFO-ONO = − .021, p < .05). No differences for ON-OF services (ßOFO-ONO = −.008, p > .05, H5a-b) and in online decisions occur (ßONO-OFO = .094, p < .05, ßONO-OFO = .017, p > .05, H6a-b). Our perceptual results are supported for ON-OF services in offline decisions. Here, congruent firms benefit by OF-ON services but still face a tradeoff between weaker and stronger quality effects.

3.4

Discussion and Conclusion

3.4.1

Overview

This study advances our cross-channel understanding of the mediation paths through which firms transform OF-ON and ON-OF services into purchase intentions. We focused on retailing, but studying the effects of major integration services for consumers is insightful for other service sectors comprising physical and online encounters (e.g., banking or healthcare, Banerjee 2014; Dahl et al. 2019). Moreover, the results help to determine whether and how important boundary conditions as diagnostic cues moderate the paths. We carefully provide major theoretical and managerial implications.

124

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

a) Perceived OF-ON services on offline purchase intention via per- b) Perceived ON-OF services on offline purchase intention via perceived quality of offline offerings ceived quality of offline offerings

c) Perceived OF-ON services on offline purchase intention via per- d) Perceived ON-OF services on offline purchase intention via perceived quality of online offerings ceived quality of online offerings

Notes: Plots are based on unstandardized coefficients. The horizontal lines denote an indirect effect of zero. Dashed lines represent the confidence bands. For any values of the moderator for which the confidence bands do not contain zero, the conditional indirect effect is significantly different from zero. Dots represent the conditional indirect effects in the mean value of moderator (ßOF-ON→OFO→OFPI = .084, p < .05; ßON-OF→OFO→OFPI = .056, p > .05; ßOF-ON→ONO→OFPI = .051, p < .05; ßON-OF→ONO→OFPI = .032, p < .05; Spiller et al. 2013).

Figure 3.4 Plots of the conditional indirect effects: perceived channel congruence. (Source: Own creation)

3.4.2

Theoretical Implications

Regarding our first research question, we show that firms indirectly participate in perceived integration services. Research has focused on the joint effects of OF-ON or ON-OF integration services but has not disentangled both (e.g., Herhausen et al. 2015; Hossain et al. 2020). In this respect, we apply an accessibility-diagnosticity theoretical rationale to the mediation paths across major sales channels. This theory provides a useful alternative to different theories used (e.g., technology adaption theory) and emphasizes which of the diagnostic cues is most relevant to omni-channel firms. Most importantly, our framework captures the perceived integration services simultaneously to reveal their mediation paths. We add to research by showing how the underlying mediation paths transform the services

3.4 Discussion and Conclusion

125

into offline and online channel purchase intentions. In general, we add to the joint perspectives by disentangling ramifications of integration services and to the OFON or ON-OF services’ perspective by considering the important cross-channel effects. In particular, we identify the most important mediation paths of the services in offline vs. online purchase decisions through perceived quality of channel offerings. Next, general and more particular implications of purchase decisions are provided. In general, all indirect and total effects of perceived OF-ON services are significant (not of ON-OF services). OF-ON services provide consumers knowledge about and ease of access to perceived quality of both offline and online channel offerings, while perceived ON-OF services fail to make both qualities more salient. Thus, it is wise to distinguish the integration services in omni-channel firms and to pay attention to perceived quality of channel offerings to prevent a disruption of integration service outcomes (Banerjee 2014). These distinct roles have been neglected in extent studies. One reason for the differences is a nonsignificant cross-channel direct effect (model 1: ßON-OF→OFO = .074, p > .05, model 2 ßON-OF→OFO = .059, p > .05; nonsignificant respective indirect effect). Adding OF-ON services to the studies on ON-OF services changes their results (underlined in alternative models, Akturk, Ketzenberg and Heim 2018). Surprisingly, perceived OF-ON services are a superior cue that increases both offline and online purchase intentions the most, with a stronger diagnosticity via perceived quality of offline (vs. online) offerings (model 1: ßOFO→OFPI = .528, p < .001, ßONO→OFPI = .375, p < .001, model 2: ßOFO→ONPI = .318, p < .001, ßONO→ONPI = .495, p < .001). For OF-ON (vs. ON-OF) services, the total effects are also stronger (model 1: ßOF-ON→OFPI = .200, p < .01, ßON-OF→OFPI = .110, p < .05, model 2: ßOF-ON→ONPI = .190, p < .01 ßON-OF→ONPI = .093, p < .05). The results may be affected by our choice of leading but former brick-and-mortar fashion firms. However, we believe that the results are noteworthy, as they enhance the studies that do not disentangle the integration services and their cross-channel effects. Theoretically, this study draws attention to the accessibility and diagnosticity of perceived integration services. Although the most important services for consumers are easily accessible to consumers, they are not equally relevant to behavioral intentions. The “why not-question”—in the case of the failure of ONOF services to make the quality of offline channel offerings more salient—can be explained by further theories. Fewer cognitive schemata and a poor retrieval between ON-OF services and offline offerings may exist. Even if consumer

126

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

often decide memory- rather than stimulus-based (Lynch, Marmorstein and Weigold 1988), motivational intensity might also affect the ON-OF services-quality of offline offerings-links. In particular, the results indicate differences in the most important mediation paths in offline and online decisions. We discuss selected implications in addition to the discussed above. • In offline purchase decisions, indirect effects of perceived OF-ON (vs. ON-OF) services are stronger via channel-specific quality of offline offerings (Campo and Breugelmans 2015; see further difference tests in the notes of Table 2.5, t = 3.612, p < .01). This is not surprising, but the cross-channel effect of OF-ON (vs. ON-OF) services via quality of online offerings is also stronger (t = 2.524, p < .05). Offline purchase intention is mostly affected by OF-ON services. For firms, these services are the strongest lever to attract consumers. Moreover, they are equally diagnostic via the quality of online and offline offerings (H1a). Firms offering OF-ON services provide consumers ease of access to the quality of offline and online offerings (even online; see the direct effects model 1: ßOF-ON→OFO = .155, p < .01, ßOF-ON→ONO = .186, p < .001). In contrast, the cross-channel effect of ON-OF services is nonsignificant. They are not diagnostic. Firms are not succeeding in steering consumers to a strategically important offline channel. • In online purchase decisions, the total effect of perceived OF-ON (vs. ONOF) services is stronger through the quality of online and offline offerings (see the difference tests in the notes of Table 2.6, t = 4.100, p < .01 and t = 4.733, p < .01). The respective direct links are also stronger. The services are not equally relatively diagnostic. Perceived OF-ON services put consumers at ease and leverage online purchase intentions (Bhargave, Mantonakis and White 2016). Moreover, adding perceived OF-ON services and quality of offline offerings to online-only studies changes their results (e.g., Herhausen et al. 2015). Cross-channel links are important: OF-ON services are relatively diagnostic via the quality of online offerings (H2a), and ON-OF services are relatively diagnostic via the channel-specific quality of online offerings for online purchase intentions only (H2b). In summary, we have studied the most important OF-ON and ON-OF services to consumers. However, they comprise different stages of the consumer journey: more informational and transactional OF-ON services vs. transactional and order fulfillment ON-OF services (see Table 3.2. Further services in the pre-purchase, purchase, and post-purchase stages exist (e.g., regarding payment/advertising,

3.4 Discussion and Conclusion

127

Pan, Wu and Olson 2017). We see advantages in future research. For example, after-sales services make additional information salient and may affect the mediation paths (Oh, Teo and Sambamurthy 2012). Regarding our second research question, consumers’ online shopping experience negatively moderates the total effects of perceived OF-ON (not ON-OF) services on offline and online purchase intention (H3a, H4a). It is a superior cue (Lee et al. 2019). This observation provides novel implications for integration services (e.g., adding to studies on OF-ON or ON-OF services only, Jin, Li and Cheng 2018). Within these interactions, only the cross-channel effects are moderated, but they reduce the total effects (e.g., the salience of online channel decreases, Shen et al. 2018). Therefore, for consumers with lower levels of online shopping experience, OF-ON services are relatively diagnostic for both purchase decisions. Increasing levels of experience diminishes the relative diagnosticity. The online shopping experience itself becomes relatively diagnostic for decision-making. However, the floodlight tests show that the conditional total effects are nonsignificant at the highest level of consumers’ online shopping experience (values of 6.0 for offline and 5.5 for online purchase intention, see Figure 3.3a-b; indirect effects mentioned are nonsignificant even at values of 5.0; see Appendix 3.5.2). Very high levels of online shopping experience do not decrease the accessibilitydiagnosticity mechanism in both online and offline decisions. Increasing levels of online shopping experience in the future will may make the services examined essential, not losing relevance (Bolton et al. 2018). Regarding our third research question, channel congruence as a diagnostic superior cue affects the relative diagnosticity mechanism. It moderates the mediation paths of perceived OF-ON and ON-OF integration services to offline (not online) purchase intention (H5a-b). In general, higher levels of channel congruence negatively affect the indirect effects via the channel-specific quality of offline offerings, while the online offerings become more salient particularly as follows. • For perceived OF-ON services, a reduced relative diagnosticity via the perceived quality of offline offerings and an increased relevance of the perceived quality of online offerings emerge. Consumers generalize information from one channel to another (Van Baal 2014). • In contrast, perceived ON-OF services affect offline purchase intention via the perceived quality of online offerings (marginal via offline offerings). Consumers are steered across channels. The channel-specific offline quality loses marginal relevance.

128

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

Perceived channel congruence is a diagnostic cue for offline decisions as congruent channels provide consumers with information and facilitate cognitive efforts (Bezes 2013). Additionally, the positive and negative moderation of the total effects of perceived OF-ON and ON-OF services are more insightful than the findings of only one mediator (see our literature review/alternative models). Congruence offers advantages depending on the decision (Van Baal 2014; Van Bruggen et al. 2010). Floodlight tests on the conditional indirect effects provide further insights for offline decisions. A marginal decreasing mediation path of ON-OF services via the perceived quality of offline offerings at low levels of channel congruence emerges (see Figure 3.4b). In contrast, the effects of both OFON and ON-OF services through the quality of online offerings increase above certain levels of congruence (over 1.5/2.0, respectively, see Figure 3.4c-d). Low levels of congruence do not increase the accessibility-diagnosticity mechanism in offline decisions. Our objective congruence measure supports the results of the perceived OFON services in offline decisions. Firms with higher (vs. competitors with lower) congruence can reinforce cross-channel effects due to the positive effects via perceived online and negative via established offline quality perceptions. Firms that synchronize channels face tradeoffs (e.g., also former online pure players like Apple, Lee et al. 2019). A careful management of perceived/objective channel congruence is necessary to benefit from the integration services in offline purchase decisions.

3.4.3

Managerial Implications

Practice have changed dramatically due to an increasing number of services to integrate channels (Hosseini et al. 2018). Managers of omni-channel firms may be aware of the various integration services and their importance and their knowledge about their effects from a consumers’ perspective may be confirmed. However, they should be interested in the mediation paths of their offered, distinct perceived integration services. This study identifies important perceived OF-ON services as major levers of offline and online purchase intentions (total toward the firm additionally). However, the mediation paths differ. • In offline decisions, perceived OF-ON services equally leverage purchase intentions through the knowledge of and ease of access to the quality of both offline and online offerings.

3.5 Limitations and Further Research

129

• In online decisions, OF-ON services more strongly leverage purchase intentions via the quality of online offerings; firms’ aim to steer consumers from offline to online channels is successful. Our results indicate a dominant role of perceived OF-ON services, which is interesting given the emergence of technologies to create seamless customer experiences across channels (Chen, Cheung and Tan 2018). Omni-channel firms in the fashion sector, in particular, should offer OF-ON services (e.g., employees with tablets that serve consumers’ needs in stores) to benefit from increasing offline and online purchases. Currently, omni-channel managers benefit by targeting consumers with a lower level of online shopping experience who value OF-ON services. However, in the future, an increasing consumers’ online shopping experience decreases these effects of the main lever. At the highest levels of this experience, particularly OFON services are seen as valuable again and affect consumers’ purchase intentions positively. The respective consumer segments need to be differentiated. Finally, managers can manage perceived channel congruence to benefit from integration services (a cost-intensive question, Van Baal 2014; Van Bruggen et al. 2010). However, only offline purchase intentions are contingent on channel congruence; online ones are not. Pushing consumer perceptions of high channel congruence supports offline purchases and in particular, the integration services’ goal to make the cross-channel more salient. The objective congruence results underline this issue for the main lever OF-ON services only.

3.5

Limitations and Further Research

This study has certain limitations that suggest future research directions. Although we paid special attention to sample selection, broadening the database would provide further insights into the role of integration services (e.g., consumer groups, or service industries with additional important OF-ON and ONOF services like travel, or even former pure online players). Furthermore, we used cross-sectional data, whereas future research could capture the changes in perceptions over time (e.g., by using cross-lagged panel models, Wu et al. 2018). Regarding the measures, options to study further integration services in the consumer journey were mentioned. We measured the services according to the literature and theory. However, experiments will allow manipulations of the accessibility of integration services (Bhargave, Mantonakis and White 2016), and measuring diagnosticity directly may provide further insights into the boundary

130

3

Study 2: Effects of Perceived Offline-Online and Online-Offline …

conditions of the mediation paths (Qiu et al. 2012). We used an intentional purchase measure (like most studies), while the findings need to be complemented with actual purchase data. Finally, due to convergence problems, the moderators cannot be tested simultaneously (Cheung and Lau 2017), which will allow the identification of the most important levers for the effects. Adapting our framework could provide further insights into the effects of integration services. Instead of mediation paths/cross-channel effects, scholars may study the role of both services for channel quality, usefulness or value. More complex, scholars may study our framework in a reciprocal design as these relationships among perceived channel quality, value or image are obvious but seldom tested in omni-channel studies (e.g., Swoboda, Weindel and Schramm-Klein 2016). Studying drivers of integration services is also advantageous (Banerjee 2014). We use cognitively rationales as most decisions are memory- rather than stimulus-based (Lynch, Marmorstein and Weigold 1988). Alternative theories are interesting. Although we focus on the major sales channels, further directions for integration exist (e.g., online-to-online services of platforms/social media, Ibrahim and Wang 2019). Incorporating them will account for further touchpoints in customer journeys.

4

Study 3: Importance of Marketing Instruments for Repurchase Intentions in Omni-Channel Retailing

4.1

Introduction

Omni-channel retailers offering a seamless experience across channels increasingly compete for consumers online, e.g., in the fashion sector, where ten of the top 20 firms worldwide are online players (EcommerceDB 2020). They have to orchestrate online- and omni-channel-specific instruments to benefit from crosschannel effects (e.g., aesthetic appeal or online-offline integration, Toufaily and Pons 2017; Yang et al. 2020). However, effective orchestration depends on the relative importance of instruments and on mechanisms that affect consumer behavior (Bleier, Harmeling and Palmatier 2019). Online trust is a known mechanism, engaged by marketing instruments and important for behavior (i.e., websites’ delivery of confidence or reliability, Kim and Peterson 2017; Ye et al. 2020). In retailing, brands are also important to consumers (e.g., Das 2016; Swoboda, Berg and Schramm-Klein 2013), and omni-channel fashion firms such as Zara and H&M see consumer-based retail brand equity as a core competence (Gielens and Steenkamp 2019; Loupiac and Goudey 2019). We therefore examine the effects of online- and omni-channel-specific instruments on consumer behavior mediated by online trust and brand equity (i.e., associations of a retailer’s website as a strong, attractive, unique, and favorable brand, Keller 2010). We moreover consider interdependencies of online trust and brand equity because they are assumed (e.g., Hollebeek and Macky 2019) and because a strong brand affects trust, while trust affects brand associations (e.g., Khan et al. 2019). Scholars have acknowledged the relative importance of various online-specific instruments but have mostly studied one or two (see Figure 4.1). For example, Awad and Ragowsky (2008) examined the effects of ease of use and usefulness

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 A. Winters, Omni-Channel Retailing, Handel und Internationales Marketing Retailing and International Marketing, https://doi.org/10.1007/978-3-658-34707-9_4

131

132

4

Study 3: Importance of Marketing Instruments for Repurchase …

on online purchase intentions through online trust, Khan et al. (2019) examined the effects of website quality on behavioral intentions mediated by brand commitment, and Al-Qeisi et al. (2014) by expectancy. These studies did not cover an effective orchestration of instruments. Fewer studies analyzed more instruments, such as the effects of aesthetic appeal, ease of use, security/privacy, information or content (e.g., mediated by perceived value or enjoyment, Bressolles, Durrieu and Deans 2015; Floh and Madlberger 2013). Online trust is most often the mediator (e.g., Chen and Dibb 2010; Toufaily and Pons 2017). These studies predict the relative importance of instruments but not of online- and omni-channel-specific ones. Moreover, few studies addressed brands. Al-Hawari (2011) studied brand awareness and image as mediators of instrumental. effects on loyalty, Bleier, Harmeling and Palmatier (2019) studied brand trustworthiness’s importance for the online instruments-purchase intention link, and scholars indicated links of brand associations to trust and vice versa (Benedicktus et al. 2010; Fazal-e-Hasan et al. 2018; Khan et al. 2019). However, brand equity and its reciprocity to trust have not yet been examined (i.e., interdependencies of both in a loop, Nagase and Kano 2017). Finally, online and offline outcomes, i.e., cross-channel effects, are seldom considered even though experiences from website interactions transfer cognitively to additional channels (e.g., Grewal and Roggeveen 2020).

Marketing instruments Trust Up to two Awad and Ragowsky (2008); Badrinarayanan et al. (2012); Badrinarayanan, Becerra and Madhavaram (2014); Bansal, Zahedi and Gefen (2016); Bashir et al. (2018); Benedicktus et al. (2010); Bock et al. (2012); Casado-Aranda, Dimoka and Sánchez-Fernández (2019); Fazal-e-Hasan et al. (2018); Javed and Wu (2020); Rahimnia and Hassanzadeh (2013); Toufaily, Souiden and Ladhari (2013); Ye et al. (2020) etc. More than Chen and Dibb (2010); Loureiro, Cavallero two and Miranda (2018); Rose et al. (2012); Shin et al. (2013); Toufaily and Pons (2017); Zhou, Lu and Wang (2009)

Mediators Brand Others Khan et al. (2019); Baek et al. Al-Qeisi et al. (2014); Bertrandie and Zielke (2020); Lin and Lee (2012); Wu (2017); Cao et al. (2018); Emrich, Paul and et al. (2013) Rudolph (2015); Fang et al. (2016); Herhausen et al. (2015); Herhausen et al. (2020); Landers et al. (2015); Lee et al. (2019); Mallapragada, Chandukala and Liu (2016); Murfield et al. (2017); Sullivan and Kim (2018); Wen and Lurie (2019); Yang et al. (2020) etc.

Al-Hawari (2011); Bleier, Harmeling and Palmatier (2019)

Bressolles, Durrieu and Deans (2015); Demangeot and Broderick (2016); Dickinger and Stangl (2013); Floh and Madlberger (2013); Hsieh et al. (2014); Kim et al. (2012); Liu, Li and Hu (2013); McDowell, Wilson and Kile Jr. (2016)

This study Note: Italic = several omni-channel outcomes, i.e. offline and online. Bold = studies on trust and brand.

Figure 4.1 Literature review on marketing instruments. (Source: Own creation)

4.1 Introduction

133

In summary, research has not sufficiently considered the relative effects of online- and omni-channel-specific instruments on consumer behavior. Scholars have called for such studies to clarify indirect and cross-channel effects (e.g., Bleier, Harmeling and Palmatier 2019; Blut, Teller and Floh 2018). Moreover, research has addressed online trust (even calls to verify its casual effects over time, Kim and Peterson 2017; Ye et al. 2020) but not brand equity as an important issue in competition with purely online players (e.g., Gielens and Steenkamp 2019; Loupiac and Goudey 2019). Brand equity may more strongly reciprocally affect behavior as consumers gain confidence in online shopping. Respective insights are beneficial for managers orchestrating online- and omni-channel-specific instruments. We aim to address these research gaps and to analyze how important instruments affect online and offline repurchase intentions, as mediated by online trust and brand equity. We study repurchase intentions, i.e., the likelihood that consumers continue to shop on a retailer’s online and offline channels (Hult et al. 2019). Thus, we show how consumers’ experience with a retailer’s website affects repurchases (Sullivan and Kim 2018). We further ask whether reciprocal relationships between online trust and brand equity exist and how they affect repurchase intentions. In doing so, we offer two important contributions to theory and practice. First, analyzing the relative effects of online- vs. omni-channel-specific instruments provides novel insights into the application of these instruments. Scholars studied many instruments but used formative measures without relative evidence (Dickinger and Stangl 2013; Shin et al. 2013) or addressed online instruments only (Toufaily and Pons 2017; Loureiro, Cavallero and Miranda 2018). We also show their relative cross-channel effects (Blut, Teller and Floh 2018; Hult et al. 2019). Our focus on online brand equity extends research, e.g., those focusing on trust (e.g., Bashir et al. 2018; Chen and Dibb 2010), and clarifies how online trust and brand equity contribute to repurchasing. Whereas studies often refer to technology acceptance or reasoned action/planned behavior, we contribute to the application of schema theory (Fiske and Taylor 1991). Online trust and brand nodes represent consumers’ schemata of a website stored in memory (Rajavi, Kushwaha and Steenkamp 2019). Experienced consumers activate nodes and associate instruments for evaluations and repurchases across channels; while some instruments are linked to trust and brands, others are not (Grewal et al. 2009). Assessing such links is the basis for the effective omni-channel management of marketing instruments (Hult et al. 2019). In line with theory, we apply sequential mediation modeling over time in study 1, i.e., we measure dependent, independent and mediating variables at different time points. In doing so, we follow calls in the

134

4

Study 3: Importance of Marketing Instruments for Repurchase …

literature to avoid the limitations of cross-sectional studies that may bias the estimation of cognitive mediation parameters, for example (Kim and Peterson 2017; Mitchell and Maxwell 2013). Second, we contribute to research by conceptualizing the reciprocal effects of online trust and brand equity on online and offline repurchase intentions (retailers in general in alternative models). Studies suggest that trust is a result of brand associations or that brand associations are affected by trust (Fazal-e-Hasan et al. 2018; Khan et al. 2019). We show that spreading activation of one node to another one over time emerges (e.g., Inman et al. 2004), i.e., from equity to trust in consumers’ minds and vice versa (Hollebeek and Macky 2019; Rajavi, Kushwaha and Steenkamp 2019). Online trust is vital in e-commerce research (e.g., Kim and Peterson 2017); examining reciprocity in the cross-lagged panel models will show whether online trust or brand equity has a stronger effect and respectively guide online- and omni-channel-specific instruments stronger to repurchase intentions. For managers, major reciprocal mechanisms of consumers’ repurchasing will become evident.

4.2

Conceptual Framework and Hypothesis Development

4.2.1

Overview

To address our research aims, we conceptualized six online- and omni-channelspecific instruments—aesthetic appeal, ease of use, security/privacy, customer service, online-offline integration, and channel consistency—as determinants of online and offline repurchase intentions (mediated by online trust and brand equity, see Figure 4.2). Their selection is given next. To explain the conceptualized relationships, we build on theoretical rationales and additional empirical results. A systematic process based on our literature review guides the selection of online instruments.1 We chose the most frequently studied instruments: aesthetic appeal (i.e., visual appearance/attractiveness of a website, Cai and Xu 2011), ease of use (i.e., website’s navigation, search and functional ability, Rose et al. 2012) and security/privacy (i.e., protection of financial and personal data, Toufaily, Souiden and Ladhari 2013). Customer service (i.e., website’s general reliable 1

Following the procedure of Gaur and Kumar (2018), we identified the online-specific marketing instruments analyzed in the studies in our literature review, coded and categorized them into subcategories based on their underlying similarities, e.g., “website design” or “aesthetic appearance” grouped under “aesthetic appeal”, and chose the most frequently analyzed instruments.

4.2 Conceptual Framework and Hypothesis Development

135

support, Cao, Ajjan and Hong 2018) was chosen because of its importance in recent research (e.g., Herhausen et al. 2020). A few studies in our review used omni-channel-specific instruments as an independent variable, e.g., online-offline information integration (Yang et al. 2020), click-and-collect (Murfield et al. 2017), integrated interactions (Lee et al. 2019), channel congruence (Landers et al. 2015), or assortment integration (Bertrandie and Zielke 2017). We chose two instruments that characterize omni-channel firms and consumer journeys (Acquila-Natale and Chaparro-Peláez 2020). Online-offline integration was chosen (i.e., support in online venues to interact with offline ones, Herhausen et al. 2015; e.g., searching for available products offline, click-and-collect or returning online purchased items offline). This technology-based instrument provides consumers with a seamless experience across channels and is a gradual decision (no longer a strategic one to be integrated or not, Lee et al. 2019). Moreover, channel consistency was chosen (i.e., harmonization of online and offline channels, Van Baal 2014; e.g., aligned/integrated channels, product descriptions, assortments or prices) as a major marketing instrument for omni-channel retailers (e.g., Bertrandie and Zielke 2017).

Aesthetic appeal Ease of use

Online trust

Online repurchase intention

Online retail brand equity

Offline repurchase intention

Security/privacy Customer service Online-offline integration Channel consistency Note:

Study 1,

Study 2

Covariates: gender, age, internet expertise, perceived assortment variety, perceived price fairness

Figure 4.2 Conceptual framework. (Source: Own creation)

4.2.2

Theory

Scholars often conceptualize marketing instruments as stimuli or signals sent out by firms and perceived by consumers as information cues to form attitudes or

136

4

Study 3: Importance of Marketing Instruments for Repurchase …

cognitions and to facilitate decisions (Mallapragada, Chandukala and Liu 2016; Rose et al. 2012). The extent to which consumers access and weight such salient and relevant signals affects their degree of behavioral relevance (Badrinarayanan et al. 2012; Connelly et al. 2011). Respectively, online- and omni-channelspecific instruments are not treated equally in decisions. However, consumers in repurchase decisions refer to learned and stored information in memory (Grewal and Roggeveen 2020). We therefore refer to schema theory, as it provides organizing mechanisms for experienced individuals’ knowledge about situations and objects, typically in networks of dependent nodes, concepts or associations and links between them (e.g., Fiske and Taylor 1991, p. 98; Krishnan 1996). Scholars conceptualize brand equity and trust as nodes in consumers’ memory (e.g., Keller 2010; Bock et al. 2012). They are linked to perceived marketing instruments and activated by external stimuli or information retrieved from memory (e.g., Mavlanova et al. 2016; Puligadda et al. 2012). The number and strength of respective links can be explained by the degree of activation (e.g., Anderson 1983; Keller 1993). The behavioral relevance of these processes in repurchase situations arises as nodes, and activated information affects consumers’ evaluations and decisions (often heuristically, Halkias 2015). In our context, perceived online- and omni-channel-specific instruments more or less intensely activate online trust and brand equity nodes and then affect online and offline repurchase intentions. Our rationale is also based on the premise that online trust and brand equity are interconnected (Inman, Shankar and Ferraro 2004; Rajavi, Kushwaha and Steenkamp 2019). Both are behaviorally relevant, cognitively available associations with a retailer’s website. To explain reciprocity and strength of the links between both, one can refer to the concept of spreading activation between nodes (Keller 1993; Teichert and Schöntag 2010). For example, trust toward an online store, activated by perceptions of a website’s security/privacy, becomes a source of spreading activation for brand equity. Conversely, brand equity may act as a primer for trust, i.e., causing reciprocity. The strength of spreading activation in a decision depends, for example, on the sum of weights of the common links between nodes relative to the weights of all associations linked to each node (e.g., Balachander and Ghose 2003; Lei et al. 2008). However, a general behavioral relevance emerges as experienced consumers retrieve information about online trust and brand equity in cross-channel repurchasing. Thus, two mechanisms theoretically substantiate the effects in our framework: perceived stimuli and activation of information in memory. How strong each instrument affect repurchasing and how online trust-online brand equityreciprocity do it are the subjects of the next analysis steps.

4.2 Conceptual Framework and Hypothesis Development

4.2.3

137

Hypotheses on the Effects of Marketing Instruments

We conceptualized indirect effects of online- and omni-channel-specific instruments, i.e., we test but do not hypothesize direct effects. As we study omnichannel retailing and online trust and brand equity as cognitively available, behaviorally relevant nodes, their importance for online and offline repurchases seems obvious. We therefore discuss the differences in cross-channel effects but do not hypothesize them. Next, hypotheses for each instrument are derived based on theoretical rationales in extant research and additionally on respective empirical findings. Aesthetic appeal is theoretically often conceptualized as a website’s important stimuli or signals affecting consumers affectively or cognitively (e.g., Floh and Madlberger 2013; Wu et al. 2013). It provides guidance in processing visual information (e.g., creates structure or unity), reduces cognitive effort and is behaviorally important due to consumers’ respective experience and knowledge (Bleier, Harmeling and Palmatier 2019; Wen and Lurie 2019). Such knowledge is linked to further associations in memory, e.g., online trust and brand equity, and retrieved in repurchase decisions (Loupiac and Goudey 2019; Rahimnia and Hassanzadeh 2013). We argue that in omni-channel retailing repurchasing consumers refer to such cognitions, which, in turn, affect their online and offline repurchase intentions (e.g., Grewal and Roggeveen 2020; Herhausen et al. 2015). Empirical studies emphasized the importance of aesthetic appeal. They mostly found indirect effects on online behavior through online trust, attitudes or value (Bressolles, Durrieu and Deans 2015; Rahimnia and Hassanzadeh 2013; Wu et al. 2013); few results were insignificant (Toufaily and Pons 2017). Aesthetic appeal’s effect on offline behavior was supposed (Loupiac and Goudey 2019). We propose the following: H1:

Aesthetic appeal has a positive indirect effect on repurchase intentions (online and offline) through (a) online trust and (b) online brand equity.

Scholars conceptualize the ease of use of websites as a stimulus or cue that simplifies consumers’ cognitive processes in two ways (e.g., Hsieh et al. 2014; Liu, Li and Hu 2013): to find information or enact a transaction and as flow experiences to control interactions (Bressolles, Durrieu and Deans 2015; Landers et al. 2015). The website’s usability is therefore associated with trustworthiness and positive experiences (Demangeot and Broderick 2016; Landers et al. 2015). This link causes respective behavioral outcomes (Badrinarayanan et al. 2012). We argue that ease of use affects online and offline repurchase intentions through associations,

138

4

Study 3: Importance of Marketing Instruments for Repurchase …

online trust and brand equity. Consumers’ experiences and associations with a channel are used to evaluate further channels in omni-channel retailing and thus affect online and offline repurchase intentions (Grewal and Roggeveen 2020). Empirically, websites’ ease of use has been linked to online (not offline) behavior. Only indirect effects were shown, mediated by online trust, enjoyment or attitudes (e.g., Bressolles, Durrieu and Deans 2015; Chen and Dibb 2010; Landers et al. 2015). Seldom have insignificant results through online trust or brand image emerged (e.g., Al-Hawari 2011; Toufaily and Pons 2017). Following our theoretical rationales, we carefully hypothesize the following: H2:

Ease of use has a positive indirect effect on repurchase intentions (online and offline) through (a) online trust and (b) online brand equity.

Security/privacy is an important signal to provide consumers’ confidence in online shopping and to facilitate their judgments on personal information safety and protection (e.g., Casado-Aranda, Dimoka and Sánchez-Fernández 2019; Kim et al. 2012). Such judgments affect consumers’ long-term knowledge, trust and behavior (Bashir et al. 2018; Toufaily, Souiden and Ladhari 2013). We argue that these mechanisms activate online brand equity as well. Repurchasing consumers retrieve this knowledge about an omni-channel retailer and draw inferences from online to offline repurchasing (Loupiac and Goudey 2019). Almost all studies showed positive indirect effects of security/privacy on online (not offline) behavior. Provided mediators include online trust and brand image, which supports our rationale (e.g., Bressolles, Durrieu and Deans 2015; Bashir et al. 2018; Sullivan and Kim 2018). Following our theoretical rationale, we carefully hypothesize the following: H3:

Security/privacy has a positive indirect effect on repurchase intentions (online and offline) through (a) online trust and (b) online brand equity.

Scholars have recently discussed customer service, especially electronic service quality, as an important instrument (Hogreve et al. 2019; Herhausen et al. 2020). Service perceptions typically facilitate consumers’ problem solving or thinking (e.g., Cao, Ajjan and Hong 2018; Fang et al. 2016). Reliable services prevent consumers’ cognitive constraints or enhance their problem-solving ability and thus increase trust and the retailer’s brand equity (Al-Hawari 2011; Toufaily and Pons 2017; White et al. 2013). Repurchasing consumers refer to such associations. As associations formed by knowledge on one channel transfer to other channels in omni-channel retailing (e.g., Loupiac and Goudey 2019), online and offline repurchase intentions will be affected by these associations.

4.2 Conceptual Framework and Hypothesis Development

139

Empirical studies have mostly found indirect effects of customer service on online behavior through brand image, online trust, or values (e.g., Toufaily and Pons 2017; Fang et al. 2016); they have also found direct effects on offline behavior (Hogreve, Bilstein and Hoerner 2019; Herhausen et al. 2020). We hypothesize the following: H4:

Customer service has a positive indirect effect on repurchase intentions (online and offline) through (a) online trust and (b) online brand equity.

Online-offline integration and channel integration in general are theorized as salient signals or information cues facilitating cognitive processing (e.g., crosschannel movements, flows or associations, Lee et al. 2019; Yang et al. 2020). Channel integration increases omni-channel retailers’ reliability and trustworthiness or brand associations (e.g., Schramm-Klein et al. 2011; Zhang et al. 2018). We similarly argue that for repurchasing consumers, online-offline integration activates associations for both online trust and brand equity. Moreover, integration induces the whole omni-channel system and therefore affects cross-channel repurchase intentions (Herhausen et al. 2015). Empirically, integration has been studied mostly as a single independent variable, indirectly affecting online and offline behavior and variously mediated, e.g., by online trust or brand associations (e.g., Javed and Wu 2020; Schramm-Klein et al. 2011; Zhang et al. 2018). We propose the following: H5:

Online-offline integration has a positive indirect effect on repurchase intentions (online and offline) through (a) online trust and (b) online brand equity.

Perceived channel consistency and respective effects have mostly been theorized cognitively, supporting our rationale. For example, it activates experience schemata and categories or helps to eliminate friction points in consumers’ minds (e.g., improving cognitive processing, Emrich, Paul and Rudolph 2015; Lee et al. 2019). Badrinarayanan et al. (2012) provided a schematic theoretical rationale for congruence effects on trust and purchase intentions. We argue similarly for trust and brand equity, following the rationale that retrieved consistency, i.e., consumers’ mental pattern and structure, affects online and offline repurchase intentions (Bertrandie and Zielke 2017; Van Baal 2014). Empirical studies have seldom analyzed the relative effects of channel consistency on additional instruments but have showed its positive links to online

140

4

Study 3: Importance of Marketing Instruments for Repurchase …

purchase intentions mediated by online trust (e.g., Badrinarayanan, Becerra and Madhavaram 2014). However, it may also directly affect consumer behavior (e.g., retail patronage, Emrich and Verhoef 2015, or loyalty intentions, Van Baal 2014). We hypothesize the following: H6:

4.2.4

Channel consistency has a positive indirect effect on repurchase intentions (online and offline) through (a) online trust and (b) online brand equity.

Hypotheses on Reciprocal Effects

Schema theory substantiates reciprocity between online trust and brand equity. Both conceptually represent interconnected nodes in the memory of experienced consumers and are activated in repurchase situations (e.g., Rajavi, Kushwaha and Steenkamp 2019). Through these links, activation can spread from online trust to brand equity and vice versa (Keller 1993; Inman, Shankar and Ferraro 2004). Studies have shown that the spread of activation tends to occur automatically, with short decision processes—perhaps particularly online—and merge between related nodes. Trust and brands are conceptually closely related (e.g., Chaudhuri and Holbrook 2001), and reciprocity is likely to occur in both omni-channel online and offline repurchase situations (e.g., Hollebeek and Macky 2019). Studies in our literature review have not conceptualized reciprocity empirically or refer to related cognitive processes. However, they have indicated possible reciprocal links by conceptualizing the effects of online brand associations on trust (Fazal-e-Hasan et al. 2018; Khan et al. 2019) or those of online trust on brand attitudes (Frasquet et al. 2017; Vernuccio et al. 2012). We propose the following: H7:

Online trust and online brand equity have a positive reciprocal relationship.

We also refer to schema theory to substantiate the relative reciprocal total (the sum of direct and indirect) effects of online brand equity and online trust on crosschannel repurchase intentions. Their direct role depends on their access, relevance and retrieval for repurchasing in online and (crosswise) offline channels. Spreading activation in a decision situation depends on the weights of the links between nodes (Inman, Shankar and Ferraro 2004) and further links each node have. For example, if online trust has few links to other nodes in consumers’ omni-channel networks while online brand equity has many strong links, more activation will occur for brand equity than vice versa. We consider both nodes to be important

4.3 Empirical Studies

141

for online and offline repurchases and cannot reliably hypothesize their weights. However, we theoretically assume that consumers activate both nodes and link them for evaluation (e.g., Badrinarayanan et al. 2012; Grewal, Levy and Kumar 2009). The reciprocal effects of online trust and brand equity have not yet been studied. Empirically, direct effects of online trust (Das 2016; Loureiro, Cavallero and Miranda 2018; Sullivan and Kim 2018 and Darke et al. 2016; Xiao et al. 2019) and online brand associations (Bleier, Harmeling and Palmatier 2019; Khan et al. 2019 and Baek et al. 2020; Toufaily, Souiden and Ladhari 2013) on online and offline behavior have been shown. We hypothesize the following: H8:

(a) Online trust and (b) online brand equity have a positive total effect on repurchase intentions (online and offline).

4.3

Empirical Studies

4.3.1

Overview

Two empirical studies were conducted. Study 1 analyzed the effects of marketing instruments on online and offline repurchase intentions in a sequential mediation design. Study 2 analyzed the effects of reciprocity between online trust and online brand equity in a cross-lagged panel design.

4.3.2

Study 1: Sequential Mediation Design

4.3.2.1 Sample The German fashion sector was chosen for both studies for several reasons. Fashion is the third largest retail sector and has the highest share of online sales in Germany (24.9% according to the national retail association, compared to 24.3% for the larger electronics sector and 8.4% for the largest food sector, HDE/IFH 2020). This sector offers a good selection of several well-known omni-channel retailers, which helps to avoid single firm-specific results (Landers et al. 2015). The fashion sector is not very concentrated, with over 40 firms accounting for twothirds of total sales (while one electronic firm dominates with 40% of total sales), and brands are important for consumers, who have considerable online experience.

