Ethical leadership and staff retention in Uganda’s health care sector: The mediating effect of job resources

Abstract This paper examined the mediating effect of job resources on the relationship between ethical leadership and staff retention in Uganda. Utilising a sample of 214 health care workers, a cross-sectional research design was adopted. SPSS v. 21 and Analysis of Moment Structure v. 21 were used in the data analysis. Job resources as a whole was established to be a partial mediator in the relationship between ethical leadership and staff retention. Ethical leadership and job resources individually predicted staff retention. Overall, 66% of the variance in staff retention is explained. As a limitation, data were collected in a cross-sectional survey, which might have confounded the causality among the studied variables. Future researchers could conduct longitudinal studies to address this limitation. The findings suggest that leaders need to be ethical in their dealings if they are to enhance the retention function of their staff in their respective organisations. Health care managers should endeavour to provide job resources in order to mitigate the effects of job overload so as to improve on staff retention in their organisations. This study finds support for job resources as a mediator in the relationship between ethical leadership and staff retention in Uganda’s health care sector. This is consistent with the requirements for the examination of process variables in relationships so as to make concrete inferences.


PUBLIC INTEREST STATEMENT
The examination of the job resources as a process variable that mediates the relationship between ethical leadership and staff retention is very interesting. First, it is relevant for researchers in the further development of the theories, more so the JD-R model that is still being interrogated. The theoretical interrogation should be able to help us understand how organisations can deal with the challenge of staff retention in the wake of heightened firm competition for talented and productive staff. Second, conclusions derived in this study can be applied to deal with issues of employee retention in organisations.

Introduction
Interest in employee retention in workplaces is continuing to attract many researchers and practitioners in the recent past (Oki, 2014;Pienaar & Bester, 2015). This is partly due to the evidence provided in the nature of the benefits that staff retention extends to organisations in form of increased productivity (Siebert, 2006), employee heightened performance (Tugade & Fredrickson, 2004), to mention but a few. This study is in the context of a developing country and most specifically focussing on Uganda. In Uganda, the retention of health care workers, remains one of the daunting challenges (Kana, 2009;Karan, DeUgarte, & Barry, 2016). This has been exacerbated by the brain drain challenge where most critical staff, such as medical doctors have left civil service engagement for better opportunities abroad (Kizito et al., 2015;Naicker, Plange-Rhule, Tutt, & Eastwood, 2009;Soucat, Scheffler, & Gebreyesus, 2013). Even where the Government of Uganda has intervened with selective remuneration perks for the health care staff where health workers enjoy a different pay scale from their counter parts in the Public Service, allowances in form of hard-to-reach, and other monetary benefits, substantial retention remains unattained especially in rural areas of Uganda.
Therefore, establishing appropriate remuneration packages to retain human resources in Uganda's health care sector, remains a challenging matter (Kajungu & Mugisha, 2015). The consequences of poor staff retention of health care staff in Uganda are clear: the health worker to population ratio in Uganda is 1:1,298 compared to the World Health Organisation (WHO) guidelines of 1:439 (GoU, n.d.). Further, arising out from poor retention, the doctor to patient ratio was estimated at 1:24,725 in 2013, with a nurse to patient ratio of 1:11,000. WHO recommends one physician per 1,000 people. For example, Trinidad, with 1.3 million people, has a doctor to patient ratio 12 times better than Uganda, with 39 million people. WHO ranks Trinidad's health system at 68 and Uganda's at 149 (PoU, 2013).
One of the key variables that has a high potential in explaining staff retention is ethical leadership. Brown, Treviño, and Harrison (2005, p. 120) define ethical leadership as "the demonstration of normatively appropriate conduct through personal actions and interpersonal relationships, and the promotion of such conduct to followers through two way communication, reinforcement, and decision making". Studies indicate that ethical leadership though a nascent concept has a lot of potential in explaining how organisations can attain high productivity (Brown & Treviño, 2006;Brown et al., 2005). Brown et al. (2005) consider ethical leaders as those who are honest, trustworthy, fair and caring, those that structure work environments fairly. Consistent with Brown et al. (2005), we see leaders' fair and moral behaviour as a central component of ethical leadership in light of morality and fairness dimensions. Some of the significances of ethical leadership in organisations are: it lowers employee turnover (Elçi, Şener, & Alpkan, 2013), and if well utilised, has the potential of nearly obliterating employee misconduct (Mayer, Kuenzi, & Greenbaum, 2010). Further, it has been noted that it stimulates organisational innovativeness (Pucetaite, 2014).
Job resources are defined as those aspects of the job that are assumed to play either an intrinsic motivational role in fostering workers' personal development, or an extrinsic motivational role as an instrument for achieving work goals, reducing the physiological and psychological costs associated with job demands (Bakker, 2011). Research indicates that job resources can play an intrinsic motivational role in nurturing individual growth, learning and development, or through an extrinsic motivational role; further, they also help employees accomplish working goals (Bakker, Demerouti, de Boer, & Schaufeli, 2003;Karatepe & Olugbade, 2009). Equally important, studies have indicated that job resources are linked to positive organisational outcomes (Coetzer & Rothmann, 2007;Hakanen, Bakker, & Demerouti, 2005;Xanthopoulou, 2009).
In line with this study, researchers have also established that ethical leadership leads to staff retention in organisations where it is practised (Lee-Kelley, Blackman, & Hurst, 2007;Rai, 2013;Simons & Roberson, 2003;Yener, Yaldiran, & Ergun, 2012). Nonetheless, even when it has been suggested that job resources and ethical leadership are important drivers and outcomes of organisation's progress, there is no independent research to date that has demonstrated empirically how these concept are linked in health sector within Uganda's rural districts. This is one of the research gaps that the current research addresses. Additionally, although preliminary research on ethical leadership is encouraging, there is a dearth of research investigating the underlying processual mechanism to explain the relationship between ethical leadership and employee retention in organisations. Such research is theoretically important because it presents insights into what ethical leaders essentially do in influencing employee retention levels in organisations. Basically, this has implications to designing organisations in order to lessen the rate of employee turnover. In the present study, we examine job resources as a mediator of the relationship between ethical leadership and employee retention.
This research has a number of intended contributions. First, though there is increasing interest in ethical leadership, the majority of research has focused on prosocial outcomes (Brown et al., 2005;Piccolo et al., 2010). We expand on recent work (Mayer et al., 2010;Ponnu & Tennakoon, 2009) by examining employee retention as an outcome. Second, there is modest empirical support for the underlying mechanism linking ethical leadership to employee retention (see Elçi et al., 2013 for an exception). To address this gap in the literature, we examine job resources as a mediator. Third, although theoretical work has highlighted the link between ethical leadership and job resources (Basar & Flizoz, 2015), we seek to empirically examine this theorised relationship.

