Students’ acceptance of WhatsApp as teaching and learning tool in distance higher education in sub-Saharan Africa

Abstract Many educators are concerned about students’ use of WhatsApp for learning purposes, especially in emergency remote teaching during the COVID-19 pandemic. This study applied the unified theory of acceptance and use of technology (UTAUT) model to examine factors that predict distance students’ acceptance of WhatsApp for learning. Correlational design was employed. Questionnaires were used to collect data from 273 undergraduate and post-graduate diploma distance students in Ghana. Data were analyzed using partial least squares-structural equation modeling (PLS-SEM). The results showed that performance expectancy, effort expectancy, social influence, and facilitating conditions are significant predictors of distance students’ intentions to use WhatsApp to support learning. However, mobile phone self-efficacy without mediation was not a significant predictor of behavioral intention. Finally, use behavior was significantly predicted by behavioral intention but not facilitating conditions. It was recommended, among other things, that before adopting WhatsApp for instructional activities, management and faculties of educational institutions should ensure the availability of suitable conditions such as WhatsApp-supported mobile devices for students’ use.


PUBLIC INTEREST STATEMENT
Interactions among learners and between learners and instructors are significant for learner success in all modes of educational delivery; fully face-to-face, blended, and distance. Learners need means of communication even more in distance education where they mostly study independently or in the absence of peers and instructors. For this reason, researchers and educational technologists pursue the practice of integrating modern communication technologies in supporting teaching and learning. This study assessed factors that influence undergraduate and post-graduate distance students to use WhatsApp to interact with their peers and instructors for the purpose of seeking learning support outside occasional face-to-face tutorial sessions. The study found that students' intention to use WhatsApp for supporting learning was significantly predicted by performance expectancy, effort expectancy, social influence, and facilitating conditions. Also, learners' actual use of WhatsApp as learning-support tool was significantly predicted by behavioral intention. Based on the findings, recommendations were made for the successful adoption of WhatsApp as a learning-support tool in distance education and directions for future research.
to increase their interest and success in their study. Again, Fatah also found that when students were engaged in interactions through WhatsApp messaging as part of their learning process, they attained a significant improvement in writing skills.
In view of the aforementioned educational benefits of WhatsApp, educational practitioners continue to explore possible ways of integrating it in instructional and administrative processes for students' benefit. As observed by Smith (2015), some higher education institutions adopt WhatsApp for communicating upcoming events and other campus news to students. Thus, WhatsApp facilitates synchronous and asynchronous communication, which enhances closer social interactivity (Vrocharidou & Efthymiou, 2012). Evidently, as noted by Vrocharidou and Efthymiou, WhatsApp provides significant opportunities for interactivity in distance education. In view of opportunities WhatsApp offers through interactivity among students in higher education, some researchers take interest in how it can be utilized to enhance social and academic lives of students (Bakker et al., 2007;Flanagin, 2005).
WhatsApp usage is characterized by some features that warrant its incorporation in distance and blended learning. These are technological, pedagogical, and social affordances (Ankeny, 2019;Tang & Hew, 2017). WhatsApp's Social affordances enable users to register their presence online as real people by communicating appropriate moods, emotions, and feelings (Tang & Hew, 2017). Thus, messaging capabilities make WhatsApp useful for communication among students and between students and faculty to support discussion activities (Hou & Wu, 2011;Lauricella & Kay, 2013). The discussion feature of WhatsApp allows learners to create committed learning communities purposely to interact and share meaningful ideas in a course (Barhoumi, 2015). For instance, in a forensic science course for biology students, Cifuentes and Lents (2011) found that instant messaging encouraged students to draw closer to their professor with confidence as well as to increase their interest and success in their study. A relatively recent study showed that WhatsApp usage in a learning environment can support learning approaches such as inquiry learning that support creativity, critical thinking, collaborative skills, and critical reflection. As Garrison (2016) noted, forming a learning community fosters social presence, a key factor that influences the construction of knowledge and skills. (2016) The technological affordances of WhatsApp are its features that enable users to perform desired tasks easily (Burden & Atkinson, 2008;Tang & Hew, 2017). These include a text field for typing messages, tapping to access or send messages, making or answering voice or video calls, the ability to capture images through a device camera, etc.
Pedagogical affordances are the abilities that WhatsApp offers to teachers and learners to perform teaching and learning activities (Kreijns & Kirschner, 2001). These include posting text messages, making voice and video calls, which enable learners to solicit assistance from peers or teachers to clarify difficult course contents. These affordances further enhance interaction, collaboration, and flexible learning (Ankeny, 2019). The aforementioned affordances and effectiveness of WhatsApp for educational use, coupled with it being most popular among higher education students, informed the choice of WhatsApp for this study.

