Modelling perceived e-learning service quality, student satisfaction and loyalty. A higher education perspective

Abstract Blended e-learning has become a common phenomenon in higher education globally. Most affluent economies embraced e-learning by design through strategic moves to augment their competitiveness. However, in most emerging economies, e-learning implementation was impulsively reactive to the Covid-19ʹs demands. Thus, the study examined the impact of perceived e-learning service quality on students’ satisfaction and loyalty in a developing country. The expectation-confirmation thty -35 eory informs the development of the conceptual framework. A causal research design enshrined in the positivism research philosophy was adopted. The research population was made up of students enrolled in public and private universities in Zimbabwe. Data was collected through person-administered survey and a stratified sample of 354 students was obtained. The results from structural equation modelling (SEM) revealed significant positive relationships between perceived e-learning service quality dimensions and e-learning student satisfaction. It was also reflected that student satisfaction positively affected student loyalty with e-learning (P < 0.05). It was therefore concluded that system quality, information quality and service quality significantly influence student satisfaction and loyalty with e-learning. The study thus recommended that the higher education industry should design e-learning systems that enhance easy access, easy navigation and user flexibility.


Introduction
The Covid-19 pandemic principally forced service providers across different sectors to consider online servicescapes to remain afloat and competitive. As such, Higher Education Institutions (HEIs) have not been exempted from the crisis. Electronic learning (e-learning) has become the panacea for universities to continue with service delivery amid the deadly pandemic (Alotaibi, 2019). E-learning is defined as any form of learning that is facilitated by the web (Wu, 2016). The learners' perceptions of e-learning service quality have been identified as pivotal to the sustainability of HEIs in extant e-learning literature. The relationship between student loyalty and its antecedents has been extensively investigated in the traditional environment where learning encounters take place in physical structures (Martinez-Arguelles & Batalla-Busquets, 2016).
It is noteworthy to establish that the education delivery mechanism has undergone fundamental changes in the current Covid-19 context and more avenues for research have been presented (Lee & Jeon, 2020). E-learning is also increasingly becoming popular in Higher Education, driven by the concerns for continued service, improvement of operational efficiency and expansion of markets (Lwoga & Sife, 2018;Wu, 2016). Ubiquity in learning and flexibility in lecture delivery have also promoted the need for e-learning. There has been a notable growth in development of educational technologies, which make it easier for the student to connect to a virtual class from anywhere in the world (Marandu et al., 2019).
However, although a notable increase in e-learning is phenomenal, the extent to which e-learning relates to student satisfaction and loyalty is still a grey area for research. The situation is validated by scarcity of research work in developing nations. Sparse e-learning literature in Africa and some developing countries, on e-learning and student loyalty, provides compelling evidence of the slow internet infrastructural development inherent in developing economies (Eli-Chukwu et al., 2022;Mwiya et al., 2019). Moyo and McKenna (2021) cites poor funding as one of the key constraints limiting the growth of academic research in Africa. Conversely, Mwiya et al. (2019) indicated that scarcity of research in developing countries is a proxy to poor ICT development for education. However, Makudza et al. (2022) argue that the Covid-19 pandemic drove African countries to value e-business, which forced them into reactive e-business implementation. Eli-Chukwu et al. (2022) also support the views of Makudza et al. (2022) that most educational institutions reactively implemented e-learning and further contribute that among the challenges of e-learning is research which informs technical implementation of e-learning. Only few research efforts have been directed on how key dimensions of perceived e-learning service quality influence student outcomes especially in developing economies (Agarwal et al., 2021;Pham et al., 2019;Tere et al., 2020;C. Y. Li et al., 2018).
It is against that backdrop that this study seeks to contribute to the body of knowledge by offering research-based solutions to e-learning satisfaction and loyalty. The study sought to investigate the impact of perceived e-learning service quality on student outcomes in the HEIs amid the Covid-19 pandemic. The study's model presents that e-learning system quality, e-learning information quality and e-learning service quality influence e-learning student satisfaction. Conversely, e-learning student satisfaction is modelled as an antecedent of e-learning student loyalty.
The results bear huge practical implications to the higher education industry. Given that the study was motivated by sparse literature on the e-learning success and a void in understanding students' perceptions of e-learning services, this study brings rich empirical findings to HEIs which could be used in strategy formulation and long-term planning on e-learning development initiatives. The study lends an insight to the HEIs on the key dimensions of e-learning quality which promote student satisfaction and loyalty. Thus, the paper provides a validated framework for service improvement in HEIs. In terms of theoretical significance, most e-learning success models tested the impact of e-learning factors on user satisfaction directly from the updated ; Hammouri & Abu-Shanab, 2018;Hermita et al., 2019;Rajasekaran et al., 2022). However, the study became one of the few to directly operationalise e-learning student satisfaction and e-learning student loyalty.