Ethical leadership and staff retention
The framework linking ethical leadership to staff retention through job resources in this study is explained by two theoretical perspectives: the Leader Member and the Conservation of Resources theory. The hallmark of the Leader-Member Exchange (LMX) theory is advancing leadership as an exchange relationship that develops or builds within the vertical dyad (supervisor-supervisee) over time during task making activities. This process leads to development of high exchange relationships. In the corresponding relationship, the extent of autonomy a supervisor grants a subordinate to discuss his/her role, projects successive behaviour on the part of both supervisor and subordinate (Dansereau, Graen, & Haga, 1975), LMX, high-exchange relationships typify high-level of trust, fondness, and respect in what amounts to (employer-employee relationships) and entail expectations of mutual exchange. The leader grants subordinates' desired outcomes, like attractive tasks, extra tasks, and generously proportioned rewards, and the subordinates reciprocate with increased desire and commitment to work and loyalty to the principal or the leader. Exchange relationships will thus develop progressively over time and reinforced by the conduct of the leader and the subordinates (Yukl, O'Donnell, & Taber, 2009). Subsequently, it is likely that employees may come to a decision to continue working for the organisation, hence increasing employee retention. LMX therefore proposes that if leader displays ethical leadership to subordinates, employee retention is likely to be high. Therefore, we propose that: H 1 : Ethical leadership is positively linked to staff retention.

