Employee retention is a multifaceted phenomenon that has been defined in numerous ways by scholars (Aburub 2020). Sengputa (2023) describes employee retention as the length of time employers are able to retain their employees. In contrast, Chong et al. (2021) view employee retention as the commitment of employees to their current employer. For this study, employee retention is a strategic initiative by organisations to retain employees for extended periods, thereby minimising the costs associated with recruiting, training and preparing new employees. This definition integrates the perspectives of Maphanga and Olivier (2025) and Pirrolas and Correia (2022), along with Schaap and Olckers (2020), and provides a practical and theoretical framework for analysing retention in the context of administrative staff. Khan et al. (2019) emphasise that employee retention is the lifeblood of organisational success.

Mzini (2019) describes employee retention within TVET colleges as the institution’s ability to retain experienced staff and reduce turnover caused by resignations, retirements or career changes. Similarly, De Koker, Schutte and Barkhuizen (2022) characterise employee retention as the capacity of TVET colleges to retain skilled employees in a competitive educational environment. Mabaso and Moloi (2016) highlight the critical role of employee retention in the institutional performance of TVET colleges, while (Netswera & Rankhumise 2005) regard it as a tool for attracting and retaining highly skilled employees, essential for providing quality support to both management and staff. Consequently, colleges must develop effective retention strategies to secure and maintain a talented workforce (Musakuro & De Klerk 2021).

Employees’ decisions to remain with or leave an organisation are shaped by a range of factors, often evaluated through a cost–benefit lens (Kaliyamoorthy et al. 2018). Similarly, Ghani et al. (2022) emphasise the multitude of drivers affecting these choices, positioning employee retention as a strategic imperative for organisations seeking to maintain their competitiveness.

According to Bhebhe and Maphosa (2016), retention-related factors significantly influence employees’ intentions to resign. This trend may be linked to the increasing availability of alternative employment opportunities, which can prompt employees to leave their current positions when faced with any form of dissatisfaction.

Sanchez (2023) highlights the importance of understanding the factors that influence employee retention, noting their significant impact on an organisation’s ability to attract and retain top talent. Akala (2012) contends that numerous retention factors exert a positive influence on employees within an organisation. Sinha and Sinha (2012) argue that effective retention strategies must extend beyond traditional benefits and address the diverse needs and aspirations of employees across different generations, experience levels and educational backgrounds.

Consequently, it is vital for organisations to gain a thorough understanding of the variables that underpin employee retention and to develop effective retention strategies accordingly. The present study identified five key retention factors: job satisfaction, performance appraisal, job security, career development and rewards and recognition, as being particularly influential within organisations. These factors are highlighted in the literature as crucial for retention. Job satisfaction boosts organisational commitment (Schaap & Olckers 2020), while performance appraisal, along with rewards and recognition, motivates employees and reinforces retention (Hatidja et al. 2024). Job security enhances loyalty and reduces turnover intentions (Ngui & Katua 2023), and career development opportunities increase engagement and organisational commitment (Schaap & Olckers 2020). These key retention factors were chosen for their empirical backing and practical significance in retaining administrative staff in organisations. Although widely studied in academic contexts, retention factors remain underexplored among administrative staff at South African TVET colleges, addressing a key knowledge gap with theoretical and practical contributions.

To conceptualise the theoretical foundation to explain the factors influencing employee retention, Herzberg’s two-factor theory was adopted. This theory is particularly well suited to the context of employee retention, as it addresses the roles of both job satisfaction and dissatisfaction.

Herzberg’s two-factor theory was developed in 1959 by the American psychologist Frederick Herzberg to explain factors influencing employee motivation and retention. Giese and Avoseh (2018) cited Herzberg’s theory of motivation, which distinguishes between factors that enhance or diminish employee satisfaction, namely, hygiene factors and motivators. Herzberg’s theory serves as the central theoretical framework (Hur 2018). According to Mitsakis and Galanakis (2022), motivators are intrinsic factors that drive employees to perform at their best, whereas hygiene factors, such as the need to avoid unpleasant circumstances, are extrinsic. Similarly, Yousaf (2020) regards motivator factors as essential for boosting employee performance, productivity and overall job satisfaction; their absence, however, can decrease job satisfaction (Abdulhamidova 2021).