Perceived e-learning service quality
Perceived e-learning service quality is defined as the students' perceptions about the overall e-learning quality they receive from the HEIs (Al Mulhem, 2020). A number of scholars have conceptualized perceived service quality using a number of models dominant in extant services marketing literature. The most common models used across research contexts in different Higher Education services are the SERVQUAL, SERVPERF, HEDPERF, HEDQUAL, SITEQUAL, DINESERV, E-S-QUAL and E-S-REC-QUAL among others (Al Mulhem, 2020;Pham & Tran, 2020). In e-learning studies, most researchers have borrowed from SERVQUAL, SITEQUAL, E-S-QUAL and E-RecS-QUAL models (Parasuraman et al. (2005), the Technology Acceptance Model (TAM; Davis, 1989) and the DeLone and McLean (1992. Researchers have adopted decomposed models from any of the aforementioned models with the goal of improving explanatory power the research models as well as addressing dominant criticisms from empirical studies (Jeyaraj, 2020). In this study, perceived e-learning service quality was conceptualized using variables proposed by the updated DeLone and McLean (2003), measured in terms of the summation of e-learning system quality, e-learning information quality and e-learning service quality (Al-Fraihat et al., 2020). These variables were adopted from the updated DeLone and McLean (2003). These variables were distilled from this model because of their vast application in e-learning success studies as determinants of student satisfaction, e-learning use and continued use intentions (Al-Fraihat et al., 2020;Alzahrani et al., 2019;Lee & Jeon, 2020;Y. Li et al., 2021;Mtebe & Raphael, 2018;Pang et al., 2020).

E-learning system quality
System quality is defined as the user's perception of the performance of the information system itself (DeLone & McLean, 2003. In an e-learning context, system quality is assessed in terms of both the hardware available to the student and the various software applications for use and e-learning needs (Pham et al., 2019). It focuses on the performance characteristics of the system under investigation (Yosep, 2015). The system processing capabilities of e-learning systems are measured in terms of ease of access, availability, log in, security of users, flexibility, appeal and design, ease of navigation, downloading speed and availability (Lee & Jeon, 2020). Cheng (2020) defines information quality as the students' perception of the quality of information they retrieve from an e-learning website. It is also defined as the quality of the content that students download from the e-learning system measured in terms of readability, understandability, format (text, audio and video), adequacy, up datedness, relevancy and transferability (DeLone & McLean, 2003. Information quality is a measure of perceived effectiveness of the system output or content that is downloaded or available on the e-learning website (Chang, 2013). Nugroho et al. (2019) also adds that it is a measure of value that the information provides to the student online. Key attributes of information quality include availability, currency, organization, authenticity, timeliness and relevance. Other dimensions such as adequacy, quality, mode of presentation and sequencing were also deemed important to students' evaluation of information resources online (DeLone & McLean, 2016). Their significance has been reflected in a number of research papers (Marandu et al., 2019).

E-learning service quality
Early internet service quality conceptions were based on the E-S-QUAL model (Parasuraman et al., 2005) and since then, different industries have conceptualized and operationalized online service quality in context-based ways. In e-learning services sector, service quality has been conceptualized and operationalised as the support quality that learners receive from the technical IT support departments as well as the ability of the e-learning website to give clear instructions, steps followed and online troubleshooting manuals or prompts (Eom & Ashill, 2018).
It has also been defined as the perceived effectiveness of Information Technology (IT) and administrative support service staff and the e-learning site's ability to troubleshoot students' problems they face during e-learning sessions online (Al Mulhem, 2020). With the emergence of the internet and the web, online service quality has been an interesting topic for academic researchers. E-learning service quality is a measure of the functionality of the systems (Pham et al., 2019). In this study, e-learning service quality was adopted from the DeLone and McLean (2003). Service quality was introduced as a new construct to the DeLone and McLean (1992) model after an influx of suggestions and recommendations to the model from empirical studies. The dimensions used in the majority of information systems success studies are responsiveness, accuracy, reliability, technical competence, and empathy of the support service providers as well as the ability of the e-learning system to convey clear instructions to users (Yosep, 2015). Y. Li et al. (2021) also support that an e-learning website's ability to rectify students' problems online also represents an important facet of e-learning service quality.
The traditional relationship between service quality and customer satisfaction is well grounded in literature. However, with the emergence of online servicescapes, new insights from varying contexts across service industries have been emerging. In the e-learning services context, DeLone and McLean (2003) adopted service quality to their original model, highlighting its significance on user's satisfaction with an information system. A number of models have been proposed to investigate perceived service quality. In the e-learning context, many researchers have used the DeLone and McLean (2003) model, the SERVQUAL (disconfirmation) model (Parasuraman et al., 1988), the SERVPERF model (performance only; Cronin & Taylor, 1992) and the E-S-QUAL model (Parasuraman et al., 2005).

