An extension of Trust and TAM model with TPB in the adoption of digital payment: An empirical study in Vietnam

2.1 TAM model

Perceived usefulness

Perceived usefulness, which reflects an individual’s strong belief in the benefits of technology, is considered a critical factor in enhancing performance (Davis, 1989; Taylor & Todd, 1995a). Digital payment systems are deemed useful for customers when they offer substantial services. Despite previous unsatisfactory experiences, customers are likely to continue using digital payment methods if they find them beneficial (Bhattacherjee, 2001; V. Van Diep, 2017). Perceived usefulness is the most frequently utilized independent variable in prior research for assessing people’s readiness to adopt DPs both at individual and organizational levels (Kabir et al., 2017). Therefore, we propose the following hypothesis:

Perceived ease of use

Perceived ease of use is defined as “the degree to which an individual believes that using a particular system would be free of effort” (Davis, 1989). Innovative technology systems that are perceived as easier to use and less complex are more likely to gain acceptance and be adopted by users. Digital payment systems are perceived as easy to use when users find them simple to understand, quick to learn, and straightforward to operate. As a result, perceived ease of use is recognized as a critical factor influencing users’ acceptance and adoption of new technology (V. Van Diep, 2017; Kabir et al., 2017; Tavera-Mesías et al., 2023). Additionally, a more vital perception of ease of use can enhance consumer confidence in the expected benefits of using technological products (Daragmeh et al., 2021; Doanh et al., 2022). Accordingly, we propose the following hypothesis:

2.2 TPB model

2.2.1 Attitude

Attitude refers to an individual’s favorable or unfavorable feelings about engaging in a particular behavior (Davis, 1989; Taylor & Todd, 1995a). Additionally, a positive or negative attitude directly impacts the strength of behavioral beliefs regarding the anticipated significant outcomes (Wu & Chen, 2005). Therefore, it is more likely for customers to take action to use digital payments if they develop a positive opinion about the adoption of a digital payment method. In line with the above argument, we propose the following hypothesis:

2.2.2 Perceived behavioral control

Perceived behavioral control represents an individual’s perception of the ease or difficulty involved in carrying out a particular behavior. It relates to beliefs about the presence of factors that may either facilitate or obstruct the performance of the behavior (Ajzen, 1991, 2002; Rachmawati & Rahardi, 2023). In the context of digital payments, perceived behavioral control refers to a consumer’s perception of having the necessary resources, knowledge, and opportunities to adopt digital payment methods. We propose the following hypothesis:

2.2.3 Subjective norm

Subjective norm refers to an individual’s perception of social pressure to either engage in or refrain from a particular behavior (Ajzen, 1991). In other words, subjective norm relates to an individual’s normative beliefs about the expectations of others (Rachmawati & Rahardi, 2023; Wu & Chen, 2005). In our study, subjective norm is defined as consumers’ beliefs about the influence that someone important to them may have on their decision to adopt digital payment methods. Based on this, we propose the following hypothesis:

2.3 Trust

Trust is the confidence one party has in the intentions and actions of the other party (Aldaabseh & Aljarah, 2021; Siagian et al., 2022). Trust often encompasses three key dimensions: ability, integrity, and benevolence (Benamati et al., 2010). Ability refers to the knowledge and skills required of digital payment service providers to fulfill their tasks effectively. Integrity signifies that these providers consistently keep their promises, while benevolence indicates that they genuinely care about the users’ interests, not just their own. Trust will be a crucial potential influencer in examining the adoption of digital payments. In exploring the factors influencing digital payment adoption, we refer to the relationships depicted in Figure 1, which presents the conceptual framework guiding our analysis. This framework emphasizes the interplay between technological acceptance and trust

Figure 1. Conceptual framework.

2.3.1 Trust and TAM relationship

The relationship between PU, PEU, and trust has been widely discussed in the literature (Gefen, 2004; Gefen et al., 2003; Pavlou, 2003). Wu & Chen (2005) followed Gefen et al. (2003) suggest a model of trust and TAM for the initial adoption of online tax. It shows that trust is an antecedent of PU, and PEU is an antecedent of trust.

