The unified theory of acceptance and the use of technology-2 (UTAUT2) is developed by extending the UTAUT theory, which provides an integrated view by excluding the concurrency and iterations of similar elements and integrating the constructs from eight popular theories and models, i.e., theory of reasoned action (TRA), technology acceptance model (TAM), innovation diffusion theory (IDT), theory of planned behaviour (TPB), and social cognitive theory (Hussain et al., 2019). Three additional factors (price value, hedonic motivation, and habit) were added with four original elements of the UTAUT theory (e.g., performance expectancy [PE], effort expectancy [EE], SI, and facilitating condition [FC]), which resulted in the establishment of the UTAUT2 (Venkatesh et al., 2012). Compared to the original UTAUT, the expansion included in the UTAUT2 contributed to a considerable improvement in the variance, as indicated through the consumers’ intentions to adopt (56–74%) and actual usage behavior (40–52%) (Tandon et al., 2016). Specifically, the UTAUT2 has been proven to be highly efficient at interpreting mobilepayment technology adoption intentions and actual usage behavior among users (Gupta and Arora, 2019). Moreover, a considerable number of earlier studies employed the UTAUT2 to assess the factors that contributed to the adoption of NFC M-payments (Slade et al., 2015; Morosan and DeFranco, 2016; Khalilzadeh et al., 2017; Karjaluoto et al., 2020; Liébana-Cabanillas et al., 2021). Given that it is a widely used theory in addressing both external and internal drivers of any new technology adoption (Koenig-Lewis et al., 2015). This study adopted the UTAUT2 model as a theoretical underpinning.

Numerous earlier studies (Morosan and DeFranco, 2016; Oliveira et al., 2016; Gupta and Arora, 2019; Hussain et al., 2019) found that the price value was not a significant indicator of adoption intention behavior in the context of upgraded technology in contactless payments. Based on literature findings and arguments, price value was not considered a factor in this study. To explore the impact of habit, users require a high level of experience and continuous usage records after the adoption of new technology (Alalwan et al., 2017). Furthermore, a new technology that has not gained widespread market acceptability may be faced with obstacles in building a habit among users (Oliveira et al., 2016; Lin et al., 2020). Considering the fact that many of the participants in this study were not using the NFC M-payments on a regular basis, the habit was not considered a determinant of users’ intention to adopt the NFC M-payments.

The expansion of the UTAUT2 model has been suggested through the incorporation of other determinants associated with regions and cultures (Sobti, 2019). To illustrate, researchers (Khalilzadeh et al., 2017; Karjaluoto et al., 2020; Liébana-Cabanillas et al., 2021; Penney et al., 2021) modified the UTAUT2 through PR as a determinant that strongly encourages the mobile payment adoption. Following previous researchers’ recommendations, this study extended the UTAUT2 model by including PR along with the other determinants.

To evaluate the adoption intention and actual acquisition of NFC M-payments among Malaysian youth, a cross-sectional research methodology and quantitative technique were employed in this study. People who aged from 18 to 40 years old were categorized as youth and working adults by the International Labor Organization (International Labor Office [ILO], 2013). According to the most recent official census in 2020, 12.1 million Malaysians who aged 15–40 years old represented 37.12% of the country total population (Department of Statistics Malaysia [DOSM], 2020). Moreover, convenience sampling was employed to gather the most accessible respondents. With the use of G-Power version 3.1.9, the size of the sample for this study was determined to be 153 through an effect size of 0.15 and power of 0.95 for seven determinants (Faul et al., 2007). Nevertheless, data were obtained from 370 Malaysian youth to manage any potential problems caused by the small sample size. Following that, the online survey was conducted through a Google Form created with the study questionnaire. The online questionnaire was distributed to respondents through emails and social media (e.g., Facebook and WhatsApp).

