Testing the predictive power of PLS through cross-validation in banking

Depósito de investigación de la Universidad de Sevilla https://idus.us.es/ “This is an Accepted Manuscript of an article published by Elsevier in Journal of Busines Research on 2016, available at: https://doi.org/10.1016/j.jbusres.2016.04.016.” Testing the predictive power of PLS through cross-validation in banking Nuria Reguera-Alvarado, University of Seville Antonio Blanco-Oliver, University of Seville David Martín-Ruiz, University of Seville Submission: February, 2016 The authors are grateful for the comments by anonymous reviewers, Rafael Pino-Mejías (University of Seville) and Francisco Bravo (University of Seville), on prior drafts of this article. Send correspondence to Nuria Reguera-Alvarado, University of Seville, Faculty of Economics and Business, Department of Accounting, Av. Ramón y Cajal, s/n, 41018, Seville (Spain), e-mail: [email protected]. Con formato: Fuente de párrafo predeter. 1 Testing the predictive power of PLS through cross-validation in banking Abstract This paper aims to predict the relationship quality (satisfaction, trust, and commitment) of a customer with a Spanish saving bank by using both the quality of service operations and the usage of e-banking, on the relationship quality (satisfaction, trust, and commitment) of a customer with a Spanish saving bank. This paper aims to analyze the effects of the quality of service operations, combined with the usage of e-banking, on the relationship quality (satisfaction, trust, and commitment) of a customer with a Spanish saving bank. A competitive strategy follows testing three alternative models employing PLS path modeling with a primary dataset obtained through both an online survey and internal bank databases of almost one thousand customers. This study includes the predictive validation of models using hold out samples and testing for casual asymmetry. The results indicate that both the quality of service operations and the customer’s use of e-banking successfully and independently contribute to predict the customer’s relationship quality. The face-to-face service encounter is still the major contributor of relationship quality, but the use of e-banking tools also improves the relationship quality of customers in the banking industry. This model is consistent across a multi-group of customers. Keywords: banking; quality of service operations; relationship quality; mediation analysis; e-banking. Con formato: Fuente: Sin Negrita 2 Banks are constantly seeking new ways to add value to their services because financial services compete in a global marketplace with generally undifferentiated products. In this regard, technological changes are causing banks to rethink their strategies for services offered to customers, providing new ways to differentiate themselves from the competition (Hossain & Shirely, 2010). The process of developing and enhancing relationships traditionally takes place through personal interaction. However, the information revolution provides alternative ways of creating and maintaining relationships with customers (Rust & Kannan, 2003). Indeed, the impact of technology on relationship quality (RQ) brings new opportunities for researchers and practitioners alike (Lang & Colgate, 2003). The former one channel branch network has changed to a “multi-channel” configuration where branches coexist with other service delivery channels (Martínez, Ortega & Román, 2007). By developing technological channels, banks can lower costs, attract new customers, and increase income from existing customers. At the same time, financial customers gain in convenience, since they can carry out operations without time and place constraints, as well as obtaining lower commissions or better interest rates. Nevertheless, activities to promote ebanking should take into account that most customers do not rely on a single channel, but a combination of channels (Martínez et al., 2007). Some evidence exists of the role that RQ plays in adopting e-banking. For instance, Rotchanakitumnuai and Speece (2003) find that trust is the key factor influencing the adoption of electronic banking. Similarly, as satisfaction has a significant impact on trust and commitment, they altogether enhance the likelihood to use ebanking (Rexha, Kingshott & Aw, 2003). On the other hand, customers using e-banking represent a very attractive segment as they own more financial products than traditional customers. How does this usage level of e-banking affect the customer relationship with the bank? What is the relative influence of face-to-face operations when a multichannel setting is in place? Do significant differences 3 exist between customers showing high levels