This study aimed to identify the factors of empowerment in higher education institutions in Ghana by developing a scale to measure empowerment. This study was guided by the research question: What are the key elements of empowerment in higher education institutions?

This study used a quantitative approach to collect respondents’ data using a self-developed questionnaire. After completing the literature review, items that measured perceptions about empowerment were developed. These items were developed after considering the empowerment constructs identified by previous scholars. The initial item pool comprised 18 items. After reviewing feedback from a panel on the items, the item pool was reduced to 14 items. A six-point Likert scale response format was used to reduce the effect of the central tendency.

The content and face validity of the items were evaluated using a panel of seven people who did not form part of the study’s sample to assess and identify the items that should be retained, revised or removed from the questionnaire. This panel comprised three subject experts, a statistician and three colleagues. The purpose of the study was outlined to the panel members. They were requested to comment on the items and their relatedness to empowerment and outline any difficulties they experienced in completing the questionnaire.

A sample of the population of the four universities was invited to participate in the study. The sample comprised 482 of 10 890 employees at higher education institutions in Ghana. A total of 386 responses were received, and a response rate of 80% was deemed sufficient for data analysis. The respondents were categorised by gender, age, educational level and the job category. Concerning gender, males constituted 60.6% of the total sample. Most respondents (19.2%) were in the age category of 18–25 years, and the least represented age category was employees aged 56–60 years (1.8%). The percentage of employees in a public university who had up to a postgraduate level qualification was 33.1%; this was followed by graduates (30%) and undergraduates (25%). Regarding job level, 44.5% were academics (senior staff) and 55.5% were administrative staff.

The questionnaire consisted of two parts: Section A included items related to demographic information, and Section B comprised the Empowerment questionnaire. The demographic questions collected information on the age, gender, educational level and job category of the participants. Section B consisted of 14 items and measured respondents’ perceptions of empowerment. The items were developed by referring to the empowerment constructs identified in previous studies. The survey was conducted in 2023 after permission was granted by the four Ghanaian universities. The items were designed to reflect situations within a higher education institution setting, and a 6-point Likert scale (1 = totally disagree; 6 = totally agree) was used.

Ethical clearance to conduct this study was obtained from the University of South Africa Human Resource Management Ethics Review Committee on 28 October 2021 (Reference No. 2021_HRM_008). Participants consented to participate and completed the survey online. The data were secured by password protection and stored in a locked location.

The Statistical Package for Social Sciences (SPSS), version 22.0 of 2013, was used to analyse the dataset. The EFA used principal axis factoring as an extraction method and Varimax for the rotation. The results of the Kaiser–Meyer–Olkin (KMO) and Bartlett’s tests (Bartlett 1954) determined whether the data were suitable for factor analysis. The KMO value should be greater than 0.70; an average value falls between 0.70 and 0.80, a good value falls between 0.80 and 0.90 and an excellent value falls between 0.90 and 1.00. According to Sharma (1996), KMO values between 0.50 and 0.60 are flawed but acceptable. These tests confirmed that the scale items correlated adequately, and eigenvalues greater than 1.00 were used as the factor criterion. A cut-off score of 0.50 was used for factor loadings. The reliability of the empowerment measure was assessed by using a Cronbach’s alpha coefficient cut-off score of 0.70. Murphy and Davidshofer (1988) state that Cronbach values < 0.60 are unacceptable; 0.70 is low, between 0.80 and 0.90 is moderate, and > 0.90 is high.

Descriptive statistics provide an overview of the distribution of responses for the identified empowerment factors: meaningfulness, autonomy and/or decision-making and competence. Table 2 presents each factor’s mean, standard deviation, skewness and kurtosis values.

The mean scores indicate the overall level of agreement among respondents, with meaningfulness (M = 5.09, standard deviation [SD] = 0.83) and competence (M = 4.99, SD = 0.88) receiving relatively high ratings. At the same time, autonomy and/or decision-making (M = 4.54, SD = 1.02) was rated slightly lower. The standard deviations suggest that the responses for autonomy and/or decision-making were more varied than the other two factors.