Statement of the problem and significance of the study
Research on the application of WhatsApp in education mostly focused on how to use it to support learning and its impact on students' learning (Barhoumi, 2015;Cifuentes & Lents, 2011;Fattah, 2015). However, factors that determine students' acceptance of WhatsApp to support learning had not been researched. This study sought to fill the gap in the literature by assessing factors that predict students' acceptance of WhatsApp messaging for supporting learning in a distance education context. Knowledge of such factors is essential for policy decisions and the successful adoption of WhatsApp as a learning-support tool in higher education. Without such knowledge, efforts to integrate WhatsApp as a learning-support tool among distance learners may prove futile.

Research objectives
(1) To assess the significance of UTAUT predictor variables in determining distance students' acceptance of WhatsApp messaging for supporting learning.
(2) To assess the nature of relationship between mobile self-efficacy and behavioral intentions of distance students in accepting WhatsApp for supporting learning.

Theoretical framework
This study was underpinned by the unified theory of acceptance and use of technology (UTAUT) model (Venkatesh et al., 2003) as its theoretical framework. The model was originally formulated for assessing the likelihood that a target population will successfully accept and utilize an information technology or information system introduced to them in an organizational setting. It also helps to understand the drivers that influence acceptance of new technologies in order to proactively design interventions targeted at populations of users that may be less inclined to adopt the technology (Venkatesh et al., 2003). The formulators reviewed and compared conceptual and empirical similarities across eight prominent technology acceptance models and subsequently formulated the UTAUT model.
According to the UTAUT, the determinants of behavioral intention are performance expectancy, effort expectancy, and social influence. Both behavioral intention and facilitating conditions are direct determinants of use behavior. The variables of gender, age, experience, and voluntariness of use were said to moderate the key relationships in the model (Venkatesh et al., 2003). Thus, the moderators either increase or decrease the effects of the independent variables on the dependent variables. Venkatesh et al. (2003) define the key constructs of the UTAUT model as follows. Performance expectancy is the extent to which a prospective user of a new technology believes that using the technology will result in improvement in job performance. Effort expectancy refers to how easy or difficult a prospective user perceives the use of a new technology to be for him or her. Social influence is defined as the extent of one's perception that people who are important to him or her believe that he or she should use new technology. Facilitating conditions is an individual's perception of the existence of technical support and other resources provided by an organization to assist in the use of a given technology.
The UTAUT model was chosen as the theoretical framework in this study for a number of reasons. First, it explains as much as 70% of the variance in behavioral intention to use an information technology or information system (Venkatesh et al., 2003). According to Venkatesh et al., this percentage of variance far outweighs that explained by each of the models from which the UTAUT model was derived. Thus, the UTAUT model is more efficient to explain behavioral intentions to use technology.
Secondly, the UTAUT model is relatively new and some researchers have validated its ability to predict user acceptance of information technology in a variety of contexts (AL-Youssef, 2015). Thus, it is deemed viable for the context of this research as well.
Also, constructs used in the UTAUT model were adopted from eight theories in social psychology and sociology. Hence, it is considered to be more comprehensive in measuring behavioral intentions to use technology.
Finally, the UTAUT model was chosen over UTAUT2 (Venkatesh et al., 2012) in that the current study was set in an organizational environment other than consumer environment. Thus, the use of WhatsApp for teaching and learning was motivated by institutional policy, but not hedonic motivation. Also, prior studies  had provided empirical evidence that the target population already owned mobile phones that supported WhatsApp and had been using the same for general communication. Hence, price value, a construct of UTAUT2, was not relevant in the context of this current study.
The preceding hypotheses formulated on the basis of existing literature led to a proposed conceptual model to guide this study. The model named mobile self-efficacy and linear relationships-based unified theory of acceptance and use of technology (MSELR-UTAUT) model is illustrated in Figure 1.
Mobile self-efficacy, a new construct the researchers included in the MSELR-UTAUT model, has been found to play important roles in the adoption of mobile devices to supplement education. Nikou and Economides (2017) as cited in Chao (2019) define mobile self-efficacy as perceptions of an individual about his or her ability to use mobile devices to perform some particular tasks. Chao (2019) analyzed the effect of mobile self-efficacy on behavioral intention mediated by a chain of two constructs. Chao found that mobile self-efficacy had a significant effect on an endogenous variable Perceived Enjoyment, which in turn influenced Satisfaction. Finally, Satisfaction was also found to influence Behavioral Intention. Thus, Chao argued that Self-efficacy indirectly influenced Behavioral Intention to accept and use mobile learning. There is no empirical evidence regarding the direct effect of mobile self-efficacy on behavioral intention, but this study seeks to unravel, if any, to fill the gap in the literature.

