Perceived Organizational Support Meta Analysis 1

575554 research-article2015 JOMXXX10.1177/0149206315575554Journal of ManagementKurtessis et al. / Perceived Organizational Support Perceived Organizational Support: A Meta-Analytic Evaluation of Organizational Support Theory Organizational support theory (OST) proposes that employees form a generalized perception concerning the extent to which the organization values their contributions and cares about their well-being (perceived organizational support, or POS). Based on hypotheses involving social exchange, attribution, and self-enhancement, we carried out a meta-analytic assessment of OST using results from 558 studies. OST was generally successful in its predictions concerning both the antecedents of POS (leadership, employee–organization context, human resource practices, and working conditions) and its consequences (employee’s orientation toward the organization and work, employee performance, and well-being). Notably, OST successfully predicted the relative magnitudes of different relationships, influences of process variables, and mediational effects. General implications of the findings for OST and research on POS are discussed. 1854 1855 Keywords: perceived organizational support; social exchange, attributions; self-enhancement; socio-emotional needs; organizational commitment; psychological well-being According to organizational support theory (OST; Eisenberger, Huntington, Hutchison, & Sowa, 1986; Eisenberger & Stinglhamber, 2011; Shore & Shore, 1995), employees develop a general perception concerning the extent to which the organization values their contributions and cares about their well-being (perceived organizational support, or POS). OST has attracted considerable interest because of the potential value of viewing the employee– organization relationship from the employees’ viewpoint, the clarity of the POS construct, and the strong associations of POS with affective organizational commitment, job satisfaction, and other attitudinal outcomes. Rhoades and Eisenberger’s (2002) meta-analytic review indicated that POS is related to the major hypothesized antecedents of POS (fairness, human resource [HR] practices, and supervisor support), attitudinal consequences (e.g., affective organizational commitment, job satisfaction), and job performance. A more recent metaanalysis by Riggle, Edmonson, and Hansen (2009) confirmed the attitudinal outcome findings with more studies. However, no quantitative analysis has addressed many important recent POS findings on such topics as leadership, organizational context, positive orientation toward the organization, and employee well-being. Using OST, we organize and theoretically integrate the POS literature to clarify the role of POS in employee–organization relationships. OST, emphasizing social exchange, attribution, and self-enhancement, has been considerably elaborated in the past few years (e.g., Aselage & Eisenberger, 2009; Baran, Shanock, & Miller, 2012; Eisenberger & Stinglhamber, 2011; Wayne et al., 2009), providing the opportunity for the present theory-driven meta-analysis of the POS literature based on OST. In the following section, we describe OST, emphasizing the major propositions assessed in the meta-analysis. This is followed by our meta-analytic methods, hypotheses, and findings. Organizational Support Theory Assessed According to OST, POS strongly depends on employees’ attributions concerning the organization’s intent behind their receipt of favorable or unfavorable treatment. In turn, POS initiates a social exchange process wherein employees feel obligated to help the organization achieve its goals and objectives and expect that increased efforts on the organization’s behalf will lead to greater rewards. POS also fulfills socioemotional needs, resulting in greater identification and commitment to the organization, an increased desire to help the organization succeed, and greater psychological well-being. Employee Attributions According to OST, POS should be enhanced to the degree that employees attribute favorable treatment received from the organization to positive regard. The organization’s discretion when providing favorable treatment, as opposed to such extraneous factors as a tight job 1856   Journal of Management / July 2017 market or government regulations, should enhance POS (Eisenberger, Cummings, Armeli, & Lynch, 1997). Similarly, providing benefits that employees make use of, and therefore are specific to employee needs (Gouldner, 1960), should increase POS. Social Exchange OST invokes social exchange theory wherein employment is viewed as the trade of effort and loyalty by the employee for tangible benefits and social resources from the organization (Cropanzano & Mitchell, 2005). POS should elicit the norm of reciprocity, leading to a felt obligation to help the organization, as well as the expectation that increased performance on behalf of the organization will be noticed and rewarded. As a result, employees with high POS should engage in greater job-related efforts, resulting in enhanced in-role job performance and extrarole performance helpful to the organization. With