142

4

Study 3: Importance of Marketing Instruments for Repurchase …

The largest retailers offer various touchpoints (Acquila-Natale and ChaparroPeláez 2020), while offline and online channels dominate the interpurchase period of under 40 days (e.g., Hult et al. 2019; HDE/IFH 2020). Pretests were conducted to select retailers from which consumers purchase often online and offline. First, a list of the twelve top-selling fashion retailers in Germany was compiled. Based on awareness data from a first pretest (N = 130, quota sample, face-to-face interviews), the eight best-known and most frequently used retailers were chosen. We selected six top-selling omni-channel fashion firms in 2018 by eliminating those offering broader assortments, i.e., not only fashion articles. In a second pretest, we then asked a convenience sample of 50 consumers regarding their online and offline purchase experiences to ensure that most consumers regularly shop at one of these firms. The four chosen firms are omni-channel retailers (according to Oh, Teo and Sambamurthy 2012). In the survey, respondents were asked to evaluate known physical stores in typical midsized cities and websites used through an app on a tablet or laptop. We aimed to develop a quota sample for 500 consumers appropriate for crosslagged modeling (i.e., national distribution according to age and gender). A total of 700 respondents who regularly made offline and online fashion purchases were chosen from a consumer panel. We contacted them in a screening phase in two waves following the quotas until 500 agreed to participate. The same respondents were surveyed over a ten-month period in 2018–2019 and in three waves conducted 4–5 months apart. Eighteen trained interviewers conducted face-to-face in-home interviews using standardized questionnaires because of their higher data quality and lower nonresponse bias than web surveys (Bennink, Moors and Gelissen 2013). Low-prize lottery vouchers were used as incentives for participants in all waves (e.g., Beal 2015). In the screening phase, we also asked about unaided and aided awareness of fashion retailers and about whether the consumers purchased offline and online in the last year (Kuehnl et al. 2019). The first or second retailer that fit our pretests was randomly chosen for each respondent to be evaluated in all waves. 493 respondents met the requirements and were again asked in the second and third waves of the survey whether they had made additional purchases online and offline at the retailer. Eighty-seven consumers who made no purchases or who purchased on one channel only were eliminated; moreover, eight incomplete cases emerged. 398 respondents were included in the studies. Mahalanobis distance identified 21 outliers. 377 observations per wave remained. Compared to the planned quotas, the 29–49 (15–29) age group was overrepresented (underrepresented, see Table 4.1). Tests revealed deviation from multivariate normality. A mean-adjusted maximum likelihood estimator (MLM), which provides robust chi-square tests (Maydeu-Olivares 2017), was chosen to test the hypotheses.

4.3 Empirical Studies

143

Table 4.1 Sample Realized quota sample (in %)

Planned quota sample (in %)

Male

Male

Female

Total

Female

Total

Fashion sector (N = 377) Age 15–29

10.1

9.3

19.4

11.7

10.7

22.4

Age 29–49

18.6

18.2

36.8

17.4

16.8

34.2

Age over 50

20.2

23.6

43.8

21.2

22.2

43.4

Total

48.9

51.1

100.0

50.3

49.7

100.0

Source: Own creation

4.3.2.2 Measurement Scales from previous studies were used (seven-point Likert-type scales, ranging from 1 for strongly disagree to 7 for strongly agree; see Table 4.2). Three items each measured the websites’ aesthetic appeal (Cai and Xu 2011), ease of use (Wells et al. 2011) and online-offline integration (Oh, Teo and Sambamurthy 2012). Security/privacy was measured with two items (Montoya-Weiss, Voss and Grewal 2003). Customer service (Montoya-Weiss, Voss and Grewal 2003) and channel consistency (Oh, Teo and Sambamurthy 2012, Van Baal 2014) were measured with four items each. Online trust was measured with three items (adapted from Eisingerich and Bell 2008), and brand equity was measured with four items (Verhoef, Langerak and Donkers 2007). Online and offline repurchase intentions were each measured with three items (adapted from Maxham III and Netemeyer 2002). Covariates were used because repurchase intentions can vary according to gender (0/1 = male/female) and age (Hult et al. 2019). General Internet expertise was controlled for because it can affect online and omni-channel use and purchase (“How would you characterize your level of expertise with the Internet?”, Montoya-Weiss, Voss and Grewal 2003). Perceived assortment variety and price were controlled, as they may be relevant for cross-channel repurchases for experienced consumers (“the variety of offerings is sufficient”; “the price level of the offerings is fair”, Blut, Teller and Floh 2018). The reliability of measurements was ensured across the three time points (see Table 4.2 and Table 4.3, e.g., corrected item-to-total correlations and Cronbach’s alpha). The values for construct and convergent validity were above the common thresholds. The average variance extracted exceeded the squared correlations of the constructs and supported discriminant validity. The fit values were satisfactory (Hair et al. 2018, p. 93).

AES2 AES2

The website is visually appealing

I like the pictures/images used in this website

5.0/1.2 4.9/1.2

It looks easy to find what I am looking for in this website EAS2

It is easy to order articles on this website

SEC2

I felt secure in providing credit card information for online transactions 4.5/1.2 4.5/1.2

COS2

[Retailer] provides reliable service through its website

3.8/1.4

3.7/1.4

[Retailer] provides convenient service through its website COS1

Customer service (Montoya-Weiss, Voss and Grewal 2003)

SEC1

I believe that my personal data are well protected on this website

Security/privacy (adapted from Montoya-Weiss, Voss and Grewal 2003)

EAS3

5.1/1.2

EAS1

Navigating this website is easy for me

4.8/1.4

4.7/1.4

.898

.778 .807

.963

.963 .500

.870

.789

.933 .730

.878

.962

.955 .756

.836

.740

.927

.927

.803

.747

.842

.858

.920

.915

.952

λ

Model 2 CR

.907

.879

.878

.878

.901 .814 .901 .814

.907

.879

.952 .965 .952 .965

.951

λ

Model 1 CR

.957

.787

(continued)

.787

.907 .912 .918 .912 .918

.957

.962 .962 .968 .962 .969

.897

.951

KMO ItTC α

4

Ease of use (adapted from Wells, Parboteeah and Valacich 2011)

AES1

I like the look and feel of the website

4.8/1.4

Time point one

Aesthetic appeal (Cai and Xu 2011)

MV/Std FL

Item

Construct

Table 4.2 Study 1: reliability and validity

144 Study 3: Importance of Marketing Instruments for Repurchase …

4.6/1.1

[Retailer] provides a high level of overall service through COS4 its website

4.7/1.7 4.8/1.6

I can easy purchase items through [retailer’s] website and OFI1 pick them up in its stores

I can easy return purchases made through website at [retailer’s] stores 4.0/1.4 4.1/1.3 4.2/1.4

CCO1

Product/service descriptions of [retailer] are consistent in CCO2 both the physical store and the online store

Prices in the offline store and the online store are identical CCO3

The offline store and the online store of [retailer] are aligned to each other

Channel consistency (Oh, Teo and Sambamurthy 2012, Van Baal 2014)

OFI2

4.7/1.7

The website allows me to search for products available in OFI3 the physical store

Online-offline integration (adapted from Oh, Teo and Sambamurthy 2012)

4.4/1.2

COS3

[Retailer] provides helpful assistance through its website

.910

.777

.873 .842

.946

.970

.850

.740

.821

.906

.922

.849

.724

.865

.967

.785

.900

λ

Model 2 CR

.869

.869

.950 .949 .950 .949

.967

.786

.900

λ

Model 1 CR

.908

.782

(continued)

.906

.783

.912 .913 .868 .913 .869

.948

KMO ItTC α

.868 .749

.763

.936

MV/Std FL

Item

Construct

Table 4.2 (continued)

4.3 Empirical Studies 145

ONT1 ONT2 ONT3

[Retailer’s] online store can be trusted at all times

[Retailer’s] online store is honest and truthful

I have confidence in [retailer’s] online store

ONB2 ONB3 ONB4

[Retailer’s] online store represents a well-known brand

[Retailer’s] online store represents an attractive brand

[Retailer’s] online store represents a unique brand

3.6/1.5

4.2/1.5

4.2/1.5

4.2/1.5

Online repurchase intention (adapted from Maxham III Time point three and Netemeyer 2002)

ONB1

[Retailer’s] online store represents a strong brand

3.5/1.5

3.5/1.4

3.5/1.5

.785

.918

.958

.946 .830

.957

.973

.958 .784

.838

.769

.886

.915

.907

.940

.952

.940

.792

.958

.844

λ

Model 2 CR

.960

.960

.974 .971 .974 .971

.957

.844

λ

Model 1 CR

.773

.907

.969

(continued)

.772

.907

.970

.945 .987 .950 .946 .950

.974

KMO ItTC α

4

Online brand equity (Verhoef, Langerak and Donkers 2007)

Time point two

Online trust (adapted from Eisingerich and Bell 2008)

4.4/1.4

CCO4

Product assortments in the offline store and the online store are identical

MV/Std FL

Item

Construct

Table 4.2 (continued)

146 Study 3: Importance of Marketing Instruments for Repurchase …

2.9/1.6 3.7/1.9

In the future, I will be more likely to buy from [retailer’s] OFRPI2 offline store than from other offline stores

It is very likely that I will buy from retailer’s offline store OFRPI3

.863

.827

.728

.664

.745

.713

.720

.843

.839

.777

.844 .808

.820

λ

Model 1 CR

.858

.849 .825

.738

λ

Model 2 CR

Source: Own creation

Confirmatory model fits: Model 1: CFI .971, TLI .966, RMSEA .048, SRMR .039, χ2 (341) = 651.289, SCF = 1.07 Model 2: CFI .969, TLI .964, RMSEA .050, SRMR .039, χ2 (341) = 672.506, SCF = 1.08 Notes: MV/Std. = Mean values and standard deviations, FL = Factor loadings (exploratory), KMO = Kaiser–Meyer–Olkin Criterion (≥ .5), ItTC = Item-to-Total Correlation (≥ .3), α = Cronbach’s alpha (≥ .7). All items measured on 7-point Likert-type scales: 1 = strongly disagree, 7 = strongly agree

3.5/1.9

In the future, I intend to continue buying from [retailer’s] OFRPI1 offline store

Offline repurchase intention (adapted from Maxham III and Netemeyer 2002) .727 .718

.773

It is very likely that I will buy from retailer’s online store

ONRPI3 3.1/1.9

.809

.690

KMO ItTC α

In the future, I will be more likely to buy from [retailer’s] ONRPI2 2.6/1.5 online store than from other online stores

MV/Std FL .823 .728

Item

In the future, I intend to continue buying from [retailer’s] ONRPI1 2.8/1.8 online store

Construct

Table 4.2 (continued)

4.3 Empirical Studies 147

148

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.3 Study 1: discriminant validity Constructs

1

2

3

4

5

6

7

8

9

Model 1 1

Aesthetic appeal (1)

.870

2

Ease of use (1)

.413

.752

3

Security/privacy (1)

.234

.191

.927

4

Customer service .307 (1)

.582

.160

.722

5

Online-offline integration (1)

.099

.252

.031

.076

.864

6

Channel consistency (1)

.176

.135

.076

.160

.074

.725

7

Online trust (2)

.160

.106

.197

.197

.059

.084

.927

8

Online brand equity (2)

.289

.116

.138

.138

.066

.140

.333

.950

9

Online repurchase intention (3)

.148

.086

.116

.116

.068

.121

.360

.271

.643

Model 2 1

Aesthetic appeal (1)

.870

2

Ease of use (1)

.413

.752

3

Security/privacy (1)

.234

.192

.927

4

Customer service .307 (1)

.584

.160

.722

5

Online-offline integration (1)

.099

.252

.031

.076

.864

6

Channel consistency (1)

.177

.135

.076

.160

.075

.726 (continued)

4.3 Empirical Studies

149

Table 4.3 (continued) Constructs

1

2

3

4

5

6

7

8

9

Model 1 7

Online trust (2)

.160

.068

.197

.197

.059

.084

.927

8

Online brand equity (2)

.289

.116

.146

.146

.066

.141

.323

.815

9

Online repurchase intention (3)

.156

.078

.082

.082

.077

.243

.333

.249

.653

Notes: AVE = Average Variance Extracted (≥ .5, Fornell and Larcker 1981), (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM. Values in italics represent squared correlations between constructs, values in bold represent the AVE of the construct Source: Own creation

Collecting data at different time points and using an appropriate questionnaire design reduce the potential threat of CMV ex ante (e.g., Fuller et al. 2016). An additional single-factor test showed lower fit values than the proposed models (see Appendix 4.1).

4.3.2.3 Method To reduce the possibility of omitted variables, endogeneity was tested with the IV method (Antonakis et al. 2014). Single-item overall perceptions of offline marketing instruments related to each online- and omni-channel instrument were included as IVs because they are theoretically related in the context of omnichannel retailing (Blut, Teller and Floh 2018). F-tests showed that the IVs were strong predictors. In addition to the efficient models, consistent models including the IVs were calculated and did not significantly differ (see Appendix 4.2; z-values were all < 1.96). Endogeneity does not seem to be a serious problem. To test the hypotheses, sequential mediation modeling using Mplus was applied. Sequential mediations allow for temporal precedence since the directional influences operate over time and provide better estimates of longitudinal parameters than cross-sectional models. They require the measurement of variables at different points in time; mediator Mt is measured after its presumed cause Xt−1 and prior to its presumed effect Yt+1 (Mitchell and Maxwell 2013).

150

4.3.3

4

Study 3: Importance of Marketing Instruments for Repurchase …

Study 2: Cross-lagged Panel Design

The same sampling and scales as in study 1 were used to measure online trust and brand equity as well as online and offline repurchase intentions (with identical, satisfactory values for reliability, validity and discriminant validity; see Table 4.4 to Table 4.6). Age, gender and general Internet expertise were again used as covariates. Regarding endogeneity, the instruments are omitted variables in study 2 but can hardly be used due to complexity and theoretical constraints (i.e., simultaneous computing in cross-lagged models and effects on online trust and brand equity). However, as a proposal, we tested additional brand attachment as an IV for online brand equity (which is theoretically a strong predictor for brand equity, Park et al. 2010) and offline trust, as it was shown to be strongly associated with online trust, Bock et al. 2012). The F-test showed that the IVs are predictors; however, consistent models do not significantly differ from efficient models (see Appendix 4.2). We indicate a reduced probability of omitted variables. Table 4.4 Measurement invariance was determined to ensure the comparability of the results across the three time points (Van de Schoot, Lugtig and Hox 2012). The results indicated a good fit for all models (online: χ2 (380) = 1,064.331, p > 0.05, offline: χ2 (381) = 1,142.603, p > 0.05, see Appendix 4.3). For the hypothesis tests, cross-lagged structural equation modeling was applied (Allison et al. 2017). This method facilitates the analysis of reciprocal effects and is based on two assumptions: a variable Xt can be predicted by Xt−1 , and Xt can be influenced in a cross-lagged manner by Yt−1 (Zyphur et al. 2019, see Appendix 4.4). All fit values were satisfactory. Next, we discuss the results based on standardized coefficients and provide rationales for insignificant findings.

4.3.4

Results

The expected indirect-only mediation of the marketing instruments’ effects on repurchase intentions was supported (Zhao, Lynch Jr. and Chen 2010). Only channel consistency affects offline repurchase intentions indirectly and directly (β = 0.256, p > 0.001, see Table 4.7), as the following results show. The results of study 1 in models 1–2 support H1a/b. Aesthetic appeal positively affects online and offline repurchase intentions, as mediated by online trust (online: β = 0.084, p < 0.01, offline: β = 0.062, p < 0.01) and online brand equity (online: β = 0.170, p < 0.001, offline: β = 0.184, p < 0.001). Ease of use does not affect repurchase intentions through online trust (online: β = −0.059, p > 0.1, offline: β = −0.045, p > 0.1) or online brand equity (online:

4.1/2.0

2.5/1.5

3.6/2.0

OFRPI2

OFRPI3

3.1/2.0

ONRPI3

OFRPI1

2.3/1.4

ONRPI2

3.5/1.5

ONB4

3.2/1.9

4.2/1.5

ONB3

ONRPI1

4.2/1.5

ONB2

.796

.807

.873

.774

.787

.901

.781

.878

.968

.906

.971

.940

.957

FL

.733

.720

.818

.779

KMO

.727

.732

.770

.712

.715

.782

.758

.844

.913

.862

.945

.922

.939

ItTC

.855

.852

.934

.969

α

3.6/1.9

2.6/1.5

4.0/2.1

3.0/1.8

2.2/1.5

3.0/1.9

3.6/1.5

4.2/1.5

4.2/1.5

4.2/1.5

3.5/1.5

3.5/1.4

3.5/1.5

.866

.801

.840

.810

.792

.890

.785

.918

.958

.946

.957

.973

.958

FL

Time point two MV/Std.

.739

.731

.830

.784

KMO

.742

.726

.787

.778

.734

.763

.769

.886

.915

.907

.940

.952

.940

ItTC

.864

.867

.945

974

α

3.7/1.9

2.9/1.6

3.5/1.9

3.1/1.9

2.6/1.5

2.8/1.8

4.0/1.5

4.1/1.5

4.1/1.5

4.1/1.4

3.5/1.5

3.5/1.4

3.6/1.5

.863

.827

.727

.773

.809

.823

.867

.946

.953

.940

.961

.974

.967

FL

Time point three MV/Std.

.718

.728

.866

.787

KMO

.728

.664

.745

.713

.720

.690

.851

.920

.924

.913

.946

.957

.951

ItTC

Source: Own creation

Notes: RPI = Repurchase intention, MV/Std. = Mean values and standard deviations, FL = Factor loadings (exploratory), KMO = Kaiser-Meyer-Olkin Criterion (≥. 5), ItTC = Item-to-Total Correlation (≥ .3), α = Cronbach’s alpha (≥. 7). All items measured on 7-point Likert-type scales: 1 = strongly disagree, 7 = strongly agree.

Offline RPI

Online RPI

4.3/1.5

3.6/1.5

ONT3

ONB1

3.6/1.4

ONT2

Online brand equity

3.6/1.5

ONT1

Online trust

MV/Std.

Item

Construct

Time point one

Table 4.4 Study 2: reliability and validity

.843

.839

.960

.977

α

4.3 Empirical Studies 151

Offline repurchase intention

Online repurchase intention .786

ONRPI3 OFRPI2

OFRPI1

.786

ONRPI2

.892

.762 .862

ONB4 ONRPI1

.857

ONB3

.929 .975

.934

ONB2

ONB1

.866

.934

.970

.817

.866

.761

.855

.976

.929

.972

.939

.957

.870

.946

.974

.825

.790

.878

.775

.907

.968

.951

.959

.971

.958

.946

.974

CR

.806

.854

.773

.907

.969

.951

.960

.971

.958

λ

.844

.939

.978

.797

.793

.815

.758

.887

.974

.934

.964

.971

.967

λ

.825

.750

.758

.885

.973

.936

.963

.972

.968

λ

(continued)

.850

.939

.978

CR

Time point three CR

4

Online brand equity

.972

.956

ONT3

.970 .939

ONT1

Online trust

λ

CR

Time point two λ

λ CR

Time point one CR

ONT2

Item

Construct

Table 4.5 Study 2: reliability and validity II

152 Study 3: Importance of Marketing Instruments for Repurchase …

OFRPI3

Item .795

λ

CR

Time point two λ

λ CR

Time point one CR

CR .847

λ

λ CR

Time point three CR

.848

λ

Source: Own creation

Confirmatory model fits: Time point one (Model 1): CFI .993, TLI .990, RMSEA .044, SRMR .018, χ2 (32) = 55.124, SCF = 1.22 Time point two (Model 1): CFI .992, TLI .989, RMSEA .050, SRMR .015, χ2 (32) = 62.379, SCF = 1.13 Time point three (Model 1): CFI .975, TLI .964, RMSEA .090, SRMR .029, χ2 (32) = 129.967, SCF = 1.10 Time point one (Model 2): CFI .989, TLI .985, RMSEA .054, SRMR .023, χ2 (32) = 67.788, SCF = 1.21 Time point two (Model 2): CFI .987, TLI .982, RMSEA .065, SRMR .023, χ2 (32) = 83.388, SCF = 1.11 Time point three (Model 2): CFI .965, TLI .951, RMSEA .107, SRMR .035, χ2 (32) = 168.861, SCF = 1.07 Notes: CR = Composite reliability (≥ .6), λ = Standardized factor loadings (confirmatory) (≥ .5), SCF = Scaling correction factor for MLM

Construct

Table 4.5 (continued)

4.3 Empirical Studies 153

154

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.6 Study 2: discriminant validity Time point one Constructs

1

2

Time point two 3

1

2

Time point three 3

1

2

3

Model 1 1

Online trust

2

Online .366 brand equity

.914 .782

3

Online repurchase intention

.382

.554

.927

.677

.936

.324

.816

.579

.339

.692

.297

.795

.607

.311

.643

Model 2 1

Online trust

2

Online .365 brand equity

.914 .781

3

Offline repurchase intention

.366

.384

.927

.683

.936

.334

.816

.334

.383

.669

.297

.795

.437

.410

.654

Notes: AVE = Average Variance Extracted (≥ .5, Fornell and Larcker 1981), SCF = Scaling correction factor for MLM. Values in italics represent squared correlations between constructs, values in bold represent the AVE of the construct Source: Own creation

β = −0.047, p > 0.1, offline: β = −0.051, p > 0.1). For experienced consumers, perceived ease of use of a website is not a salient or relevant signal relative to other instruments. It does not activate online trust or brand equity associations in memory in online or offline channel repurchase situations. The carefully formulated H2a/b are rejected. Security/privacy positively affects repurchase intentions, and this relationship is mediated by online trust (online: β = 0.157, p < 0.001, offline: β = 0.118, p < 0.001) and online brand equity (online: β = 0.056, p < 0.05, offline: β = 0.061, p < 0.01), H3a/b are supported. Customer service positively affects repurchase intentions through online trust (online: β = 0.080, p < 0.05, offline: β = 0.060, p < 0.05) but not through online brand equity (online: β = 0.003, p > 0.1, offline: β = 0.001, p > 0.1). H4a is supported, while H4b is rejected. In line with the hypothesized mechanisms, customer service on a website is an important cue for online and offline repurchasing. However, while studies have shown its relevance for online brand image, customer service does not affect brand equity (similar to brand equity studies

4.3 Empirical Studies

155

Table 4.7 Study 1: results of the sequential mediation models Model 1: Online RPI

Model 2: Offline RPI

β

β

P

P

Direct effects Aesthetic appeal (1)

→ Online trust (2)

.166

**

.163

**

Ease of use (1)

→ Online trust (2)

−.117

ns

−.119

ns

Security/privacy (1)

→ Online trust (2)

.311

***

.312

***

Customer service (1)

→ Online trust (2)

.158

*

.158

*

Online-offline integration (1)

→ Online trust (2)

.079

ns

.082

ns

Channel consistency (1)

→ Online trust (2)

.114

*

.115

*

Aesthetic appeal (1)

→ Online brand equity (2)

.443

***

.443

***

Ease of use (1)

→ Online brand equity (2)

−.122

ns

−.123

ns

Security/privacy (1)

→ Online brand equity (2)

.146

*

.147

**

Customer service (1)

→ Online brand equity (2)

.007

ns

.003

ns

Online-offline integration (1)

→ Online brand equity (2)

.108

*

.109

*

Channel consistency (1)

→ Online brand equity (2)

.159

**

.162

**

Aesthetic appeal (1)

→ RPI (2)

−.022

ns

−.023

ns

Ease of use (1)

→ RPI (2)

.038

ns

.020

ns

Security/privacy (1)

→ RPI (2)

−.038

ns

−.084

ns

Customer service (1)

→ RPI (2)

−.077

ns

−.104

ns

Online-offline integration (1)

→ RPI (2)

.064

ns

.087

ns (continued)

156

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.7 (continued) Model 1: Online RPI

Model 2: Offline RPI

β

β

P

P

Channel consistency (1)

→ RPI (2)

.078

ns

.256

***

Online brand equity (2)

→ RPI (2)

.385

***

.417

***

Online trust (2)

→ RPI (2)

.504

***

.378

***

Indirect effects Aesthetic appeal (1)

→ Online trust (2)

→ RPI (3)

.084

**

.062

**

Ease of use (1)

→ Online trust (2)

→ RPI (3)

−.059

ns

−.045

ns

Security/privacy (1)

→ Online trust (2)

→ RPI (3)

.157

***

.118

***

Customer service (1)

→ Online trust (2)

→ RPI (3)

.080

*

.060

*

Online-offline integration (1)

→ Online trust (2)

→ RPI (3)

.040

ns

.031

ns

Channel consistency (1)

→ Online trust (2)

→ RPI (3)

.058

*

.044

*

Aesthetic appeal (1)

→ Online brand equity (2)

→ RPI (3)

.170

***

.184

***

Ease of use (1)

→ Online brand equity (2)

→ RPI (3)

−.047

ns

−.051

ns

Security/privacy (1)

→ Online brand equity (2)

→ RPI (3)

.056

*

.061

**

Customer service (1)

→ Online brand equity (2)

→ RPI (3)

.003

ns

.001

ns

Online-offline integration (1)

→ Online brand equity (2)

→ RPI (3)

.042

*

.045

*

Channel consistency (1)

→ Online brand equity (2)

→ RPI (3)

.061

**

.067

** (continued)

4.3 Empirical Studies

157

Table 4.7 (continued) Model 1: Online RPI

Model 2: Offline RPI

β

β

P

P

Total effects Aesthetic appeal (1)

→ RPI (3)

.232

***

.224

***

Ease of use (1)

→ RPI (3)

−.068

ns

−.076

ns

Security/privacy (1)

→ RPI (3)

.175

**

.095

Customer service (1)

→ RPI (3)

.005

ns

−.043

ns

Online-offline integration (1)

→ RPI (3)

.146

*

.163

**

Channel consistency (1)

→ RPI (3)

.197

**

.367

***

†(.067)

Covariates Gender (1)

→ RPI (3)

.046

ns

.056

ns

Age (1)

→ RPI (3)

−.011

ns

−.053

ns

Internet expertise (1)

→ RPI (3)

.038

ns

.040

ns

Assortment variety (1)

→ RPI (3)

.078

ns

−.036

ns

Price fairness (1)

→ RPI (3)

.015

ns

.007

ns

Structural model fit: Model 1: CFI .932, TLI .920, RMSEA .067, SRMR .128, χ2 (470) = 1254.596, SCF = 1.00 Model 2: CFI .930, TLI .918, RMSEA .068, SRMR .127, χ2 (470) = 1279.778, SCF = 1.01 Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377, β = standardized coefficients *** p < .001, ** p < .01, * p < .05, † p < .10, ns = not significant Source: Own creation

158

4

Study 3: Importance of Marketing Instruments for Repurchase …

for pure offline retailers). Surprisingly, online service cues—relative to other instruments—narrowly activate brand associations in memory. Online-offline integration does not significantly affect repurchase intentions through online trust (online: β = 0.040, p > 0.1, offline: β = 0.031, p > 0.1) but positively affects it through online brand equity (online: β = 0.042, p < 0.05, offline: β = 0.045, p < 0.05). H5a is rejected, while H5b is supported. Online-offline integration affects online and offline repurchase behavior, as hypothesized, but not trust associations. We believe that experienced consumers access this signal or cue but it does not facilitate trust—in relation to brand—nodes in mind for repurchases. However, the total effects discussed later show that online-offline integration is important for cross-channel repurchasing. Channel consistency positively affects the repurchase intentions mediated by online trust (online: β = 0.058, p < 0.05, offline: β = 0.044, p < 0.05) and online brand equity (online: β = 0.061, p < 0.01, offline: β = 0.067, p < 0.01); H6a/b are supported. The covariates gender, age, Internet expertise, assortment variety and price are insignificant. The results of study 2 in models 1–2 support H7a/b (see Table 4.8). Online trust positively affects online brand equity (online: β1−2 = 0.097, p < 0.01, β2−3 = 0.104, p < 0.01, offline: β1−2 = 0.100, p < 0.001, β2−3 = 0.108, p < 0.01), and online brand equity affects online trust (online: β1−2 = 0.086, p < 0.01, β2−3 = 0.089, p < 0.01, offline: β1−2 = 0.084, p < 0.01, β2−3 = 0.084, p < 0.01). Thus, H7a/b are supported. Online trust has a total effect on repurchase intentions over time (online: β = 0.072, p < 0.1, offline: β = 0.136, p < 0.001), as does online brand equity (online: β = 0.112, p < 0.01, offline: β = 0.140, p < 0.01). Hypotheses H8a/b are supported. Among the covariates, gender exerts positive and age negative effects, while Internet expertise has insignificant effects. As expected, stronger effects on repurchase intentions emerge for females and older consumers. Spreading activation—without further stimuli—differs because cognitive processing is known to depend on gender and accumulated experiences (Fang et al. 2016).

4.3 Empirical Studies

159

Table 4.8 Study 2: results of the cross-lagged models Model 1: Online RPI

Model 2: Offline RPI

β

β

P

P

Direct effects Online trust (1)

→ Online brand equity (2)

.097

**

.100

***

Online brand equity (1)

→ Online trust (2)

.086

**

.084

**

Online trust (1)

→ RPI (2)

.047

ns

.096

***

Online brand equity (1)

→ RPI (2)

.082

*

.103

***

Online trust (1)

→ Online trust (2)

.619

***

.614

***

Online brand equity (1)

→ Online brand equity (2)

.610

***

.611

***

RPI (1)

→ RPI (2)

.682

***

.655

***

Online trust (2)

→ Online brand equity (3)

.104

**

.108

**

Online brand equity (2)

→ Online trust (3)

.089

**

.084

**

Online trust (2)

→ RPI (3)

.051

ns

.104

***

Online brand equity (2)

→ RPI (3)

.088

**

.109

**

Online trust (2)

→ Online trust (3)

.650

***

.664

***

Online brand equity (2)

→ Online brand equity (3)

.646

***

.640

***

RPI (2)

→ RPI (3)

.657

***

.636

***

.525

***

.565

***

R2 RPI (3) Total effects Online trust (1)

→ RPI (3)

.072

†(.060)

.136

***

Online brand equity (1)

→ RPI (3)

.112

**

.140

**

Diff. in total effects

t = 1.988*

t = .333 ns (continued)

160

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.8 (continued) Model 1: Online RPI

Model 2: Offline RPI

β

β

P

P

Covariates Gender (1)

→ RPI (1)

.058

**

.090

***

Gender (2)

→ RPI (2)

.063

**

.096

***

Gender (3)

→ RPI (3)

.066

**

.100

***

Age (1)

→ RPI (1)

−.058

**

−.064

**

Age (2)

→ RPI (2)

−.064

**

−.069

**

Age (3)

→ RPI (3)

−.068

**

−.072

**

Internet expertise (1)

→ RPI (1)

−.009

ns

−.016

ns

Internet expertise (2)

→ RPI (2)

−.009

ns

−.017

ns

Internet expertise (3)

→ RPI (3)

−.010

ns

−.017

ns

Structural model fits: Model 1: CFI .928, TLI .924; RMSEA .067, SRMR .183, χ2 (662) = 1794.952, SCF = .86 Model 2: CFI .926, TLI .921; RMSEA .069, SRMR .169, χ2 (662) = 1840.577, SCF = .86 Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377. Standardized coefficients are shown *** p < .001, ** p < .01, * p < .05, † p < .10, ns = not significant Source: Own creation

4.3.5

Stability Tests and Alternative Models

Stability tests and alternative models strengthen our observations. First, we provided stability tests by removing trust as a mediator (see Table 4.9). The results for brand equity remain stable (e.g., for online purchase intentions βAES = 0.174, p < 0.001, βEAS = −0.049, p > 0.1, βIOF = 0.042, p < 0.05, βSEC = 0.058, p < 0.01, βCCO = 0.064, p < 0.01, βCOS = 0.001, p > 0.1). We removed brand equity as a mediator, and the results remain stable as well (e.g., for offline purchase intentions βAES = 0.062, p < 0.01, βEAS = −0.045, p > 0.1, βIOF = 0.028, p > 0.1, βSEC = 0.114, p < 0.001, βCCO = 0.043, p < 0.05, βCOS = 0.062, p < 0.05; see Table 4.10). Both stability tests support the hypothesis tests. Second, we provide alternative models by using repurchase intentions toward the retailer as the dependent variable (measure: “In the future, I intend to continue buying from [retailer]”; “In the future, I will be more likely to buy from [retailer] than from others”; “It is very likely that I will buy from [retailer]”, Maxham III

4.3 Empirical Studies

161

Table 4.9 Study 1: results of the alternative model (online brand equity only) Model 1: Online RPI

Model 2: Offline RPI

β

β

p

P

Direct effects Aesthetic appeal (1)

→ Online brand equity (2)

.442

***

.442

***

Ease of use (1)

→ Online brand equity (2)

−.126

ns

−.124

ns

Security/privacy (1)

→ Online brand equity (2)

.148

**

.149

**

Customer service (1)

→ Online brand equity (2)

.003

ns

.002

ns

Online-offline integration (1)

→ Online brand equity (2)

.108

*

.108

*

Channel consistency (1)

→ Online brand equity (2)

.164

**

.164

**

Aesthetic appeal (1)

→ RPI (2)

.059

ns

.036

ns

Ease of use (1)

→ RPI (2)

−.061

ns

−.048

ns

Security/privacy (1)

→ RPI (2)

.124

*

.039

ns

Customer service (1)

→ RPI (2)

−.015

ns

−.057

ns

Online-offline integration (1)

→ RPI (2)

.104

ns

.118

*

Channel consistency (1)

→ RPI (2)

.137

*

.302

***

Online brand equity (2)

→ RPI (2)

.394

***

.425

***

Indirect effects Aesthetic appeal (1)

→ Online brand equity (2)

→ RPI (3)

.174

***

.188

***

Ease of use (1)

→ Online brand equity (2)

→ RPI (3)

−.049

ns

−.101

ns (continued)

162

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.9 (continued) Model 1: Online RPI β

Model 2: Offline RPI β

p

P

Security/privacy (1)

→ Online brand equity (2)

→ RPI (3)

.058

*

.063

**

Customer service (1)

→ Online brand equity (2)

→ RPI (3)

.001

ns

.001

ns

Online-offline integration (1)

→ Online brand equity (2)

→ RPI (3)

.042

*

.046

*

Channel consistency (1)

→ Online brand equity (2)

→ RPI (3)

.064

**

.070

**

Total effects Aesthetic appeal (1)

→ RPI (3)

.233

***

.224

***

Ease of use (1)

→ RPI (3)

−.111

ns

−.053

ns

Security/privacy (1)

→ RPI (3)

.182

**

.102

Customer service (1)

→ RPI (3)

−.013

ns

−.056

ns

Online-offline integration (1)

→ RPI (3)

.146

*

.164

**

Channel consistency (1)

→ RPI (3)

.201

**

.372

***

Gender (1)

→ RPI (3)

.034

ns

.050

ns

Age (1)

→ RPI (3)

.011

ns

−.040

ns

Internet expertise (1)

→ RPI (3)

.062

ns

−.026

ns

Assortment variety (1)

→ RPI (3)

−.047

ns

−.017

ns

†(.054)

Covariates

(continued)

4.3 Empirical Studies

163

Table 4.9 (continued)

Price fairness (1)

→ RPI (3)

Model 1: Online RPI

Model 2: Offline RPI

β

β

p .077

ns

P .031

ns

Structural model fit: Model 1: CFI .924, TLI .912, RMSEA .070, SRMR .133, χ2 (396) = 1132.203, SCF = 1.02 Model 2: CFI .922, TLI .910, RMSEA .071, SRMR .134, χ2 (396) = 1154.357, SCF = 1.02 Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377, β = standardized coefficients ***p < .001, **p < .01, *p < .05, † p < .10, ns = not significant Source: Own creation Table 4.10 Study 1: results of the alternative model (online trust only) Model 1: Online RPI

Model 2: Offline RPI

β

β

p

P

Direct effects Aesthetic appeal (1)

→ Online trust (2)

.165

Ease of use (1)

→ Online trust (2)

Security/privacy (1)

→ Online trust (2)

Customer service (1)

**

.165

**

−.120

ns

−.120

ns

.306

***

.306

***

→ Online trust (2)

.166

*

.166

*

Online-offline integration (1)

→ Online trust (2)

.074

ns

.074

ns

Channel consistency (1)

→ Online trust (2)

.115

*

.116

*

Aesthetic appeal (1)

→ RPI (2)

.144

*

.155

**

Ease of use (1)

→ RPI (2)

.007

ns

.005

ns (continued)

164

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.10 (continued) Model 1: Online RPI β

Model 2: Offline RPI β

p

P

Security/privacy (1)

→ RPI (2)

.024

ns

−.016

ns

Customer service (1)

→ RPI (2)

−.079

ns

−.108

ns

Online-offline integration (1)

→ RPI (2)

.107

ns

.134

*

Channel consistency (1)

→ RPI (2)

.137

**

.320

***

Online trust (2) → RPI (2)

.504

***

.374

***

**

.062

**

Indirect effects Aesthetic appeal (1)

→ Online trust (2)

→ RPI (3)

.083

Ease of use (1)

→ Online trust (2)

→ RPI (3)

−.060

ns

−.045

ns

Security/privacy (1)

→ Online trust (2)

→ RPI (3)

.154

***

.114

***

Customer service (1)

→ Online trust (2)

→ RPI (3)

.084

*

.062

*

Online-offline integration (1)

→ Online trust (2)

→ RPI (3)

.037

ns

.028

ns

Channel consistency (1)

→ Online trust (2)

→ RPI (3)

.058

*

.043

*

***

.217

*** ns

Total effects Aesthetic appeal (1)

→ RPI (3)

.227

Ease of use (1)

→ RPI (3)

−.053

ns

−.050

Security/privacy (1)

→ RPI (3)

.178

**

.098

Customer service (1)

→ RPI (3)

.005

ns

−.045

ns

Online-offline integration (1)

→ RPI (3)

.145

*

.161

**

†(.059)

(continued)

4.3 Empirical Studies

165

Table 4.10 (continued) Model 1: Online RPI

Model 2: Offline RPI

β

β

p

P

→ RPI (3)

.195

**

.363

***

Gender (1)

→ RPI (3)

.033

ns

.036

ns

Age (1)

→ RPI (3)

−.016

ns

−.060

ns

Internet expertise (1)

→ RPI (3)

.020

ns

−.065

ns

Assortment variety (1)