Materials and methods
This study employed a correlational design with the justification that it offers researchers the opportunity to predict and explain relationships among two or more variables (Creswell, 2012;Leedy & Ormrod, 2010). A sample of 273 distance students was selected from two study centers (campuses; names withheld for confidentiality) of University of Education, Winneba (UEW) in Ghana. Selection of respondents from the two centers was based on their proximity to the location of the researchers that offered convenience in travelling for data collection. Moreover, students at the two centers had similar characteristics as those at other centers, hence they could be representative of the others. A proportional stratified random sampling technique was employed in sample selection. This ensured representativeness of the proportion of males and females as in the population of students at the centers. Thus, possible bias in the responses as a result of differences in sex was eliminated. The sample size was determined using Krejcie and Morgan's (1970) formula.
Data were collected using a structured questionnaire, which comprised three main sections. The first section introduced the purpose of the study, sought students' voluntary participation, spelled out expected roles of participants, and assured confidentiality and anonymity of responses. The second section comprised 24 statements, each associated with 5-point Likert-type scale responses ranging from 1 (strongly disagree) to 5 (strongly agree). These measured seven latent constructs: performance expectancy, effort expectancy, social influence, facilitating conditions, behavioral intention, use behavior, and mobile self-efficacy. The construct indicator statements were adapted from Venkatesh et al. (2003), Evans (2013), and Chao (2019). Those studies had already subjected the indicator items to rigorous technical processes of construction, validity, and reliability testing. For instance, Venkatesh et al. reported that internal consistency reliabilities of all the items were greater than .70. The reliability and validity estimates were further confirmed in this study to be acceptable as presented in Table 1. The third section of the questionnaire comprised eight items that solicited socio-demographic data from the participants.
Prior to data collection, the principal researcher sought written permission from the coordinators and administrators of the study centers. Hardcopy questionnaires were subsequently administered by the principal researcher. On each day, some students belonging to the sample visited the study center administrators' offices to collect, complete, and return copies of the questionnaire. The principal researcher was present to check the returned questionnaires for missing or illegible responses. Participants whose responses contained critical anomalies were prompted on the instant of submission for completion when necessary. The return rate of the instrument was 100%.
The data were subsequently coded and entered into SPSS software to create a data file. The data were analyzed descriptively using SPSS version 25. Regression analysis was performed in SmartPLS 3.2.7 . All hypotheses were tested at an error margin of 5%. This means relationships between variables were considered significant at the probability of 0.05 or less (p ≤ 0.05) and thus null hypotheses were rejected.

