Academy of Management Journal2009, Vol. 52, No. 4, 779–801. THE RELATIONSHIP BETWEEN PERCEPTIONS OF ORGANIZATIONAL POLITICS AND EMPLOYEE ATTITUDES, STRAIN, AND BEHAVIOR: A META-ANALYTIC EXAMINATION CHU-HSIANG CHANG University of South Florida CHRISTOPHER C. ROSEN University of Arkansas PAUL E. LEVY University of Akron The current study tested a model that links perceptions of organizational politics to job performance and “turnover intentions” (intentions to quit). Meta-analytic evidence supported significant, bivariate relationships between perceived politics and strain (.48), turnover intentions (.43), job satisfaction ( .57), affective commitment ( .54), task performance ( .20), and organizational citizenship behaviors toward individuals ( .16) and organizations ( .20). Additionally, results demonstrated that work atti- tudes mediated the effects of perceived politics on employee turnover intentions and that both attitudes and strain mediated the effects of perceived politics on perfor- mance. Finally, exploratory analyses provided evidence that perceived politics repre- sent a unique “hindrance stressor.” Organizational politics are ubiquitous and have or coworkers (Ferris, Russ, & Fandt, 1989). There-widespread effects on critical processes (e.g., per- fore, organizational politics are often viewed as aformance evaluation, resource allocation, and man- dysfunctional, divisive aspect of work environ-agerial decision making) that influence organi- ments (Mintzberg, 1983). The current article fo-zational effectiveness and efficiency (Kacmar & cuses on understanding how employees’ percep-Baron, 1999). Employees may engage in some tions of illegitimate, self-serving political activitieslegitimate, organizationally sanctioned political ac- (viz., perceptions of organizational politics) influ-tivities that are beneficial to work groups and or- ence individual-level work attitudes and behaviors.ganizations (see Fedor, Maslyn, Farmer, & Betten- Accumulating empirical research has providedhausen, 2008). For example, managers who are considerable evidence for linkages between percep-“good politicians” may develop large bases of so- tions of organizational politics and a variety ofcial capital and strong networks that allow them to employee outcomes, including job satisfaction, af-increase the resources that are available to their fective organizational commitment, and job anxietysubordinates (Treadway et al., 2004). On the other (see Ferris, Adams, Kolodinsky, Hochwarter, &hand, employees also demonstrate a number of il- Ammeter, 2002). However, despite the intuitive ap-legitimate political activities (e.g., coalition build- peal of the idea that perceived politics will have aning, favoritism-based pay and promotion decisions, impact on key individual-level outcomes associ-and backstabbing) that are strategically designed to ated with organizational effectiveness, research hasbenefit, protect, or enhance self-interests, often failed to consistently demonstrate such an impact.without regard for the welfare of their organization For example, Ferris et al. (2002) observed that four of nine studies (e.g., Cropanzano, Howes, Grandey, & Toth, 1997; Hochwarter, Witt, & Kacmar, 2000; We thank Brad Kirkman and the three anonymous Parker, Dipboye, & Jackson, 1995; Randall, Cropan-reviewers for their helpful insights. We would like tonote that the first two authors contributed equally to this zano, Bormann, & Birjulin, 1999) relating percep-project. We would like to thank Rosalie Hall for her tions of organizational politics to task performancehelpful comments on an earlier version of the article. We and organizational citizenship behavior (OCB) didare also grateful for the assistance of Michelle Matias and not support the expected negative linkages. Simi-Jessica Junak in preparing our manuscript. larly, four of nine studies (e.g., Cropanzano et al., 779Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download or email articles for individual use only.
780 Academy of Management Journal August1997; Harrell-Cook, Ferris, & Dulebohn, 1999; mine if one, both, or neither of these explanationsHochwarter, Perrewe, Ferris, & Guercio, 1999; Ran- accounts for the effects of perceptions of organiza-dall et al., 1999) examining linkages between polit- tional politics.ical perceptions and turnover intentions failed to A recent meta-analysis of the outcomes of per-reach statistical significance (Ferris et al., 2002). ceptions of organizational politics (see Miller,Thus, evidence linking perceptions of organization- Rutherford, & Kolodinsky, 2008) underscores someal politics to these outcomes is equivocal. of these empirical and theoretical weaknesses in Moreover, it is not clear whether these inconsis- the politics literature. Empirically, Miller et al.’stent findings exist because of statistical artifacts (2008) results failed to clearly support a linkage(e.g., low power) or because the politics-outcome between perceptions of organizational politics andrelationships are not negative. Regarding the latter performance. Moreover, their study (1) did notpoint, Ferris et al. (1989) noted that employees may present an overarching theoretical framework thatrespond to perceptions of organizational politics by explains why perceived organizational politics isincreasing involvement in their jobs. Ferris et al. linked to employee attitudes and behaviors and (2)(1989) suggested that perceived politics may lead to focused only on bivariate linkages between the con-positive outcomes when they are experienced as struct and its outcomes, without considering howopportunity stress (Schuler, 1980), which occurs outcomes of perceptions of organizational politicswhen a stressor presents an opportunity for em- relate to one another. The current research ad-ployees to gain something from the situation at dresses these shortcomings of the literature on per-hand. Employees respond to opportunity stress by ceptions of organizational politics in three ways.putting more time and effort into their jobs in an First, this study provides a comprehensive, quanti-attempt to capitalize on the situation (LePine, Pod- tative review of the relationships between per-sakoff, & LePine, 2005; Schuler, 1980). Supporting ceived organizational politics and its outcomes.this perspective, there is evidence that perceptions Meta-analysis allowed estimation of the true pop-of organizational politics are associated with desir- ulation effect size (Hunter & Schmidt, 2004) andable outcomes, including lower strain (Ferris et al., examination of whether study characteristics ex-1993), and increased job involvement (Ferris & Kac- plain variability in effect sizes. Thus, a meta-ana-mar, 1992) and performance (Maslyn & Fedor, lytic examination of the perceptions of organiza-1998; Rosen, Levy, & Hall, 2006). Thus, it is not tional politics– outcome relationships is importantclear whether inconsistent findings in the literature because it helps determine whether the past incon-are a function of study artifacts or exist because sistent findings were the result of statistical arti-perceptions of organizational politics are either notrelevant to or positively associated with certain facts or, rather, were associated with a broader is-outcomes. sue, such as the misspecification of relationships In addition, the theoretical underpinnings of the between perceptions of organizational politics andlinkages between perceptions of organizational pol- its outcomes.itics and job performance and turnover intentions Second, and perhaps more importantly, the cur-are not well understood, as existing frameworks do rent study focuses on developing and testing a the-not explain how these perceptions are associated oretically derived model that identifies the key psy-with critical employee outcomes. Rather, concep- chological mechanisms that link perceivedtual models (e.g., Aryee, Chen, & Budhwar, 2004; organizational politics to its distal outcomes. Fig-Ferris et al., 2002) specify that perceptions of or- ure 1 outlines the proposed model, which inte-ganizational politics are related directly to em- grates the organizational politics literature withployee attitudes and behaviors. Hence, knowledge theoretical frameworks that specify the causal or-of the psychological mechanisms that relate politi- dering of stress-related outcomes (Podsakoff,cal perceptions to employee outcomes is limited, LePine, & LePine, 2007; Schaubroeck, Cotton, &and there is little guidance for systematically ex- Jennings, 1989). This approach is consistent withamining these mechanisms. In addition, research previous studies (LePine et al., 2005; Podsakoff ethas failed to examine mediators that link percep- al., 2007) that have cast organizational politics as ations of organizational politics to outcomes. For hindrance stressor that prevents employees fromexample, theorists have noted that job stress and meeting personal and professional goals. We testedsocial exchange theories may explain reactions to the validity of the proposed model using meta-these perceptions (Cropanzano et al., 1997; Ferris et analytically derived correlations. Thus, the contri-al., 2002). However, the dearth of empirical re- bution of our meta-analysis is enhanced by its abil-search examining the linkages implied by these ity to not only provide information on the strengththeories has limited researchers’ ability to deter- of the bivariate relationships between constructs,