→ RPI (3)

−.070

ns

−.025

ns

Price fairness (1)

→ RPI (3)

−.004

ns

−.035

ns

Channel consistency (1) Covariates

Structural model fit: Model 1: CFI .924, TLI .912, RMSEA .073, SRMR .139, χ2 (367) = 1101.636, SCF = 1.01 Model 2: CFI .922, TLI .910, RMSEA .074, SRMR .137, χ2 (367) = 1118.415, SCF = 1.01 Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377, β = standardized coefficients ***p < .001, **p < .01, *p < .05, † p < .10, ns = not significant Source: Own creation

and Netemeyer 2002; see Table 4.11 and Table 4.12). In study 1, the results are identical to those for the offline model (e.g., for brand equity mediation βAES = 0.200, p < 0.001, βEAS = −0.055, p > 0.1, βIOF = 0.050, p < 0.05, βSEC = 0.067, p < 0.01, βCCO = 0.073, p < 0.01, βCOS = 0.001, p > 0.1). In study 2, online brand equity still positively affects online trust (e.g., β2−3 = 0.090, p < 0.01) and vice versa (e.g., β2−3 = 0.107, p < 0.01). The results support our theoretical rationale at the omni-channel retailer level. Third, we use an overall retailer brand equity measure (“[Retailer] represents a strong brand, a well-known brand, an attractive brand and a unique brand”, Verhoef, Langerak and Donkers 2007; see Table 4.13 and Table 4.14). In study 1, the results are identical, e.g., for online trust on online repurchase intentions (βAES = 0.086, p < 0.001, βEAS = −0.052, p > 0.1, βIOF = 0.032, p > 0.1, βSEC = 0.147, p < 0.001, βCCO = 0.048, p < 0.1, βCOS = 0.078, p < 0.05) or offline: βAES

166

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.11 Study 1: results of the alternative model (total repurchase intention) Model 1: Total RPI β

p

Direct effects Aesthetic appeal (1)

→ Online trust (2)

.163

**

Ease of use (1)

→ Online trust (2)

−.122

ns

Security/privacy (1)

→ Online trust (2)

.312 ***

Customer service (1)

→ Online trust (2)

.158

*

Online-offline integration (1)

→ Online trust (2)

.083

ns

Channel consistency (1)

→ Online trust (2)

.115

*

Aesthetic appeal (1)

→ Online brand equity (2)

.443 ***

Ease of use (1)

→ Online brand equity (2)

−.122

Security/privacy (1)

→ Online brand equity (2)

.147

*

Customer service (1)

→ Online brand equity (2)

.002

ns

Online-offline integration (1)

→ Online brand equity (2)

.109

*

Channel consistency (1)

→ Online brand equity (2)

.162

**

Aesthetic appeal (1)

→ RPI (2)

−.057

ns

Ease of use (1)

→ RPI (2)

.075

ns

Security/privacy (1)

→ RPI (2)

−.086

ns

Customer service (1)

→ RPI (2)

−.086

ns

Online-offline integration (1)

→ RPI (2)

.092

ns

Channel consistency (1)

→ RPI (2)

.198 ***

Online brand equity (2)

→ RPI (2)

.453 ***

Online trust (2)

→ RPI (2)

.401 ***

ns

Indirect effects Aesthetic appeal (1)

→ Online trust (2)

→ RPI (3)

.065

** ns

→ Online trust (2)

→ RPI (3) −.048

Security/privacy (1)

→ Online trust (2)

→ RPI (3)

.125 ***

Customer service (1)

→ Online trust (2)

→ RPI (3)

.063

*

Online-offline integration (1)

→ Online trust (2)

→ RPI (3)

.033

ns

Ease of use (1)

(continued)

4.3 Empirical Studies

167

Table 4.11 (continued) Model 1: Total RPI β Channel consistency (1) Aesthetic appeal (1)

→ Online trust (2)

→ RPI (3)

→ Online brand equity (2) → RPI (3)

p .046

*

.200 ***

Ease of use (1)

→ Online brand equity (2) → RPI (3) −.055

ns

Security/privacy (1)

→ Online brand equity (2) → RPI (3)

.067

**

Customer service (1)

→ Online brand equity (2) → RPI (3)

.001

ns

Online-offline integration (1)

→ Online brand equity (2) → RPI (3)

.050

*

Channel consistency (1)

→ Online brand equity (2) → RPI (3)

.073

**

→ RPI (3)

.208 ***

Total effects Aesthetic appeal (1) Ease of use (1)

→ RPI (3)

−.028

Security/privacy (1)

→ RPI (3)

.106

*

Customer service (1)

→ RPI (3)

.021

ns

Online-offline integration (1)

→ RPI (3)

.175

*

Channel consistency (1)

→ RPI (3)

.317 ***

→ RPI (3)

.064

ns

Covariates Gender (1)

ns

→ RPI (3)

−.034

ns

Internet expertise (1)

→ RPI (3)

.043

ns

Assortment variety (1)

→ RPI (3)

−.031

ns

Price fairness (1)

→ RPI (3)

−.031

ns

Age (1)

Structural model fit: CFI .930, TLI .918, RMSEA .068, SRMR .128, χ2 (470) = 1283.748, SCF = 1.01 Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377, β = standardized coefficients ***p < .001, **p < .01, *p < .05, † p < .10, ns = not significant Source: Own creation

168

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.12 Study 2: results of the alternative model (total repurchase intention) Model 1: Total RPI β

p

Direct effects Online trust (1)

→ Online brand equity (2)

.098

**

Online brand equity (1)

→ Online trust (2)

.085

**

Online trust (1)

→ RPI (2)

.096

***

Online brand equity (1)

→ RPI (2)

.139

***

Online trust (1)

→ Online trust (2)

.613

***

Online brand equity (1)

→ Online brand equity (2)

.610

***

RPI (1)

→ RPI (2)

.675

***

Online trust (2)

→ Online brand equity (3)

.107

**

Online brand equity (2)

→ Online trust (3)

.090

**

Online trust (2)

→ RPI (3)

.102

***

Online brand equity (2)

→ RPI (3)

.145

***

Online trust (2)

→ Online trust (3)

.666

***

Online brand equity (2)

→ Online brand equity (3)

.646

***

RPI (2)

→ RPI (3)

.675

***

.581

***

R2

RPI (3)

Total effects Online trust (1)

→ RPI (3)

.137

***

Online brand equity (1)

→ RPI (3)

.184

**

5.658

**

Diff. in total effects Covariates Gender (1)

→ RPI (1)

.095

***

Gender (2)

→ RPI (2)

.105

***

Gender (3)

→ RPI (3)

.107

***

Age (1)

→ RPI (1)

−.054

**

Age (2)

→ RPI (2)

−.059

**

Age (3)

→ RPI (3)

−.061

**

Internet expertise (1)

→ RPI (1)

−.007

ns

Internet expertise (2)

→ RPI (2)

−.007

ns (continued)

4.3 Empirical Studies

169

Table 4.12 (continued) Model 1: Total RPI β Internet expertise (3)

→ RPI (3)

p

−.007

ns

Structural model fits: CFI .927, TLI .922; RMSEA .068, SRMR .173, χ2 (662) = 1823.810, SCF = .86 Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377. Standardized coefficients are shown ***p < .001, **p < .01, *p < .05, † p < .10, ns = not significant Source: Own creation

= 0.062, p < 0.01, βEAS = −0.040, p > 0.1, βIOF = 0.025, p > 0.1, βSEC = 0.110, p < 0.001, βCCO = 0.037, p < 0.05, βCOS = 0.030, p < 0.05). In study 2, brand equity affects online trust and vice versa in all models. The total effect of brand equity gains importance for online repurchase (βBrand = 0.134, p < 0.01; βTrust = 0.062, p > 0.1) and for offline repurchase intentions (βBrand = 0.170, p < 0.001 vs. βTrust = 0.121, p < 0.001). We again support all hypotheses and underline the important role of brand equity in omni-channel retailing.

170

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.13 Study 1: results of the alternative model (brand equity) Model 1: Online RPI

Model 2: Offline RPI

β

β

p

p

Direct effects Aesthetic appeal (1)

→ Online trust (2)

.177

Ease of use (1)

→ Online trust (2)

Security/privacy (1)

→ Online trust (2)

Customer service (1) Online-offline integration (1)

**

.173

**

−.106

ns

−.110

ns

.302

***

.304

***

→ Online trust (2)

.161

*

.161

*

→ Online trust (2)

.067

ns

.068

ns

Channel → Online trust (2) consistency (1)

.099

†(.054)

.103

* ***

Aesthetic appeal (1)

→ Brand equity (2)

.437

***

.437

Ease of use (1)

→ Brand equity (2)

−.139

ns

−.141

ns

Security/privacy (1)

→ Brand equity (2)

.155

*

.157

**

Customer service (1)

→ Brand equity (2)

−.006

ns

.008

ns

Online-offline integration (1)

→ Brand equity (2)

.115

*

.115

*

Channel → Brand equity (2) consistency (1)

.201

***

.205

***

−.034

ns

−.045

ns

Aesthetic appeal (1)

→ RPI (2)

Ease of use (1)

→ RPI (2)

.034

ns

.021

ns

Security/privacy (1)

→ RPI (2)

−.037

ns

−.089

ns

Customer service (1)

→ RPI (2)

−.071

ns

−.099

ns

Online-offline integration (1)

→ RPI (2)

.065

ns

.084

ns

Channel → RPI (2) consistency (1)

.066

ns

.234

*** (continued)

4.3 Empirical Studies

171

Table 4.13 (continued) Model 1: Online RPI

Model 2: Offline RPI

β

β

p

p

→ RPI (2)

.416

***

.473

***

Online trust (2) → RPI (2)

.487

***

.361

***

Online brand equity (2) Indirect effects Aesthetic appeal (1)

→ Online trust (2)

→ RPI (3)

.086

**

.062

**

Ease of use (1)

→ Online trust (2)

→ RPI (3)

−.052

ns

−.040

ns

Security/privacy (1)

→ Online trust (2)

→ RPI (3)

.147

***

.110

***

Customer service (1)

→ Online trust (2)

→ RPI (3)

.078

*

.060

*

Online-offline integration (1)

→ Online trust (2)

→ RPI (3)

.032

ns

.025

ns

Channel → Online trust consistency (1) (2)

→ RPI (3)

.048

†(.053)

.037

*

Aesthetic appeal (1)

→ Brand equity (2)

→ RPI (3)

.182

***

.207

***

Ease of use (1)

→ Brand equity (2)

→ RPI (3)

−.058

ns

−.067

ns

Security/privacy (1)

→ Brand equity (2)

→ RPI (3)

.065

*

.074

**

Customer service (1)

→ Brand equity (2)

→ RPI (3)

−.002

ns

−.004

ns

Online-offline integration (1)

→ Brand equity (2)

→ RPI (3)

.048

*

.054

*

Channel → Brand equity consistency (1) (2)

→ RPI (3)

.084

**

.097

***

***

Total effects Aesthetic appeal (1)

→ RPI (3)

.232

***

.225

Ease of use (1)

→ RPI (3)

−.075

ns

−.086

ns (continued)

172

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.13 (continued) Model 1: Online RPI

Model 2: Offline RPI

β

β

p

p

Security/privacy (1)

→ RPI (3)

.175

**

.095

Customer service (1)

→ RPI (3)

.005

ns

−.058

ns

Online-offline integration (1)

→ RPI (3)

.146

*

.163

**

Channel → RPI (3) consistency (1)

.198

**

.369

***

†(.064)

Covariates Gender (1)

→ RPI (3)

.050

ns

.062

ns

Age (1)

→ RPI (3)

−.017

ns

−.055

ns

Internet expertise (1)

→ RPI (3)

.037

ns

−.037

ns

Assortment variety (1)

→ RPI (3)

−.078

ns

−.034

ns

Price fairness (1)

→ RPI (3)

.024

ns

.007

ns

Structural model fit: Model 1: CFI .876, TLI .856, RMSEA .086, SRMR .120, χ2 (596) = 2244.622, SCF = 1.01 Model 2: CFI .872, TLI .851, RMSEA .088, SRMR .121, χ2 (596) = 2324.852, SCF = 1.01 Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377, β = standardized coefficients ***p < .001, **p < .01, *p < .05, † p < .10, ns = not significant Source: Own creation

4.4

Discussion and Implications

4.4.1

Overview

This study contributes to the understanding of the relative importance of online- and omni-channel-specific instruments for consumers’ online and offline

4.4 Discussion and Implications

173

Table 4.14 Study 2: results of the alternative model (brand equity) Model 1: Online RPI Model 2: Offline RPI β

β

P

p

Direct effects Online trust (1)

→ Brand equity (2)

.065

**

.069

**

Brand equity (1)

→ Online trust (2)

.092

**

.090

***

Online trust (1)

→ RPI (2)

.042

ns

.088

***

Brand equity (1)

→ RPI (2)

.094

**

.121

***

Online trust (1)

→ Online trust (2)

.613

***

.608

***

Brand equity (1)

→ Brand equity (2)

.676

***

.680

*** ***

RPI (1)

→ RPI (2)

.678

***

.652

Online trust (2)

→ Brand equity (3)

.069

**

.073

**

Brand equity (2)

→ Online trust (3)

.096

**

.095

***

Online trust (2)

→ RPI (3)

.046

ns

.095

***

Brand equity (2)

→ RPI (3)

.101

**

.127

***

Online trust (2)

→ Online trust (3)

.643

***

.658

***

Brand equity (2)

→ Brand equity (3)

.713

***

.704

***

RPI (2)

→ RPI (3)

.654

***

.626

***

.526

***

.563

***

R2 RPI (3) Total effects Online trust (1)

→ RPI (3)

.062

ns

.121

***

Brand equity (1)

→ RPI (3)

.134

**

.170

***

Diff. in total effects

4.025**

4.202**

Covariates Gender (1)

→ RPI (1)

.056

**

.087

***

Gender (2)

→ RPI (2)

.062

**

.097

***

Gender (3)

→ RPI (3)

.065

**

.094

***

Age (1)

→ RPI (1)

−.057

**

−.063

**

Age (2)

→ RPI (2)

−.063

**

−.068

**

Age (3)

→ RPI (3)

−.067

**

−.071

**

Internet expertise (1)

→ RPI (1)

−.009

ns

−.018

ns

Internet expertise (2)

→ RPI (2)

−.009

ns

−.018

ns (continued)

174

4

Study 3: Importance of Marketing Instruments for Repurchase …

Table 4.14 (continued) Model 1: Online RPI Model 2: Offline RPI β Internet expertise (3)

→ RPI (3)

−.010

P ns

β −.019

p ns

Structural model fits: Model 1: CFI .837, TLI .830, RMSEA .091, SRMR .183, χ2 (1188) = 4863.598, SCF = .89 Model 2: CFI .828, TLI .821; RMSEA .094, SRMR .184, χ2 (1188) = 5114.794, SCF = .89 Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377. Standardized coefficients are shown ***p < .001, **p < .01, *p < .05, † p < .10, ns = not significant Source: Own creation

repurchase intentions, mediated by online trust and brand equity. Furthermore, the relative importance of the reciprocal effects of online trust and brand equity are discussed. Next, theoretical and managerial implications are presented.

4.4.2

Theoretical Implications

We Regarding our first research question, study 1 provides novel insights into the different roles of important marketing instruments for cross-channel repurchase intentions by referring to the calls in the literature (e.g., Bleier, Harmeling and Palmatier 2019; Blut, Teller and Floh 2018). It identifies four online- and omni-channel-specific instruments as the most influential. Retailers indirectly participate using these instruments, engaged by online trust and brand equity. Online trust was studied frequently, while the role of online brand equity is notable, as omni-channel retailers benefit from a strong brand (e.g., competing online pure players, Gielens and Steenkamp 2019; Loupiac and Goudey 2019). We moreover believe to contribute to the application of schema theory because we capture timedependent effects by applying sequential mediation modeling over time, i.e., we measure dependent, mediating and independent variables at different time points. Thus, we show how consumers’ perceptions of instruments affect associations and repurchase intentions (e.g., Sullivan and Kim 2018). We argue that experienced consumers link instruments with online trust and brand equity nodes in memory to evaluate omni-channel retailers in repurchase decisions (e.g., Grewal, Levy and

4.4 Discussion and Implications

175

Kumar 2009). Three theoretical implications emerge regarding the relative importance of instruments, the role of online trust vs. brand equity and differences in cross-channel effects. First, Table 4.15 visualizes the four most effective instruments for repurchasing intentions while we accentuate their total, indirect and cross-channel effects. Table 4.15 Results overview Online RPI

Offline RPI

Indirect effect Trust

Total effect

Brand equity

Indirect effect Trust

Brand equity

Total effect

Online-specific instruments Aesthetic appeal











Ease of use

ns

ns

ns

ns

ns

 ns

Security/privacy













Customer service



ns

ns



ns

ns

Omni-channel-specific instruments Online-offline integration

ns





ns





Channel consistency













Reciprocal effect of mediators Online trust





Online brand equity





Source: Own creation

• Aesthetic appeal is the major online lever. Despite some insignificant findings in the literature, we found that attractive websites affect repurchase intentions (online: β = 0.232, p < 0.001, offline: β = 0.224, p < 0.001, see the total effects in Table 4.7). It activates online trust and stronger brand equity associations in consumers’ minds (Hollebeek and Macky 2019; Rajavi, Kushwaha and Steenkamp 2019). Online and (more weakly) offline repurchase intentions are affected, which notably indicates valuable cross-channel effects in omni-channel retailing (Baek et al. 2020; Loupiac and Goudey 2019).

176

4

Study 3: Importance of Marketing Instruments for Repurchase …

• Security/privacy is an important, however less effective, online lever (total effects on repurchase intentions: online: β = 0.175, p < 0.01, offline: β = 0.095, p < 0.1). In line with extant research, it activates the trust node in consumers’ minds (e.g., Bashir et al. 2018; Sullivan and Kim 2018) as well as (more weakly) brand associations. While research attributes security/privacy exclusively to online outcomes (Bressolles, Durrieu and Deans 2015; Toufaily and Pons 2017), it also determines (more weakly) offline repurchase intentions, particularly through cross-channel effects. • Channel consistency is the most important omni-channel-specific lever (total effects on repurchase intentions: online: β = 0.197, p < 0.01, offline: β = 0.367, p < 0.001). Harmonization of online and offline channels activates online trust and brand equity nodes in consumers’ minds to the same extent (Bertrandie and Zielke 2017; Van Baal 2014). Moreover, channel consistency affects online and (more strongly) offline repurchase intentions (it also affects the latter directly; it is the only direct effect of the instruments in this study, β = 0.256, p < 0.001). • Online-offline integration is an important but weaker omni-channel-specific lever (total effects on repurchase intentions: online: β = 0.146, p < 0.05, offline: β = 0.163, p < 0.01). Notably, online-offline integration is linked only to the online brand equity node in consumers’ minds (not to online trust), but it equally determines online and offline repurchase intentions through this node. In line with extant research, omni-channel retailers benefit from offering options to complete consumers’ cross-channel purchases (e.g., Yang et al. 2020). The total effects of ease of use is insignificant in all models, while customer service affects trust and, through trust, affects cross-channel repurchase intentions quite weakly. In summary, the results show that perceived marketing instruments activate online trust and brand equity nodes differently, and thus, they are not equally retrieved in repurchase decisions (Badrinarayanan et al. 2012; Puligadda, Ross and Grewal 2012). Assessing such mechanisms is a basis for the effective omnichannel management of important marketing instruments (e.g., Bleier, Harmeling and Palmatier 2019). Second, study 1 highlights the importance of online trust and brand equity as mediators of repurchase intentions. The theory and methods in this study relate to respective calls for longitudinal studies on online trust (e.g., Kim and Peterson 2017). Online trust is known to be relevant for online repurchasing but also determines offline repurchasing (e.g., Xiao, Zhang and Fu 2019). Brand equity seems

4.4 Discussion and Implications

177

to stronger affect offline but also online repurchase intentions. However, both are important nodes in the memories of repurchasing consumers. This implication underlines the importance of examining which has stronger reciprocal effects on repurchase intentions in study 2. Third, the results are stable for online and offline repurchase intentions (for a general retailer in alternative models). The provided theoretical rationale is supported. Omni-channel retailers can attract consumers online and steer them to physical stores (e.g., Herhausen et al. 2019). The main total levers for crosschannel repurchasing are aesthetic appeal and channel consistency. However, if retailers aim to stimulate online channel repurchase intentions, aesthetic appeal is a superior lever to channel consistency, and security/privacy is superior to online-offline integration. For offline repurchase intentions, channel consistency is superior (even having a direct effect) to aesthetic appeal, closely followed by online-offline integration and then by security/privacy. Regarding the second research question, study 2 shows that there exist reciprocal relationships between online trust and brand equity and that both reciprocally affect online and offline repurchase intentions in omni-channel retailing. Two theoretical implications are highlighted. We enhance the analysis of links between brand equity and trust and from trust to brand equity by providing evidence of these reciprocal relationships. Thus, the activation of associations in consumers’ minds spreads from one node to another over time, i.e., from online trust to brand equity and vice versa (supporting assumptions and calls in previous literature, e.g., Hollebeek and Macky 2019; Rajavi, Kushwaha and Steenkamp 2019). The direct links are slightly stronger for trust to brand equity than from brand equity to trust at both time points and in both models (see Table 4.8). However, the question of whether online trust or brand equity more strongly facilitates the effect of online- and omni-channel-specific instruments on repurchase intentions can be answered by considering the total reciprocal effects. The activation of both nodes leads to stronger brand equity total effects, significantly for online repurchase intentions and not significantly for offline repurchase intentions (Hollebeek and Macky 2019). Online trust may lose relative importance in the future as consumers gain confidence in online shopping. Omnichannel retailers aiming to increase cross-channel repurchasing should strengthen online brand equity without disregarding online trust (e.g., competing online pure players, Khan et al. 2019). Instruments that affect brand equity in online repurchase situations are particularly important: aesthetic appeal, channel consistency, security/privacy, and online-offline integration. This insight is beneficial

178

4

Study 3: Importance of Marketing Instruments for Repurchase …

for managers aiming to effectively orchestrate instruments while considering all possible relationships.

4.4.3

Managerial Implications

Omni-channel retailers are competing with purely online players for consumers online and can benefit from the effective orchestration of perceived online- and omni-channel-specific instruments (e.g., Bleier, Harmeling and Palmatier 2019). Aesthetic appeal and channel consistency as well as online-offline integration and security/privacy increase consumers’ repurchase intentions the most. In particular, omni-channel-specific instruments are a source of competitive advantage for omni-channel retailers, given the competencies required to implement them (i.e., high resources, Acquila-Natale and Chaparro-Peláez 2020). This implementation might be a strategic barrier for online players when opening physical stores and applying omni-channel-specific instruments. However, for omni-channel retailers, using these instruments could also be a challenge in that they may gain or lose customers as a result. For example, consumers visiting physical stores may be disappointed if their appeal or the consistency of offers differ from those of online channels (e.g., Loupiac and Goudey 2019). Managers need to carefully orchestrate the relatively most important instruments. Effective orchestration also depends on the mechanisms that translate the instruments into consumer behavior (e.g., Bleier, Harmeling and Palmatier 2019). This study underlines the need for managers to ensure online trust, as it is a known mechanism that is activated by instruments and is important for behavior (e.g., Kim and Peterson 2017). However, online brand equity (retailers’ brand equity in alternative models) reciprocally translates the instruments into repurchase behavior more strongly than trust. Because online trust may lose importance due to consumers’ increasing online confidence, it is wise for omni-channel retailers to build strong online brand equity as a source of competitive advantage (Gielens and Steenkamp 2019; Loupiac and Goudey 2019). We recommend a continuous evaluation of the websites’ relative association strength, attractiveness or uniqueness. Mimicking the behaviors of successful, strongly brand-oriented manufacturers seems useful. Finally, the relative importance of brand equity and the most important instruments differs for cross-channel repurchase behavior. However, omni-channel retailers (vs. purely online players) can leverage benefits for their online and offline stores. Pointing to cross-channel effects requires a match of online and offline behavior, i.e., overcoming consumer information in silos (Herhausen et al. 2015) or disconnected organizational structures and processes.

4.5 Limitations and Future Research

4.5

179

Limitations and Future Research

This study has certain limitations that suggest future research directions. Although we carefully collected specific data, database expansion will allow further conclusions, e.g., analyses of other retailers, industries (e.g., electronics with new augmented reality instruments, tourism with more transaction-oriented websites, Khan et al. 2019), consumer segments (e.g., Herhausen et al. 2019) or purely online players (e.g., with fewer physical store links, Bleier, Harmeling and Palmatier 2019). Broadening our channel scope, e.g., to non-firm-controlled touchpoints, will reveal their role in the cognitive processes and effects studied (e.g., De Keyser et al. 2020). We could not observe offline channel instruments or reciprocity in one study—respective models affect results—or objective purchase data (due to model complexity or methodological limits, Yrjölä, Spence and Saarijärvi 2018). Regarding the measures, additional instruments are interesting, even though no common agreement on them or validity challenges exist (e.g., omni-channel customization, social chat bots, Herhausen et al. 2020). Instruments or trust can be measured in greater detail, e.g., differentiating cognitions vs. affect (Toufaily and Pons 2017; Ye et al. 2020). Studying online, offline or overall brand equity is interesting, as associations of a brand as strong, attractive or unique may differ in comparative competitive situations (Batra and Keller 2016). Retail brands provide functional vs. psychological values; e.g., quality/value for money vs. emotions/enhanced self-concepts may translate instruments and retail brands’ effects differently into behavior. The last argument extends our conceptual framework. Furthermore, the cognitive load of each instrument can be manipulated (Wen and Lurie 2019) and activation in minds can be automatic or controlled (e.g., depending on the decision situations, subjects or objects). Although experienced consumers decide more frequently based on memory than on stimuli, we rationally do not differentiate external and internal perceptions of instruments. Emotional or motivational theories may be insightful (e.g., Landers et al. 2015). Due to the model complexity, we do not use moderators, while consumers’ online shopping experience increases the associations in their minds and may increase the effects (e.g., Emrich and Verhoef 2015). Further touchpoints and thus consumers’ social feedback affect their associations and cognitive mechanisms (e.g., Hogreve, Bilstein and Hoerner 2019).

5

Final Remarks

5.1

Discussion and Conclusions

5.1.1

Core Results

Instead of viewing physical and online channels as independent of each other, retailers have realized that they can remain competitive and retain loyal customers by supporting image transfer between channels. They increasingly recognize that offering channel integration services and successfully orchestrating onlineand omni-channel-specific instruments further supports such interdependencies to create a seamless experience. However, retailers still face the challenge of how to more effectively engage consumers from different channels and adapt their practices to meet these challenges (Acquila-Natale and Chaparro-Peláez 2020; Hult et al. 2019). The focus of this doctoral thesis is on these specific challenges, analyzed from a consumer perspective. A review of existing literature in Section 1.2 points to several research gaps in the context of channel interdependencies and the decisions of retailers to support seamless channel transfers to retain consumers. First, regarding channel interdependencies, studies have underscored that image transfers from offline to online stores and vice versa occur in today’s retail environment (e.g., Kwon and Lennon 2009b; Badrinarayanan and Becerra 2019). However, research has predominantly focused on unidirectional (e.g., Moliner-Velázquez et al. 2019) or bidirectional effects (e.g., Badrinarayanan et al. 2012) on different outcomes. These studies neglect important reciprocal relationships, which, unobserved, lead to different

© The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 A. Winters, Omni-Channel Retailing, Handel und Internationales Marketing Retailing and International Marketing, https://doi.org/10.1007/978-3-658-34707-9_5

181

182

5

Final Remarks

conclusions and implications. Moreover, although, it is known that consumer evaluations are affected by prior experiences (e.g., Lemon and Verhoef 2016), the inclusion in effects of channel image transfers is missing. Second, research has focused on joint effects of channel integration or addressed OF-ON or ON-OF services only, but has not disentangled both (e.g., Herhausen et al. 2015; Hossain et al. 2020). Integration services were found to positively affect consumer behavior such as loyalty, repurchase intentions or word-of-mouth (e.g., Lee et al. 2019; Murfield et al. 2017). Indirect effects were mostly shown. However, studies often considered general channel evaluations that translated integration services into consumer behavior and thus did not account for important cross-channel effects (e.g., Ravula, Bhatnagar and Ghose 2020). Moreover, consumers’ online shopping experience is known to reduce effects of ON-OF services (Herhausen et al. 2015), but whether this changes the results of the mediation paths of simultaneously perceived OF-ON and ON-OF services is questionable. Finally, channel congruence is known to lead to a more holistic channel view of consumers (e.g., Hammerschmidt, Falk and Weijters 2016), but its moderating role for the mediation paths of integration services is unexplored. Third, previous studies have acknowledged the relative importance of various online- and omni-channel-specific instruments, but have mostly looked at one or two (e.g., Bashir et al. 2018; Rahimnia and Hassanzadeh 2013), which does not cover an effective orchestration of instruments. Fewer studies have analyzed more instruments (e.g., Bressolles, Durrieu and Deans 2015; Floh and Madlberger 2013) and have been able to predict their relative importance, but not that of online- and omni-channel-specific instruments simultaneously. Moreover, indirect effects through various mediators and most often through online trust have been conceptualized (e.g. Toufaily and Pons 2017). Few have studied brands (Al-Hawari 2011; Bleier, Harmeling and Palmatier 2019) as a mechanism that transforms instruments into behavior, although this is of particular importance for omni-channel retailers. Moreover, links of brand associations to trust and vice versa are assumed (e.g., Fazal-e-Hasan et al. 2018; Khan et al. 2019). Reciprocal relationships between online trust and brand equity have not yet been examined. Finally, cross-channel effects of instruments are seldom considered, although it is known that associations with a channel are used to evaluate further omni-channels (e.g., Loupiac and Goudey 2019). In summary, former research has addressed a range of retailers’ omni-channel decisions. However, specific questions have remained unanswered, leading to the existence of a number of research gaps. The present doctoral thesis has dealt with these questions in order to fill these gaps. In particular, effects of such decisions on channel-level and retailer-level outcomes have been proposed based on consumer

5.1 Discussion and Conclusions

183

evaluations of omni-channel fashion retailers. Study 1 is based on longitudinal and cross-sectional data. Study 2 refers to a cross-sectional design, whereas the hypotheses of Study 3 were tested based on longitudinal data. Thus, this doctoral thesis offers novel empirical and theoretical contributions to the current state of research on interdependencies, channel integration services and specific marketing instruments over time. In this context, the following central research questions were answered in this doctoral thesis: (1) Do reciprocal relationships between major retail sales channels exist and affect overall, offline and online channel loyalty, and how do these relationships change depending on prior offline and online experiences? (2) How do perceived OF-ON and ON-OF channel integration services influence offline and online purchase intentions through perceived quality of offline and online offerings and how are these relationships moderated by consumer’s online shopping experience and channel congruence? (3) How do important online- and omni-channel-specific instruments affect online and offline repurchase intentions, mediated by online trust and brand equity, and whether do reciprocal relationships between online trust and brand equity exist and affect online and offline repurchase intentions? The core results of the studies conducted to answer these three central research questions are described below. The first central research question was answered in Study 1 by analyzing reciprocal relationships between major offline and online purchase channels of omni-channel retailers, following respective research calls (e.g., Loupiac and Goudey 2019; Wiener, Hoßbach and Saunders 2018). The results of the cross-lagged panel models show that interdependencies between retailers’ major purchase channel images exist, which is especially important for omni-channel retailers as many consumers start buying in one channel and then switch channels (Li et al. 2018; Zhang et al. 2018). Offline and online channel images have a positive reciprocal relationship, although the offline channel image more strongly affects the online channel image than vice versa. Browsing on a retailer’s website, for example, creates positive attitudes towards physical stores of the same retail brand, while consumers in online stores update their current knowledge through perceptions of the retail brands’ physical store. The reciprocal total effects show that offline and online channel image affect offline channel, online channel and overall retailer loyalty. Surprisingly, offline channel images affect online channel loyalty

184

5

Final Remarks

the most. Considering both retailer-level and channel-level outcomes contributes to the fact that today, consumers not only choose different retailers, but also retailers’ different channels (e.g., Herhausen et al. 2015). The findings extend the results of Swoboda, Weindel and Schramm-Klein (2016) by monitoring different important outcomes in omni-channel systems. The results moreover contribute to the dominant unidirectional and bidirectional studies and provide important theoretical implications. In addition, Study 1 shows that prior experiences affect the reciprocal effects of major offline and online purchase channel images on all loyalties (e.g., Grewal and Roggeveen 2020). More favorable prior experiences strengthen the reciprocal effects, and retailers can benefit from their offline channel image effects. Retailers with less favorable prior experiences cannot use reciprocity to retain consumers in omni-channel retailing. The results of the cross-sectional models in Study 1 indicate possible consequences of the exclusion of reciprocity in cross-sectional designs. The results show no reciprocal effects between the channel images. The total effects of online and offline channel image are stronger than in the cross-lagged panel models, and offline channel image affects loyalty even more strongly. Different conceptualizations of mediations and the exclusion of interdependencies in the tested models lead to different conclusions regarding the image links between offline and online channels. These findings underline the importance of considering reciprocity, as otherwise unobserved effects occur. The second central research question was addressed in Study 2. This study adds to the joint perspectives of channel integration services by disentangling ramifications of integration services and by considering important cross-channel effects. The results show that retailers indirectly participate in offered OF-ON and ON-OF services for purchase intentions differently, answering respective research calls (e.g., Shen et al. 2018; Zhang et al. 2018). OF-ON services affect the perceived quality of both offline and online offerings, which in turn enhance offline and online purchase intentions. Thus, these services exert important cross-channel effects, which occur for example, when employees in physical stores refer to online information, making this channel more salient and consumers in turn more confident in offline channels (e.g., Bhargave, Mantonakis and White 2016). For OF-ON services, all indirect and total effects are significant. Contrary, ON-OF services enhance offline and online purchase intentions via the quality of online offerings only. The importance of distinguishing between these services becomes evident. Comparing the effects of the services, surprisingly, OF-ON services have the strongest impact and are identified as the main lever for omni-channel retailers. The results also hold in the alternative models with observed purchase

5.1 Discussion and Conclusions

185

intentions towards the retailer in general. For retailers offering a seamless experience across channels, OF-ON services are the dominant driver for consumers’ purchase intentions at both the channel-level and retailer-level. Additionally, the results of Study 2 show that consumers’ online shopping experience is an important context factor that negatively moderates the total effects of perceived OF-ON services on offline and online purchase intentions, while the mediation paths of ON-OF services are not moderated. Within the interactions, only the cross-channel effects of OF-ON services are moderated, which also reduces the total effects. This observation provides novel implications for integration services and adds to studies on OF-ON or ON-OF services only (e.g., Jin, Li and Cheng 2018; Mosquera et al. 2018). Moreover, channel congruence has a moderating effect on the mediation paths of OF-ON and ON-OF services to offline purchase intention. Higher levels of channel congruence negatively affect the indirect effects through the channel-specific quality of offline offerings. The indirect effects are positively moderated through the quality of online offerings. An objective congruence measure supports the results for OF-ON services in offline decisions. Firms with higher congruence can reinforce cross-channel effects due to the positive effects via perceived online quality perceptions and reduce the effects through established offline quality perceptions. The third central research question is dealt with in Study 3. The results of Study 3 demonstrate that omni-channel retailers benefit in terms of enhanced consumer repurchase intentions predominantly by applying four online- and omnichannel-specific instruments in online stores (contributing to respective calls from literature, e.g., Bleier, Harmeling and Palmatier 2019; Blut, Teller and Floh 2018). These instruments are aesthetic appeal, security/privacy, online-offline integration, and channel consistency. Retailers participate indirectly in using these instruments as they affect perceived online trust and brand equity, albeit to varying degrees. Online trust was studied frequently as a mechanism in e-commerce research (e.g, Kim and Peterson 2017). The role of online brand equity is not negligible, as omni-channel retailers benefit from a strong brand in competition with online pure players (Gielens and Steenkamp 2019; Loupiac and Goudey 2019). Through these mechanisms, the instruments exert valuable cross-channel effects on offline consumer behavior (e.g., Baek et al. 2020; Loupiac and Goudey 2019). Channel consistency directly affects offline repurchase intentions, while customer service affects online and offline repurchase intentions mediated through online trust only. Moreover, Study 3 provides evidence for a reciprocal relationship between the mechanisms that translate the instruments into behavior. Perceptions of online trust and brands are interconnected in consumers’ minds, supporting assumptions in previous literature (e.g., Hollebeek and Macky 2019; Rajavi, Kushwaha

186

5

Final Remarks

and Steenkamp 2019). Online trust affects online brand equity slightly stronger than vice versa at both time points. Both reciprocally affect online and offline repurchase intentions over time. Online trust is known to be relevant for online repurchasing but also has cross-channel effects on offline repurchase intentions (e.g., Xiao, Zhang and Fu 2019). The results regarding the total reciprocal effects point to stronger online brand equity (vs. online trust) effects. Online brand equity (vs. online trust) enhances online repurchase intentions, and by tendency offline repurchase intentions significantly stronger over time. In summary, this doctoral thesis provides important and valuable insights into how omni-channel retailers can benefit from channel interdependencies, integration services, and online- and omni-channel-specific marketing instruments in order to retain consumers at the channel-level and retailer-level. Moreover, insights into important context factors in omni-channel retailing are provided.

