Unravelling Paradoxical Effects of Leader-Rated Performance on Follower Turnover Intention: A Regulatory Focus Perspective

: This paper explores leader-rated performance as one such situational cue that will affect a follower’s promotion focus or prevention focus, which will increase and decrease her/his intention to leave for another organization. The hypotheses were supported with data collected from 133 supervisor-subordinate dyads in a mix of companies in Hong Kong. The ﬁndings demonstrate that the stronger the leader-rated performance, the stronger the promotion and prevention focus of followers. Insofar as people in a promotion focus are eager to seek new options for further advancement, and those in a prevention focus are vigilant to safeguard the current option. The study implies that as promotion-focused followers are preoccupied with better options for advancement, leaders should go beyond coaching and feedback and offer increasingly challenging or innovative tasks that demand individual proactivity and creativity. In so doing, promotion-focused employees are provided with more opportunities to satisfy their eagerness to seek escalating successes within the current organization.


INTRODUCTION
Leaders in the workplace often pay more attention to followers with superior performance as potential talents to be retained for further development (Graen & Uhl-Bien, 1995;Le Blanc & González-Romá, 2012) .To the followers, however, being favored as targets for retention may not correspond with a stronger intention to stay.Indeed, evidence relating leader-rated performance to follower turnover intention has not been consistent.In some studies, the relationship is negative (Lee, Idris, & Tuckey, 2019;Memon et al., 2019) .In other studies, the relationship can be positive.For example, when there is cynicism about the appraisal system (Brown, Kraimer, & Bratton, 2019), perceived unfairness (Islam et al., 2019) or office politics (Imran, Hamid, Aziz, & Wan, 2019).Apparently, the relationship is paradoxical.
Studying employees' turnover intention is important because it is a known predictor of actual turnover (Griffeth, Hom, & Gaertner, 2000).As such, it provides a forward index of the stability of a firm's human resources (Cole & Bruch, 2006).From a human resource management perspective, the turnover intention is a useful control metric that provides forward feedback for managers in their efforts to retain high performers (Lambert, Hogan, & Barton, 2001).Because high performers are targets of retention (Biron & Boon, 2013), scholars have called for more insight into the relationship between performance and turnover intention (Allen & Griffeth, 1999;Birnbaum & Somers, 1993;Kharina, Zulkarnain, & Nauly, 2018).
Earlier, a meta-analysis has noted inconsistent findings of the relationship between leader-rated performance and follower turnover intention and has called for more effort to explore uncharted areas, such as person-based factors (Zimmerman & Darnold, 2009).This paper attempts to address the paradox surrounding this relationship from a person-based perspective: the regulatory focus theory (Higgins, 1997(Higgins, , 1998) ) which is elaborated below.

THEORETICAL FOUNDATION
According to the hedonic principle, people are motivated to approach the presence of pleasure and the absence of pain (Elliot, 1999).The regulatory focus theory (Higgins, 1997(Higgins, , 1998) ) extends the hedonic principle to the domain of goal-directed behaviors.The theory posits that there are two basic motivational orientations.Individuals in a promotion focus are motivated to pursue the pleasurable presence of gains, whereas those in a prevention focus are motivated to secure the absence of losses.
According to the theory, people engage in self-regulation through a promotion or a prevention focus that caters to different motivational needs in the pursuit of individual goals (Higgins & Spiegel, 2004;Spiegel, Grant-Pillow, & Higgins, 2004).In promotion focus, hopes and ideals function like maximal goals, whereas in prevention focus, duties and obligations function like minimal goals (Brendl & Higgins, 1996;Gilang, Fakhri, Pradana, Saragih, & Khairunnisa, 2018).In other words, people in promotion focus seek to maximize the pleasurable end state of advancing success, while those in prevention focus tend to minimize the unpleasurable end state of failing to perform as prescribed.
Extended to goal-directed behaviors, promotion-focused individuals are preoccupied with ideals for advancement, whereas prevention-focused ones are concerned with duty-bound obligations (Liberman, Idson, Camacho, & Higgins, 1999).In goal pursuits at work, individuals with promotion focus tend to approach advancement by changing the current option with a new and perceivably better alternative, whereas those with prevention focus tend to avoid loss of security by staying with the original option (Liberman et al., 1999;Li, 2017;Scholer, Zou, Fujita, Stroessner, & Higgins, 2010).Hence, it is more likely for promotion-focused employees to consider alternative jobs for advancements and prevention-focused employees to safeguard existing jobs for stability.
Regulatory focus is treated as a dispositional trait as well as an induced state.As an induced state, situational cues can activate regulatory focus that, in turn, can affect one's motivation and action in those situations (Brockner & Higgins, 2001), in particular cues conveyed by one's leader who holds position power (Neubert, Wu, & Roberts, 2013).This paper explores leader-rated performance as one such situational cue that will affect a follower's promotion focus or prevention focus, which in turn will increase and decrease her/his intention to leave for another organization.

