RELATIONSHIPS BETWEEN SUPERVISOR TURNOVER INTENTION AND SUBORDINATE PERCEIVED ORGANISATIONAL SUPPORT AND POSITIVE MOOD

—————————— Jamie S. Donsbach was Doctoral Student in the Psychology Department at the University at Albany, USA, and is currently working as Consultant, The Group for Organizational Effectiveness, USA; Linda Rhoades Shanock, Assistant Professor, Psychology and Organizational Science, University of North Carolina at Charlotte, USA. Correspondence concerning this article should be addressed to: Linda Rhoades Shanock, Psychology Department, University of North Carolina at Charlotte, 9201 University City Blvd., Charlotte, NC 28223 USA. E-mail: lshanock@uncc. RELATIONSHIPS BETWEEN SUPERVISOR TURNOVER INTENTION AND SUBORDINATE PERCEIVED ORGANISATIONAL SUPPORT AND POSITIVE MOOD

In recent years, researchers have begun to recognise the utility of studying the relationships between phenomena that exist at different levels of the organisation (Kozlowski & Klein, 2000). Consistent with this trend, organisational researchers have explored how attitudes and behaviours of individuals at one level of the organisation influence the attitudes and behaviours of individuals at other levels. One area of interest has been the influence of supervisor-level attitudes with subordinate-level attitudes and behaviours (e.g., Shanock & Eisenberger, 2006;Tepper & Taylor, 2003).
Researchers in the area of perceived organisational support (POS) recently began studying the relationship between supervisor and subordinate level variables relevant to POS. POS refers to employees' global beliefs concerning the extent to which the organisation values their contributions and cares about their well-being (Eisenberger, Huntington, Hutchison, & Sowa, 1986;). The recent multi-level research in this area is based on organisational support theory (OST: Eisenberger et al., 1986), a derivative of social exchange theory. OST supposes that employees who the organisation could influence subordinates' POS. In the present study, we build on the recent multilevel POS research by examining for the first time the notion that supervisors' turnover intention may contribute to subordinates' POS and by considering how supervisors' intent to quit may contribute to subordinates' positive mood at work through reduced POS.