Demographic data
The participants comprised 133 (49%) females and 140 (51%) males, aged from 20 to 48 years (M = 30.4, SD = 5.03). Age was non-normally distributed, with skewness of 1.32 (SE = 0.15) and kurtosis of 2.02 (SE = 0.29). The participants further comprised 89 (33%) Diploma, 118 (43%) Bachelor's degree, and 66 (24%) Post-Graduate Diploma students. Other characteristics of the participants are presented in Table 1. Table 1 shows further participants' characteristics with associated numbers and percentages of students. As shown, 269 (98.5%) students owned WhatsApp-supported mobile phones as against 4(1.5%) who owned none. This shows that WhatsApp-supported mobile phones are common among undergraduate and post-graduate distance students at the study centers. Also, 272 (99.6%) students indicated that they had been using WhatsApp, with the exception of one, who indicated he/she had not been using WhatsApp prior to the study. Lastly, 119 (43.6%) of the participants had been using WhatsApp for more than 6 years, whiles 144 (52%) had WhatsApp usage experience of 1-6 years inclusive. Only one student indicated no prior experience in WhatsApp usage. These characteristics indicate the participants had the requisite resources and experience to accept and use WhatsApp to support their distance learning.

Measurement model analysis
The measurement models of the research model were of a reflective type. Hence, they were assessed by testing the individual indicator reliability, internal consistency reliability, convergent validity, discriminant validity, and cross-loadings as recommended by Hair et al. (2014). Individual indicator reliability was assessed using outer loadings of individual measurement items. Internal consistency reliability was assessed using rho_A and composite reliability. Also, convergent validity was assessed using Average Variance Extracted (AVE). Finally, discriminant validity was assessed using Heterotrait-Monotrait (HTMT) ratio. The aforementioned measurement model assessments were performed by initially running PLS Algorithm in SmartPLS version 3.2.7 . The initial output of the PLS algorithm is shown in Figure 2. Figure 3 shows outer loadings for each of the individual items in the measurement models. Hair et al. (2014) recommended that outer loadings are considered significant when they are 0.708 or higher. It was observed in Figure 3 that one indicator item of mobile self-efficacy (MSE2) had an outer loading of 0.689. This value is below the recommended threshold; thus, it is not significant. Subsequently, MSE2 was deleted from the model and the PLS Algorithm re-ran. The output of the PLS Algorithm after MSE2 deletion is shown in Figure 4.
The output of the PLS Algorithm after removal of MSE2 shows that the outer loadings of all the items in the measurement models ranged from 0.722 (UB3) to 0.962 (MSE1). These values were all greater than the recommended minimum significant value of 0.708 (Hair et al., 2014). Thus, the items of each construct were ascertained to have significant convergent validity.
Next in the assessment of the measurement model is the consideration of internal consistency reliability and convergent validity. Hair et al. (2014) recommend that the use of composite reliability for the assessment of internal consistency reliability is more robust because it overcomes the limitation of Cronbach's alpha. Furthermore, Hair et al. recommend the use of Average Variance Extracted (AVE) for the assessment of convergent validity. The results of the PLS algorithm for the assessment of internal consistency reliability and convergent validity are shown in Table 2. Table 2 indicates that the range of values for composite reliability of the measurement model was 0.848 to 0.947. These values were all greater than the minimum recommended value of 0.7 (Hair et al., 2014). Also, the AVE values ranged from 0.650 to 0.857. Thus, the AVE values were all greater than 0.5 which is the minimum recommended value (Hair et al.). The range of values for composite reliability and AVE in Table 1 established that the measurement models had acceptable internal consistency reliability and convergent validity.
Next, the measurement model was assessed for discriminant validity. This was done using Heterotrait-Monontrait (HTMT) ratio as recommended by Henseler et al. (2015). The values of the HTMT ratio for the measurement model as obtained from the PLS algorithm are shown in Table 3.  Table 3 that all the HTMT ratios of the measurement model are less than the recommended upper limit of 0.85 for the strict sense (Henseler et al., 2015). Hence, the HTMT ratios confirmed that the measurement model passed the discriminant validity test. The next section presents an analysis of the structural model.

Structural model analysis
The second stage in the assessment of PLS-SEM is to assess the structural model after the measurement models pass all the relevant tests as discussed in the preceding section. The structural model assessment involves five steps as outlined by Hair et al. (2014). These involve tests for collinearity, significance, and relevance of path coefficients in the structural model, level of R 2 values, effect sizes of f 2 , predictive relevance (Q 2 ), and q 2 effect sizes.