Student e-learning outcomes: student satisfaction and student loyalty
The student-centered approach is an emerging strategy meant to improve the quality of Higher Education services that has attracted much interest around the world (Pham et al., 2019). The strategy utilises customer orientation with the learning service experiences and is designed to maximize student satisfaction and loyalty with the university. Universities should do their best to deliver the highest service for students, "their customers" (Martinez-Arguelles et al., 2013). Delivering the best educational service results in student satisfaction, which delivers student loyalty. Student outcomes are defined as the ultimate measure of success of an e-learning service offered to students in terms of enjoyment, goal accomplishment and future behavioural intentions (Pham & Tran, 2020). According to Lee and Jeon (2020), an evaluation of student outcomes on an e-learning service experience is imperative for universities and acts as a basis for improving service delivery.
E-learning student satisfaction refers to the degree to which the student believes that e-learning services evoke positive feelings (Chang, 2013). It is also defined as an emotional, attitudinal and judgmental response to e-learning experiences that students are subjected to during successive e-learning sessions (Pham et al., 2019). Student satisfaction is a subjective judgment by the student which sums whether they were happy with the service or not (Pham et al., 2019). Satisfaction is an attitudinal feeling towards the performance of a product or a service (Eom & Ashill, 2018). According to Bhattacherjee (2001), customer satisfaction is an emotional attitude towards a transaction between an individual and a provider. Satisfaction is affected by the quality of service, product quality, price, and contextual and personal determinants.
Generally, customer satisfaction is the objective of every customer-oriented organisation because of its positive relationship with customer loyalty, the same understanding applies for Higher Education Institution (Lukic & Lukic, 2018). According to Martinez-Arguelles and Batalla-Busquets (2016), student satisfaction is a key determinant of student loyalty which comes in form of positive word of mouth to prospective students, active participation in the learning process, active role in the alumni as well as coming back to the HEI for further studies. Martinez-Arguelles and Batalla-Busquets (2016) further claim that e-learning student satisfaction improves the quality of educational achievement among learners in an e-learning service context. E-learning student loyalty is the desire, trust and long-term commitment by a student to repatronise and continue using e-learning services offered by the university and their propensity to spread positive word of mouth to prospective and current students to enroll in online programmes with his or her university (Pham & Tran, 2020). Student loyalty has been defined in relation to customer experience as the degree to which a student is willing to use e-learning systems in future and the willingness to recommend the use of e-learning systems to other students who are contemplating to study or who are not using it (Makudza et al., 2020). Some scholars, for example, Chang (2013) earlier on defined and operationalized e-learning student loyalty as continuance intentions. Student satisfaction is an attitude whilst student loyalty is a behavior (Lwoga & Sife, 2018) and the relationship between the two has been proposed and reflected by a number of scholars.
According to Martinez-Arguelles and Batalla-Busquets (2016), e-learning student loyalty creates a valuable source of competitive advantage for higher education institutions. Further, e-learning student loyalty ensures that the university has cost effective marketing strategy towards enhancing the brand image of the university as students spread good recommendations to their peers (Pham & Tran, 2020). According to Pham et al. (2019), loyal students are an indispensable factor for maintaining the financial stability of a HEI through tuition that can be directed towards sustainable development programmes. From the customer orientation perspective, it is expensive for educational institutions to bring new students to their programmes than to retain existing ones, thus student loyalty has been found critical to the sustainability of a HEI (Kilburn et al., 2016). It is also supported by Lukic and Lukic (2018) that student loyalty is key for the continuity of a HEI, therefore its antecedents are worth of an investigation.