Retention studies show that motivators such as achievement, recognition and growth increase job satisfaction and commitment, reducing turnover (Shinde 2025). Career development and rewards strengthen retention (El Mountasser & Sahraoui 2025), while hygiene factors like job security and fair performance appraisals reduce dissatisfaction (Abdulhamidova 2021). A study by Thant and Chang (2021) found that insufficient hygiene factors, such as company policy, supervision, salary and working conditions, lead to dissatisfaction, even though they do not actively motivate employees. Conversely, the presence of intrinsic factors such as recognition, responsibility and opportunities for advancement in the workplace leads to increased happiness among employees (Raj & Ling-Meng 2024). Employees with high self-esteem tend to be more satisfied due to intrinsic workplace rewards. It is therefore important for decision makers to align the organisational goals and objectives by ensuring work is both rewarding and satisfying for employees (Reduan, Sabli & Adenan 2023).

Tezel (2023) critiques Herzberg’s theory, highlighting the existence of diverse sources of job satisfaction and dissatisfaction. Kinyanjui (2017) concluded that motivation factors such as the working environment, compensation and job security are key determinants in retaining employees within an organisation.

Applying Herberg’s two-factor theory to the context of TVET college employees is significant, as it demonstrates the importance for management not only to address hygiene factors to encourage employee retention, but also to integrate both motivation and hygiene elements in order to reduce employee turnover. It is evident that ensuring the presence of these factors is vital for preventing dissatisfaction and supporting retention within TVET colleges (Nikisi et al. 2025).

According to Varpio et al. (2020), a conceptual framework provides an overview of principles and ideas that enable the researcher to understand and develop the circumstances under investigation, as well as to communicate them effectively. Various approaches have been applied in the study of employee retention, with each addressing different dimensions of intentional employee retention. The conceptual framework underpinning this study is grounded in Herzberg’s two-factor theory, developed in 1959 by the American psychologist Frederick Herzberg. This theory proposes that employee retention is influenced by two sets of factors: motivator factors, such as job satisfaction, career development and rewards and recognition, and hygiene factors, such as job security (Alrawahi et al. 2020).

The proposed conceptual model framework presented in Figure 1 illustrates that the retention of administrative employees is influenced by factors such as career development, job security, performance appraisal, rewards and recognition and job satisfaction, each of which may exert both positive and negative effects. When employees experience dissatisfaction with management, they are more inclined to leave the organisation, whereas fair treatment increases the likelihood of them remaining with the organisation. It is evident from the literature review by Behera (2024), Rahman et al. (2017), Dias et al. (2022) and Sanchez (2023) that employee satisfaction and engagement can positively influence retention strategies, thereby contributing to favourable outcomes for employee retention.

Drawing on the factors identified in the underlying theories, a conceptual framework comprising five hypotheses was developed. These hypotheses were formulated to empirically test the relationships between the identified factors and employee retention.

This study was guided by two core philosophical assumptions: objectivist ontology and positivist epistemology (Kirongo & Odoyo 2020). Ontological and epistemological positions are central to the research process, as they provide clear guidance on how studies are conceptualised and conducted (Ylönen & Aven 2023). The adoption of objectivism was based on the premise that employee retention is a real phenomenon that can be objectively examined by identifying the factors that influence it. A positivist epistemological stance was therefore applied, enabling the measurement of employee retention factors through empirical, data-driven methods designed to test the relationships between independent and dependent variables.

This study adopted a quantitative research design, using a survey to collect data on the factors influencing employee retention at the selected HEI.

As noted by Rahman (2020), quantitative research enables the collection of data from larger samples, thereby providing a more comprehensive understanding of a particular phenomenon. A descriptive research design was employed to ensure representativeness and to capture respondents’ opinions (Ansari et al. 2022). This approach assisted the researcher in obtaining and accurately describing the respondents’ views.

The target population for this study comprised the administrative employees of Nkangala TVET College. According to the institution’s HR department, the college employed 300 administrative staff members (N = 300) across five campuses and the head office. From this population, 131 employees were excluded, resulting in a final target sample of 169 respondents. A stratified sampling technique was employed to ensure adequate representation of each subgroup and to enhance the efficiency of the sample design (Howell et al. 2020). The Raosoft sample calculator was used to determine the appropriate sample size based on a 5% margin of error and a 95% confidence interval. The study sample included only administrative staff members with more than 5 years of service (Raosoft 2023). Stratum-specific weights were applied to determine the proportion of participants from each subgroup, ensuring balanced representation. Table 1 outlines the stratified sampling distribution across the six campuses.

This study employed a structured, closed-ended questionnaire to collect data from respondents. The questionnaire was divided into three sections:

To enhance the robustness and validity of the research instrument, the questionnaire was reviewed by a statistician and study leaders and then piloted with 10 administrative employees, distinct from the main study’s respondent pool. The survey was distributed physically by a fieldworker to sub-campuses, Witbank, Mpondozankomo and the head office. However, due to a low response rate, only 45 questionnaires were eventually returned, and an online approach was subsequently adopted. In total, 119 completed questionnaires were received, yielding a response rate of 70.41%. The rate is considered satisfactory and aligns with Mailu et al. (2021), who regard such a level of participation as acceptable even within relatively small respondent groups.