Regarding the distribution, skewness and kurtosis values help assess normality. Meaningfulness (skewness = -2.06, kurtosis = 6.37) and competence (skewness = -1.56, kurtosis = 3.35) exhibited negative skewness, indicating that most respondents rated these factors highly. Their kurtosis values suggest a leptokurtic distribution (higher peak), meaning that the responses were concentrated around the mean. In contrast, autonomy and/or decision-making (skewness = –0.78, kurtosis = 0.30) is closer to a normal distribution but still slightly skewed negatively, showing a slight tendency towards higher ratings.

The descriptive statistics suggest that respondents generally perceived meaningfulness and competence more positively, while autonomy and/or decision-making showed greater response variability. The kurtosis and skewness values indicate the normality of the data (Kline 2011).

Table 3 presents the results to confirm the suitability of the data for factor analysis. The value for the KMO measures of sampling adequacy was 0.93, well above the threshold of 0.60 (Kaiser 1970). The significance level of Bartlett’s test of sphericity was 0 and below the threshold of p < 0.05 (Bartlett 1954). According to the results of the Kaiser–Olkin measure of sampling adequacy and Bartlett’s test of sphericity, factor analyses were appropriate. In addition, items 10 and 15 were cross-loaded and could not be allocated to a specific factor and were thus also deleted from further analysis. According to Hair et al. (2019) and Tabachnick and Fidell (2019), items with low communalities may fail to load meaningfully onto factors and can distort the factor solution. Field (2018) further suggests that such items do not effectively measure the intended construct, making their removal necessary to improve the model’s interpretability and validity.

Concerning the reliability of the factors identified, a coefficient alpha of 0.6 or more is acceptable (Bryman et al. 2014). The coefficients ranged from 0.787 to 0.893, which exceeded the 0.70 threshold value proposed by Hair et al. (2019). The following reliability scores were determined for each of the empowerment factors: meaningfulness (7 items): α = 0.893; autonomy (5 items): α = 0.859 and competence (2 items): α = 0.787.

The principal-axis factor analysis revealed three factors with eigenvalues exceeding 1.0, which cumulatively explained 58.73% of the variance in empowerment. The factor loadings were all above 0.50 (Veth et al. 2018) and were considered suitable for inclusion in a factor. The factor loadings in the rotated factor matrix, as illustrated in Table 4, were analysed in line with the theory and labelled as follows:

Factor 1: Meaningfulness

Factor 2: Autonomy and/or decision-making

Factor 3: Competence

Confirmatory factor analysis, using AMOS version 26 software, was performed to confirm the reliability and construct validity of the employee empowerment scale. The analysis assessed the measurement model fit and examined key validity indicators, including factor loadings, Cronbach’s alpha, CR and AVE.

The model fit was evaluated using several goodness-of-fit indices, including the chi-square statistic (χ²), comparative fit index (CFI), Tucker–Lewis index (TLI), root mean square error of approximation (RMSEA) and standardised root mean square residual (SRMR). Factor loadings were examined to ensure that all items were significantly loaded onto their respective constructs, with a recommended threshold of 0.50 or higher (Hair et al. 2019).

Cronbach’s alpha (α) and CR values were computed to establish construct reliability. Cronbach’s alpha assessed internal consistency, with acceptable reliability indicated by values above 0.70 (Nunnally & Bernstein 1994). Similarly, CR values above 0.70 confirmed adequate reliability, offering a more suitable measure within the structural equation modelling (SEM) framework, as it accounts for individual item factor loadings rather than assuming equal contributions (Raykov 1997). While Cronbach’s alpha provides a widely recognised benchmark, CR offers a more precise reliability estimate for latent constructs (Fornell & Larcker 1981; Hair et al. 2014).

