Perceived Supervisor Support and Turnover Intention: A Meta-Analysis Study

Employees intending to resign are commonly due to material matters, such as lack of salary or incentives. Evidently, moral matters, e.g., supervisor support, implies to affect the turnover intention in an organization. Studies pertaining the effect of these two variables to intention to resign report mixed result. For example, several studies report direct and significant effect of these two variables to intention to resign, while otherstudies report the contrary. This study aimed to investigate the direct influence of perceived supervisor support on employee turnover intention through a meta-analysis method. There are 17 articles and 20 studies, ranging from 2007 to 2019 being analysed, of which 60% were published in accredited international journals, and the remaining 40% were published in international journals. Aggregate number of respondents of the studies amounted to 10,079 people from all industrial sectors, such as manufacturing, healthcare, hospitality industry, education, and social services in several countries, such as the Philippines, China, India, Pakistan, Turkey, Norway, and the United States. This study employed data from the standardized beta coefficient using an open-source program data analysis application of Mplus. The study results found that perceived supervisor support did not directly affect employee turnover intention. The study can be used as a reference for similar studies to compare other interrelated variables.


Introduction
Employee turnover intention has been receiving attention in the recent years (Chen, et al., 2012;Yousaf, 2008;Afzal et al., 2019;Presbitero & Teng-Calleja, 2019;Price, 2001). Studies regarding employees' turnover intention have found that lack of salary as a major factor, and therefore, employees tend to pursue other places that offer higher income. Afiveyearstudy by Society for Human Resource Management or SHRM (2019) discovered that 58% of the participants chose to resign due to influence of management policy and managerial role. Pfeffer and Sutto (2006) have previously found similar finding and they also observed that manager and organization's policy were vital to prevent employees' turnover.
Employee turnover intention demands the company to ensure, and further facilitate, employee ownership and integration in the company in order to maintain organizational performance (Collins & Smith, 2006). According to a study by Trevor and Nyberg (2008), a relatively high turnover rate negatively affected company performance, such as reducing performance quality, consistency, and service stability. It aligns with a study by Daghfous et al. (2013), that also found turnover intention highly affected the company performance and profit.
Scholars such as Carmeli and Weisberg (2006) and Addae et al. (2006) had attempted to define turnover intention. Carmeli and Weisberg (2006) explained three elements of turnover intention: thoughts of quitting, the intention to search for another job elsewhere and the intention to quit, which is built upon previous studies such as from Moble (1977) and Tett and Meyer (1993) who argued that employee's intention to resign involved a cognitive process is a deliberate decision with predetermined reasons. In several studies, resign decision directly affected turnover decision (Addae et al., 2006;Bedeian et al., 1991).
Understanding turnover intention is a critical to organization due to its potential risk to performance and profit (Addae et al., 2006). A relatively high turnover rate negatively impacts the company, causing imbalance and uncertainty in the working environment. Sunyoto (2012) mentioned that the desire to change jobs might be caused by dissatisfaction in the current job. Dissatisfied employees will foster their intention to resign. Work satisfaction will push employees to work better and stay in the company. This argument is supported by Robbins and Timothy (2015), asserting that low work satisfaction is the predictor of employee resignation. To avoid such impact, Low et al. (2001) argued that resign decision's predictors should be mitigated.
Precedent studies had tried to examine antecedents to discover things affecting employee turnover intention, such as those of Hancock et al. (2013) and Tracey andHinkin (2008). De Clercq et al. (2016 found that perceived supervisor's support, a level received by members from their superiors who consider their well-being, have contributed to major emotional and psychological power to employee to handle work-related stress (Hoonakker et al., 2013) and will effectively increase teamwork to be more solid (Astuti et al., 2020). Furthermore, some studies further point a negative association between perceived supervisor's support to turnover intention (Kuvaas & Dysvik, 2010) and actual turnover (Eisenberger et al., 2002;Kammeyer-Mueller & Wanberg, 2003). Ertürk (2014) argued that the interaction between perceived supervisor's support and the exchange relation quality between supervisors and subordinates might predict employee's intention to resign.
Perceived supervisor support can be described as subordinates' general view regarding the extent of supervisors appreciating their performance and caring for their overall well-being (Eisenberger et al., 2002). Supervisor's support was found by Joo (2010) to enhance higher organizational commitment, through facilitating employee's learning culture. Further, the unique contribution from perceived supervisor's support, regardless of perceived organizational support and human resource practice perception, was found by Kuvaas and Dysvik (2010). These authors concluded that managerial support was crucial in shaping employee attitudes, affecting their commitment and turnover intention. Maertz et al. (2007), learned that supervisor's support perceived by employees had an independent effect on turnover intention and was not mediated through organizational support.
Despite supervisor's support being reported in scholarly articles as effective in reducing turnover intention, however the level of effectiveness across studies has not been investigated. Such limitation hampered a more elaboration to emphasize the crucial role of adopting a strategy in providing more perceived superior's support to reduce turnover intention in working places (Krishnan & Singh, 2010). As such, this study is intended to fill the gap by examining in-deptheffect of perceived supervisor's support on turnover intention, using a meta-analysis method. According to Koricheva and Gurevitch (2013), has been widely used in over 20 studies and has contributed to synthesize a new understanding of a particular topic. As such, the method is relevant to the objective of this study.

