What drives consumers to use P2P payment systems? An analytical approach based on the stimulus–organism–response (S-O-R) model

3.1 S-O-R framework

Numerous behavioral models have been proposed in the previous scientific literature to explain the antecedents of intention to use and the behavior of the users regarding the adoption of different financial technologies as well as in mobile payment systems (Irimia-Diéguez et al., 2023; Liébana-Cabanillas et al., 2022). Most research has employed classical theories such as the theory of reasoned action (TRA) (Fishbein and Ajzen, 1977), technology acceptance model (TAM) (Davis, 1989), theory of planned behavior (TPB) (Ajzen, 1991), diffusion of innovation (DOI) (Rogersà Rogers, 2003), unified theory of acceptance and use of technology (UTAUT) (Venkatesh et al., 2003) as well as its subsequent modifications (Lee et al., 2022; Goel et al., 2022; Shaikh et al., 2023). Although previous studies have made important contributions in the field of mobile payments by applying these theories and have improved knowledge (Al-Okaily et al., 2023; Dash et al., 2023), we consider that the use of the S-O-R model would enrich both theory and practice in this field of study where, so far, few authors have used it as a reference in their research (Yuan et al., 2020).

The S-O-R model was proposed by Mehrabian and Russell (1974) and later used by Donovan and Rossiter (1982) to define the environment as a stimulus, which causes an evaluation by the consumer, and consequently some behavioral responses. This psychological model explains human behavior through a three-stage process as shown in Figure 1; the perceptions of external or environmental stimuli associated with a purchase decision (S) influence the individual's emotional and cognitive states resulting from perceptions, psychological factors, feelings and thoughts (O), which drive conscious or unconscious responses (R).

This model has been used in numerous investigations in many different scientific fields such as e-commerce (Le et al., 2022), website design (Molinillo et al., 2021), social commerce (Herzallah et al., 2021), showrooming (Shankar et al., 2021), online reviews based on text and photos using automatic processing patterns and conscious perceptions (Bigne et al., 2020) and even in the analysis of consumer behavior in payment systems (e.g., Yuan et al., 2020; Sun et al., 2021; Zhou et al., 2022).

3.2 Research model and hypotheses

4.1 Data collection

In this work, we use a primary source of data obtained from a survey made with 701 Spanish smartphone users, who are considered potential users of the P2P payment systems. All the users who participated in the survey had experience in using their cell phone for commercial activities, either for shopping or payment. To collect the data, non-probability snowball sampling was employed through a mailing list and social networks.

The final sample, after eliminating questionnaires that took too little time or exceeded the recommended time, amounted to 701 participants, of whom 46.22% were male and 53.78% were female. 42.37% were between 18 and 24 years old, 51.21% were between 25 and 44 years old and 6.28% were over 44 years old. 4.28% had doctoral studies, 49.93% had university studies, 26.68% had secondary school studies, 15.83% had primary school studies and the remaining 3.28% had no studies at all.

To confirm whether our sample was suitable, the inverse square root method (Kock and Hadaya, 2018) was applied. The minimum sample size required for this study is 249 (where the significance value = 0.05 and a minimum path coefficient = 0.10, potential power = 90%). Therefore, the estimation for sample size confirms that the sample size is adequate for this kind of research.

4.2 Measures

The data for the empirical evaluation of the research model were collected through a self-administered online survey. The constructs of the model were measured through reflective measurement scales validated in previous research and adapted to the proposed payment context. Subjective norms were measured using 4 items adapted from Agarwal and Prasad (1998) and Taylor and Todd (1995). Perceived quality was measured with 7 items from Parasuraman et al. (1988) and Lai et al. (2007). Personal innovativeness was measured with 4 items adapted from Agarwal and Prasad (1998) and Ramos-de-Luna et al. (2016). Perceived usefulness was measured with 4 items from the research of Bhattacherjee and Premkumar (2004). Perceived trust was measured with 5 items adapted from Pavlou (2002). Perceived risk was measured with 4 items adapted from Jarvenpaa et al. (2000) as well as Wakefield and Whitten (2006) Finally, intention to use was measured with 3 items adapted from Venkatesh and Bala (2008). In all the cases, 7-point Likert-type scales were used (1 = strongly disagree; 7 = strongly agree) (see Table A2). As the original scales were in English, and the target population was Spanish, in order to maintain the accuracy of the original scales, the questionnaire was translated into Spanish by a professional service so that the level of comprehension was guaranteed.