This study employed both descriptive and inferential statistics to analyse the data. Descriptive statistics were presented using frequency distributions and percentages. The research data were first coded and recorded in Microsoft Excel before being imported into the latest version of the Statistical Package for the Social Sciences (SPSS, version 29.0). For inferential analysis, the study applied analysis of variance techniques and Pearson’s correlation and linear regression analysis coefficient to examine the relationships between employee retention and the identified retention factors. In addition, exploratory factor analysis (EFA) was employed as a data-driven approach to identify patterns within the variables and to explain the correlation among them (Mohameda et al. 2023).

A gatekeeper letter was obtained from the Nkangala TVET College (Ref: R0007/2023), granting permission to conduct the study. Full ethical approval was subsequently secured from the Tshwane University of Technology (TUT) Human Research Ethics Committee (HREC) under the Faculty of Management Sciences. The ethics approval number is REC2024=04=017 (MS). The study complied with institutional and international ethical standards. Participation was voluntary and based on informed consent, with respondents assured of their right to withdraw without penalty. Anonymity, confidentiality and secure data storage were maintained, and the information was used exclusively for academic purposes.

The demographic variables analysed in this study included gender, age, level of education and work experience (see Table 2). The results show that the majority of respondents were women (57.98%), and most respondents were aged 30–39 years (43.70%). In terms of education, 36.13% of the administrative employees held a national diploma. About work experience, most respondents (65.55%) had been employed at Nkangala TVET College for between 5–10 years.

Pearson’s correlation coefficient was employed to examine the relationship between employee retention and its influencing factors. As shown in Table 3, all correlations with employee retention were significant at p < 0.05, indicating meaningful relationships. The strongest positive correlation was found between employee retention and rewards and recognition (r = 0.589, p = 0.001), followed by job security (r = 0.444, p = 0.001), job satisfaction (r = 0.425, p = 0.01) and career development (r = 0.418, p = 0.001). A weaker but statistically significant relationship was observed between employee retention and performance appraisal (r = 0.288, p = 0.002).

According to Ullah et al. (2023), EFA employs measures, such as the Kaiser-Meyer-Olkin (KMO) test, Bartlett’s test of sphericity, communalities, total variance explained, rotated component matrix, composite reliability, as well as convergent validity and discriminant validity assessments, to determine the factorability of data.

In this study, the KMO value was 0.800, exceeding the threshold of 0.70, thereby confirming the adequacy of the dataset for factor analysis. Furthermore, Bartlett’s test of sphericity produced a p-value of 0.001, which falls below the 5% significance level, indicating that the correlation among variables was suitable for factor analysis (Sudershan Reddy & Kulshrestha 2019).

Communalities represent the proportion of variance in each variable explained by the extracted factors (Karimian & Chahartangi 2024). A communality value above 0.5 is generally considered acceptable, indicating weak and less reliable items. In this study, items ER2, ER5, ER6, ER7, ER9, ER12 and ER13 recorded communality values below 0.5 and were therefore excluded from further analysis.

As shown in Table 4, the remaining 36 items displayed communality values above 0.5, confirming their suitability for inclusion in the factor structure.

Table 5 presents the distribution of variance across the 36 potential factors. Notably, eight components initially recorded an eigenvalue greater than 1.0, a widely used criterion for identifying factors that account for a meaningful proportion of variance (Sappaile, Abeng & Nuridayanti 2023). Factors with eigenvalues below 1.0 explain less variance than a single item, rendering them less useful in the analysis. In this study, the six extracted factors accounted for approximately 60.45% of the total variance across the 36 variables, thereby capturing a substantial portion of the dataset’s variability. The first component explained 17.42% of the total variance (eigenvalue of 11.20), the second 12.57% (eigenvalue of 4.60), the third 11.44% (eigenvalue of 2.77), the fourth 9.94% (eigenvalue of 2.21), the fifth 7.81% (eigenvalue of 1.66) and the sixth 7.25% (eigenvalue of 1.45).

The rotation component matrix consists of rotated component loadings, which indicate both the weight of each variable on a component and the correlation between variables and their respective components (Shrestha 2021). For an acceptable rotation method, items should load above 0.5, as loadings below the threshold are considered weak and are typically removed (Howard 2023). In this study, small coefficients with absolute values below 0.5 were suppressed. As shown in Table 6, 17 items (ER2, ER5, ER6, ER7, ER8, ER9, ER12, ER13, RR1, RR2, RR3, RR4, JS_04, JS_05, JS_06, CD_1 and CD_2) were removed, resulting in five components. Notably, all rewards and recognition constructs (RR1–RR4) loaded below 0.5, indicating that they did not statistically contribute to employee retention in this context.