Supervisor turnover intention and subordinate POS
The theoretical basis for the hypotheses that supervisors' intentions to leave the organisation influence subordinate POS is drawn from the logic of recent trickle down approaches based on social exchange. Like organisational support theory, Masterson's (2001) analysis of social exchange in organisations emphasised the obligation of employees to repay favourable treatment received from the organisation. Masterson suggested that, in the case of service employees, their receipt of favourable treatment from the organisation would have a "trickle down" effect on their treatment of customers. Masterson found that service employees who perceived they were treated fairly responded by treating customers well. Tepper and Taylor (2003) extended this approach to the relationship between supervisors and their subordinates. Tepper and Taylor (2003) argued that supervisors who perceived they were treated fairly by the organisation could reciprocate by treating subordinates more favourably. Accordingly, Tepper and Taylor (2003) reported that supervisors' perception that they had received fair treatment was positively related to their subordinates' ratings of extra-role behaviours exhibited by their supervisors, including help with difficult assignments, showing respect, and help in skill building. These findings agree with the implication of organisational support theory that supervisors who have favourable experiences with the organisation would feel an obligation to repay the organisation, and suggest that such a felt obligation would result in increased support of other employees. Consistent with this view, Shanock and Eisenberger (2006) found that the higher the supervisor's POS, the higher their subordinates PSS, POS, and in-role and extra-role performance. The trickle-down models of Masterson (2001), Tepper and Taylor (2003), and Shanock and Eisenberger (2006) are a good start in providing evidence that supervisors' attitudes and experiences with the organisation may affect the treatment of those below them. Those studies suggest the need for an expanded view of the consequences of supervisor attitudes towards their job and the organisation to include supportive actions towards others not previously considered relevant, such as subordinates. High POS supervisors would meet their increased obligation to help the organisation reach its objectives in part by providing support and resources to aid subordinates (Shanock & Eisenberger, 2006).
The present study builds on this logic by examining the extent to which supervisors' intentions to leave the organisation relates negatively to their subordinates' POS. Individuals with intentions to leave the organisation are often unsatisfied and frustrated with the job or the organisation (Johns, 2001). Although turnover intention has been found to relate to a number of negative outcomes such as reduced performance and extra-role behaviours such as helping other employees (Johns, 2001), to date no study to our knowledge has considered how the turnover intention of supervisors may affect those below them. Supervisors who intend to leave the organisation are likely to no longer feel obligated to help the organisation reach its goals. As such, supervisors may not feel as obligated to provide help, resources and support to subordinates if they intend to leave the organisation as if they were staying. Because supervisor actions tend to be viewed as indicative of the overall organisation's benevolent or malevolent intent (Levinson, 1965), we expect that supervisors intention to leave will have a negative relationship with subordinate POS.
Hypothesis 1: Supervisors' turnover intentions will be negatively related to subordinates' POS.
Subordinates' POS and positive mood Employee mood has been described as a general emotional state on a given day, or the state component of affectivity which is influenced by environmental factors such as one's work environment (George, 1991;George & Brief, 1992). Positive mood is the emotional state described as feelings of enthusiasm, alertness, and excitement (Watson, Clark, & Tellegen, 1988) or a zest for life (Burke, Brief, George, Roberson, & Webster, 1989). Positive mood has been a focus of interest to organisational researchers because it has been found to relate to favourable outcomes such as increased helping behaviour (e.g., Eisenberger et al., 2001) and creative thinking (George & Brief, 1992). Brief and Weiss (2001) and George (2000) suggest that leaders' actions affect how lower-level workers feel, but to our knowledge no study has yet considered how supervisors' own attitudes towards the organisation (e.g., intention to leave the company) may affect the mood of those below them. To date, much of the research in this area has investigated leaders' emotional expression. For example, the charismatic leadership literature demonstrates that leaders who use emotions to appeal to their workers stimulate powerful emotions in their followers (Conger & Kanungo, 1994). In a laboratory study, Lewis (2000) found that followers of leaders who expressed anger reported feeling more nervous and less relaxed than followers of leaders expressing sadness or no emotion. Followers who observed leaders express sadness reported less enthusiasm and greater fatigue than followers who observed leaders expressing anger or no emotion. Like the notion that the emotional expression of leaders affects follower emotions, it may be that leaders also influence follower mood through the help, support and resources they provide which would lead to a positive perception of the organisation overall (higher POS).
Previous research suggests that POS influences employees' affective reactions to their job, including positive mood . In a meta-analysis of the antecedents and consequences of POS, Rhoades and Eisenberger (2002) found a large effect size between employee POS and job related affect (i.e., positive mood). This relationship may occur in part because POS serves the function of fulfilling socio-emotional needs such as esteem, approval, and respect (Eisenberger et al., 1986. Also, POS is thought to contribute to employees' feelings of competence and worth, which in turn enhances positive mood George & Brief, 1992).
Consistent with previous findings, we expect that there will be a positive relationship between subordinate POS and positive mood. In the present study, we also hypothesise that supervisor turnover intention will relate to positive mood indirectly through its relationship with subordinate POS.
Hypothesis 2. Subordinates' POS will be positively related to subordinates' positive mood.
Hypothesis 3. Subordinates' POS will mediate the relationship between supervisors' turnover intentions and positive mood The full hypothesised model is shown in Figure 1.

Sample and procedure
Six hundred nine non-supervisory corrections employees of a state correctional agency in the mid-eastern region of the United States were given the opportunity to complete surveys. The employees voluntarily completed the survey during muster, a 10-15 minute period prior to their shift. Surveys were administered by the correctional agency's Human Resource Department personnel. To encourage candidness, employees were given verbal and written assurances that their individual responses would be confidential and that data would be presented to the organisation in aggregate only. So that subordinate responses could be matched to their supervisor responses, employees were asked to provide their names and their supervisor's names on a detachable portion of the survey. Surveys were returned immediately after they were completed or were returned in sealed envelopes to Human Resources. Human Resource personnel detached the identifying information, assigned identification numbers to each participant, and sent the surveys to the researchers with identification numbers only. Ninety-seven employees returned questionnaires with their own and their supervisor's names for a response rate of 16%.
Like subordinates, the 92 supervisors employed at the organisation were given the opportunity to complete their survey during muster. To help ensure a good supervisor response rate, supervisors identified by subordinate respondents, but who did not initially return a survey, were given a second opportunity to complete a supervisor survey at another muster. Surveys were returned by 44 of the 92 employed supervisors (48%). Supervisor and subordinate surveys were then matched. Fifteen subordinate surveys did not have corresponding supervisor surveys and thus were not included, resulting in a final subordinate sample of 82 (97 minus 15). Similarly, the 10 surveys completed by supervisors whose subordinates did not complete surveys were not included, resulting in a final supervisor sample of 34 (44 minus 10). Thus, once supervisor and subordinate responses were matched, the final sample was 82 employees with 34 corresponding supervisors. The number of supervisors exceeds the criterion of 30 suggested by Hofmann (1997) for level-2 units in hierarchical linear modeling. The average number of employee responses per supervisor was 2.15, ranging from 1 to 13 responses per supervisor.
The participating employees were from three corrections facilities belonging to the same correctional agency. Of the subordinate respondents, 41% were from the first site, 37% were from the second site, and 13% were from the third site. There were no significant differences in subordinate POS or mood across the three sites. Of the supervisors, 44% were from the first site, 35% from the second site, and 21% were from the third site. There were no significant differences in supervisors' turnover intentions across the three sites. On average, the subordinate respondents had been with the organisation for 79.8 months (SD = 50.1) and the mean age was 44.1 years (SD = 10.1). The mean tenure for supervisors was 80.0 months (SD = 38.3) and the mean age was 46.5 years (SD = 7.8). On average, 42% of the subordinates and 18% of the supervisors were female, and 45% of the subordinates and 82% of the supervisors were male.