Literature Search Stages
The literature search was carried out for approximately five months, starting on September 2, 2019. There are x stages for literature selection. First, the researcher defined searched criteria such as publication year, keywords, and database. The publication year is limited to 2007 -2019, because research that examined the effect of supervisor support on turnover intention started from 2007 and ended with the article search process in 2019. The researchers then used keywords search such as "perceived supervisor's support" and "turnover intention". Literature was identified from www.lib.ugm.ac.id page on scientific article directories such as JSTOR, Emerald Insight, SAGE Journals, and Science Direct. There two selection criteria were: study conducted on perceived supervisor's support and turnover intention that includes an assessment tool and produce data on the effect of perceived supervisor support and turnover intention. The researchers entered data on the number of subjects, gender, gender ratio, standardized beta coefficient, and tvalue. Based on the article search results, 17 articles were obtained that are relevant to 20 studies from 2007 to 2019. This study used a sample of 17 articles with 20 studies. The number of articles published in 2007-2010 was three, six articles in 2011-2014, and eight articles in 2015-2019. 70% of the sample articles used have been published in international journals indexed by Scopus, and 30% are articles published in international journals. The following a summary of the articles used. In the variable perceived supervisor support, the scale included the superior's concern for its members' welfare. Some of the scales used were adapted from the scale of Eisenberger et al. (1986), consisting of four question items and a scale from Tate, Whatley, and Clugston (1997), consisting of three question items. An example of questions asked was my boss cares about the opinion I put forward. Besides, in the turnover intention variable, the scale used was adapted from the Dysvik and Kuvaas (2010) scale, which consists of five question items, the scale from Bond et al. (1998) and Qiu (2010), which consists of one question item, and the Mobley et al. (1978) with a three-item scale. An example of a question given to a respondent was how often employees think about quitting their job. The total number of subjects from the entire sample of articles was 10,079 people. All sample articles obtained were then examined to obtain standardized beta coefficient value data for each article. The following are sources of literature in this study.

Data Analysis Procedure
The researchers used standardized beta coefficients to measure metric sizes in a metaanalysis (Farley, et al., 1995;Rosenthal & DiMatteo, 2001). Furthermore, Peterson and Baron (2005) saw a phenomenon in which many meta-analysis studies used correlation coefficients and very few meta-analysis studies used regression coefficients, although there have been enough studies that describe its benefits. The definition of standardized beta coefficients compares the strength of influence on the independent variable against the dependent variable (Freedman, 2009). The following is the equation.
β1-β20 x y β1 = study 1...20 In the data analysis homogeneity test, the open-source program Mplus was used. Furthermore, the heterogeneity was tested using the following equation.