Finally, two focus groups were set up, the first with the heads of the payment methods departments of five Spanish financial institutions and the second with ten customers who regularly use mobile payment systems. After these meetings, the questionnaire was established and proposed to a group of users as a pre-test to analyze the validity and reliability of the proposed scales.

4.3 Statistical methodology

Due to the characteristics of the indicators obtained from the survey, we have the opportunity of building constructs that can be related to the indicators by using PLS-SEM. Due to its maximum interest and applicability, PLS-SEM has been employed and recommended in previous studies which focused on the prediction of the constructs and the complex interrelationships between latent variables (Hair et al., 2011). To this end, Smart PLS 3.2.7 software was used to conduct the research model (Ringle et al., 2015).

The analysis and interpretation of PLS results has two main stages. The first stage, the measurement model, determines whether the indicators and constructs have been measured correctly (outer model) and the second stage, the structural model (inner model), determines whether the relations between the constructs are significant or not.

5.1 Common method bias

Common method bias (CMB) happens when variations in responses are caused by the instrument rather than the actual predispositions of the respondents that the instrument attempts to uncover (Kock, 2015). A full collinearity test based on variance inflation factors (VIFs) was used to detect a potential CMB situation (Kock, 2015) following the guidelines described by Kock and Lynn (2012). Kock (2015) indicates that when a VIF achieves a value greater than 3.3, there would be an indication of pathological collinearity. The present model, with a maximum VIF of 2.352 (Table 1), may be considered free of CMB.

5.2 Measurement (outer) model assessment

The measurement model section assesses the reliability and validity to prove the feasibility of the individual items and constructs employed in the research model. The results, reported in Table 2, display that all the standard factor loadings are larger than 0.7 Thus, all the items are considered satisfactory. In addition, the descriptive statistics of all the items are shown in Table A1 in the Appendix. As can be observed, all the values show that the data are not extremely non- normal since the absolute values of skewness and kurtosis are under /2/.

As shown in Table 2, all the constructs comply with Cronbach's Alpha, Dijkstra–Henseler's rho (rho_A) and composite reliability (C.R.) due to having figures over the critical 0.7 level, showing that all the constructs have convergence or internal consistency. Finally, it can be observed that all the constructs satisfy the criterion of 0.5 points (equal to or above) demanded by the average variance extracted (AVE), fulfilling the convergent validity of the constructs and dimensions.

Nevertheless, it is also necessary to assess the discriminant validity of the constructs. Under the Fornell-Larcker framework, the AVE should be greater than the variance shared between the construct and other constructs in the model. As shown in Table 3, the diagonal elements are significantly greater than the off-diagonal elements in the corresponding rows and columns for all the constructs obtaining an adequate discriminant validity (Hair et al., 2021). Additionally, Table 3 also shows that all the constructs have a Heterotrait-Monotrait ratio of correlations (HTMT) value lower than 0.85 which indicates the existence of discriminant validity for all the constructs.

5.3 Structural (inner) model assessment

The evaluation of the structural model is based on the sign, magnitude and significance of the structural path coefficients and on the R2 values (Hair et al., 2021). In order to ascertain the statistical significance of the path coefficients, a bootstrapping procedure is performed with 5,000 resamples (Hair et al., 2011). A percentile bootstrap 95% confidence interval is also employed. As can be seen in Table 4, all the hypotheses were supported.

The results of the structural model reveal that all the hypotheses have significant relationships with their respective variables. The direct relationships are robust and acceptable because they have t-values greater than 0.226 and p-values lower than 0.10 except H8 whose level of significance is slightly lower. Due to this fact, the mediator effect of perceived quality in the relationship between personal innovativeness and perceived trust was analyzed as shown in Figure 3.

According to Nitzl et al. (2016), a complementary partial mediation of perceived quality in the relationship between personal innovativeness and perceived trust is found, due to the fact that both the direct effect (c´ = 0.110) and the indirect effect (a x b = 0.261) are significant and the sign of the mediation (a x b x c’) is positive. The strength of this partial mediation can be evaluated by calculating the VAF value which determines the extent to which the mediation process explains the dependent variable's variance. Since the VAF = 70.35% a partial mediation is confirmed.