5.1.2

Theoretical Implications

The three studies presented in this doctoral thesis provide valuable contributions to theory and existing research. In the following, the main theoretical implications of all three studies are summarized. Study 1 contributes to the understanding of channel interdependencies between omni-channel retailers’ major offline and online purchase channels. Whereas past studies have referred to cognitive dissonance or reasoned action theory to explain channel interdependencies (e.g., Badrinarayanan, Becerra and Madhavaram 2014; Kwon and Lennon 2009b), this study contributes to the application of categorization theory in an omni-channel context. This theory is able to explain that cognitive connections in consumers’ minds exist between perceptions of major offline and online purchase channels. Theoretically, positive image transfers are shown for attitudinal channel images and for associative brand equity constructs over time. This finding is theoretically remarkable and extends initial attempts to understand reciprocity (e.g., Kwon and Lennon 2009a; Swoboda, Weindel and Schramm-Klein 2016). Consumers, for example, perceive online stores more positively by updating their online channel image based on positive offline perceptions of a retailer. The results of the reciprocal effects support the theoretical assumption that both offline and online channel images help omni-channel retailers to retain loyal consumers. Classifying a channel into a retail brand category allows maximal inferences to be drawn about this and related channels for consumer evaluation due to the high discrimination ability with the category (Liu et al. 2017). Theoretically, consumers seem to map accessible relational structures

5.1 Discussion and Conclusions

187

and establish links between both images. Moreover, consumers rely on the most knowledgeable or representative category member, namely the offline channel in the case of former brick-and-mortar retailers, to draw the strongest inferences for decision-making (e.g., Loken 2006; Sohn 2017). In today’s retail environment, physical stores are still the predominant channels for consumers to draw inferences in decisions. This finding theoretically supports the still crucial role that offline retailing plays, as outlined in Section 1.1. However, cross-channel effects are important, because offline channel loyalty is affected by the online channel image of omni-channel retailers (e.g., Loupiac and Goudey 2019). These results are to be expected when analyzing former brick-and-mortar retailers (see also Swoboda, Weindel and Schramm-Klein 2016). In contrast, the results might change for former online pure players that open physical stores. Here, the online channel is likely to be the reference channel for evaluation, and stronger online channel image effects could be expected. However, for omni-channel retailers, understanding the interplay between their major sales channels from a consumer perspective is of paramount importance. Omni-channel retailers redirect consumers between these channels more than multi-channel retailers do, forcing even greater categorization effects (e.g., Wiener, Hoßbach and Saunders 2018). Unidirectional studies have indicated such categorization effects (e.g., Chu et al. 2017; Grosso, Castaldo and Grewal 2018), but these studies have often analyzed offline and online channel images in isolation, without considering cross-channel or important reciprocal effects (Mitchell and Maxwell 2013). Identical relationships tested with cross-sectional data show that such designs do not have inference capability. The online channel image does not reciprocally affect its offline counterpart. Cross-sectional designs, often used in unidirectional studies, theoretically imply a more piecemeal information processing by consumers, which disregards certain causal relationships. Therefore, a theory-based conceptualization of reciprocity is valuable, as otherwise, unobserved effects occur (Chu et al. 2017). The need for reciprocal approaches becomes obvious and crosslagged panel models with inference capability should be applied more often in future research (e.g., Wiedermann and von Eye 2015; Zyphur et al. 2019). Moreover, Study 1 shows that for retailers with which consumers have more (vs. less) favorable prior experiences, the reciprocal effects on all loyalties change. Consumers consider experience and knowledge to achieve efficient processing to support categorization and reciprocal links between images of channels in decisions (Lemon and Verhoef 2016; Yang et al. 2020). Theoretically, the use of favorable knowledge is facilitated, which leads to stronger responses. Prior experiences change current and future evaluations and should be conceptualized, as

188

5

Final Remarks

otherwise, certain consumer inferences used to evaluate retailers are neglected (e.g., Grewal and Roggeveen 2020; Loupiac and Goudey 2019). Study 2 advances the understanding of important cross-channel effects of the mediation paths through which retailers can transform OF-ON and ON-OF services into purchase intentions. This study is based on accessibility-diagnosticity theory to explain the mediation paths of the channel integration services across major sales channels (e.g., Feldman and Lynch 1988) as a useful alternative to further theories used in previous literature (e.g., technology adaptation theory, Herhausen et al. 2015). This theory suggests that the likelihood of using a cue or input for decision-making depends on the specific input’s accessibility and relative diagnosticity (Lynch, Marmorstein and Weigold 1988) and is particularly suitable as it emphasizes which of the diagnostic cues consumers are faced with in omni-channel retail decisions is most relevant. Theoretically, the results allow the conclusion that in decisions, all available cues are utilized, although their relevance for evaluation differs. OF-ON services provide consumers with knowledge about and ease of access to the perceived quality of both offline and online channel offerings. ON-OF services fail to make both qualities more salient and transparent, as only the quality of online offerings is enhanced. It is therefore important to distinguish theoretically and empirically between the mediation paths of integration services for omni-channel retailers, otherwise unobserved paths will emerge. The theoretical consideration of the perceived quality of channel offerings helps retailers to prevent a disruption of integration service outcomes (Banerjee 2014). Surprisingly, OF-ON services are a superior cue than ON-OF services as they increase both offline and online purchase intentions the most. Theoretically, this study draws attention to the accessibility and diagnosticity of perceived integration services. Although most important services for consumers are easily accessible to consumers, they are not equally relevant in decision-making (Lynch, Marmorstein and Weigold 1988; Menon and Raghubir 2003). Moreover, the results contribute to research as they show a changing relevance of integration services based on further diagnostic cues. For consumers with lower levels of online shopping experience, OF-ON services are relatively diagnostic for both purchase decisions. Increasing levels of experience diminishes the relative diagnosticity. Consumers’ online shopping experience itself affects the purchase intentions and becomes a higher order cue that is relatively diagnostic for decision-making. It can therefore theoretically be seen as a superior cue that consumers use for evaluations. Nevertheless, even if such a higher-order cue is available, the lower-order cue still operates. OF-ON services are attenuated with the availability of the higher-order cue (Feldman and Lynch 1988). Moreover, channel congruence serves as a further diagnostic superior cue by affecting the

5.1 Discussion and Conclusions

189

relative diagnosticity mechanism. For perceived OF-ON services, higher levels of channel congruence lead to a reduced relative diagnosticity via the perceived quality of offline offerings and an increased relevance of the perceived quality of online offerings. This observation allows the theoretical conclusion that consumers of omni-channel retailers generalize information from one channel to another (Van Baal 2014). Higher levels of perceived channel congruence promote a more holistic evaluation of channels and their relevance (Bezes 2013; Hammerschmidt, Falk and Weijters 2016). Study 3 contributes to the understanding of the relative importance of onlineand omni-channel-specific marketing instruments for online and offline repurchase intentions, mediated by online trust and brand equity. In contrast to studies in this context that refer to technology acceptance theory or apply a reasoned action and planned behavior theoretical rational (e.g., Loureiro, Cavallero and Miranda 2018; Hsieh et al. 2014), this study refers to schema and associative network theory. These theories provide organizing mechanisms for the knowledge of experienced consumers. This knowledge refers to situations and objects, typically in networks of dependent nodes or associations, and the links between them (e.g., Fiske and Taylor 1991, p. 98; Krishnan 1996). Because time-dependent effects are captured by applying sequential mediation modeling over time and by using cross-lagged panel models, this study is able to apply such theoretical rationales. In doing so, it answers research calls, that claim limitations of cross-sectional studies as they bias the estimation of mediation parameters (Kim and Peterson 2017; Mitchell and Maxwell 2013). Theoretically, the results show that respective nodes in consumers’ memories are linked to perceived marketing instruments and are activated by external stimuli or information retrieved from memory (e.g., Mavlanova, Benbunan-Fich and Lang 2016; Puligadda, Ross and Grewal 2012). Marketing instruments as stimuli or signals perceived by consumers in online stores are information cues that form positive attitudes or cognitions. They provide guidance for processing information and facilitate decisions (Mallapragada, Chandukala and Liu 2016; Rose et al. 2012). The extent to which consumers access and weight the online- and omni-channel-specific instruments affects the degree of consumers’ behavioral relevance (Badrinarayanan et al. 2012; Connelly et al. 2011). This study provides evidence that perceived instruments activate online trust and brand equity nodes stored in associative networks. However, differences in the activation occur, which supports the assumption that effective orchestration depends on important mechanisms. Online trust and brand equity nodes transform instruments into online and cross-channel offline repurchase intentions, which underlines that experiences from website interactions transfer cognitively to additional channels (e.g., Grewal and Roggeveen 2020).

190

5

Final Remarks

Additionally, this study advances extant research by analyzing the links between online trust and brand equity. Evidence for a reciprocal relationship suggests that online trust and brand equity are interconnected (Inman, Shankar and Ferraro 2004; Rajavi, Kushwaha and Steenkamp 2019). Associations in consumers’ minds spread from one node to another node over time and can be theoretically explained by the concept of spreading activation between nodes (Keller 1993; Teichert and Schöntag 2010). Both are behaviorally relevant, and the activation of nodes by marketing instruments leads to stronger total effects of brand equity on online and offline repurchase intentions. Omni-channel retailers aiming to increase crosschannel repurchasing should therefore strengthen online brand equity without disregarding online trust. The need for longitudinal research becomes evident: effects of online trust and brand equity cannot be fully understood without analyzing consumer perceptions over time (Swoboda, Weindel and Schramm-Klein 2016; Ye et al. 2020). In summary, this doctoral thesis extends current knowledge about the crosschannel effects of different decisions that omni-channel retailers face in today’s retailing environment by explicitly addressing their impact on behavioral outcomes at the channel-level and retailer-level and responding to the respective research calls from the literature. For this propose, cognitive theory argumentation is employed in the three studies of this doctoral thesis. The theories applied are suitable for explaining how consumers create categories and cognitive structures, process information when confronted with stimuli and cues, and how they guide their behavior across offline and online channels of omni-channel retailers.

5.1.3

Managerial Implications

In the following, the managerial implications for each of the three studies described above are discussed. The general management implications of this doctoral thesis are outlined subsequently. Study 1 reveals that carefully managed reciprocal channel images help retailers who find it still challenging to retain loyal consumers to do so through a combination of major sales channels (Hult et al. 2019). Careful attention must be paid to the reciprocity between channels, otherwise misinterpretations of market studies, short conclusions, or ineffective investments are likely. The strongest image transfers from offline to online channels emerge. Omni-channel retailers can particularly leverage this benefit in competition with online pure players. The stronger reciprocal effect of offline channel image on all loyalties underlines the

5.1 Discussion and Conclusions

191

still important role that physical stores play in omni-channel systems as highlighted in Section 1.1. To control reciprocal image transfers, it is recommended that instruments such as integration services or further communication touchpoints should be offered to consumers to steer them to important sales channels (Lee et al. 2019). However, not all retailers can benefit from reciprocity, as consumers’ prior experience affects current and future evaluations of stores (Grewal and Roggeveen 2020). Competitive advantages emerge for retailers with which consumers have more favorable prior offline and online experiences, as these retailers can benefit from the image transfers. Retailers with less favorable prior experiences need careful strategic considerations of how to overcome these competitive disadvantages. Strengthening experience in the target channel by using further touchpoints or monitoring short- and long-term effects of prior experiences is recommended (Lemon and Verhoef 2016). Study 2 underlines the important role of offering channel integration services to consumers. This study has stressed the importance of OF-ON services in contrast to ON-OF services, which is interesting for managers given the emergence of technologies to create seamless customer experiences across channels (e.g., Gao et al. 2019). OF-ON services help both to redirect consumers to major purchase channels of an omni-channel retailer, and enhance purchase intentions the most. It is recommended that omni-channel retailers in the fashion sector should offer OF-ON services, to complement their physical channel with their online channel. For example, retailers could provide employees with tablets to assist in-store consumers by alerting them to offers and information available online (e.g., Bhargave, Mantonakis and White 2016). Moreover, today, consumers also rely on their online shopping experience to guide their evaluations and behavior. Retailers benefit from most important OF-ON services only by targeting consumers with a lower level of online shopping experience. Consumers with higher levels of online shopping experience see OF-ON services just as further services added to existing ones (e.g. Falk et al. 2007). In the future, consumers will have higher levels of online shopping experience due to the continuing trend of e-commerce across sectors described in Section 1.1, and such individual difference should be monitored across different consumer segments (e.g., Herhausen et al. 2019). Finally, managers should control their channel congruence, because offering important OF-ON services while having high perceived levels of congruence increases the perceived quality of online offerings, but reduces that of offline offerings. Consumers are redirected to online stores, which subsequently increases offline purchase intentions in a cross-channel manner. Because this is a cost-intensive option (Van Baal 2014; Van Bruggen et al. 2010), the tradeoff must be weighted out.

192

5

Final Remarks

Study 3 emphasizes the need for omni-channel retailers to effectively orchestrate perceived online- and omni-channel-specific marketing instruments in online stores. From a managerial point of view, the most important instruments identified for omni-channel retailers are aesthetic appeal, channel consistency, online-offline integration, and security/privacy. These instruments promote the repurchase intentions of consumers the most and should be offered in online stores. In particular, omni-channel-specific instruments are an important source of competitive advantage for omni-channel retailers over online pure players due to the required implementation capabilities (e.g., Luo, Fan and Zhang 2015). In addition, confirmed cross-channel effects of instruments on offline repurchase intentions indicate that an online store can be the starting point leading to a consumer’s experience with the retailer’s physical store (e.g., Loupiac and Goudey 2019). For managers, an effective orchestration of instruments depends on trust, which is especially critical in the online context (e.g., Kim and Peterson 2017). However, for omni-channel retailers, it is demonstrated that online brand equity is a further critical mechanism that needs to be addressed to support the effects of online- and omni-channel-specific instruments on repurchase intentions. In addition, online trust is shown to depend on brand-related associations and vice versa (e.g., Hollebeek and Macky 2019; Khan et al. 2019). Over time, consumers rely more strongly on their brand associations to determine their repurchase behavior. This finding is particularly important as consumer confidence in online stores is likely to increase in the future, and omni-channel retailers have competitive advantages to strengthen their retail brand positioning by being able to point consumers to their entire store network (Gielens and Steenkamp 2019; Loupiac and Goudey 2019). In summary, the three studies in this doctoral thesis have important practical implications for omni-channel retailers that increasingly need to coordinate processes and technologies across major offline and online sales channels to provide consumers with seamless experiences. As summarized in an overview of the analysis steps of this doctoral thesis in Figure 5.1, retailers need to understand the interplay between their major offline and online sales channels to retain consumers. To be successful, they must also provide and manage a broad range of stimuli and cues that influence consumer behavior and responses (e.g., Becker and Jaakkola 2020; Gao, Melero and Sese 2019). The three studies conducted in this doctoral thesis point to important general managerial implications, which are subsequently highlighted. This doctoral thesis has shown that managers of omni-channel firms should place special emphasis on channel interdependencies between major offline and online sales channels. They profit the most from offline to online channel image

5.1 Discussion and Conclusions

193

Loyalty

I. Channel interdependencies

II. Channel integration services Prior experience (t-n)

III. Specific marketing instruments

Purchase intention Repurchase intention

Consumers’ online shopping experience and channel congruence

Figure 5.1 Overview of analysis steps and resulting effects on behavioral outcomes. (Source: Own creation)

transfers to retain consumers, in dependence of favorable prior experiences. This finding is valuable, as today, former brick-and-mortar retailers, like H&M and Zara are closing physical stores, possibly ignoring their role for online stores (Ailawadi and Farris 2017; Valentini, Neslin and Montaguti 2020). Only omni-channel retailers can take advantage of the benefits of their physical stores compared to online pure players. To manage these transfers, however, these retailers must overcome frequent silos of consumer information to match consumer behavior across offline and online channels, and they must operate online and offline divisions dependently (e.g., Herhausen et al. 2015). To further support such transfers, retailers have started to integrate multiple business functions across channels (Banerjee 2014; Gao, Melero and Sese 2019). In this doctoral thesis, OF-ON services have been identified to be most useful in increasing purchase intentions of consumers in physical and online stores. Retailers that offer OF-ON services succeed in steering consumers to their online and strategically important physical channels, which is important because offline retailing is still the anchor of the fashion sector, as described in Section 1.1. In addition, this thesis has affirmed that important context factors need to be considered to benefit from this important lever. It should be noted that the recent spread of the coronavirus might change the types of experiences that consumers are likely to value. The current crisis could alter the role of OF-ON and ON-OF services as consumers become accustomed to new ways of shopping (Roggeveen and Sethuraman 2020). The recent

194

5

Final Remarks

shifts and developments point to a further increase in online sales, and for omnichannel managers, it is now even more important to strategically use imperative marketing instruments to strengthen consumer online trust and brand equity and ultimately repurchase intentions. Facing competition from other retailers, online pure players, and online marketplaces, a retailer’s online store must be aesthetically pleasing, and should provide secure payment methods, and protect privacy. Moreover, omni-channel retailers can best leverage their physical store presence by offering consistent information across online and offline stores and by applying online-offline integration. The omni-channel-specific instruments are identified as especially important in competition with online retailers due to the specific competencies and resources needed to implement them (e.g., Acquila-Natale and Chaparro-Peláez 2020; Yrjölä, Spence and Saarijärvi 2018).

5.2

Further Research

In addition to valuable theoretical and managerial implications, this doctoral thesis also offers starting points for further research. Limitations and avenues for further research have already been highlighted at the ends of each of the three studies. In this section, these limitations are integrated and discussed in the context of this thesis, while general areas for further research are also presented. They are structured in terms of data basis, measurement, conceptual framework, theory, and methodology. First, the data basis of the three studies could be enhanced in future studies. Although the fashion sector was consciously chosen for this doctoral thesis due to several reasons outlined in Section 1.1., studying other sectors will enhance the generalizability of the presented findings. For example, initial findings point to the still important role of physical stores in banking (e.g., Cambra-Fierro et al. 2020), which should be further examined. The effects of integration services for consumers in other sectors that comprise physical and online stores would be insightful (e.g., banking or healthcare, Banerjee 2014; Dahl, Milne and Peltier 2019). Further specific marketing instruments exist in the electronics sector, or for more transaction-oriented websites, in the travel sector, and should be examined (e.g., Khan et al. 2019). While previous experiences have been considered an important individual factor that changes evaluations of consumers in omnichannel retailing, carefully studying specific consumer segments might change the results. Differentiating, for example, between store- or online-focused consumers might allow segment-specific consumer behavior to be monitored (e.g., Herhausen et al. 2019). In addition to exploring effects for online pure players

5.2 Further Research

195

with fewer connections to physical stores, research into the immediate impacts of the spread of the novel coronavirus was mentioned (Roggeveen and Sethuraman 2020). The interdependencies between channels and the need to support them with appropriate stimuli and cues may increase in the future due to the crisis. Furthermore, the channel scope of this doctoral thesis lies on major offline and online purchase channels that are brand-owned and directly controllable. Broadening the channel scope to non-firm-controlled or social touchpoints will yield richer insights into how perceptions and evaluations of an omni-channel retailer are built (e.g., Swoboda and Winters 2020), however, this requires new empirical methods (e.g., De Keyser et al. 2020). Examples for interdependencies between further retail-brand-owned, retail-partner-owned and social touchpoints are illustrated in Figure 5.2. In addition to important interdependencies between offline and online stores, interdependencies are likely to exist between all points of interaction between consumers and retailers, especially in a future fully integrated retail environment. For example, a consumer learns about a product on his or her social network, then uses a search engine to compare prices, buys the product in the offline store of an omni-channel retailer, and writes a review on his or her experience with that retailer (Grewal and Roggeveen 2020).



… Loyalty programs

TV Social media

Newsletter

Promotions



Retail brand-owned touchpoints

Social network

Showrooms Media Public transport

Packaging Voice commerce

Drive-in/ through



Reviews

Word-ofmouth Online stores

Chatbots Search engines



Cars Offline stores

Workplace Smart home

Email

Platforms Pick-up station

Retail partner-owned touchpoints

Social touchpoints

Figure 5.2 Possible interdependencies between further touchpoints. (Source: Own creation)

Second, although existing scales from previous research have been used, future research could consider refining certain measurements of the previously presented empirical studies. In the three studies, consumer behavior was analyzed based on self-reported behavioral intentions. Future research, should supplement the used

196

5

Final Remarks

intention measures with objective purchases, for example, using secondary financial performance data (e.g., Herhausen et al. 2020). Although channel image is carefully captured in Study 1, no common agreement on a measure exists and future research may study reciprocal effects of more fine-grained constructs (e.g., marketing mix, Blut, Teller and Floh 2018). Furthermore, retailers increasingly use instruments based on technological innovations, which could be considered in Study 3 in addition to the proposed instruments (e.g., omni-channel customization, social chat bots, Becker et al. 2020; Sousa and Amorim 2018). With regard to the mechanisms in Study 3, the mediator online trust can be further differentiated into cognitive and affective dimensions to understand emotional bonds that serve individuals as the basis for trust (Toufaily and Pons 2017; Ye et al. 2020). Studying online, offline or overall brand equity would be interesting, as associations of a brand as strong, attractive or unique may differ in competitive situations (e.g., Batra and Keller 2016). For the moderators in Study 1, the multi-group cross-lagged models were based on the data of respondents who answered with reference to different retailers and were categorized as low/high in terms of prior offline (online) experience. Future research should modify this measure to take into account further dynamics between channels (e.g., continuous moderations, Melis et al. 2015; Grewal and Roggeveen 2020). Third, extending the conceptual frameworks of the three studies seems promising. Studying reciprocity between further communication touchpoints or their antecedent roles is promising in our framework but is challenging in cross-lagged panel models. New methods need to be applied to capture reciprocal relationships between more than two channels or touchpoints simultaneously (e.g., De Keyser et al. 2020). Informational, transactional, and fulfillment-specific integration services are used in Study 2. Although, these services are most important from the consumer perspective and comprise different stages of the consumer journey, future work could differentiate their effects at the pre-purchase, purchase, and post-purchase stages. In a similar vein, further directions for integration exist. Online-to-online services of platforms or social media touchpoints should be further analyzed (e.g., Ibrahim and Wang 2019). In Study 2, instead of crosschannel effects, scholars should apply a reciprocal design, as these relationships among perceived channel quality, value, or image are obvious but seldom tested in omni-channel studies (e.g., Swoboda, Weindel and Schramm-Klein 2016). Research has placed particular emphasis on studying channel integration quality, but analyzing important drivers of the integration services is also advantageous (Banerjee 2014; Cao and Li 2018). Due to the model complexity, no moderators were used in Study 3. Future research should account for moderating effects.

5.2 Further Research

197

Consumers’ online shopping experience, for example, might increase the associations in consumers’ minds and could increase the effects for repurchase intentions (e.g., Emrich and Verhoef 2015). Fourth, regarding theory, throughout this doctoral thesis cognitive theoretical rationales were applied. However, it would be interesting to explain the effects across channels affectively. Particularly, online effects might become stronger, if the emotional responses of consumers are conceptualized (e.g., Liu, Li and Hu 2013; Verhagen, van Dolen and Merikivi 2019). A combination that incorporates cognitive and affective components could be a fruitful approach to further explaining the resulting effects (e.g., Ye et al. 2020). Moreover, most consumer decisions are memory-based (Lynch, Marmorstein and Weigold 1988), but activation in consumers’ mind can be more automatic or controlled. Research on the automation of the activation processes outlined in this doctoral thesis should be revealing in order to clarify whether consumer decisions are unconscious or whether conscious processes take place (e.g., Feldman 1981). Finally, regarding methodology, in Studies 1 and 3, longitudinal data were used and analyzed with appropriate sequential mediation analysis and cross-lagged panel models. These models are able to establish causality of the obtained estimates (e.g., Wiedermann and von Eye 2015). Nevertheless, the longitudinal data could be extended by further waves, allowing the inclusion of further antecedences for the constructs under investigation. Importantly, the results of Study 2 are based on cross-sectional data without time-dependent effects. This issue may be ameliorated by applying sequential mediation analyses to account for effects over time (Mitchell and Maxwell 2013). Moreover, future studies with experimental designs will need to be undertaken to manipulate the accessibility of integration services (Bhargave, Mantonakis and White 2016) or the activation of marketing instruments by manipulating their cognitive load (e.g., Wen and Lurie 2019). Likewise, measuring diagnosticity directly may provide further insights for mediation paths (e.g., Qiu, Pang and Lim 2012). Applying new methods, like fMRI, will help detect the brain mechanisms that are activated when consumers evaluate retailers (e.g., initially conducted for online trust, Casado-Aranda, Dimoka and Sánchez-Fernández 2019). Finally and importantly, this doctoral thesis contributed insights at the national level in Germany, while future research across nations could explore cross-cultural differences in responses to stimuli and cues in omni-channel retailing (e.g., Shavitt and Barnes 2020).

Appendix

1. Acknowledgements This doctoral thesis consists of three studies. At the time, this thesis has been prepared, these three studies were in the process of being evaluated by a scientific journal and were not yet finally published. The first study was under review at Journal of Business Research, the second study was under review at Decision Support Systems and the third study was under review at Journal of Business Research. Each of the journals is ranked B according to the VHB-JOURQUAL 3 journal ranking of the German Academic Association for Business Research (VHB 2020). At the time of publication of this doctoral thesis, later versions of the articles have already been accepted for publication or are published. The author retains the right to use the articles for his or her scientific career by including them in his or her doctoral thesis. In the following, reference is made to the publication of the later versions of the articles. Study 1: Swoboda, B. & Winters, A. (2021). Reciprocity within Major Retail Purchase Channels and their Effects on Overall, Offline and Online Loyalty. Journal of Business Research, 125 (3), 279–294. Study 2: Swoboda, B. & Winters, A. (2021). Effects of the Most Useful Offline-Online and Online-Offline Channel Integration Services for Consumers. Decision Support Systems, 145 (June).

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 A. Winters, Omni-Channel Retailing, Handel und Internationales Marketing Retailing and International Marketing, https://doi.org/10.1007/978-3-658-34707-9

199

200

Appendix

2. Study 1: Reciprocity within Major Retail Purchase Channels and their Effects on Overall, Offline and Online Loyalty 2.1. Sample Selection and Manipulation Check

Table A.1 Sales growth 2015–2017 Retailer 1

Retailer 2

Retailer 3

Retailer 4

Retailer 5

Retailer 6

Retailer 7

Retailer 8

 Sales 2017–2015

− 1.74 %

5.09 %

− 14.50 %

− 14.29 %

21.05 %

− 6.01 %

− 8.13 %

− 3.64 %

 Sales 2017–2016

− 2.35 %

.21 %

− 4.50 %

− .85 %

8.58 %

− 2.67 %

− 6.81 %

− 4.79 %

 Sales 2017–2015

.63 %

4.87 %

− 10.47 %

− 13.55 %

11.48 %

− 3.43 %

− 1.41 %

1.21 %

Source: Own creation

To guide the manipulation of retailers with witch consumers have more vs. less favorable prior experiences, we conducted a series of t-tests followed by ANOVAs (see Table A.1, A.2 and Table A.3). The t-tests and ANOVAS revealed significant differences between the groups.

Table A.2 Independent sample t-tests: prior offline and online experiences Prior Offline Experiences

Prior Online Experiences

Dependent variable

Group

MV/Std.

t-test (2-sided)

p

MV/Std.

t-test (2-sided)

p

Offline channel image

less favorable

3.3/1.6

− 9.5

.000

3.2/1.5

− 11.3

.000

more favorable

4.5/1.3

Online channel image

less favorable

3.9/1.4

− 12.2

.000

more favorable

4.7/1.2

4.6/1.3 − 6.4

.000

3.6/1.3 4.9/1.1

Notes: MV/Std.=mean value and standard deviation, N = 475, equal variances assumed. Source: Own creation

Appendix

201

Table A.3 ANOVA Fashion Dependent variable

Effect

MS

F

p

Partial η2

Offline channel image

OFP

258.4

156.1

.000

.215

ONP

195.5

118.0

.000

.171

.8

.5

.083

.001

OFP x ONP Online channel image

Error

1.7

OFP

74.9

54.6

.000

.087

ONP

228.1

166.4

.000

.226

OFP x ONP

6.5

4.8

.030

.008

Error

1.4

Notes: MS = mean square, OFP = Offline experience, ONP = Online experience. Source: Own creation

2.2. Common-method Bias-testing Collecting data at different time points and using an appropriate questionnaire design diminishes the potential threat of CMV within our data set ex ante. Respondents were told that the study was anonymous and confidential and that their answers could neither be right or wrong. The study started with the dependent measures (Chang et al. 2010). A posteriori we accounted for CMV by calculating a single-factor test using confirmatory factor analysis (Table A.4). Table A.5 and Table A.6 show the results for the marker variable technique following the latent variable approach of Williams et al. (2010). We used selfefficacy as a marker variable. First, it is an ideal marker because it is theoretically unrelated to our constructs. Second, it is similar to our constructs in content and format, thus it might be equivalently vulnerable to the same causes of CMV (Simmering et al. 2015). The marker variable technique consists of three consecutive phases. The correlations between the latent constructs are not biased through the marker variable (phase I, Method-C vs. -R). The results of the following reliability decomposition (phase II) indicate that the amount of method variance, associated with the measurement of the substantive latent constructs, is less than 14.44 percent (between 3.88 and 14.44 percent). Since previous literature (e.g., Williams, Hartman and Cavazotte 2010) found impacts up to 19.7 percent, the possibility of CMV seems to be reduced. This is also supported by the third phase, which shows only a minor impact of the marker-based method variance on construct correlations (Table A.7)

202

Appendix

Table A.4 Study 1: results of the single-factor test CFI TLI RMSEA SRMR χ2 (df)

 χ2 (df)

p-value of difference

Overall retailer loyalty model Time point one Proposed model

.936 .917 .130

.047

Single factor model

.660 .584 .291

.132

546.749 (51) 2133.773 (3) .000 2680.522 (54)

Time point two Proposed model

.948 .933 .118

.042

Single factor model

.689 .619 .281

.124

459.111 (51) 2055.225 (3) .000 2514.336 (54)

Time point three Proposed model

.957 .944 .111

.041

Single factor model

.648 .614 .293

.125

414.077 (51) 2289.094 (3) .000 2703.171 (54)

Offline channel loyalty model Time point one Proposed model

.934 .915 .131

.046

Single factor model

.657 .581 .290

.133

554.436 (51) 2116.100 (3) .000 2670.536 (54)

Time point two Proposed model

.949 .934 .117

.043

Single factor model

.686 .616 .283

.127

453.417 (51) 2082.440 (3) .000 2535.857 (54)

Time point three Proposed model

.955 .942 .114

.043

Single factor model

.672 .599 .301

.127

432.814 (51) 2429.596 (3) .000 2862.410 (54) (continued)

Appendix

203

Table A.4 (continued) CFI TLI RMSEA SRMR χ2 (df)

 χ2 (df)

p-value of difference

Online channel loyalty model Time point one Proposed model

.961 .950 .096

.033

Single factor model

.601 .513 .299

.145

321.629 (51) 2512.496 (3) .000 2834.125 (54)

Time point two Proposed model

.970 .961 .087

.026

Single factor model

.620 .536 .301

.154

273.642 (51) 2594.207 (3) .000 2867.849 (54)

Time point three Proposed model

.975 .968 .082

.025

Single factor model

.614 .528 .318

.150

249.878 (51) 2936.644 (3) .000 3186.522 (54)

Notes: Difference tests were conducted using χ2 tests of difference. Source: Own creation

Table A.5 Study 1: results of the model comparisons (phase I) Time point 1 χ2

df

CFA

594.642

84

Baseline

592.235

Method-C

588.106

Method-U Method-R

Model

CFI

TLI

RMSEA

SRMR

SCF

.939

.923

.103

.040

1.11

92

.940

.931

.097

.055

1.13

91

.940

.931

.098

.044

1.13

573.449

80

.941

.922

.104

.040

1.14

587.437

94

.941

.934

.096

.043

1.13

LOY

Chi-square differences of model comparison tests Models

χ2

df

p (continued)

204

Appendix

Table A.5 (continued) Time point 1 Baseline with Method-C

4.129

1

**

Method-C with Method-U

14.657

9

ns

Method-C with Method-R

.669

3

ns

OFLOY CFA

604.372

84

.937

.921

.104

.040

1.12

Baseline

603.063

92

.938

.929

.098

.053

1.14

Method-C

598.312

81

.939

.929

.098

.043

1.14

Method-U

583.295

80

.939

.920

.105

.040

1.16

Method-R

597.914

94

.939

.923

.097

.042

1.14

.078

.030

1.19

Chi-square differences of model comparison tests Models

df

p

4.751

1

**

Method-C with Method-U

15.017

9

ns

Method-C with Method-R

.398

3

ns

CFA

375.880

84

.962

.953

Baseline

402.700

92

.960

.954

.077

.053

1.20

Method-C

398.002

91

.960

.954

.077

.038

1.20

Method-U

375.687

80

.962

.950

.080

.031

1.22

Method-R

398.644

94

.961

.956

.075

.037

1.20

TLI

RMSEA

SRMR

SCF

.094

.037

1.15

Baseline with Method-C

χ2

ONLOY

Chi-square differences of model comparison tests Models

df

p

4.698

1

**

Method-C with Method-U

22.315

11

ns

Method-C with Method-R

.642

3

ns

Baseline with Method-C

χ2

Time point 2 χ2

df

CFA

507.952

84

.951 .939

Baseline

510.061

92

.952 .945

.089

.062

1.17

Method-C

500.103

91

.953 .945

.088

.038

1.17

Method-U

484.241

80

.953 .938

.094

.037

1.19

Model

CFI

LOY

(continued)

Appendix

205

Table A.5 (continued) Time point 2 Method-R

499.583

94

.952 .947

.087

.058

1.18

Chi-square differences of model comparison tests Models Baseline with Method-C

χ2 9.958

Method-C with Method-U 15.862 Method-C with Method-R

.520

df 1 11

p ** ns

3

ns

OFLOY CFA

503.711

84

.951 .939

.093

.038

1.15

Baseline

504.814

92

.952 .946

.088

.062

1.17

Method-C

494.598

91

.953 .946

.088

.038

1.17

Method-U

477.566

80

.954 .940

.093

.037

1.19

Method-R

494.140

94

.955 .948

.086

.038

1.17

Chi-square differences of model comparison tests Models

χ2

df

p

Baseline with Method-C

10.216

1

**

Method-C with Method-U

17.032

11

ns

Method-C with Method-R .458

3

ns

ONLOY CFA

321.011

84

.971 .964

.070

.026

1.18

Baseline

342.700

92

.969 .965

.069

.070

1.20

Method-C

330.009

91

.971 .966

.068

.035

1.20

Method-U

308.109

80

.972 .963

.070

.025

1.22

Method-R

330.845

94

.971 .968

.066

.035

1.19

TLI

RMSEA

SRMR

SCF 1.13

Chi-square differences of model comparison tests Models

χ2

df

p

Baseline with Method-C

12.691

1

***

Method-C with Method-U

21.900

11

ns

Method-C with Method-R

.836

3

ns

Time point 3 χ2

df

CFA

479.184

84

.957

.946

.091

.036

Baseline

482.084

92

.958

.952

.086

.054

Model

CFI

LOY 1.14 (continued)

206

Appendix

Table A.5 (continued) Time point 3 Method-C

477.467

91

.958

.952

.086

.039

1.14

Method-U

458.475

80

.959

.946

.091

.036

1.16

Method-R

477.816

94

.958

.953

.084

.039

1.14

.093

.037

1.10

Chi-square differences of model comparison tests Models

df

p

4.617

1

**

Method-C with Method-U

18.992

11

ns

Method-C with Method-R

.349

3

ns

CFA

495.141

84

.956

.945

Baseline

496.255

92

.957

.951

.088

.055

1.12

Method-C

490.390

91

.958

.951

.088

.039

1.13

Method-U

472.445

80

.958

.945

.093

.037

1.14

Method-R

489.591

94

.958

.953

.086

.039

1.12

.069

.025

1.15

Baseline with Method-C

χ2

OFLOY

Chi-square differences of model comparison tests Models Baseline with Method-C

χ2

df

p

5.865

1

**

Method-C with Method-U

17.945

11

ns

Method-C with Method-R

.799

3

ns

CFA

318.494

84

.974

.968

Baseline

320.480

92

.974

.971

.066

.056

1.16

Method-C

312.238

91

.975

.971

.065

.028

1.16

Method-U

298.253

80

.976

.968

.069

.024

1.17

Method-R

312.262

94

.975

.973

.064

.028

1.16

ONLOY

Chi-square differences of model comparison tests Models Baseline with Method-C

χ2

df

p

8.242

1

**

Method-C with Method-U

13.985

11

ns

Method-C with Method-R

.024

3

ns

Notes: LOY = Overall retailer loyalty model, OFLOY = Offline channel loyalty model, ONLOY = Online channel loyalty model, SCF = Scaling correction factor for MLM, ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001. Source: Own creation

Appendix

207

Table A.6 Study 1: results of the reliability decomposition (phase II) Time point 1

Latent variable

Reliability baseline model

Decomposed reliability from method-C model

Total reliability

Substantive reliability

Method reliability

% reliability marker variable

Overall retailer loyalty model Offline channel image

.998

.883

.115

11.55

Online channel image

.998

.897

.114

11.45

Loyalty

.995

.856

.139

13.96

Offline channel loyalty model Offline channel image

.998

.879

.119

11.96

Online channel image

.998

.879

.119

11.91

Loyalty

.995

.852

.144

14.44

Online channel loyalty model Offline channel image

.998

.865

.133

13.31

Online channel image

.998

.872

.126

12.63

Loyalty

.996

.852

.144

14.44

Time point 2

Latent variable

Reliability baseline model

Decomposed reliability from method-C model

Total reliability

Substantive reliability

Method reliability

% reliability marker variable

Overall retailer loyalty model Offline channel image

.998

.929

.065

6.55

Online channel image

.998

.931

.067

6.70

Loyalty

.995

.915

.081

8.10 (continued)

208

Appendix

Table A.6 (continued) Time point 1

Latent variable

Reliability baseline model

Decomposed reliability from method-C model

Total reliability

Substantive reliability

Method reliability

% reliability marker variable

Offline channel loyalty model Offline channel image

.998

.935

.063

6.30

Online channel image

.998

.934

.064

6.41

Loyalty

.995

.920

.076

7.62

Online channel loyalty model Offline channel image

.998

.959

.039

3.88

Online channel image

.998

.958

.039

3.95

Loyalty

.994

.947

.047

4.68

Time point 3

Latent variable

Reliability baseline model

Decomposed reliability from method-C model

Total reliability

Substantive reliability

Method reliability

% reliability marker variable

Overall retailer loyalty model Offline channel image

.999

.939

.056

5.62

Online channel image

.998

.941

.057

5.68

Loyalty

.995

.928

.068

6.81

Offline channel loyalty model Offline channel image

.998

.947

.048

4.82

Online channel image

.998

.949

.049

4.87 (continued)

Appendix

209

Table A.6 (continued) Time point 1 Reliability baseline model

Decomposed reliability from method-C model

Latent variable

Total reliability

Substantive reliability

Method reliability

Loyalty

.996

.939

.057

5.72

% reliability marker variable

Online channel loyalty model Offline channel image

.998

.904

.091

9.15

Online channel image

.998

.906

.092

9.23

Loyalty

.996

.892

.104

10.42

Source: Own creation Table A.7 Study 1: results of the sensitivity analyses (phase III) Time point 1 Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

Overall retailer loyalty model OF with LOY

.816

.816

.813

.868

.869

ON with LOY

.552

.552

.548

.669

.671

OF with ON

.603

.603

.598

.707

.709

SELF with OF

− .086

.000

.000

.000

.000

SELF with ON

− .037

.000

.000

.000

.000

SELF with LOY

− .078

.000

.000

.000

.000

Offline channel loyalty model OF with LOY

.812

.812

.810

.864

.865

ON with LOY

.541

.541

.538

.657

.660

OF with ON

.603

.603

.598

.707

.709

SELF with OF

− .086

.000

.000

.000

.000

SELF with ON

− .037

.000

.000

.000

.000

SELF with LOY

− .062

.000

.000

.000

.000 (continued)

210

Appendix

Table A.7 (continued) Time point 1 Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

Online channel loyalty model OF with LOY

.622

.622

.616

.741

.743

ON with LOY

.590

.590

.585

.713

.715

OF with ON

.598

.598

.594

.715

.717

SELF with OF

− .085

.000

.000

.000

.000

SELF with ON

− .037

.000

.000

.000

.000

SELF with LOY

− .100

.000

.000

.000

.000

Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

Time point 2

Overall retailer loyalty model OF with LOY

.842

.842

.839

.889

.890

ON with LOY

.626

.626

.619

.736

.738

OF with ON

.638

.638

.631

.742

.744

SELF with OF

− .076

.000

.000

.000

.000

SELF with ON

− .079

.000

.000

.000

.000

SELF with LOY

− .092

.000

.000

.000

.000

Offline channel loyalty model OF with LOY

.837

.837

.835

.889

.886

ON with LOY

.578

.578

.571

.736

.703

OF with ON

.638

.638

.631

.742

.743

SELF with OF

− .077

.000

.000

.000

.000

SELF with ON

− .079

.000

.000

.000

.000

SELF with LOY

− .085

.000

.000

.000

.000

Online channel loyalty model OF with LOY

.625

.626

.615

.885

.749

ON with LOY

.612

.612

.598

.701

.741

OF with ON

.635

.635

.629

.742

.740

SELF with OF

− .075

.000

.000

.000

.000

SELF with ON

− .079

.000

.000

.000

.000

SELF with LOY

− .156

.000

.000

.000

.000 (continued)

Appendix

211

Table A.7 (continued) Time point 1 Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

Method-C

Method-S (.05)

Method-S (.01)

Time point 3 Construct correlations

CFA

Baseline

Overall retailer loyalty model OF with LOY

.826

.826

.824

.881

.883

ON with LOY

.588

.588

.585

.714

.716

OF with ON

.657

.657

.652

.766

.768

SELF with OF

− .083

.000

.000

.000

.000

SELF with ON

− .069

.000

.000

.000

.000

SELF with LOY

− .047

.000

.000

.000

.000

Offline channel loyalty model OF with LOY

.804

.804

.802

.866

.867

ON with LOY

.559

.559

.556

.693

.695

OFF with ON

.657

.657

.652

.766

.768

SELF with OF

− .084

.000

.000

.000

.000

SELF with ON

− .069

.000

.000

.000

.000

SELF with LOY

− .043

.000

.000

.000

.000

Online channel loyalty model OF with LOY

.632

.632

.625

.753

.756

ON with LOY

.578

.577

.570

.715

.718

OF with ON

.655

.655

.649

.764

.766

SELF with OF

− .082

.000

.000

.000

.000

SELF with ON

− .069

.000

.000

.000

.000

SELF with LOY

− .101

.000

.000

.000

.000

Notes: LOY = Loyalty, OF = Offline channel image, ON = Online channel image, SELF = Self-efficacy. Source: Own creation

212

Appendix

2.3. Unobserved Heterogeneity To reduce the possibility of unobserved heterogeneity we tested for endogeneity by using the IV approach with offline and online channel attributes. In a first step, we checked whether the instruments are strong predictors for offline and online channel image using F-tests. As the calculated F-values exceeded the recommended threshold of 10 (see Table A.8), they can be considered to be a strong predictors (Antonakis et al. 2014). Additionally to the efficient (proposed) models (Antonakis et al. 2010), we estimated consistent models which included the two IVs (see Table A.9) and tested if there was a change in path estimates (Hausman 1978). Table A.8 Study 1: F-test of strong instrumental variables Overall retailer loyalty model

Offline channel loyalty model

Online channel loyalty model F-value

F-value

F-value

IV1 → Offline channel image

1166.569

1108.417

902.822

IV2 → Online channel image

1834.044

1907.477

1436.660

Notes: IV = Instrumental variable, F-value > 10 indicates strong predictor. Source: Own creation

2.4. Measurement Invariance Tests First, we computed measurement invariance across the three time points by applying confirmatory factor analysis. First, we assessed the model fit of the baseline model—which estimates factor loadings and intercepts freely—to assure configural invariance. Second, we estimate the metric invariant model in which factor loadings are constrained to be equal across time points. The goodness-of-fit statistics are then compared to those of the baseline model. To determine metric invariance (MI), we applied several differences-in-fit indices (e.g., chi square difference tests and CFI). MI was ascertained between the time points (Table A.10).