Job resources and staff retention
This study uses the JD-R model to examine how job resources influence staff retention among the Ugandan health care workers. On the basis of this model, we hypothesise that the work environment influences employees' behaviour. The Job Demands-Resources (JD-R) model (Demerouti, Bakker, Nachreiner, & Schaufeli, 2001) proposes that while the employees in diverse organisations may be faced with different working environments, the features of these environments may constantly be classified in two general groups-job demands and job resources-thus comprising an overarching or embracing model with universal application to diverse work settings, regardless of the specific demands and resources involved. Job demands denote physical, psychological, social, or organisational aspects of the job that entail sustained physical and/or psychological (cognitive and emotional) effort and are consequently related to certain physiological and/or psychological costs (Bakker & Demerouti, 2007). Examples are a high work pressure, role overload, poor environmental conditions and problems related to reorganisation. Job resources define those physical, psychological, social, or organisational aspects of the job that are either/or: (1) functional in attaining work objectives; (2) reduce job demands and the accompanying physiological and psychological costs; (3) trigger personal growth and development. Resources may be situated at the level of the organisation at large (for instance, pay, career opportunities, job security), at the interpersonal level (e.g., supervisor and co-worker support, team climate), at the level of the organisation of work (say, role clarity, participation in decision-making), and at the task level (for instance, performance feedback, skill variety, task significance, task identity, autonomy, see- Hackman and Oldham (1975), Bakker, Schaufeli, Leiter, and Taris (2008).
A second assumption in the JD-R model is that work characteristics may induce two different processes. First, high job demands (i.e., work overload) are ordinarily associated with physiological and psychological costs for employees, especially when high effort on the part of the employee is required to meet the demands (Nahrgang, Morgeson, & Hofmann, 2011). Also, poor or missing job resources inhibit actual goal attainment, which is likely to cause failure and frustration that culminate into reduced staff retention. When the external environment is devoid of sufficient job resources, employees cannot reduce the potentially negative influence of high job demands and they cannot achieve their work goals. Numerous studies have supported the JD-R model and its underlying processes (see Demerouti & Bakker, 2011). However, rising empirical evidence suggests that job resources are involved not only in the motivational process, but also in the energetic process, signifying that lack of job resources increase job strain thereby lowering staff retention.
We expect job resources to facilitate actual goal achievement, including dealing with the challenges caused by job demands, which evoke feelings of success that further enhance staff retention. Armstrong (2014) and Mullins (2002) argue that work environment influences employees' behaviour. In line with the physiological and psychological costs for employees, especially when high effort on the part of the employee is required to meet the job demands, we predict that job resources (and not job demands) will have a positive effect on staff retention. Based on this, we hypothesise that job resources are positively related to staff retention. Therefore, we propose that: H 2 : Job resources are positively related to staff retention. Brown et al. (2005) conceptualise ethical leadership as: "the demonstration of normatively appropriate conduct through personal actions and interpersonal relationships, and the promotion of such conduct to followers through two-way communication, reinforcement, and decision-making" (p. 120). The job resources of health care workers may be located at the level of the organisation (for instance, pay, job security); interpersonal and social relations (e.g., co-worker support); the organisation of work (such as role clarity, participation in decision making); or at the level of the task (say, skill variety, task identity, task significance, autonomy, or performance feedback) (Buys & Rothmann, 2009;Shahkarami, Shafiee, & Hamedani, 2016).

Ethical leadership and job resources
Specifically, studies have established a relationship between ethical leadership and autonomy (Shahidul, 2015). In other words, ethical leadership allows employees considerable autonomy over work behaviours and influence on decisions in their workgroups. Further, drawing on social learning theory, it clear that ethical leadership predicts employees' conflict resolution efficacy, which is a job resource, and subsequently increases employees' ability to deal with conflict situations in the workplace; that is, relationship, task, and process conflict- (Babalola, Stouten, Euwema, & Ovadje, 2016).

Mediation of job resources on the relationship between ethical leadership and staff retention
Despite the voluminous support found for the strong main effects of ethical leadership on staff retention (Lee-Kelley et al., 2007;Rai, 2013;Simons & Roberson, 2003;Yener et al., 2012), relatively little research has been conducted on the mediating potential of job resources on the ethical leadership-staff retention sequence. Further, although preliminary research on ethical leadership is encouraging, there is a dearth of research investigating the underlying mechanism to explain the relationship between ethical leadership and employee retention in organisations, (with the exception of Walumbwa, Morrison, & Christensen, 2012). Such research is theoretically important because it presents insights into what ethical leaders essentially do to influence employee retention levels in organisations. Basically, this has implications for crafting or designing organisations to lessen the rate of employee turnover.
Previous studies have positioned job resources as significant in the mediation literature (Loi, Hang-yue & Foley, 2006;Suazo & Turnley 2010). Loi et al. (2006) found that perceived organisational support (POS), which is job resource, mediates the relationship between procedural and distributive justice and organisational commitment and intention to leave. Also, Suazo and Turnley (2010) established that POS mediated the relationship between psychological contract breach (PCB) and four individual differences: positive affectivity, reciprocation wariness, equity sensitivity, and Protestant Work Ethic. However, the extent to which job resources mediate the relationship between ethical leadership and staff retention in Uganda's health care sector has not been investigated to date. Based on this, we hypothesise that: H 4 : Job resources mediate in the relationship between ethical leadership and staff retention 3. Methododology

Study population, sample size and sampling procedure
A total of 460 workers from two hospitals of Iganga and Bugiri District Referral Hospitals participated in the study. From this population, using a formula suggested by (Yamane, 1967), a sample size of 214 was selected through simple random sampling technique. This technique, allowed all respondents equal chance of participation in the study (Zikmund, Babin, Carr, & Griffin, 2010). Out of 214 respondents, we retrieved 201 usable questionnaires constituting a response rate of 93.9%. Most participants were male (67%). Of the participants, about 30% were younger than 29 years, 40% were between 30 and 39 years old, while 30% were aged between 40 and 49 years. Some participants (36%) had a Diploma, 50% degree, and 14% had masters' qualification. With respect to their tenure, 15% had worked for less than 5 years in their hospital. Thirty four percent had a working experience ranging between 6 and 10 years, 20%, between 11 and 15 years, while 31%, had over 16 years working experience.