2009 Chang, Rosen, and Levy 781 FIGURE 1 Proposed and Alternative Models of Effects of Perceptions of Organizational Politics on Employee Outcomesbut also explain how the focal constructs are related BACKGROUND AND THEORY(Viswesvaran & Ones, 1995). Theorists have provided two explanations that Finally, we explore whether perceptions of or- link perceptions of organizational politics to nega-ganizational politics can be distinguished fromother hindrance stressors. Although some research- tive work outcomes. First, Ferris et al. (1989) sug-ers have argued that it is best to treat various stres- gested that politics are a source of stress that elicitssors as distinct yet related constructs (Schaubroeck strain responses from employees. Other theoristset al., 1989), the hindrance stressor literature im- have suggested that perceptions of organizationalplies that perceptions of organizational politics, politics are detrimental to the maintenance ofrole ambiguity, and role conflict are all indicators healthy employee-organization exchange relation-of a higher-order hindrance stressor factor (LePine ships (Aryee et al., 2004; Hall, Hochwarter, Ferris,et al., 2005; Podsakoff et al., 2007). Thus, we con- & Bowen, 2004). Below, we review these explana-ducted exploratory analyses to compare relation- tions of the effects of perceived organizational pol-ships among perceptions of organizational politics, itics and apply Schaubroeck et al.’s (1989) frame-role stressors, and outcomes and to evaluate mod- work of work stress to tie these perspectivesels based on different conceptualizations of the together. Finally, we develop a model based on thehindrance stressor construct. hindrance stressor literature to link perceptions of
782 Academy of Management Journal Augustorganizational politics to proximal (strain and atti- gate latent constructs. The morale construct repre-tudes) and distal (performance and turnover inten- sented general employee attitudes and was com-tions) outcomes. prised of job satisfaction and affective commitment (see Harrison, Newman, & Roth, 2006), and the performance construct consisted of task perfor-Stress-Based Effects of Perceptions of mance and OCB (see Rotundo & Sackett, 2002),Organizational Politics which captured behaviors related to both the tech- Drawing on research conceptualizing job stress nical cores of organizations and behaviors that con-as a subjective experience associated with uncer- tribute to the psychosocial contexts of workplacestainty and ambiguity (e.g., Schuler, 1980), Ferris et (Organ, 1997). Rosen et al. (2006) suggested thatal. (1989) proposed that perceptions of organiza- lower morale reflects judgments that reward allo-tional politics represent a stressor that is directly cation processes are arbitrary and unfair. Employ-related to attitudinal and behavioral reactions. Fer- ees holding less favorable attitudes also feel lessris et al. speculated that perceptions of organiza- obligated to reciprocate with behaviors that en-tional politics trigger a primary appraisal (Lazarus hance the well-being of their organization. Thus,& Folkman, 1984) that a work context is threatening Rosen et al. provided evidence, albeit indirectly,and put pressure on employees to engage in poli- that morale is part of the social exchange mecha-ticking to meet their goals. Highly political organi- nism that links perceptions of organizational poli-zations tend to reward employees who (1) engage in tics to performance.strong influence tactics, (2) take credit for the workof others, (3) are members of powerful coalitions, Current Study: Model and Hypothesesand (4) have connections to high-ranking allies. Asorganizations reward these activities, demands are The stress and social exchange perspectives areplaced on workers to engage in political behaviors useful to understanding reactions to perceptions ofto compete for resources. According to the job de- organizational politics. Nonetheless, research fallsmands–resource model of work stress (Demerouti, short in describing the mechanisms that link suchBakker, Nachreiner, & Schaufeli, 2001), employees perceptions to outcomes. For example, Ferris etwho perceive that job demands exceed their coping al.’s (2002) model specifies that job anxiety, jobresources feel overwhelmed. This emotional strain satisfaction, affective commitment, performance,requires additional coping efforts, which are taken and turnover intentions are direct outcomes of per-away from resources that could otherwise be de- ceptions of organizational politics, with each reac-voted to job performance. Excessive strain also im- tion occurring at the same time. However, there arepacts employee health (Dragano, Verde, & Siegrist, reasons to believe that some reactions to political2005) and eventually drives employees to search perceptions precede others. In particular, workfor less stressful work environments. stress research (Schaubroeck et al., 1989) suggests that job anxiety, job satisfaction, and affective com- mitment are antecedents to turnover intentions andA Social Exchange Perspective on the Effects of performance. Therefore, we suggest that percep-Perceptions of Organizational Politics tions of organizational politics have indirect effects In highly political organizations, rewards are tied on turnover intentions and performance throughto relationships, power, and other less objective more immediate outcomes (viz., strain and morale).factors. As a result, “the immediate environment As such, previous studies examining only the di-becomes unpredictable because the unwritten rules rect effects of perceptions of organizational politicsfor success change as the power of those playing on performance and turnover intentions may havethe political game varies” (Hall et al., 2004: 244). misspecified these linkages, thus biasing the studyTherefore, it is difficult for employees to predict if results (Duncan, 1975).their behaviors will lead to rewards in political The stress and social exchange perspectives em-work contexts, and they are likely to perceive ploy a similar logic useful for understanding em-weaker relationships between performance and the ployees’ reactions to perceptions of organizationalattainment of desired outcomes (Aryee et al., 2004; politics. Particularly, both perspectives suggest thatCropanzano et al., 1997). Supporting this perspec- these perceptions are associated with ambiguitytive, Rosen et al. (2006) demonstrated that percep- and uncertainty in a work environment that resultstions of organizational politics are associated with in psychological strain and lower morale. However,performance through employee morale. In their neither perspective describes how these outcomesstudy, as in the present study, employee morale of perceptions of organizational politics relate toand job performance were conceptualized as aggre- each other and whether these outcomes have a