regard to affective organizational commitment, employees seek balance in their relationship with the organization by developing favorable attitudes and behaviors consistent with POS. Thus, felt obligation resulting from POS has been found to be positively related to affective organizational commitment (Eisenberger, Armeli, Rexwinkel, Lynch, & Rhoades, 2001). Self-Enhancement OST is often mischaracterized as predominantly a social exchange theory. However, OST also emphasizes self-enhancement processes. POS is assumed to fulfill socioemotional needs (approval, esteem, affiliation, and emotional support), leading to identification with the organization. OST holds that affective organizational commitment arises both from self-enhancement and social exchange. Concerning self-enhancement, the organizational identification arising from POS may lead to affective organizational commitment through the development of shared values (Meyer, Becker, & Van Dick, 2006) and the promotion of stronger relational bonds between employees and organizational representatives. In this meta-analysis, we examine 31 hypotheses, derived from OST, as related to attribution, social exchange, and self-enhancement. Because there are many hypotheses, we have combined the presentation of the hypotheses, their rationale, and empirical findings into a single Hypotheses and Results section that follows the methodology description. Method Reviews, such as Kepes, McDaniel, Brannick, and Banks (2013), and guidelines, such as the Meta-Analysis Reporting Standards (MARS; American Psychological Association, 2008), provide detailed discussions of the methodological and procedural information that should be reported to ensure that meta-analyses are transparent and replicable. We have used these recommendations to guide the information and level of detail provided in this section. Literature Search and Inclusion Criteria We identified relevant published and unpublished studies (e.g., dissertations, theses, technical reports) through systematic searches (Rothstein, 2012) using the databases ABI-Inform,    1857 Academic Search Complete, APA PsycNET, Business Source Complete, Defense Technical Information Center, Digital Dissertations, Educational Administration Abstracts, Google Scholar, JSTOR, ProQuest Research Library, Psychology and Behavioral Sciences Collection, Science Direct, and Social Science Full Text. We used the following key terms in our search: perceived organizational support, organizational support, perceived support, and POS. We also used the common British Commonwealth spelling organisational. These searches identified studies from 1986, when the POS construct was introduced by Eisenberger and his colleagues, through 2011. Second, we searched previous meta-analyses by Rhoades and Eisenberger (2002) and Riggle et al. (2009). Third, we collected studies that cited, or were cited by, one of several major source articles on POS (Eisenberger et al., 1986, 1997; Eisenberger, Fasolo, & Davis-LaMastro, 1990; Rhoades & Eisenberger, 2002). Fourth, we attempted to identify and collect additional unpublished data and manuscripts by contacting leading researchers in this area. Using these search methods, we collected a total of 743 documents. Studies were identified for inclusion by examining their abstract, correlation matrix, study method and procedure, and full text of each study, with the following exclusions. First, we excluded studies that failed to measure POS based on Eisenberger et al.’s (1986) definition, such as the use of the initials POS for other constructs. Second, studies were excluded for applying POS to specific types of support (e.g., POS for creativity; Zhou & George, 2001) that are distinct from global POS (Eisenberger & Stinglhamber, 2011). Third, we excluded studies using combinations of POS with other measures (e.g., supervisor support and engagement; Ward, 2006). Fourth, we excluded studies failing to provide adequate statistical information or clearly identify the correlate of POS. Fifth, we excluded data from unpublished studies (e.g., dissertations) that were later published using the same data. However, when the published version of a study and the unpublished version of the same study reported results for different variables, we included the unique results from each manuscript in our metaanalysis. Finally, we excluded foreign-language studies where an English translation was not available. We then narrowed the database to studies containing variables relevant to our hypotheses. Based on these exclusion criteria, our final database included 492 papers containing 558 studies (see online Appendix C for information about each study and online Appendix D for a list of references). Coding Procedure All coders had a background in psychology (i.e., were psychology students or held a degree in psychology) and were provided with training, written coding instructions, and examples of accurate coding. The coders practiced until they coded without error. Each study was first independently evaluated for inclusion by two coders based on the criteria noted above. Relevant studies were then independently coded by two members of the research team, and coding was compared for disagreements. Using a subset of 50 articles, coder agreement was calculated to be 97.6%. Disagreements about inclusion of a study or specific coding were resolved by referring to the original source or other relevant literature (e.g., scale validation studies), and discrepancies were discussed until a consensus was achieved. We categorized correlates of POS based on information provided in the studies or, when necessary, the scale source. For performance-related constructs, we distinguished between self- and supervisor ratings. 