The relationship between e-learning system quality and e-learning student satisfaction
The relationship between system quality and student satisfaction has been explored by a number of scholars who adopted the DeLone and McLean (1992McLean ( , 2003 model. Al-Fraihat et al. (2020) support the existence of a positive correlation between system quality and user satisfaction. They alluded to important dimensions of system quality such as reliability, usability, maintainability and trust as key to satisfaction of users in e-learning. Yosep (2015) also concurs that the efficiency and performance of an e-learning system drives users to use an information system, hence pinpointed a significant positive relationship between system quality and learner's satisfaction. Alzahrani et al. (2019) also indicated that perceived e-learning quality mainly hinges on system quality. In their research in Malaysia, they alluded to resource utilization, reliability, response time, human factors, aggregation of details, system trust and accuracy were important antecedents of good system quality. They adopted the DeLone and McLean (2003) to investigate the impact of e-learning quality dimensions on user satisfaction, use and net benefits. From their SEM analysis, they concluded that system quality positively influences user satisfaction with e-learning. Their findings were consistent with those of DeLone and McLean (2016).
According to Pang et al. (2020), there is a significant positive relationship between system quality and user satisfaction. Their study sought to explore the determinants of Knowledge Sharing Platforms (KPS) and continued usage intentions amongst Chinese consumers in Korea. The model was rooted in the DeLone and McLean (2003) and the expectation confirmation model (Bhattacherjee, 2001). The results of their study indicated that system quality in one of the key antecedents of online user satisfaction. Their claims are consistent with the results of Chang (2013), who also used the DeLone and McLean model to model the determinants of students' continuance intentions with a digital library system in Taiwan. The findings were also supported by the results of Alzahrani et al. (2019). Al-Fraihat et al. (2020) also examined the relationship between e-learning system quality and e-learning student satisfaction. Their study was operationalised at the University of Warwick, UK by adopting DeLone and McLean (2003), TAM (Marandu et al., 2019), e-learning quality model (Attwell, 2006) and the Ozkan and Koseler (2009) user satisfaction model. The study of Y. Li et al. (2021) was instrumental in developing a comprehensive e-learning model through adoption of a number of theories which were developed to examine success of information systems. Their findings reflected that a positive and significant relationship exists between technical system quality and perceived user satisfaction. Considering the foregoing discussion, it was hypothesised that; H 1 : There is a significant positive relationship between e-learning system quality and e-learning student satisfaction in the HEIs in Zimbabwe. Alzahrani et al. (2019) also claim a positive relationship between content quality factors and user's satisfaction in online learning at four universities in Malaysia. They indicated that perceived e-learning quality also depends on the information quality. Factors such as format, understandability, readability, relevancy, up datedness, and detail were important antecedents of good information quality. Their results were consistent with the findings of DeLone and McLean (2016), who also found that information quality had a positive impact on user satisfaction in digital library systems in Taiwan, China. Pang et al. (2020) also concur on this relationship between information quality and learners' satisfaction.

The relationship between e-learning information quality and e-learning student satisfaction
Furthermore, Al Mulhem (2020) supports the positive relationship between information quality and student satisfaction. Information quality increases user's satisfaction and adds up to stickiness of the e-learning site which motivates students to come again on future visits (Pham et al., 2019). The relationship between e-learning information quality and learner satisfaction has also been explored from a Saudi Arabian context. Al Mulhem (2020) used the DeLone and McLean e-learning quality model to find their influence on students' satisfaction. The model was also extended with organizational factors (change management and top management support), which were proposed to affect e-learning quality. Student satisfaction was also proposed to influence e-learning quality. The study concluded that system quality positively affected students' satisfaction with e-learning. Alotaibi (2019) also found positive impact of information quality on student satisfaction in Saudi Arabian universities. The findings corroborate prior findings by Yosep (2015), DeLone and McLean (2016), and Lee and Jeon (2020) also claim that information systems which provide high quality information resources are considered useful when they assist users to satisfy all their requirements. Their study investigated the antecedents of user satisfaction, use and net benefits to learners using a Mobile Learning Management System (MLMS) at a cyber university. Their results indicated that information quality positively affected the user's satisfaction with the MLMS. They affirmed results by DeLone and McLean (2016) and Pang et al. (2020). In the light of these claims, the study also hypothesised as that; H 2 : There is a significant positive relationship between e-learning information quality and e-learning student satisfaction in the HEIs in Zimbabwe.