HYPOTHESES DEVELOPMENT
In the workplace, individuals are sensitive to situational cues that shape their self-regulations and behavioral inclinations (James, James, & Ashe, 1990;Scott & Bruce, 1994), especially cues coming from their leaders who are regarded as role models with position power (Mayer, Kuenzi, Greenbaum, Bardes, & Salvador, 2009;Tierney & Farmer, 2004).Evidence suggests that promotion focus is evoked when attainment of ideal goals is emphasized by situational cues (Forster, Friedman, & Liberman, 2004).
Promotion focus is consistent with the notion of discrepancy production of the gap between ideal self-guides and actual self (Higgins, Roney, Crowe, & Hymes, 1994).In other words, promotion-focused individuals are eager to increase outcome expectancies to pursue higher goals (Higgins et al., 1994).Insofar as leader-rated performance is an important cue to followers, the level of leader-rated performance may influence the strength of individual followers' promotion focus, such that the higher the leader-rated performance, the stronger the promotion focus.H1a: Leader-rated performance is positively related to a follower's promotion focus.
Conversely, in a non-loss prevention focus, a person is motivated to reduce the gap between ought self-guides and actual self (Higgins et al., 1994).Ought self-guides are concerned with the absence of negative outcomes in connection with duties and obligations (Higgins, 1997(Higgins, , 1998)).Prevention-focused individuals are concerned with the completion of prescribed goals.With this concern, they tend to operate with vigilance to complete the current obligation (Higgins et al., 1994).
Extended to work, prevention-focused individuals tend to be vigilant in maintaining a match between prescribed standards and actual performance (Liberman et al., 1999;Scholer et al., 2010).Inasmuch as leaders' cues about performance standards are regarded by followers as ought-to-do goals, the level of leader-rated performance may alter the strength of individual followers' prevention focus, such that the higher the leader-rated performance, the stronger the prevention focus.
H1b: Leader-rated performance is positively related to employees' prevention focus.
Turnover intentions often involve a process of evaluating the current job against viable alternatives to determine which one is more attractive (Mobley, 1977;Steel, 2002).In promotion focus, individuals are motivated to pursue goals of ideals and aspirations, with a preoccupation for advancement and change for better options (Crowe & Higgins, 1997;Higgins, 1997Higgins, , 1998)).
People in a promotion focus are more open to seeking changes and new alternatives (Liberman et al., 1999).Job alternatives are possibilities for advancement that psychologically pull employees away from their current company (Bretz, Boudreau, & Judge, 1994).Given that promotion-focused employees are preoccupied with better alternatives for advancement, they tend to be eager to consider other job opportunities that promise greater potential for growth.Thus, H2a: Promotion focus is positively related to turnover intention.
In contrast, people in a prevention focus are motivated by a need for stability and tend to avoid change by safeguarding the original alternative (Liberman et al., 1999).Hence, employees in a prevention focus are concerned with securing the current job.As such, they should pay less attention to the thought of quitting.H2b: Prevention focus is negatively related to turnover intention.
Insofar as leader-rated performance is positively related to followers' promotion focus and prevention, which in turn increases and decreases turnover intention respectively, it is likely that promotion focus and prevention focus will mediate the positive and negative effects of leader-rated performance on turnover intention of individual followers.H3a: Promotion focus mediates the positive effect of leader-rated performance on turnover intention.
H3b: Prevention focus mediates the negative effect of leader-rated performance on turnover intention.
The hypothesized structural model is depicted below.