of usage of electronic banking and those showing low levels of usage of this channel? Do any other differences exist among customers due to socio-demographic or economic variables? Actually, few authors have investigated whether the usage of IT-based channels may have detrimental or positive effects on a firm’s RQ with its customers (Lang & Colgate, 2003). The main research objective of this paper is to analyze the predictive role that of both the use of e-banking and in combination with the quality of face-to-face service operations have in the prediction of the customers’ RQ variables (satisfaction, trust, commitment). To address this research aim, the proposal of three alternative models follows a competitive strategy. The first model considers the customer’s level of usage of e-banking as a mediator between the quality of service operations and RQ. The second model explores if the customer’s level of usage of e-banking moderates the relationship between the quality of service operations and RQ. Finally, the third model examines what is the independent effect of the customer’s level of usage of e-banking and the quality of service operations on RQ. The methodology employs PLS path modeling with a primary dataset provided by a major savings bank in Spain, including the predictive validation of models using hold out samples and testing for casual asymmetry. To carry out the above objective this paper used partial least squares (PLS) path modeling (Henseler, Hubona & Ray, 2016), including the predictive validation of models using 10-fold cross-validation procedure and testing for casual asymmetry. Through this research design and procedures this study seeks foster and enhance the use of PLS from a predictive point of view. 4 2.1. Service qQuality and the use of e-banking Perceived quality is one of the highly debated topics in marketing theory (Spreng & Mackoy, 1996). In today’s world, the key to sustainable competitive advantage for service companies lies in delivering high quality service that will result in satisfied and loyal customers (Shemwell, Yavas & Bilgin, 1998), because service quality is an essential pillar in service value creation (Martín, Gremler, Washburn & Cepeda, 2008). Since the competition finds it difficult to imitate service quality (Parasuraman & Grewal, 2000), this component represents a source of differentiation (Berry, 1995) and competitive advantage (Reichheld & Sasser, 1990). Banks delivering superior service quality can have a distinct marketing edge since improved levels of service quality yield higher revenues, increased cross-selling ratios, higher customer retention (Bennett & Higgins, 1988), and an expanded market share (Bowen & Hedges, 1993). Banks should focus on delivering service quality as a core competitive strategy (Chaoprasert & Elsey, 2004) in order to improve customer loyalty (Newman, 2001), and attract new customers through positive word of mouth (Wang, Lo & Hui, 2003). Meanwhile, the recent technological revolution has changed the way to access many services and, in consequence, how customers evaluate service quality through electronic channels. In this sense, researchers nowadays try to conceptualize and measure electronic service quality in an attempt to understand how customers perceive and evaluate online services (Barrera & Cepeda, 2014).People currently use Information Technology in almost every field of life including the banking sector (Rasheed & Latif, 2011). A definition of eservice quality is “the customer’s overall evaluation and judgment of the excellence and quality of e-service offering in the virtual marketplace” (Santos, 2003). When face-to-face encounters are difficult to arrange, the relationship develops through technological interfaces. 5 Studies show that availability, efficiency fulfillment, and privacy positively influence the intention to use e-banking (Parasuraman, 2000; Liao & Cheung, 2002; Akinci, Aksoy & Atilgan, 2004). The physical separation of the bank office and the customer, as well as the overall environment of perceived insecurity when carrying out electronic transactions on the Internet, provide unique challenges to online banks in the developing of e-business relationships. Indeed, the use of technology also raises concerns of confidentiality, anxiety, and stress for some customers (Bitner, Brown & Meuter, 2000). 