1858   Journal of Management / July 2017 Outlier Search We searched for outliers in each meta-analytic distribution using a two-step approach. First, we computed the sample-adjusted meta-analytic deviancy statistic (SAMD; Beal, Corey, & Dunlap, 2002; Huffcutt & Arthur, 1995) and used a scree plot to visually search for outliers. Second, because the SAMD statistic may overidentify outliers (Beal et al., 2002), we examined whether the effect sizes identified as outliers were three or more standard deviations from the mean construct correlation. Applying these criteria, we found only two potential outliers. These involved the relationship between POS and organizational citizenship behavior (OCB) directed toward individuals (OCB-I). Excluding these outliers made little difference, ρ = .19 and .20, with and without outliers, respectively. Because caution has been urged in removing outliers (e.g., Cortina, 2003) and these two did not substantially impact results, we retained them. Data Analysis We used the random-effects meta-analytic procedures and formulas described by Hunter and Schmidt (2004). At least three studies were required to be considered sufficient to provide data for a hypothesis. Using the zero-order effect sizes (in the form of a correlation) and the study sample size, we computed sample-size weighted mean correlations. If a study reported effect sizes for multiple independent samples, all of the relevant correlations were included as separate effect sizes, and if a study reported effect sizes for multiple dependent samples, a single mean correlation and reliability coefficient was computed (Hunter & Schmidt, 2004). The overall mean correlations were corrected for unreliability using full artifact distributions as described by Hunter and Schmidt (2004). Our analyses corrected for sampling error, but we did not correct for range restriction because the population variance of the variables was unknown. For each effect size, the coefficient alpha was coded as the reliability for POS and all other constructs, and when no coefficient alpha was reported or a single-item measure was used in a study, no reliability estimate from that study was included in the artifact distribution. We also conducted our analyses on a subset of variables with the largest number of studies using partial artifact distributions. Because the results for our analyses using full and partial artifact distributions were nearly identical, we report only the results from the full artifact analysis. Using the formulas provided by Hunter and Schmidt (2004), we computed two tests of effect size heterogeneity. First, we computed the percentage of variability in the effect sizes due to sampling error. Second, we computed the 95% credibility intervals around the corrected effect sizes. Last, we computed the 95% confidence intervals around each population effect size using the standard error of the weighted mean observed correlation. These values represent the extent to which sampling error remains in the effect size estimate. Hypotheses and Results Here we present hypotheses and results for proposed antecedents and outcomes of POS (see listing in Table 1). Although many of the estimated effect sizes are based on large numbers of studies, this was not always the case. The number of studies used to assess a given relationship ranged from 237 (affective organizational commitment) to 3 (e.g., reward 1859    Table 1 Organization of Variables Included in the Meta-Analysis Category Antecedents Treatment by organization members Employee–organization relationship quality Human resource practices and job conditions Outcomes Orientation toward the organization and work Subjective well-being Behavioral outcomes Variables Abusive supervision Consideration and initiating structure Leader–member exchange Supervisor and coworker supportiveness Transformational and transactional leadership Fairness Perceived organizational politics Psychological contracts Value congruence with the organization Job security, flexible and family supportive work practices, and developmental opportunities Work role characteristics Affective organizational commitment Economic and social exchange with the organization Felt obligation and normative commitment Job involvement Organizational identification Performance–reward expectancy Trust Positive psychological wellbeing Negative psychological wellbeing In-role performance Organizational citizenship behavior Counterproductive work behavior Withdrawal expectancy). Because estimates based on few studies are less robust than estimates based