METHOD
Data were collected from 133 leader-follower dyads in a mix of companies in Hong Kong.Following the method of temporal separation to minimize spillover bias when filling out questionnaires (Podsakoff, MacKenzie, & Podsakoff, 2012;Podsakoff, MacKenzie, Lee, & Podsakoff, 2003), data for the independent variable, the mediators and the dependent variables were gathered in different weeks.

Measures
The supervisor-rated performance was assessed with three items with the highest factor loadings on a four-item scale (Francesco & Chen, 2004).For measuring regulatory focus, three items for promotion focus and four items for prevention focus that had the highest loadings on their respective factors were adopted from an eighteen-item scale for assessing workplace regulatory focus (Neubert et al., 2008).Turnover Intention was rated with a three-item scale of for scoring how frequently an employee thinks of leaving the current organization (Cammann, Fichman, Jenkins, & Klesh, 1979;Nadiri & Tanova, 2010).All items in the measures were rated on seven-point Likert-type scales.In this study, the Cronbach's alphas of the measures were: Supervisor-rated Performance (α = .92),Promotion Focus (α = .80),Prevention Focus (α = .77),and Turnover Intention (α = .95).Control variables: To preclude spurious explanations due to omitted variables, the author used four control variables.First, in research involving supervisor-rated performance, one of the potential confounders is the time working with one's supervisor (e.g., in (Andrews, Kacmar, & Harris, 2009;Clark, Halbesleben, Lester, & Heintz, 2010)).Thus, Tenure with Supervisor was controlled for.It was measured on a reverse scale with 0 = more than six months, 1 = less than six months and 2 = never.Second, Gender has contributed to a disturbance in the study of regulatory focus and supervisor-rated performance (Bowen, Swim, & Jacobs, 2000;Harris, Kacmar, & Zivnuska, 2007) .It was controlled for and scored in terms of 0 = female and 1 = male.Third, the author controlled for Employment Status in view of its plausible disturbance involving motivation at work (e.g., in (Martin & Hafer, 1995;Maynard, Thorsteinson, & Parfyonova, 2006)).It was scored on a reverse scale with 0 = full time, 1= part-time, 2 = contract basis and 3 = freelance.Lastly, an employee's managerial role may bias effects on turnover intention (Cotton & Tuttle, 1986;Moncarz, Zhao, & Kay, 2008).Hence, the author included Managerial Role as a control variable in the employee-reported data.It was rated with 0 = manage staff and 1 = do not manage staff.In sum, this study controlled for Tenure with Supervisor, Gender, Employment Status, Managerial Role to preclude potential confounding effects.Analytic strategy: This study used the Bayesian method of analysis because of its advantages over frequentist methods in dealing with models involving complex psychometrics with modest sample sizes (Arminger & Muthén, 1998;Rupp, Dey, & Zumbo, 2004).Because Bayesian probability refers directly to the coefficients of parameters rather than to null hypotheses of frequentist methods, it has been favored as a technique for improving the nature of statistical inference (Zyphur & Oswald, 2013) and for analyzing data in organizational research (Kruschke, Aguinis, & Joo, 2012).In this study, the author used the Bayesian estimator in the Mplus 7 program (Muthén & Muthén, 2012), which operates on default settings of an uninformative prior and 30000 iterations in Markov Chain Monte Carlo (MCMC) estimation (Asparouhov & Muthén, 2010).The uninformative prior setting suits the present study because informative priors cannot be assumed due to the inconsistent findings in the existing research.In addition, the MCMC process can handle potential skewness and nonnormality in posterior distributions that often pester the testing of indirect effects in mediation models (Yuan & MacKinnon, 2009).The Bayesian method of analysis has been used by researchers to test mediation models involving cognitive psychology in both empirical and experimental settings (e.g., (Preacher & Selig, 2012;Zhang, Wedel, & Pieters, 2009)).In sum, the Bayesian method was chosen for its practicality for testing indirect effects involving multiple mediators and complex psychometric data based on the modest sample size of the present study.