2.2. Predicting rRelationship quality Given that the core service offered by banks may be essentially the same, developing longterm relationships with customers can provide successful differentiation. In financial services, the view of loyalty is in relation to the length of time a customer remains with its service provider, the number of services used and the frequency of service usage (Fragata & Moustakas, 2013). By keeping a loyal base of customers, banks obtain three benefits. Firstly, the cost of acquiring customers should decrease, since replacing deserters with new customers as a result of word-of-mouth is cheaper than attracting customers through marketing efforts. Secondly, loyal customers are less price-sensitive, which enables the obtaining of higher profit margins if necessary. Finally, a loyal customer is likely to respond better to a bank’s cross-selling strategies, generating more profits for the company (Barcellos, Fossati & Silva, 2009). Banks understand that customers are likely to be loyal if they receive a superior value from them than from their competitors (Dawes & Swailes, 1999). On the other hand, banks can obtain higher profits if they are able to position themselves better than the competition within a specific market (Davies, Moutinho & Curry, 1995). To enhance loyalty, banks must satisfy customer demands, which are increasing given the growing competition in internet banking. 6 Indeed, both the banks providing e-banking services and their customers can benefit from a high quality relationship (Alawneh, 2012). The focus of RQ is on long-term relationships rather than on short-term transactions. RQ refers to the overall assessment of the strength of a relationship between two parties (Palmatier, Dant, Grewal & Evans, 2006). Commonly, RQ represents a multidimensional construct, and a high-order combination of customer satisfaction, trust and commitment (Hennig-Thurau, Gwinner & Gremler, 2002; Ulaga & Eggert, 2006). According to Garbarino and Johnson (1999), satisfaction, trust, and commitment have different roles in determining the future of the customer relationship. Satisfaction is “an evaluation of the perceived discrepancy between prior expectations and the actual performance of the product” (Oliver, 1999). The satisfaction component of RQ is about the users’ evaluation of the relationship with the service provider. Similarly, dissatisfaction among customers using e-banking services might happen due to technological failure, technology design problems, or service design problems, which may result in a negative perception of the service’s functional quality (Meuter, Ostrom, Roundtree & Bitner, 2000). Dissatisfied customers are more prone to defect than satisfied customers (Barcellos et al., 2009). Customer commitment is the enduring desire to maintain a valued relationship (Palmatier et al., 2006). A committed customer invests time and effort in the relationship with a service provider, shows affection for this relationship, and is willing to maintain it in the future. Committed customers are more likely to become long-term loyal customers, since a committed partner wants the relationship to endure indefinitely and is willing to work at maintaining it. Commitment has an instrumental component of some sort of investment, an attitudinal component or psychological attachment, and a temporal dimension indicating that the relationship exists over time (Morgan & Hunt, 1994). As commitment entails 7 vulnerability, the parties only seek trustworthy partners. Trust is the cornerstone of a strategic partnership. Trust exists when one party has confidence in an exchange partner’s reliability and integrity (Morgan & Hunt, 1994). The literature on trust suggests that a trustworthy party is reliable, consistent, competent, honest, fair, responsible, helpful and benevolent. As a result, trust is an important indicator of RQ, since a relationship where one party is not willing to rely on its partner is limited (Moorman, Deshpandé & Zaltman, 1993). Trust is difficult to foster, and can be extremely hard to build after being damaged (Schneidermann, 2000). In the context of this study, trust refers to the customer’s willingness to rely on the bank providing e-banking services to perform several transactions. Privacy is a key element in building trust in e-banking services (Kim, Ferris & Rao, 2009). 2.3. Model development The major objective of this research is to combine the quality of service operations (i.e., the service delivery process at the bank’s office) with the usage level of e-banking services, in order to examine and predict the joint effect on the relationship between the customer and the bank. The combination of both channels should improve the prediction of the RQ between them. In order to do so, three alternative models consider a different role of the usage level of e-banking (LUEB). According to Figure 1, the first model (model A) examines LUEB as a mediating variable in the relationship between quality of service operations (QSOP) and a customer’s RQ. The second model (model B) considers LUEB as a moderator in the