= [ − ] × 100%
In general, the percentage of about 75% shows high heterogeneity, 50% shows moderate heterogeneity, and 25% shows low heterogeneity. A score of 0% of I 2 indicates no heterogeneity, whereas a higher I 2 score indicates a greater degree of heterogeneity. This study used the Fixed-Effect Model (FEM), in which the study was assumed to be homogeneous, meaning that there are no differences in both theory and dimensions in both the independent and dependent variables (Hedges & Olkin, 1985).
In addition to the heterogeneity test, the next data analysis was to look at the beta coefficient data characteristics using the Mplus program. The following is a table of the characteristics of the beta coefficient of 17 articles from 2007 to 2019, which includes 20 studies.  Table 3 shows the characteristics of the 20 studies that constitute the study samples using the regression model. The first characteristic obtained was the range of standardized beta coefficient values moving from -0.05 to 0.48, with the average value obtained for all articles of 0.499. Meanwhile, the median calculation result obtained is 0.457. The variance value obtained in this study was 0.023, while the skewness and kurtosis values were 0.527 and -1.110.  (2019), which in their study obtained a beta coefficient of 0.09 with a t value of 1.03 (p> 1.96).
Several previous studies obtained beta coefficient values under 0.6, for example, in a study by Nichols et al. (2016) with 270 subjects. This study obtained a beta coefficient of -0.60. This study also had the highest beta coefficient value of 20 studies used as study samples and negatively impacted. Others, the acquisition coefficient value obtained the beta coefficient value below 0.5.
Several other studies also obtained beta coefficient values that tended to be small, below or less than 0.1, for example, as in the study conducted by Presbytero and Calle (2019), this study was conducted on 338 subjects with a beta coefficient value of 0.09 and t value of 1.03. This study presented a minimal and insignificant result from the variable perceived supervisor support for turnover intention. A similar study that obtained a beta coefficient value of less than 0.1 is a study conducted by Afzal et al. (2019), which involved 304 subjects. This study obtained a beta coefficient value of -0.08. It indicates that perceived supervisor support is very weak in influencing employee turnover intention.
Another indicator to see the weakness of perceived supervisor support in influencing turnover intention is to look at the t value of each sample study. According to Weston and Gore (2006), the value of t should meet the criteria, p> 1.96. Table 1 shows that there were study results with t values under 1.96 (Gordon et al., 2018;Presbitero & Teng-Calleja, 2019;Tymon & Stumpf, 2011).
The absence of a direct impact of the two variables proposed in this study can be explained further. For example, because supervisors themselves need direct organizational support to facilitate them to meet the needs of their members. Starting from this, it indicates that perceived supervisor support will indirectly affect turnover intention.
Supervisor support must go through other variables in order to influence employee turnover intention. Such as the study results obtained by Maertz et al. (2007), where to reduce employees' turnover intention, it was necessary to have perceived supervisor support through organizational support and normative commitment. The same thing was also found in a study conducted by Kuvaas and Dysvik (2010). In shaping employee attitudes, perceived supervisor support affected employee commitment, which affected the intention to stop working. Another study that supports the indirect effect of perceived supervisor support is the study results of Kurtessis et al. (2017) on organizational support theory. Perceived supervisor support should reduce turnover intention if it increases perceived organizational support. Supervisors are seen as agents of the organization and ensure employees get support so that supervisors are seen as contributing to organizational support. Perceived organizational support that is produced because of the role of perceived supervisor support will strengthen the sense of responsibility of each individual in the organization to help companies achieve goals and foster affective organizational commitment Eisenberger et al., 1986;Shore & Shore, 1995).

Conclusion
The meta-analysis study in this study found no direct influence between perceived supervisor support on employee turnover intention. However, in several other studies, supervisor support has an indirect effect on turnover intention. This means that other variables are still needed as mediators or moderators in influencing turnover intentions. This study also found that majority of the included studies were carried out in western countries, which might explain influence of cultural differences in viewing perceived supervisor support in countries where the society is more collectivist, which is more concerned with group goals than personal goals. For example, in collectivist societies like Indonesia, supervisor support can be a source of strength to make employees more committed. This can be explained because people with collectivist culture such as Indonesia perceive leaders as a figure who can help achieve their goals. Therefore, if supervisors can build the trust of their subordinates, their subordinates will believe that their leader can help achieve good performance.
The study results inform consideration for further research to explorein-depth influence of supervisor's support to employee turnover. Particular attention should be given to understand contextual influence of community culture that may affect these aspecets of study. In addition, researchers can also consider other research variables such as moderating variables that may influence dynamic interaction between perceived supervisor support and employee turnover intention.

Funding
The author did not receive funding for the study, manuscript writing, and publication of this article.

Author's Contribution
TA and AFH discussed and agreed on using meta-analysis approach for this study. TA collected the article samples needed to be discussed with AFH. TA then wrote the manuscript with direct guidance and supervision of AFH.

Conflict of Interests
The author declares no conflict of interest in this study.