In line with social cognitive theory, perceived ease of use (PEU) can generally be argued to positively impact a person’s favorable expectations regarding accepting innovative technology (Bandura, 1986). This is because cognition-based trust is primarily founded on initial impressions of behavior. In the context of online services, PEU represents the initial feeling or expectation that influences further online transactions (Wu & Chen, 2005). PEU is hypothesized to positively influence trust because it helps foster a favorable impression of e-vendors during the initial adoption of online services. This positive impression can make customers more willing to invest in and commit to the buyer-seller relationship (Gefen et al., 2003; Wu & Chen, 2005). In line with the above argument, we propose the following hypothesis:

Trust is a key determinant of perceived usefulness, particularly in the digital payments environment, as it assures consumers that they will experience the expected benefits from service providers. In addition, trust is recognized to positively influence PU because it allows consumers to feel comfortable being vulnerable with e-vendors, ensuring they receive the expected value from their interactions and services (Pavlou, 2003). When consumers initially trust their e-vendors and perceive that adopting an online tax will enhance their job performance, they are more likely to believe that the online tax is beneficial (Wu & Chen, 2005). Hence, we propose the following hypothesis:

2.3.2 Trust and TPB relationship

The relationship between trust and TPB can be explored from various perspectives. Trust is hypothesized as a common antecedent influencing attitude, perceived behavioral control, and subjective norm.

Trust in an e-vendor is considered a significant behavioral belief that directly influences a customer’s attitude toward purchasing behavior. When an e-vendor is perceived as trustworthy, it increases the likelihood that the consumer will benefit from and avoid potential risks associated with adopting online services (Pavlou, 2003). Many previous studies have shown a positive relationship between trust and attitude (Agag & El-Masry, 2016; Chawla & Joshi, 2019; Wu & Chen, 2005). Trust is clearly a crucial factor influencing attitudes toward adopting digital payment behavior. Thus, the following hypothesis is proposed:

Trust can strengthen perceived behavioral control in online transactions by making the interactions between customers and e-vendors more predictable and consistent (Pavlou, 2003). Trust impacts perceived behavioral control by bolstering self-efficacy and fostering conditions that facilitate successful interactions (Wu & Chen, 2005). When customers trust a digital payment provider that meets their expectations, this trust is likely to enhance their perceived behavioral control over digital payment transactions. Based on the preceding arguments, the hypothesis can be stated as follows:

Peer and superior influences shape users’ subjective norms concerning IT usage. Consequently, it can be inferred that trust in peers and superiors concerning their beliefs about IT usage should play a significant role in shaping subjective norms (Taylor & Todd, 1995a). Similarly, trust in e-vendors, particularly regarding their reputation, brand name, and service quality, may positively influence subjective norms related to online transaction behavior (Wu & Chen, 2005). Therefore, whether trust exerts direct or indirect influences on subjective norms remains a crucial antecedent of subjective norms in digital payments. Hence, we propose the following hypothesis:

3.1 Sample profile

This study utilized both quantitative and qualitative research across four phases to explore the factors influencing digital payment adoption in Vietnam’s northern mountainous region. The survey scales were developed based on theoretical frameworks and relevant literature, with input from experts in fintech and e-commerce to ensure content validity. A pilot test with 100 participants from Thai Nguyen and Cao Bang helped refine the questionnaire. The primary survey was conducted in Lang Son and Cao Bang provinces from December 2023 to January 2024. These provinces were chosen to explore the factors limiting digital payment adoption, particularly in rural areas, as part of a broader investigation into the challenges of digital payment adoption in the northern mountainous region of Vietnam.

Following the sample size guidelines from Hair et al. (2019), we aimed for at least 140 participants, with the final sample comprising 800 respondents evenly split between the two provinces. Participants were informed about the study and provided consent before the interviews. Each respondent received a small gift valued at approximately $2 to encourage participation. Of the initial surveys, 509 were valid, yielding a 63.63% response rate. Participants provided personal information, including their name, address, age, education, income, and occupation, which interviewers recorded. The interviewees reviewed their responses and confirmed their accuracy by signing the survey form. Afterward, they signed a summary information table and received a gift from the project.

3.2 Method estimation

Structural Equation Modeling (SEM) is a second-generation statistical analysis technique designed to examine multidimensional relationships between multiple variables within a model (Anderson et al., 2004). SEM has been widely utilized across various fields, including sociology (Lavee, 1988; Lorence & Mortimer, 1985), psychology (Anderson et al., 2004), and management (Tharenou et al., 1994).

The complex theoretical model created using this method is generally associated with the data collected for validation purposes. This association is known as model-data fit. Any theoretical model can be evaluated using available empirical data for this type of fit. SEM is recognized as an extensive sample method, usually necessitating a minimum sample size of 200 (Dash & Paul, 2021; Hayes et al., 2017). There are two primary models: the path analysis and the measurement model. While some advanced models, such as multilevel and growth models, are also considered, this study will focus on i) the measurement model and ii) path analysis.