RESULTS
The sample respondents comprised middle and junior managers who reported to supervisors.There was a fairly balanced mix of gender (55% female and 45% male) and age groups (46% aged 26-35 and 30% aged 36-45).Respondents came from a diversity of industries including finance, management, human resource, technology, engineering, sales and marketing.Thus, the sample represented a broad spectrum of organizational characteristics that would not skew the sample toward any specific gender, age bracket and industry type.
The means, standard deviations and bivariate correlations among the variables in the study are shown below (Table 1).Confirmatory Factor Analysis (CFA) The author followed the two-step analytic approach recommended by SEM scholars (Anderson & Gerbing, 1988).The first step involves testing a measurement model that provides a confirmatory assessment of convergent and discriminant validity of the factors under investigation (Bentler, 1978;Campbell & Fiske, 1959).Based on satisfactory CFA in the first step, the second step proceeds to test a structural model that specifies the hypothesized relationships between constructs (Anderson & Gerbing, 1988).
The author conducted CFA using the Bayesian estimator of Mplus 7 (Muthén & Muthén, 2012) to test whether the indicators loaded on their purported latent variables.The CFA yielded goodness-of-fit (i.e., PP p-value =.06; PSR < 1.05; autocorrelations <.10; parameter estimates within the 95% credibility interval did not contain zero) (Table 2).Also, all indicators loaded onto their underlying factors with statistical significance.Factor correlation estimates ranged from .10 to .52,indicating that the factors were distinct from each other.Hence, the CFA results confirmed the convergent and discriminant validities of the four-factor measurement model (Mplus outputs are available from the author upon request).
The author tested a competing CFA model by loading all items to one factor.Results of the one-factor test revealed model misfit with significant PP p-value (.00) and inferior standard diagnostics (i.e., factor loading parameter estimates within the 95% credibility interval covered zero) (Table 2).DIC was larger than that of the four-factor CFA model (5051.06vs 4415.64),indicating that the four-factor model was the superior model, and so it was adopted for specifying the structural model.