The relationship between e-learning service quality and e-learning student satisfaction
The relationship between e-learning service quality and learner satisfaction has been investigated by a plethora of researchers. In Saudi Arabia, Al Mulhem (2020) validated the DeLone and McLean (2003) model by modelling the effect of e-learning quality factors on students' satisfaction. The model was also extended with organizational factors (change management and top management support), which were predicted to influence e-learning quality. The study reported that service quality positively affected students' satisfaction. These findings confirmed prior results by Yosep (2015), DeLone and McLean (2016), and Pang et al. (2020).
Similarly, the influence of e-learning service quality on student's satisfaction was investigated by Lukic and Lukic (2018). From their studies in the Western Balkans, their model sought to determine the relationship between perceived service quality, student satisfaction and future behavioural intentions in Higher Education sector. Results from SEM reflected the positively significant estimates on the causal path between perceived service quality and student's satisfaction. Their findings supported the earlier claims by Dehghan et al. (2014), from their enquiries in the Higher Education service sector of Michigan, USA. However, Pham et al. (2019) argues that their study included none Information Systems (IS) service quality factors hence their findings warrant further enquiry. Pham et al. (2019) developed and validated their e-learning success model in Vietnam. They proposed a model in which e-learning instructor and course materials quality, e-learning system quality, e-learning administrative and support service quality were predictors of e-learning service quality. Their findings confirmed the positive relationship between e-learning service quality and student satisfaction. Their study also validated e-learning service quality as a second order construct, a key finding distinct from earlier studies. Their findings are still consistent with similar studies (Al-Fraihat et al., 2020;Lukic & Lukic, 2018;Yosep, 2015) despite this difference. Al-Fraihat et al. (2020) have also proposed the positive relationship between e-learning service quality and e-learning student satisfaction. The study was conducted at the University of Warwick, United Kingdom by combining the DeLone and McLean (2003), TAM (Marandu et al., 2019), e-learning quality model (Attwell, 2006) and the Ozkan and Koseler (2009) user satisfaction model. The study conceptualized service quality as a direct antecedent of perceived satisfaction with e-learning. Using the SEM, their findings of the study indicated that a positive and significant relationship exists between e-learning service quality and perceived user satisfaction. The results concur with the findings of Lukic and Lukic (2018), Pham et al. (2019), and Pang et al. (2020). Thus, this study hypothesised that; H 3 : There is a significant positive relationship between e-learning service quality and e-learning student satisfaction in the HEIs in Zimbabwe.

The relationship between e-learning student satisfaction and e-learning student loyalty
Satisfaction is a subjective, emotional response about the experience relating to a transaction between an individual and an organization (Pham et al., 2019). The relationship between customer satisfaction and loyalty has been extensively explored in mainstream services literature. However, in e-learning contexts, the area has not attracted much enquiry. Lukic and Lukic (2018) allude to the positive link between student satisfaction and student loyalty in Higher Education. They also reiterate that student loyalty has long and short-term benefits to the universities. Loyal students bring a positive impact during the teaching process by their active participation and commitment which uplifts the learning environment for other learners. It is also reflected that loyal students recommend their university to other prospective students and they return to the university being the choice that they continue with their studies (Martinez-Arguelles & Batalla-Busquets, 2016; Pham et al., 2019).
Pinpointing the significance of student satisfaction, Cidral et al. (2020), recommended that HEIs need to implement student-centered approaches in order to sustain intense competition in the industry. Furthermore, Lee and Jeon (2020) add that universities should provide the best quality of services to drive student satisfaction. Student satisfaction has been the most cited antecedent of student loyalty (Pham et al., 2019). According to Kilburn et al. (2016), high levels of student satisfaction are likely to influence student loyalty with both the service and the service provider in the long term. Pham et al. (2019) also direct the importance of students' influence, which culminates after graduation when they spread positive word of mouth, towards a strong alum. From their studies in Vietnam, their findings reflected a direct positive and significant influence of e-learning student satisfaction and e-learning student loyalty in Vietnam. Their model received credit for conceptualizing student outcomes of e-learning with specificity (e-learning student delight and e-learning student loyalty. According to Pham and Tran (2020), high levels of e-learning student satisfaction positively influences e-learning student loyalty with both the service and the Higher Education The impact of e-learning student satisfaction has also been examined from a Southern African perspective. Mwiya et al. (2019) investigated the relationship in the Zambian public universities. The study employed the SERVQUAL model to model the service quality dimensions which influenced customer satisfaction. The study also proposed that customer satisfaction in higher education positively influences behavioural intention (loyalty and positive word of mouth). The findings reflected that customer satisfaction had a positive relationship with customer loyalty. Customer satisfaction positively influenced word of mouth among Zambian students. The results are cognizant of the similar findings by Martinez-Arguelles and Batalla-Busquets (2016) and Kilburn et al. (2016). Thus, in this study, it was hypothesised that; H 4 : There is a significant positive relationship between e-learning student satisfaction and e-learning student loyalty in the HEIs in Zimbabwe.

Population and sampling
According to Zimbabwe Council of Higher Education (ZIMCHE; 2020), Zimbabwe has twenty-four (24) fully registered and operational universities who adopted e-learning during the Covid-19 pandemic. For the purposes of this study, students who were registered at the four selected public and private universities in Harare and Bindura were chosen to provide research data. The study adopted stratified random sampling technique to select elements who participated in the study. The basis for adoption of stratified random sampling was to ensure proportionate representation of all universities in the sampling frame. Secondly, it was done to accurately represent both undergraduate and postgraduate learners from four distinct campuses. To do so, a proportionate sampling procedurewas applied for stratification purposes. In each stratum, simple random sampling was eventually applied. In line with the requirements of Confirmatory Factor Analysis (CFA), a sample of 420 students was used in the study (Hair et al., 2019;Kline, 2016). Data was collected using hand administered questionnaires to overcome low response rates which are observed in online surveys in Zimbabwe (Makudza et al., 2020).