2009 Chang, Rosen, and Levy 783meaningful impact on more distal reactions. Fortu- We propose that perceptions of organizationalnately, the work stress literature provides insight politics have both direct and indirect effects onregarding the causal ordering of these reactions to morale. In turn, psychological strain and moralestressors. Following Mobley, Horner, and Holling- link perceptions of organizational politics to moresworth’s (1978) model of turnover, Schaubroeck et distal outcomes. In other words, employees’ perfor-al. (1989) specified that role stressors lead to in- mance suffers because they must focus time andcreased job strain, which is associated with lower effort on coping with the strain associated withjob satisfaction and affective commitment and, sub- perceptions of organizational politics. In addition,sequently, increased turnover intentions. Podsakoff employees are likely to reduce the time and effortet al. (2007) and LePine et al. (2005) employed that they put into their jobs in response to per-similar mediational chains to explain the effects of ceived disequilibrium in the exchange relation-hindrance stressors on turnover and task perfor- ship, which is reflected by lower morale. Finally,mance. Moreover, Cropanzano, Rupp, and Byrne employees will attempt to remove themselves from(2003) demonstrated that the effects of strain work situations appraised as unfavorable or threatening.through morale and impact OCBs, in addition to In summary, we hypothesize the following:task performance. Together, these studies provide Hypothesis 1. Perceptions of organizationalcomplementary approaches to understanding the politics has a positive relationship with psy-effects of stressors. chological strain. We incorporate these perspectives into a modelthat conceptualizes perceptions of organizational Hypothesis 2. Perceptions of organizationalpolitics as a hindrance stressor reflecting job de- politics has a (a) direct negative relationshipmands that interfere with employees’ ability to with morale (b) partially mediated by psycho-achieve career goals. Hindrance stressors are logical strain.broadly defined as constraints that impede individ- Hypothesis 3. Perceptions of organizationaluals’ work achievements and are not usually asso- politics has a positive relationship with turn-ciated with potential net gains for them (LePine et over intentions.al., 2005). In addition to perceptions of organiza-tional politics, researchers include role stressors, Hypothesis 4. Perceptions of organizationalbureaucracy, and daily hassles under the umbrella politics has a negative relationship with jobof hindrance stressors. Collectively, research has performance.shown that these stressors elicit strain, reduce mo- Hypothesis 5. The relationship between per-rale, motivation, and performance, and increase ceptions of organizational politics and turn-employee withdrawal (Boswell, Olson-Buchanan, over intentions is mediated by (a) psychologi-& LePine, 2004; Cavanaugh, Boswell, Roehling, & cal strain and (b) morale.Boudreau, 2000; Podsakoff et al., 2007). In keepingwith previous research examining the effects of Hypothesis 6. The relationship between per-perceptions of organizational politics and hin- ceptions of organizational politics and job per-drance stressors, we argue that politics hamper em- formance is mediated by (a) psychologicalployees’ ability to attain personal and professional strain and (b) morale.goals, which results in a primary appraisal of thework context that evokes strain and reduces mo- Exploratory Analyses: Comparing Politics torale. In accordance with the causal ordering sup- Other Hindrance Stressorsported by previous studies, we also propose thatstrain is a more proximal outcome than morale. In Schaubroeck et al.’s (1989) model, role ambi-This proposition derives from both the stress and guity and role conflict represent distinct, yet re-social exchange perspectives. Work stress research- lated, stressors. More recently, researchers (e.g.,ers (Schaubroeck et al., 1989) have suggested that LePine et al., 2005; Podsakoff et al., 2007) havepsychological strain influences employees’ overall suggested that a unified hindrance stressor con-attitudes toward their jobs, as employees consider struct encompasses perceptions of organizationaltheir jobs to be the root of the problem. Strain is politics and role stressors. Perceived politics andalso purported to reflect a negative evaluation of role stressors certainly share the similarity of inter-the employee-organization exchange relationship fering with employees’ ability to achieve personal(Cropanzano et al., 1997). Thus, as strain increases, and professional goals. However, conceptualizingemployees’ morale and sense of obligation toward these three constructs as indicators of a unifiedtheir organization decline (Cropanzano et al., hindrance stressor construct entails an assumption2003). that perceptions of organizational politics and role
784 Academy of Management Journal Auguststressors are analogous and demonstrate similar re- rated performance: “performance,” “productivity,”lationships with each other and with outcomes. “task/job performance,” “organizational citizen-Unfortunately, this assumption has not been empir- ship behavior,” “OCB,” “OCBI” [OCB toward indi-ically tested. Therefore, we provide supplemental viduals], “OCBO” [OCB toward organizations], andanalyses that, first, compare relationships among “contextual performance”). Second, we manuallyperceptions of organizational politics, role stres- searched the 1989 –2007 issues of eight high-qual-sors, and outcomes, and second, explore whether ity journals that have published articles related topolitical perceptions and role stressors are best organizational politics: the Academy of Manage-conceptualized as a unified construct or as a set of ment Journal, Journal of Applied Psychology, Jour-diversified yet related stressors. Figure 2 graphi- nal of Management, Journal of Organizationalcally depicts these two contrasting patterns. Third, Behavior, Journal of Vocational Behavior, Organi-our additional analyses also explore whether per- zational Behavior and Human Decision Processes,ceptions of organizational politics and role stres- Organization Science, and Personnel Psychology.sors demonstrate similar patterns of relationships Third, we compared the reference list derived fromwith distal outcomes. Similar relationships with the sources described so far with the lists of twoeach other and with outcomes and a better-fitting qualitative reviews of research on perceptions ofmodel based on a unified hindrance stressor con- organizational politics (Ferris et al., 2002; Kacmarstruct would provide further evidence for the uni- & Baron, 1999). Finally, we contacted researchersfied approach (Podsakoff et al., 2007). On the other in the field for “file-drawer studies” and posted ahand, differing relationship patterns and a better fit call for unpublished papers on discussion lists forfor the diversified model would imply that percep- the Society for Industrial/Organizational Psychol-tions of organizational politics may have meaning- ogy and the Academy of Management. In total, weful differences with other role-based hindrance identified 57 relevant papers dealing with 70 sep-stressors. arate samples that we could include in the meta-analyses. Research Question. Are perceptions of organi- Three inclusion criteria were used. First, we in- zational politics and role stressors (role ambi- cluded studies in the meta-analysis if they investi- guity and role conflict) distinct forms of hin- gated relationships between perceptions of organi- drance stressors? zational politics and at least one of the dependent variables. Second, we included studies that mea- METHODS sured politics perceptions and excluded studies that measured other operationalizations of organi-Literature Search and Inclusion Criteria zational politics. The majority of studies used vari- To identify studies that could be used in this ations of the perceptions of organizational politicsmeta-analysis, we first conducted a computerized measure (Kacmar & Ferris, 1991); three exceptionssearch of three databases (PSYCINFO, ABI-INFORM, were Anderson (1994), Christiansen, Villanova,and Business Source Premier) for the years between and Mikulay (1997), and Drory (1993). Kacmar and1989 (the year that Ferris and colleagues proposed Baron’s (1999) qualitative review suggested that thethe perceptions of organizational politics con- three scales in those studies assess perceptions ofstruct) and 2007. We combined keywords associ- organizational political climate. In addition, ourated with politics (i.e., “organizational politics,” own content analysis revealed that items from“politics perceptions,” and “perceived politics”) these scales have counterparts in the perceptions ofwith keywords related to outcomes (general out- organizational politics measure. Thus, we includedcomes: “outcome,” “consequence,” and “result”; these in the current meta-analysis.2 Finally, we in-strain1: “strain,” “stress,” “stressor,” “anger,” “anx- cluded studies reporting relationships betweeniety,” “depress[ion],” “frustration,” “tension,” and perceptions of organizational politics and depen-“burnout”; morale: “job/work satisfaction,” “organ-izational/work commitment,” “affective commit- 2ment”; turnover intentions: “turnover,” “intent to We also conducted meta-analyses without these studies for the applicable analyses and found the resultsturnover,” “withdrawal cognitions”; supervisor- showed minor, nonsignificant fluctuations. After remov- ing these studies, we observed these changes in relation- ships: (a) perceptions of organizational politics and 1 Two-thirds of the samples we found (14 out of 21) strain, from .48 to .47; (b) politics and job satisfaction,used the Work Tension Scale by House and Rizzo (1972) from .57 to .58; (c) politics and affective commitment,to assess psychological strain associated with tension from .54 to .55; and (d) politics and withdrawal in-experienced at work. tentions, from .43 to .44.