Before examining the path analysis among latent variables (also called factors or constructs), we first assess the measurement of these unobserved variables. Since these variables are not directly observable, they are represented through a set of measured variables from which the latent constructs are derived. Each latent variable is quantified using observed indicators tested for reliability and validity. SEM uses Confirmatory Factor Analysis (CFA) to evaluate the measurement model. In this case, the model fit is assessed to validate the measurement model. Once the model fit is confirmed, path models among the latent variables are evaluated (Hair et al., 2019; Hayes et al., 2017). In our research, before employing SEM, we followed the recommendations of Doanh et al. (2024) and Huy et al. (2024) by conducting exploratory factor analysis (EFA) to identify latent variables. EFA aims to uncover the latent constructs underlying a set of observed variables. EFA is used when the research goal is to determine the nature and number of common factors among these variables (Omura et al., 2022). Subsequently, we performed confirmatory factor analysis (CFA) by examining factor loadings, composite reliability (CR) indexes (Joreskog et al., 1971), and average variance extracted (AVE). While not a parameter estimate, AVE helps evaluate how much of the variance in indicators is explained by the latent factor. Calculated using factor loadings and residual error variances for a latent factor, AVE values of 0.50 or higher are generally acceptable. Composite Reliability (CR), like AVE, is not a parameter estimate but is valuable for interpreting the CFA model. Generally, CR values of 0.70 or higher are considered acceptable; however, values between 0.60 and 0.70 may indicate questionable reliability but are not necessarily unacceptable (Fornell & Larcker, 1981; Joreskog et al., 1971). This methodology, commonly referenced in the literature, rigorously assesses our research constructs and ensures the validity and reliability of our findings.

The path model is a form of multiple regression model estimated simultaneously, illustrating mediation, moderation, and interaction effects among variables. It defines the structural relationships between latent variables based on their associations with observed indicators (measured variables). Once the measurement models of latent constructs are validated through CFA, these paths can be interpreted as causal or covariance-based. It can assess the unidimensionality, validity, and reliability of an unobserved latent construct (factor) (Dash & Paul, 2021; Hair et al., 2019).

Some of the fit indices in the SEM model are used to test and compare models, such as Chi-square, Comparative fit index (CFI), Tucker–Lewis Index (TLI), and Root mean squared error of approximation (RMSEA). Overall model fit is evaluated using the chi-square statistic, which reflects the discrepancy between the sample data and the model’s specified covariance matrices. This statistic is widely regarded as an indicator of model fit quality, with a non-significant value at the 0.05 level being preferable. Additionally, the chi-square to degrees of freedom ratio (CMIN/df ) provides a more straightforward assessment, with a value of 3 or less (sometimes up to 5) typically indicating a good fit. The Comparative Fit Index (CFI) is an incremental fit index that compares the focal model to a baseline model, often called the null or independent model. CFI tends to be less affected by sample size than the chi-square test. A CFI value of 0.95 or above typically signifies a good fit to the data, although some may consider a threshold of 0.90 acceptable. The Tucker-Lewis Index (TLI) is a comparative or incremental fit index. Generally, a TLI value of 0.95 or higher indicates a good model fit to the data, though some guidelines permit a lower cutoff of 0.90. The Root Mean Square Error of Approximation (RMSEA) is an absolute fit index that penalizes model complexity, favoring more parsimonious models. Generally, an RMSEA value of 0.06 or lower indicates a good fit to the data, although some guidelines may relax this threshold to 0.08 or even 0.10 (Bagozzi & Yi, 1988; Dash & Paul, 2021).

We selected SEM as the analysis method for this study due to its robustness in testing theories, as highlighted by Steenkamp & Baumgartner (2000). This makes SEM particularly suited to our research focus. Our research model encompasses multidimensional constructs that are not directly observable but are measured through observable indicators. Thus, SEM, which emphasizes construct operationalization (Bagozzi, 1994), is an appropriate and practical approach for our investigation. Furthermore, our application of SEM is consistent with prior research (Thongsri et al., 2019).