Operationalisation
A seven-point Likert scale from strongly disagree to strongly agree was adopted in measuring e-learning system quality, e-learning information quality, e-learning service quality, e-learning student satisfaction and e-learning student loyalty consistent with previous studies (Al-Fraihat et al., 2020;Alzahrani et al., 2019). The study adopted the performance only measurement (Cronin & Taylor, 1992) to assess perceived e-learning service quality and its outcomes, in line with supporting literature. Table 1 shows the sources of measurment scales used in this study.

Ethical compliance
The study was compliant to research ethics as governed by the Zimbabwe Ezekiel Guti University ethics requirements. Participants were informed of the goal, benefits and intended outcome of the study. Voluntary participation of respondents was upheld and respondents were informed of their freedom to withdraw from the research at any point, should they wish to do so. Anonymity of respondents was maintained throughout the study. Information obtained from participants was confidentially kept and used only for academic reasons which were disclosed to all participants of the study.

Sample characterisation
From the 420 distributed questionnaires, 354 were returned giving rise to a response rate of 84.28%. However, 321 valid responses were retained for conclusive analysis. The demographic characterization showed that 61.7% were females and 38.3% were males. Furthermore, there were more participants from Universities A (33.3%) and B (35.5%) than there were from Universities C (15.3%) and D (15.9%). The study also indicated that young participants aged from 17-21 years and 22-26 years dominated the sample, with high incidences of 27.4%, 32.4% respectively.

Confirmatory Factor Analysis (CFA)
A two-step data analysis procedure was employed in this study. The data was subjected to Confirmatory Factor Analysis (CFA) and structural equation modelling (SEM; Anderson & Gerbing, 1988). A maximum likelihood estimation using the variance-covariance matrix was used to estimate the models in this study (Byrne, 2013). The measurement model was examined on a number of criterions.
First, the standardized factor loadings were inspected for unidimensionality. Hair et al. (2016) recommended that items with weak standardized loadings (less than 0.5) should be deleted. Byrne (2013) also advises that loadings <0.3 are not significant, those >0.4 are more important, and factor loadings greater than 0.5 are significant, and in CFA, loadings >0.7 are considered very significant. In CFA, loadings equal or greater than 0.7 confirm the validity of the measurement model in explaining the latent variables (J. F. Hair et al., 2010;Holmes-Smith, 2001). The lowest standardized loading in the CFA model was on ELS3 (0.70) and the highest loading was 0.9 on ESVQ3 and ESVQ4 (Table 2).
Secondly, the Average Variance Explained (AVEs) were also assessed for convergent validity. According to Hair et al. (2016), these explain the proportion of variance shared by indicators on a latent factor and the threshold for AVE is 0.5, which means the items should at least share 50% of the total variance of a construct. It is shown in Table 2 that the construct with the lowest AVE was E-learning student loyalty (0.53) and the highest was E-learning service quality (0.77), therefore the requirement of convergent validity was satisfied.

Construct
Source Scale Discriminant validity was also checked using the Heterotrait-Monotrait (HTMT) ratio as it has been recommended over the Fornell-Larcker and cross loadings criterion. According to Henseler et al. (2015), the two methods should be verified by the HTMT method, a recent method of testing discriminant validity using the ratio of the mean of correlations of unrelated items relative to the means of correlations of items measuring related constructs. Kline (2016) recommends HTMT values below 0.85 (HTMT.85) whilst Gold et al. (2001) suggests that those below 0.90 (HTMT.90) as still reflective of discriminant validity. HTMT.85 was used in this study and all scores in Table 3 confirmed that there were no discriminant validity problems. Lastly, the model fit for the measurement models was examined to check the discrepancy between the variance-covariance structure of the hypothesised models and the variancecovariance matrix of the sample data (Byrne, 2013;Hair et al., 2016). Two groups of model fit criterion namely the Absolute fit and Incremental fit indices were used in this study. The CMIN (Chi-square), degrees of freedom, normed chi-square, Goodness of Fit Index (GFI), Standardised  Root Mean Residual (SRMR), Root Mean Square Error of Approximation (RMSEA) belong to the absolute fit indices because they are not compared to any model at all. The Comparative Fit Index (CFI), Incremental Fit Index (IFI) and the Tucker Lewis Index (TLI) were the incremental fit indices used to examine model fit in this study. The modification indices and residuals matrices were inspected and showed no significant problems to deny an acceptable model fit. By the recommendations of Byrne (2013), the CFA model reflected a good fit as shown in Table 4. Figure 2 presents the measurment model.