Figure 4 shows the significance of the hypothesized relationships and the explanatory capacity of the proposed model. A large portion of the variance of the latent variance is explained. Additionally, the standardized root mean square residual (SRMR) value was 0.043 (lower than the maximum acceptable value of 0.08 recommended by Schuberth et al., 2022). Thus, the proposed model has goodness of fit.

5.4 Assessment of the predictive validity

This study also performs a predictive analysis to verify the predictive ability level of the proposed research model. According to Shmueli and Koppius (2011), this analysis has the capacity to generate accurate predictions of new observations, temporal, interpretable or transversal. The current PLS predict algorithm of the SmartPLS software version 3.2.7 (Ringle et al., 2015) was applied to value the proposed research model's predictive ability. The values obtained through these statistics are root mean squared error (RMSE), the mean absolute error (MAE) and the Q² predict values. These enable assessing the predictive performance of their PLS path model for the constructs (LV Prediction Summary) and indicators (MV Prediction Summary).

In the PLS predict routine, firstly a k-fold cross-validation was executed, setting k = 24 subgroups, with the aim of meeting the minimum size of N = 30 for the hold-out sample (Hair et al., 2021), repeating this procedure ten times. Since all the values of the skewness for prediction errors of result indicators were under /1/, both for the PLS and the LM analyses, RMSE was selected as a basis of the predictive power assessment. Hence, since over half of the RMSE values of PLS-SEM are lower than those of the LM analysis, a medium predictive relevance performance is found (Table 5).

5.5 Importance-performance map analysis (IPMA)

The Importance-Performance Map Analysis, also known as the Importance-Performance Matrix, is a complementary analysis to let us know which constructs and items are most important and have a higher performance when determining a construct (Ringle and Sarstedt, 2016). Thus, this analysis allows for prioritizing the constructs as well as the corresponding items to improve the targeted construct. In our current research model, IPMA was performed employing intention to use Paytech services as the target endogenous construct and this was established before executing the IPMA analysis.

Figures 5 and 6 display the diagrams that will let us analyze these relations from a more practical and intuitive approach. The first diagram focuses on the construct and the second one on the items of the proposed research model. To better understand the figures, it incorporates two supplementary lines that divide the importance-performance graph into four spaces or quadrants (Ringle and Sarstedt, 2016). The horizontal line represents the performance and the vertical line the importance represented by the average values of both dimensions. Hence, the graph has 4 quadrants that include the values above and below the mean: 1) Upper-right quadrant (+Importance; + Performance); 2) Lower-right quadrant ((+Importance; − Performance); 3) Lower-left quadrant (− Importance; − Performance) and 4) Upper-left quadrant (−Importance; + Performance).

As may be observed in Figure 5, the variables perceived usefulness, subjective norms and personal innovativeness are in quadrant 1 (upper-right). Therefore, this analysis reveals that these constructs have the greatest (namely, over the average) levels of importance and performance. They can be considered as key factors in influencing consumer intention to use Paytech services. The variables perceived quality and perceived trust are in quadrant 3 (upper-left) meaning that users are paying little attention to them although they have a high performance for the target construct.

In Figure 6, we can see which indicators are the most important and have the highest performance. All the items are in the same quadrant as their construct. Thus, we reach the same conclusion.

5.6 The moderating effect of perceived risk

In order to analyze the moderator effect of perceived risk on the relationship between usefulness and intention to use and trust and intention to use, the following procedure was applied. First, the measurement invariance assessment procedure (MICOM) (Henseler et al., 2016) was applied to determine whether, under different conditions of observation and study of the phenomena, the measurement models yield measurements of the same attribute. The MICOM procedure comprises three steps: (1) configural invariance, (2) compositional invariance and (3) the equality of composite mean values and variances.

Configurational invariance is the condition that both the items and the model in both groups are the same. If this condition is not met, it is not possible to conduct a multi-group analysis. Compositional invariance seeks to assess that respondents in both groups understood the same meaning to each question when answering the questionnaire. If this condition is met, “partial measurement invariance” is established, which is a sufficient condition for conducting a multi-group analysis. In our research, partial invariance is confirmed because steps 1 and 2 conform to the limits established in the literature, while step 3 does not find optimal values for all the variables. However, in practical applications, the full measurement invariance is often not fulfilled (Steenkamp and Baugarner, 1998).