Collinearity assessment
Collinearity assessment is used to determine the existence or otherwise of a phenomenon known as Common Method Bias (CMB) in the structural model. CMB occurs when indicators share a certain level of common variation as a result of the existence of some form of influence or clues that make respondents answer the questionnaire in the same way.
A full collinearity test is recommended for the assessment of a structural model for the existence of both vertical and lateral collinearity simultaneously (Kock & Lynn, 2012). In the process of full collinearity test, SmartPLS software automatically generates Variance Inflation Factors (VIFs) through the PLS Algorithm. The results of Inner VIF values from the PLS Algorithm are shown in Table 4.  The acceptable threshold of VIF values for a structural model to be free of common method bias is less than or equal to 3.3, otherwise, the model is said to contain "pathological collinearity" and CMB (Hair et al., 2017, Kock, 2015. From Table 4, it is clear that the structural model was devoid of collinearity and CMB because all the VIF values were below the recommended maximum threshold of 3.3.

Significance and relevance of path coefficients
In assessing the significance of relationships in the structural model of this study, a bootstrapping procedure was run in SmartPLS 3.2.7  with subsamples of 5000, two-tailed test type and significance level of 0.05 as recommended by Hair et al. (2014). The results are presented in Figure 4 and Table 5. (2005). Table 5 indicates that five of the hypothesized relationships were significant. These are prediction of behavioral intention by performance expectancy (β = 0.252, p ≤ 0.01); effort expectancy (β = 0.128, p < 0.05); social influence (β = 0.343, p ≤ 0.01); and facilitating conditions (β = 0.359, p ≤ 0.01). Also, the prediction of use behavior by behavioral intention is significant (β = 0.432, p ≤ 0.01). However, two other hypothesized relationships were non-significant. These are prediction of behavioral intention by mobile self-efficacy (β = −0.091, p > 0.05) and prediction of use behavior by facilitating conditions (β = 0.166, p > 0.05). Table 5 were relevant in establishing the existence or otherwise of significant relationships between the exogenous variables and their counterpart endogenous variables. However, these values did not provide information on the strength of the effects each exogenous variable had on its associated endogenous variable. This limitation was addressed using the f 2 effect size as recommended by Hair et al. (2014). A recommendation for assessment of the f 2 effect size values in a model was made by Cohen (1988) that the values 0.02, 0.15, and 0.35 represent small, medium, and large effect sizes, respectively. Contrary to Cohen's recommendation, Kenny (2018) argues that these effect size values are unrealistic. Kenny's argument was based on a review by Aguinis, Beaty, Boik, and Pierce (2005) as cited in Kenny (2018) which showed that the average effect size in the moderation test is only 0.009. Kenny (2018) therefore recommended new realistic values of effect sizes thus 0.005, 0.01, and 0.025 for small, medium, and large, respectively. Applying Kenny's classification of f 2 effect sizes to the significant relationships in Table 5, it is obvious that effort expectancy had a medium effect size (0.017) on behavioral intention. Also, performance expectancy, social influence, and facilitating conditions had large effect sizes (0.095, 0.195, and 0.186 respectively) on behavioral intention. Finally, behavioral intention also had a large effect size on use behavior (0.171).

Coefficient of determination (R 2 value)
The structural model in this study was also evaluated using the coefficient of determination, R 2 . This value measures the predictive accuracy of a structural model. It represents the combined effects of exogenous variables on their related endogenous variable. The coefficient of determination explains how much of the variance in an endogenous variable is predicted by its antecedent exogenous variables. The values of R 2 range between 0 and 1 inclusive. The closer the R 2 value gets to 1, the higher its level of predictive accuracy. Conversely, the closer the R 2 value gets to 0, the lower its predictive accuracy. The R 2 values of the endogenous constructs of the structural model are shown in Table 6.
From Table 6, the R 2 value of 0.630 for behavioral intention indicates that the exogenous constructs contributed up to 63% of the variance in behavioral intention and its adjusted value for model comparison is 62.3%. Likewise, the model explained up to 30% of the variance in use behavior.