Followed Fan et al. (2016); Kline (2018), we deploy SEM, which consists of five main steps: model specification, identification, parameter estimation, model evaluation, and model modification. The model specification defines hypothesized relationships among variables based on prior knowledge, while model identification checks if the model is over-identified, just-identified, or under-identified, as coefficients can only be estimated in just-identified or over-identified models. Parameter estimation then calculates these coefficients. The model evaluation assesses how well the model fits using quantitative indices to measure the overall goodness of fit. If necessary, model modification adjusts the model to improve fit, often in a post hoc fashion. Finally, validation enhances the model’s reliability and stability. In our study, SEM analysis was conducted using JASP (https://jasp-stats.org/).

3.3 Data description

Table 1 presents the demographic features of the respondents to the survey. According to the study’s gender review, 42.6% are male, and 57.4% are female. Most respondents are between the ages of 31 and 40 (229 individuals, accounting for 45%), followed by the age range of 41 to 50 years (143 individuals, accounting for 28.1%). In terms of culture, the majority are ethnic minorities such as Tay and Nung, accounting for 78.2%, while the remaining belong to the Kinh and other groups. Regarding educational level, most interviewers had completed college or university degrees, accounting for 56.7%, followed by those who graduated high school (24.6%). Those with education levels below secondary school and secondary school constituted a small proportion (10.4%). Most families have 4 to 5 members, accounting for 61.1%. The average monthly income of people is meager; the number with income equal to or greater than 5 million VND/month accounts for only 29.1%, and the rest all have income below 5 million. One of the reasons is that their primary income comes from their profession, in which data shows that 44.6% are purely farmers, the remaining 55.4% are workers working away from home, and commune-level civil servants.

4.1 Measurement model

Before proceeding with SEM model testing to evaluate the hypotheses, we conducted EFA and CFA to assess the variables’ construct and confirm the reliability and validity of the measurement model. Table 2 shows that the total Cronbach’s alpha is 0.938 greater than 0.6, and all observed variables have Cronbach’s alpha coefficients ≥ 0.935. The results also indicate that seven variables were extracted from 30 observed factors; the eigenvalue value is 1.053 (>1) with a variance of 71.62%. The results of the EFA in Table 2 show that the factor loadings of the items ranged from 0.661 to 0.853, and the construct of items is consistent with the literature. The test coefficient KMO = 0.916 satisfies the 0.5 < KMO< 1, showing that the exploratory factor analysis is appropriate for our data. The Chi-square statistic of the Bartlett test reached 9630.515, with the p-value = 0.000, showing that the data is suitable and reliable for performing the EFA method.

CFA allows for testing the validity and accuracy of specific models constructed based on data and theoretical foundations. We assess the goodness-of-fit of the measurement model using various tests, including the Chi-square test (χ²), Comparative fit index (CFI), Tucker–Lewis Index (TLI), and Root mean squared error of approximation (RMSEA), composite reliability (CR) and the average variance extracted (AVE). As shown in Table 3, the composite reliability (CR) of latent variables exceeds 0.836, and the AVE values range from 0.560 to 0.687. In addition, the results indicate that the measurement model exhibits favorable fit index values (χ²/df = 2.486, CFI = 0.943, TLI = 0.931, RMSEA = 0.05, SRMR = 0.03). These findings support our proposed model’s suitability, reliability, and validity; the data exhibit a good model fit.

4.3 Structural equation model

Table 4 and Figure 2 present the model’s estimation results using the SEM method. The assessment revealed favorable fit index values: χ²/df = 3.014, CFI = 0.909; TLI = 0.895, RMSEA = 0.06, SRMR= 0.05. These results indicate that the empirical findings are reliable and valid.

Figure 2. The results of the structural equation model.

As a result, all hypotheses are positive and statistically significant at the 1% level. It means that all hypotheses are supported.

4.4 Additional analysis

It is crucial to emphasize the importance of addressing the Skewness-Kurtosis test. To tackle this issue, we consulted the studies by Barnes et al. (2001) and Vieira (2011). Additionally, following the recommendation of Diamantopoulos & Siguaw (2000) we conducted a model cross-validation analysis.

Our analysis used multivariate normality tests, such as the Mardia mSkewness and Mardia mKurtosis tests, to evaluate our model’s performance. The findings reveal that our model satisfied the criteria set by both the mKurtosis and mKurtosis tests. This means that our research model is appropriate for maximum likelihood (ML) estimation and demonstrates a good fit.

5.1 Discussion results

This study aims to utilize an extended version of the Trust and TAM model, incorporating the TPB, to better understand the acceptance behavior toward adopting digital payments in the Northern mountainous area of Vietnam. The empirical results from the SEM model reveal that the standardized path coefficients are all positively significant except for the path from PU to ATT.

