Subjective norm refers to people’s perceptions of how their behavior as users will be judged (Park and Kim, 2018). It is an important behavioral intention in the Theory of Reasoned Action and the Theory of Planned Behavior (Ajzen, 1991). The existing TAM did not consider SN. However, Venkatesh and Davis (2000) found that in actual social environments, people employed their perceived ease of use and usefulness. Hence, they proposed TAM2, which considers the factors affecting behavioral intention, along with PU. Sathye et al. (2018) explained that SN has a direct effect on the intention to use mobile payment systems, and an indirect effect through the medium of PU. Therefore, this study proposed the following hypotheses:

Perceived ease of use refers to the degree to which an individual feels comfortable to use a particular system, and is an important predictor of PU, as well as users’ intentions for accepting a particular technology (Davis, 1989). Several studies on mobile easy payment services found a positive relationship between PEOU and the intention to use mobile payment services. Cao et al. (2016), Shankar and Datta (2018), and Daragmeh et al. (2021) found that PEOU had the greatest effect on PU. Therefore, this study proposed the following hypotheses:

Perceived usefulness is a variable that directly predicts users’ intention to accept a particular technology (Al-Maroof and Al-Emran, 2018), and indicates the degree to which users believe that using the system will improve their performance (Davis, 1989). Several studies on mobile easy payment services suggest that PU has a positive effect on UI (De Luna et al., 2019; Singh et al., 2020). Therefore, this study proposes the following hypothesis:

Risk perception is an individual’s subjective intuition for judging and recognizing a dangerous situation under specific circumstances. Early research on risk perception was primarily conducted from an economic point of view, but since the mid-19th century, many studies have been conducted in the field of consumer behavior (Bae and Jeong, 2021). Risk perception is determined by the characteristics of various national cultures, and since the COVID-19 pandemic outbreak, people’s anxieties and fears have increased. Therefore, it can be assumed that the uncertainties resulting from the spread of COVID-19 acts as individuals’ psychological risk factor, and affects their behavioral intentions (Kim, 2020).

Owing to the social distancing norms and national lockdowns implemented in some countries after the COVID-19 outbreak, consumers have been forced to use contactless-based mobile payment services, and thereby, their use has significantly increased. As people might be variously aware of the risks of COVID-19, it is difficult to find a reliable empirical study on how this affects the UI of mobile easy payment services. Therefore, this study assumes the following hypotheses to determine the influence of C19RP on perceiving the risk of COVID-19 infection, and to find out how C19RP is a modulator variable that affects the UI of mobile easy payment services:

The PLS-SEM was used to establish this study’s research model. It is based on structural equation modeling (SEM), developed by combining path, regression, and factor analyses. SEM proposes two approaches for estimating the relationships between variables. The first is a covariance-based structural equation model (CB-SEM) for verifying the conceptual theory. This is a method of confirming (or rejecting) a set of systematic relationships. The second is the application of PLS-SEM for predicting a specific causal relationship. This is mostly used for developing a theory for maximizing predictability in exploratory research (Abbasi et al., 2021; Yoo et al., 2021).

With regard to sample size, while CB-SEM requires a minimum sample size of 200 participants, PLS-SEM requires 30–100 participants. Therefore, it can be conveniently applied to studies with small samples. While a larger sample size increases the estimation’s accuracy, CB-SEM and PLS-SEM yield similar results when the sample size comprises 250 participants or more (Shin, 2018). Considering that this study’s sample size was 227, either the CB-SEM or PLS-SEM could have been applied. The PLS-SEM model was chosen for interpreting the relationship between the two.

WarpPLS 7.0, a statistical program used for PLS-SEM analysis, was used to analyze the PLS-SEM model in this study as well. It performs both linear and non-linear analysis for determining whether the statistical assumption of the paths between the variables is linear or non-linear (WarpPLS, 2020; Sung, 2021). A non-linear relationship analysis using WarpPLS, rather than the existing PLS-SEM, yields a more accurate analysis than a linear relationship analysis method. As a result, one could observe significant non-linearity between the variables. An attempt was also made to derive an interpretation of the relationship (Kim, 2014). Therefore, this study first examined the definition of linearity, and then clarified the concept of non-linearity. First, “linear” indicates a relationship between different values. The definition of a linear relationship between two variables x and y is:

where the coefficients a, b, and c are constants, independent of the variables x and y, and the relationship between the two variables defined in this manner satisfies all linear characteristics. When defining a non-linear relationship between two variables on the basis of the concept of linearity, “non-linear” refers to a relationship between two variables. It, therefore, indicates any relationship between x and y that is not expressed by Eq. 1. For example, all cases that include terms other than the linear terms regarding variables x and y, such as axn + by + c = 0 and axy = b, can be labeled as non-linear relationships (Jang and Kim, 1994).