.103

.032

.121

.040

Online CI (1) → Offline CI (2)

Offline CI (1) → LOY (2)

Online CI (1) → LOY (2)

**

***

*

***





Online IV (1) → Online CI (1)

Offline CI (1) → Online CI (2)





.040

.098

.031

.089

.925

.887

***

***

***

***

***

***

p

β





.038

.120

.034

.097

**

***

*

***





p

β

p

β

Offline IV (1) → Offline CI (1)

Direct effects

Efficient model

Consistent model

Efficient model β

.042

.117

.034

.085

.926

.889

Consistent model

Model 2: Offline channel loyalty

Model 1: Overall retailer loyalty

Table A.9 Study 1: results of the efficient and consistent models

***

***

***

***

***

***

p

β





.027

.072

.035

.098

Efficient model

*

***

**

***





p

β

**

**

***

***

***

***

(continued)

.034

.075

.031

.095

.909

.941

p

Consistent model

Model 3: Online channel loyalty

Appendix 213

.734

.708

.115

.035

.133

.045

Online CI (1) → Online CI (2)

LOY (1) → LOY (2)

Offline CI (2) → Online CI (3)

Online CI (2) → Offline CI (3)

Offline CI (2) → LOY (3)

Online CI (2) → LOY (3)

*

***

**

***

***

***

***

p

.047

.116

.033

.105

.775

.739

.821

p

***

***

***

***

***

***

***

β

.043

.131

.038

.108

.702

.737

.850

**

***

*

***

***

***

***

p

β

.852

β

Offline CI (1) → Offline CI (2)

Efficient model

Consistent model

Efficient model β

.048

.137

.037

.098

.750

.743

.819

Consistent model

Model 2: Offline channel loyalty

Model 1: Overall retailer loyalty

Table A.9 (continued)

p

***

***

***

***

***

***

***

β

.032

.081

.039

.108

.738

.737

.848

Efficient model p

*

***

**

***

***

***

***

β

p

**

***

***

***

***

***

***

(continued)

.041

.087

.035

.107

.758

.724

.840

Consistent model

Model 3: Online channel loyalty

214 Appendix

.758

.879

LOY (2) → LOY (3)

R2 LOY (3)

Covariates

t = 5.036**

.068

Online CI (1) → LOY (3)

Diff. in total effects

.209

Offline CI (1) → LOY (3)

Total effects

.841

Online CI (2) → Online CI (3)

*

**

***

***

***

***

p

t = 9.209**

.071

.178

.852

.803

.846

.900

p

**

***

***

***

***

***

.072

.206

.867

.754

.846

.910

t = 9.217**

β

**

***

***

***

***

***

p

β

.913

β

Offline CI (2) → Offline CI (3)

Efficient model

Consistent model

Efficient model

.066

.193

.834

.770

.850

.896

t = 9.798**

β

Consistent model

Model 2: Offline channel loyalty

Model 1: Overall retailer loyalty

Table A.9 (continued)

p

**

***

***

***

***

***

.048

.131

.788

.812

.845

.909

t = 3.795**

β

Efficient model p

*

***

***

***

***

***

.061

.138

.757

.816

.841

.906

p

**

***

***

***

***

***

(continued)

t = 6.627**

β

Consistent model

Model 3: Online channel loyalty

Appendix 215

.003

.003

.032

.031

.032

.024

Gender (2) → LOY (2)

Gender (3) → LOY (3)

Age (1) → LOY (1)

Age (2) → LOY (2)

Age (3) → LOY (3)

Internet expertise (1) → LOY (1)

p

*

**

**

**

ns

ns

ns

.030

.037

.034

.033

.005

.004

.004

p ns

**

***

***

***

ns

ns

β

ns

− .010

.007

.011

.011

ns

ns

ns

ns

ns

− .010

.011

ns

− .010

p

β

.003

β

Gender (1) → LOY (1)

Efficient model

Consistent model

Efficient model β

.014

.016

.015

.014

− .008

− .007

− .006

Consistent model

Model 2: Offline channel loyalty

Model 1: Overall retailer loyalty

Table A.9 (continued)

p

ns

ns

ns

ns

ns

ns

ns

β

.040

.004

.004

.004

.012

.013

.012

Efficient model

ns

ns

ns

ns

ns

ns

**

p

β

p

**

ns

ns

ns

ns

ns

ns

(continued)

.042

.007

.007

.007

.016

.014

.013

Consistent model

Model 3: Online channel loyalty

216 Appendix

.157

.144

.154

Familiarity (1) → LOY (1)

Familiarity (2) → LOY (2)

Familiarity (3) → LOY (3)

***

***

***

*

*

.151

.136

.140

.033

.030

p

***

***

***

**

**

β

.163

.154

.170

.006

.006

ns

ns

***

***

***

p

β

.169

.155

.159

.015

.014

ns

ns

***

***

***

p

β

.039

.035

.039

.040

.037

Efficient model

**

**

**

**

**

p

β

.047

.042

.044

.045

.031

**

**

***

***

***

p

Consistent model

Model 3: Online channel loyalty

Source: Own creation

Structural model fits: Model 1: Efficient model: CFI .935, TLI .931, RMSEA .066, SRMR .079, χ2 (1009) = 3506.910, SCF = .87. Consistent model: CFI .914, TLI .911, RMSEA .067, SRMR .196, χ2 (1410) = 5064.092, SCF = .87. Model 2: Efficient model:CFI .933, TLI .930, RMSEA .066, SRMR .078, χ2 (1009) = 3547.310, SCF = .89. Consistent model: CFI .912, TLI .908, RMSEA .068, SRMR .193, χ2 (1410) = 5122.172, SCF = .88. Model 3: Efficient model: CFI .949, TLI .946, RMSEA .056, SRMR .063, χ2 (1009) = 2835.315, SCF = .88. Consistent model: CFI .932, TLI .929, RMSEA .058, SRMR .159, χ2 (1410) = 4153.234, SCF = .88. Notes: IV = Instrumental variable, CI = Channel image, LOY = Loyalty, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 573. Standardized coefficients are shown. Differences between total effects have been tested using t-tests. ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001.

.024

Internet expertise (3) → LOY (3)

p

β

.023

β

Internet expertise (2) → LOY (2)

Efficient model

Consistent model

Efficient model

Consistent model

Model 2: Offline channel loyalty

Model 1: Overall retailer loyalty

Table A.9 (continued)

Appendix 217

218

Appendix

Table A.10 Study 1: measurement invariance across time points Model

χ2 /df (p-value)

χ2 -Difference (p-value)

CFI (CFI)

TLI (TLI)

.965

.958

RMSEA (RMSEA)

SCF

.057

1.18

.057

1.16

1.16

Overall retailer loyalty model Model 1:

1488.396/522

Configural invariance

(.000)

Model 2:

1527.592/540

Full metric invariance Model 3: Partial metric invariancea

(.000) 1520.006/539 (.000)

(−) 32.085 (.021) 22.045 (.183)

.965 (−) .965 (−)

(−)

(−) .959

(.001)

(−)

.959

.056

(.001)

(.001)

.954

.058

1.16

1.18

Offline channel loyalty model Model 1:

1510.927/522

Configural invariance

(.000)

Model 2:

1506.523/540

Full metric invariance

(.000)

.962

13.678 (.750)

.965

.959

.056

(.003)

(.005)

(.002)

.974

.969

.046

1.65

1.82

Online channel loyalty model Model 1:

1158.333/522

Configural invariance

(.000)

Model 2:

1155.085/540

9.570

.976

.972

.045

(.000)

(.945)

(.002)

(.003)

(.001)

Full metric invariance

Notes: SCF = Scaling correction factor for MLM. a Factor loading freed for the following item: LOY1 time point one. Source: Own creation

2.5. Description of the Cross-lagged Panel Model Cross-lagged models apply to the most appropriate methods for studying causality in longitudinal data and reciprocal relationships over time (Allison, Williams and Moral-Benito 2017). The stability of the constructs is controlled by the inclusion

Appendix

219

Offline channel image (1)

Offline channel image (2)

Offline channel image (3)

Online channel image (1)

Online channel image (2)

Online channel image (3)

Loyalty (1)

Loyalty (2)

Loyalty (3)

Figure A.1 Cross-lagged panel model. (Source: Own creation)

of autoregressive relationships between a variable and its prior state, because the prior state of a variable determines its current state (Zyphur et al. 2019). All constructs were measured three times. Burkholder and Harlow (2003) advice for the including of disturbance correlations in cross-lagged designs with respect to the indicators. Disturbance correlations were modeled between the same indicators across the three time points. We also included disturbance correlations between all constructs at time point two and integrated them at time point three and are constrained and thus estimated equally (Allison, Williams and Moral-Benito 2017) (Figure A.1), (Table A.11, A.12, A.13, A.14, A.15, A.16, A.17, A.18, A.19, A.20, A.21, A.22, A.23).

2.6. Alternative Cross-lagged Panel Models

Table A.11 Study 1: reliability and validity of alternative general models I Time point one Construct

Item

MV/Std.

FL

KMO

ItTC

α

Offline retail brand equity

OFB1

4.0/1.7

.924

.841

.886

.944

Online retail brand equity

OFB2

4.1/1.6

.949

.907

OFB3

4.1/1.6

.876

.847

OFB4

3.5/1.6

.848

.824

ONB1

4.2/1.5

.939

ONB2

4.2/1.5

.963

.954

ONB3

4.2/1.5

.962

.927

.843

.919

.963

(continued)

220

Appendix

Table A.11 (continued) Time point one Overall retailer loyalty

Offline channel loyalty

Online channel loyalty

ONB4

3.7/1.6

.882

.864

LOY1

4.7/2.1

.763

LOY2

2.9/1.6

.907

.829

LOY3

3.6/1.8

.854

.807

.791

.725

LOY4

2.4/1.6

.815

OFLOY1

4.7/2.1

.761

OFLOY2

2.9/1.7

.916

.839

OFLOY3

3.6/1.8

.858

.810

OFLOY4

2.4/1.6

.806

ONLOY1

3.0/1.8

.715

ONLOY2

2.5/1.5

.829

.854

ONLOY3

2.9/1.7

.758

.830

ONLOY4

2.2/1.5

.687

.782

.896

.749 .781

.723

.897

.742 .810

.807

.918

Time point two Construct

Item

MV/Std.

FL

KMO

ItTC

α

Offline retail brand equity

OFB1

4.0/1.6

.888

.818

.854

.942

OFB2

4.2/1.6

.949

.906

OFB3

4.2/1.5

.871

.841

OFB4

3.7/1.6

.882

ONB1

4.1/1.5

.916

ONB2

4.2/1.5

.957

.922

ONB3

4.2/1.6

.912

.887

ONB4

3.8/1.6

.881

.859

LOY1

4.5/2.0

.874

LOY2

3.0/1.6

.907

.837

LOY3

3.6/1.8

.886

.837

Online retail brand equity

Overall retailer loyalty

Offline channel loyalty

.851 .836

.790

.886

.715

LOY4

2.6/1.6

.840

OFLOY1

4.6/2.0

.753

OFLOY2

3.1/1.6

.929

.859

OFLOY3

3.6/1.8

.880

.834

OFLOY4

2.7/1.7

.841

.779

.954

.904

.775 .792

.721

.908

(continued)

Appendix

221

Table A.11 (continued) Time point two Online channel loyalty

ONLOY1

3.0/1.8

.831

.828

.792

.917

ONLOY2

2.5/1.5

.896

.843

ONLOY3

2.9/1.6

.895

.847

ONLOY4

2.3/1.4

.824

.778

Construct

Item

MV/Std.

FL

KMO

ItTC

α

Offline retail brand equity

OFB1

4.0/1.7

.907

.788

.875

.948

OFB2

4.1/1.6

.948

.910

OFB3

4.2/1.5

.908

.878

Time point three

Online retail brand equity

Overall retailer loyalty

Offline channel loyalty

Online channel loyalty

OFB4

3.8/1.6

.861

ONB1

4.2/1.6

.933

.836

ONB2

4.2/1.5

.946

.911

ONB3

4.1/1.5

.921

.892

ONB4

3.9/1.6

.858

.838

LOY1

4.5/2.0

.751

LOY2

3.0/1.6

.909

.846

LOY3

3.5/1.8

.909

.858

.855

.804

.901

.722

LOY4

2.7/1.6

.850

OFLOY1

4.5/2.0

.757

OFLOY2

3.1/1.6

.938

.876

OFLOY3

3.5/1.8

.914

.865

OFLOY4

2.8/1.7

.844

ONLOY1

2.9/1.8

.817

ONLOY2

2.5/1.4

.932

.883

ONLOY3

2.7/1.6

.932

.887

ONLOY4

2.3/1.5

.850

.808l

.953

.910

.789 .810

.729

.916

.789 .835

.789

.929

Notes: OFB = Offline retail brand equity, ONB = Online retail brand equity, LOY = Overall retailer loyalty, OFLOY = Offline channel loyalty, ONLOY = Online channel loyalty, MV/Std. = Mean values and Standard deviations, FL = Factor loadings (exploratory), KMO = Kaiser-Meyer-Olkin Criterion (≥.5), ItTC = Item-to-Total Correlation (≥.3), α = Cronbach’s alpha (≥.7). Source: Own creation

222

Appendix

Table A.12 Study 1: reliability and validity of alternative general models II Time point one Construct

Item

CR

λ

CR

λ

CR

Offline retail brand equity

OFB1

.944

.942

.944

.942

.944

Online retail brand equity

Overall retailer loyalty

Offline channel loyalty

Online channel loyalty

λ .940

OFB2

.950

.948

.949

OFB3

.864

.867

.863

OFB4

.837

.832

ONB1

.954

.933

.953

.934

.833 .954

.934

ONB2

.960

.961

.961

ONB3

.912

.910

.910

ONB4

.857

.849

.851

LOY1

.902

.773

LOY2

.868

LOY3

.876

LOY4

.809

OFLOY1

.901

.766

OFLOY2

.880

OFLOY3

.867

OFLOY4

.815

ONLOY1

.921

.839

ONLOY2

.905

ONLOY3

.866

ONLOY4

.840

Time point two Construct Offline retail brand equity

Online retail brand equity

Overall retailer loyalty

Item

CR

OFB1

.943

λ

CR

.916

.943

λ

CR

.917

.943

λ .913

OFB2

.950

.949

.952

OFB3

.863

.863

.863

OFB4

.859

.860

ONB1

.949

.937

.954

.937

.859 .954

.937

ONB2

.962

.963

.962

ONB3

.894

.895

.895

ONB4

.863

.864

.864

LOY1

.908

.747

LOY2

.888

LOY3

.886

LOY4

.848 (continued)

Appendix

223

Table A.12 (continued) Time point two Offline channel loyalty

Online channel loyalty

OFLOY1

.912

.739

OFLOY2

.915

OFLOY3

.872

OFLOY4

.864

ONLOY1

.919

.824

ONLOY2

.883

ONLOY3

.892

ONLOY4

.842

Time point three Construct

Item

CR

λ

CR

λ

CR

Offline retail brand equity

OFB1

.948

.931

.948

.931

.947

Overall retailer loyalty

Offline channel loyalty

Online channel loyalty

.929

OFB2

.960

.959

.961

OFB3

.888

.888

.888

OFB4 Online retail brand equity

λ

ONB1

.838 .953

.941

.838 .953

.941

.836 .953

.941

ONB2

.950

.950

.949

ONB3

.910

.911

.911

ONB4

.853

.853

.853

LOY1

.915

.754

LOY2

.896

LOY3

.903

LOY4

.857

OFLOY1

.920

.744

OFLOY2

.928

OFLOY3

.895

OFLOY4

.869

ONLOY1

.937

.816

ONLOY2

.924

ONLOY3

.922

ONLOY4

.961

Confirmatory model fits: Notes: OFB = Offline retail brand equity, ONB = Online retail brand equity, LOY = Overall retailer loyalty, OFLOY = Offline channel loyalty, ONLOY = Online channel loyalty, CR = Composite reliability (≥.6), λ = Standardized factor loadings (confirmatory) (≥.5), SCF = Scaling correction factor for MLM. Source: Own creation

Offline retail brand equity

Online retail brand equity

Overall retailer loyalty

Offline retail brand equity

Online retail brand equity

Offline channel loyalty

Offline retail brand equity

Online retail brand equity

1

2

3

1

2

3

1

2

Constructs

.626

.809

.602

.626

.812

.627

.626

.809

1

Time point one

.840

.458

.840

.477

.840

2

.702

.698

3

.669

.805

.602

.669

.805

.624

.669

.804

1

Time point two

Table A.13 Study 1: discriminant validity of alternative general models

.840

.430

.839

.491

.840

2

.730

.718

3

.656

.820

.584

.656

.819

.598

.656 a

.819

1

Time point three

.837

.425

.837

.460

.837

2

(continued)

.753

.737

3

224 Appendix

Online channel loyalty

Time point one

.406

1 .381

2 .748

3 .364

1

Time point two .381

2 .743

3 .368

1

Time point three .376

2

Source: Own creation

Confirmatory model fits: Time point one (Overall retailer loyalty): CFI .928, TLI .906, RMSEA .126, SRMR .043, χ2 (51) = 512.237, SCF = 1.30. Time point two Overall retailer loyalty): CFI .929, TLI .909, RMSEA .118, SRMR .043, χ2 (51) = 461.836, SCF = 1.47. Time point three (Overall retailer loyalty): CFI .934, TLI .915, RMSEA .115, SRMR .043, χ2 (51) = 439.065, SCF = 1.49. Time point one (Offline channel loyalty): CFI .925, TLI .903, RMSEA .118, SRMR.042, χ2 (51) = 528.337, SCF = 1.30. Time point two (Offline channel loyalty): CFI .931, TLI .911, RMSEA .118, SRMR. 046, χ2 (51) = 455.070, SCF = 1.46. Time point three (Offline channel loyalty): CFI .934, TLI .915, RMSEA .116, SRMR .043, χ2 (51) = 445.438, SCF = 1.48. Time point one (Online channel loyalty): CFI .941, TLI .924, RMSEA .112, SRMR .034, χ2 (51) = 414.711, SCF = 1.34. Time point two (Online channel loyalty): CFI .947, TLI .931, RMSEA .100, SRMR .031, χ2 (51) = 345.742, SCF = 1.49. Time point three (Online channel loyalty): CFI .948, TLI .932, RMSEA .101, SRMR .033, χ2 (51) = 351.207, SCF = 1.54. Notes: AVE = Average Variance Extracted (≥.5), SCF = Scaling correction factor for MLM, values in italics represent squared correlations between constructs, values in bold represent the AVE of the construct.

3

Constructs

Table A.13 (continued)

.783

3

Appendix 225

226

Appendix

Table A.14 Study 1: results of the single-factor test of alternative general models CFI

TLI

RMSEA

SRMR

χ2 (df)

 χ2 (df)

p-value of difference

1131.658 (3)

.000

924.782 (3)

.000

1059.019 (3)

.000

1168.055 (3)

.000

1002.742 (3)

.000

1131.355 (3)

.000

Overall retailer loyalty model Time point one Proposed model

.928

.906

.126

.043

512.237 (51)

Single factor model

.750

.695

.227

.081

1643.895 (54)

Time point two Proposed model

.929

.909

.118

.043

461.836 (51)

Single factor model

.770

.720

.207

.078

1386.618 (54)

Time point three Proposed model

.934

.915

.115

.043

439.065 (51)

Single factor model

.756

.702

.216

.083

1498.084 (54)

Offline channel loyalty model Time point one Proposed model

.925

.903

.118

.042

528.337 (51)

Single factor model

.743

.686

.230

.083

1696.392 (54)

Time point two Proposed model

.931

.911

.118

.046

455.070 (51)

Single factor model

.761

.708

.213

.082

1457.812 (54)

.915

.116

.043

445.438 (51)

Time point three Proposed model

.934

(continued)

Appendix

227

Table A.14 (continued)

Single factor model

CFI

TLI

RMSEA

SRMR

χ2 (df)

.745

.689

.222

.087

1576.793 (54)

 χ2 (df)

p-value of difference

1449.606 (3)

.000

1194.821 (3)

.000

1383.830 (3)

.000

Online channel loyalty model Time point one Proposed model

.941

.924

.112

.034

414.711 (51)

Single factor model

.709

.644

.241

.106

1846.317 (54)

Time point two Proposed model

.947

.931

.100

.031

345.742 (51)

Single factor model

.731

.672

.219

.110

1540.563 (54)

Time point three Proposed model

.948

.932

.101

.033

351.207 (51)

Single factor model

.707

.642

.233

.111

1735.037 (51)

Notes: Difference tests were conducted using χ2 tests of difference. Source: Own creation

580.954 555.835 580.978

Method-C

Method-U

Method-R

25.119 .024

Method-C with Method-U

Method-C with Method-R

601.005 596.334 571.819 596.362

Baseline

Method-C

Method-U

Method-R

Chi-square differences of model comparison tests

601.788

CFA

OFLOY

4.985

Baseline with Method-C

Models

χ2

585.939

Chi-square differences of model comparison tests

584.616

94

80

81

92

84

3

11

1

df

94

80

91

92

84

df

Time point 1

Baseline

χ2

CFA

LOY

Model

p

ns

ns

**

.930

.931

.930

.929

.928

.932

.933

.931

.931

.930

CFI

.922

.910

.919

.919

.910

.924

.913

.921

.921

.912

TLI

Table A.15 Study 1: results of the model comparisons (phase I) for alternative general models

.097

.104

.098

.098

.104

.095

.102

.097

.097

.102

RMSEA

.038

.038

.039

.051

.038

.040

.038

.040

.053

.039

SRMR

(continued)

1.22

1.25

1.22

1.21

1.20

1.22

1.25

1.22

1.22

1.21

SCF

228 Appendix

.028

Method-C with Method-R

487.931 465.622 488.962

Method-C

Method-U

Method-R

CFA

LOY

1

3

11

1

df

94

80

91

92

84

3

11

546.241

84

df

Time point 2

Method-C with Method-R χ2

1.031

Method-C with Method-U

Model

4.769 22.309

Baseline with Method-C

Models

χ2

492.700

Baseline

Chi-square differences of model comparison tests

486.244

CFA

ONLOY

4.671 24.515

df

Time point 1

Method-C with Method-U

χ2

Baseline with Method-C

Models

Table A.15 (continued)

.942

.943

.944

.943

.942

.933

CFI

ns

ns

**

p

ns

ns

**

p

.928

.916

TLI

.936

.927

.934

.934

.091

.098

RMSEA

.086

.092

.087

.087

.033

.038

SRMR

.035

.033

.036

.052

1.24

(continued)

1.30

SCF

1.24

1.28

1.25

1.25

Appendix 229

550.878

Method-R

23.755 1.495

Method-C with Method-U

Method-C with Method-R

538.381 515.076 539.132

Method-C

Method-U

Method-R

8.192 23.305

Baseline with Method-C

Method-C with Method-U

Models

χ2

546.573

Baseline

Chi-square differences of model comparison tests

537.396

CFA

OFLOY

8.084

Baseline with Method-C

Models

χ2

525.628

Method-U

Chi-square differences of model comparison tests

549.383

Method-C

11

1

df

94

80

91

92

84

3

11

1

df

94

80

91

92

Time point 2 557.467

Baseline

Table A.15 (continued)

ns

**

p

ns

ns

**

p

.936

.937

.936

.934

.935

.934

.935

.933

.932

.928

.918

.926

.925

.918

.926

.915

.923

.923

.091

.097

.093

.093

.097

.092

.099

.094

.094

.042

.040

.042

.056

.041

.041

.038

.041

.056

(continued)

1.29

1.33

1.30

1.29

1.29

1.30

1.34

1.30

1.30

230 Appendix

432.413 401.673 434.854

Method-C

Method-U

Method-R

527.253 533.363 530.063 506.636

CFA

Baseline

Method-C

Method-U

LOY

84

3

11

1

df

94

80

91

92

80

91

92

84

df

Time point 3

2.441

Method-C with Method-R χ2

30.74

Method-C with Method-U

Model

11.109

Baseline with Method-C

Models

χ2

443.522

Chi-square differences of model comparison tests

416.714

Baseline

3

Time point 2 .751

CFA

ONLOY

Method-C with Method-R

Table A.15 (continued)

.949

.948

.951

.948

.946

.940

.938

.938

.937

CFI

ns

ns

***

p

ns .937

.921

.928

.929

.922

TLI

.942

.936

.940

.939

.083

.097

.092

.092

.096

RMSEA

.080

.084

.081

.082

.030

.036

.038

.048

.037

SRMR

.042

.030

.042

.065

1.32

(continued)

1.34

1.31

1.31

1.31

SCF

1.32

1.36

1.33

1.32

Appendix 231

23.427 1.766

Method-C with Method-U

Method-C with Method-R

535.110 512.343 536.073

Method-C

Method-U

Method-R

4.040 22.767

Baseline with Method-C

Method-C with Method-U

Models

χ2

539.150

Baseline

Chi-square differences of model comparison tests

534.593

CFA

OFLOY

3.300

χ2

11

1

df

94

80

91

92

84

3

11

1

df

94

Time point 3 531.829

Baseline with Method-C

Models

Chi-square differences of model comparison tests

Method-R

Table A.15 (continued)

p

ns

*

p

ns

ns

**

.938

.940

.938

.938

.937

.938

.931

.921

.929

.929

.921

.931

.091

.097

.092

.092

.097

.090

.037

.036

.038

.048

.037

.038

(continued)

1.31

1.34

1.31

1.31

1.31

1.31

232 Appendix

440.129 417.370 442.961

Method-C

Method-U

Method-R

22.759 2.832

Method-C with Method-U

Method-C with Method-R

84

3

11

1

df

94

80

91

92

ns

ns

**

p

ns .949

.950

.951

.950

.949

.937

.944

.936

.942

.942

.085

.080

.086

.082

.082

.030

.032

.029

.033

.052

1.36

1.35

1.38

1.35

1.35

Source: Own creation

Notes: LOY = Overall retailer loyalty model, OFLOY = Offline channel loyalty model, ONLOY = Online channel loyalty model, SCF = Scaling correction factor for MLM, ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001.

5.792

Baseline with Method-C

Models

χ2

445.921

Chi-square differences of model comparison tests

434.555

Baseline

3

Time point 3 .963

CFA

ONLOY

Method-C with Method-R

Table A.15 (continued)

Appendix 233

234

Appendix

Table A.16 Study 1: results of the reliability and decomposition (phase II) for alternative general models Time point 1

Latent variable

Reliability baseline model

Decomposed reliability from method-C model

Total reliability

Substantive reliability

Method reliability

% reliability marker variable

Overall retailer loyalty model Offline retail brand equity

.997

.878

.115

12.04

Online retail brand equity

.998

.883

.115

11.52

Loyalty

.995

.860

.119

13.54

Offline channel loyalty model Offline RBE

.997

.876

.121

12.18

Online RBE

.997

.880

.117

11.77

Loyalty

.995

.858

.137

13.77

Online channel loyalty model Offline retail brand equity

.997

.879

.118

11.80

Online retail brand equity

.997

.884

.114

11.41

Loyalty

.995

.869

.126

12.67

Reliability baseline model

Decomposed reliability from method-C model

Total reliability

Substantive reliability

Method reliability

Time point 2

Latent variable

% reliability marker variable

Overall retailer loyalty model Offline retail brand equity

.997

.915

.082

8.21

Online retail brand equity

.997

.918

.079

7.91

Loyalty

.995

.904

.091

9.14

Offline channel loyalty model (continued)

Appendix

235

Table A.16 (continued) Time point 1 Reliability baseline model

Decomposed reliability from method-C model

Latent variable

Total reliability

Substantive reliability

Method reliability

Offline retail brand equity

.997

.918

.079

7.88

Online retail brand equity

.997

.921

.076

7.60

Loyalty

.995

.910

.086

8.64

% reliability marker variable

Online channel loyalty model Offline RBE

.996

.950

.046

4.62

Online RBE

.997

.952

.044

4.45

Loyalty

.995

.944

.050

5.06

Time point 3

Latent variable

Reliability baseline model

Decomposed reliability from method-C model

Total reliability

Substantive reliability

Method reliability

% reliability marker variable

Overall retailer loyalty model Offline retail brand equity

.997

.931

.061

6.10

Online retail brand equity

.997

.937

.060

6.02

Loyalty

.995

.928

.068

6.80

Offline channel loyalty model Offline retail brand equity

.997

.934

.063

6.36

Online retail brand equity

.997

.935

.062

6.24

Loyalty

.996

.927

.069

6.90

Online channel loyalty model Offline retail brand equity

.997

.867

.130

13.06

Online retail brand equity

.997

.906

.092

9.24

Loyalty

.996

.892

.104

10.49

Source: Own creation

236

Appendix

Table A.17 Study 1: results of the sensitivity analyses (phase III) for alternative general models Time point 1 Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

Overall retailer loyalty model OFB with LOY

.792

.791

.789

.850

.851

ONB with LOY

.691

.691

.687

.775

.777

OFB with ONB

.791

.791

.789

.848

.849

SELF with OFB

− .074

.000

.000

.000

.000

SELF with ONB

− .063

.000

.000

.000

.000

SELF with LOY

− .078

.000

.000

.000

.000

Offline channel loyalty model OFB with LOY

.776

.776

.774

.837

.838

ONB with LOY

.677

.677

.674

.763

.765

OFB with ONB

.791

.791

.789

.848

.849

SELF with OFB

− .074

.000

.000

.000

.000

SELF with ONB

− .063

.000

.000

.000

.000

SELF with LOY

− .061

.000

.000

.000

.000

Online channel loyalty model OFB with LOY

.637

.637

.632

.741

.743

ONB with LOY

.617

.617

.611

.725

.727

OFB with ONB

.791

.791

.789

.847

.848 (continued)

Appendix

237

Table A.17 (continued) Time point 1 Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

SELF with OFB

− .074

.000

.000

.000

.000

SELF with ONB

− .063

.000

.000

.000

.000

SELF with LOY

− .099

.000

.000

.000

.000

Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

Time point 2

Overall retailer loyalty model OFB with LOY

.790

.790

.787

.849

.851

ONB with LOY

.701

.701

.696

.788

.790

OFB with ONB

.818

.818

.817

.867

.868

SELF with OFB

− .040

.000

.000

.000

.000

SELF with ONB

− .065

.000

.000

.000

.000

SELF with LOY

− .091

.000

.000

.000

.000

Offline channel loyalty model OF with LOY

.775

.776

.773

.842

.839

ON with LOY

.655

.656

.650

.750

.756

OFF with ON

.818

.818

.817

.868

.868

SELF with OF

− .040

.000

.000

.000

.000

SELF with ON

− .065

.000

.000

.000

.000

SELF with LOY

− .085

.000

.000

.000

.000

Online channel loyalty model (continued)

238

Appendix

Table A.17 (continued) Time point 1 Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

OFB with LOY

.603

.603

.594

.719

.721

ONB with LOY

.617

.617

.605

.733

.735

OFB with ONB

.818

.818

.817

.863

.864

SELF with OFB

− .039

.000

.000

.000

.000

SELF with ONB

− .065

.000

.000

.000

.000

SELF with LOY

− .155

.000

.000

.000

.000

Method-C

Method-S (.05)

Method-S (.01)

Time point 3 Construct correlations

CFA

Baseline

Overall retailer loyalty model OFB with LOY

.774

.774

.773

.841

.843

ONB with LOY

.680

.680

.678

.779

.781

OFB with ONB

.810

.810

.809

.868

.869

SELF with OFB

− .043

.000

.000

.000

.000

SELF with ONB

− .069

.000

.000

.000

.000

SELF with LOY

− .046

.000

.000

.000

.000

Offline channel loyalty model OFB with LOY

.764

.764

.763

.834

.835

ONB with LOY

.654

.654

.652

.759

.761

OFB with ONB

.811

.811

.809

.868

.869 (continued)

Appendix

239

Table A.17 (continued) Time point 1 Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

SELF with OFB

− .043

.000

.000

.000

.000

SELF with ONB

− .069

.000

.000

.000

.000

SELF with LOY

− .043

.000

.000

.000

.000

Online channel loyalty model OFB with LOY

.607

.607

.603

.732

.735

ONB with LOY

.613

.613

.607

.739

.742

OFB with ONB

.811

.811

.809

.868

.869

SELF with OFB

− .043

.000

.000

.000

.000

SELF with ONB

− .069

.000

.000

.000

.000

SELF with LOY

− .100

.000

.000

.000

.000

Notes: LOY = Loyalty, OFB = Offline retail brand equity, ONB = Online retail brand equity, SELF = Self-efficacy. Source: Own creation Table A.18 Study 1: F-test of strong instrumental variables for alternative general models Overall retailer loyalty model

Offline channel loyalty model

Online channel loyalty model

F-value

F-value

F-value

IV1 → Offline retail brand equity

779.177

961.369

714.042

IV2 → Online retail brand equity

1436.975

824.761

1351.195

Notes: IV = Instrumental variable, F-value > 10 indicates strong predictor. Source: Own creation

.149

.062

.205

.058

Online RBE (1) → Offline RBE (2)

Offline RBE (1) → LOY (2)

Online RBE (1) → LOY (2)

**

***

***

***





Online IV (1) → Online RBE (1)

Offline RBE (1) → Online RBE (2)





Offline IV (1) → Offline RBE (1)

Direct effects

.049

.135

.060

.130

.835

.746

**

***

***

***

***

***

p





β

.030

.220

.059

.150

β

β

p

Consistent model Efficient model

ns

***

**

***





p

Model 2: Offline channel loyalty

Efficient model

Model 1: Overall retailer loyalty

β

.022

.178

.058

.132

.845

.772

ns

***

***

***

***

***

p





β

.065

.051

.064

.148

*

*

**

***





p

β

.069

.048

.058

.138

.873

.849

(continued)