Measures
Employees at all three facilities completed the same measures. All variables were measured with 5-point Likert-type scale of agreement (1 = strongly disagree; 5 = strongly agree). The anchors for the positive mood scale were slightly different and are described below.

Turnover intention
This two-item scale measures an employee's intention to leave the organisation. The items are from the Michigan Organizational Assessment Questionnaire (Cammann, Fichman, Jenkins, & Klesh, 1983). A sample item is "I will probably look for a new job in the next year".

Perceived organisational support
Prior studies surveying many occupations and organisations provide evidence for the high internal reliability and unidimensionality of the Survey of Perceived Organisational Support (Eisenberger, Fasolo, & Davis-LaMastro, 1990;Eisenberger et al., 1986;. Therefore, Rhoades and Eisenberger (2002) state that POS may be assessed with just a few items from the SPOS, as long as the items encompass the facets of the definition of POS (valuation of employees' contribution and care about employees' well-being). Consistent with this recommendation, we used six items from the SPOS to assess subordinate POS that represented both facets of the definition of POS. Sample items include "My work organisation values my contribution to its well-being" and "My work organisation strongly considers my goals and values" and were the same used by Roch and Shanock (2006).

Positive mood
Adapted from the Brief, Burke, George, and Robinson (1988) measure of mood, this 5-item scale measures how the employee typically feels on an average day at work. For each mood item (e.g., elated, active, enthusiastic) respondents indicated how they typically felt on a 5-point Likert-type scale (1 = very little; 5 = very much).

Descriptive statistics and correlations
Means, standard deviations, internal reliabilities, and intercorrelations among the variables are reported in Table 1. All measures showed high internal reliabilities, with coefficient alphas ranging from .72 to .81. The pattern of correlations was consistent with the hypothesised relationships. That is, supervisors' turnover intentions had a statistically significant positive relationship with the potential mediator, subordinates' perceived organisational support, and with subordinates' positive mood, the outcome variable. Also, subordinates' perceived organisational support had a statistically significant positive relationship with subordinates' positive mood.

Discriminant validity of subordinate constructs
We conducted a factor analysis to provide evidence of the distinctiveness of the two subordinate constructs (POS and positive mood). Given the modest sample size, we conducted an exploratory factor analysis using principal axis factoring and an oblique rotation. The scree plot showed two clear breaks and two eigenvalues above 1.0 for two factors, providing initial evidence that POS and positive mood are best represented as two factors. There were also no problems with cross-loading items, however one POS item (This organisation shows little concern for me, loading = .24) and one positive mood item (elated, loading = .19) loaded below the typically accepted cutoff of .30 (Tabachnick & Fidell, 2001). All other items loaded well above the cutoff on the appropriate factor. Because the items were from established measures, we retained them for the remainder of the analyses. However, the continued suitability of these items may be worthy of examination in future research. One possibility is that the item suitability may depend on the context. For example, it may be that in the current context of a correctional facility, the word 'elated' is not highly relevant to describe the mood of employees.

