Customer satisfaction as mediator in the relationship between higher education service quality and student customer loyalty Open Access

The delivery of superior service quality to students is paramount for universities worldwide (Mattah et al., 2018; Phonthanukitithaworn et al., 2022). With students increasingly viewed as customers (Ashwin et al., 2023; George, 2007; Gupta et al., 2023), the need to understand student experiences has garnered significant attention from researchers and practitioners (Oliveira et al., 2023; Singh et al., 2021). This is because improved experiences could lead to student retention/loyalty (Annamdevula and Bellamkonda, 2016) and hence the achievement of sustainable development through universities (Lăzăroiu, 2017).

Central to this discourse is the seminal work of Parasuraman et al. (1985, 1988) which conceptualized service quality as a multidimensional construct of several variables including; tangibility, reliability, responsiveness, assurance and empathy. This is famously known as the service quality (SERVQUAL) model with each of the variables being referred to as dimensions of service quality.

While numerous studies (e.g. (Annamdevula and Bellamkonda, 2016; Bwachele et al., 2023),) investigate the impact of these individual SERVQUAL dimensions on student outcomes, a notable methodological gap persists. Most studies take the five SERVQUAL model dimensions in Parasuraman et al. (1988) as proxies for service quality. For instance, Ananda and Devesh (2019) and Pakurár et al. (2019) treat each of the service quality dimensions as an independent variable. Our work is novel because it argues that this ignores potential multicollinearity in the model and increases the likelihood of biased estimates.

To solve this, the five dimensions must be considered as lower order constructs (LOCs) which collectively measure service quality as a higher order construct (HOC). We advocate for the treatment of each of the service quality dimensions as first order constructs and the actual service quality as a second order construct. This helps reduce the multicollinearity problem. Hence, our study fills a methodological gap and addresses a theoretical issue in the service quality model.

A HOC is a composite construct formed by combining two or more LOCs (Crocetta et al., 2021). It helps researchers to conceptualize service quality as a multidimensional yet unified construct (Falk et al., 2010; Vatolkina et al., 2020). This hierarchical model mitigates multicollinearity by accounting for interdependencies among the service quality dimensions while preserving their individual contributions (Hair et al., 2024).

According to Tsagris and Pandis (2021) multicollinearity occurs when independent variables are highly correlated. It either inflates or deflates the standard errors of the coefficients. The coefficients might falsely become nonsignificant or significant. Multicollinearity can also change the sign of the coefficient so that a positive effect becomes negative and vice versa (Johnston et al., 2018). Overall, multicollinearity leads to unreliable regression results (Vatcheva et al., 2016; Voss, 2005). Hence, better specification of a HOC and LOCs becomes useful at resolving this challenge.

The relationship between LOCs and the HOCs can be either reflective or formative (Bollen and Lennox, 1991; Edwards and Bagozzi, 2000; Jarvis et al., 2003). It is reflective when the HOC represents the effects of the LOCs or formative when the LOCs collectively define the HOC. The HOCs can exist at multiple levels of abstraction (Crocetta et al., 2021). For example; second order, third order and so on. Various scholars address the definition of first-, second- and higher-order constructs (for example Edwards (2001) and Law et al., (1998)).

To begin with, researchers study abstract ideas called constructs–like service quality. These are measured through more specific concepts called first-order constructs. Taking the SERVQUAL model as an example, the first-order constructs could include; tangibles, reliability, responsiveness, assurance and empathy. When these first-order constructs are combined under a broader concept, that becomes a second-order construct. In this case, perceived service quality encompasses all the five SERVQUAL dimensions. This hierarchy can continue to a third-order construct formed by combining second-order constructs. For example, combining perceived service quality with student engagement with learning materials (another second-order construct) could form a broader third-order construct like comprehensive student experience. By employing this hierarchical approach which many studies ignore, our study makes a theoretical and methodological contribution.

Our study measures service quality as a second order construct and the SERVQUAL dimensions as first order constructs. This is unlike most extant literature which treat the dimensions of service quality as independent variables. This methodological gap is consequential, because treating each dimension of service quality as an independent variable can lead to multicollinearity, thereby potentially confounding the interpretation of results. Hence, our study seeks to address this gap by emphasizing the novelty of a framework that conceptualizes service quality as a higher-order construct, integrating the dimensions delineated by Parasuraman et al. (1985).