5.1.1 TAM model

Perceived usefulness (PU) negatively influences attitudes (ATT) towards adopting digital payments at the 5% level, as supported by the standard regression coefficient value of -0.131. Therefore, the hypothesis H₁ has been rejected. This result contrasts previous studies (Ariffin et al., 2021; Chawla & Joshi, 2019; Mabkhot et al., 2023; Ranpariya & Joshi, 2024; Wu & Chen, 2005). This indicates that users do not feel digital payments are useful. This may come from their daily habits of using cash instead of digital payment methods. Therefore, it is essential to keep expanding the spread of information and communication regarding the practical advantages of digital and cashless payments. Additionally, implementing safeguards and strict measures to prevent fraud and property theft on online platforms is equally important.

Our results also show the positive influence of perceived usefulness (PU) at the 1% significance level on DPA, thereby supporting H₉. Conclusively, the results obtained are supported by the results of previous studies (Aji et al., 2020; Chawla & Joshi, 2019; Nguyen & Ao, 2022; Ranpariya & Joshi, 2024; Siagian et al., 2022). This means that when consumers achieve significant benefits from digital payment systems, they will trust DP more and continue to use it. Therefore, service providers should focus on further developing and improving the quality of DP to increase customers’ perception of its usefulness. This will help increase the number of users adopting DPs.

The findings of H₂ align with previous studies, showing that perceived usefulness positively and significantly influences users’ attitudes toward using digital payment (Ariffin et al., 2021; Mabkhot et al., 2023); however, in contrary to Chawla & Joshi (2019). This study’s findings indicate that consumers found it easy to conduct payment transactions using a digital payment system. The system not only streamlined their payment processing but also enhanced their transaction efficiency, leading to a positive attitude toward adopting digital payments.

Furthermore, the finding of H₃ is consistent with previous research, which demonstrated that perceived ease of use positively and significantly influences perceived usefulness at the 1% level (Aji et al., 2020; Chawla & Joshi, 2019; Ranpariya & Joshi, 2024; Siagian et al., 2022; Wu & Chen, 2005). It proved that perceived ease of use increases the perceived usefulness of adopting digital payments. Therefore, service providers should focus on further developing and improving their payment system to be simpler to understand, quick to learn, and straightforward to operate digital payment services.

5.1.2 TPB model

Attitude (ATT) was also determined to impact DPA at the 1% significance level positively; hence, hypothesis H₁₀ is supported. It means that customers have a positive opinion about adopting digital payment methods. This study is also in accordance with the results of research by Ariffin & Lim (2020); Ayudya & Wibowo (2018); Chawla & Joshi (2019); Ranpariya & Joshi (2024).

Perceived behavioral control (PBC) is supposed to positively affect the use of DP at the 1% level. This hypothesis is supported by the value of the standard regression coefficient (0.178). Therefore, the acceptance of hypothesis H₁₁ has been verified. This finding reveals that consumers nowadays have enough resources, knowledge, and opportunities to adopt a digital payment method. This result was consistent with the results of Ariffin et al. (2021); Ariffin & Lim (2020); Ayudya & Wibowo (2018); Mabkhot et al. (2023).

Similarly, subjective norm (SN) also positively impacts users’ adoption of digital payments at the 1% significance level. Hence, hypothesis H₁₂ is supported. The result aligns with the viewpoints of Aji et al. (2020); Ariffin et al. (2021); Jusoh & Jing (2019) but not consistent with Ayudya & Wibowo (2018). This insight demonstrates that people’s willingness to adopt digital payments as a new mode of transaction is significantly influenced by their peers’ opinions and behaviors. Reliable information from trusted sources like relatives, neighbors, or friends boosts consumers’ trust in digital payment methods, leading to a stronger intention to adopt them.

5.1.3 Trust and TAM relationship

Perceived ease of use (PEU) is found to positively and significantly influence trust (TR) towards adopting digital payments, and hypothesis H₄ is supported. The result in light of the previous studies, which also showed the importance of PEU in enhancing user trust (Chawla & Joshi, 2019; Gefen et al., 2003; Wu & Chen, 2005). It reveals that consumers had a favorable impression of e-vendors and trust when using digital payment transactions.

Moreover, trust (TR) positively impacts perceived usefulness (PU). This relationship was proved by the value of the standard regression coefficient of 0.478 at the 1% significant level. Hence, the hypothesis H₅ is supported. It confirmed that consumers feel comfortable being vulnerable with digital payment service providers and can receive the expected value from their actions and services. Then, they will adopt more and more digital payment methods for their daily buy and sell activities. Previous studies support this result (Gefen et al., 2003; Wu & Chen, 2005).