Once the measurement invariance was verified, the structural model was estimated, using the PLS multigroup analysis method (Henseler et al., 2009). To this end, the sample was divided into two groups, users with low perceived risk and those with a high perceived risk. Table 6 shows the results of the test of differences between the two groups for the relationships proposed in H11. As shown in Table 6, both relationships differ significantly between the two groups created (p<0.10%). Therefore, the results indicate that the risk perception moderates the relationships between both usefulness and trust with the intention to use the proposed P2P payment system.

We observe that, on the one hand, users with high perceived risk have a lower intensity in the relationship between perceived usefulness and intention to use (β _{HIGH RISK} = 0.572; p = 0.000; β _{LOW RISK} = 0.684; p = 0.000) as they probably value other elements derived from their own experience. On the other hand, for users with a high perception of risk, the relationship between trust and intention will have a greater intensity than in the case of users with a low perception of risk (β _{HIGH RISK} = 0.298; p = 0.000; β _{LOW RISK} = 0.144; p = 0.012).

6.1 Conclusion

Fintech is considered by many researchers and practitioners as one of the most important innovations in the financial sector, due to its significant potential for change and disruption as well as its implications for all players in that sector (Bajwa et al., 2022). Within these technological improvements, the use of cell phones and their practical applications in the financial sector are crucial. In this vein, mobile applications have generated new opportunities, such as improvements in the provision of advice and customer services, access to almost any financial service and an increase in the number of PoS terminals due to the use of smartphones incorporating NFC technologies.

In this sense, as reported, mobile payment systems in general (Balachandran et al., 2022) and P2P in particular (Lara-Rubio et al., 2021), seriously threaten the use of banking and cash for everyday purchases. This fact should imply an effort in analyzing users' behavior as well as their background and barriers (Belanche et al., 2022).

Among the different P2P payment systems, it is proposed to analyze the case of Bizum due to the growing interest that this new P2P payment system has aroused not only among users and Fintech companies, but also in previous academic studies. Due to this huge interest, we consider the research carried out to be innovative and timely.

This research proposes the development of a conceptual framework using the S-O-R model, where the proposed stimuli are subjective norms, the perceived quality of Bizum itself, as well as the user's level of innovation, while the body is identified with perceived usefulness and perceived trust as antecedents of the response, identified as the intention to use Bizum.

6.2 Theoretical implications

Our contributions to the research on P2P payment systems are, firstly, to better understand the crucial role of P2P payment systems nowadays. Their economic and social significance in our society is highlighted due to the enormous development of smart phones and the advantages that their use entails.

Secondly, many different theories have been extensively applied in the adoption of technology but not yet in Fintech services adoption studies. Our work is one of the first to propose and empirically evaluate a behavioral model that uses the S-O-R framework to study the intention of individuals about the use of a specific P2P payment tool. Consequently, this is a contribution to the improvement of the scientific knowledge and the academic literature on consumer behavior.

Thirdly, this study also performs an analysis of the out-of-sample predictive power of the proposed research model. From a methodological view, not only explanatory analyses but also predictive studies on fintech adoption are required to validly extrapolate the results. Thus, our work seeks to advance within this emerging line of research.

Fourthly, this study improves the understanding of the effects on perceived usefulness and perceived trust through three fundamental aspects in the analysis of technology adoption: Subjective norms through the influence of the potential user's immediate environment of P2P payment systems, perceived quality through the potential user's evaluation of the tool itself, as well as the individual's level of personal innovativeness toward Bizum. The best antecedents on perceived usefulness and on perceived trust were shown to be, respectively, subjective norms and perceived quality, which reinforces the findings of previous works in other online contexts (i.e., Yasin et al., 2019; Gao and Waechter, 2017). The present study shows that both characteristics had a considerable influence and of a similar intensity. Thus, this research contributes to improving the understanding of the importance of both perceived usefulness and perceived trust by extending the results of previous studies to the P2P context. In addition, our results demonstrate the importance of perceived usefulness as the main antecedent of intention to use. This relationship corroborates the importance of this variable in innovations related to financial technologies (i.e., Singh et al., 2021; Saheb et al., 2022). Consequently, we can affirm that utility aspects have been decisive in assessing potential users' intention to use this type of payment systems. This result is also corroborated by the IPMA analysis where perceived usefulness, subjective norms and personal innovativeness are shown to have the greatest importance and performance.