As stated by Haji-Othman and Yusuff (2022), composite reliability (CR) as a measure of internal consistency is used to estimate reliability, while average variance extracted (AVE) is used to assess convergent validity. Huang et al. (2013) suggest that desirable outcomes are demonstrated when AVE and CR values exceed 0.50 and 0.70, respectively. In this study, the calculations of CR and AVE were performed following Aydoğdu (2023), as presented in Table 7. The CR values ranged from 0.795 to 0.950, while AVE values ranged from 0.532 to 0.732. These results confirm that all factors met the required thresholds and that strong convergent validity was achieved.

Lim (2024) explained that discriminant validity ensures that each study measure is distinct, not captured by other constructs. As reflected in Table 7, the correlations between the constructs were all less than 1, and each correlation coefficient was lower than the square root of AVE. These results confirm the discriminant validity of the constructs.

Regression model summary in Table 8 shows that job satisfaction, performance appraisal, job security, career development and rewards and recognition at Nkangala TVET College were predictors of employee retention. Prediction strength was evaluated using the R-squared value. The higher the R-squared value, the stronger the predictive power of the model (Ozili 2023). Keer, Lohiya and Chouhan (2023) maintain that a good model may have a low R-square, while a biased model can have a high R-square. The R-squared in Table 8 revealed that the model only explained approximately 41.8% of the variance within the retention rate. This implies that the model has a moderate correlation, and the remaining 58.2% of the variance is unexplained, which suggests that other factors not included in the model also contribute to variations in employee retention. But it is still regarded as acceptable.

Regression analysis was employed to predict the relationship between employee retention and its influencing factors. As shown in Table 9, certain variables had a significant effect on employee retention within the organisation (B = 18.927, p = 0.001), indicating that the overall regression model was significant despite the insignificance of some individual factors. The strongest contributor was found between employee retention rewards and recognition, showing a positive relationship with employee retention (B = 0.820, p = 0.001), followed by career development (B = 0.196, p = 0.039), and job security (B = 0.320, p = 0.045). These findings revealed statistically significant predictors, given that their p-values were below the 0.05 threshold. However, job satisfaction (B = 0.90, p = 0.356) and performance appraisal (B = -0.88, p = 0.239) p-values were above 0.05 and considered statistically insignificant. These results support the hypothesis that at least one beta coefficient differs from zero, indicating the model’s explanatory value. Job security, career development and rewards and recognition were significantly related to employee retention, suggesting that administrative employees experiencing these factors are more likely to remain with the organisation. In contrast, job satisfaction and performance appraisal were statistically insignificant in relation to retention, suggesting employees experiencing these factors are more likely to leave the organisation.

Oke, Akinkunmi and Etebefia (2019) regard tolerance as the inverse of the variance inflation factor (VIF). A VIF value > 5 indicates the presence of multicollinearity. Ideally, the VIF should not exceed 5, and the tolerance should not drop below 0.2. Kyriazos and Poga (2023) state that when multicollinearity is detected, it is significant to identify two highly correlated variables, one of which must be removed from the model. As per the collinearity statistics shown in Table 9, tolerance ranged between 0.460 and 0.716, and the VIF ranged from 2.174 to 1.396, which are well within acceptable limits. Therefore, no significant multicollinearity was observed in the model.

As with similar studies, this research was constrained by the use of stratified sampling with a relatively small sample size of 119 administrative employees. Consequently, the extent to which the findings can be generalised to a wider population beyond the institution is limited. Furthermore, the college does not have independent remuneration policies, as it adheres to the guidelines set by the Department of Higher Education, Science and Innovation. While the findings were suitable for the purposes of this study, they may not be generalisable to all colleges in South Africa. In addition, the study employed a single research approach, which may have restricted a deeper exploration of employee experiences and perceptions. These limitations present opportunities for future research, particularly through the application of qualitative methods that could provide richer insights.

This study offers valuable insights for management and practice within TVET colleges. It emphasises the importance of prioritising rewards and recognition (the strongest driver), alongside career development and job security, as core components of HR strategies directed at retaining administrative employees. Retention initiatives should focus on these high-impact factors, which significantly influence employees’ decisions to remain with the institution. While job satisfaction and performance appraisal showed limited influence, the findings underscore the need for targeted training and professional development initiatives that align with the key retention drivers identified. By fostering employee growth and promoting a sense of belonging, TVET colleges can enhance organisational commitment. These insights are particularly valuable for decision-makers within TVET colleges and HEIs, as they provide a foundation for developing effective employee retention strategies and investing in initiatives that improve job satisfaction and career development.