focal relationship between QSOP and RQ. Finally, the third model (model C) represents both LUEB and QSOP as independent variables influencing RQ. The first step is to establish which model has the stronger predictive power of the 14 4.3. Analysis of the model’s predictive capacity This work applies a 10-foldcross-validation by using R software to analyze the PLS model’s predictive capability. Table 6 summarizes the results of the validation procedure in terms of GoF, R2inner, and R2RQ (Henseler & Sarstedt, 2013). In accordance with Table 6, the accuracy performance of each one of the three PLS models in each test sub-sample is similar to the training sample, which supports the robustness and stability of their predictive capacity and highlights the theoretical and methodological relevance of the findings. According to Table 6, the average accuracy ability of Model A is 55.7% (std. dev. = 0.05), 68.9% (std. dev. = 0.09) and 80.7% (std. dev. = 0.08) in terms of GoF, R2inner, and R2RQ, respectively. Nevertheless, Model C is the one which obtains the best accuracy performance: average GoF = 55.8% (std. dev. = 0.05), average R2inner=69.0% (std. dev. = 0.09), and average R2RQ=81.1% (std. dev. = 0.08). Thus, before implementing this validation procedure, Model C obtained the best predictive capacity, its continuous performance being better than Model A’s under this approach. The implications from a methodological view of the results of the validation procedure are especially important, since they enhance the use of the PLS method with a predictive orientation due to its high predictive performance (higher than 80% in terms of the R-squared of the dependent construct) and contribute to foster the utility of PLS as a predictive statistical method. [Table 6 here] 4.4. Testing the causal asymmetry Although this issue is not central to the paper, in order to confirm and complement the PLS results with respect to the relationship between the constructs, this study implements a fuzzyset qualitative comparative analysis (fsQCA) (Fiss, 2011; Woodside, 2013) since set-theoretic 15 approaches are the means to gain deeper insights into management issues (Seny Kan, Adegbite, Omari & Abdellatif, 2015). In this sense, the analysis of the data using fsQCA allows exploring the multipath relationship in greater detail, showing how a certain set of conditions (endogenous constructs, QSOP and LUEB) relate to the outcome of interest (the exogenous construct, RQ). The fsQCA really tests the asymmetric causality through the use of the software package fsQCA 2.5., which requires several steps. The first step demands transforming the scale data (the unstandardized latent variable scores) to fuzzy-set membership values from 0 (full non-membership) to 1 (full membership) using the calibrate procedure (Ragin, 2008). Considering (i) the cross-over point as the median of the unstandardized latent variable scores, (ii) the threshold for full membership at the 95th percentile, and (iii) the threshold for full non-membership at the 5th percentile, Table 7 shows the results of the truth-table for presence/absence (complex, parsimonious, and standard solutions). Moreover, this table suggests that each one of the two endogenous constructs (conditions) is a core condition (Fiss, 2011), that is, a condition that has a strong link to the exogenous construct (outcome), which leads to a high RQ fit with single and overall consistencies higher than 0.80. According to the values of both consistency and coverage indexes, the findings confirm the fit from causal asymmetry analysis, which supports the predictive capacity of both QSOP and LUEB for RQ in the banking industry. [Table 7 here] 4.5. PLS mMulti-group analysis According to Henseler, ringle & Sarstedt (2015), the second analysis phase consists of a multi-group analysis using PLS-SEM. The MICOM (measurement invariance of composite models) procedure provides the method for studying the invariance prior to the multi-group 16 analysis. After confirming the existence of invariance, the next step is to apply the multigroup analysis, comparing the explained variance for each group. Therefore, tThe present research also tests the existence of significant differences across several clusters by using the PLS multi-group analysis. The multi-group study follows Model C due to its better performance according to the results of the previous sections. The first analysis studies the relationships between QSOP-RQ and LUEB-RQ through the comparison of the gender-based differences among customers (women or men). Table 8 shows that no statistically significant differences between women and men regarding the QSOP-RQ path exist. In contrast, the results show that the influence of the