Fifthly, the proposed model shows the moderating effect of perceived risk. Namely, this moderating effect was demonstrated in the two proposed relationships. Our results show that perceived usefulness has more influence on intention among users who have a lower perception of risk, whilst in the relationship between trust and intention, the effect is higher for users who have a higher perception of risk. This is an important contribution that improves the understanding of the effect of risk on payment systems and addresses a gap identified in previous studies. The results are in line with recent work that identified the effect of risk related to intention (Daragmeh et al., 2021; Mehta et al., 2021; Qalati et al., 2021; Poon and Tung, 2022).

Finally, the mediating effect of perceived quality on the relationship between personal innovativeness and perceived trust was analyzed. This conclusion is fundamental to understand the important role that the innovation of potential users plays in the adoption of a technology and, more specifically, the importance that it has on trust.

6.3 Managerial implications

This research also has important implications for the different sectors involved in the design, marketing and sale of these banking services. This P2P payment sector is a highly competitive market with great growth and potential; these financial solutions will reach out to the next generations since the majority are “digital natives” ready for innovative payment services. In addition, these P2P solutions are highly appreciated among end users due to involving a perception of belonging to a digital community. As a consequence of the networking effects, the various competitors are trying to win the race to dominate this “winners-take-all” market (Wirtz et al., 2018). Furthermore, the knowledge of the factors fostering the adoption of these payment services would allow an international expansion of the financial providers without the inconvenience of establishing offices or companies in other countries. Other significant advantage is that P2P services can promote financial inclusion by allowing access to financial services to the entire population regardless of their location (e.g. rural areas) or of their age (e.g. elder people).

Thus, financial providers need to identify the competitive advantages that will allow them to lead P2P services in their markets. To this vein, firstly, P2P payment systems need to improve usability among users by ensuring that the agility and speed of transactions is adequate through the applications designed for this purpose.

Secondly, in order to improve user confidence, it will be essential to pay attention to the perceived quality of the payment system analyzed, through the reliability, accuracy and information provided, which are fundamental aspects for the potential users of the payment tools studied. As trust is the second most important factor influencing take-up, promotional campaigns should also focus on activities aimed at building consumer trust.

Thirdly, the results showed significant differences in perceived risk in the relationships between usefulness and trust with intention to use. In this respect, it will be necessary for P2P payment tools to implement rigorous security measures to keep not only users' transactions protected but also their personal data by ensuring the integrity of the information.

Finally, companies must take into account the changes in consumer behavior that are occurring due to social changes in the face of the COVID-19 pandemic (Daragmeh et al., 2021; Liébana-Cabanillas et al., 2022). In recent months, new payment systems have been established to reduce the contact between buyer and seller, which has improved the growth prospects of contactless payment systems and P2P systems themselves, so companies should encourage their use because of the advantages that they offer compared to other traditional payment systems.

6.4 Limitations and further research

Like any research, this study also has some limitations that can be considered a justified starting point for future research. Firstly, with regard to the sample, only cell phone users with experience of purchasing through these terminals were used, but their experience with payment systems prior to P2P payment systems was not assessed. It would be interesting to contrast the effect of usage experience in future research and to analyze its importance in the decision process for a continuous use, as well as to review the drivers and barriers between users and non-users in a more exhaustive way.

Additionally, the research was conducted at a single point in time, which means that the study lacks the dynamism of including observations and measuring them at consecutive time intervals. To overcome this limitation, it is proposed to carry out a longitudinal study and to analyze over time the variations that may occur. A longitudinal approach would make it possible to check the robustness of the relationships and constructs established and to verify, from a temporal perspective, the evolution of the effects of the variables analyzed.

Regarding the proposed theoretical framework, in spite of being adequate due to the empirical evidence that numerous studies have obtained from it, it would be interesting to introduce new drivers in the analysis of the users' intentions, as well as in the measurement of the real behavioral use of P2P payment systems.

In addition, we propose a comparative analysis between the different services offered by Bizum, such as payments to merchants or donations among others.

To close, it would be also interesting to analyze the possible moderating effects of other variables related to payment systems included by other, recent research, as well as the effects of different cultural factors through a cross-national study.