5.1.4 Trust and TPB relationship

Our research findings corroborate those of (Ariffin & Lim, 2020; Wu & Chen, 2005), who similarly suggested that trust (TR) positively influences attitude (ATT) towards adopting digital payments, with a coefficient of 0.751 and a p-value of 0.000. Based on these findings, the hypothesis H₆ is confirmed. This supports the previous findings on trust and attitude, showing that when consumers believe in the services and information of e-vendors, they will increase the likelihood of adopting digital payment methods.

The positive significance of H₇ proves that trust (TR) boosts perceived behavioral control (PBC) in online transactions by making interactions between customers and e-vendors more stable and predictable. This result is consistent with the studies of Ariffin & Lim (2020); Wu & Chen (2005). It confirms that when customers trust a digital payment provider that fulfills their expectations, this trust will likely increase their perceived control over digital payment transactions.

Notably, the results have highlighted the significant role of trust (TR) in influencing subjective norm (SN) toward adopting digital payments. This trust has a positive impact on subjective norm, with a coefficient of 0.821 and a p-value of 0.000. This result is not in line with research conducted by Ariffin & Lim (2020) but consistent with Wu & Chen (2005). It shows that customers in this region had trust in e-vendors, particularly in terms of their reputation, brand recognition, and service quality, which may positively affect subjective norm related to digital payment transaction behavior.

Among the control variables, the results show that age is not statistically significant, suggesting that there is no difference in the adoption of DPs based on the respondents’ age. Education is found to have a positive significance with DPA at a 10% level and in light with Cao et al. (2018). It proves that users with higher education levels are more likely to use DP systems to transfer money, shop, and pay for utility services.

5.2 Theoretical contributions

The previous body of literature concerning intentions to adopt digital payments has primarily relied on established theories such as TRA, TAM, TPB, UTAUT, and UTATU2 (Kabir et al., 2017; Susanto et al., 2022). In line with this existing research, the analysis in this study makes a noteworthy contribution to the existing literature by extending the version of the Trust and TAM model, incorporating the TPB, to better understand the acceptance behavior toward adopting digital payments. Our empirical findings indicate that only the standardized path coefficient from PU to ATT is negative, while others are all positively significant. These findings improve our capacity to analyze the issue more effectively and provide valuable recommendations to increase the adoption of digital payments in the Northern mountainous regions of Vietnam.

5.3 Practical implications

While digital payments are primarily enabled by Internet and communication technologies, this study highlights the importance of addressing both technological and trust-related issues to enhance citizens’ decision to use these services. The Technology Acceptance Model (TAM) identifies perceived usefulness (PU) and perceived ease of use (PEU) alongside trust as critical factors influencing behavioral intention. Each factor significantly affects adoption through mediators such as attitude, perceived behavioral control, and subjective norms.

Careful attention must be given to both technological design and trust components to encourage citizens to adopt digital payments. Additionally, as noted earlier, novice users tend to prioritize trust in non-technological aspects over PEU and usefulness when forming their attitudes. This suggests that trust is more critical in shaping a user’s attitude toward digital payments than the technology’s perceived ease of use or usefulness. Major trust concerns include privacy protection, accuracy of information, and unauthorized access, among others.

Fundamentally, trust has been empirically identified as a precursor to perceived usefulness (PU), which in turn influences attitude. This has important practical implications for improving attitudes toward digital payments. Providers of digital payment services should focus on developing trust-building mechanisms to attract novice users. Examples of such mechanisms include guarantees, increased familiarity through advertising, reliable customer service, and incentives for usage.

Once trust is established, enhancing the perceived usefulness of digital payments becomes crucial for attracting new users. This requires careful design aligning with users’ needs to communicate the service’s benefits effectively. Without initially addressing trust, even a well-designed digital payment system with high perceived usefulness may struggle to engage novice users.

5.4 Limitations

While this study enhances the existing literature by integrating multiple models to examine users’ adoption of digital payments, it is essential to acknowledge its limitations. First, we used perceived usefulness to capture the overall benefits of digital payments for users. However, these benefits should be distinguished between economic and non-economic aspects, such as user satisfaction and service quality. Second, the number of farmers with low levels of education is less than that of others working away from home and commune-level civil servants. Farmers with low levels of education may be less likely to have adequate information to trust and adopt digital payment methods than cash. Therefore, future research should explore the impact of adopting digital payments on farmer income.