Discussion
The results of data analysis showed that the intention of distance education students to accept WhatsApp for supporting their learning is positively and significantly predicted by performance expectancy. This finding is consistent with several other studies (Almogheerah, 2020c;Ameen & Wills, 2018;Humaid & Ibrahim, 2019;Israel & Velu, 2019;Jambulingam, 2013;Khechine et al., 2014;Lakhal et al., 2013;Nikolopoulou et al., 2020;Tan, 2013;Venkatesh et al., 2003). All these researches, though conducted in different contexts and on a variety of technologies, found that performance expectancy predicted behavioral intention for the adoption of new technologies. Nevertheless, the findings of this study contradicted other studies (Attuquayefio & Addo, 2014;Yueh et al., 2015) that found no significant relationship between performance expectancy and behavioral intentions to accept new technologies. The significant positive relationship between performance expectancy and behavioral intention could be attributed to the fact that the students were in need of learning-support systems besides the face-to-face tutorials they attended on weekends at 2-week intervals, hence they were ever ready to accept any provision that would address their learning needs. Also, they were optimistic that engaging in WhatsApp chat to discuss their course contents besides tutorial sessions could be  an opportunity to improve upon their learning. Hence, students' high expectation that WhatsApp group chat would help improve their learning positively influences their intention to adopt the same in their learning.
A second finding of the study was that a positive and significant relationship exists between effort expectancy and behavioral intentions of distance education students in acceptance of WhatsApp to support learning. This finding is consistent with Venkatesh et al. (2003) and other later researches (Almogheerah, 2020c(Almogheerah, , 2020cAmeen & Wills, 2018;Attuquayefio & Addo, 2014;Chaka & Govender, 2017). This finding could be attributed to the fact that the students were already familiar with the use of WhatsApp. Students' familiarity with WhatsApp usage enhanced their belief that using WhatsApp chat to support their learning would be effortless because they had already been engaging in WhatsApp chat with friends and relatives for non-academic purposes (see Table 1). Hence, the students readily intended to accept the technology.
Thirdly, the study also found a non-significant relationship between mobile self-efficacy and behavioral intention. This is consistent with some previous studies (Venkatesh, 2000;Venkatesh et al., 2003) yet inconsistent with others (Chao, 2019;. Venkatesh (2000) maintained that computer self-efficacy is an indirect predictor of behavioral intention fully mediated by perceived ease of use (effort expectancy). This implies that without mediating effect of effort expectancy, self-efficacy would not determine behavioral intentions; a clear case of consistency with the finding of this current study as mobile self-efficacy was not mediated by effort expectancy. Moreover, Venkatesh et al. (2003) postulated and empirically confirmed that computer self-efficacy does not significantly determine behavioral intentions to accept computer technology.  however found significant relationship between mobile self-efficacy and behavioral intention due to the mediating effect of effort expectancy.
A fourth finding is that social influence has a significant positive relationship with the behavioral intention of distance education students in acceptance of WhatsApp as a learning-support tool. This positive relationship implies that the more distance education students strongly perceive that important referent others believe the should use WhatsApp to support their learning, the stronger the intention they develop towards acceptance of WhatsApp for supporting learning, and vice versa. This finding is inconsistent with Venkatesh et al. (2003). Venkatesh et al. reported that the effect of social influence on behavioral intention is not significant unless moderated by other factors. It is clear in this study that social influence determines behavioral intention even in the absence of moderators. Though this finding is inconsistent with the original UTAUT postulation, it is consistent with many later studies (Bere, 2014;Humaid & Ibrahim, 2019;Khechine et al., 2014;Yueh et al., 2015). All these studies reported that social influence has a significant effect on behavioral intention even in the absence of moderators.
Another finding is that facilitating conditions have positive significant relationship with behavioral intention of distance education students to accept WhatsApp for supporting their learning. This finding is contrary to the original postulation of the UTAUT model by Venkatesh et al. (2003) which was later supported by Israel and Velu (2019). The aforementioned authors postulated and empirically confirmed non-significant relationship between facilitating conditions and behavioral intentions. In another respect, Venkatesh et al. concluded that the influence of facilitating conditions on behavioral intentions is only significant if effort expectancy is absent in the model. However, the current study has refuted that claim and showed that facilitating conditions significantly predict behavioral intentions directly, even in the presence of effort expectancy. Nevertheless, the finding of this current study is consistent with many other studies (Almogheerah, 2020c;Chaka & Govender, 2017;Humaid & Ibrahim, 2019;Khechine et al., 2014;Raman et al., 2014;Toh, 2013). All these studies reported significant relationship between facilitating conditions and behavioral intentions to acceptance of technologies.
Again, this study found non-significant relationship between facilitating conditions and use behavior. This is consistent with Venkatesh et al. (2003). These authors found in the original UTAUT model that the relationship between facilitating conditions and use behavior was significant only when moderated by age and experience. Hence, without moderating effect of age and experience, the relationship between facilitating conditions and use behavior would not be significant. This is the case in this study because moderator variables were excluded from the model. Contrarily, this study is inconsistent with Humaid and Ibrahim (2019), Almogheerah (2020c), and Tan (2013). In the case of Humaid and Ibrahim, facilitating conditions significantly determined use behavior even though the relationship was not moderated by age and experience. Somewhat similarly, Almogheerah (2020c) and Tan (2013) found significant effect of facilitating conditions on use behavior, in the absence of moderating variables.
The explanation of the non-significant effect of facilitating conditions on use behavior in this study is that the use of WhatsApp by distance education students is not dependent on provision of support-systems by the university. This is understandable because the students had already been using WhatsApp (see Table 1) before the researchers proposed its usage as a learning-support tool.
Finally, behavioral intention was found to significantly and positively predict use behavior. This is consistent with some earlier studies (Almogheerah, 2020c;Tan, 2013;Venkatesh et al., 2003). A possible explanation of this relationship is that when people form a stronger intention to adopt a new technology, they are more likely to use the technology once they have access.