Specifically, a reflective-reflective higher order construct model (Sarstedt et al., 2019) is used where the LOCs (that is, the service quality dimensions) are postulated to be highly correlated and the HOC (service quality in this case) represents the cause explaining the correlations in line with a type I higher order construct (Hair et al., 2024).

We then assess the mediation role of customer satisfaction in the relationship between higher education service quality (measured as a second order construct) and customer loyalty. The research questions are as follows:

This significance of this study is in its theoretical, methodological and practical contributions for university administrators and policymakers. It offers insights for institutions seeking to optimize service delivery, enhance student satisfaction and foster retention and loyalty.

The next section reviews the relevant literature. Section 3 outlines the methodology while Section 4 presents the empirical results. This is followed by a discussion of their implications in Section 5. Section 6 concludes the study by summarizing the main findings. Section 7 highlights policy recommendations and suggests avenues for future research.

This study is underpinned by the service quality (SERVQUAL) model proposed by Parasuraman et al. (1985). We propose an extension of this model by treating service quality as a higher-order construct. This approach is becoming common in service quality studies in other sectors (e.g. (De Leon et al., 2020; Narayan et al., 2008; Nunkoo et al., 2017; Untachai, 2013)). It is guided by the theoretical assumption that the dimensions of service quality in Parasuraman et al. (1988) are predictors or indicators of an underlying, overarching service quality construct.

Another recent theoretical perspective is the higher education service quality (HESQUAL) model which according to Schijns (2021) was developed by Teeroovengadum et al. (2016). This model measures service quality as a second order construct. While useful in addressing multicollinearity, it is too specific to the higher education sector. A generalized model like the original SERVQUAL model provides broader application to various sectors while also being useful to higher education.

Myriads of studies exist on service quality–customer satisfaction and/or loyalty/retention nexus (such as (Amin, 2016; Kaura et al., 2015; Salamah et al., 2022)). However, few (such as (Fernandes and Solimun, 2018; Slack and Singh, 2020)) assess the mediation role of customer satisfaction. Even fewer treat the dimensions of service quality as LOCs and service quality as a HOC.

Prentice et al. (2020) assess the effects of artificial intelligence (AI) and service quality on customer satisfaction as well as loyalty in Portuguese hotels. Both AI and service quality predicted customer satisfaction and loyalty. While the study was not in higher education, it provides insights into how service quality influences customer loyalty and satisfaction. The use of AI in education has potential to influence not just education service quality but humanity as a whole (Peters et al., 2024). However, Prentice and others studied not only a different sector but also a different country with better social economic status than Zambia.

Nguyen et al. (2024) assesses the influence of service quality on student satisfaction and loyalty in Vietnam. They use programming, academics industry interaction and facilities as measures of university service quality. However, these measures deviate from the SERVQUAL model measures and make comparability of results with other studies difficulty.

Borishade et al. (2021) assess the relationships among service quality, student satisfaction and student loyalty among private university students in Nigeria. Service quality in the study is measured as a higher order construct and had a relationship with student customer loyalty. Student satisfaction was also found to be a mediator in this relationship. While this study considered service quality as a higher order construct, it left out public university students.

Pham et al. (2018) study e-learning services quality, loyalty and satisfaction among online learning students in the USA. A positive correlation was found between service quality and loyalty as well as between satisfaction and e-learning loyalty. Hence, we hypothesize the following:

Islam et al. (2021) study the relationships among dimensions of service quality, customer satisfaction and loyalty among private sector bank customers in Bangladesh. Structural equation modelling revealed that the dimensions of service quality had a positive influence on customer satisfaction. Customer satisfaction and customer loyalty also had a positive relationship. Although not in the higher education sector, the study provides some insight into the potential influence of university service quality and student customer satisfaction on customer loyalty.

Pham et al. (2019) examine the relationships among service quality, student satisfaction and student loyalty in the e-learning in Vietnam and recommend that service quality should be a second order construct. The study finds that service quality has a positive influence on student satisfaction with e-learning which also showed a positive influence on student loyalty. Service quality also had a positive effect on student loyalty. This was also supported by Darawong and Sandmaung (2019) who found that various dimensions of service quality have a positive influence on student customer satisfaction. However, both studies were in the context of Asia – leaving out perspectives from Africa. Further, the first study focused on e-learning and therefore left out traditional learning service quality.