Data collection and respondents
This study was conducted in two rural government hospitals staff (Iganga and Bugiri District Referral Hospitals) in Uganda. First, the Human Resource (HR) department of these hospitals were contacted.
Using the HR department, we secured one contact person in each hospital who assisted in the administration of the tool. The hospital administration and the research team agreed to protect employees' identity. As a result, the tool did not contain personal data or sensitive information that allowed respondents' identification. Further, research participation in the study was voluntary.
All cadres in the two hospitals participated in the study. The data collection process was carried out in partnership with the respective HR departments. The respondents were given questionnaires along with cover letters explaining the study. The questionnaires could be filled out during working hours.

Job resources
These were assessed with the scale proposed by Van Veldhoven, de Jonge, Broersen, Kompier, and Meijman (2002). The scale contains five subscales for job resources: autonomy (five items; for instance, "Are you allowed to decide by yourself how to perform your work?"), social support from supervisor (five items; "If necessary, can you ask your superior for help?"), social support from colleagues (five items; "Can you count on your colleagues when you encounter difficulties in your work?"), decision latitude (six items; "Are you free to decide what work activities to enroll in?"), and challenge (five items; "My work is very interesting."). Items are scored on a 6-point rating scale, ranging from 1 ("never") to 5 ("always"). The compound "Job resources" score was computed as the sum of the three sub-scores. This is a practice which was previously used in several other studies (e.g., Hu, Schaufeli, & Taris, 2011).

Ethical leadership
This was assessed with the scale proposed by De Hoogh and Den Hartog (2008). This scale was modified to address the study, some of the previous dimensions, such as despotic leadership, and optimism about the future, were not included. Their non inclusion was based on the reasoning that they are suitable for political surveys (Stevens, 2007). The new scale contains three subscales for ethical leadership: morality and fairness (six items: for instance, "My supervisor makes sure that his/her actions are always ethical"), role clarification (five items; "my supervisor explains who is responsible for what"), and power sharing (nine items; "my supervisor allows subordinates to have influence on critical decisions"). Items are scored on a 6-point Likert like rating scale, ranging from 1 ("disagree very much") to 6 ("agree very much"). The index score "Ethical Leadership" score was computed as the sum of the three sub-scores.

Staff retention
This was assessed with the scale proposed by Döckel, Basson, and Coetzee (2006). The scale contains five subscales for staff retention: compensation (18 items: for instance, "I am comfortable with my current salary"), job characteristics (four items; "The job requires me to use a number of complex or high level skills"), supervisor support (six items; "My supervisor looks for opportunities to praise positive employee performance, both privately and in front of others"), training and development (eleven items: "I think my school places the right amount of importance on training) career opportunities (11 items; "My chances for being promoted are good"), and work life balance (nine items; "I often feel like there is too much work to do"). Items were scored on a 6-point Likert like rating scale, ranging from 1 ("disagree very much") to 6 ("agree very much"). The composite "Staff Retention" score was computed as the sum of the four sub-scores.

Controlling for common methods variance and common methods bias
The influence of Common Methods Variance (CMV) has been an extensively cited concern in social research (Podsakoff, MacKenzie, & Podsakoff, 2003). CMV suggests that variance in observed scores is partly linked to a methods effect. CMB refers to the extent to which correlations are altered (inflated) due to methods effect (Meade, Watson, & Kroustalis, 2007). This study employed a questionnaire to enlist for information. Questionnaire-based behavioural studies are affected by the problem of CMV (Gorrell, Ford, Madden, Holdridge, & Eaglestone, 2011). This study does not substantially have this problem because the researchers performed a confirmatory factor analysis (CFA) (Marsh & Hocevar, 1988). The CFA procedure gives unique estimates of trait, method, and error variance present in the correlation or covariance matrix (Bagozzi & Yi, 1990). The CFA procedure is based on the assumption that the first unrotated factor derived from a principal components analysis extracts the greatest proportion of CMV across all variables (Kemery & Dunlap, 1986). To determine whether any meaningful relationships exist after accounting for CMV, the first unrotated factor was partialled out of all the correlations among variables and the relationships of interest was again examined for significance.

Confirmatory factor analyses
In order to demonstrate the construct validity of the measures, CFA was performed using the AMOS v. 21 software package (Arbuckle, 2009;Kline, 2011). Prior to CFAs, data were examined for multicollinearity, non-normality, and outliers. Initially, correlations higher than .85 were suggestive of multicollinearity, while a skewness index over 3 and a kurtosis index higher than 10, were standards for non-normal data (Weston & Gore, 2006). Scores with 3 standard deviations afar from the mean were regarded outliers and eliminated from the analysis. The inspection of the data revealed tolerable multicollinearity issues. The distribution of the variables met the requirements of normality.
We used maximum likelihood procedure to estimate the CFA models. A total of three models were estimated. Model 1, composed of three first order factors, that is, autonomy, social support from supervisors and challenge, each one derived from their respective items. Model 2, composed of three factors in which ethical leadership (morality and fairness, role clarification and power sharing) were derived from their respective items (first order factors). Staff retention was modelled as second order factors composed of five (i.e., job characteristics, supervisor support, training and development, career opportunities and work life balance).
The models were compared by means of both absolute and relative indices (Byrne, 2010;Kline, 2011). Three absolute values goodness were calculated: the χ 2 goodness of fit statistic, the Root Mean Square Error of Approximation (RMSEA), and the Goodness of Fit Index (GFI). Relative indexes, the Normed Fit Index (NFI), the Comparative Fit Index (CFI), and the Tucker-Lewis Index (TLI) were evaluated. Following the recommendations of Byrne (2010), and (Kline, 2011), well-fit models should have a non-significant χ 2 , RMSEA smaller than .06, and values bigger than .95 for the rest of the indices.