Implications for theory
This study found that performance expectancy, effort expectancy, and social influence are significant predictors of behavioral intentions. These findings corroborate postulations of the original UTAUT model. This study also indicated that facilitating conditions significantly predict behavioral intentions. This contradicts the original UTAUT postulation about facilitating conditions and behavioral intentions. Thus, the study contributes new knowledge about predictors of behavioral intention in acceptance of technology.
Finally, mobile self-efficacy was found to be non-significant in prediction of behavioral intentions without mediation of effort expectancy. This corroborates with earlier findings in both UTAUT and TAM3.

Implications for practice
Before adoption of WhatsApp for supporting learning in higher education, administrators and faculty should collaborate to ensure that determinants of behavioral intentions are adequately catered for. For instance, authorities of the educational institution should provide facilitating conditions such as helping students who do not own WhatsApp-supported mobile devices to acquire some. This can be achieved by the university awarding contracts to individuals or organizations to supply mobile devices at a reduced cost to the students. Once provisions are made by the university to support implementation, the students will be willing to accept the technology. Furthermore, technical support should be provided by the institution to both learners and lecturers to facilitate efficient utilization.
Also, prior to implementation, educational institutions seeking to support distance learning with WhatsApp chat should sensitize the students about the educational benefits of WhatsApp chat in supporting learning. By doing so, the students will anticipate performance gains that will eventually influence their decision to accept the technology.
Finally, instructional designers of distance education courses and programs should specifically incorporate instructional activities involving the use of WhatsApp chat in the courses that lack electronic means of communication among students and faculty. Once the students are made aware that WhatsApp messaging is officially part of their instructional strategies, they will have belief that university authorities expect them to use it. This will subsequently influence their intention to accept WhatsApp for pedagogical use.

Limitations of the study and recommendations for future studies
This study was associated with some limitations. Firstly, the participants were selected from two study centers. This might affect external validity of the findings. Hence, the results must be generalized with caution. It is therefore recommended that future research expands the scope of the study to include more study centers. This would further enhance external validity of the results. Secondly, the moderating variables of age, gender, and experience were excluded from this study. As a result, the possible moderating effects on the relationships between exogenous and endogenous constructs were not observed in this study. Other studies may include gender, age, and experience as moderators in the MSELR-UTAUT model to assess the significance of their effects on the relationships between the exogenous and endogenous constructs. Thirdly, this study utilized the UTAUT model in assessing determining factors of students' acceptance of WhatsApp usage for teaching and learning purposes in a context mandated by instructors. Other studies may utilize the UTAUT2 model to replicate this study in a context where students would voluntarily opt to engage in or initiate use of WhatsApp for learning purposes.