Khoo et al. (2017) study the relationship between service quality and student customer satisfaction with non-academic services in Singapore. Data were drawn from a survey of 324 responses from students of two private universities. A positive correlation was found between perceived service quality and perceived student satisfaction. This study however left out the context of public universities and considered only non-academic services.

Ali et al. (2016) study the effect of service quality on satisfaction and loyalty among international students in Malaysia. Structural equation modeling was used on data from 241 students in public universities. It is found that service quality influences student satisfaction which in turn influences student loyalty. This study however used different measures of service quality from the ones proposed in (Parasuraman et al., 1985).

Given the studies reviewed, it is hypothesized that:

Teeroovengadum et al. (2019) validate the higher education service quality (HESQUAL) scale and test relationships among student loyalty, service quality, image, perceived value and satisfaction. A sample of 501 responses is collected from students in Mauritius. It was found that service quality influences student satisfaction. Most importantly, the authors conclude that measuring service quality as a higher-order model is worthwhile. However, this was not tested in the context of Zambia.

Fernandes and Solimun (2018) investigate whether or not customer satisfaction mediates the relationship between three variables (service orientation, the service quality and marketing mix strategy) and customer loyalty. They conclude that indeed, customer satisfaction is a mediator. However, this study was not in the education sector. This is similar to the findings of Anabila et al. (2022). However, this study too was not in the context of higher education and considered slightly different variables – customer delight, hotel service quality and hotel customer loyalty.

Moslehpour et al. (2020) used data from 197 international students in Taiwanese’ higher education institutions. The study assesses the mediation role of student satisfaction in the service quality – university reputation nexus. The existence of such a mediation role was confirmed. However, the study was not in the context of Zambia. Hence the following hypothesis is proposed:

A quantitative design is employed (Creswell, 2012; Saunders et al., 2019) in this study. Specifically, mediation analysis (Hayes, 2018) was used to investigate the mediating role of customer satisfaction in the nexus between service quality and student customer loyalty.

A systematic random sampling technique was used to recruit students enrolled in two universities (one private and one public) in Zambia with every kth person targeted (Arrogante, 2022). Random sampling was adopted from existing studies such as Rajab et al. (2011) and Mwiya et al. (2022). The first reason for choosing the location of the study was accessibility of the respondents to the researchers. The second was because of the funding constraints that universities in this area face amidst escalating demand for tertiary education (Moshtari and Safarpour, 2024). This situation is reflective of challenges faced in diverse geographical locations of the world (Hazelkorn et al., 2022; Moshtari and Safarpour, 2024). Overall, this approach allowed us to target individuals relevant to the research questions.

A total of 340 responses were obtained using a self-administered online questionnaire shared with respondents’ onsite via a link. Here, the data collectors stood at a point and every 5th person who passed by was asked to fill in the questionnaire. This is how systematic random sampling was assured. The sample size was determined with reference to existing literature (Hair et al., 2021; Kaulu, 2022; Wolf et al., 2013) which suggest that even a sample size of 200 is sufficient for structural equation modelling.

We chose higher education service quality because education is the bedrock of all industries and economic development (Lauder, 2020). Without quality education, nations would lack effective graduates (Hanushek, 2020; Masaiti, 2013). The sample for this study is from Zambia. In Zambia, universities grapple with funding constraints amidst escalating demand for tertiary education. For instance, in 2022, universities in Zambia enrolled 156,044 students – this marks an 11.1% increase from 2021 (Higher Education Authority, 2022). Despite growing enrollments, funding has not kept pace (Masaiti et al., 2024). This exacerbates already existing challenges such as poor lecturer–student ratios, limited digital learning and physical infrastructure and brain drain as lecturers seek employment with well-funded institutions globally (Nyashanu et al., 2023). Most of these challenges mirror those faced by higher education institutions globally (Hazelkorn et al., 2022), where balancing financial sustainability with quality education delivery is paramount (Jacob and Gokbel, 2018). Hence, insights from Zambia’s experience can contribute significantly to international literature on enhancing service quality and fostering student loyalty in resource-constrained settings.