on large numbers of studies, we urge readers to examine the sample sizes reported in the tables and consider the more provisional nature of findings that are based on smaller samples. We encourage the reader to interpret estimates in light of credibility intervals, which indicate the range of effect sizes within which 95% of studies would be expected to fall (with narrower credibility intervals reflecting less heterogeneity) and the percentage of variance explained by sampling error and other statistical artifacts. When the percentage of variance explained is small or the credibility intervals are wide, estimates may not be robust and substantive moderators may be present. For a review, we direct the reader to Kepes et al. (2013). Because this is a theory-driven review of OST, we focus on partially corrected effect size estimates, as such estimates control for error and most accurately represent the theoretical constructs (Hunter & Schmidt, 2004), although our results tables also present the uncorrected estimates. Relationships are reported as statistically significant when the high and low confidence limits do not include zero. We conclude that effect size is larger than another effect size when the confidence limits of the two effect sizes do not overlap. To interpret effect size, we use Cohen’s effect size rules of thumb (e.g., Cohen, 1992; cf. Bosco, Aguinis, Singh, Field, & Pierce, in press) that a correlation of .50 is considered strong, .30 is moderate, and .10 is small. 1860   Journal of Management / July 2017 Table 2 Meta-Analytic Results for Treatment by Organization Members Variable Supervisor support Coworker support Team support Abusive supervision Consideration (leadership) Initiating structure (leadership) Transformational leadership Transactional leadership Leader–member exchange r .53 .40 .31 –.30 .41 .29 .51 .18 .47 95% CI [.52, .54] [.38, .41] [.26, .35] [–.37, –.23] [.37, .45] [.24, .34] [.47, .55] [.08, .27] [.46, .48] ρ 95% CrI k N % Variance Explained .60 .47 .34 –.34 .46 .33 .56 .21 .53 [.36, .84] [.17, .77] [–.14, .82] [–.34, –.34] [.29, .63] [.06, .60] [.37, .76] [–.08, .51] [.27, .78] 64 16 8 3 4 4 8 3 59 22,371 12,773 1,626 662 1,433 1,433 1,353 427 16,481 10.7 4.7 7.6 100.0 24.5 13.6 26.8 30.7 13.6 Note: r = uncorrected mean correlation; ρ = corrected mean correlation; k = number of studies; N = sample size; 95% CI = 95% confidence interval around uncorrected mean correlation; 95% CrI = 95% credibility interval around corrected mean correlation; % variance explained = percentage of variability in effect sizes that would be expected based on sampling error. Antecedents of POS Treatment by Organization Members Favorable treatment by organization members should enhance employees’ view that the organization is positively oriented toward them. Because they ascribe humanlike characteristics to the organization and view organizational members as agents of the organization, employees will tend to attribute role-related actions taken by members of the organization to the organization itself (Eisenberger et al., 1986). Thus, behaviors by organizational members that meet socioemotional needs should have an especially strong influence on POS. The results for such favorable treatment are presented in Table 2. Supervisor and coworker supportiveness. Perceived supervisor support refers to employees’ view that their supervisor values their contributions and cares about their well-being (Eisenberger, Stinglhamber, Vandenberghe, Sucharski, & Rhoades, 2002). Although the supportiveness of all organizational members should be related to POS, some members more closely embody the organization and are seen as acting on its behalf (Eisenberger et al., 2010). Supervisors and others in leadership roles play a key role in providing organizational rewards and resources to employees, and thus should be viewed as a greater source of organizational support than coworkers (Wayne, Shore, & Liden, 1997). Therefore, supportive behavior from supervisors should be more closely related to POS than supportive behavior from coworkers and teammates, who would be seen as less representative of the organization. Hypothesis 1: Perceived supervisor support is positively related to POS. Hypothesis 2: Perceived supervisor support is more strongly related to POS than is perceived coworker support or perceived team support. Supervisor support was strongly related to POS, ρ = .60. Based on confidence intervals that do not overlap, supervisor support is more closely related to POS than is coworker    1861 support, ρ = .47, or team support, ρ = .34. Consistent with OST, support from higher-status organizational members is more strongly interpreted by employees as organizational support. Abusive supervision. Abusive supervision refers to subordinates’ subjective assessment of the supervisor’s hostile behavior (Tepper, 2000). Because supervisors embody and represent their organizations to some degree, abusive supervision should reduce POS. Hypothesis 3: Abusive supervision is negatively related to POS. Abusive supervision was negatively related to POS, ρ = –.34. Thus, hostile supervisor behavior reflects poorly on the