Mediation analyses
In order to test for mediation, two structural equation models using AMOS v. 21 were estimated. The first model is the theoretical model (mediated model) that postulates that job resources mediate the relationship between ethical leadership and staff retention. The second model (nonmediated) portrays a situation where job resources do not mediate the relationship between ethical leadership and staff retention. As such both direct and indirect effects (mediated through job resources) of both ethical leadership and staff retention are modelled. This approach to mediation testing is consistent with the recommendation of Anderson and Gerbing (1988) on testing for mediation.
The non-mediated model did not adequately fit the data. As the non-mediated model did not adequately fit, a χ 2 difference test between the two models (i.e., mediated model and the nonmediated model) was not done. Consequently, we resorted to other criteria to compare the two models. Morgan and Hunt (1994) suggested a four criteria in a SEM model assessment: (1) total model fit as measured by CFI; (2) percentage of theorised significant paths; (3) amount of variance explained as measured by squared multiple correlations (SMC); and (4) parsimony evaluated by the parsimonious NFI. In probing the mediated and non-mediated models on these four criteria, results showed that the mediated model was better depiction of the data.
For instance, three of the five (about 60%) of the mediated model's paths are supported at the p < 0.01 level of significance. To the contrary, two of five (40%) of the non-mediated model's paths are supported at the p < 0.01 level of significance. In addition, the mediated model explained more variance as measured by SMC in the endogenous variable (staff retention). The mediated model is more parsimonious compared to a non-mediated model. The higher NFI and CFI show that the mediated model offers a better fit of the data. Based on these criteria, it was affirmed that the mediated model was a more accurate and useful depiction of the relationships among the study variables. Accordingly, the hypotheses were tested based on the mediated model.

Preliminary analyses
The reliability of the scales was good with alpha values over .70 (Nunnally & Bernstein, 1994). As expected, all the variables were found to be positively and significantly related showing small and moderate correlation coefficients.

Measurement models
Prior to estimation of the hypothesised staff retention model, that is the mediating effect of job resources in the relationship between ethical leadership and staff retention, it was essential to estimate the measurement models for ethical leadership, job resources and staff retention. This was done to test the effectiveness of the measure. Below are the results of the Confirmatory Factor Analysis/Measurement Models.

Measurement model for ethical leadership
The ethical leadership measurement model in Figure A1 in appendix A, shows an NFI of 0.984, which is a demonstration of strong convergent validity. The χ 2 statistic of 18.027 was not significant at the 0.05 level: its p-value was 0.81, and degrees of freedom (11), signifying that the model adequately fitted the data. This was validated by other fit indices: RMSEA (.054), TLI (.970), IFI (.984), GFI (.978), and AGFI (.944). According to Fornell and Larcker (1981), the average variance extracted (AVE) as a measure for discriminant validity, should be above 0.5. In our study, the AVE, was 0.51. Analysis of the observed factor loadings in relation to their standard errors showed evidence of a link between ethical leadership and its respective constructs (Schreiber, Stage, King, Nora, & Barlow, 2006). Assessment of item reliability was based on the multiple regressions (R 2 ). Save for ELM 2, ELM 5 and ER 3, all the other regression weights (R 2 ) values were above 0.5, which is an indication of satisfactory reliability for each item (Kline, 2005). Accordingly, each item was a reliable factor for ethical leadership.

Measurement model for job resources
The job resources measurement model in Figure A2 in appendix A shows an NFI of 0.972, which is a demonstration of strong convergent validity. The χ 2 statistic of 11.558 was not significant at the 0.05 level; its p-value was 0.398, and degrees of freedom (11), signifying that the model adequately fitted the data. This was validated by other fit indices: RMSEA (.015), TLI (.997), IFI (.999), CFI (.999), GFI (.985), and AGFI (.963). According to Fornell and Larcker (1981), the AVE as a measure for discriminant validity, should be above 0.5. In our study, the AVE, was 0.50, which meets the suggested minimum. Analysis of the observed factor loadings in relation to their standard errors showed evidence of a link between job resources and its respective constructs (Kline, 2005;Koufteros, 1999;Schreiber et al., 2006). Assessment of item reliability was based on the multiple regressions (R 2 ). Save for CTY 1, and CTY 5, all the other regression weights (R 2 ) values were above 0.5, which is an indication of satisfactory reliability for each item (Kline, 2005). Accordingly, each item was a reliable factor for job resources. Therefore, results from CFA indicate that job resources is a three dimensional construct comprising of autonomy, social support from supervisors and challenge. Social support from colleagues and decision latitude constructs were not confirmed factors of job resources in Uganda's health care settting.