usage of e-banking on RQ is greater in men, but this difference is not statistically significant at 10% either (pvalue 0.11). The second multi-group analysis considers the place where the office is located (city center or suburbs) in order to seek differences of behavior between banking customers. The findings establish that the influence of QSOP on RQ is more important for the customers from the city center, this difference being significant at 5%. Conversely, the effect of LUEB on RQ is lower for the city center customers than those of suburban offices. Finally, the last multi-group compares the big savers group with the rest of the customers. For this cluster, the use of e-banking is more limited since their banking operations require personalized advice and preferential treatment from investment advisors in order to know the investment alternatives that the bank can offer them. Accordingly, LUEB has less influence on RQ than for the rest of the banking customers, this difference being high and substantially significant (p-value 0.00). [Table 8 here] 17 5. Concluding Remarks, Discussion, and Limitations A business’ key resource to achieve success lies in its customers. Firms need to develop sustainable competitive advantages to satisfy and keep loyal and profitable customers. Technological advances have caused a huge expansion of Internet use, which is changing the way to access many services and influencing how the relationships between firms and customers develop. In particular, the use of e-banking in the banking industry represents an important mechanism to improve the value of the service through higher convenience and lower prices. E-banking allows customers to perform many operations without having to be physically at the bank office, to use a service which is always open and has no waiting times. In addition, this channel enables lower prices because of its cost-efficiency and increased service productivity. All the same, the traditional channel remains active for a majority of customers. This study focuses on predicting the RQ in a bank with a multichannel setting. To do so, this research considers both the quality of service operations and the customer’s level of usage of e-banking as predictors of RQ, testing three alternative models. In this vein, Model A considers the mediating effect that LUEB has on the relationship between QSOP and RQ, while Model B tests a moderating influence of LUEB in this relationship. Model C analyzes the direct effect of both QSOP and LUEB on RQ. According to each model’s predictive capacity, the results show that Model C achieves the highest accuracy performance in both the model developing and the validation step. So, from a methodological point of view, Model C outperforms the other two models in terms of R2 and GOF. Additionally, from a theoretical perspective these findings confirm that in banking both QSOP and LUEB are two relevant and independent predicting variables of RQ. They are different channels to access the service. Therefore, the main implication of this paper is for banks; they must substantially foster both the quality of face-to-face service operations and the use of e-banking to improve the quality 18 of their relationship with customers. This research contributes to the literature in three ways. First, results show that the quality of service operations successfully predicts the level of a customer’s RQ. This finding is in consonance with the previous literature which indicates that banks can build satisfaction, trust and commitment from continuous successful service encounters with customers. The weight of QSOP on RQ is similar to that of previous research addressing this construct as an endogenous variable (Miguel-Dávila et al., 2010). Secondly, the findings also suggest that the customer’s use of e-banking significantly influences the development of RQ with the bank. Consequently, the use of technology is an important instrument to improve RQ with those customers who are more likely to use this channel. In this regard, the results of the multi-group analysis show that for big savers the use of e-banking does not influence their RQ, as they require the power of personalized attention in banking services. This result also confirms the idea that there are different clusters of customers in terms of needs, behaviors and motivation when it comes to the use of e-banking (Martínez et al., 2007). Thirdly, from a methodological point of view, these findings support the use of PLS as a predictive method, obtaining a powerful accuracy performance of the exogenous construct (RQ in banking). Additionally, this article also advances in the implementation of a validation procedure (10-fold cross-validation) in the R environment in order to test the robustness of the model’s accuracy ability. This use of R software and a cross-validation procedure in order to fit a predictive-oriented PLS model is new. Finally, the study has some limitations that also lead to future lines of research. First, the data set does not contain information from customers who defected from the bank. In other words, actual customer loyalty is not available in the study; a very interesting line for future research would be to address how RQ predicts actual loyalty behavior in this context. 