organization. This relationship was weaker in magnitude than the POS–supervisor support relationship. This may be, in part, a measurement issue because abusive supervision measures use extreme negative wording. Alternatively, this weaker relationship with POS may be due to the tendency to attribute personally oriented abuse more to the personality of the supervisor than to the organization’s general supportiveness (Eisenberger et al., 2010). Leader consideration and initiating structure. Leader consideration, the extent to which the leader is supportive and shows concern for subordinates’ well-being, is usually contrasted with initiating structure, in which the leader communicates clear work role expectations to subordinates. Although followers associate both types of leadership with effectiveness (Judge, Piccolo, & Ilies, 2004), consideration should be more strongly related to POS since it more directly conveys high regard for the work group. Hypothesis 4: Leader consideration is more strongly related to POS than is leader initiating structure. POS was more strongly related to leader consideration, ρ = .46, than to initiating structure, ρ = .33, although both types of leadership were related to POS. Leader behaviors that convey caring, concern, and support for followers appear to be effective ways to enhance POS. However, taking the time and effort to clarify role responsibilities also communicates to followers their leaders’ positive valuation by leaders. Transformational and transactional leadership. Transformational and transactional leadership (Bass, 1990) should differ in the extent to which each type of leadership fulfills socioemotional needs and lead to POS. Transformational leadership includes individualized consideration and inspirational motivation. Because individualized consideration involves concern with followers’ needs, transformational leadership should enhance their subjective well-being. Inspirational motivation provides followers with purpose and efficacy, thus conveying positive valuation of their contributions to the organization. In contrast, transactional leadership involves the use of rewards to motivate followers’ performance and corrective action to prevent or mend errors and should be more associated with the short-term trade of effort for wages, conveying less positive valuation of employees. Hypothesis 5: Transformational leadership is more strongly related to POS than is transactional leadership. 1862   Journal of Management / July 2017 Table 3 Meta-Analytic Results for Employee–Organization Relationship Quality Variable r Value congruence with organization .41 Psychological contracts Psychological contract breach –.60 Psychological contract fulfillment .37 Procedural justice .59 Distributive justice .51 Interactional justice .47 Organizational politics –.75 95% CI ρ 95% CrI k N % Variance Explained [.37, .45] .50 [–0.14, 1.00] 7 1,946 3.3 [–.63, –.58] [.35, .39] [.58, .59] [.50, .53] [.45, .49] [–.76, –.74] –.67 .42 .66 .57 .52 –.83 [–0.95, –0.38] [0.15, 0.69] [0.37, 0.95] [0.34, 0.80] [0.20, 0.85] [–1.00, –0.54] 16 9 72 46 21 12 3,464 9,315 25,070 16,929 6,835 4,980 9.6 4.5 6.5 11.0 7.7 2.5 Note: r = uncorrected mean correlation; ρ = corrected mean correlation; k = number of studies; N = sample size; 95% CI = 95% confidence interval around uncorrected mean correlation; 95% CrI = 95% credibility interval around corrected mean correlation; % variance explained = percentage of variability in effect sizes that would be expected based on sampling error. POS was more strongly related to transformational leadership, ρ = .56, than transactional leadership, ρ = .21. Consistent with OST, transformational leaders, who display a concern for the needs of subordinates and inspire subordinates to dedicate themselves to organizational goals, more strongly convey POS than do transactional leaders. Leader–member exchange (LMX). A high-quality LMX relationship is characterized by mutual trust and respect wherein the leader provides substantial support and resources and the subordinate reciprocates with commensurate effort and performance (Liden, Sparrowe, & Wayne, 1997). Because leaders are representatives of the organization, LMX should be positively associated with POS. Hypothesis 6: LMX is positively related to POS. LMX was strongly related to POS, ρ = .53, similar in magnitude to the strong relationships found with leader consideration and transformational leadership. Thus, various forms of leadership that convey positive valuation for employees as individuals or members of a group are highly related to POS. Employee–Organization Relationship Quality Contextual factors that convey the organization’s regard for employees influence POS. Such factors include common values shared with employees (value congruence), fulfillment or breach of the perceived obligations to employees (psychological contracts), fairness of treatment, and perceived organizational politics. These results are presented in Table 3. Value congruence. High employee–organization value congruence occurs when employees have beliefs and priorities that are similar to those of their employers. Employees who do not share