**

**

***

***

***

***

p

Consistent model

Model 3: Online channel loyalty Consistent model Efficient model

Table A.19 Study 1: results of the efficient and consistent alternative general models

240 Appendix

.839

.725

.578

.158

.066

Offline RBE (1) → Offline RBE (2)

Online RBE (1) → Online RBE (2)

LOY (1) → LOY (2)

Offline RBE (2) → Online RBE (3)

Online RBE (2) → Offline RBE (3)

***

***

***

***

***

.063

.142

.693

.717

.807

***

***

***

***

***

p

β

.063

.160

.584

.723

.842

β

β

p

Consistent model Efficient model

**

***

***

***

***

p

Model 2: Offline channel loyalty

Efficient model

Model 1: Overall retailer loyalty

Table A.19 (continued)

β

.061

.145

.667

.716

.812

***

***

***

***

***

p

β

.068

.157

.722

.726

.836

Consistent model Efficient model

**

***

***

***

***

p

β

.061

.149

.743

.721

.823

(continued)

***

***

***

***

***

p

Consistent model

Model 3: Online channel loyalty

Appendix 241

.215

.062

.873

.784

.617

Offline RBE (2) → LOY (3)

Online RBE (2) → LOY (3)

Offline RBE (2) → Offline RBE (3)

Online RBE (2) → Online RBE (3)

LOY (2) → LOY (3)

***

***

***

**

***

.706

.785

.858

.049

.152

***

***

***

**

***

p

β

.631

.780

.876

.032

.230

β

β

p

Consistent model Efficient model

***

***

***

ns

***

p

Model 2: Offline channel loyalty

Efficient model

Model 1: Overall retailer loyalty

Table A.19 (continued)

β

.679

.781

.861

.024

.197

***

***

***

ns

***

p

β

.797

.782

.870

.071

.055

Consistent model Efficient model

*

*

***

***

***

p

β

.803

.782

.866

.079

.055

(continued)

***

***

***

*

***

p

Consistent model

Model 3: Online channel loyalty

242 Appendix

LOY (3)

Gender (1) → LOY (1)

.001

t = 7.379**

Diff. in total effects

Covariates

.095

Online RBE (1) → LOY (3)

ns

**

***

***

.316

.874

.002

t = 11.600**

.075

.224

.848

ns

**

***

***

p

.056

.337

.872

− .015 ns

† (.064)

***

***

p

t = 9.560**

β

β

p

β

Offline RBE (1) → LOY (3)

Total effects

R2

Consistent model Efficient model

Model 2: Offline channel loyalty

Efficient model

Model 1: Overall retailer loyalty

Table A.19 (continued)

.044

.284

.843

− .015

t = 22.678**

β

ns

ns

***

***

p

.107

.097

.792

.013

ns

**

*

***

p

t = 1.125 ns

β

Consistent model Efficient model

.115

.094

.765

.013

ns

**

**

***

p

(continued)

t = 1.566 ns

β

Consistent model

Model 3: Online channel loyalty

Appendix 243

.001

.001

.034

.033

.033

Gender (2) → LOY (2)

Gender (3) → LOY (3)

Age (1) → LOY (1)

Age (2) → LOY (2)

Age (3) → LOY (3)

*

*

*

ns

ns

.036

.034

.033

.002

.002

p

*

*

*

ns

ns

.012

.012

ns

ns

ns

ns

− .016

.012

ns

p

− .016

β

β

β

p

Consistent model Efficient model

Model 2: Offline channel loyalty

Efficient model

Model 1: Overall retailer loyalty

Table A.19 (continued)

.015

.014

.014

− .017

− .015

β p

ns

ns

ns

ns

ns

ns

ns

ns

− .002

− .002

− .002

ns

ns

.012

p

.014

β

Consistent model Efficient model

.015

.013

− .001

− .001

− .001

β

ns

ns

ns

ns

ns

(continued)

p

Consistent model

Model 3: Online channel loyalty

244 Appendix

.017

.016

.016

.256

Internet expertise (1) → LOY (1)

Internet expertise (2) → LOY (2)

Internet expertise (3) → LOY (3)

Familiarity (1) → LOY (1)

***

ns

ns

ns

.227

.025

.024

.024

*

*

*

***

p

ns

− .002

***

ns

− .002

.260

ns

p

− .002

β

β

β

p

Consistent model Efficient model

Model 2: Offline channel loyalty

Efficient model

Model 1: Overall retailer loyalty

Table A.19 (continued)

β

.242

.006

.005

.005

ns

ns

ns

***

p

β

.072

.037

.034

.037

Consistent model Efficient model

*

*

*

***

p

β

.080

.040

.037

.038

**

**

**

(continued)

***

p

Consistent model

Model 3: Online channel loyalty

Appendix 245

.245

Familiarity (3) → LOY (3)

***

***

.237

.219

***

***

β

.243

.231

***

***

p

β

.250

.232

***

***

p

β

.073

.065

Consistent model Efficient model

***

***

p

β

.084

.075

***

***

p

Consistent model

Model 3: Online channel loyalty

Source: Own creation

Structural model fits: Model 1: Efficient model: CFI .918, TLI .914, RMSEA .070, SRMR .062, χ2 (1009) = 3847.824, SCF = .90. Consistent model: CFI .895, TLI .890, RMSEA .071, SRMR .184, χ2 (1410) = 5495.237, SCF = .89. Model 2: Efficient model: CFI .916, TLI .912, RMSEA .061, SRMR .070, χ2 (1009) = 3856.525, SCF = .91. Consistent model: CFI .893, TLI .888, RMSEA .072, SRMR .182, χ2 (1410) = 5505.123, SCF = .90. Model 3: Efficient model: CFI .930, TLI .926, RMSEA .063, SRMR .052, χ2 (1009) = 3270.553, SCF = .91. Consistent model: CFI .911, TLI .907, RMSEA .064, SRMR .153, χ2 (1410) = 4662.513, SCF = .89. Notes: IV = Instrumental variable, RBE = Retail brand equity, LOY = Loyalty, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 573, Standardized coefficients are shown. Differences between total effects have been tested using t-tests. ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001.

.231

Familiarity (2) → LOY (2)

p

β

β

p

Consistent model Efficient model

Model 2: Offline channel loyalty

Efficient model

Model 1: Overall retailer loyalty

Table A.19 (continued)

246 Appendix

Appendix

247

Table A.20 Study 1: measurement invariance across time points of alternative general models Model

χ2 /df (p-value)

χ2 -Difference (p-value)

CFI (CFI)

TLI (TLI)

RMSEA (RMSEA)

SCF

.060

1.26

.060

1.25.

.059

1.25.

Overall retailer loyalty model Model 1: Configural invariance Model 2: Full metric invariance Model 3: Partial metric invariancea

1592.531/522

.956

(.000) 1632.544/540 (.000) 1626.427/539 (.000)

(−) 31.708 (.024)

.955 (.001)

24.574 (.105)

.947 (−) .947 (−)

.955 (.001)

(−)

(−) .947

(−)

(.001)

Offline channel loyalty model Model 1: Configural invariance Model 2: Full metric invariance Model 3: Partial metric invarianceb

1640.451/522

.950

.939

.061

1.23

.061

1.22

.061

1.22

(.000) 1678.392/540 (.000) 1672.844/539 (.000)

29.772 (.040)

.949

.940

(.001)

(.001)

23.275 (.141)

(−)

.949

.941

(.001)

(.002)

.955

.946

.057

1.22

.955

.947

.056

1.20

(.001)

(.001)

(−)

Online channel loyalty model Model 1: Configural invariance Model 2: Full metric invariance

1486.909/522 (.000) 1514.486/540 (.000)

18.628 (.415)

(−)

Notes: SCF = Scaling correction factor for MLM. a Factor loading freed for the following item: LOY1 time point one. b Factor loading freed for the following item: LOY4 time point one. Source: Own creation

ns

***

***

***

−.006

Online RBE (1) → LOY (2)

Offline .795 RBE (1) → Offline RBE (2)

Online .726 RBE (1) → Online RBE (2)

LOY (1) → LOY (2)

.555

***

.234

Offline RBE (1) → LOY (2)

***

.067

Online RBE (1) → Offline RBE (2)

***

.114

.441

.681

.871

.230

.268

.064

.223

***

***

***

***

***

***

***

p

β

β

p

More favorable

Less favorable

Model 1: More vs. less favorable prior offline experiences (Overall retailer loyalty)

Offline RBE (1) → Online RBE (2)

Direct effects

Prior Offline Experiences

ns

ns

*

*

ns

*

***

p

β

.132

.728

.758

.121

.518

.122

.126

***

***

***

*

***

***

***

p

Diff. Less favorable test β

.151

.749

.821

.188

.551

.118

.144

***

***

***

***

***

***

***

p

ns

ns

*

*

ns

ns

ns

p

β

.584

.789

.563

.198

.023

.331

.058

**

***

***

***

***

ns

***

p

β

.648

.826

.615

.185

.043

.326

.068

p

*

*

(continued)

***

*** ns

***

** ns

ns ns

*** ns

*** ns

p

More favorable Diff. test

Model 3: More vs. less favorable prior offline experiences (Online channel loyalty)

More favorable Diff. Less favorable test

Model 2: More vs. less favorable prior offline experiences (Offline channel loyalty)

Table A.21 Study 1: results of alternative moderator models

2.7. Alternative Cross-sectional Models

248 Appendix

***

***

***

Offline .849 RBE (2) → Offline RBE (3)

Online .744 RBE (2) → Online RBE (3)

LOY (2) → LOY (3)

Total effects

R2 LOY (3)

ns

−.011

Online RBE (2) → LOY (3)

.848

.630

***

.252

Offline RBE (2) → LOY (3)

***

***

.070

Online RBE (2) → Offline RBE (3)

***

.193

.908

.592

.824

.892

.077

.268

.066

.128

***

***

***

***

**

***

***

***

p

β

β

p

More favorable

Less favorable

Model 1: More vs. less favorable prior offline experiences (Overall retailer loyalty)

Offline RBE (2) → Online RBE (3)

Prior Offline Experiences

Table A.21 (continued)

p

ns

*

ns

*

*

ns

*

β

.844

.137

.788

.807

.005

.685

.128

.137

***

***

***

***

ns

***

**

***

p

Diff. Less favorable test β

.901

.134

.807

.836

.138

.642

.125

.149

***

***

***

***

***

***

***

***

p

ns

ns

ns

*

ns

ns

ns

p

β

.681

.415

.858

.595

.360

.120

.348

.064

**

***

***

***

***

***

† (.071)

***

p

β

.633

.410

.880

.627

.355

.107

.345

.069

p ** ns

(continued)

***

*** ns

*** ns

*** ns

*** ns

* ns

*** ns

p

More favorable Diff. test

Model 3: More vs. less favorable prior offline experiences (Online channel loyalty)

More favorable Diff. Less favorable test

Model 2: More vs. less favorable prior offline experiences (Offline channel loyalty)

Appendix 249

**

**

−.024

−.025

.012

.011

.011

Gender (2) → LOY (2)

Gender (3) → LOY (3)

Age (1) → LOY (1)

Age (2) → LOY (2)

Age (3) → LOY (3)

ns

ns

ns

**

−.027

***

Gender (1) → LOY (1)

Covariates

.185

Online RBE (1) → LOY (3)

.012

.012

.011

−.025

−.025

−.024

.193

.649

ns

ns

ns

**

**

**

***

***

p

β

***

p

β

.532

More favorable

Less favorable

Model 1: More vs. less favorable prior offline experiences (Overall retailer loyalty)

Offline RBE (1) → LOY (3)

Prior Offline Experiences

Table A.21 (continued)

p

*

**

β

.005

.005

.005

.007

.006

.007

.104

.590

ns

ns

ns

ns

ns

ns

**

***

p

Diff. Less favorable test β

.005

.005

.005

.006

.006

.006

.205

.621

p *

ns

ns

ns

ns

ns

ns

*** **

***

p

*

*

*

−.032

−.034

ns

ns

**

**

−.033

p

ns

.005

.005

.406

.098

.006

β

.005

.006

.005

.404

.107

−.038

−.035

−.034

β

p * ns

(continued)

*

*

*

ns

ns

ns

*** ns

p

More favorable Diff. test

Model 3: More vs. less favorable prior offline experiences (Online channel loyalty)

More favorable Diff. Less favorable test

Model 2: More vs. less favorable prior offline experiences (Offline channel loyalty)

250 Appendix

† (.070) −.014

−.013

.278

.232

.244

Internet expertise (3) → LOY (3)

Familiarity (1) → LOY (1)

Familiarity (2) → LOY (2)

Familiarity (3) → LOY (3)

***

***

.201

.196

.194

† (.069) −.013

−.013

Internet expertise (2) → LOY (2)

***

† (.069) −.013

−.015

***

***

***

† (.070)

† (.069)

† (.068)

p

β

p

β

More favorable

Less favorable

Model 1: More vs. less favorable prior offline experiences (Overall retailer loyalty)

Internet expertise (1) → LOY (1)

Prior Offline Experiences

Table A.21 (continued)

p

.386

.365

.431

−.017

−.017

−.019

β

β

***

***

.310

.302

.303

ns −.017

ns −.018

ns

ns

ns

***

***

***

p

p

β

.104

.093

.106

.007

.006

.007

ns

ns

ns

***

***

***

p

β

.088

.080

.080

.007

.007

.006

ns

ns

ns

p

(continued)

***

***

***

p

More favorable Diff. test

Model 3: More vs. less favorable prior offline experiences (Online channel loyalty)

More favorable Diff. Less favorable test ns −.017

***

p

Diff. Less favorable test

Model 2: More vs. less favorable prior offline experiences (Offline channel loyalty)

Appendix 251

.706

.118

LOY (1) → LOY (2)

.134

Online RBE (1) → LOY (2)

Online RBE (1) → Online RBE (2)

.533

Offline RBE (1) → LOY (2)

.769

.108

Online RBE (1) → Offline RBE (2)

Offline RBE (1) → Offline RBE (2)

.115

**

***

***

**

***

**

***

.120

.726

.766

.152

.531

.116

.107

***

***

***

***

***

***

***

p

ns

ns

ns

ns

ns

ns

ns

p

β

.163

.713

.747

.137

.475

.130

.104

***

***

***

**

***

***

***

p

β

p

β

.159

.736

.750

.162

.471

.144

.098

***

***

***

***

***

***

***

p

ns

ns

ns

ns

ns

ns

ns

p

Diff. test

β

More favorable

Less favorable

More favorable

Less favorable

Diff. test

Model 5: More vs. less favorable prior online experiences (Offline channel loyalty)

Model 4: More vs. less favorable prior online experiences (Overall retailer loyalty)

Offline RBE (1) → Online RBE (2)

Direct effects

Prior Online Experiences

Table A.21 (continued)

β

.576

.708

.651

.166

.097

.106

.105

Less favorable

***

***

***

***

***

***

***

p

β

.485

.669

.648

.216

.116

.219

.214

***

***

***

***

***

***

***

p

More favorable

(continued)

ns

ns

ns

ns

ns

ns

ns

p

Diff. test

Model 6: More vs. less favorable prior online experiences (Online channel loyalty)

252 Appendix

Offline RBE (1) → LOY (3)

Total effects

.499

.820

.127

LOY (2) → LOY (3)

R2 LOY (3)

.792

Online RBE (2) → Online RBE (3)

.142

Online RBE (2) → LOY (3)

.791

.538

Offline RBE (2) → LOY (3)

Offline RBE (2) → Offline RBE (3)

.118

Online RBE (2) → Offline RBE (3)

***

***

**

***

***

**

***

**

***

p

.531

.872

.131

.809

.816

.163

.577

.130

.116

***

***

**

***

***

**

***

**

***

p

**

ns

ns

ns

ns

*

ns

ns

p

β

.504

.823

.160

.800

.768

.067

.563

.142

.110

***

***

***

***

***

ns

***

***

***

p

β

.121

β

.566

.861

.153

.817

.791

.047

.653

.157

.105

***

***

***

***

***

ns

***

***

***

p

**

ns

ns

ns

ns

ns

ns

ns

p

Diff. test

β

More favorable

Less favorable

More favorable

Less favorable

Diff. test

Model 5: More vs. less favorable prior online experiences (Offline channel loyalty)

Model 4: More vs. less favorable prior online experiences (Overall retailer loyalty)

Offline RBE (2) → Online RBE (3)

Prior Online Experiences

Table A.21 (continued)

β

.138

.521

.470

.797

.565

.206

.108

.350

.106

Less favorable

***

***

***

***

***

***

***

***

***

p

β

.204

.719

.481

.828

.565

.290

.133

.402

.096

***

***

***

***

***

***

***

***

***

p

More favorable

(continued)

**

ns

ns

ns

ns

ns

ns

p

Diff. test

Model 6: More vs. less favorable prior online experiences (Online channel loyalty)

Appendix 253

ns

ns

ns

ns

−.009

−.009

.020

.022

.022

−.019

Gender (2) → LOY (2)

Gender (3) → LOY (3)

Age (1) → LOY (1)

Age (2) → LOY (2)

Age (3) → LOY (3)

Internet expertise (1) → LOY (1)

ns

ns

ns

−.008

***

p

−.023

.019

.019

.019

−.008

−.008

−.008

.206

ns

ns

ns

ns

ns

ns

ns

***

p **

p .143

ns

−.006

ns

ns

−.006

−.029

ns

−.006

*

*

ns

*

p

−.017

−.016

−.016

β

β

.175

.153

−.031

−.006

−.006

−.006

−.015

−.015

−.015

β

p

ns

ns

ns

ns

*

*

*

ns

ns

p

Diff. test

β

More favorable

Less favorable

More favorable

Less favorable

Diff. test

Model 5: More vs. less favorable prior online experiences (Offline channel loyalty)

Model 4: More vs. less favorable prior online experiences (Overall retailer loyalty)

Gender (1) → LOY (1)

Covariates

Online RBE (1) → LOY (3)

Prior Online Experiences

Table A.21 (continued)

.246

.003

−.038

−.034

−.031

−.021

−.018

−.017

β

Less favorable

ns

*

*

*

ns

ns

ns

***

p .327

.004

−.040

−.034

−.037

−.022

−.019

−.021

β

**

p

Diff. test

(continued)

ns

*

*

*

ns

ns

ns

***

p

More favorable

Model 6: More vs. less favorable prior online experiences (Online channel loyalty)

254 Appendix

.428

.400

.437

Familiarity (1) → LOY (1)

Familiarity (2) → LOY (2)

Familiarity (3) → LOY (3)

.390

.382

.367

−.022

ns

ns

***

***

***

p

p

.449

.399

.446

−.031

−.029

β

ns

ns

***

***

***

p

.371

.378

.362

−.029

−.029

β

ns

ns

***

***

***

p

p

β

.136

.110

.117

.004

.003

Less favorable

ns

ns

***

***

***

p

β

.134

.117

.123

.004

.003

ns

ns

***

***

***

p

More favorable

p

Diff. test

Model 6: More vs. less favorable prior online experiences (Online channel loyalty)

Source: Own creation

Structural model fit: Model 1: CFI .898, TLI .894, RMSEA .087, SRMR .104, χ2 (1975) = 6251.789, SCF = .75. Model 2: CFI .923, TLI .919, RMSEA .078, SRMR .100, χ2 (1900) = 4086.386, SCF = .79. Model 3: CFI .903, TLI .899, RMSEA .080, SRMR .111, χ2 (1975) = 5783.164, SCF = .76. Model 4: CFI .888, TLI .884, RMSEA .085, SRMR .109, χ2 (1968) = 6031.754, SCF = .76. Model 5: CFI .886, TLI .881, RMSEA .085, SRMR .108, χ2 (1968) = 6025.829, SCF = .76. Model 6: CFI .893, TLI .888, RMSEA .081, SRMR .108, χ2 (1968) = 5652.805, SCF = .77. Notes: RBE = Retail brand equity, LOY = Loyalty, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, Standardized coefficients are shown. N-Offline (284 less favorable, 285 more favorable), N-Online (285 less favorable, 283 more favorable). Differences in effects between groups have been tested using χ2 tests of difference. ns = not significant; † p < .10; *p < .05; **p < .01; ***p < .001.

***

***

***

ns

−.021

Internet expertise (3) → LOY (3)

−.022

β

ns

p

−.021

Diff. test

β

More favorable

Less favorable

More favorable

Less favorable

Diff. test

Model 5: More vs. less favorable prior online experiences (Offline channel loyalty)

Model 4: More vs. less favorable prior online experiences (Overall retailer loyalty)

Internet expertise (2) → LOY (2)

Prior Online Experiences

Table A.21 (continued)

Appendix 255

256

Appendix

Table A.22 Study 2: F-test of strong instrumental variables Overall retailer loyalty model

Offline channel loyalty model

Online channel loyalty model

F-value

F-value

F-value

238.995

243.950

224.434

Online Trust → Online 178.770 channel attributes

181.166

167.786

Offline Trust → Offline channel attributes

Notes: F-value > 10 indicates strong predictor. Source: Own creation

.068

.740

.813

→ Offline CI

→ LOY

→ LOY

→ Offline CI

→ Online CI

Online CI

Offline CI

Online CI

Offline CA

Online CA

Offline CI

Total effects

→ LOY

.599

→ Online CI

Offline CI

.612

.546

.065



→ Offline CI

Online Trust

R2 loyalty

.214



→ Online CI

Offline Trust

Direct effects

***

***

***

***

ns

***

ns

*





.673

.594

.618

.655

.051

.561

.070

.341

.593

.749

***

***

***

***

ns

***

ns

***

***

***

p





β

.628

.546

.813

.740

.051

.616

.064

.215

***

***

***

***

ns

***

ns

**





p

β

β p

Consistent model

Efficient model

Consistent model β

.684

.596

.618

.656

.034

.577

.071

.340

.593

.750

Efficient model

Model 2: Offline channel loyalty

Model 1: Overall retailer loyalty

Table A.23 Study 2: results of the efficient and consistent models

***

***

***

***

ns

***

ns

***

***

***

p





β

.330

.418

.811

.726

.284

.306

.068

.226

Consistent model

***

***

***

***

***

***

ns

*





p

β

***

***

***

***

***

***

ns

***

***

***

p

(continued)

.440

.457

.614

.634

.264

.283

.081

.349

.594

.743

Efficient model

Model 3: Online channel loyalty

Appendix 257

.200

.327

.034

.304

− .032

− .067

***

ns

*

ns

***

t = 9.686**

.186

.327

− .034

− .074

.017

t = 7.746**

β

ns

*

ns

*

***

p .316

.015

.305

− .031

− .071

***

ns

*

ns

***

p

t = 10.302**

β

.075

.359

.266

.012

− .116

β

Consistent model

***

ns

***

*

***

p

.072

.450

.254

.012

− .141

β

Efficient model

Model 3: Online channel loyalty

***

ns

***

Source: Own creation

*

***

p

Structural model fits: Model 1: Consistent model: CFI .918, TLI .914, RMSEA .070, SRMR .062, χ2 (1009) = 3847.824, SCF = .90. Efficient model: CFI .912, TLI .899, RMSEA .086, SRMR .106, χ 2 (248) = 945.579, SCF = 1.00. Model 2: Consistent model: CFI .916, TLI .912, RMSEA .061, SRMR .070, χ2 (1009) = 3856.525, SCF = .91. Efficient model: CFI .912, TLI .899, RMSEA .086, SRMR .107, χ 2 (248) = 943.997, SCF = 1.00. Model 3: Consistent model: CFI .930, TLI .926, RMSEA .063, SRMR .052, χ2 (1009) = 3270.553, SCF = .91. Efficient model: CFI .917, TLI .904, RMSEA .083, SRMR .105, χ 2 (248) = 894.901, SCF = .98. Notes: CI = Channel image, CA = Channel attributes, LOY = Loyalty, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 573, Standardized coefficients are shown. Differences between total effects have been tested using t-tests. ns = not significant; † p < .10; * p < .05; ** p < .01; *** p < .001.

***

ns

− .035

Internet expertise

.326

*

Familiarity

ns

**

.037

t = 7.746**

− .070

→ LOY

p

β

β p

Consistent model

Efficient model

Consistent model

Efficient model

Model 2: Offline channel loyalty

Model 1: Overall retailer loyalty

Age

Gender

Covariates

Diff. in total effects

Online CI

Table A.23 (continued)

258 Appendix

Appendix

259

3. Study 2: Effects of Perceived Offline-Online and Online-Offline Channel Integration Services 3.1. Common Method Variance CMV was adressed because the data were collected based on respondents’ selfreported perceptions. In this regard, CMV was addressed a priori by using an appropriate questionnaire design. First, the respondents were told that the study was anonymous and confidential and that there were no right or wrong answers. Moreover, the study started with the measures of the dependent variables (Chang, Van Witteloostuijn and Eden 2010). A posteriori CMV was accounted for by calculating a single-factor test using confirmatory factor analysis. The results show that the models with all items loading on a single factor had a significantly worse fit than the proposed models did (see Table A.24). Table A.24 Results of the single-factor tests CFI

TLI

RMSEA SRMR χ2 (df)

 χ2 (df)

p-value of difference

Offline purchase intention Proposed .957 .944 .053 model

.049

Single factor model

.160

.454 .395 .175

345.046 (114) 2834.239 (24) .000 3179.285 (138)

Online purchase intention Proposed .960 .947 .053 model

.050

Single factor model

.161

.481 .425 .176

345.222 (114) 2868.781 (24) .000 3214.003 (138)

Notes: Difference tests were conducted using χ2 tests of difference. Source: Own creation

Table A.25, A.26 and A.27 show the results for the marker variable technique (Lindell and Whitney 2001) following the latent variable approach of Williams, Hartman and Cavazotte (2010). Self-efficacy was used as a marker variable. First, it is an ideal marker because it is theoretically unrelated to the studies’ constructs. Second, it is similar to the studies’ constructs in content and format (e.g., the same Likert-type scale). Thus, it might be equivalently vulnerable to the same causes of CMV (Simmering et al. 2015). CMV was tested for by using a confirmatory

260

Appendix

factor analysis model, because it is preferable to a full structural equation model, as it is least restrictive in terms of latent variable relations (no risk of path model misspecification, Williams and McGonagle 2016). These technique consists of three consecutive phases. The first phase helps to ensure that the correlations between the latent constructs are not biased through the marker variable (phase I, Method-U vs. -R). The results of the following reliability decomposition (phase II) indicates that the amount of method variance, associated with the measurement of the substantive latent constructs, is less than 16.80 % (between 9.93 % and 16.80 %). Since previous literature (e.g., Williams, Hartman and Cavazotte 2010) found impacts up to 19.7 %, the bias in observed relations is reduced. The results of the sensitivity analysis (phase III) support that the marker-based method variance has a low effect on constructs. Table A.25 Results of the model comparisons (phase I) Offline purchase intention

χ2

df

TLI

RMSEA

SRMR

CFA

334.462

120

.964

.954

.050

.040

Baseline

352.517

130

.962

.956

.049

.047

Method-C

343.512

129

.964

.957

.048

.043

Method-U

322.267

115

.965

.953

.050

.040

Method-R

342.370

125

.964

.956

.048

.040

CFI

Chi-square differences of model comparison tests Models Baseline with Method-C

χ2

df

p

9.005

1

***

Method-C with Method-U

21.245

15

ns

Method-C with Method-R

1.142

5

ns

Online purchase intention

χ2

df

TLI

RMSEA

SRMR

CFA

362.343

120

.967

.958

.049

.040

Baseline

339.894

130

.966

.960

.047

.046

Method-C

331.292

129

.968

.961

.047

.042

Method-U

311.732

115

.968

.958

.049

.039

Method-R

311.843

125

.970

.963

.047

.040

CFI

Chi-square differences of model comparison tests Models Baseline with Method-C Method-C with Method-U

χ2

df

p

8.602

1

***

19.560

15

ns (continued)

Appendix

261

Table A.25 (continued) Offline purchase intention

χ2

Method-C with Method-R

df

9.449

CFI 5

TLI

RMSEA

SRMR

ns

* p < .05; ** p < .01; *** p < .001; ns = not significant. Source: Own creation

Table A.26 Results of the reliability decomposition (phase II)

Latent variable

Reliability baseline model

Decomposed reliability from method-U model

Total reliability

Substantive reliability

Method reliability

% reliability marker variable

Offline purchase intention Perceived offline-to-online services

.989

.850

.141

14.22

Perceived online-to-offline services

.988

.844

.146

14.77

Perceived quality of offline offerings

.974

.820

.153

15.72

Perceived quality of online offerings

.987

.835

.153

15.53

Offline purchase intention

.997

.887

.110

11.01

Online purchase intention Perceived offline-to-online services

.989

.859

.132

13.37

Perceived online-to-offline services

.988

.853

.137

13.91

Perceived quality of offline offerings

.973

.813

.163

16.80

Perceived quality of online offerings

.987

.844

.145

14.69

Online purchase intention

.998

.899

.099

9.93

Source: Own creation

262

Appendix

Table A.27 Results of the sensitivity analyses (phase III) Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

Offline purchase intention OF-ON with ON-OF

.472

.472

.472

.478

.479

OF-ON with OFO

.203

.203

.197

.205

.206

OF-ON with ONO

.254

.254

.250

.254

.253

OF-ON with OFPI

.119

.119

.116

.120

.121

ON-OF with OFO

.145

.144

.145

.153

.149

ON-OF with ONO

.195

.195

.197

.203

.201

ON-OF with OFPI

.131

.131

.133

.131

.135

OFO with ONO

.694

.694

.690

.701

.702

OFO with OFPI

.556

.556

.553

.584

.583

ONO with OFPI

.403

.403

.400

.413

.410

SELF with OF-ON

-.034

.000

.000

.000

.000

SELF with ON-OF

.034

.000

.000

.000

.000

SELF with OFO

−.124

.000

.000

.000

.000

SELF with ONO

−.065

.000

.000

.000

.000

SELF with OFPI

−.041

.000

.000

.000

.000

Online purchase intention OF-ON with ON-OF

.472

.472

.472

.479

.481

OF-ON with OFO

.204

.204

.198

.209

.210 (continued)

Appendix

263

Table A.27 (continued) Construct correlations

CFA

Baseline

Method-C

Method-S (.05)

Method-S (.01)

OF-ON with ONO

.255

.255

.252

.256

.255

OF-ON with ONPI

.175

.175

.172

.181

.181

ON-OF with OFO

.137

.136

.137

.147

.143

ON-OF with ONO

.194

.194

.197

.195

.203

ON-OF with ONPI

.168

.168

.169

.171

.178

OFO with ONO

.696

.696

.692

OFO with ONPI

.331

.331

.326

.332

.327

ONO with ONPI

.526

.526

.524

.529

.528

SELF with OF-ON

−.034

.000

.000

.000

.000

SELF with ON-OF

.034

.000

.000

.000

.000

SELF with OFO

−.121

.000

.000

.000

.000

SELF with ONO

−.065

.000

.000

.000

.000

SELF with ONPI

−.044

.000

.000

.000

.000

712

.701

Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, OFPI = Offline purchase intention, ONPI = Online purchase intention, SELF = Self-efficacy. Source: Own creation

264

Appendix

3.2. Endogeneity Test Endogeneity was tested by using the IV approach. Offline store accessibility and online service quality are included as IVs for integration services. In a first step, whether the two instruments are strong predictors of OF-ON services and ON-OF services using F-tests was ensured (see Table A.28). As the calculated Fvalues exceeded the recommended threshold of 10, the IVs can be considered to be strong predictors. Additionally to the efficient (proposed) models consistent models which included the IVs were estimated (Antonakis et al. 2010, see Table 29). Whether a change in path estimates emerges was tested (Hausman 1978). The respective z-values were below the critical value of 1.96. The probability of endogeneity seems to be reduced (Table A.29, A.30).

Table A.28 F-test of strong instrumental variables Model 1

Model 2

F-value

F-value

IV1 → Perceived offline-to-online services

49.072

49.945

IV2 → Perceived online-to-offline services

14.484

14.597

Notes: IV = Instrumental variable, F-value > 10 indicates strong predictor. Source: Own creation

3.3. Test for Measurement Invariance Measurement equivalence as tested to ensure comparability across the objective congruence groups. First, configural invariance by estimating a baseline model in which the factor loadings and intercepts are freely estimated was ensured. Second, a test for metric invariance by fixing the factor loadings of each item was accomplished. A comparison of the configural and the metric model shows that all deviations are within limits. We relied on differences in the comparative fit indices (following Chen 2007) to ensure measurement invariance. Full metric invariance was ascertained. The results indicate a good fit for all models and thus support our proposition that metric invariance holds for all latent constructs (Table A.31).

.186 .074 .118 .528 .375 − .076 − .001

→ ONO

→ OFO

→ ONO

→ OFPI

→ OFPI

→ OFPI

→ OFPI

OF-ON

ON-OF

ON-OF

OFO

ONO

OF-ON

ON-OF

→ OFPI

→ OFPI

→ OFO

→ ONO

ON-OF

OF-ON

Total effects

→ OFPI

→ OFPI

→ ONO

OF-ON

ON-OF

→ OFPI

→ OFO

OF-ON

.200

.059

.051

.092

.108

.155

→ OFO

OF-ON

Indirect effects



→ ON-OF

IV2

.708

.867

.053

.038

.095

.091

.146

.037

.033

.067

.079

.001

− .073



→ OF-ON

IV1

Direct effects

**

**

*

ns

**

*

ns

ns

***

***

**

ns

***

t = 4.527**

t = .271 ns

t = .519 ns

.229

.056

.050

.106

.123

.004

− .093

.373

.528

.115

.072

.225

.185

.085

.291

.709

.870

.050

.036

.114

.108

.096

.324

.175

.036

.032

.081

.094

.004

− .089

b

β

Diff. test

Consistent model p

β b

Proposed / efficient model

Table A.29 Results of the efficient and consistent models: offline purchase intention

***

*

ns

***

**

ns

ns

***

***

**

ns

***

**

*

***

p

(continued)

t = 4.703**

t = .364 ns

t = .595 ns

Diff. test

Appendix 265

→ OFPI

.014

.013

.225 − .004

.085

.070

− .049

.110

ns

ns

*

*

.014

− .049

.084

.106

.012

− .004

.224

.067

ns

ns

*

*

Source: Own creation

Structural model fits: Proposed / efficient model: CFI .957; TLI .944; RMSEA .053; SRMR .049; χ2 (114) = 345.046. Consistent model: CFI .895; TLI .868; RMSEA .076; SRMR .104; χ2 (143) = 742.658. Notes: IV = Instrumental variable, OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, OFPI = Offline purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests. ***p < .001; **p < .01; *p < .05; ns = not significant.

General Internet expertise

Age

Gender

Covariates

ON-OF

Table A.29 (continued)

266 Appendix

.059 .107 .318 .495 − .021 .009

→ ONO

→ OFO

→ ONO

→ ONPI

→ ONPI

→ ONPI

→ ONPI

OF-ON

ON-OF

ON-OF

OFO

ONO

OF-ON

ON-OF

→ ONPI

→ ONPI

→ OFO

→ ONO

ON-OF

ON-OF

OF-ON

→ ONPI

→ ONPI

→ ONO

OF-ON

Total effects

→ ONPI

→ OFO

OF-ON

.190

.068

.024

.124

.065

.159 .194

→ OFO

OF-ON

Indirect effects



→ ON-OF

IV2

.681

.047

.028

.097

.088

.174

.055

.019

.114

.060

.010

− .025

1.169



→ OF-ON

IV1

Direct effects

**

**

*

ns

***

*

ns

ns

***

***

*

ns

***

t = 4.233**

t = 2.300*

t = 2.079*

.214

.066

.023

.141

.073

.012

− .024

.491

.315

.105

.058

.234

.189

.084

.293

β

Diff. test

Consistent model

p

β b

Proposed / efficient model

Table A.30 Results of the efficient and consistent models: online purchase intention

1.164

.674

.045

.028

.116

.104

.094

.328

.205

.052

.019

.135

.070

.012

− .028

b

***

*

ns

***

**

ns

ns

***

***

*

ns

***

**

*

***

p

(continued)

t = 5.155**

t = 1.969*

t = 2.951*

Diff. test

Appendix 267

→ ONPI

.080

.087

.035 − .019

.011

.074

− .168

.093

*

***

ns

*

.079

− .169

.010

.089

.086

− .019

.033

.071

*

***

ns

*

Source: Own creation

Structural model fits: Proposed / efficient model: CFI .957; TLI .944; RMSEA .053; SRMR .049; χ2 (114) = 345.046. Consistent model: CFI .895; TLI .868; RMSEA .076; SRMR .104; χ2 (143) = 742.658. Notes: IV = Instrumental variable, OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, ONPI = Online purchase intention, β = standardized coefficients, b = unstandardized coefficients; Differences between total indirect and indirect standardized effects have been tested using t-tests. ***p < .001; **p < .01; *p < .05; ns = not significant.

General Internet expertise

Age

Gender

Covariates

ON-OF

Table A.30 (continued)

268 Appendix

Appendix

269

Table A.31 Measurement invariance Model

χ2 /df (p-value)

χ2 -Difference CFI (p-value) (CFI)

TLI (TLI)

.961

.949

RMSEA (RMSEA)

Offline purchase intention Model 1:

376.628/160

Configural invariance Model 2:

(.000)

(−)

381.873/170 14.245

Full metric invariance

(.000)

(.162)

(−)

.060 (−)

.960

.951

.054

(.001)

(.002)

(.006)

.966

.956

.058

Online purchase intention Model 1: Configural invariance Model 2: Full metric invariance

351.878/160 (.000) 368.858/170 16.981 (.000)

(.075)

.965

.957

.057

(.001)

(.001)

(.001)

Source: Own creation

3.4. Description of the Latent Moderated Structural Equation Method The LMS approach allows for testing moderated mediation with latent interactions. Two models need to be estimated, because the LMS approach does not provide conventional fit indices for assessing overall model fit (Muthén 2012). In a first step, a model without the latent interaction term is estimated. After ensuring adequate fit for the first model without the latent interaction and the significance of factor loadings, a profound measurement of constructs can be ensured. In the second step, the latent interactions were estimated (Cheung and Lau 2017; Maslowsky et al. 2015). As the computations are heavier than for models without latent variable interactions, numerical integration is needed (Muthén 2012). Therefore, an expectation-maximization algorithm is used to estimate the interactions. In adopting this algorithm, the LMS approach performs an iterative estimation of the model parameters and stops when the log-likelihood function of the observed variables is maximized. During the whole process, no products of indicators are created (Cheung and Lau 2017). To reduce model complexity, regression scores were applied for the integration services in the channel congruence moderation models (DiStefano et al. 2009). To present the standardized interaction effects, a standardization approach by considering the moderator function was

270

Appendix

used (Muthén 2012).The standardization process was also used for the first model without latent interactions to ensure comparability.