2009 Chang, Rosen, and Levy 785 FIGURE 2 Conceptualizations of Hindrance Stressordent variables that were calculated from an original were cross-referenced, and only one effect size wassample. When the same sample was used in multi- included. Correlations were considered as separateple studies, sample characteristics and effect sizes entries when they represented relationships be-
786 Academy of Management Journal Augusttween perceptions of organizational politics and (1) tween all the variables using the current and pre-distinctive outcome variables and (2) one depen- vious meta-analytic results. Selected meta-analysesdent variable, but from different samples (Arthur, published since 1995 provided estimates for rela-Bennett, & Huffcutt, 2001). We aggregated correla- tionships among nonpolitics variables. Table 1 pre-tions representing relationships between percep- sents this correlation matrix. We performed struc-tions of organizational politics and different mea- tural equation modeling (SEM) based on thissures of the same outcome variable. These criteria correlation matrix to evaluate the fit of the pro-resulted in 21 effect sizes for strain, 45 for job posed model. We adopted Shadish (1996) and Vi-satisfaction, 33 for affective commitment, 27 for swesvaran and Ones’s (1995) procedures for modelturnover intentions, 14 for task performance, 9 for testing. Unless otherwise noted, the structuralOCBI, and 9 for OCBO. model used manifest indicators without correction for measurement error, as these corrections were done through meta-analysis. Finally, because noMeta-analytic and Model-Testing Procedures published meta-analysis estimates the relation- We first used meta-analysis to summarize rela- ships between turnover intentions and OCBI andtionships between perceptions of organizational OCBO, we used primary studies found for the cur-politics and each of the outcome variables. When rent meta-analysis to estimate these relationshipsavailable, we also examined whether the publica- (see Harrison et al., 2006). Specifically, seven of thetion status of a study, the employment status of the nine samples that we found in our meta-analysissampled population (full-time employees vs. em- for the relationships between perceptions of organ-ployed students), and the country from which the izational politics and OCBI and OCBO includeddata were collected, accounted for differences in turnover intentions as an outcome variable. Rela-effect sizes among studies. Following Arthur et al.’s tionships between turnover intentions and OCBs(2001) strategy, we calculated a sample-weighted were extracted from these seven studies and meta-mean correlation. We then computed the percent- analyzed to provide estimates for the meta-analyticage of variance accounted for by sampling error correlation matrix.(Hunter & Schmidt, 2004) and performed the chi- For the exploratory analyses, we performed ansquare test for the homogeneity of observed corre- additional literature search for meta-analytic corre-lation coefficients across studies (Rosenthal, 1991). lations involving role stressors (i.e., role ambiguityThe 95% confidence interval around the sample- and role conflict) and the outcome variables in-weighted mean correlation was then computed us- ¨ cluded in the current study (e.g., Ortqvist & Win-ing different formulas depending on chi-square test cent, 2006; Podsakoff, MacKenzie, & Bommer,results (Whitener, 1990). We then corrected for un- 1996). We then incorporated these correlations intoreliability of measures to derive the population cor- the meta-analytic correlation matrix (Table 1) andrelation coefficient. We used interrater reliability tested the proposed model using the two differentestimates from Viswesvaran, Ones, and Schmidt conceptualizations of the hindrance stressor con-(1996) to correct for measurement error in the rela- struct (i.e., as a unified vs. diversified construct).tionships between perceptions of organizational Because no published meta-analysis estimated thepolitics and supervisor-rated performance. The relationships between perceptions of organization-variance and standard deviation of the population al politics and the two role stressors, we used avail-estimate were then calculated to determine the able primary studies to estimate these relationships95% credibility interval. We also calculated the (six effect sizes for the perceptions of organization-Q-statistic to examine variance in the corrected al politics–role ambiguity relationship [ .52];population estimate. When the credibility interval four effect sizes for the perceptions of organiza-included zero or Q was significant, we performed tional politics–role conflict relationship [ .58]).subgroup analyses to examine the moderating ef-fects of study characteristics (Cortina, 2003). Themoderator analyses included an examination of RESULTSboth publication status and sample employment Bivariate Relationshipsstatus, as well as a cross-cultural comparison be-tween U.S. and Israeli samples, none of which were Table 2 shows the results of the meta-analysis forexamined in Miller et al.’s (2008) meta-analysis of the relationships between politics and strain, jobthe outcomes of perceptions of organizational satisfaction, affective commitment, turnover inten-politics. tions, task performance, OCBI, and OCBO. All of Next, we built a correlation matrix containing the the 95% confidence intervals excluded zero, indi-corrected population correlation coefficients be- cating that each correlation was statistically signif-
2009 Chang, Rosen, and Levy 787 TABLE 1 Meta-analytic Correlations between Perceptions of Organizational Politics, Strain, Morale, and Performancea, b 1 2 3 4 5 6 71. Perceptions of organizational politics2. Strain .48b k 21 N 7,1403. Job satisfaction .57c .45i k 45 72 N 16,640 22,1064. Affective commitment .54d .31j .68o k 33 32 36 N 11,633 10,808 12,2695. Task performance .20e .21k .30p .18t k 14 30 242 87 N 3,397 6,769 44,518 20,9736. OCB, individual .16f .23l .26q .25u .66x k 9 15 22 42 14 N 1,913 5,194 5,549 10,747 4,8317. OCB, organization .20g .25m .24r .25v .72y .72aa k 9 16 20 42 13 44 N 1,913 3,893 5,189 10,747 4,958 10,6478. Turnover intentions .43h .31n .58s .58w .16z .21ab .22ac k 27 63 70 97 38 7 7 N 8,439 21,056 23,603 41,002 7,643 1,344 1,344 a All correlations were corrected for attenuation due to unreliability. If more than one study reported on the same relationship, we usedthe estimate reflecting the greatest amount of data (in most cases, it was the most recent data). “k” is the number of effect sizes; N is totalobservations. b The letter superscripts in the body of the table indicate the source of the meta-analytic correlations as follows: “i,” “j,” “n,” “o,” “s,”Podsakoff, LePine, and LePine (2007); “k,” LePine, Podsakoff, and LePine (2005); “l,” “m,” Matias, Chang, and Johnson (2007); “p” Judge,Thoresen, Bono, and Patton (2001); “q,” “r,” Organ and Ryan (1995); “t,” “u,” “v,” Harrison, Newman, and Roth (2006); “w,” Cooper-Hakinand Viswasvaran (2005); “x,” “y,” “aa,” Hoffman, Blair, Meriac, and Woehr (2007); “z,” Zimmerman and Darnold (2009). Original analyses in the current paper include “b,” “c,” “d,” “e,” “f,” “g,” “h,” “ab,” and “ac.” Detailed information for relationships “b”through “h” can be found in Table 2. For more detailed results for relationships “ab” and “ac,” please contact the first author.icant (p .05). In addition, all of the 95% credibil- ror and measurement unreliability, which ac-ity intervals excluded zero, suggesting that all counted for 21 percent of the variance inbivariate relationships were in the anticipated di- correlations, the population correlation estimaterections. Although the significant Q-statistics indi- was .57. Unpublished studies yielded larger ef-cated that there were between-study moderators for fect sizes ( .61) than published studies (relationships between politics and nonperfor- .57; Z 2.87, p .01), and U.S. samples re-mance outcomes, these moderators were likely to sponded more negatively to perceptions of organi-affect only the magnitude, rather than the direction, zational politics ( .58) than Israeli samplesof the relationships, as the credibility interval ex- ( .46; Z 5.93, p .001). Similarly, affectivecluded zero. commitment had a sample-weighted mean correla- Proximal outcomes. The sample-weighted mean tion coefficient of .43 with perceptions of organ-correlation between perceptions of organizational izational politics. Sampling error and measurementpolitics and strain was .39. Sampling and measure- unreliability accounted for 14 percent of the vari-ment error accounted for 16 percent of the variance ance in the correlations between perceptions ofin correlations. After correcting for sampling and organizational politics and commitment, and themeasurement error, the population correlation was.48. The Q was significant. However, the three be- corrected population correlation was .54. The un-tween-study moderators that were tested did not published studies also had larger effect sizes (account for differences between effect sizes, as the .62) than the published studies ( .52; Zsubgroup analysis yielded nonsignificant results. 6.35, p .001), and perceptions of organizationalThe sample-weighted mean correlation between politics were more strongly related to commitmentperceptions of organizational politics and job satis- in U.S. samples ( .56) than in Israeli samplesfaction was .47. After correction for sampling er- ( .34; Z 9.81, p .001).
788 Academy of Management Journal August TABLE 2 Meta-analytic Results for Bivariate Relationships between Perceptions of Organizational Politics and Outcome Variablesa 95% CI 95% CV Variables and Moderators k N r s.d. %s.e. Lower Upper Lower Upper Q ZStrain Overall 21 7,140 .39 .48 .15 11.99 .34 .45 .19 .76 130.99*** Publication status 0.34 Unpublished studies 3 742 .40 .49 .10 27.50 .30 .52 .29 .69 10.41** Published studies 18 6,398 .39 .48 .15 10.99 .33 .46 .18 .77 117.48*** Sample 0.43 Employed student samples 5 1,135 .37 .46 .09 35.59 .29 .46 .28 .63 11.82* Employee samples 16 6,005 .40 .48 .15 9.94 .33 .47 .18 .78 117.91*** Country 0.77 Israel 3 541 .39 .50 .01 91.45 .32 .46 .48 .51 3.01 United States 15 5,676 .40 .49 .16 9.44 .33 .48 .18 .79 115.59***Job satisfaction Overall 45 16,640 .47 .57 .13 11.23 .51 .44 .83 .32 218.98*** Publication status 2.87** Unpublished studies 7 2,597 .50 .61 .08 19.05 .57 .44 .77 .46 14.21* Published studies 38 14,043 .46 .57 .14 10.68 .50 .43 .84 .30 211.70*** Sample 0.52 Employed student samples 8 1,861 .46 .54 .00 84.68 .50 .42 .54 .54 7.83 Employee samples 37 14,779 .47 .58 .14 9.44 .51 .43 .86 .31 198.48*** Country 5.93*** Israel 7 1,414 .35 .46 .06 54.03 .41 .28 .60 .33 11.30 United States 35 14,671 .48 .58 .13 10.43 .52 .44 .82 .33 203.04***Affective commitment Overall 33 11,633 .43 .54 .16 13.58 .47 .38 .86 .22 243.00*** Publication status 6.35*** Unpublished studies 5 2,271 .50 .62 .11 13.67 .59 .42 .83 .41 30.47*** Published studies 28 9,362 .41 .52 .17 10.49 .46 .36 .84 .19 199.59*** Sample 1.01 Employed student samples 6 1,525 .42 .52 .05 57.30 .46 .38 .62 .41 10.06 Employee samples 27 10,108 .43 .54 .17 8.38 .48 .37 .88 .20 219.54*** Country 9.81*** Israel 7 1,414 .26 .34 .11 38.13 .34 .18 .55 .13 17.37* US 26 10,219 .45 .56 .15 10.14 .50 .40 .85 .27 176.64***Turnover intentions Overall 27 8,439 .36 .43 .11 20.28 .32 .40 .21 .66 110.45*** Publication status 1.78 Unpublished studies 3 895 .40 .48 .03 75.05 .35 .46 .42 .54 3.69 Published studies 24 7,544 .36 .43 .12 19.07 .31 .40 .19 .66 103.63*** Sample 1.54 Employed student samples 4 1,080 .38 .47 .04 60.61 .33 .44 .39 .56 5.36 Employee samples 23 7,359 .36 .43 .12 18.38 .31 .40 .19 .66 100.08***Task performance Overall 14 3,397 .13 .20 .07 63.63 .16 .09 .34 .05 21.81OCB, individual Overall 9 1,913 .14 .16 .03 95.11 .18 .09 .20 .13 9.37OCB, organization Overall 9 1,913 .16 .20 .00 100.00 .20 .12 .20 .20 7.57 a k is the number of effect sizes; N is the total number of subjects; r is the mean sample-weighted correlation; is the estimate of the fullycorrected population correlation; s.d. is the standard deviation of the estimate of the fully corrected population correlation; %s.e. is thepercentage of observed variance accounted for by sampling and measurement error; 95% CI is the 95% confidence interval aroundthe mean sample-weighted correlation; 95% CV is the 95% credibility interval around the corrected mean population correlation; Qis the chi-square test for the homogeneity of true correlations across studies; and Z is the test for the significance of the differencebetween the sample-weighted correlations. * p .05 ** p .01 *** p .001
2009 Chang, Rosen, and Levy 789 Distal outcomes. The sample-weighted mean than .08 (Hu & Bentler, 1999). The overall measure-correlation between perceptions of organizational ment model had reasonably good fit to the datapolitics and turnover intentions was .36. After cor- (CFI .98, TLI .94, RMSEA .12, SRMR .02).recting for sampling and measurement artifacts, We then tested the proposed model in whichwhich explained 24 percent of the variance in cor- effects of perceptions of organizational politics onrelations, the estimated population correlation was turnover intentions and job performance were fully.43. Because of the significant Q, we tested for mediated by strain and morale. As reported in Ta-between-study moderators and found that neither ble 3, the model fit the data (CFI .97, TLI .94,publication status nor sample type moderated the RMSEA .09, SRMR .02), and all paths weremagnitude of this relationship. This result sup- significant, except for the path linking strain toported Hypothesis 3. The sample-weighted mean turnover intentions ( .02). We then tested forcorrelation between perceptions of organizational partial mediation effects by including direct pathspolitics and task performance was .13. Sampling from perceptions of organizational politics to theerror and measurement unreliability accounted for two distal outcome variables in the next three al-64 percent of the variance in observed correlations ternative models (see Figure 1). As reported in Ta-across studies. The corrected population correla- ble 3, the three alternative models had essentiallytion was .20. For the relationship between per- the same fit indexes as the theoretical model. Fur-ceptions of organizational politics and OCB toward ther analyses revealed that adding a path betweenindividuals (OCBI), the sample-weighted mean cor- perceptions of organizational politics and turnoverrelation was .14, and the corrected population intentions (alternative model 1: 2 0.59, n.s.), 1correlation was .16. Sampling error and measure- or between politics and performance (alternativement unreliability accounted for 86 percent of the model 2: 2 1.24, n.s.) did not improve model 1variance observed in correlations. The sample- fit over that of the theoretical model. When bothweighted mean correlation between perceptions of paths were freely estimated (alternative model 3:organizational politics and OCB toward one’s or- 2 2 2.35, n.s.), none of the direct paths wereganization (OCBO) was .16, the corrected popu- significant, nor did model fit improve significantly.lation correlation was .20, and sampling error and Thus, the theoretical model received support, andmeasurement unreliability accounted for 100 per- the results provided evidence that strain and mo-cent of the variance observed in these correlations. rale fully mediated the effects of perceptions ofThus, perceptions of organizational politics had organizational politics on performance and turn-negative relationships with all three supervisor- over intentions.rated performance measures, which supported Hy- As shown in Figure 3, perceptions of organiza-pothesis 4. Additionally, none of the 95% credibil- tional politics were