Measurement model for staff retention
The staff retention measurement model in Figure A3 in appendix A shows an NFI of 0.947, which is a demonstration of strong convergent validity. The χ 2 statistic of 35.971 was not significant at the 0.05 level: its p-value was 0.174, and degrees of freedom (29), signifying that the model adequately fitted the data. This was validated by other fit indices: RMSEA (.033), TLI (.983), IFI (.989), CFI (.989), GFI (.969), and AGFI (.942). According to Fornell and Larcker (1981), the AVE as a measure for discriminant validity, should be above 0.5. In our study, the AVE was 0.493, a figure close to the suggested minimum. Analysis of the observed factor loadings in relation to their standard errors showed evidence of a link between staff retention and its respective constructs (Kline, 2005;Koufteros, 1999;Schreiber et al., 2006). Assessment of item reliability was based on the multiple regressions weights (R 2 ). Save for COT 13, SUS 4, WKL 1, all the other regression weights (R 2 ) values were above 0.5, which is an indication of satisfactory reliability for each item (Kline, 2005). Accordingly, each item was a reliable factor for staff retention (Table 1).

Testing the hypotheses
Prior to conducting the SEM analysis for hypotheses testing, we assessed the hypothesised measurement models for ethical leadership, job resources and staff retention, which indicated acceptable fit of the data of the population studied. We therefore continued to estimate the structural model using SEM. The results of the analysis are shown in Figure A4.
An analysis of the SEM staff retention model in Figure A4 demonstrates an acceptable fit. The standardised regression weights show that ethical leadership positively and significantly predicts staff retention ( β = 0.341; p < 0.001). Job Resources also predicts staff retention ( β = 0.685; p < 0.001). On the basis of this, H 1 that stated that ethical leadership is positively related to staff retention, is supported. Also, H 2 ( job resources is positively related to staff retention) is supported. Ethical leadership also predicted job resources ( β = .168; p < 0.01), hence, H 3 is supported.
As H 1, H 2 , and H 3 were supported, it became necessary to test H 4 that stated, "Job resources mediate the relationship between ethical leadership and staff retention". Mediation results are displayed in Table 2 below. The need to explore why a relationship exists between ethical leadership and staff retention necessitated this mediation test. In this study, job resources offer a mechanism in explaining the relationship between ethical leadership and staff retention. To determine the mediation effect in the SEM model, we considered the direct and indirect effects (see Table 3). SEM is recommended for mediation analysis because of its versatility and ability to analyse complicated relations among the variables (Gefen, Rigdon, & Straub, 2011;Hair, Black, Babin, & Anderson, 2010;Kline, 2011;Richter, Sinkovics, Ringle, & Schlägel, 2016). Analysing the direct and indirect effects, including the mediating effect is considered as an initial step for assessing the complicated measurement and structural model for a study (Hair et al., 2010;Sobel, 1982). According to Hair et al. (2010), in performing such modelling analyses, the mediating relations among the variables ought to be analysed using path analysis between related variables. Further, testing mediation through SEM allows the researcher to take care of measurement errors by estimating the common and unique variance independently. This, technique consequently is likely to increase the probability of discovering any indirect effect if it does exist.  Figure A4 show that the standardised total effect of ethical leadership on staff retention is significant (β = 0.456; p < 0.001). Nonetheless, when the paths coefficients of the relationship between ethical leadership and staff retention was controlled through the introduction of the mediator (job resources), the original significant path between ethical leadership and staff retention reduced from β = 0.456; p < 0.001 to β = 0.341; p < 0.001, but the model remained significant. The difference is the indirect effect (β = 0.115). This signifies partial mediation of job resources on the relationship between ethical leadership and staff retention. This result support H 4 , that-job resources mediate the relationship between ethical leadership and staff retention. By and large, the mediated model explains 66% of the variance in staff retention.