3.5. Further Tests and Models 3.5.1. Results of Alternative Moderator Models (Table A.32 and Table A.33)

.216 .516 .313 − .024 − .034

→ OFO

→ ONO

→ OFPI

→ OFPI

→ OFPI

→ OFPI

ON-OF

ON-OF

OFO

ONO

OF-ON

ON-OF

→ OFPI

→ OFPI

→ OFPI

→ ONO

→ OFO

→ ONO

OF-ON

ON-OF

.069 .085

→ OFPI

→ OFPI

ON-OF

.067

.017

.002

OF-ON

Total effects

ON-OF

→ OFPI

→ OFO

OF-ON

.067

.033

→ ONO

OF-ON

Indirect effects

.130 .005

→ OFO

OF-ON

Direct effects

.072

.067

.058

.015

.002

.066

− .029

− .024

.701

.936

.082

.016

.002

.070 ns

ns

ns

ns

ns

*

ns

ns

ns

ns

ns

***

***

**

p

.035

.221

.021

.013

.121

.100

.064

− .126

.400

.470

.053

.028

.303

.213

.654

.846

.027

.013

.170

.109

.028

.203

.017

.011

.111

.092

.053

− .115

b

ns

***

ns

ns

***

**

ns

ns

***

***

ns

ns

***

**

p

β

b

β

Model 1 Higher objective channel congruence

Lower objective channel congruence

Table A.32 Results of the multigroup models: offline purchase intention

ns

**

ns

ns

ns

ns

**

ns

**

**

Difference between groups

− .050

.065

Difference within groups: lower

(continued)

− .008

− .021*

Difference within groups: higher

Appendix 271

.069

− .012

− .131 .075

.516

.182 ns

*

* .001 − .036

.011

.082

− .043

.033 ns

ns

ns

Source: Own creation

Structural model fit Model 1: CFI .940; TLI .927; RMSEA .061; SRMR .066; χ2 (248) = 577.613. Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, OFPI = Offline purchase intention, Nlow = 322, Nhigh = 400, β = standardized coefficients, b = unstandardized coefficients. ***p < .001; **p < .01; *p < .05; ns = not significant.

General Internet expertise

Age

Gender

Covariates

Table A.32 (continued)

272 Appendix

.026 .209 .314 .378 − .016 .066

→ OFO

→ ONO

→ ONPI

→ ONPI

→ ONPI

→ ONPI

ON-OF

ON-OF

OFO

ONO

OF-ON

ON-OF

→ ONPI

→ ONPI

→ ONPI

→ ONO

→ OFO

OF-ON

ON-OF

ON-ON → ONO .042 .087

→ ONPI

→ ONPI

ON-OF

.079

.008

.013

OF-ON

Total effects

→ ONPI

→ OFO

OF-ON

.038

.008

→ ONO

OF-ON

Indirect effects

.122

→ OFO

OF-ON

Direct effects

.087

.048

.079

.002

.004

.044

.066

− .018

1.017

.708

.078

.012

.004

.062

*

ns

ns

ns

ns

ns

*

ns

ns

ns

ns

ns

***

***

p

.030

.246

.023

.006

.170

.076

.008

− .056

.548

.362

.043

.017

.310

.211

1.202

.904

.021

.007

.171

.102

.028

.203

.092

.111

.011

.017

.009

− .068

b

ns

***

ns

ns

***

*

ns

ns

***

***

ns

ns

***

**

p

β

b

β

Model 2 Higher objective channel congruence

Lower objective channel congruence

Table A.33 Results of the multigroup models: online purchase intention

ns

**

ns

ns

*

ns

*

ns

**

**

Difference between groups

.071*

− .025

Difference within groups: lower

(continued)

.017

.094*

Difference within groups: higher

Appendix 273

.053

− .029

− .258 .049

.303

.092 ns

*

ns .110

− .109

− .045 .121

− .012

− .150 *

*

ns

Source: Own creation

Structural model fit Model 2: CFI .947; TLI .936; RMSEA .059; SRMR .067; χ2 (248) = 554.429. Notes: OF-ON = Perceived offline-to-online services, ON-OF = Perceived online-to-offline services, OFO = Perceived quality of offline offerings, ONO = Perceived quality of online offerings, ONPI = Online purchase intention, Nlow = 322, Nhigh = 400, β = standardized coefficients, b = unstandardized coefficients. ***p < .001; **p < .01; *p < .05; ns = not significant.

General Internet expertise

Age

Gender

Covariates

Table A.33 (continued)

274 Appendix

Appendix

275

3.5.2. Additional Plots of Conditional Indirect Effects (Figure A.2)

a) OF-ON services on offline purchase intention through the online b) OF-ON services on online purchase intention through the online offerings offerings

Notes: Plots are based on unstandardized coefficients. The horizontal lines denote an indirect effect of zero. Dahsed lines represent the confidence bands. For any values of the moderator for which the confidence bands do not contain zero, the conditional indirect effect is significantly different from zero. Dots represent the conditional indirect effects in the mean value of moderator (ßOF-ON→ONO→OFPI = .058, p < .05; ßOF-ON→ONO→ONPI = .121, p > .01; Spiller et al. 2013).

Figure A.2 Plots of the conditional indirect effects: consumers’ online shopping experience. (Source: Own creation)

4. Study 3: Importance of Marketing Instruments for Repurchase Intentions in Omni-channel Retailing 4.1. Common Method Variance A data collection at different time points reduces the potential threat of CMV in our data set ex ante Fuller et al. 2016. Additionally, we used an appropriate questionnaire design. First, the respondents were told that the study was anonymous and confidential and that there were no right or wrong answers. Moreover, the study started with the measures of the dependent variables Chang, Van Witteloostuijn and Eden 2010. We calculated a single-factor test using confirmatory factor analysis. The results show that the models with all items loading on a single factor had a significantly worse fit than our proposed models did (see Table A.34 and A.35).

276

Appendix

Table A.34 Study 1: results of the single-factor tests CFI

TLI

RMSEA SRMR χ2 (df)

 χ2 (df)

p-value of difference

Online repurchase intention Proposed .971 .966 .048 model

.039

Single factor model

.148

.367 .319 .214

635.921 (341) 6270.502 (36) .000 6906.423 (377)

Offline repurchase intention Proposed .969 .963 .050 model

.039

Single factor model

.149

.366 .317 .215

658.961 (341) 6283.010 (36) .000 6941.971 (377)

Notes: Difference tests were conducted using χ2 tests of difference. Source: Own creation Table A.35 Study 2: results of the single-factor tests CFI

TLI

RMSEA

SRMR

χ2 (df)

 χ2 (df)

p-value of difference

1032.683 (3)

.000

1285.345 (3)

.000

1217.705 (3)

.000

Online repurchase intention Time point one Proposed model

.993

.990

.044

.018

55.124 (32)

Single factor model

.690

.601

.282

.146

1087.807 (35)

Time point two Proposed model

.992

.989

.050

.015

62.379 (32)

Single factor model

.664

.568

.315

.161

1347.724 (35)

.964

.090

.029

129.967 (32)

Time point three Proposed model

.975

(continued)

Appendix

277

Table A.35 (continued)

Single factor model

CFI

TLI

RMSEA

SRMR

χ2 (df)

.661

.564

.315

.167

1347.672 (35)

 χ2 (df)

p-value of difference

1144.547 (3)

.000

1476.049 (3)

.000

1375.561 (3)

.000

Offline repurchase intention Time point one Proposed model

.989

.985

.054

.023

67.788 (32)

Single factor model

.646

.545

.299

.163

1212.335 (35)

Time point two Proposed model

.987

.982

.065

.023

83.388 (32)

Single factor model

.610

.498

.340

.153

1559.437 (35)

Time point three Proposed model

.965

.951

.107

.035

168.861 (32)

Single factor model

.617

.508

.338

.186

1544.422 (35)

Notes: Difference tests were conducted using χ2 tests of difference. Source: Own creation

278

Appendix

4.2. Endogeneity Test In order to reduce possible biases from endogeneity the IV approach was used. Whether the results of the studies change, if the exogenous variables are endogenized by including IVs for each marketing instrument was checked. The IV’s for study 1 are measured with one item: “The offline store is visually appealing; The physical store has a very good overall layout design; I believe that my personal data are well protected in this physical store; [Retailer] provides reliable service through its offline store; The physical store allows consumers to inform themselves about the online store; The employees are helpful when using the online store” (e.g., adapted from Oh, Teo and Sambamurthy 2012; Montoya-Weiss, Voss and Grewal 2003). For study 2, brand attachment as an IV for online brand equity was used, which is theoretically a strong predictor for brand equity with one item (“I consider [retailer] as my first choice”, e.g., Keller 2010; Park et al. 2010). Offline trust is used as an IV for online trust, as it was shown to be strongly associated with online trust (“[Retailer’s] offline store can be trusted at all times“, Bock et al. 2012). First, F-tests proved that the IVs are strong predictors of the analyzed variables (see Table A.36). The IVs are included in the models to calculate consistent models in addition to the efficient (proposed) models (Antonakis et al. 2010, see Table A.37). Second, regarding the path estimates it was verified whether changes emerged (Hausman 1978). Respective z-values were below the critical value of 1.96 and it is concluded that the probability of endogeneity seems to be reduced (Table A.38 and A.39). Table A.36 Study 1: F-test of strong instrumental variable

Model 1 IV1 → Aesthetic appeal (1)

Model 2

F-value

F-value

271.218

282.904

IV2 → Ease of use (1)

129.195

131.099

IV3 → Security/privacy (1)

580.644

617.397

IV4 → Customer service (1)

130.788

135.595

IV5 → Online-offline integration (1)

623.908

646.227

24.066

24.248

IV6 → Channel consistency (1)

Notes: IV = Instrumental variable, F-value > 10 indicates strong predictor. Source: Own creation

− .117 .311

→ Online trust (2)

→ Online trust (2)



→ Channel consistency (1)

IV6

Security/privacy (1)



→ Online-offline integration (1)

IV5

Ease of use (1)



→ Customer service (1)

IV4

.166



→ Security/privacy (1)

IV3

→ Online trust (2)



→ Ease of use (1)

IV2

Aesthetic appeal (1)



→ Aesthetic appeal (1)

IV1

Direct effects

***

ns

**

.312

− .119

.163

***

ns

**













.172

.190

.698

.316

.830

.288

.474

.312

− .068

β

***

ns

**

***

***

***

***

***

***

p

β

β p

Model 1: Online RPI

Model 2: Offline RPI

Model 1: Online RPI p

Consistent model

Proposed / efficient model

Table A.37 Study 1: results of the efficient and consistent models

.170

.192

.698

.316

.830

.288

.474

.313

− .069

β

(continued)

***

ns

***

***

***

***

***

***

***

p

Model 2: Offline RPI

Appendix 279

.158 .079 .114 .443 − .122 .146 .007 .108 .159 − .022 .038 − .038

Customer service → Online trust (2) (1)

→ Online trust (2)

→ Online trust (2)

→ Online brand equity (2)

→ Online brand equity (2)

→ Online brand equity (2)

→ Online brand equity (2)

→ RPI (2)

→ RPI (2)

→ RPI (2)

Channel consistency (1)

Aesthetic appeal (1)

Ease of use (1)

Security/privacy (1)

Customer service → Online brand (1) equity (2)

→ Online brand equity (2)

Online-offline integration (1)

Online-offline integration (1)

Channel consistency (1)

Aesthetic appeal (1)

Ease of use (1)

Security/privacy (1)

Table A.37 (continued)

ns

ns

ns

**

*

ns

*

ns

***

*

ns

*

− .084

.020

− .023

.162

.109

.003

.147

− .123

.443

.115

.082

.158

ns

ns

ns

**

*

ns

**

ns

***

*

ns

*

− .049

.025

− .015

.162

.104

.006

.153

− .072

.440

.114

.078

.138

ns

ns

ns

***

**

ns

***

ns

***

**

ns

***

− .073

.008

− .013

.165

.104

− .009

.154

− .073

.440

.114

.081

.139

(continued)

ns

ns

ns

***

**

ns

***

ns

***

**

ns

***

280 Appendix

.078 .385 .504

→ RPI (2)

→ RPI (2)

→ RPI (2)

Channel consistency (1)

Online brand equity (2)

Online trust (2) → RPI (3) → RPI (3) → RPI (3) → RPI (3) → RPI (3) → RPI (3)

→ Online trust (2)

→ Online trust (2)

→ Online trust (2)

→ Online trust (2)

→ Online trust (2)

→ Online trust (2)

Aesthetic appeal (1)

Ease of use (1)

Security/privacy (1)

Customer service (1)

Online-offline integration (1)

Channel consistency (1)

.058

.040

.080

.157

− .059

.084

.064

→ RPI (2)

Online-offline integration (1)

Indirect effects

− .077

Customer service → RPI (2) (1)

Table A.37 (continued)

*

ns

*

***

ns

**

***

***

ns

ns

ns

.044

.031

.060

.118

− .045

.062

.378

.417

.256

.087

− .104

*

ns

*

***

ns

**

***

***

***

ns

ns

.057

.039

.070

.158

− .034

.087

.506

.387

.082

.062

− .059

**

ns

***

***

ns

***

***

***

ns

ns

ns

.042

.030

.051

.116

− .025

.063

.370

.413

.263

.088

− .090

(continued)

*

ns

***

***

ns

*

***

***

***

ns

ns

Appendix 281

→ RPI (3) → RPI (3) → RPI (3) → RPI (3) → RPI (3)

→ Online brand equity (2)

→ Online brand equity (2)

→ Online brand equity (2)

Security/privacy (1)

Customer service → Online brand (1) equity (2)

→ Online brand equity (2)

Ease of use (1)

Online-offline integration (1)

Channel consistency (1) .232 − .068 .175

.005 .146

→ RPI (3)

→ RPI (3)

→ RPI (3)

→ RPI (3)

→ RPI (3)

Ease of use (1)

Security/privacy (1)

Customer service (1)

Online-offline integration (1)

.061

.042

.003

.056

− .047

.170

Aesthetic appeal (1)

Total effects

→ RPI (3)

→ Online brand equity (2)

Aesthetic appeal (1)

Table A.37 (continued)

*

ns

**

ns

***

**

*

ns

*

ns

***

.163

− .043

.095

− .076

.224

.067

.045

.001

.061

− .051

.184

**

ns

†(.067)

ns

***

**

*

ns

**

ns

***

.142

.008

.168

− .037

.242

.063

.040

− .002

.059

− .028

.171

**

ns

***

ns

***

***

**

ns

**

ns

***

.161

− .043

.106

− .048

.231

.068

.043

− .004

.064

− .030

.182

(continued)

***

ns

*

ns

***

***

*

ns

**

ns

***

282 Appendix

.038 − .078 .015

→ RPI (3)

→ RPI (3)

→ RPI (3)

→ RPI (3)

Internet expertise (1)

Assortment variety (1)

Price fairness (1)

ns

ns

ns

ns

ns

**

.007

− .036

.040

− .053

.056

.367

ns

ns

ns

ns

ns

***

.013

− .084

.040

− .013

.049

.202

ns

ns

ns

ns

ns

***

.009

− .042

.043

− .054

.055

.373

ns

ns

ns

ns

ns

***

Source: Own creation

Structural model fit: Proposed / efficient model: Model 1: CFI .932, TLI .920, RMSEA .067, SRMR .128, χ2 (470) = 1254.596, SCF = 1.00. Model 2: CFI .930, TLI .918, RMSEA .068, SRMR .127, χ2 (470) = 1279.778, SCF = 1.01. Consistent model: Model 1: CFI .868, TLI .853, RMSEA .092, SRMR .202, χ2 (653) = 2716.112, SCF = .80. Model 2: CFI .865, TLI .850, RMSEA .093, SRMR .204, χ2 (653) = 2780.131, SCF = .80. Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377, β = standardized coefficients. ***p < .001, **p < .01, *p < .05, † p < .10, ns = not significant

.046 − .011

→ RPI (3)

.197

Age (1)

→ RPI (3)

Gender (1)

Covariates

Channel consistency (1)

Table A.37 (continued)

Appendix 283

284

Appendix

Table A.38 Study 2: F-test of strong instrumental variables

Model 1

Model 2

F-value

F-value

IV1 → Online trust

939.908

1156.615

IV2 → Online brand equity

498.705

546.090

Notes: IV = Instrumental variable, F-value > 10 indicates strong predictor. Source: Own creation

4.3. Test for Measurement Invariance Measurement equivalence was tested to ensure comparability across the three time points. First, configural invariance was ensured by estimating a baseline model in which the factor loadings and intercepts are freely estimated. Second, metric invariance was tested by fixing the factor loadings of each item. A comparison of configural and the metric model shows that all deviations are within limits. Additionally differences in the comparative fit indices were compared (following Chen 2007) to ensure measurement invariance. Partial metric invariance was ascertained by freely estimating some of the factor loadings (Table A.40).

.086 .047 .082 .619 .610

.682

→ Online trust (2)

→ RPI (2)

→ RPI (2)

→ Online trust (2)

→ Online brand equity (2)

→ RPI (2)

Online brand equity (1)

Online trust (1)

Online brand equity (1)

Online trust (1)

Online brand equity (1)

RPI (1)

***

***

***

*

*

**

.097

→ Online brand equity (2)

Online trust (1)

**



→ Online brand equity (1)

IV2





→ Online trust − (1)

.655

.611

.614

.103

.096

.084

.100

***

***

***

***

***

**

***

p

β

.683

.623

.633

.083

.047

.090

.108

.467

.788

***

***

***

**

*

***

***

***

***

p

β

β p

Model 1: Online RPI

Model 1: Online RPI

IV1

Direct effects

Consistent model Model 2: Offline RPI

Proposed / efficient model

Table A.39 Study 2: results of the efficient and consistent models

β

.656

.623

.627

.103

.098

.089

.111

.467

.789

(continued)

***

***

***

***

***

***

***

***

***

p

Model 2: Offline RPI

Appendix 285

.051 .088 .650 .646

.657

→ RPI (3)

→ RPI (3)

→ Online trust (3)

→ Online brand equity (3)

→ RPI (3)

Online trust (2)

Online brand equity (2)

Online trust (2)

Online brand equity (2)

RPI (2)

t = 1.988*

.112

→ RPI (3)

Online brand equity (1)

Diff. in total effects

.072

→ RPI (3)

Online trust (1)

Total effects

.525

.089

→ Online trust (3)

Online brand equity (2)

R2 RPI (3)

.104

→ Online brand equity (3)

Online trust (2)

Table A.39 (continued)

**

†(.060)

***

***

***

***

**

*

**

**

t = .333 ns

.140

.136

.565

.636

.640

.664

.109

.104

.084

.108

**

***

***

***

***

***

**

***

**

**

t = 2.429**

.113

.072

.526

.659

.638

.648

.087

.050

.091

.112

**

*

***

***

***

***

**

*

***

***

t = .135 ns

.140

.139

.564

.637

.632

.662

.106

.104

.091

.116

(continued)

***

***

***

***

***

***

***

***

***

***

286 Appendix

.063 .066 − .058 − .064 − .068 − .009 − .009 − .010

→ RPI (2)

→ RPI (3)

→ RPI (1)

→ RPI (2)

→ RPI (3)

→ RPI (1)

→ RPI (2)

→ RPI (3)

Gender (2)

Gender (3)

Age (1)

Age (2)

Age (3)

Internet expertise (1)

Internet expertise (2)

Internet expertise (3)

ns

ns

ns

**

**

**

**

**

**

− .017

− .017

− .016

− .072

− .069

− .064

.100

.096

.090

ns

ns

ns

**

**

**

***

***

***

− .010

− .010

− .009

− .068

− .064

− .057

.066

.063

.057

ns

ns

ns

**

**

**

**

**

**

− .017

− .017

− .016

− .073

− .069

− .064

.100

.096

.090

Source: Own creation

Structural model fits: Proposed / efficient model: Model 1: CFI .928, TLI .924, RMSEA .067, SRMR .183, χ2 (662) = 1794.952, SCF = .86. Model 2: CFI .926, TLI .921, RMSEA .069, SRMR .169, χ2 (662) = 1840.577, SCF = .86. Consistent model: Model 1: CFI .904, TLI .898, RMSEA .080, SRMR .202, χ2 (721) = 2470.852, SCF = .78. Model 2: CFI .893, TLI .886, RMSEA .086, SRMR .193, χ2 (721) = 2708.541, SCF = .78. Notes: RPI = Repurchase intention, (1, 2, 3) = Time points, SCF = Scaling correction factor for MLM, N = 377. Standardized coefficients are shown. ***p < .001, **p < .01, *p < .05, † p < .10, ns = not significant

.058

→ RPI (1)

Gender (1)

Covariates

Table A.39 (continued)

ns

ns

ns

**

**

**

***

***

***

Appendix 287

288

Appendix

Table A.40 Study 2: measurement invariance across time points Model

χ2 /df (p-value)

χ2 -Difference (p-value)

CFI (CFI)

TLI (TLI)

RMSEA (RMSEA)

SCF

Online repurchase intention Model 1: Configural invariance Model 2: Full metric invariance Model 3: Partial metric invariancea

1,041.474/369

.944

.934

.070

1.16

.943

.935

.069

1.15

(.001)

(.001)

(.001)

(.000) 1,070.549/383 (.000) 1,064.331/380 (.000)

24.731 (.037) 19.677 (.050)

.943

.935

.069

(.000)

(.000)

(.000)

1.15

.938

.927

.073

1.16

.936

.927

.073

1.14

(.002)

(.000)

(.000)

.937

.928

.073

(.001)

(.001)

(.000)

Offline repurchase intention Model 1: Configural invariance Model 2: Full metric invariance Model 3: Partial metric invarianceb

1,119.750/369 (.000) 1,155.412/383 (.000) 1,142.603/381 (.000)

32.836 (.003) 16.785 (.158)

1.14

Notes: SCF = Scaling correction factor for MLM. a Factor loading freed for the following item: ONB4 time point two, ONRPI2 time point one, ONRPI2 time point two. b Factor loading freed for the following item: OFRPI1 time point one, OFRPI1 time point two. Source: Own creation

4.4. Description of the Cross-lagged Panel Model Cross-lagged panel models are appropriate for studying causality in longitudinal data because reciprocal relationships between variables can be conceptualized over time (Allison, Williams and Moral-Benito 2017). Autoregressive relationships between a variable and its prior state have to be modelled (Zyphur et al. 2019). The constructs are measured at three time points. The advice of Burkholder and Harlow (2003) was followed to include disturbance correlations in the

Appendix

289

cross-lagged design. These correlations were modelled between the same indicators across the three time points. Disturbance correlations are also included between all constructs at time point two and are then integrated at time point three. They are constrained and thus estimated equally (Allison, Williams and Moral-Benito 2017) (Figure A.3).

Online trust (1)

Online trust (2)

Online trust (3)

Online retail brand equity (1)

Online retail brand equity (2)

Online retail brand equity (3)

Repurchase intention (1)

Repurchase intention (2)

Repurchase intention (3)

Figure A.3 Cross-lagged panel model. (Source: Own creation)

References

Acquila-Natale, E. & Chaparro-Peláez, J. (2020). The Long Road to Omnichannel Retailing: An Assessment of Channel Integration Levels across Fashion and Apparel Retailers. European Journal of International Management, 14(6), 1–16. Acquila-Natale, E. & Iglesias-Pradas, S. (2020). How to Measure Quality in Multi-Channel Retailing and Not Die Trying. Journal of Business Research, 109(March), 38–48. Aghekyan-Simonian, M., Forsythe, S., Kwon, W. S., & Chattaraman, V. (2012). The Role of Product Brand Image and Online Store Image on Perceived Risks and Online Purchase Intentions for Apparel. Journal of Retailing and Consumer Services, 19(3), 325–331. Ailawadi, K. L. & Farris, P. W. (2017). Managing Multi– and Omni–Channel Distribution: Metrics and Research Directions. Journal of Retailing, 93(1), 120–135. Akturk, M. S., Ketzenberg, M., & Heim, G. R. (2018). Assessing Impacts of Introducing Ship-to-Store Service on Sales and Returns in Omnichannel Retailing: A Data Analytics Study. Journal of Operations Management, 61(July), 15–45. Al-Hawari, M. A. (2011). Do Online Services Contribute to Establishing Brand Equity within the Retail Banking Context? Journal of Relationship Marketing, 10(3), 145–166. Al-Qeisi, K., Dennis, C., Alamanos, E., & Jayawardhena, C. (2014). Website Design Quality and Usage Behavior: Unified Theory of Acceptance and Use of Technology. Journal of Business Research, 67(11), 2282–2290. Alba, J. W. & Hutchinson, J. W. (1987). Dimensions of Consumer Expertise. Journal of Consumer Research, 13(4), 411–454. Allison, P. D., Williams, R., & Moral-Benito, E. (2017). Maximum Likelihood for CrossLagged Panel Models with Fixed Effects. Socius, 3(June), 1–17. Anderson, J. R. (1983). A Spreading Activation Theory of Memory. Journal of Verbal Learning and Verbal Behavior, 22(3), 261–295. Anselmsson, J., Burt, S., & Tunca, B. (2017). An Integrated Retailer Image and Brand Equity Framework: Re-Examining, Extending, and Restructuring Retailer Brand Equity. Journal of Retailing and Consumer Services, 38(September), 194–203. Antonakis, J., Bendahan, S., Jacquart, P., & Lalive, R. (2014). Causality and Endogeneity: Problems and Solutions. In The Oxford Handbook of Leadership and Organizations, David D. Day (ed). Oxford: Oxford University Press, 93–117. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2021 A. Winters, Omni-Channel Retailing, Handel und Internationales Marketing Retailing and International Marketing, https://doi.org/10.1007/978-3-658-34707-9

291

292

References

Antonakis, J., Bendahan, S., Jacquart, P., & Lalive, R. (2010). On Making Causal Claims: A Review and Recommendations. The Leadership Quarterly, 21(6), 1086–1120. Awad, N. F. & Ragowsky, A. (2008). Establishing Trust in Electronic Commerce through Online Word of Mouth: An Examination across Genders. Journal of Management Information Systems, 24(4), 101–121. Badrinarayanan, V. & Becerra, E. P. (2019). Shoppers’ Attachment with Retail Stores: Antecedents and Impact on Patronage Intentions. Journal of Retailing and Consumer Services, 50(September), 371–378. Badrinarayanan, V., Becerra, E. P., Kim, C.-H., & Madhavaram, S. (2012). Transference and Congruence Effects on Purchase Intentions in Online Stores of Multi-Channel Retailers: Initial Evidence from the U.S. and South Korea. Journal of the Academy of Marketing Science, 40(4), 539–557. Badrinarayanan, V., Becerra, E. P., & Madhavaram, S. (2014). Influence of Congruity in Store-Attribute Dimensions and Self-Image on Purchase Intentions in Online Stores of Multichannel Retailers. Journal of Retailing and Consumer Services, 21(6), 1013–1020. Baek, E., Choo, H. J., Wei, X., & Yoon, S.-Y. (2020). Understanding the Virtual Tours of Retail Stores: How Can Store Brand Experience Promote Visit Intentions? International Journal of Retail & Distribution Management, 48(7), 649–666. Balachander, S. & Ghose, S. (2003). Reciprocal Spillover Effects: A Strategic Benefit of Brand Extensions. Journal of Marketing, 67(1), 4–13. Banerjee, M. (2014). Misalignment and Its Influence on Integration Quality in Multichannel Services. Journal of Service Research, 17(4), 460–474. Bansal, G., Zahedi, F. M., & Gefen, D. (2016). Do Context and Personality Matter? Trust and Privacy Concerns in Disclosing Private Information Online. Information & Management, 53(1), 1–21. Barrutia, J. M. & Gilsanz, A. (2013). Electronic Service Quality and Value: Do Consumer Knowledge-Related Resources Matter? Journal of Service Research, 16(2), 231–246. Bashir, S., Anwar, S., Awan, Z., Qureshi, T. W., & Memon, A. B. (2018). A Holistic Understanding of the Prospects of Financial Loss to Enhance Shopper’s Trust to Search, Recommend, Speak Positive and Frequently Visit an Online Shop. Journal of Retailing and Consumer Services, 42(May), 169–174. Batra, R. & Keller, K. L. (2016). Integrating Marketing Communications: New Findings, New Lessons, and New Ideas. Journal of Marketing, 80(6), 122–145. Beal, D. J. (2015). ESM 2.0: State of the Art and Future Potential of Experience Sampling Methods in Organizational Research. Annual Review of Organizational Psychology and Organizational Behavior, 2(1), 383–407. Becker, J. U., Spann, M., & Barrot, C. (2020). Impact of Proactive Postsales Service and CrossSelling Activities on Customer Churn and Service Calls. Journal of Service Research, 23(1), 53–69. Becker, L. & Jaakkola, E. (2020). Customer Experience: Fundamental Premises and Implications for Research. Journal of the Academy of Marketing Science, 48(January), 630–648. Bell, D. R., Gallino, S., & Moreno, A. (2018). Offline Showrooms in Omnichannel Retail: Demand and Operational Benefits. Management Science, 64(4), 1629–1651.

References

293

Bendoly, E., Blocher, J. D., Bretthauer, K. M., Krishnan, S., & Venkataramanan, M. A. (2005). Online/In-Store Integration and Customer Retention. Journal of Service Research, 7(4), 313–327. Benedicktus, R. L., Brady, M. K., Darke, P. R., & Voorhees, C. M. (2010). Conveying Trustworthiness to Online Consumers: Reactions to Consensus, Physical Store Presence, Brand Familiarity, and Generalized Suspicion. Journal of Retailing, 86(4), 322–335. Bennink, M., Moors, G., & Gelissen, J. (2013). Exploring Response Differences between Face-to-Face and Web Surveys: A Qualitative Comparative Analysis of the Dutch European Values Survey 2008. Field Methods, 25(4), 319–338. Berg, B. (2013). Retail Branding and Store Loyalty: Analysis in the Context of Reciprocity, Store Accessibility, and Retail Format. Wiesbaden: Gabler. Bertrandie, L. & Zielke, S. (2017). The Effects of Multi-Channel Assortment Integration on Customer Confusion. The International Review of Retail, Distribution and Consumer Research, 27(5), 437–449. BEVH/Beyonddata (2019). Interaktiver Handel. Berlin: Bundesverband E-Commerce und Versandhandel Deutschland e.V. BEVH/Beyonddata (2020). Interaktiver Handel. Berlin: Bundesverband E-Commerce und Versandhandel Deutschland e.V. BEVH/GIM (2016). Interaktiver Handel. Berlin: Bundesverband E-Commerce und Versandhandel Deutschland e.V. BEVH/GIM (2015). Interaktiver Handel. Berlin: Bundesverband E-Commerce und Versandhandel Deutschland e.V. BEVH/GIM (2017). Interaktiver Handel. Berlin: Bundesverband E-Commerce und Versandhandel Deutschland e.V. BEVH/GIM (2018). Interaktiver Handel. Berlin: Bundesverband E-Commerce und Versandhandel Deutschland e.V. Bezes, C. (2014). Definition and Psychometric Validation of a Measurement Index Common to Website and Store Images. Journal of Business Research, 67(12), 2559–2578. Bezes, C. (2013). Effect of Channel Congruence on a Retailer’s Image. International Journal of Retail & Distribution Management, 41(4), 254–273. Bhargave, R., Mantonakis, A., & White, K. (2016). The Cue-of-the-Cloud Effect: When Reminders of Online Information Availability Increase Purchase Intentions and Choice. Journal of Marketing Research, 53(5), 699–711. Bleier, A., Harmeling, C. M., & Palmatier, R. W. (2019). Creating Effective Online Customer Experiences. Journal of Marketing, 83(2), 98–119. Blut, M., Teller, C., & Floh, A. (2018). Testing Retail Marketing-Mix Effects on Patronage: A Meta-Analysis. Journal of Retailing, 94(2), 113–135. Bock, G.-W., Lee, J., Kuan, H.-H., & Kim, J.-H. (2012). The Progression of Online Trust in the Multi-Channel Retailer Context and the Role of Product Uncertainty. Decision Support Systems, 53(1), 97–107. Bolton, R. N., McColl-Kennedy, J. R., Cheung, L., Gallan, A., Orsingher, C., Witell, L., & Zaki, M. (2018). Customer Experience Challenges: Bringing Together Digital, Physical and Social Realms. Journal of Service Management, 29(5), 776–808. Bressolles, G., Durrieu, F., & Deans, K. R. (2015). An Examination of the Online ServiceProfit Chain. International Journal of Retail & Distribution Management, 43(8), 727–751.

294

References

Breugelmans, E. & Campo, K. (2016). Cross-Channel Effects of Price Promotions: An Empirical Analysis of the Multi-Channel Grocery Retail Sector. Journal of Retailing, 92(3), 333–351. Burkholder, G. J. & Harlow, L. L. (2003). An Illustration of a Longitudinal Cross-Lagged Design for Larger Structural Equation Models. Structural Equation Modeling, 10(3), 465– 486. Cai, S. & Xu, Y. (2011). Designing Not Just for Pleasure: Effects of Web Site Aesthetics on Consumer Shopping Value. International Journal of Electronic Commerce, 15(4), 159– 188. Cambra-Fierro, J., Melero-Polo, I., Patrício, L., & Sese, F. J. (2020). Channel Habits and the Development of Successful Customer-Firm Relationships in Services. Journal of Service Research, 23(4), 456–475. Campo, K. & Breugelmans, E. (2015). Buying Groceries in Brick and Click Stores: Category Allocation Decisions and the Moderating Effect of Online Buying Experience. Journal of Interactive Marketing, 31(August), 63–78. Cao, L. & Li, L. (2018). Determinants of Retailers’ Cross-Channel Integration: An Innovation Diffusion Perspective on Omni-Channel Retailing. Journal of Interactive Marketing, 44(November), 1–16. Cao, L. & Li, L. (2015). The Impact of Cross-Channel Integration on Retailers’ Sales Growth. Journal of Retailing, 91(2), 198–216. Cao, Y., Ajjan, H., & Hong, P. (2018). Post-Purchase Shipping and Customer Service Experiences in Online Shopping and Their Impact on Customer Satisfaction. Asia Pacific Journal of Marketing and Logistics, 30(2), 400–416. Carlson, J. & O’Cass, A. (2011). Managing Web Site Performance Taking Account of the Contingency Role of Branding in Multi-Channel Retailing. Journal of Consumer Marketing, 28(7), 524–531. Carlson, J., O’Cass, A., & Ahrholdt, D. (2015). Assessing Customers’ Perceived Value of the Online Channel of Multichannel Retailers: A Two Country Examination. Journal of Retailing and Consumer Services, 27(November), 90–102. Casado-Aranda, L.-A., Dimoka, A., & Sánchez-Fernández, J. (2019). Consumer Processing of Online Trust Signals: A Neuroimaging Study. Journal of Interactive Marketing, 47(August), 159–180. Chang, E.-C. & Tseng, Y.-F. (2013). Research Note: E-Store Image, Perceived Value and Perceived Risk. Journal of Business Research, 66(7), 864–870. Chang, S.-J., Van Witteloostuijn, A., & Eden, L. (2010). From the Editors: Common Method Variance in International Business Research. Journal of International Business Studies, 41(2), 178–184. Chaudhuri, A. & Holbrook, M. B. (2001). The Chain of Effects from Brand Trust and Brand Affect to Brand Performance: The Role of Brand Loyalty. Journal of Marketing, 65(2), 81–93. Chen, F. F. (2007). Sensitivity of Goodness of Fit Indexes to Lack of Measurement Invariance. Structural Equation Modeling, 14(3), 464–504. Chen, G., Gully, S. M., & Eden, D. (2001). Validation of a New General Self-Efficacy Scale. Organizational Research Methods, 4(1), 62–83. Chen, J. & Dibb, S. (2010). Consumer Trust in the Online Retail Context: Exploring the Antecedents and Consequences. Psychology & Marketing, 27(4), 323–346.

References

295

Chen, Y., Cheung, C. M. K., & Tan, C.-W. (2018). Omnichannel Business Research: Opportunities and Challenges. Decision Support Systems, 109(May), 1–4. Cheung, G. W. & Lau, R. S. (2017). Accuracy of Parameter Estimates and Confidence Intervals in Moderated Mediation Models: A Comparison of Regression and Latent Moderated Structural Equations. Organizational Research Methods, 20(4), 746–769. Chiu, H.-C., Hsieh, Y.-C., Roan, J., Tseng, K.-J., & Hsieh, J.-K. (2011). The Challenge for Multichannel Services: Cross-Channel Free-Riding Behavior. Electronic Commerce Research and Applications, 10(2), 268–277. Chowdhury, J., Reardon, J., & Srivastava, R. (1998). Alternative Modes of Measuring Store Image: An Empirical Assessment of Structured Versus Unstructured Measures. Journal of Marketing Theory and Practice, 6(2), 72–86. Chu, S.-Y. C., Wu, C., Wu, K., & Chen, Y.-F. (2017). Does Established Offline Store Drive Online Purchase Intention? The International Journal of Business and Information, 11(4), 432–465. Collier, J. E. & Kimes, S. E. (2013). Only If It Is Convenient: Understanding How Convenience Influences Self-Service Technology Evaluation. Journal of Service Research, 16(1), 39– 51. Connelly, B. L., Certo, S. T., Ireland, R. D., & Reutzel, C. R. (2011). Signaling Theory: A Review and Assessment. Journal of Management, 37(1), 39–67. Dahl, A. J., Milne, G. R., & Peltier, J. W. (2019). Digital Health Information Seeking in an Omni-Channel Environment: A Shared Decision-Making and Service-Dominant Logic Perspective. Journal of Business Research, in press. Darke, P. R., Brady, M. K., Benedicktus, R. L., & Wilson, A. E. (2016). Feeling Close from Afar: The Role of Psychological Distance in Offsetting Distrust in Unfamiliar Online Retailers. Journal of Retailing, 92(3), 287–299. Das, G. (2016). Antecedents and Consequences of Trust: An E-Tail Branding Perspective. International Journal of Retail & Distribution Management, 44(7), 713–730. De Kerviler, G., Demoulin, N. T. M., & Zidda, P. (2016). Adoption of In-Store Mobile Payment: Are Perceived Risk and Convenience the Only Drivers? Journal of Retailing and Consumer Services, 31(July), 334–344. De Keyser, A., Verleye, K., Lemon, K. N., Keiningham, T. L., & Klaus, P. (2020). Moving the Customer Experience Field Forward: Introducing the Touchpoints, Context, Qualities (TCQ) Nomenclature. Journal of Service Research, 23(4), 433–455. Demangeot, C. & Broderick, A. J. (2016). Engaging Customers During a Website Visit: A Model of Website Customer Engagement. International Journal of Retail & Distribution Management, 44(8), 814–839. Destatis. (2020). Turnover in Retail Trade Sectors. https://www.destatis.de/DE/Themen/Wir tschaft/Grosshandel-Einzelhandel/_inhalt.html. Accessed October, 30, 2020. Diamantopoulos, A., Smith, G., & Grime, I. (2005). The Impact of Brand Extensions on Brand Personality: Experimental Evidence. European Journal of Marketing, 39(1/2), 129–149. Dickinger, A. & Stangl, B. (2013). Website Performance and Behavioral Consequences: A Formative Measurement Approach. Journal of Business Research, 66(6), 771–777. DiStefano, C., Zhu, M., & Mindrila, D. (2009). Understanding and Using Factor Scores: Considerations for the Applied Researcher. Practical Assessment, Research & Evaluation, 14(20), 1–11.