associated with increased psy-ity intervals included zero, and all three chological strain ( .48, p .05), supportingQ-statistics were nonsignificant, indicating that themagnitudes of these relationships were not affected Hypothesis 1. Supporting Hypothesis 2, percep-by between-study moderators. tions of organizational politics were related to mo- rale both directly ( .57, p .05) and indirectly through strain ( .20, p .05). In terms of mediation effects, we found support for HypothesisModel Testing 5b: effects of morale fully mediated perceptions of To test the proposed model, we first created la- organizational politics on turnover intentions (tent constructs to represent morale and job perfor- .70, p .05). Hypothesis 5a was supported by amance. Job satisfaction and affective commitment more extended mediational chain; higher percep-served as indicators of morale. The job performance tions of organizational politics were associatedconstruct included task performance, OCBI, and with increased strain and then reduced morale,OCBO as indicators. To evaluate these latent con- which in turn related to increased turnover inten-structs, we conducted a confirmatory factor analy- tions. Supporting Hypothesis 6, the relationshipsis (CFA) using maximum likelihood estimation in between perceptions of organizational politics andMplus 4.2 (Muthen & Muthen, 2005). The following ´ ´ performance was fully mediated by both strain (criteria were used to assess model fit: comparative .14, p .05) and morale ( .28, p .05).fit (CFI) and Tucker-Lewis (TLI) index values Overall, our results demonstrated that strain andgreater than .90, a root-mean-square error of ap- morale fully mediated the effects of perceptions ofproximation (RMSEA) of less than .06, and a stan- organizational politics on turnover intentions anddardized root-mean-square residual (SRMR) of less performance.
790 Academy of Management Journal August TABLE 3 Fit Statistics for Alternative Models 2 2 Model df CFI TLI RMSEA SRMRPerceptions of organizational politics as the only predictor a Theoretical 587.00 15 .97 .94 .09 .02 Alternative model 1b 586.41 14 .97 .94 .09 .02 0.59e Alternative model 2c 585.76 14 .97 .95 .08 .02 1.24e Alternative model 3d 584.45 13 .97 .95 .08 .02 2.35eTwo conceptualizations of hindrance stressor f Unified hindrance stressor 1,526.87 29 .94 .90 .10 .03 Diversified hindrance stressors: Perceptions 1,333.76 25 .93 .89 .10 .03 193.11k*** of organizational politics, role ambiguity, and role conflict Alternative model 4g 1,332.60 23 .93 .88 .11 .03 1.16l Alternative model 5h 1,266.28 23 .94 .89 .11 .02 67.48l*** Alternative model 6i 1,270.00 24 .94 .89 .10 .03 3.72m Alternative model 7j 1,128.11 22 .94 .89 .10 .03 141.89n*** a CFI comparative fit index; TLI Tucker-Lewis index; RMSEA root-mean-square error of approximation; SRMR standardizedroot-mean-square residual. N 5,160. b Estimate of the direct effect from perceptions of organizational politics to turnover intentions with other paths held constant. c Estimate of the direct effect from perceptions of organizational politics to performance with other paths held constant. d Estimate of the direct effects from perceptions of organizational politics to distal outcomes with other paths held constant. e Model fit compared with the theoretical model. f N 4,865. g Estimate of the direct effects from perceptions of organizational politics to distal outcomes with other paths held constant. h Estimate of the direct effects from role ambiguity to distal outcomes with other paths held constant. i Estimate of the direct effects from role ambiguity to turnover intentions with other paths held constant. j Estimate of the direct effects from role ambiguity to turnover intentions and role conflict to distal outcomes with other paths heldconstant. k Model fit compared with the unified hindrance stressor model. l Model fit compared with the diversified hindrance stressors model. m Model fit compared with alternative model 5. n Model fit compared with alternative model 6. * p .05 ** p .01 *** p .001Perceptions of Organizational Politics, Role for almost all the variables except for strain, OCB1,Ambiguity, and Role Conflict and turnover intentions. These results indicate that perceptions of organizational politics had different To explore the distinctiveness of perceptions oforganizational politics from other hindrance stres- relationships with role stressors, and that percep-sors, we first compared the bivariate meta-analytic tions of organizational politics– outcome relation-estimates of relationships between politics, role ships were typically stronger, if not comparable, toconflict, and role ambiguity, and between these the role stressor– outcome relationships.stressors and outcome variables (see Table 4). Per- Next, we explored two different conceptualiza-ceptions of organizational politics had a stronger tions of the hindrance stressor construct within therelationship with role conflict than with role ambi- context of the proposed model (Figure 2). Table 3guity ( .58 vs. .52, Z 3.04, p .001). For summarizes the results of these nested model tests.stressor-outcome relationships, perceptions of or- The diversified model had better fit ( 24 196.32,ganizational politics had significantly stronger as- p .001) than the unified model (CFI .94, TLIsociations with strain, job satisfaction, affective .90, RMSEA .10, SRMR .03), suggesting thatcommitment, and OCB0, than role ambiguity. Ad- perceptions of organizational politics and roleditionally, the perceptions of organizational poli- stressors are best viewed as distinct yet relatedtics– outcome relationships were significantly stressors. We also evaluated possible partial medi-stronger than role conflict– outcome relationships ation effects (see Table 3). Although the effects of
2009 Chang, Rosen, and Levy 791 FIGURE 3 Final Model of Effects of Perceptions of Organizational Politics on Employee Outcomesa a All path coefficients and loadings are significant at p .05 except for the italicized coefficient, for which p .05; numbers inparentheses represent the lower and upper bounds for the 95% confidence interval for path coefficients.perceptions of organizational politics on the two strain. The 95% confidence intervals for the effectsdistal outcomes appeared to be fully mediated (al- of politics and role ambiguity overlapped, but role 2ternative model 4: 2 1.16, n.s.), the effects of conflict had the strongest association with strain.role ambiguity were partially mediated (alternative Paths from perceptions of organizational politics 2model 5: 2 67.48, p .001). However, only ( .43), role ambiguity ( .22), and rolethe direct path from role ambiguity to turnover conflict ( .09) to morale were all significant.intentions was significant. Thus, we dropped the None of their confidence intervals overlapped, in-other direct path from the model, which did not dicating that the politics-morale link was stronger 2impact the model fit (alternative model 6: 1 than the other two links. In addition, though strain3.72, n.s.). Finally, we added paths from role con- and morale fully mediated the effects of percep-flict to both outcomes, and both paths were signif- tions of organizational politics on distal outcomes,icant. This final model (Figure 4) showed improve- role ambiguity had a direct link with turnover in-ment in fit over alternative model 6 ( 22 141.89, tentions ( .10), and role conflict had directp .001) and included three direct paths from role associations with turnover intentions ( .16) andambiguity and role conflict to distal outcomes. performance ( .06). Overall, these significant Perceptions of organizational politics ( .18), differences in path coefficients and distinct pat-role ambiguity ( .20), and role conflict ( .33) terns of mediation effects help distinguish percep-each had significant, positive relationships with tions of organizational politics from role stressors.