Assumptions of structural equation modelling (SEM)
The structural model examines the relationships between latent variables. However, J. F. Hair et al. (2010) recommend the satisfaction of certain assumptions prior to model estimation. Univariate normality was tested using the skewness and kurtosis values. In line with Kline (2013), the range of skewness and kurtosis data ranged from −2 to +2. Furthermore, multivariate normality was tested using the Mardia's criterion of p(p + 2), which should be less than the multivariate value estimated in SPSS AMOS, where p = number of observed variables (Raycov & Malcoulides, 2008). In this study, 19 variables were used for final analysis, hence 19(19 + 2) = 399. This was greater than the models' multivariate estimate (146.18).
Multicollinearity was also examined. In this study, correlations between both predictor and criterion variables ranged between 0.494 and 0.746. To verify that, Tolerance and the Variance Inflation Factor (VIF) were examined. The VIF explains how much larger the standard error would be relative to if the predictors were uncorrelated. According to Hair et al. (2016), the VIF should be less than 5 and Tolerance should be greater than 0.2. The VIF range between 1.4 and 2.4 whilst Tolerance ranged from 0.406 and 0.691.

Structural equation modelling
The structural model was assessed by evaluating the model fit (absolute and incremental fit), significance of path estimates and the predictive power of the model (R square). These provided the basis on which the hypothesised relationships were tested.
The structural model used both absolute and incremental fit indices for assessment. The Chi-Square (CMIN) is one of the main absolute fit indices used in SEM. A significant CMIN normally means that there is a significant discrepancy between the covariance matrix of the hypothesised model and the covariance matrix of the sample data (Hair et al., 2019). However, as sample size increases the chi-square usually leads to rejection of a good fitting model, thus the chi-square normalized by degrees of freedom was used (χ2/df). Other absolute fit indices used in this study were the Goodness of Fit (GFI), Root Mean Square Error of Approximation (RMSEA), and the Standardised Root Mean Residual (SRMR). The IFI, TLI and CFI were the incremental fit criterion  used to assess the structural model. Table 5 shows that the structural modelattained an acceptable fit.
Secondly, the relevance and significance of structural relationships was assessed using the path coefficients. There was support that the structural model was grounded from theory as there were significant estimates for the paths as hypothesised. All paths had positive estimates and t-values that were greater than 1.96 were judged to be significant from zero (at 95% confidence interval) and p values were less than 0.05. Figure 3 shows the path estimates for the structural model.
Thirdly, the structural model was evaluated using its predictive power (R Square). This refers to the magnitude of the variance which is explained by a model (Hair et al., 2016). The explanatory power of the model that was proposed in this study was 82% and 60% for e-learning student satisfaction and e-learning student loyalty respectively, as illustrated in Figure 3.

Results and discussion
The purpose of the study was to determine the impact of perceived e-learning service quality on students' outcomes; hence it was inevitable to test the hypotheses proposed in the study. Table 6 summarises the research hypotheses and their outcomes.
H 1 : There is significant positive relationship between e-learning system quality and e-learning student satisfaction in the HEIs in Zimbabwe. Table 6 above, the resulting path estimate of 0.40 confirmed a positive relationship drawn between e-learning system quality and e-learning student satisfaction. The relationship was statistically significant as evidenced by a t-value of 10.276 which is greater than 1.96 (at 95% confidence interval) and a p-value with three asterisks (***) confirms that p was less than 0.001. As a result, H 1 was supported.