Convergent and discriminant validity ensure the structural validity of the empowerment construct. Convergent validity was measured using AVE, with values above 0.50 indicating that the construct explained more than half of the variance in its indicators (Fornell & Larcker 1981). The refined factor structure was then subjected to CFA to further validate the scale. Out of the 14 items, three items (5, 11 and 12) were eliminated and not included in the CFA. These items exhibited commonalities below 0.5, indicating that they did not share sufficient variance with their respective factors and contributed weakly to the overall measurement structure. This adjustment allowed for a clearer identification of the underlying constructs and enhanced the reliability and validity of the measurement model. Average variance extracted ensured that items within a construct were highly correlated and measured the same underlying concept. Discriminant validity was established by comparing the square root of AVE with the inter-construct correlations, ensuring that each construct was distinct. By incorporating these measures, the study adhered to best practices in construct validation, ensuring that the empowerment construct demonstrated strong internal consistency, convergent validity and discriminant validity, thereby supporting the robustness of the theoretical model (Hair et al. 2014). The CFA results provided empirical support for the three-factor model of meaningfulness, autonomy and/or decision-making and competence, confirming that the scale was a reliable and valid measure of employee empowerment in higher education institutions.

The model fit indices provide an overall evaluation of the alignment between the measurement model for the empowerment scale and the observed data. Figure 1 provides a graphical representation of the alignment between the constructs of empowerment. The chi-square statistic (CMIN) of 201.400 with 68 degrees of freedom (DF) is significant, which is expected because of its sensitivity to the sample size. To account for this limitation, the relative chi-square (CMIN/DF) is assessed, with a value of 2.962 falling within the acceptable range of 1 to 3, thus confirming that the underlying structure of the data is adequately represented by the model.

Further support for model fit comes from the CFI, the normed fit index (NFI) and the TLI, which measure how well the proposed model fits compared to a null model. The CFI value of 0.940 is slightly below the recommended threshold of 0.95 but still acceptable. Similarly, the NFI and TLI values of 0.922 meet the acceptable benchmark of 0.90, reinforcing the model’s adequacy. These indices indicate that the specified factor structure provides a reasonably good representation of the data, with only minor deviations from an ideal fit.

The absolute fit indices, including the SRMR and the RMSEA, offer additional insights into the model performance. The SRMR value of 0.058, below the 0.08 threshold, suggests an excellent fit, indicating minimal residual discrepancies between the observed and predicted covariance matrices. The RMSEA value of 0.064 is slightly above the ideal threshold of 0.06 but remains within the acceptable range, implying that the model adequately approximates the population-level data. These fit indices demonstrate that the empowerment scale exhibits a satisfactory model fit, supporting its reliability and validity for measuring the empowerment construct. Table 5 presents the model fit indices for the empowerment scale.

Ensuring the reliability and validity of the measurement constructs is essential for establishing the robustness of the research model. Reliability assesses the consistency of the measurement scale, while validity ensures that the constructs accurately capture the intended theoretical concepts. In this section, CR (AVE), maximum shared variance (MSV) and Cronbach’s alpha are reported. The MSV was examined to assess discriminant validity, ensuring that the constructs are distinct. These metrics collectively comprehensively assess the measurement model’s quality and appropriateness.

To assess the construct validity and reliability, CR, AVE, MSV and maximum reliability (MaxR[H]) were calculated using Gaskin’s statistical tool (Gaskin 2016). This tool facilitated an efficient evaluation of convergent and discriminant validity, ensuring that the measurement model met acceptable validity and reliability thresholds. Table 6 presents the results of CR, AVE, MSV and the correlations among the three empowerment factors: meaningfulness, competence and autonomy and/or decision-making. The CR values for meaningfulness (0.871), competence (0.741) and autonomy (0.831) exceed the recommended threshold of 0.70 (Hair et al. 2019), indicating strong internal consistency. Additionally, the maximum reliability, MaxR(H), values are slightly higher but remain close to their respective CR values, further supporting the reliability of the constructs. These findings suggest that the empowerment scale is reliable for measuring empowerment in higher education institutions.

The AVE values indicate the extent to which a construct explains the variance in its indicators and were used to assess convergent validity. The AVE values for meaningfulness (0.575), competence (0.589) and autonomy (0.552) exceed the minimum recommended threshold of 0.50 (Fornell & Larcker 1981), confirming that each construct captures a sufficient proportion of variance from its items. As all AVE values surpass 0.50, it can be concluded that the items measuring each construct are well related and contribute meaningfully to their respective latent variables.

Discriminant validity was determined by comparing the square root of the AVE (bolded diagonal values) with the inter-construct correlations. The square root of AVE for meaningfulness (0.758), competence (0.768) and autonomy (0.743) is larger than their correlations with other constructs, indicating that each construct is distinct from the others (Rönkkö & Cho 2022). The MSV for competence (0.558) is close to its AVE (0.589), suggesting substantial shared variance with other constructs but not to the extent that it compromises discriminant validity. Similarly, meaningfulness (MSV = 0.558) and autonomy (MSV = 0.482) exhibit an acceptable level of distinction.

The correlation analysis reveals significant positive relationships among the three factors. Meaningfulness and competence (r = 0.747) show a strong association, suggesting that employees who find their work meaningful tend to feel more competent. Meaningfulness and autonomy (r = 0.694) also exhibit a strong relationship, indicating that employees who perceive their tasks as meaningful are more likely to experience autonomy in decision-making. Meanwhile, competence and autonomy (r = 0.581) show a moderate correlation, implying that employees who feel competent in their roles also experience a certain level of decision-making autonomy. These findings highlight the interconnected nature of the three dimensions of empowerment while affirming their distinctiveness as separate constructs.

A KMO of 0.935 was obtained. All the items had factor loadings equal to or greater than 0.50. The total variance was 58.73%. All the factors had Cronbach’s alpha values > 0.70, which indicated high internal consistency. The items discriminated sufficiently with no item having a median close to one of the extremes.

The study contributes to understanding empowerment within the context of higher education in several ways. It confirms the role of the meaningfulness of work as a significant dimension of empowerment, resonating with the findings of Pradhan and Mishra (2019). Regarding autonomy, the power to make decisions should be delegated to staff. This confirms the findings of various studies on how power, information, resources and rewards are shared to enable staff to make autonomous decisions (Fernandez & Moldogaziev 2013; Gomez & Rosen 2001). Competence plays an important role in the efficient functioning of departments and colleges and refers to staff’s content knowledge, cognitive processing, behavioural skills and personality traits (Potnuru, Sahoo & Sharma 2019).

Human Resource practitioners and managers should assess the empowerment of employees to manage performance. Key performance indicators should consider the extent of empowerment afforded to staff. Higher education institutions could use the results of this study as input to leadership development programmes. The ability of leaders to disseminate information, delegate authority, create forums for joint decision-making and create opportunities for staff development is critical to empower staff to change the status quo. Performance management should focus on intrinsic motivation and cognitions such as meaning, competence and self-determination (Angelovska et al. 2018).

Overall, the study adds to the existing theories of empowerment. It explains which dimensions of empowerment are the most important to higher education staff. Government institutions are known for their bureaucracy and rigid hierarchical structures. Structural and psychological empowerment are critical for facilitating changes in educational institutions.

Culture, leadership, workload, supervisor support and other institutional variables must also be considered influencers on empowerment. It is recommended that future studies refine the measurement of empowerment to include contextual factors.

Empowerment dynamics change constantly. As a result, future studies could investigate how online work influences empowerment dynamics in an educational institution.

Using a measurement instrument that integrates structural and psychological empowerment would provide a more comprehensive understanding of how empowerment could be promoted. This would further assist in developing strategies to advance empowerment in the workplace. The COVID-19 pandemic has brought many challenges, and more research is needed on how remote work influences perceptions of empowerment.