Such a non-linear relationship analysis is primarily performed when a relationship that is otherwise difficult to explain using only linear relationship analysis appears. However, if the relationship between the variables is non-linear, and a commonly used linear relationship analysis is applied, the complex relationship between the variables cannot be analyzed completely. This results in reduced efficiency, which is a common challenge in this analysis. Therefore, when conducting research on complex social phenomena, it is necessary to examine the non-linear relationships between variables. This allows for obtaining more accurate results (Sung, 2021). Seo (2020) confirmed a U-shaped non-linear relationship between stress-levels emerging from restricted viewing due to COVID-19 and the intention of professional baseball fans. An appropriate stress-level might have a positive effect on promoting the intention to watch, by increasing the value of watching the game due to restrictions on visiting the stadium. A negative effect of the reluctance to visit a stadium because of perceiving it as an adverse situation, due to the risk of infection, was also found. Jang et al. (2021) explained that a V-shaped non-linear relationship exists between time required for commuting and the satisfaction thereof. In general, the longer the commuting time, the lower the satisfaction thereof. However, non-linear relationship analysis revealed that after a certain period, the number of household members, work experience, satisfaction with neighbors, and health variables affected satisfaction. Due to a high relative influence, it was interpreted as showing a V-shaped characteristic between commuting time and satisfaction with commuting time.

In this study, Warp 3 Stable was selected among the analysis options provided by WarpPLS, and structural model analysis was used for deriving the coefficient of determination (R2) value representative of the dependent variable’s explanatory power. If the value of R2 is 0.25 or higher, the fitness of the research model is “high.” The R2 for the dependent variables PU and UI was 0.59 and 0.48, respectively. Thus, Goodness of Fit (GoF) can be evaluated as “high.” It is also suggested that if the Tenenhaus GoF is 0.25 or more, as a measure of the GoF of other research models, the research model has a “medium” fit. If it is 0.36 or more, the research model has a “high” fit (Rahman et al., 2013). The GoF value of this research model is 0.670, which indicates a high fit. Further research on the model’s fitness criteria can be found in Table 4. The GoF of this research model meets the requisite criteria.

The model was verified using PLS-SEM and the WarpPLS (7.0) software, and the linear and non-linear relationships of each linear path were analyzed. To secure the statistical significance of each path, the p-values of the path coefficients must be less than 0.1 (WarpPLS, 2020). Testing the hypothesis revealed that all p values were less than 0.1, as shown in Figure 2. This indicated the statistically significant results in all paths.

Examining the path coefficient and effect size revealed that PEOU was the path to PU, and PU was the path to UI. Statistical significance was found to be high in that order. Table 5 shows the results of the path hypotheses tests, as described above.

The analysis of the existing structural equation model assumes that all paths are linear (Yang and Kim, 2015). We used WarpPLS software to verify whether the research hypothesis had a linear or non-linear relationship. The results of the verification showed that the paths for all hypotheses were non-linear, as shown in Table 6.

This section explains the analysis of the degree of difference in the pathways according to C19RP, which was set as a moderator variable. We reviewed the type of meaningful modulating effect of C19RP on each of the pathways, as shown in Table 7. After examining the results of hypothesis testing on the control effect of C19RP, all hypotheses with p-values of <0.1 were adopted. On closely reviewing the verification results, it was found that the path with the greatest difference in C19RP was from the PU to the UI. Thus, it can be concluded that the lower the group, the more sensitive the mobile easy payment service’s UI for PU. In all other pathways, lower C19RP was found to be more sensitive.

As shown in Figure 3, in the path from the SN to the UI, the group with a high C19RP was compared with that with a low C19RP. It was confirmed that according to SNs, the UI was high. With the implementation of social distancing, an atmosphere that encourages alternatives to face-to-face payment methods has been created. In this context, it shows that groups with high C19RP are used to mobile easy payment services. Therefore, the group with a high C19RP is more likely to use mobile easy payment services, than the group with a low C19RP, after the onset of COVID-19. This is due to social influences.

As shown in Figure 4, the high C19RP group, on the path from the SN to the PU, had a low C19RP. When compared with the group, it was confirmed that according to the SN, the PU was high. Although the high C19RP group showed a U-shaped non-linear relationship, when compared with the low C19RP group, perception according to the SN and PU was found to be high. This is because, in the case of a group with a high C19RP, information about COVID-19 is largely obtained from television or the internet, as well as from family and friends. Therefore, the high C19RP group is more likely to use mobile easy payment services than the low C19RP group, owing to social influences of the surrounding people and media. This might be interpreted as greater recognition of its usefulness as a means for preventing infection.

As shown in Figure 5, in the path of PEOU-to-UI, the high C19RP group had higher UI according to PEOU. A group with a high C19RP perceives that face-to-face, traditional payment methods increase the risk of contracting COVID-19, and that, to prevent infection, they must actively use contactless mobile easy payment services. Therefore, it can be interpreted that this group uses mobile easy payment services more than the group with low-risk perception. Conversely, the group with a low C19RP was less likely to perceive fear of COVID-19. Although they recognize that contactless services are easy to use and reduce the risk of infection, it can be deduced that the effect on UI is not as large as that of the group with a high C19RP.

As shown in Figure 6, in the path from the PEOU to the PU, the high C19RP group had higher PU and PEOU than the group with low-risk perception. The high C19RP group perceived remote payment methods as being more capable of preventing COVID-19 infection than conventional in-person payment methods, and experienced no difficulty in using them. Therefore, it can be deduced that this group secured the usefulness of mobile easy payment services to some extent, as compared with the low C19RP group.

As shown in Figure 7, the high C19RP group had higher UI for PU than the low C19RP group. It was confirmed that the high C19RP group was already using a mobile service to some extent due to the fear of COVID-19 infection. It can be deduced that greater recognition of the usefulness of such payment services correspond to a higher positive effect on UI. Conversely, the UI of the low C19RP group was not affected significantly, even though they perceived the service as useful. Therefore, it can be deduced that in this case, the greater recognition of mobile easy payment services as useful and faster leads to a more positive effect on UI.