As indicated in
These results are not a new phenomenon in e-learning literature across research contexts. Pham et al. (2019), Al-Fraihat et al. (2020), and Lee and Jeon (2020) also found the relationship quite significant from their studies in Vietnam, United Kingdom (UK) and Korea respectively. This Notes: The SEM diagram shows standardized estimates. H 2 : There is a significant positive relationship between e-learning information quality and e-learning student satisfaction in the HEIs in Zimbabwe.
As presented in Table 6 above, this hypothesis was also tested. The resulting path estimate of 0.51 confirmed a positive relationship drawn between e-learning information quality and e-learning student satisfaction. The relationship was statistically significant as evidenced by a t-value of 12.179, which is greater than 1.96 (at 95% confidence interval) and a p-value with three asterisks (***) indicated that p was less than 0.001. Resultantly, the hypothesis was accepted.
The study confirmed the importance of e-learning information quality in determining student satisfaction with e-learning. The study corroborates that information quality is a significant predictor of e-learning student satisfaction, in line with findings reported by Alzahrani et al. (2019) The results of this study indicate that students prioritise the availability of information that is in readable format, timely, accessible, understandable and up-to-date.
H 3 : There is a significant positive relationship between e-learning service quality and e-learning student satisfaction in the HEIs in Zimbabwe.
The above hypothesis was also proposed in this study. The resulting path estimate of 0.12 confirmed a positive relationship between e-learning service quality and e-learning student satisfaction. The relationship was also statistically significant as confirmed by a t-value (5.601), which is greater than 1.96 (at 95% confidence interval) and a p-value with three asterisks (***) ratified that p was less than 0.001. These findings influenced the support for the hypothesis.
These findings corroborate with earlier results by McLean (2003, 2016), Cheng (2020), Pham et al. (2019), and Umukoro and Tiamiyu (2017) as well as Alzahrani et al. (2019). These researchers found a positive relationship between support service quality and user satisfaction in their respective enquiries. Therefore, H 3 was accepted and the study concluded that there is a positive relationship between e-learning service quality and e-learning student satisfaction. It is important to note that students value the assistance they receive especially when troubleshooting on the e-learning portals.
H 4 : There is a significant positive relationship between e-learning student satisfaction and e-learning student loyalty in the HEIs in Zimbabwe.
As presented in Table 6, the resulting path estimate of 0.78 confirmed a positive relationship between drawn between e-learning student satisfaction and e-learning student loyalty. The relationship was statistically significant as reflected by a t-value of 8.894, which is greater than 1.96 (at 95% confidence interval) and a p-value with three asterisks (***) confirmed that the p-value was less than 0.001. As a result, H 4 was accepted.
between customer satisfaction and customer loyalty, very few researchers have dwelled on the influence of student satisfaction on their loyalty in a higher education online context. These results add empirical support from a Zimbabwean context.

Conclusions
In light of the above findings, the study concludes that the e-learning quality dimensions (system quality, information quality and service quality) all have a positive relationship with student satisfaction with e-learning. The study also concludes that student satisfaction with e-learning cultivates long term student loyalty with both e-learning use and university commitment.

Recommendations
Given the conclusions, the study recommends that HEIs should view e-learning system quality as a critical dimension of e-learning quality. HEIs should focus on designing e-learning systems so that they produce the best access, availability, integration, navigation, ease of downloading material. Administrators should always upgrade their e-learning technologies to provide a seamless and quick service to students without letting students endure torrid sessions.
Furthermore, HEIs should also constantly upgrade the quality of information that is downloaded from e-learning sites. Timeliness, understandability, availability, currency, relevance and adequacy are all important dimensions of information quality that Faculty Administrators should always evaluate. More so, the HEIs should also check the format, and form (text, audio and video) that is provided to learners.
The study further recommends that HEIs may consider improving knowledge and motivation to IT professionals who assist learners online. Most non-marketing professionals' understanding of the importance of service quality is far-fetched, thus HEIs should offer all support service personnel marketing courses to equip them with empathy, courtesy, responsiveness and kindness needed when serving learners (Y. Li et al., 2021). HEIs may also consider upskilling the IT staff especially in these times where technological developments are ubiquitous. This equips the staff with necessary tools to troubleshoot learners' challenges with e-learning.
The study finally recommends that HEIs in Zimbabwe should envisage building long-term student loyalty through offering best e-learning quality across all dimensions discussed. They thus should consider a holistic service system, in both online and offline contexts, to revamp learners' satisfaction and build long-term loyalty which impacts on future behavioural intentions. Satisfied learners are more active in current learning environment as well as supporting the HEI as alumni when they complete their studies. Thus e-learning student loyalty benefits the HEIs in the short and long-term.

Limitations and future study
The study adopted the Updated DeLone and McLean (2003) and the Expectations-Confirmation Model (ECM) to develop the research model. Future researchers can employ other technology adoption theoretical frameworks such as the Technology Adoption Model (TAM), Unified Theory of Acceptance and Use of Technology (UTAUT) Venkatesh et al. (2003) or combining frameworks to improve the predictive power of technology success models in a higher education setting. Secondly, the study employed quantitative methods to examine relationships. However, future researchers might dwell on qualitative methods using the born in other paradigms. Future enquiries can adopt in-depth interviews and student focus group sessions to gain richer perceptions from the participants. Ultimately, student perceptions were used to generalise on the success of e-learning systems in Zimbabwean universities. However, for better inclusion, future researchers need to enquire from other important e-learning stakeholders. This includes instructors, facilitators, IT staff as well as faculty administrators across the twenty-four (24) universities in Zimbabwe.
Although the study successfully modelled perceived e-learning service quality, student satisfaction and loyalty; the study also suffered some limitations. The model presented and tested in this study has never been empirically tested before. That presented the limitation on model integrity in various situations and target groups. To minimise the negativity of the limitation, the study employed reliability and validity tests, which proved that the model was sound. COVID-19 restrictions also caused mobility limitations during the research. However, the threat was minimal as adherence to COVID-19 preventative guidelines helped the study to gather sufficient responses to validate the model.