792 Academy of Management Journal August TABLE 4 distinguishable from other role-based hindrance Comparison of Meta-analytic Relationships among stressors. In keeping with Miller et al.’s (2008)Hindrance Stressors and between Hindrance Stressors meta-analysis of the outcomes of perceptions of and Employee Attitudes and Behaviorsa organizational politics, the results of the current Outcome and Hindrance study demonstrated that such perceptions have Stressors k N Z strong, positive relationships with strain and turn- over intentions and strong, negative relationshipsPerceptions of organizational politics with job satisfaction and affective commitment. Role ambiguityb .52 6 3,504 However, the current study extends previous re- Role conflictb .58 4 1,941 3.04*** search by providing unequivocal support for a re-Strain lationship between perceptions of organizational Perceptions of organizational .48 21 7,140 politics and aspects of job performance that were politics Role ambiguityc .43 8 1,435 2.05* not clearly supported (viz., task performance) or Role conflictc .52 7 1,220 1.84 tested (viz., OCB) in Miller et al.’s (2008) meta-Job satisfaction analysis. Perceptions of organizational .57 45 16,640 Beyond the basic bivariate estimates, our results politics Role ambiguityc .48 42 10,062 10.34*** also provided compelling evidence supporting a Role conflictc .49 39 9,780 9.23*** theoretically derived model that integrates theAffective commitment stress- and social exchange– based explanations of Perceptions of organizational .54 33 11,633 the effects of perceptions of organizational politics. politics Role ambiguityc .39 12 3,774 10.07*** In particular, perceptions of organizational politics Role conflictc .30 9 3,225 12.94*** were associated with increased psychologicalTask performance strain, which was associated directly with reduced Perceptions of organizational .20 14 3,397 performance, as well as indirectly with increased politics Role ambiguityc .22 18 4,301 1.05 turnover intentions through reduced morale. Polit- Role conflictc .14 16 4,057 2.66** ical perceptions also had a direct, negative linkOCB, individual with employee morale, which was related to in- Perceptions of organizational .16 9 1,913 creased turnover intentions and reduced perfor- politics Role ambiguityd .16 10 2,651 0.00 mance. These findings revealed that strain and mo- Role conflictd .15 7 2,351 0.33 rale fully mediate the effects of perceptions ofOCB, organization organizational politics on important employee re- Perceptions of organizational .20 9 1,913 actions. In addition, they indicated that the stress politics Role ambiguityd .12 7 2,456 2.69*** and social exchange perspectives complement each Role conflictd .14 6 2,156 1.97* other. Thus, simultaneously considering the medi-Turnover intentions ating effects of morale and strain provides a more Perceptions of organizational .43 27 8,439 complete picture of the intrapersonal processes politics Role ambiguityc .44 8 1,188 0.29 that relate perceptions of organizational politics to Role conflictc .45 8 1,188 0.53 distal employee outcomes. Interestingly, results of the exploratory analyses a The letter superscripts in the body of the table indicate the suggested that political perceptions are distinctsource of the meta-analytic correlations as follows: “b,” current ¨study; “c,” Ortqvist and Wincent (2006); “d,” Podsakoff, Mac- from at least two other hindrance stressors––roleKenzie, and Bommer (1996). ambiguity and role conflict. Perceptions of organi- * p .05 zational politics had different relationships with ** p .01 those role stressors. Also, bivariate relationships *** p .001 between political perceptions and outcomes were almost always stronger than or comparable with the DISCUSSION role stressor– outcome relationships. Finally, when The current research had three goals: (1) to ad- considered together, perceived politics had adress inconsistencies in the research findings on unique pattern of associations with employee out-perceptions of organizational politics, (2) to exam- comes: the effects of perceptions of organizationaline a model that incorporated stress and social politics on distal outcomes were fully mediated byexchange explanations of reactions to perceptions strain and morale, whereas the effects of role stres-of organizational politics, and (3) to explore sors on distal outcomes were only partiallywhether perceptions of organizational politics were mediated.