19 Secondly, as the study focuses on one bank, a generalization of the results is limited. Third, part of the data collection process uses an on-line survey (information regarding the customer’s perceptions) that leaves out those customers who do not use e-banking services, leading to partially-biased information. Fourth, the study focuses only on quality of service operations, meaning that other dimensions of service quality are beyond the scope of the study (the outcome of service encounters, the service environment, and the personal relationship). Including them, as well as other customer-based variables (i.e., timesensitivity), would enrich the study. Akinci, S., Aksoy, S., & Atilgan, E. (2004). Adoption of Internet Banking among Sophisticated Consumer Segments in an Advanced Developing Country. International Journal of Bank Marketing, 22(3), 212-232. Alawneh, A.A. (2012). Assessing the Dimensions of Relationship Quality in B2C E-Banking Services: An Empirical Comparative Study. International Journal of Computer Science Issues, 9 (6), 290-302. Al-Hawari, M., Hartley, N., & Ward, T. (2005). Measuring Banks’ Automated Service Quality: A Confirmatory Factor Analysis Approach. Marketing Bulletin, 16, 1-19. Barcellos, M.L., Fossati, K., & Silva, C.A. (2009). The Influence of Banking Service Customers’ Satisfaction Level on the Perception of Switching Costs and on Behavioural Loyalty. The Flagship Research Journal of International Conference of the Production and Operations Management Society, 2(1), 1-13. Barczak, G., Scholder, P., & Pilling, B.K. (1997). Developing Typologies of Customer Motives for Use of Technologically Based Banking Services. Journal of Business Research, 38, 131-139. 20 Barrera, R., & Cepeda, G. (2014), Simultaneous Measurement of Quality in Different Online Services. The Service Industries Journal, 34(2), 123-144. Bennett, D., & Higgins, M. (1988). Quality Means More Than Smiles. ABA Banking Journal, June, 46. Berry, Leonard L. (1995). On Great Service: A Framework for Action. New York: The Free Press. Bitner, M.J., Brown, S.W., & Meuter, M.L. (2000). Technology Infusion in Service Encounters. Journal of the Academy of Marketing Science, 28, 138-149. Bowen, J.W., & Hedges, R.B. (1993). Increasing Service Quality in Retail Banking. Journal of Retail Banking, 15, 21-8. Brady, M., & Cronin, J. (2001). Some New Thoughts on Conceptualizing Perceived Service Quality: A Hierarchical Approach. Journal of Marketing, 65(3), 34–49. Chaoprasert, C., & Elsey, B. (2004). Service Quality Improvement in Thai Retail Banking and its Management Implications. ABAC Journal, 24(1), 47-66. Chin, W.W. (2010). How to write up and report PLS analyses. In Handbook of Partial Least Squares: Concepts, Methods and Application. Esposito Vinzi, V.; Chin, W.W.; Henseler, J.; Wang, H. (Eds.), Springer, Germany, 645-689. Davies, F., Moutinho, L., & Curry, B. (1995). Construction and Testing of a Knowledge- Based System in Retail Bank Marketing. International Journal of Bank Marketing, 13(2), 4-14. Dawes, J., & Swailes, S. (1999). Retention Sans Frontiers: Issues For Financial Service Retailers. International Journal of Bank Marketing, 17(1), 36-43. Fiss, P. C. (2011). Building Better Causal Theories: A Fuzzy Set Approach to Typologies in Organization Research. Academy of Management Journal, 54(2), 393-420. Fragata, A., & Moustakas, E. (2013). Investigating the Determinants of e-Banking Loyalty for 21 Large Business Customers: Two Empirical Models. Journal of Economics, Business, and Management, 1(2), 204-208. Garbarino, E., & Johnson, M.S. (1999). The Different Roles of Satisfaction, Trust, and Commitment in Customer Relationships. Journal of Marketing, 63, 70-87. Hair, J. F. Ringle, C. M., & Arstedt, M. (2011). PLS-SEM: Indeed a silver bullet. Journal of Marketing Theory and Practice, 19(2), 137-149. Hennig-Thurau, T., Gwinner, K., & Gremler, D. (2002). Understanding Relationship Marketing Outcomes: an Integration of Relational Benefits and Relationship Quality. Journal of Service Research, 4 (3), 230-247. Henseler, J., Hubona, G., & Ray, P.A. (2016). Using PLS Path Modeling in New Technology Research: Updated Guidelines. Industrial