Relationships of supervisors' turnover intention with subordinates' POS and positive mood
Hierarchical linear modeling (HLM Version 6.0; Raudenbush, Bryk, Cheong, & Congdon, 2004) was used to assess the degree to which supervisors' turnover intention was related to subordinates' POS and mood, and whether subordinates' POS mediated the relationship between supervisors' turnover intention with subordinates' positive mood. Hierarchical linear modeling allows for the appropriate modeling of nested data as was the case in the present study (Hofmann, Griffin, & Gavin, 2000). In the present study, subordinates were nested within supervisors, that is, subordinates within any given work group share the same supervisor and are thus nested rather than independent. HLM accounted for this nesting and allowed us to examine the hypothesised influences of a supervisor-level variable (supervisor turnover intention) on subordinate-level variables (subordinates POS and positive mood) as well as relationships among subordinate-level variables (subordinate POS relationship with positive mood). Table 2 shows the HLM models used to test our predictions. Consistent with Hofmann, Morgeson, and Gerras (2003) and Shanock and Eisenberger (2006), we report a series of models 1) to test our hypotheses and 2) to test for cross-level mediation using the steps that correspond to those required to show mediation in standard hierarchical regression (Kenny, Kashy, & Bolger, 1998). The regression coefficient for each of the relationships tested with HLM is designated by the parameter γ 01 . In support of hypothesis 1, Model 1 demonstrated that supervisors' turnover intentions had a significant negative relationship with the potential mediator, subordinates' perceived organisational support (γ 01 = -.16, p < .05). Model 2 shows support for hypothesis 2, which predicted that subordinates' POS would be positively related to subordinates' mood (γ 10 = .29, p < .01). Because this model is testing two variables at the subordinate level, the regression coefficient for this relationship, as tested by HLM is γ 10.
subordinate POS and positive mood, controlling for turnover intention), and 3) the demonstration of a significant indirect effect. Because step 1 was successfully met in Model 1 above, we continued with the additional steps needed to demonstrate mediation.
To test for POS' potential mediation of the relationships of supervisors' turnover intentions with subordinates' positive mood, we additionally needed to show that the mediator, POS, predicted positive mood, controlling for supervisors' turnover intentions. Thus, POS was added to the model in which supervisors' turnover intensions predicted subordinates' mood (see Model 4 in Table 2). When conducting such cross-level mediation analyses, Hofmann and Gavin (1998) recommend grand-mean centring the potential mediator (here, POS) upon its inclusion in the model. Grand-mean centring the mediator allows an appropriate assessment of the relationship of the mediator with the outcome variable (Hofmann & Gavin, 1998). Therefore, POS was grandmean centred when added to each model. POS significantly predicted mood controlling for supervisors' turnover intentions (γ 10 = .26, p < .05).
The final step of mediation, assessing the significance of the indirect effect of supervisors' turnover intentions on subordinates' mood through POS, was tested using the z-prime method (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002). The z-prime method and Kenny et al.'s (1998) procedure are similar in that they use an identical formula to calculate an indirect effect of the independent variable on the outcome variable through the mediator. However, as demonstrated by MacKinnon et al. (2002), the z-prime method provides more power and a lesser Type 1 error rate than the Kenny et al. (1998) approach. This is because the two approaches differ in the statistical distribution used to determine whether the indirect effect is significant. Specifically, because the estimate of the indirect effect is not normally distributed, Mackinnon et al.'s (2002) z-prime method uses a critical value of .97 for the test of significance, rather than 1.96, for the Z. The indirect effect of turnover intention was significant (Z = -1.65, p < .05). This result supports Hypothesis 3 by demonstrating the mediating role of POS on the relationship of supervisors' job turnover intentions with subordinates' positive mood.

Discussion
In summary, supervisor turnover intention was negatively related to subordinates' POS. It appears that supervisors who plan to leave the organisation may have a negative effect on their subordinates' perceptions of the organisation as a caring and supportive place to work. In addition, subordinates' POS mediated the relationship between supervisor turnover intention and subordinates' positive mood, suggesting that reduced POS experienced by subordinates of a supervisor who intends to leave may result in reduced positive mood at work. Given that positive mood has been shown to result in favourable outcomes such as innovation and helping behaviour (cf , this finding is of practical importance to organisations.

Theoretical contributions
The findings of our study are consistent with recent multilevel research, which reasons that employee attitudes may influence the attitudes of others within the organisation, particularly when these attitudes and behaviours are passed down the organisational hierarchy. For example, "trickle down" models of the transmission of supervisor attitudes to subordinates emphasise social exchange theory (Masterson, 2001;Shanock & Eisenberger, 2006;Tepper & Taylor, 2003). Our study contributes to the literature on trickle down social exchange effects by suggesting that the turnover intentions of supervisors may have an impact on perceptions of organisational support and ultimately the mood of employees at lower levels of the organisation. To our knowledge, research to date has not considered how a supervisor's desire to leave an organisation may affect those under her command. We hope that our study will help encourage social exchange accounts such as organisational support theory to consider what happens in the organisation when a supervisor decides to leave but has not yet left the organisation. It is likely that the supervisor's social exchange relationship with the organisation is weakened, and the supervisor would no longer feel obligated to help the organisation reach its goals, which for supervisors includes providing support to subordinates.
Future research could extend our study with the assessment of felt obligation of the supervisor as an explanatory mechanism for the relationship of turnover intention with subordinate POS, and also consider other implications for the organisation and for subordinates of a supervisor's intention to quit. Future research may also examine moderating effects of the relationship between supervisor turnover intention and subordinate perceptions and behaviours. The extent to which supervisor turnover intention has a negative effect on subordinate outcomes may depend on the reason the supervisor intends to leave. Supervisors who are leaving due to a spousal relocation are likely to act differently during their remaining tenure than supervisors who are leaving because they have negative perceptions of the company.