296

References

EcommerceDB. (2020). Store Ranking & Overview. https://ecommercedb.com/en/ranking/ ww/fashion. Accessed June, 17, 2020. EHI (2019). Omnichannel-Commerce 2019. Köln: EHI Retail Institute. Eisingerich, A. B. & Bell, S. J. (2008). Perceived Service Quality and Customer Trust: Does Enhancing Customers’ Service Knowledge Matter? Journal of Service Research, 10(3), 256–268. Emrich, O., Paul, M., & Rudolph, T. (2015). Shopping Benefits of Multichannel Assortment Integration and the Moderating Role of Retailer Type. Journal of Retailing, 91(2), 326– 342. Emrich, O. & Verhoef, P. C. (2015). The Impact of a Homogenous Versus a Prototypical Web Design on Online Retail Patronage for Multichannel Providers. International Journal of Research in Marketing, 32(4), 363–374. EY Parthenon. (2019). The King Is Dead, Long Live the King! Multichannel, Omnichannel, Cross-Channel: Does Cross-Channel Retail Still Have a Chance? https://ey.com/public ation. Accessed February, 19, 2020. Falk, T., Schepers, J., Hammerschmidt, M., & Bauer, H. H. (2007). Identifying Cross-Channel Dissynergies for Multichannel Service Providers. Journal of Service Research, 10(2), 143–160. Fang, J., Wen, C., George, B., & Prybutok, V. R. (2016). Consumer Heterogeneity, Perceived Value, and Repurchase Decision-Making in Online Shopping: The Role of Gender, Age, and Shopping Motives. Journal of Electronic Commerce Research, 17(2), 116–131. Fazal-e-Hasan, S. M., Ahmadi, H., Mortimer, G., Grimmer, M., & Kelly, L. (2018). Examining the Role of Consumer Hope in Explaining the Impact of Perceived Brand Value on Customer-Brand Relationship Outcomes in an Online Retailing Environment. Journal of Retailing and Consumer Services, 41(March), 101–111. Feldman, J. M. (1981). Beyond Attribution Theory: Cognitive Processes in Performance Appraisal. Journal of Applied Psychology, 66(2), 127–148. Feldman, J. M. & Lynch, J. G. (1988). Self-Generated Validity and Other Effects of Measurement on Belief, Attitude, Intention, and Behavior. Journal of Applied Psychology, 73(3), 421–435. Fiske, S. T. & Taylor, S. E. (1991). Social Cognition. New York: Mcgraw-Hill Book Company. Floh, A. & Madlberger, M. (2013). The Role of Atmospheric Cues in Online Impulse-Buying Behavior. Electronic Commerce Research and Applications, 12(6), 425–439. Forbes. (2019). Omnichannel Is Dead. The Future Is Harmonized Retail. https://forbes.com. Accessed January, 13, 2020. Fornell, C. & Larcker, D. F. (1981). Structural Equation Models with Unobservable Variables and Measurement Error: Algebra and Statistics. Journal of Marketing Research, 18(3), 375–381. Frasquet, M., Descals, A. M., & Ruiz-Molina, M. E. (2017). Understanding Loyalty in Multichannel Retailing: The Role of Brand Trust and Brand Attachment. International Journal of Retail & Distribution Management, 45(6), 608–625. Frasquet, M. & Miquel, M.-J. (2017). Do Channel Integration Efforts Pay-Off in Terms of Online and Offline Customer Loyalty? International Journal of Retail & Distribution Management, 45(7/8), 859–873.

References

297

Fuentes-Blasco, M., Moliner-Velázquez, B., Servera-Francés, D., & Gil-Saura, I. (2017). Role of Marketing and Technological Innovation on Store Equity, Satisfaction and Word-ofMouth in Retailing. Journal of Product & Brand Management, 26(6), 650–666. Fuller, C. M., Simmering, M. J., Atinc, G., Atinc, Y., & Babin, B. J. (2016). Common Methods Variance Detection in Business Research. Journal of Business Research, 69(8), 3192– 3198. Gallino, S. & Moreno, A. (2014). Integration of Online and Offline Channels in Retail: The Impact of Sharing Reliable Inventory Availability Information. Management Science, 60(6), 1434–1451. Gallino, S., Moreno, A., & Stamatopoulos, I. (2017). Channel Integration, Sales Dispersion, and Inventory Management. Management Science, 63(9), 2813–2831. Gao, F. & Su, X. (2017). Omnichannel Retail Operations with Buy-Online-and-Pick-Up-InStore. Management Science, 63(8), 2478–2492. Gao, L., Melero, I., & Sese, F. J. (2019). Multichannel Integration Along the Customer Journey: A Systematic Review and Research Agenda. The Service Industries Journal, 15(August), 1087–1118. Gaur, A. & Kumar, M. (2018). A Systematic Approach to Conducting Review Studies: An Assessment of Content Analysis in 25 Years of IB Research. Journal of World Business, 53(2), 280–289. Gensler, S., Verhoef, P. C., & Böhm, M. (2012). Understanding Consumers’ Multichannel Choices across the Different Stages of the Buying Process. Marketing Letters, 23(4), 987–1003. Gielens, K. & Steenkamp, J.-B. E. M. (2019). Branding in the Era of Digital (Dis)Intermediation. International Journal of Research in Marketing, 36(3), 367–384. Grewal, D., Levy, M., & Kumar, V. (2009). Customer Experience Management in Retailing: An Organizing Framework. Journal of Retailing, 85(1), 1–14. Grewal, D., Noble, S. M., Roggeveen, A. L., & Nordfalt, J. (2020). The Future of In-Store Technology. Journal of the Academy of Marketing Science, 48(1), 96–113. Grewal, D. & Roggeveen, A. L. (2020). Understanding Retail Experiences and Customer Journey Management. Journal of Retailing, 96(1), 3–8. Grosso, M., Castaldo, S., & Grewal, A. (2018). How Store Attributes Impact Shoppers’ Loyalty in Emerging Countries: An Investigation in the Indian Retail Sector. Journal of Retailing and Consumer Services, 40(January), 117–124. Hair, J. F., Black, W. C., Babin, B. J., E., A. R., & Tatham, R. L. (2018). Multivariate Data Analysis. Hampshire: Cengage. Halkias, G. (2015). Mental Representation of Brands: A Schema-Based Approach to Consumers’ Organization of Market Knowledge. Journal of Product & Brand Management, 24(5), 438–448. Hammerschmidt, M., Falk, T., & Weijters, B. (2016). Channels in the Mirror: An Alignable Model for Assessing Customer Satisfaction in Concurrent Channel Systems. Journal of Service Research, 19(1), 88–101. Hamouda, M. (2019). Omni-Channel Banking Integration Quality and Perceived Value as Drivers of Consumers’ Satisfaction and Loyalty. Journal of Enterprise Information Management, 32(4), 608–625. Hausman, J. A. (1978). Specification Tests in Econometrics. Econometrica, 46(6), 1251–1271.

298

References

HDE (2020). Konjunkturumfrage Frühjahr 2020. Berlin: Handelsverband Deutschland (HDE). HDE/IFH (2019). Online Monitor. Berlin: Handelsverband Deutschland (HDE). HDE/IFH (2020). Online Monitor. Berlin: Handelsverband Deutschland (HDE). Heerwegh, D. (2009). Mode Differences between Face-to-Face and Web-Surveys: An Experimental Investigation of Data Quality and Social Desirability Effects. International Journal of Public Opinion Research, 21(1), 111–121. Herhausen, D., Binder, J., Schoegel, M., & Herrmann, A. (2015). Integrating Bricks with Clicks: Retailer-Level and Channel-Level Outcomes of Online-Offline Channel Integration. Journal of Retailing, 91(2), 309–325. Herhausen, D., Emrich, O., Grewal, D., Kipfelsberger, P., & Schoegel, M. (2020). Face Forward: How Employees’ Digital Presence on Service Websites Affects Customer Perceptions of Website and Employee Service Quality. Journal of Marketing Research, 57(5), 917–936. Herhausen, D., Kleinlercher, K., Verhoef, P. C., Emrich, O., & Rudolph, T. (2019). Loyalty Formation for Different Customer Journey Segments. Journal of Retailing, 95(3), 9–29. Hogreve, J., Bilstein, N., & Hoerner, K. (2019). Service Recovery on Stage: Effects of Social Media Recovery on Virtually Present Others. Journal of Service Research, 22(4), 421–439. Hollebeek, L. D. & Macky, K. (2019). Digital Content Marketing’s Role in Fostering Consumer Engagement, Trust, and Value: Framework, Fundamental Propositions, and Implications. Journal of Interactive Marketing, 45(February), 27–41. Homburg, C., Jozi´c, D., & Kuehnl, C. (2017). Customer Experience Management: Toward Implementing an Evolving Marketing Concept. Journal of the Academy of Marketing Science, 45(3), 377–401. Hossain, T. M. T., Akter, S., Kattiyapornpong, U., & Dwivedi, Y. (2020). Reconceptualizing Integration Quality Dynamics for Omnichannel Marketing. Industrial Marketing Management, 87(May), 225–241. Hosseini, S., Merz, M., Röglinger, M., & Wenninger, A. (2018). Mindfully Going OmniChannel: An Economic Decision Model for Evaluating Omni-Channel Strategies. Decision Support Systems, 109(May), 74–88. Hox, J. J., Moerbeek, M., & Van de Schoot, R. (2017). Multilevel Analysis: Techniques and Applications. New York: Routledge. Hsieh, J.-K., Hsieh, Y.-C., Chiu, H.-C., & Yang, Y.-R. (2014). Customer Response to Web Site Atmospherics: Task-Relevant Cues, Situational Involvement and PAD. Journal of Interactive Marketing, 28(3), 225–236. Hsu, M. K., Huang, Y., & Swanson, S. (2010). Grocery Store Image, Travel Distance, Satisfaction and Behavioral Intentions: Evidence from a Midwest College Town. International Journal of Retail & Distribution Management, 38(2), 115–132. Hult, G. T. M., Sharma, P. N., Morgeson III, F. V., & Zhang, Y. (2019). Antecedents and Consequences of Customer Satisfaction: Do They Differ across Online and Offline Purchases? Journal of Retailing, 95(1), 10–23. Hunneman, A., Verhoef, P. C., & Sloot, L. M. (2017). The Moderating Role of Shopping Trip Type in Store Satisfaction Formation. Journal of Business Research, 78(9), 133–142. Ibrahim, N. F. & Wang, X. (2019). A Text Analytics Approach for Online Retailing Service Improvement: Evidence from Twitter. Decision Support Systems, 121(June), 37–50.

References

299

Inman, J. J., Shankar, V., & Ferraro, R. (2004). The Roles of Channel-Category Associations and Geodemographics in Channel Patronage. Journal of Marketing, 68(2), 51–71. Jara, M., Vyt, D., Mevel, O., Morvan, T., & Morvan, N. (2018). Measuring Customers Benefits of Click and Collect. Journal of Services Marketing, 32(4), 430–442. Javed, M. K. & Wu, M. (2020). Effects of Online Retailer after Delivery Services on Repurchase Intention: An Empirical Analysis of Customers’ Past Experience and Future Confidence with the Retailer. Journal of Retailing and Consumer Services, 54(May), 1–7. Jin, M., Li, G., & Cheng, T. C. E. (2018). Buy Online and Pick Up In-Store: Design of the Service Area. European Journal of Operational Research, 268(2), 613–623. Jindal, R. P., Gauri, D. K., Li, W., & Ma, Y. (2021). Omnichannel Battle between Amazon and Walmart: Is the Focus on Delivery the Best Strategy? Journal of Business Research, 122(January), 270–280. Jinfeng, W. & Zhilong, T. (2009). The Impact of Selected Store Image Dimensions on Retailer Equity: Evidence from 10 Chinese Hypermarkets. Journal of Retailing and Consumer Services, 16(6), 486–494. Keaveney, S. M. & Hunt, K. A. (1992). Conceptualization and Operationalization of a Retail Store Image: A Case of Rival Middle-Level Theories. Journal of the Academy of Marketing Science, 20(2), 165–175. Keller, K. L. (2010). Brand Equity Management in a Multichannel, Multimedia Retail Environment. Journal of Interactive Marketing, 24(2), 58–70. Keller, K. L. (1993). Conceptualizing, Measuring, and Managing Customer-Based Brand Equity. Journal of Marketing, 57(1), 1–22. Kent, R. J. & Allen, C. T. (1994). Competitive Interference Effects in Consumer Memory for Advertising: The Role of Brand Familiarity. Journal of Marketing, 58(3), 97–105. Khan, I., Hollebeek, L. D., Fatma, M., Islam, J. U., & Rahman, Z. (2019). Brand Engagement and Experience in Online Services. Journal of Services Marketing, 34(2), 163–175. Kim, C., Galliers, R. D., Shin, N., Ryoo, J.-H., & Kim, J. (2012). Factors Influencing Internet Shopping Value and Customer Repurchase Intention. Electronic Commerce Research and Applications, 11(4), 374–387. Kim, Y. & Peterson, R. A. (2017). A Meta-Analysis of Online Trust Relationships in ECommerce. Journal of Interactive Marketing, 38(May), 44–54. Klein, A. & Moosbrugger, H. (2000). Maximum Likelihood Estimation of Latent Interaction Effects with the LMS Method. Psychometrika, 65(4), 457–474. Kleinlercher, K., Emrich, O., Herhausen, D., Verhoef, P. C., & Rudolph, T. (2018). Websites as Information Hubs: How Informational Channel Integration and Shopping Benefit Density Interact in Steering Customers to the Physical Store. Journal of the Association for Consumer Research, 3(3), 330–342. Krishnan, H. S. (1996). Characteristics of Memory Associations: A Consumer-Based Brand Equity Perspective. Journal of International Research in Marketing, 13(4), 389–405. Kuehnl, C., Jozic, D., & Homburg, C. (2019). Effective Customer Journey Design: Consumers’ Conception, Measurement, and Consequences. Journal of the Academy of Marketing Science, 47(3), 551–568. Kumar, V., Anand, A., & Song, H. (2017). Future of Retailer Profitability: An Organizing Framework. Journal of Retailing, 93(1), 96–119. Kwon, W.-S. & Lennon, S. J. (2009a). Reciprocal Effects between Multichannel Retailers’ Offline and Online Brand Images. Journal of Retailing, 85(3), 376–390.

300

References

Kwon, W.-S. & Lennon, S. J. (2009b). What Induces Online Loyalty? Online Versus Offline Brand Images. Journal of Business Research, 62(5), 557–564. Landers, V. M., Beatty, S. E., Wang, S., & Mothersbaugh, D. L. (2015). The Effect of Online Versus Offline Retailer-Brand Image Incongruity on the Flow Experience. Journal of Marketing Theory and Practice, 23(4), 370–387. Lee, Z. W. Y., Chan, T. K. H., Chong, A. Y.-L., & Thadani, D. R. (2019). Customer Engagement through Omnichannel Retailing: The Effects of Channel Integration Quality. Industrial Marketing Management, 77(2), 90–101. Lei, J., Dawar, N., & Lemmink, J. (2008). Negative Spillover in Brand Portfolios: Exploring the Antecedents of Asymmetric Effects. Journal of Marketing, 72(3), 111–123. Lemon, K. N. & Verhoef, P. C. (2016). Understanding Customer Experience Throughout the Customer Journey. Journal of Marketing, 80(November), 69–96. Li, Y., Liu, H., Lim, E. T. K., Goh, J. M., Yang, F., & Lee, M. K. O. (2018). Customer’s Reaction to Cross-Channel Integration in Omnichannel Retailing: The Mediating Roles of Retailer Uncertainty, Identity Attractiveness, and Switching Costs. Decision Support Systems, 109(May), 50–60. Lin, M.-Q. & Lee, B. C. Y. (2012). The Influence of Website Environment on Brand Loyalty: Brand Trust and Brand Affect as Mediators. International Journal of Electronic Business Management, 10(4), 308–321. Lindell, M. K. & Whitney, D. J. (2001). Accounting for Common Method Variance in CrossSectional Research Designs. Journal of Applied Psychology, 86(1), 114–121. Liu, Y., Li, H., & Hu, F. (2013). Website Attributes in Urging Online Impulse Purchase: An Empirical Investigation on Consumer Perceptions. Decision Support Systems, 55(3), 829–837. Liu, Y., Li, K. J., Chen, H., & Balachander, S. (2017). The Effects of Products’ Aesthetic Design on Demand and Marketing-Mix Effectiveness: The Role of Segment Prototypicality and Brand Consistency. Journal of Marketing, 81(1), 83–102. Loken, B. (2006). Consumer Psychology: Categorization, Inferences, Affect, and Persuasion. Annual Review of Psychology, 57(January), 453–485. Loupiac, P. & Goudey, A. (2019). How Website Browsing Impacts Expectations of Store Features. International Journal of Retail & Distribution Management, 48(1), 92–108. Loureiro, S. M. C., Cavallero, L., & Miranda, F. J. (2018). Fashion Brands on Retail Websites: Customer Performance Expectancy and E-Word-of-Mouth. Journal of Retailing and Consumer Services, 41(March), 131–141. Luo, J., Fan, M., & Zhang, H. (2015). Information Technology, Cross-Channel Capabilities, and Managerial Actions: Evidence from the Apparel Industry. Journal of the Association for Information Systems, 17(5), 308–327. Lynch, J. G., Marmorstein, H., & Weigold, M. F. (1988). Choices from Sets Including Remembered Brands: Use of Recalled Attributes and Prior Overall Evaluations. Journal of Consumer Research, 15(2), 169–184. Mallapragada, G., Chandukala, S. R., & Liu, Q. (2016). Exploring the Effects of “What”(Product) and “Where”(Website) Characteristics on Online Shopping Behavior. Journal of Marketing, 80(2), 21–38. Martineau, P. (1958). The Personality of the Retail Store. Harvard Business Review, 36(1), 47–55.

References

301

Maslowsky, J., Jager, J., & Hemken, D. (2015). Estimating and Interpreting Latent Variable Interactions: A Tutorial for Applying the Latent Moderated Structural Equations Method. International Journal of Behavioral Development, 39(1), 87–96. Mavlanova, T., Benbunan-Fich, R., & Lang, G. (2016). The Role of External and Internal Signals in E-Commerce. Decision Support Systems, 87(July), 59–68. Maxham III, J. G. & Netemeyer, R. G. (2002). A Longitudinal Study of Complaining Customers’ Evaluations of Multiple Service Failures and Recovery Efforts. Journal of Marketing, 66(4), 57–71. Maydeu-Olivares, A. (2017). Maximum Likelihood Estimation of Structural Equation Models for Continuous Data: Standard Errors and Goodness of Fit. Structural Equation Modeling: A Multidisciplinary Journal, 24(3), 383–394. McDowell, W. C., Wilson, R. C., & Kile Jr., C. O. (2016). An Examination of Retail Website Design and Conversion Rate. Journal of Business Research, 69(11), 4837–4842. Melis, K., Campo, K., Breugelmans, E., & Lamey, L. (2015). The Impact of the Multi-Channel Retail Mix on Online Store Choice: Does Online Experience Matter? Journal of Retailing, 91(2), 272–288. Menon, G. & Raghubir, P. (2003). Ease-of-Retrieval as an Automatic Input in Judgements: A Mere-Accessibility Framework? Journal of Consumer Research, 30(2), 230–243. Mervis, C. B. & Rosch, E. (1981). Categorization of Natural Objects. Annual Review of Psychology, 32(1), 89–115. Mitchell, M. A. & Maxwell, S. E. (2013). A Comparison of the Cross-Sectional and Sequential Designs When Assessing Longitudinal Mediation. Multivariate Behavioral Research, 48(3), 301–339. Moliner-Velázquez, B., Fuentes-Blasco, M., Servera-Francés, D., & Gil-Saura, I. (2019). From Retail Innovation and Image to Loyalty: Moderating Effects of Product Type. Service Business, 13(1), 199–224. Montoya-Weiss, M. M., Voss, G. B., & Grewal, D. (2003). Determinants of Online Channel Use and Overall Satisfaction with a Relational, Multichannel Service Provider. Journal of the Academy of Marketing Science, 31(4), 448–458. Morales, A., Kahn, B. E., McAlister, L., & Broniarczyk, S., M. (2005). Perceptions of Assortment Variety: The Effects of Congruency between Consumers’ Internal and Retailers’ External Organization. Journal of Retailing, 81(2), 159–169. Mosquera, A., Olarte-Pascual, C., Ayensa, E. J., & Murillo, Y. S. (2018). The Role of Technology in an Omnichannel Physical Store: Assessing the Moderating Effect of Gender. Spanish Journal of Marketing, 22(1), 63–82. Murfield, M., Boone, C. A., Rutner, P., & Thomas, R. (2017). Investigating Logistics Service Quality in Omni-Channel Retailing. International Journal of Physical Distribution & Logistics Management, 47(4), 263–296. Murray, J., Elms, J., & Teller, C. (2017). Examining the Role of Store Design on Consumers’ Cross-Sectional Perceptions of Retail Brand Loyalty. Journal of Retailing and Consumer Services, 38(September), 147–156. Muthén, B. (2012). Latent Variable Interactions. http://www.statmodel.com. Accessed November, 13, 2018. Nagase, M. & Kano, Y. (2017). Identifiability of Nonrecursive Structural Equation Models. Statistics & Probability Letters, 122(March), 109–117.

302

References

Oh, L.-B. & Teo, H.-H. (2010). Consumer Value Co-Creation in a Hybrid Commerce ServiceDelivery System. International Journal of Electronic Commerce, 14(3), 35–62. Oh, L.-B., Teo, H.-H., & Sambamurthy, V. (2012). The Effects of Retail Channel Integration through the Use of Information Technologies on Firm Performance. Journal of Operations Management, 30(5), 368–381. Oliver, R. L. (1999). Whence Consumer Loyalty? Journal of Marketing, 63(4), 34–44. Pan, Y., Wu, D., & Olson, D. L. (2017). Online to Offline (O2O) Service Recommendation Method Based on Multi-Dimensional Similarity Measurement. Decision Support Systems, 103(November), 1–8. Park, C. W., Macinnis, D. J., Priester, J., Eisingerich, A. B., & Iacobucci, D. (2010). Brand Attachment and Brand Attitude Strength: Conceptual and Empirical Differentiation of Two Critical Brand Equity Drivers. Journal of Marketing, 74(6), 1–17. Patrício, L., Fisk, R. P., & Falcão e Cunha, J. (2008). Designing Multi-Interface Service Experiences: The Service Experience Blueprint. Journal of Service Research, 10(4), 318– 334. Patten, E., Ozuem, W., Howell, K., & Lancaster, G. (2020). Minding the Competition: The Drivers for Multichannel Service Quality in Fashion Retailing. Journal of Retailing and Consumer Services, 53(March), 1–9. Pauwels, K. & Neslin, S. A. (2015). Building with Bricks and Mortar: The Revenue Impact of Opening Physical Stores in a Multichannel Environment. Journal of Retailing, 91(2), 182–197. Planet Retail. (2019). Planet Retail Data. http://planetretail.net. Accessed May, 1, 2019. Puligadda, S., Ross, W. T. J., & Grewal, R. (2012). Individual Differences in Brand Schematicity. Journal of Marketing Research, 49(1), 115–130. Qiu, L., Pang, J., & Lim, K. H. (2012). Effects of Conflicting Aggregated Rating on eWom Review Credibility and Diagnosticity: The Moderating Role of Review Valence. Decision Support Systems, 54(1), 631–643. Rahimnia, F. & Hassanzadeh, J. F. (2013). The Impact of Website Content Dimension and E-Trust on E-Marketing Effectiveness: The Case of Iranian Commercial Saffron Corporations. Information & Management, 50(5), 240–247. Rajavi, K., Kushwaha, T., & Steenkamp, J.-B. E. M. (2019). In Brands We Trust? A Multicategory, Multicountry Investigation of Sensitivity of Consumers’ Trust in Brands to Marketing-Mix Activities. Journal of Consumer Research, 46(4), 651–670. Ravula, P., Bhatnagar, A., & Ghose, S. (2020). Antecedents and Consequences of CrossEffects: An Empirical Analysis of Omni-Coupons. International Journal of Research in Marketing, 37(3), 405–420. Roggeveen, A. L. & Sethuraman, R. (2020). How the COVID Pandemic May Change the World of Retailing. Journal of Retailing, 96(2), 169–171. Rose, S., Clark, M., Samouel, P., & Hair, N. (2012). Online Customer Experience in ERetailing: An Empirical Model of Antecedents and Outcomes. Journal of Retailing, 88(2), 308–322. Saghiri, S., Wilding, R., Mena, C., & Bourlakis, M. (2017). Toward a Three-Dimensional Framework for Omni-Channel. Journal of Business Research, 77(August), 53–67. Satorra, A. & Bentler, P. M. (2010). Ensuring Positiveness of the Scaled Difference Chi-Square Test Statistic. Psychometrika, 75(2), 243–248.

References

303

Schlosser, A. E., Barnett White, T., & Lloyd, S. M. (2006). Converting Web Site Visitors into Buyers: How Web Site Investment Increases Consumer Trusting Beliefs and Online Purchase Intentions. Journal of Marketing, 70(2), 133–148. Schramm-Klein, H., Wagner, G., Steinmann, S., & Morschett, D. (2011). Cross-Channel Integration—Is It Valued by Customers? International Review of Retail, Distribution and Consumer Research, 21(5), 501–511. Schwarz, N. (2004). Metacognitive Experiences in Consumer Judgment and Decision Making. Journal of Consumer Psychology, 14(4), 332–348. Seck, A. M. & Philippe, J. (2013). Service Encounter in Multi-Channel Distribution Context: Virtual and Face-to-Face Interactions and Consumer Satisfaction. Service Industries Journal, 33(6), 565–579. Shavitt, S. & Barnes, A. J. (2020). Culture and the Consumer Journey. Journal of Retailing, 96(1), 40–54. Shen, X.-L., Li, Y.-J., Sun, Y., & Wang, N. (2018). Channel Integration Quality, Perceived Fluency and Omnichannel Service Usage: The Moderating Roles of Internal and External Usage Experience. Decision Support Systems, 109(May), 61–73. Shin, J. I., Chung, K. H., Oh, J. S., & Lee, C. W. (2013). The Effect of Site Quality on Repurchase Intention in Internet Shopping through Mediating Variables: The Case of University Students in South Korea. International Journal of Information Management, 33(3), 453–463. Simmering, M. J., Fuller, C. M., Richardson, H. A., Ocal, Y., & Atinc, G. M. (2015). Marker Variable Choice, Reporting, and Interpretation in the Detection of Common Method Variance: A Review and Demonstration. Organizational Research Methods, 18(3), 473–511. Sirohi, N., McLaughlin, E. W., & Wittink, D. R. (1998). A Model of Consumer Perceptions and Store Loyalty Intentions for a Supermarket Retailer. Journal of Retailing, 74(2), 223–245. Sohn, S. (2017). Consumer Processing of Mobile Online Stores: Sources and Effects of Processing Fluency. Journal of Retailing and Consumer Services, 36(May), 137–147. Sousa, R. & Amorim, M. (2018). Architectures for Multichannel Front-Office Service Delivery Models. International Journal of Operations & Production Management, 38(3), 828–851. Spiller, S. A., Fitzsimons, G. J., Lynch, J. G., & McClelland, G. H. (2013). Spotlights, Floodlights, and the Magic Number Zero: Simple Effects Tests in Moderated Regression. Journal of Marketing Research, 50(2), 277–288. Srinivasan, S. S., Anderson, R. E., & Ponnavolu, K. (2002). Customer Loyalty in E-Commerce: An Exploration of Its Antecedents and Consequences. Journal of Retailing, 78(1), 41–50. Sullivan, Y. W. & Kim, D. J. (2018). Assessing the Effects of Consumers’ Product Evaluations and Trust on Repurchase Intention in E-Commerce Environments. International Journal of Information Management, 39(April), 199–219. Swoboda, B., Berg, B., & Schramm-Klein, H. (2013). Reciprocal Effects of the Corporate Reputation and Store Equity of Retailers. Journal of Retailing, 89(4), 447–459. Swoboda, B., Weindel, J., & Schramm-Klein, H. (2016). Crosswise and Reciprocal Interdependencies within Retailers’ Multichannel Structures. The International Review of Retail, Distribution and Consumer Research, 26(4), 347–374.

304

References

Swoboda, B. & Winters, A. (2020). Vollständige Integration – Touchpoints über alle Kanäle. In Handelsmonitor: Mega-Trends 2030+. Der Handel auf dem Weg in ein neues Zeitalter, Thomas Foscht, Dirk Moreschett and Hanna Schramm-Klein (eds). Frankfurt/Main: Deutscher Fachverlag, 37–65. Tagashira, T. & Minami, C. (2019). The Effect of Cross-Channel Integration on Cost Efficiency. Journal of Interactive Marketing, 47(August), 68–83. Teichert, T. A. & Schöntag, K. (2010). Exploring Consumer Knowledge Structures Using Associative Network Analysis. Psychology & Marketing, 27(4), 369–398. Teller, C. & Reutterer, T. (2008). The Evolving Concept of Retail Attractiveness: What Makes Retail Agglomerations Attractive When Customers Shop at Them? Journal of Retailing and Consumer Services, 15(3), 127–143. Toufaily, E. & Pons, F. (2017). Impact of Customers’ Assessment of Website Attributes on E-Relationship in the Securities Brokerage Industry: A Multichannel Perspective. Journal of Retailing and Consumer Services, 34(September), 58–69. Toufaily, E., Souiden, N., & Ladhari, R. (2013). Consumer Trust toward Retail Websites: Comparison between Pure Click and Click-and-Brick Retailers. Journal of Retailing and Consumer Services, 20(6), 538–548. Valentini, S., Neslin, S. A., & Montaguti, E. (2020). Identifying Omnichannel Deal Prone Segments, Their Antecedents, and Their Consequences. Journal of Retailing, 96(3), 310– 327. Van Baal, S. (2014). Should Retailers Harmonize Marketing Variables across Their Distribution Channels? An Investigation of Cross-Channel Effects in Multi-Channel Retailing. Journal of Retailing and Consumer Services, 21(6), 1038–1046. Van Bruggen, G. H., Antio, K. D., Jap, S. D., Reinartz, W. J., & Pallas, F. (2010). Managing Marketing Channel Multiplicity. Journal of Service Research, 13(3), 331–340. Van de Schoot, R., Lugtig, P., & Hox, J. (2012). A Checklist for Testing Measurement Invariance. European Journal of Developmental Psychology, 9(4), 486–492. Van Nierop, J. E. M., Leeflang, P. S. H., Teerling, M. L., & Huizingh, K. R. E. (2011). The Impact of the Introduction and Use of an Informational Website on Offline Customer Buying Behavior. International Journal of Research in Marketing, 28(2), 155–165. Verhagen, T. & van Dolen, W. (2009). Online Purchase Intentions: A Multi-Channel Store Image Perspective. Information & Management, 46(2), 77–82. Verhagen, T., van Dolen, W., & Merikivi, J. (2019). The Influence of In-Store Personnel on Online Store Value: An Analogical Transfer Perspective. Psychology & Marketing, 36(3), 161–174. Verhoef, P. C., Kannan, P. K., & Inman, J. J. (2015). From Multi-Channel Retailing to OmniChannel Retailing: Introduction to the Special Issue on Multi-Channel Retailing. Journal of Retailing, 91(2), 174–181. Verhoef, P. C., Langerak, F., & Donkers, B. (2007). Understanding Brand and Dealer Retention in the New Car Market: The Moderating Role of Brand Tier. Journal of Retailing, 83(1), 97–113. Verhoef, P. C., Neslin, S. A., & Vroomen, B. (2007). Multichannel Customer Management: Understanding the Research-Shopper Phenomenon. International Journal of Research in Marketing, 24(2), 129–148.

References

305

Vernuccio, M., Barbarossa, C., Giraldi, A., & Ceccotti, F. (2012). Determinants of E-Brand Attitude: A Structural Modeling Approach. Journal of Brand Management, 19(6), 500– 512. VHB. (2020). Liste der Fachzeitschriften in VHB-Jourqual 3. https://vhbonline.org/vhb4you/ vhb-jourqual/vhb-jourqual-3/gesamtliste. Accessed October, 20, 2020. Vyt, D., Jara, M., & Cliquet, G. (2017). Grocery Pickup Creation of Value: Customers’ Benefits vs. Spatial Dimension. Journal of Retailing and Consumer Services, 39(November), 145– 153. Wang, K. & Goldfarb, A. (2017). Can Offline Stores Drive Online Sales? Journal of Marketing Research, 54(5), 706–719. Wang, S., Beatty, S. E., & Mothersbaugh, D. L. (2009). Congruity’s Role in Website Attitude Formation. Journal of Business Research, 62(6), 609–615. Weindel, J. (2016). Retail Brand Equity and Loyalty: Analysis in the Context of Sector-Specific Antecedents, Perceived Value, and Multichannel Retailing. Wiesbaden: Gabler. Wells, J. D., Parboteeah, V., & Valacich, J. S. (2011). Online Impulse Buying: Understanding the Interplay between Consumer Impulsiveness and Website Quality. Journal of the Association for Information Systems, 12(1), 32–56. Wen, N. & Lurie, N. H. (2019). More Than Aesthetic: Visual Boundaries and Perceived Variety. Journal of Retailing, 95(3), 86–98. White, R. C., Joseph-Mathews, S., & Voorhees, C. M. (2013). The Effects of Service on Multichannel Retailers’ Brand Equity. Journal of Services Marketing, 27(4), 259–270. Wiedermann, W. & von Eye, A. (2015). Direction of Effects in Mediation Analysis. Psychological Methods, 20(2), 221–244. Wiener, M., Hoßbach, N., & Saunders, C. (2018). Omnichannel Businesses in the Publishing and Retailing Industries: Synergies and Tensions between Coexisting Online and Offline Business Models. Decision Support Systems, 109(May), 15–26. Williams, L. J., Hartman, N., & Cavazotte, F. (2010). Method Variance and Marker Variables: A Review and Comprehensive CFA Marker Technique. Organizational Research Methods, 13(3), 477–514. Williams, L. J. & McGonagle, A. K. (2016). Four Research Designs and a Comprehensive Analysis Strategy for Investigating Common Method Variance with Self-Report Measures Using Latent Variables. Journal of Business and Psychology, 31(3), 339–359. Wu, W.-Y., Lee, C.-L., Fu, C.-S., & Wang, H.-C. (2013). How Can Online Store Layout Design and Atmosphere Influence Consumer Shopping Intention on a Website? International Journal of Retail & Distribution Management, 42(1), 4–24. Wu, W., Carroll, I. A., & Chen, P.-Y. (2018). A Single-Level Random-Effects Cross-Lagged Panel Model for Longitudinal Mediation Analysis. Behavior Research Methods, 50(5), 2111–2124. Xiao, L., Zhang, Y., & Fu, B. (2019). Exploring the Moderators and Causal Process of Trust Transfer in Online-to-Offline Commerce. Journal of Business Research, 98(May), 214– 226. Yang, S., Lu, Y., & Chau, P. Y. K. (2013). Why Do Consumers Adopt Online Channel? An Empirical Investigation of Two Channel Extension Mechanisms. Decision Support Systems, 54(January), 858–869.

306

References

Yang, S., Lu, Y., Chau, P. Y. K., & Gupta, S. (2017). Role of Channel Integration on the Service Quality, Satisfaction, and Repurchase Intention in a Multi-Channel (Online-Cum-Mobile) Retail Environment. International Journal of Mobile Communications, 15(1), 1–25. Yang, Y., Gong, Y., Land, L. P. W., & Chesney, T. (2020). Understanding the Effects of Physical Experience and Information Integration on Consumer Use of Online to Offline Commerce. International Journal of Information Management, 51(April), 1–45. Ye, C., Hofacker, C. F., Peloza, J., & Allen, A. (2020). How Online Trust Evolves over Time: The Role of Social Perception. Psychology & Marketing, 37(11), 1539–1553. Yoo, B., Donthu, N., & Lee, S. (2000). An Examination of Selected Marketing Mix Elements and Brand Equity. Journal of the Academy of Marketing Science, 28(2), 195–211. Yrjölä, M., Spence, M. T., & Saarijärvi, H. (2018). Omni-Channel Retailing: Propositions, Examples and Solutions. The International Review of Retail, Distribution and Consumer Research, 28(3), 259–276. Zhang, J. & Bloemer, J. M. (2008). The Impact of Value Congruence on Consumer-Service Brand Relationships. Journal of Service Research, 11(2), 161–178. Zhang, M., Ren, C., Wang, G. A., & He, Z. (2018). The Impact of Channel Integration on Consumer Responses in Omni-Channel Retailing: The Mediating Effect of Consumer Empowerment. Electronic Commerce Research and Applications, 28(March-April), 181– 193. Zhao, X., Lynch Jr., J. G., & Chen, Q. (2010). Reconsidering Baron and Kenny: Myths and Truths About Mediation Analysis. Journal of Consumer Research, 37(2), 197–206. Zhou, T., Lu, Y., & Wang, B. (2009). The Relative Importance of Website Design Quality and Service Quality in Determining Consumers’ Online Repurchase Behavior. Information Systems Management, 26(4), 327–337. Zyphur, M. J., Allison, P. D., Tay, L., Voelkle, M. C., Preacher, K. J., Zhang, Z., Hamaker, E. L., Shamsollahi, A., Pierides, D. C., Koval, P., & Diener, E. (2019). From Data to Causes I: Building a General Cross-Lagged Panel Model (GCLM). Organizational Research Methods, 23(4), 651–687.