Test of the Structural Model
The structural model specified that there is an indirect effect of supervisor-rated performance on employees' turnover intention, which is positive via promotion focus and negative via prevention focus.The author used the Bayesian estimator available in Mplus 7 to test the hypothesized relationships of the structural model.
Following the recommendation of Antonakis, Bendahan, Jacquart, and Lalive (2010) for checking endogeneity in mediation models, the author conducted a Hausman (1978) endogeneity test using the Wald χ 2 test method available in the Mplus program (Muthén & Muthén, 2012).Constraining the covariances of Promotion Focus-Turnover Intention and Prevention Focus-Turnover Intention to zero, the tests yielded nonsignificant χ 2 p-values for the constrained covariances (Promotion Focus-Turnover Intention p > .05;Prevention Focus-Turnover Intention, p > .05),and in the Wald tests of parameter constraint (Promotion Focus-Turnover Intention =.85; Prevention Focus-Turnover Intention =.39).The test results revealed that the correlations of disturbance between Turnover Intention with Promotion Focus and Prevention Focus were zero, thereby indicating that endogeneity trouble of the two mediators was not significant.In addition, the correlation of disturbance (or covariance) between the independent and the dependent variables was not significant (β = .02,p > .05).In other words, significant spurious effects in the structural relationships were not found.
After controlling for the three instruments and the control variable, the structural model was tested using the Bayesian estimator of Mplus 7. First, the author tested whether Supervisor-rated Performance directly affected Turnover Intention without the mediation of regulatory focus.Testing this direct effect is important because a significant direct effect would diminish the mediating role of regulatory focus and suggest partial mediation.After removing the mediators, the test showed that the direct effect of Supervisor-rated Performance on Turnover Intention was not significant (β = .10,p >.05), indicating the necessity of considering a mediation model (Competing Model One in Table 3).Also, in the event that the mediation effects turned out to be significant, the effect of Supervisor-rated Performance on Turnover Intention would be fully mediated by regulatory focus.
Next, the hypothesized structural mediation model was tested.The test results confirmed goodness-of-fit of the original model (i.e., PP p-value =.08; PSR < 1.05; autocorrelations < .10;parameter estimates within the 95% credibility interval did not contain zero) (Table 3).Specifically, there were positive and significant relationships between Supervisor-rated Performance and Promotion Focus (β =.33, p < .01),Supervisor-rated Performance and Prevention Focus (β =.19, p < .01),Promotion Focus and Turnover Intention (β =.80, p < .01),Prevention Focus and Turnover Intention (β = -.91,p < .01),thus supporting H1a, 1b, 2a and 2b.The indirect effect of Supervisor-rated Performance on Turnover Intention via Promotion Focus (Hypothesis 3a) was significant and positive (β =.26, p < .01),and via Prevention Focus (H3b) was significant and negative (β = -.16,p < .01).Because the direct effect was not significant, the results demonstrated that the impact of Supervisor-rated Performance on Turnover Intention was fully mediated by regulatory focus.In essence, all the research hypotheses in the structural model were supported by the observed data.

Test of Competing Models
The author tested setting Turnover Intention as the independent variable and Supervisor-rated Performance as the dependent variable.The test is pertinent because it addressed an important question about reverse causality, viz., in some studies, turnover intention affected performance (Carraher & Buckley, 2008;Hui, Wong, & Tjosvold, 2007).The competing model yielded a larger DIC than that of the original model (4375.88vs 4373.55) with parameter estimates within the 95% credibility interval covering zero.(Competing Model Two in Table 3).Thus, the original structural model was superior to the second competing model.
The next competing model examined whether the removal of the three instruments and the control variable affected the model's goodness-of-fit by checking the DIC.The DIC is an index introduced by Spiegelhalter et al. (2002) that identifies a better-fitting model with a lower DIC based on the principle of parsimony (i.e., if removing a free parameter does not improve model fitness with a lower DIC, the free parameter should not be removed).After deleting the instrumental and control variables from the model, test results yielded a higher DIC than that of the original model (4384.02 vs 4373.55).The removal of the instrumental and control variables from the competing model did not increase model fitness.Thus, the original model showed a better fit by including the instrumental and control variables for precluding confounding effects.In sum, the original structural model satisfied the prerequisites for model goodness-of-fit as assessed by the standard diagnostics of the Bayesian estimator.In addition, it exhibited superior fit indices compared to the three competing models.All structural relationships were significant, thereby supporting the research hypotheses, as shown in the observed structural model below (Figure 2).For cross-referencing, the scatterplot for checking the original model's PP p-value and posterior distributions of parameter estimates for the indirect effects are also given underneath (Figures 3 and 4).This study examined, on a leader-follower dyadic level, the effects of leader-rated performance on follower turnover intention through the mediation of follower promotion focus and prevention focus.As hypothesized, leader-rated performance is positively related to follower promotion focus and prevention focus, which in turn increased and decreased turnover intention respectively, thus suggesting an apparent paradox.
In the workplace, leaders play a key role in identifying superior performers as talents to be retained for development.According to the relationship-based leader-member exchange theory (Graen & Uhl-Bien, 1995), leaders favor high performers as "in-group" and devote more time and effort to develop them.They spend more time with "in-group" followers in order to continue to groom them through feedback, coaching and support (Graen & Uhl-Bien, 1995).
However, more time and effort given to favored followers does not always equal their less intention to quit.Other person-based factors such as a follower's affective commitment to the current organization or willingness to follow one's leader to another organization (Becker, Ertz, & Buettgen, 2019), also matter for followers' intention to quit for another job.Insofar as person-based factors play a role, a relationship-based view per se is not sufficient for explaining why favored "in-group" followers intend to quit.This study contributes by adopting a person-based approach to unravel the paradoxical relationship between leader-rated performance and follower turnover intention.Underpinned by the regulatory focus theory (Higgins, 1997(Higgins, , 1998)), findings in this study indicate that it is the follower's regulatory focus that explains the positive and negative effects of leader-rated performance on turnover intention.
As this study demonstrates, the higher the leader-rated performance, the stronger are the promotion and prevention focus of followers.Insofar as people in a promotion focus are eager to seek new options for further advancement, and those in a prevention focus are vigilant to safeguard the current option (Liberman et al., 1999;Scholer et al., 2010), a stronger promotion focus and prevention focus will increase and decrease turnover intention respectively.In essence, the person-based perspective of regulatory focus theory sheds more light on the paradoxical relationship.