Practical implications
Understanding the potential influence of supervisor intention to leave on subordinates' perceptions and mood also has practical implications. As demonstrated by previous research, employees who feel supported and are in a heightened mood perform better, are more creative, are more committed to the organisation, and are less likely to leave the organisation (George & Brief, 1992;. Our results suggest that one way to enhance employees' POS and mood is to ensure that employees at supervisory-level positions have a desire to stay with the organisation. Upon finding out that a supervisor plans to leave the organisation, the organisation might want to take special care to make sure that the subordinates of these supervisors continue to receive the resources and support they need to do the job.
Also, succession planning for supervisory positions as part of an overall talent management strategy may help. With a succession plan in place, human resource professionals or other organisational representatives could let subordinates know who their new supervisor will be as soon as possible after a supervisor decides to leave. This may reduce ambiguity for the future and may help mitigate negative effects on mood by serving as an indication that the organisation cares about the subordinates' well-being. Another strategy would be for organisations to head off the potential negative effects of supervisors deciding to leave by discussing with employees the positive aspects of the organisation and the supervisor's reasons for leaving upon hearing that the supervisor intends to leave.
Finally, using regularly scheduled diagnostic tools about supervisors' satisfaction with their job could help organisations identify dissatisfied supervisors and explore alternatives that would alleviate their discontent before the supervisor decides to leave. For example, supervisors who are dissatisfied with factors related to their specific department or position might be transferred to one with a better fit. In this situation, such a strategy would be beneficial for supervisors as well as their subordinates.

Limitations
The primary limitation of this study is the small sample size resulting from a subordinate response rate of 16% and the need to match supervisor and subordinate responses. The supervisors had a much higher response rate (48%), however, the process of matching subordinates to supervisors resulted in the loss of several participants at each level. One implication of the level of usable responses is that we must caution that our results may not be representative of the organisation as a whole. Having a small sample size also limits the statistical power necessary for showing significant relationships if they do exist (Tabachnick & Fidell, 2001). According to a multilevel power analysis (Snijders & Bosker, 1993), to obtain a reasonable power level of .80 we would either need to increase our level 1 (subordinate) sample size from an average of 2 responses per supervisor to an average of 4 responses per supervisor or increase our level 2 (supervisor) sample size to 60 supervisors. However, despite our sample size, we were able to find support for our hypotheses. The only relationship in our model that was not significant was the direct relationship of supervisors' turnover intentions with subordinate positive mood (p = .06). The significance of that relationship is not vital given that we proposed and found support for an indirect relationship of supervisor turnover intention with positive mood.
Another limitation is that our study employed a cross-sectional design. Although the relationship of supervisors' turnover intentions with subordinates' mood was mediated by POS, we can not rule out reverse causality: that subordinates' mood and POS influence the extent to which the supervisors intend to remain with the organisation or that subordinates' mood may influence their ratings of POS. However, previous theory and research on POS and mood and on the relationship of supervisor attitudes with subordinate outcomes is consistent with the direction of our models Shanock & Eisenberger, 2006).
A third limitation is that the POS and positive mood measures both came from the subordinates, which could result in common method variance (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). The relationship between POS and mood could be due more to common response styles used by the subordinates as they completed the measures and less to their standing on the target constructs. Although this is a possibility, previous studies that measure both POS and mood provide evidence to suggest that these constructs are related yet distinct .

Conclusion
In conclusion, our results suggest that supervisors' own turnover intentions may have a negative impact on subordinates' perceptions of the overall organisation (POS) and on subordinates' positive mood on a daily basis at work as mediated by POS. We used organisational support theory to suggest that supervisors who plan to leave the organisation may have a reduced sense of felt obligation to help the organisation reach its goals, with negative implications for subordinates. This theoretical explanation adds to the social exchange perspective employed in recent multilevel research involving supervisors and subordinates which has suggested that supervisors with positive attitudes toward the organisation will feel obligated to treat subordinates better, resulting in better subordinate POS. Subordinate POS and positive mood have the potential to positively influence important organisational outcomes such as commitment, performance, creativity, and citizenship behaviour.