The research instrument was adapted from Mwiya et al. (2017). Adapting entails taking existing questions from previous studies non-verbatim and making slight changes to fit the needs of a current study (Sousa et al., 2017). This study adapted questions in order to ensure the use of well-established and validated scales specifically designed for the education sector to measure student perceptions. The instrument consisted of a battery of questions on a five-point Likert scale as shown in Appendix D.

The collected data was first cleaned for potential missing values, outliers and inconsistencies. Descriptive statistical tests were then done. This was followed by a two-stage measurement and structural model assessment approach as per extant literature (Lehmann and Gupta, 1989).

The measurement model was assessed using confirmatory factor analysis (CFA) in AMOS. Stage one was a model containing the service quality constructs only (the independent variables) (Appendix A). The factor loadings for EMP1, EMP2 and TAN 3 were below 0.6 and hence these were deleted. Stage two involved assessment of the model including the dependent and mediator variables (See Appendices B and C.)

The model fit was tested using the chi-square/degrees of freedom (CMIN/DF) ratio, comparative fit index (CFI), standardized root mean square residual (SRMR) and root mean square error of approximation (RMSEA). The discriminant validity of the constructs was measured using the Fornell–Larcker criterion. The convergent validity of the constructs was tested using average variance extracted (AVE). Construct reliability was established using composite reliability (CR) scores. The various validity and reliability tests were adopted from a variety of literature (for example (Hair et al., 2020; Hu and Bentler, 1999; McNeish and Wolf, 2023)).

Structural equation modeling (SEM) was employed at the structural model assessment stage using SPSS AMOS Version 23. The SEM allowed for the simultaneous estimation of multiple relationships (direct and indirect effects) within a single model (Beran and Violato, 2010; Fan et al., 2016; Gunzler et al., 2013; Tarka, 2018) with the help of bootstrapping. This technique allowed us to test the hypotheses.

Both males and females were represented in the sample at 48.8 and 51.2% cent respectively. Most respondents (92.6%) were below the age of 31. A majority (81.2%) came from business schools – which reflects the dominance of business studies in the universities from which data was collected. Similarly, 81.8% of the respondents were in full time mode of study – reflecting the dominance of this mode of study in the universities. A majority of the respondents (73.5%) reported being at a private university while 26.5% reported being at a public university. These results are shown in Table 1.

We began with a model containing the service quality constructs only (the independent variables). In accordance with model fit benchmarks recommended by Hu and Bentler (1999), this model’s CMIN/DF was excellent (2.635), the CFI was acceptable (0.921), the SRMR was excellent (0.052) and the RMSEA was acceptable (0.069). The model validity tests showed an AVE of 0.837 and a CR of 0.962. Hence, no validity concerns existed at this stage.

The assessment was then moved to stage two where CFA was conducted on a model containing student customer loyalty (the dependent variable) and customer satisfaction (the mediator) plus the stage one CFA service quality dimensions (independent variables).

Stage two model assessment on the full model showed that factor loadings were all above 0.6. In accordance with model fit benchmarks recommended by Hu and Bentler (1999), the model CMIN/DF was excellent (2.425), the CFI was acceptable (0.931), the SRMR was excellent (0.052) and the RMSEA was acceptable (0.065).

The CR scores of the constructs were all above the 0.7 threshold (Hair et al., 2020) and hence construct reliability was established.

As per Table 2, Convergent validity was established by the AVE whose values were all above the 0.5 threshold (Hair et al., 2020). The AVE is a measure of the amount of variance that is captured by the latent construct relative to the amount of variance due to measurement error. Discriminant validity was established using the Fornell–Larcker criterion (Fornell and Larcker, 1981). This requires that each construct’s square root of the AVE must be greater than the correlations between that construct and all other constructs in the model. In Table 2, the square roots of AVE are shown in the diagonal and satisfy this requirement. Given these results, the model was then used in structural model assessment.

The study assesses the mediating role of student customer satisfaction (SAT) in the relationship between universities’ service quality (SERVQUAL) and student customer loyalty (INT). Structural model assessment was done via mediation analysis through bootstrapping in AMOS. The number of bootstrap samples was set at 5,000 and the bias corrected confidence level was set at 95.