With minimal modifications, the survey questionnaire was derived from previously tested and verified surveys conducted in other studies. The questionnaire was structured using a clear and unbiased method to ensure that participants would be able to understand and answer the questionnaire without any ambiguity. Furthermore, to measure PE six items were adopted from Wei et al. (2021) and Karjaluoto et al. (2020), while the EE consisted of six items, which were adopted from Venkatesh et al. (2003), Karjaluoto et al. (2020), and Wei et al. (2021). The five items that gauged HM were implemented by Karjaluoto et al. (2020) and Liébana-Cabanillas et al. (2021). The PR was assessed using five items adopted from Karjaluoto et al. (2020) and Sobti (2019). For the measurement of SI, five items from Venkatesh et al. (2003), Rahi and Ghani (2019); Dutot (2015), and Wei et al. (2021) were consolidated. The FC was measured with five items taken from Morosan and DeFranco (2016) and Wei et al. (2021). Moreover, six items to evaluate INFC were acquired from Dutot (2015). Finally, four items for measuring ANFC were obtained from Morosan and DeFranco (2016). Overall, all the items were requested to be answered through a five-point Likert scale, in which its five items ranged from “strongly disagree” to “strongly agree.”

Harman’s single-factor test is the most often used strategy among academics to manage Common Method Variance (CMV), given that this test affirms that CMV has no impact on the study model (Chang et al., 2010). In this study, the single-factor test resulted in the highest factor accounting for 48.974%, which was lower than the recommended maximum limit of 50%. This result also indicated the insignificant effect of CMV on this study (Podsakoff et al., 2012).

Although the multivariate normal distribution is not necessary for the PLS approach, Peng et al. (2012) cautioned against making broad generalizations about PLS capacity for evaluating a model, given that this action might violate the criteria of multivariate normality. The multivariate data normality for this study dataset was evaluated using Web Power, which was an online tool for validating data normality.¹ Upon the assessment of the multivariate coefficient (output of skewness and kurtosis computation), p-values were found to be below the recommended threshold value of 0.05 (Cain et al., 2017), proving that the dataset was not normal.

Given that this study was an exploratory type with non-normality issues, variance-based partial least squares structural equation modelling (PLS-SEM) was employed to analyse the study dataset. The PLS-SEM is a causal modeling approach that minimizes the explained variance of endogenous latent components (Hair et al., 2011). As highlighted by Hair et al. (2019), the data analysis comprises (a) descriptive analysis (mean and standard deviation [SD]), (b) internal reliability (Cronbach’s alpha and composite reliability [CR]), (c) convergent validity using average variance extracted (AVE), (e) discriminant validity (Fornell–Larcker criterion, and loadings and cross-loadings), (f) path coefficient (β), (g) predictive relevance (Q²), and (h) coefficient of determination (r²).

Table 1 illustrates the demographic characteristics of the participants. The majority of the 370 respondents (63.5%) were male. Following that, 60.0% of the participants were aged between 21 and 25 years old, while 18.9% of them were under the age of 21, and 29.1% of them were aged between 26 and 30 years old. Furthermore, the monthly income under RM2,500 was the most common amount among the participants (67.6%), followed by the monthly income between RM2,501 and RM5,000 (23.8%). While only 10.8% of the respondents were married, 88.9% of them were single. The majority of participants (65.7%) had completed Bachelor’s degree or equivalent, while 17.6% of them had fulfilled diploma/technical school level education. Moreover, 89.7% of respondents stated that they reside in cities, while 10.3% of them stated that they reside in rural regions.

Cronbach’s alpha provides a reliability analysis based on the inter-correlations between the indicators. As demonstrated in Table 2, all of the indicators showed Cronbach’s alpha values of higher than 0.9, indicating that they are dependable. Furthermore, with a cut-off value of higher than 0.7, CR is employed as an alternative to analyse internal consistency (Hair et al., 2019). The CR values for all the components remained higher than 0.842 as shown in Table 2, indicating a high degree of internal consistency and reliability. Moreover, the AVE is a standard indicator for assessing convergent validity, which determines to what extent can the variation in the indicators be explained by the latent variables. All constructs with the AVE values of higher than 0.654 (Table 2) have proven a strong convergent validity due to their ability to explain over 65.4% of the variance in their indicators on average (Hair et al., 2017).