their values with the organization may find it more difficult to communicate    1863 and cooperate with organizational members (Erdogan, Kraimer, & Liden, 2004), resulting in lessened organizational support. We included three types of value congruence measures in our review: (a) direct measures where employees rated the alignment of their values and those of the organization, (b) difference scores reflecting the difference between the employee’s values and the organization’s values, and (c) correlations between the employee’s values and the organization’s values. There were too few studies (k = 7) to differentiate these methods of operationalization, so all were included in a single effect size estimate. Hypothesis 7: Employee–organization value congruence is positively related to POS. Consistent with OST, employee–organization value congruence was strongly related to POS, ρ = .50. In addition to this direct relationship between value congruence and POS, preliminary evidence indicates that POS may compensate for low value congruence in influencing employee job satisfaction and career satisfaction (Erdogan et al., 2004). Psychological contracts. The psychological contract reflects employees’ beliefs about their social exchange relationships with their organization, mutual obligations, and the extent to which these obligations are fulfilled (Rousseau, 1995). The obligations could be based on explicit promises made by the organization or implicit expectations held by employees. Because most organizational obligations are viewed by employees as voluntary implicit or explicit promises (Morrison & Robinson, 1997; Rousseau, 1995), contract breach should have a strong negative relationship with POS (Aselage & Eisenberger, 2009). Hypothesis 8: Psychological contract breach is negatively related to POS. As predicted, psychological contract breach, ρ = –.67, and contract fulfillment, ρ = .42, were related to POS. Additionally, the absolute values of these confidence intervals do not overlap, which suggests that, at least with regard to psychological contracts, the negative event of a contract breach has a stronger influence on POS than the positive event of contract fulfillment. Fairness. Fair procedures should make a major contribution to POS because they are generally seen as under the organization’s discretionary control (Moorman, Blakely, & Niehoff, 1998). Distributive justice involves the fair distribution of outcomes, procedural justice concerns processes used to determine outcome distribution, and interactional justice refers to the way one is treated during the enactment of organizational decisions. Procedural justice reflects policies and practices, such as employee participation and bias in decision making, that are relatively stable and generally under the overall organization’s control. Distributive and interactional fairness should be less related to POS. Distributive justice involves specific pay decisions more subject to factors outside of the organization’s control. Interactional justice is based on the actions of specific organizational agents, such as supervisors, who differ in the extent to which they embody the organization (Shoss, Eisenberger, Restubog, & Zagenczyk, 2013). Interactional justice also tends to be more episodic and less stable than procedural justice (D. Ferris, Spence, Brown, & Heller, 2012) and thus may be more attributable to specific events. 1864   Journal of Management / July 2017 Hypothesis 9: Procedural justice is more strongly related to POS than distributive or interactional justice. Although all three forms of fairness were strongly related to POS, POS was more strongly related to procedural justice, ρ = .66, than to distributive justice, ρ = .57, and interactional justice, ρ = .52. These results suggest that employees place an added emphasis on general enduring characteristics of the organization as indications of the organization’s favorable or unfavorable orientation toward them. Perceived organizational politics. Organizational politics involves perceptions of favoritism, lack of meritocracy and of frank discussion, and the presence of cliques or in-groups (G. Ferris & Kacmar, 1992), all of which influence organizational policies and practices. Such a political organizational orientation would suggest to employees that the organization cares too little about their welfare to rein in harmful behaviors. Hypothesis 10: Perceptions of organizational politics are negatively related to POS. As expected, perceptions of organizational politics were strongly and negatively related to POS, ρ = –.83. Because of this strong relationship, and because this estimate is based on only 12 studies, we reviewed 10 recent studies for evidence that POS and politics are distinct constructs (see online Appendix A for a table detailing the relevant findings of each study). Of the eight studies reporting confirmatory factor analyses (CFAs), all but one found POS to be clearly distinct from organizational politics. Distinct from POS, organizational politics therefore seems to be highly damaging to POS because organization members who are willing to damage the organization’s future to