Discussion
The present study examined whether job resources mediate the association between ethical leadership and staff retention. Based on the JD-R model and LMX theory, this study proposed that job resources, defined as those physical, psychological, social, or organisational aspects of the job that are either/or: (1) functional in attaining work objectives; (2) reduce job demands and the accompanying physiological and psychological costs; (3) trigger personal growth and development   Bakker et al. (2008) carry the effect of three ethical leadership facets (i.e., morality and fairness, role clarification, and power sharing) on staff retention.
Results confirmed the mediation hypothesis. Job resources variable was a significant and a partial mediator for ethical leadership on staff retention. Specifically, the influences ethical leadership (morality and fairness, role clarification, and power sharing) on staff retention were carried by job resources. As expected, all the associations were positive. This means that leaders who are ethical, honest, fair in their actions, provide the opportunity for employee creativity, improved interpersonal relationships, and performance of diverse tasks. These feelings of availability of job resources, in turn, relate to higher levels of staff retention. In line with the assumptions of LMX theory, the results suggest that ethical leadership and job resources created a leader-resource caravan process (Dansereau et al., 1975). According to this, the availability of ethical leadership acts as a caravan passageway that triggers a process of accumulation of job resources. In doing so, high levels of job resources foster the motivational process, proposed by the JD-R model, which increases staff retention. Similarly, the job demands, such as a high work pressure, role overload, poor environmental conditions and problems related to reorganisation are mitigated by the job resources (autonomy, social support from supervisors and challenge), leading to increased staff retention.
Further, the results demonstrate that better job resources (autonomy, social support from supervisors and challenge) must intervene to improve staff retention when initially, ethical leadership is operational. In support of the original hypothesis, the findings indicate that the relationship between ethical leadership (which causes the scores observed on morality/fairness, role clarification and power sharing) and job resources is robust and significant. The findings are in line with the assertion that effective ethical leadership will heighten the availability and usage of job resources in organisations. Undeniably, the findings of Bakker et al. (2008), job resources (such as performance feedback, skill variety, task significance, task identity, autonomy can ease functional attainment of work objectives), reduce job demands and the accompanying physiological and psychological costs, can trigger personal growth and development among others. These suggest that these job resources must be borne by ethical leaders if staff retention is to be realised in organisations.
These results are consistent with previous studies that positioned job resources as significant in the mediation literature. Loi et al. (2006) found that POS, which constitute job resources, mediate the relationship between procedural and distributive justice and organisational commitment and intention to leave. Also, Suazo and Turnley (2010) established that POS mediated the relationship between PCB and four individual differences: positive affectivity, reciprocation wariness, equity sensitivity, and protestant work ethic. The influence of the leader was also a significant factor in how employees viewed organisational support (Dawley et al., 2010). Based on this, the mediation effect of job resources, is alluded to.
The findings of this study indicating that ethical leadership influence job resources is insightful. Therefore, job resources account for the relationship between ethical leadership and staff retention. In the absence of job resources as a mediator, the direct relationship between ethical leadership and staff retention can plausibly be regarded as unexplained. Accordingly, this study supports those claims that process or third variables account for significant variations in exogenous variables (Mathieu, DeShon, & Bergh, 2008;Ndofor, Sirmon, & He, 2011). The results of this study agree with Mihelič, Lipičnik, and Tekavčič (2010), who reviewed literature and established the need for conceptual clarity on ethical leadership concept. Therefore, proper grasp of the ethical leadership concept would better explain how staff retention can be augmented in organisations. In a way, the findings add to the conceptual development of ethical leadership and job resources and their potential in predicting staff retention in organisations.
The findings of this study suggest that whilst ethical leadership is necessary for staff retention, the total effect on staff retention is improved by ethical leadership and job resources, which is consistent with the LMX theory (Buch, Thompson, & Kuvaas, 2016), and the JD-R model (Demerouti et al., 2001). The LMX theory suggests that leadership is an exchange relationship that develops or builds within the vertical dyad (supervisor-supervisee) over time during task making activities. This process leads to development of high exchange relationships which makes the employees to reciprocate by continuing to work for the organisation. Similarly, the JD-R model proposes that while the employees in diverse organisations may be faced with different working environments, the features of these environments may constantly be classified in two general groups-job demands and job resourcesthus comprising an overarching or embracing model with universal application to diverse work settings, regardless of the specific demands and resources involved. The job demands, such as a high work pressure, role overload, poor environmental conditions, and problems related to reorganisation are mitigated by the job resources (autonomy, social support from supervisors, and challenge), leading to increased staff retention. It is also in tandem with the findings of Bakker et al. (2003), Karatepe andOlugbade (2009), andCoetzer andRothmann (2007) who found out that job resources can play an intrinsic motivational role in nurturing individual growth, learning and development, or through an extrinsic motivational role that helps individuals accomplish working goals.