Management & Data Systems, 116(1), 2-20. Henseler, J., Ringle, C.M., & Sarstedt, M. (2015). A New Criterion for Assessing Discriminant Validity in Variance-Based Structural Equation Modeling. Journal of the Academy of Marketing Science, 43(1), 115-135. Henseler, J., Dijkstra, T.K., Sarstedt, M., Ringle, C.M., Diamantopoulos, A., Straub, D.W., Ketchen, D.J., Hair, J.F., Hult, G.T.M., & Calantone, R.J. (2014). Common beliefs and reality about PLS: Comments on Rönkkö & Evermann (2013). Organizational Research Methods, 17(2), 182-209. Henseler, J., & Sarstedt, M. (2013). Goodness-Of-Fit Indices for Partial Least Squares Path Modeling. Computational Statistics, 28(2), 565-580. Hossain, M., & Sherloy, L. (2010). Customer Perception on Service Quality in Retail Banking In Middle East: The Case of Qatar. International Journal of Islamic and Middle Eastern Finance and Management, 2(4), 338-350. Iacobucci, D., Saldanha, N., & Deng, X. (2007). A Meditation on Mediation: Evidence That Structural Equations Models Perform Better Than Regressions. Journal of Consumer 22 Psychology, 17 (2), 140-154. Jarvis, C. B., MacKenzie, S. B., & Podsakoff, P. M. (2003). A Critical Review of Construct Indicators and Measurement Model Misspecification In Marketing And Consumer Research. Journal of Consumer Research, 30, 199-218. Kan, A.K., Adegbite, E., Omari, S.E., & Abdellatif, M. (2015). On the Use of Qualitative Comparative Analysis in Management. Journal of Business Research, 69(4), 1458-1463. Kim, D.J., Ferrin, D.L, & Rao, H.R. (2009). Trust and Satisfaction, Two Stepping Stones for Successful E-Commerce Relationships: A Longitudinal Exploration. Information Systems Research, 20, 2, 237-257. Lang, B., & Colgate, M. (2003). Relationship Quality, On-Line Banking, and the Information Technology Gap. International Journal of Bank Marketing, 21(1), 29-37. Liao, Z., & Cheung, M.T. (2002). Internet-based E-banking and Consumer Attitudes: an Empirical Study. Information & Management, 39, 283-295. McDougall, G., & Levesque, T. (1994). A Revised View of Service Quality Dimensions: an Empirical Investigation. Journal of Professional Services Marketing, 11, 189-209. Martin, D., Gremler, D., Washburn, J., & Cepeda, G. (2008). Service Value Revisited: Specifying A Higher-Order, Formative Measure. Journal of Business Research, 61, 1278– 1291. Martínez, M., Ortega, J.M. & Román, M.V. (2007). Application of the Latent Class Regression Methodology to the Analysis of Internet Use for Banking Transactions in the European Union. Journal of Business Research, 60, 137-145. Meuter, M., Ostrom, A., Roundtree, R., & Bitner, M. (2000). Self-Service, Technologies: Understanding Customer Satisfaction with Technology-Based Service Encounters. Journal of Marketing 64, 50-64. Miguel-Dávila, J.A., Cabeza-García, L., Valdunciel, L., & Flórez, M. (2010). Operations in 23 Banking: the Service Quality and Effects on Satisfaction and Loyalty. The Service Industries Journal, 30, 2163-2182. Moorman, C., Desphandé R., & Zaltman G. (1993). Factors Affecting Trust in Market Research Relationships. Journal of Marketing, 57, 81-101. Morgan R.M., & Hunt, S.D. (1994). The Commitment-trust Theory of Relationship Marketing. Journal of Marketing, 58, 20-38. Newman, K. (2001). Interrogating Servqual: a Critical Assessment of Service Quality Measurement in a High Street Retail Bank. International Journal of Bank Marketing, 19, 126-139. Oliver, R. (1999). Whence Customer Loyalty? Journal of Marketing, 63, 33-44. Palmatier, Robert W., Dant, R. P., Grewal, D., & Evans, K. R. (2006). Factors Influencing the Effectiveness of Relationship Marketing: A Meta-Analysis. Journal of Marketing, 70(4), 136-153. Parasuraman, A., Zeithaml, V.A., & Berry, L. (1991). Refinement and Reassessment of the Servqual Scale. Journal of Retailing, 6(4), 420-450. Parasuraman, A. (2000). Technology Readiness Index (TRI): A Multiple-Item Scale to Measure Readiness to Embrace New Technologies. Journal of Service Research, 2(4), 307- 320. Parasuraman, A. & Grewal, D. (2000). Serving Customers and Consumers Effectively in the Twenty-First Century: a Conceptual Framework and Overview. Journal of the Academy of Marketing Science, 28, 9-16. Podsakoff, P.M., MacKenzie, S.B., & Lee, J.Y. (2003). Common Method Biases in Behavioral Research: A Critical Review of the Literature and Recommended Remedies. Journal of Applied Psychology, 88(5), 879-903. Ragin, C.C. (2008). Redesigning Social Inquiry: Fuzzy Sets and Beyond (pp. 190-212). Con formato: Fuente: Cursiva 30 Table 5. Effects on endogenous variables for each PLS model Model A Effects on endogenous variables Theoretical sense (support) Direct effect (path coefficient) t-Value (bootstrap) Percentile 95% confidence intervals Explained variance LUEB (R2 = 81.40%) QSOP (a1) + (Yes) 0.76*** 33.60 [0.71;0.80] Sig 57.76% RQ (R2 = 57.90%) QSOP (c´) + (Yes) 0.75*** 25.42 [0.69;0.80] Sig 66.75% LUEB (b1) + (Yes) 0.19*** 5.96 [0.13;0.26] Sig 14.44% QSOP: Service quality; LUEB: Level of use of e-bank; RQ: Relationship quality.