IMPLICATIONS FOR MANAGERS AND RESEARCHERS
In the workplace, followers often appraise a situation based on the words and behaviors of leaders for deciding what to do next (Tierney & Farmer, 2004).Situational cues from leaders are particularly important for followers because they signal position power and organizational endorsement (Wallace & Chen, 2006).One such cue is leader-rated performance, the substance of which often manifests itself to followers in the way leaders provide day-to-day coaching and feedback (Graen & Uhl-Bien, 1995).In order to cater to followers' motivational needs based on their regulatory focus, leaders should pay more attention to the potential repercussions of evoking followers' regulatory focus.
As promotion-focused followers are preoccupied with better options for advancement, leaders should go beyond coaching and feedback and offer increasingly challenging or innovative tasks that demand individual proactivity and creativity.In so doing, promotion-focused employees are provided with more opportunities to satisfy their eagerness to seek escalating successes within the current organization.
For prevention-focused followers, positive feedback and reward for performance may satisfy motivation to complete current job responsibilities.Inasmuch as they are prone to operate with vigilance to avert failure and mistakes, tasks that require efficiency and accuracy should provide more person-job fit.
For future research, the regulatory focus perspective may be extended to other work-based factors affecting turnover intention.Leadership styles, job crafting, work demands and job-person fit are some examples of low hanging fruits for researchers.

LIMITATIONS OF THIS STUDY
The present study has its limitations.First, the study is cross-sectional in nature and can only answer the research question based on a snapshot of circumstances in the workplace.As such, it lacks the breadth of monitoring the hypothesized structural model over time.Second, the study presumes that leaders, in general, treat followers with superior performance favorably as their "in-group" members.It does not consider situations where leaders are at odds with better-performing followers due to career-path conflict or personal value incongruence.Lastly, in structural equation modelling, there may be more than one model that fits the observed data.While steps have been taken to test competing models to identify a model with a better fit, the result does not exhaust the possibility of other fitting models.

Figure 1
Figure 1 Self-Contradicting Effects of Performance on Turnover Intention

Figure 2 Figure 3
Figure 2 Structural Equation Modeling Results

Figure 4
Figure 4 Bayesian Posterior Parameters Distribution of Specific Indirect Effect

Table 1
MEANS, STANDARD DEVIATIONS, AND CORRELATIONS OF STUDY VARIABLES (N = 133 DYADS)