further their personal welfare may have little interest in generally supporting other organization members. HR Practices and Job Conditions Here we focus on HR practices, work role characteristics, and working conditions that can be enhanced by the organization to make the context and nature of work more pleasant. These results are presented in Table 4. Most employees believe that organizations have considerable control over HR practices (Eisenberger et al., 1997). This perception reflects the substantial control that many organizations do have over such practices, such that perceptions of their favorableness should be strongly related to POS. Hypothesis 11: Job security, flexible work practices, family supportive work practices, and developmental opportunities are positively related to POS. Developmental opportunities were strongly related to POS, ρ = .57, and job security was moderately related to POS, ρ = .42. Flexible work schedules, ρ = .21, and perceptions of family supportive organizational practices, ρ = .26, were also related to POS, although less so. These findings are consistent with the findings of Eisenberger et al. (1997), indicating that training (developmental) opportunity was the job condition most strongly viewed as under the discretionary control of organizations and, presumably, most indicative of POS. By contrast, flexible and family-supportive work practices may be seen as supportive mainly by 1865    Table 4 Meta-Analytic Results for Favorable Human Resource Practices and Job Conditions Variable Employee security and benefits Developmental opportunities Job security Flexible work schedule Perceptions of family supportive organizational practices Benefits used Work role characteristics Enriching job characteristics Autonomy Participation in decision making Role stressors   Role ambiguity   Role conflict   Role overload r 95% CI ρ 95% CrI k N % Variance Explained .50 .36 .16 .22 [.48, .51] [.34, .39] [.08, .24] [.19, .24] .57 .42 .21 .26 [.39, .75] [.13, .71] [.10, .31] [–.13, .64] 15 11 3 6 12,712 4,098 627 4,425 8.7 11.2 72.2 4.3 .07 [.02, .12] .08 [.03, .12] 5 1,550 88.9 .52 .42 .49 [.47, .57] [.40, .43] [.46, .51] .65 .51 .56 [.65, .65] [.34, .68] [.25, .87] 5 15 13 832 10,603 3,640 98.1 13.2 9.6 –.31 –.38 –.24 [–.32, –.29] [–.40, –.37] [–.25, –.22] –.36 –.45 –.29 [–.84, .13] [–.78, –.12] [–.51, –.06] 31 26 19 12,757 10,270 11,951 4.2 8.2 13.0 Note: r = uncorrected mean correlation; ρ = corrected mean correlation; k = number of studies; N = sample size; 95% CI = 95% confidence interval around uncorrected mean correlation; 95% CrI = 95% credibility interval around corrected mean correlation; % variance explained = percentage of variability in effect sizes that would be expected based on sampling error. those employees who need and want those benefits, weakening the relationship between these practices and POS among workers without these needs. Gouldner (1960) suggested that the receipt of benefits specific to an individual’s needs would create a greater obligation to reciprocate because such benefits convey a greater concern for the individual’s well-being. As an example, the substantial number of employees who have little desire or need for flexible and family-oriented work practices may associate these benefits to only a small degree with POS. This argument suggests that actual benefit use, rather than just the availability of benefits, should be positively associated with POS. Hypothesis 12: Benefit use is positively related to POS. Benefit use was related to POS, ρ = .08, as hypothesized. The weakness of this relationship should be judged cautiously in view of the small number of studies (k = 5) but may indicate the need for a more nuanced approach in future research studies to better understand the conditions under which benefits contribute to POS. For example, many organizations offer health benefits that cover only a small proportion of costs and thus may be viewed as indicating low POS. From a measurement perspective, because employees may engage in different degrees of benefit use, measuring use in an all-or-none fashion may be insufficient. Both sufficiency and use of benefits should be taken into account in future research. Work role characteristics relevant to POS include job enrichments, such as skill variety, task identity, autonomy, and feedback (Hackman & Oldham, 1980), and role stressors, such 1866   Journal of Management / July 2017 as ambiguity, conflict, and overload (Kahn, Wolfe, Quinn, Snoek, & Rosenthal, 1964). Eisenberger et al. (1997) reported that job enrichment conditions were believed by employees to be under greater organizational control than role stressors. Therefore, job enrichment characteristics should be more strongly related to POS than role stressors. Hypothesis 13: Job enrichment conditions are more strongly related to POS than work role stressors. Job enrichment conditions as a group were closely related to POS, ρ = .65. Considering specific job conditions, POS was highly associated with autonomy, ρ = .51, and participation in decision making, ρ = .56. As predicted, POS was less related to role stressors, including ambiguity, ρ = –.36; conflict, ρ = –.45; and overload, ρ = –.29. According to Eisenberger and Stinglhamber (2011), although stress and pressure are highly unpleasant for employees, in most occupations, employees attribute stress and pressure more to the nature of the job rather than to discretionary actions by employers. This suggests the importance of attribution for POS when employees experience favorable or unfavorable work experiences. Relative Importance of Supervisor Support, Fairness, and Dispositional Affectivity We examined the distinct contributions of three conceptually important antecedents of POS: supervisor support, fairness, and dispositional affectivity. Supervisor support, involving communication of the leader’s positive valuation of subordinates and care about their well-being, is the component of all leadership behaviors most directly related to POS and with the strongest relationship to POS. We selected organizational fairness because it is widely considered a fundamental aspect of employees’ experience that makes an important contribution to POS (Rhoades & Eisenberger, 2002). Finally, we included affectivity because, although receiving relatively little attention in OST, affectivity has been found to be related to job attitudes (e.g., Thoreson, Kaplan, Barsky, Warren, & deChermont, 2003) and may substantially influence POS. The inclusion of affectivity allowed us to test the incremental effects of fairness and supervisor support on POS beyond this broad personal disposition. To construct the correlation matrix shown in Table 5, we either (a) used correlations from other meta-analyses or (b) computed correlations using studies from our meta-analysis. In computing correlations involving fairness, we used the intercorrelations among the justice dimensions reported by Colquitt et al. (2013) to create a composite of fairness dimensions. We could not find a meta-analysis of the relationship between negative affectivity and supervisor support, and there was not a study in our meta-analysis that reported this relationship. Thus, we used .30 as an estimate for this effect, based on similar effects found by Thoresen et al. (2003). Because we were unable to find a precise estimate of this relationship, we ran analyses using estimates of .20 and .40 to test whether this influenced the conclusions drawn from our analyses. We tested a multiple regression model predicting POS from supervisor support, fairness perceptions, and affectivity. We used as the sample size the harmonic mean of the samples sizes on which the correlations in Table 5 were based, which was 3,159. Our 1867    Table 5 Intercorrelations Among Variables 1. Supervisor support 2. Fairness 3. Positive affectivity 4. Negative affectivity 5. POS 1 2 3 4 5 — .61a .33a –.30e .60 — .26b –.22b .70 — –.36c .34d — –.43d — Note: POS = perceived organizational support. aBased on correlations reported among studies included in this meta-analysis. bBased on meta-analytic correlations reported by Barsky and Kaplan (2007). cBased on an estimate reported by Thoresen, Kaplan, Barsky, Warren, and de Chermont (2003). dBased on meta-analytic correlations reported by Rhoades and Eisenberger (2002). eEstimated based on corrected correlations found between negative affectivity and job satisfaction (–.34), organizational commitment (–.28), and turnover intentions (.22) (Thoresen et al., 2003). Models were also tested using .20 and .40 as the correlation between negative affectivity and supervisor support. Table 6 Analyses of Relative Importance Variable Supervisor support Fairness perceptions Positive affectivity Negative affectivity β Weights From Multiple Regressiona Raw Relative Weight Relative Weight as % of R2 .20 .51 .06 –.24 .161 .302 .039 .097 26.8 50.5 6.5 16.2 Note: N = 3,159, which is the harmonic mean of the samples sizes on which the correlations in Table 5 are based. All coefficients were statistically significant at the .05 alpha level. aModel R2 = .60. multiple regression results (see Table 6) indicate that fairness perceptions emerged as the strongest predictor of POS, β = .51, followed by negative affectivity, β = –.24; supervisor support, β = .20; and positive affectivity, β = .06. All weights were statistically significant at p < .05. We also conducted a relative-weights analysis using techniques described by Johnson (2000) and LeBreton and Tonidandel (2008). The relative weights analysis yielded similar results (see Table 6), although supervisor support had a stronger relative weight than negative affectivity. Analyses were also run using –.20 and –.40 as the correlation between supervisor support and negative affectivity. This did not influence the ordering among the antecedents in their relative importance, although the relative weights for negative affectivity and supervisor support increased as their intercorrelation decreased. These results suggest that fairness has a strong unique effect on POS, with supervisor support and negative affectivity having lesser unique effects. The nearly comparable effect size for negative affectivity and supervisor support suggests the need to examine the influence of negative affectivity on POS in more detail.