Implications
This study has a number of theoretical implications. First, we find support for a mediator in the relationship between of ethical leadership and employee retention levels in organisations. Although substantial empirical research has considered the relationship between ethical leadership and employee retention (Lee-Kelley et al., 2007;Rai, 2013;Simons & Roberson, 2003;Yener et al., 2012), we at present do not have a concrete understanding of the process underlying the relationship between ethical leadership and staff retention. In this study, we find that job resources mediate the relationship between ethical leadership and staff retention. Thus, we find that when leaders are moral people (for instance being people of integrity, exhibiting concern for others, being honest, fair in their actions, truthful) and exercise morality in a range of their scheduled and regular activities like communicating, rewarding, punishing, coordinating, planning among others; they are better equipped to appropriately offer job resources for employees' benefit (Armstrong, 2014;Mullins, 2002). This ultimately heightens their desire to stay with the organisation (Bakker & Demerouti, 2007). These job resources include autonomy, social support from supervisors and challenge. Consequently, job resources can be derived from the organisation (e.g., through the pay-salaries, job security), interpersonal and social relations (such as supervisor and co-worker support), organisation of work (like participation in decision making), and the task (such as autonomy, feedback) (Bakker & Demerouti, 2007). When resources are provided, employees operations are improved and they are motivated to stay longer with the organisation, thereby increasing their retention levels. Job resources generate a motivational mechanism through which the resources motivate employees to heighten commitment to the organisation thereby lowering turnover rates (Bakker & Demerouti, 2007).
The findings of job resources as a significant mediator in the relationship between ethical leadership and staff retention are a response to the calls by Rosenberg (1968) and Friedrich (1982) that a relationship study devoid of addressing a mediating process of a third variable yields inconclusive results. Consequently, this study makes a valuable contribution on how and why ethical leadership influences staff retention in organisations. Since job resource variable has been found as a significant mediator in this study, it is advisable that its importance in the staff retention literature be acknowledged.
Drawing from the LMX theory, this study contributes valuable insights into the practical application of ethical leadership theory in the place of work. Earlier research on ethical leadership focused on identifying theoretical models and describing what constitutes ethics in leadership (Brown & Mitchell, 2010). Lately, the study of this leadership theory has stirred to the fore into the empirical arena of distinguishing whether a relationship between ethical leadership and employee retention is prevalent. Studies (Lee-Kelley et al., 2007;Rai, 2013;Simons & Roberson, 2003;Yener et al., 2012) have found significant associations between ethical leadership and employee retention. The findings from this study support the theoretical assertion that ethical leadership is necessary for employee retention in the setting of the health care sector. We demonstrate that encouraging and practicing ethical leadership has significant benefits for the leader, the employees and the organisation as a whole. Due to limited empirical studies undertaken in this area, especially in Uganda's health care sector, the results of this study emphasise the importance of ethical leadership.
Previous studies have independently showed that ethical leadership predicts staff retention (Lee-Kelley et al., 2007;Rai, 2013;Simons & Roberson, 2003;Yener et al., 2012). Similarly, studies have established that job resources enhance staff retention in organisations (Bakker et al., 2003;Karatepe & Olugbade, 2009). Nonetheless, even when it has been proved that job resources and ethical leadership are important drivers and outcomes of organisation's progress, there is no independent research to date that has demonstrated empirically how these concepts are linked in health sector within Uganda's rural districts. Therefore, the study contributes to the staff retention literature by empirically demonstrating that ethical leadership and job resources are linked in explaining staff retention within the Uganda's health care sector.
Finally, though there is increasing interest in ethical leadership, the majority of research has focused on prosocial outcomes (Brown et al., 2005;Piccolo et al., 2010;Walumbwa & Schaubroeck, 2009). We expand on recent work (Ponnu & Tennakoon, 2009;Mayer et al., 2010) by examining employee retention as an outcome of ethical leadership.
The findings have implications for policy and management. First, findings suggest that leaders need to be ethical in their dealings if they are to enhance the retention function of their staff in their respective organisations. Similarly, though ethical leadership is regularly emphasised by corporate managers as necessary for staff retention, it cannot be separated from job resources. Therefore, health care managers should endeavour to provide job resources in order to mitigate the effects of job overload so as to improve on staff retention in their organisations. Efforts to improve on staff retention must therefore consider effective ethical leadership and job resources.

Limitations and future directions of research
The study was not devoid of limitations. First, this study relied on self-report which could have resulted into common method bias problem. This was worsened by the fact that respondents were asked to respond to more than one construct in the same survey at the same time. This poses a problem in terms of generalisability of the results. Second, data were collected in a cross-sectional survey, which might have confounded the causality among the studied variables (ethical leadership, job resources, and staff retention). Future researchers could conduct longitudinal studies to address this glaring limitation. Third, all participants were from one sector (health care sector) in Uganda. Future studies could collect data from people working in other industries in different contexts to examine the generalisability of our findings. Despite these limitations, policy makers of Uganda and perhaps other developing nations dealing with health care issues and even scholars interested in the field of HR management might find this study insightful.

Conclusions
Ethical leadership and job resources are significant predictors of staff retention in organisations. Whilst the direct relationship between ethical leadership and staff retention was established as significant, when the paths coefficients of the relationship between ethical leadership and staff retention was controlled through the introduction of the mediator (job resources), the original significant path between ethical leadership and staff retention reduced, but the model remained significant-suggesting a partial mediation. This concludes that the entire effect does not only go through the main hypothesised predictor variable (ethical leadership) but also, through job resources. Therefore, the relationship between ethical leadership and staff retention, lowers on  Figure A3. Staff retention measurement model.