The indirect effect of service quality (SERVQUAL) on student customer loyalty (INT) through student customer satisfaction (SAT) was significant (β = 0.803, p = 0.000, CI = 0.609 to1.032). This suggests that student customer satisfaction mediates the relationship between university service quality and student customer loyalty and provides support for H3. The direct effect of SERVQUAL on SAT was significant (β = 1.087, p = 0.001, CI = 0.927 to 1.253). This suggests that university service quality positively affects student satisfaction. Hence H2 was supported. The direct effect of SERVQUAL on IN was insignificant (β = 0.166, p = 0.154, CI = −0.73 to 0.402). This suggests that university service quality does not have a direct effect on student customer loyalty. Hence H1 was not supported. It also suggests that customer satisfaction fully mediates the relationship between service quality and customer loyalty. Table 3 summarizes the mediation analysis results.

The study assesses the mediating role of student customer satisfaction (SAT) in the relationship between universities’ service quality (SERVQUAL) as a second order construct and student customer loyalty (INT).

The results show that customer satisfaction fully mediates the relationship between service quality (as a second order construct) and student customer loyalty. Hence, H3 was supported. This is in line with Moslehpour et al. (2020). However, most existing studies did not measure service quality as a second order construct. Hence, more studies are needed to assess how service quality as a second order construct would influence customer loyalty via customer satisfaction.

This study also found that university service quality positively affects student satisfaction. Therefore, H2 was supported. This is similar to the finding of Ali et al. (2016), Khoo et al. (2017), Pham et al. (2019) and Darawong and Sandmaung (2019). It also supports the applicability of the SERVQUAL model in the higher education context. This result also provides a broader understanding of student behavior, student decision-making and the usefulness of service experiences in shaping satisfaction.

The current study however found that university service quality does not have a direct effect on student customer loyalty. Hence H1 was not supported. While this is in contrast with the findings of several studies such as Borishade et al. (2021) and Pham et al. (2018), it makes theoretical sense because in the absence of customer satisfaction, it is difficult to have customer loyalty. However, the above studies do not arrive at the same conclusion likely because of the methodological gap established by this study.

While this study used Parasuraman et al. (1985), traditional five service quality dimensions, it can also be understood within Zambia’s evolving higher education context. Digitalization and post-COVID remote learning have reshaped student perceptions (Nyashanu et al., 2023; Sumi and Kabir, 2021), influencing dimensions like responsiveness (e.g. virtual communication, online systems) and assurance (e.g. remote teaching competence). Cultural factors – such as authority, communal values and communication norms – also affect how students judge empathy and reliability. Though not explicitly modeled, these influences are evident in literature, highlighting the need to either adapt the SERVQUAL model to reflect the digital and cultural dynamics of higher education in Zambia in future studies or interpret results in that light.

This study has assessed the mediating role of student customer satisfaction (SAT) in the relationship between universities’ service quality (SERVQUAL) and student customer loyalty (INT). Service quality as a second order construct influenced customer satisfaction but had no influence on customer loyalty. However, service quality had an indirect effect on customer loyalty through customer satisfaction – implying that customer satisfaction fully mediates the relationship between university service quality and student customer loyalty. Several recommendations are drawn from this.

Several practical implications arise from this study. First, because the study found that service quality influences student loyalty only through satisfaction, Zambian university administrators must recognize satisfaction as the strategic driver of loyalty. Second, while the study effectively draws on SERVQUAL and the Higher Education Service Quality model, it highlights their continued relevance in a developing country context. The contextual insights generated offer localized value even in the absence of broader consumer behavior theories.

In terms of recommendations, future research should continue to model service quality as a second-order construct to address potential multicollinearity. Universities in Zambia should focus on improving student satisfaction, as it is the key driver of loyalty. This includes simplifying administrative processes, reducing delays in service delivery and ensuring that support services are accessible, friendly and responsive to students’ needs.

Given the increasing role of digital platforms in higher education, institutions should invest in staff training for digital service delivery such as prompt SMS, email and other essential communication with students. To improve service quality, institutions should embed student feedback mechanisms in operations–such as surveys, complaint tracking and suggestion systems. These tools can help identify areas for improvement and ensure that student concerns are addressed systematically.

Finally, future studies should consider incorporating emerging service quality dimensions such as digital access, inclusiveness and cultural sensitivity. These context-specific factors may provide deeper insights into student expectations and help tailor service models to the realities of Zambia’s higher education environment.

The supplementary material for this article can be found online.