The Fornell–Larcker criteria are evaluated by emphasizing that the square root of the AVE of each predictor should have a greater value than the highest association with any other construct (Hair et al., 2017). As shown in Table 3, all of the elements fulfilled the Fornell–Larcker criteria required parameters. Additionally in Table 3, the Heterotrait–Monotrait Ratio (HTMT) values of all factors were reported below the threshold of 0.9 as recommended by Henseler et al. (2015). Cross-loading analyses were also performed to determine the discriminant validity. To assess the model appropriateness, an indicator’s outer loading on the associated construct should be more significant than any of its cross-loadings (e.g., correlation) on other constructs (Hair et al., 2017). The cross-loading values presented in Table 4 illustrated that all of the items fulfilled the criteria through their maximum loading value with respective components.

The first section of Table 5 presents the analysis of the correlations between INFC and PE, EE, HM, PR, and SI, which demonstrated that PE (β-value = 0.348, and p-value = 0.000), HM (β-value = 0.251, and p-value = 0.000), and SI (β-value = 0.298, and p-value = 0.000) showed positive beta coefficients and statistically significant p-values. It was proven from the results that PE, HM, and SI showed favorable substantial impacts on INFC, which supported the hypotheses H1, H3, and H5. Furthermore, the path analysis demonstrated that though EE showed a positive path coefficient (0.097), a non-significant p-value of 0.065 was also present. This result indicated that EE had no substantial influence on INFC, which refuted the hypothesis H2. Similarly, PR showed a non-significant p-value of 0.392, which indicated that PR had no considerable influence on INFC. In line with the research model, PR should have a negative association with INFC. Given that data analysis illustrated a positive path coefficient value of PR (β-value = 0.008), indicating that hypothesis H4 was not supported. In the second part of the path analysis, Table 5 presents the correlations of FC and INFC with ANFC. The results indicated that both FC (β-value = 0.311, p-value = 0.000) and INFC (β-value = 0.457, p-value = 0.000) had positive path coefficients and statistically significant p-values. Thus, FC and INFC were suggested to have substantial favorable influences on ANFC, which was in line with hypotheses H6 and H7.

The coefficient of determination (r²) assesses the study model predictive power, which demonstrates the number of variances in endogenous constructs that are represented by all the exogenous constructs associated with it (Hair et al., 2017). The values of r² that exceed or are equal to 0.75, 0.50, or 0.25 are categorized as significant, moderate, or weak, respectively (Hair et al., 2017). Overall, the components PE, HM, and SI presented a significant proportion (r² = 74.9%) of the variation in adoption intention of NFC M-payment. Simultaneously, FC and INFC were able to explain a moderate proportion (r² = 52.6%) of the variance in the actual adoption of NFC M-payments. As a predictive accuracy criterion, the Q² value should be examined besides the r² values (Hair et al., 2017). Moreover, Q² values higher than zero in the structural model for a specific reflective endogenous latent variable illustrated the model predictive relevance for the dependent variable (Hair et al., 2017). The values of Q² (e.g., 0.587 for INFC and 0.373 for ANFC) presented in Table 5 were found to have values higher than zero, indicating that all of the components had a significant predictive relevance.

The mediating effects of INFC on the relationships between ANFC and PE, EE, HM, PR, and SI are presented in Table 6. The result indicated that INFC positively and significantly mediated the association between ANFC and PE (β-value = 0.159, and p-value = 0.000), HM (β-value = 0.115, and p-value = 0.000), and SI (β-value = 0.136, and p-value = 0.000). Furthermore, the mediation effects of INFC in the association of ANFC and EE had a non-significant p-value (0.081), revealing that INFC had no significant mediating impact in the relationship. Simultaneously, mediation analysis of INFC in the association of ANFC and PR resulted in a p-value of 0.394, which was larger than 0.05. This result indicated that INFC had no significant mediating influence in the relationship between ANFC and PR.

An importance-performance matrix analysis was conducted using the ANFC. All seven constructs aimed to determine the factors that were more significant in the adoption of NFC M-payments consumers. The importance-performance matrix analysis (Table 7) illustrated that even though EE had no significant influence on ANFC, it showed the highest performance value (79.054). The second-highest performance value (77.791) was for PE, followed by FC (73.362), INFC (72.877), HM (71.365), and SI (66.464). For the total effect evaluation, the most crucial factor was INFC with the total effect value of 0.457, followed by FC (0.311), PE (0.159), and SI (0.136).