*p < 0.10; **p < 0.05; ***p <0.01 (based on t(4999), two-tailed test); t(0.10; 4999) = 1.65; t(0.05; 4999) = 1.96; t(0.01; 4999) = 2.58. Sig. denotes a significant direct effect at 0.10. Model C Effects on endogenous variables Theoretical sense (support) Direct effect (path coefficient) t-Value (bootstrap) Percentile 95% confidence intervals Explained variance RQ (R2 = 82.00%) QSOP (a1) + (Yes) 0.76*** 25.63 [0.70;0.81] Sig 68.40% LUEB (b1) + (Yes) 0.19*** 5.55 [0.12;0.25] Sig 14.44% QSOP: Service quality; LUEB: Level of use of e-bank; RQ: Relationship quality.*p < 0.10; **p < 0.05; ***p <0.01 (based on t(4999), two-tailed test); t(0.10; 4999) = 1.65; t(0.05; 4999) = 1.96; t(0.01; 4999) = 2.58. Sig. denotes a significant direct effect at 0.10. 31 Table 6. Results of 10-fold cross validation for each Model Model A k-fold R2inner Com R2average R2QSOP R2LUEB R2RQ 1 59.4% 0.40 46.4% - 45.8% 72.9% 2 77.2% 0.52 43.3% - 68.9% 85.6% 3 72.5% 0.48 45.8% - 62.8% 82.3% 4 64.1% 0.43 42.3% - 43.7% 84.5% 5 76.1% 0.51 46.1% - 63.1% 89.1% 6 76.7% 0.51 49.9% - 66.6% 86.8% 7 74.4% 0.50 46.8% - 63.4% 85.3% 8 57.7% 0.39 44.2% - 51.9% 63.4% 9 56.0% 0.37 41.3% - 40.4% 71.6% 10 75.0% 0.50 45.6% - 64.7% 85.4% Mean 68.9% 0.46 45.2% - 57.1% 80.7% Std. dev. 0.09 0.06 0.03 - 0.11 0.08 Model C k-fold R2inner Com R2average R2LUEB R2QSOP R2RQ 1 59.5% 0.40 46.3% - 45.8% 73.2% 2 76.9% 0.51 43.3% - 68.1% 85.8% 3 73.0% 0.49 45.8% - 63.1% 82.9% 4 64.7% 0.43 42.3% - 44.5% 84.9% 5 76.2% 0.51 46.1% - 63.0% 89.3% 6 77.0% 0.51 49.9% - 66.7% 87.3% 7 74.2% 0.50 46.9% - 62.6% 85.8% 8 57.7% 0.39 44.2% - 51.4% 64.1% 9 55.9% 0.37 41.3% - 39.9% 71.9% 10 75.0% 0.50 45.7% - 64.4% 85.7% Mean 69.0% 0.46 45.2% - 56.9% 81.1% Std. dev. 0.09 0.06 0.03 - 0.10 0.08 Tabla con formato Tabla con formato 32 Table 7. Truth-table fsQCA Raw coverage Unique coverage Consistency Quality of Service Operations (QSOP) 0.90 0.09 0.92 Level of use of e-bank (LUEB) 0.85 0.05 0.86 Note: frequency cutoff: 80.00; consistency cutoff: 0.83; solution coverage: 0.95; solution consistency: 0.85. 33 Table 8. Results of multi-group analysis Relationship Groups Comparison Path coefficients Difference P-value parametric P-value permutation QSOP → RQ Women & men Men 0.75 0.06 0.84 0.32 Women 0.81 City center & suburb offices City center 0.81 0.009 0.005 0.14 Suburb 0.72 Big savers & other customers Big savers 0.88 0.14 0.97 0.06 Other customers 0.74 LUEB → RQ Women & men Men 0.21 0.09 0.11 0.19 Women 0.11 City center & suburb offices City center 0.14 0.08 0.90 0.21 Suburb 0.22 Big savers & other customers Big savers 0.02 0.20 0.00 0.02 Other customers 0.22 QSOP → RQ LUEB → RQ WOMEN & MEN Path coefficients for men 0.75 (p-value 0.00) 0.21 (p-value 0.00) Path coefficients for women 0.81 (p-value 0.00) 0.11 (p-value 0.08) Path coefficients differences 0.06 (p-value 0.84) p-value permutation 0.32 0.09 (p-value 0.11) p-value permutation 0.19 CITY CENTER & SUBURB OFFICES Path coefficients for city center 0.81 (p-value 0.00) 0.14 (p-value 0.00) Path coefficients for suburbs 0.72 (p-value 0.00) 0.22 (p-value 0.08) Path coefficients differences 0.09 (p-value 0.05) 0.08 (p-value 0.90) 34 p-value permutation 0.14 p-value permutation 0.21 BIG SAVERS & OTHER CUSTOMERS Path coefficients for big savers 0.88 (p-value 0.00) 0.02 (p-value 0.76) Path coefficients for other customers 0.74 (p-value 0.00) 0.22 (p-value 0.08) Path coefficients differences 0.14 (p-value 0.97) p-value permutation 0.06 0.20 (p-value 0.00) p-value permutation 0.02 Figure 1. Structural Models Model A SRMR composite model = 0.05 VAF=16.14% *p<0.10; **p<0.05; ***p<0.01. QSOP: Quality of service operations; LUEB: Level of use of e-bank; RQ: Relationship quality.*p < 0.10; **p < 0.05; ***p <0.01 LUEB R2 = 57.90% % RQ R2 = 81.40% QSOP a1= 0.76*** c´= 0.75 *** b1 = 0.19*** Con formato: Izquierda 35 Model B SRMR composite model = 0.05 VAF=75% *p<0.10; **p<0.05; ***p<0.01. QSOP: Quality of service operations; LUEB: Level of use of e-bank; RQ: Relationship quality.*p < 0.10; **p < 0.05; ***p <0.01 Model C SRMR composite model = 0.05 VAF: Not available *p<0.10; **p<0.05; ***p<0.01. QSOP: Quality of service operations; LUEB: Level of use of e-bank; RQ: Relationship quality.*p < 0.10; **p < 0.05; ***p <0.01 LUEB RQ R2 = 82% QSOP c´= 0.74 *** LUEB RQ R2 = 82% QSOP c´= 0.76 *** b1 = 0.19*** QSOP x LUEB m = -0.02 b1 = 0.18*** Con formato: Izquierda, Sangría: Primera línea: 0 cm Con formato: Izquierda, Sangría: Primera línea: 0 cm 36