This study developed a sophisticated framework that extended on the well-known theoretical model (e.g., UTAUT2) to further analyse other crucial components and constraints of NFC M-payments adoption, including the possibility for consumers to respond to those factors. Based on the statistical analysis, the proposed model showed a high level of explanatory power in terms of forecasting customer adoption intention and actual use of NFC M-payments. The study conceptual framework has theoretically augmented the UTUAT2 model by incorporating an important factor, such as the PR, which was found to be rarely investigated in the earlier studies of NFC M-payments adoption in Malaysia. Moreover, the impact of PR was found to contradict the previous findings on contactless payment adoption by consumers from other countries. This factor could be considered an important contribution of this study to the existing literature, which created opportunities for further investigation in other circumstances.

The primary goal of this study is to determine and validate several components that strongly influenced behavioral intention and the actual adoption of NFC M-payments among young users. Given the strong influence of PE on INFC, service providers should work on improving the functionality and efficiency of the services to ensure that the M-payment services could be adopted and used more productively to fulfill the consumer-specific purposes. Financial corporates should launch additional promotional efforts advertising NFC M-payments, highlighting their effectiveness and convenience of use (e.g., faster and simpler purchasing, higher productivity, improved transaction performance, and secure transactions at any time and place) and could foster the users’ attention. Given that SI and HM were two crucial drivers for NFC M-payments adoption among users, service providers should consider various methods to increase the appeal of their services to consumers through the addition of sophisticated add-ons, entertaining features, and freebies with promotional deals to encourage customers to use their services more frequently and generate positive word-of-mouth. Moreover, marketers should consider appointing popular social influencers to campaign the compelling advantages of using the services to boost popularity. Service providers should identify the early adopters and encourage the use of NFC M-payments as they may serve as references in the future to enable more widespread adoption. Marketers should focus on the social media marketing methods to promote the word-of-mouth and favorable recommendations since the social media networks have improved this interactions between users. Based on the study’s emphasis on FC as a strong driver of actual acquisition of NFC M-payments, providers should deploy an advanced machine learning algorithm to simplify transaction procedures and make facilitating features easily accessible to consumers (e.g., strong network coverage, platform-independent application, coverage for all business sectors, increasing merchants, partnering all commercial banks, and financial institutes among others). Although PR was reported to be a non-significant factor in users’ adoption of NFC M-payments, this study emphasized the significance of concentrating and extending certified resources to develop a more secure system infrastructure and environment. Because consumers are worried about their privacy and personal information, mobile payment providers must use enhanced encryption tools to support transactions more securely (Flavián et al., 2020). Furthermore, in the event of scamming assaults, a money-back guarantee approach might be another successful marketing strategy that would be more advantageous in resolving risk barriers. Service providers should also assure the availability of on-demand and on-time expert assistance for any concerns that may occur upon the use of the system.

Although this study represented a successful endeavor in the domain of NFC M-payments acceptance and use, several constraints were also present. First, the study used a limited sample of 370 Malaysian young users of NFC M-payments, which restricted the generalizability of the findings. The demographic characterization of the study dataset also revealed that the majority of respondents were young in age and well-educated, and considered to have received adequate technology and Internet knowledge and skills. As a result, several concerns were present regarding the application of the results to other groups of the contemporary population with distinct demographic features (e.g., income, age, gender, education level, and technology experience). Therefore, it was suggested that future research extends the findings of this study by including an analogous research model with a larger sample size focusing on various demographic populations from the cross-country perspective on the distinct cultural, environmental, economic, and technical contexts. This action could be made by investigating the more determinants of behavioral intents of adopting NFC M-payments. Furthermore, the current study findings were based on the cross-sectional data, which raised concerns regarding the long-term relevance of these findings. This issue might be attributed to the fact that customers’ attitudes, perceptions, and awareness of these types of technologies have a higher possibility to transform over time. Finally, further longitudinal studies are hoped to create a better understanding of the issues and the solutions to stabilize or modify the influence of the elements over time.