2009 Chang, Rosen, and Levy 793 FIGURE 4 Final Model of Effects of Perceptions of Organizational Politics, Role Ambiguity, and Role Conflict on Employee Outcomesa a All path coefficients and loadings are significant at p .05, except for the italicized coefficient, for which p .05; numbers inparentheses represent the lower and upper bounds for the 95% confidence interval for path coefficients.Theoretical Implications themselves in their work or by increasing the extent to which they are involved in their jobs. Similarly, This research offers a number of important theo- our findings suggest that previous studies showingretical contributions. First, the significant relation-ships between perceptions of organizational poli- positive associations between perceptions of organ-tics and the focal outcomes (i.e., psychological izational politics and desirable outcomes (e.g., Fer-strain, morale, turnover intentions, and perfor- ris et al., 1993; Ferris & Kacmar, 1992) representmance) provide additional support for the notion exceptions in the literature and may be due tothat, in response to perceptions of organizational statistical artifacts. In general, employee percep-politics, employees are likely to withdraw from an tions of self-serving, illegitimate political activitiesorganization in order to avoid political “games.” at work have consistently negative relationshipsMoreover, our findings clearly linked perceptions with employee attitudes and behaviors.of organizational politics with task performance Regarding the moderator analyses, we found lit-and OCB, indicating that, overall, perceived poli- tle evidence for publication bias (Rosenthal, 1979),tics represent an aversive aspect of the work envi- as unpublished studies had either stronger effectronment. These findings also counter theoretical sizes than published ones, or comparable effectarguments (see Ferris et al., 1989; Ferris & Kacmar, sizes. In addition, effect sizes from studies with1992) suggesting that employees may respond to employed students were similar to those from stud-perceptions of organizational politics by immersing ies with full-time employees. This pattern sug-
794 Academy of Management Journal Augustgested that the employment relationships that de- ways of arranging the outcomes of perceptions ofvelop between employed students and their organizational politics that go beyond treating thememployers may be as meaningful and important as all as direct outcomes, as has been implied by pre-relationships that develop between full-time em- vious research (e.g., Ferris et al., 2002; Miller et al.,ployees and their organizations and that percep- 2008). Interestingly, the effects of perceptions oftions of organizational politics have similar impli- organizational politics on turnover intentions andcations for both groups. performance worked through slightly different However, we did observe cross-cultural differ- pathways. In particular, the psychological strainences between U.S. and Israeli samples, in that elicited by perceptions of organizational politicsperceptions of organizational politics had stronger was associated with decreased morale, which wasrelationships with morale for U.S. employees. This related to higher turnover intent. This pattern im-pattern is consistent with Vigoda’s (2001) proposi- plies that the effects of perceptions of organization-tion that, because of their experiences with geopo- al politics on turnover intentions may take longerlitical conflict, Israeli employees are better condi- to unfold and may involve a more rational process.tioned for coping with the interpersonal conflict Turnover researchers have conceptualized the typ-associated with organizational politics. Similarly, ical voluntary turnover process as initiated by lowRomm and Drory (1988) suggested that Israelis morale and involving multiple decision pointshave greater familiarity with political processes in- (Griffeth, Hom, & Gartner, 2000). An alternativeside work and outside of work, which may be as- viewpoint, the unfolding model of turnover (Lee &sociated with (1) greater tolerance for political ac- Mitchell, 1994), suggests that a new situation ortivities as a means of getting ahead and (2) a belief event, in addition to morale, triggers employeethat organizational politics are normative and mor- turnover. However, according to this model, em-ally legitimate. These ideas are consistent with the ployees may still follow extensive decision-makingnotion that national culture influences the mental paths leading to turnover, and three out of the fourprograms that guide employees’ interpretations and proposed paths involve job satisfaction. Thus, ourreactions to different aspects of their jobs (Hof- results are consistent with previous work concern-stede, 1980). As such, these cross-cultural differ- ing the formulation of turnover intentions. They areences may contribute to understanding of the gen- also consistent with other studies (e.g., Podsakoff eteralizability of theories linking perceptions of al., 2007) in demonstrating that the effects of hin-organizational politics to employee outcomes. For drance stressors on turnover intentions work firstexample, our findings imply that U.S. and Israeli through strain and then through morale.employees may have different expectations thatguide the evaluations of their organizational ex- On the other hand, the effect of perceptions ofchange relationships. Thus, to the extent that low organizational politics on performance workedlevels of politics are central to employees’ work through both strain and morale simultaneously.expectations, perceptions of organizational politics The pathway through strain coincides with theo-will be salient and represent a more serious viola- ries suggesting that strain and other negative affec-tion of their social exchange relationships. How- tive experiences have an impact on motivation andever, additional research is necessary to determine performance (Lord & Kanfer, 2002). For example,whether differences in exchange expectations ac- the resource allocation perspective (Kanfer & Ack-count for cross-cultural disparities in relationships erman, 1989) implies that strain drains mental re-between perceptions of organizational politics and sources that could otherwise be devoted to self-morale. regulatory activities associated with job Another contribution of this study is that it sub- performance. According to Kanfer and Ackerman,stantiated our arguments that multiple pathways self-regulatory activities, such as goal striving andlink perceptions of organizational politics to em- feedback monitoring, require effortful processingployee outcomes. Model-testing results demon- and mental resources. As employees experiencestrate that mediators proposed by work stress (i.e., strain associated with perceptions of organizationaljob anxiety) and social exchange (i.e., morale) the- politics, they may devote energy to coping withories explain relationships between politics and their negative affect, thereby reducing the resourcesboth performance and withdrawal intentions. they can spare for regulating performance. In addi-These findings were supportive of theory and pro- tion, perceptions of organizational politics werevided evidence that both perspectives contribute to associated with lower morale, which led to reducedunderstanding of how social context affects atti- job performance. Our results imply that, as a resulttudes and behaviors. Additionally, our results of perceiving politics, employees may begin tohighlight the importance of considering alternative view their organizations as risky investments and
2009 Chang, Rosen, and Levy 795may demonstrate lower levels of morale, and also their unique effects. Moreover, unlike the effects ofdecrease their contributions to their jobs. perceptions of politics, the effects of role stressors Finally, the results of exploratory analyses on turnover intentions and performance were notshowed that perceptions of organizational politics completely accounted for by the stress- and socialare distinct from both role ambiguity and role con- exchange– based paths, which suggests that addi-flict. In addition to the stronger bivariate correla- tional mediating mechanisms explain the effects oftions with outcomes demonstrated by political per- role stressors, but not those of perceptions of organ-ceptions, the mechanisms underlying the effects of izational politics.perceptions of organizational politics on turnoverintentions and performance were dissimilar tothose underlying the effects of role stressors. These Managerial Implicationsfindings point to the possibility that perceptions oforganizational politics may be qualitatively differ- Our results have several practical implications.ent from role-based hindrance stressors. In partic- First, leaders should recognize that, though someular, perceptions of organizational politics repre- political activities may be essential to the function-sent evaluations of social aspects of organizational ing of work groups (Fedor et al., 2008), their ownsettings (i.e., witnessing members politicking and political activities may have unanticipated con-receiving rewards), rather than the assessments of sequences at the individual level. For instance,personal situations (i.e., comparing individuals’ job leaders often make “idiosyncratic deals” withdemands to their coping resources) that character- employees as a means of optimizing individualize role stressors (Boswell et al., 2004). In addition, performance and reducing turnover (Rousseau, Ho,employees experience role ambiguity and conflict & Greenberg, 2006). The current research demon-because they are concerned with fulfilling their strates that if these activities are perceived as po-roles as stipulated by their organization, whereas litical (i.e., based on favoritism and self-interest),perceptions of organizational politics are associ- then they may have extensive negative effects onated with observing behaviors that are self-serving organization members. Therefore, it is importantand threatening to the well-being of other employ- for top management to make decisions that balanceees (Kacmar & Baron, 1999). Thus, although per- the costs and benefits of engaging in behaviors thatceived politics had seemingly similar effects on may be perceived as political.outcomes as the broadly defined hindrance stressor Next, we demonstrated that employees respondconstruct examined in previous studies (e.g., negatively to work conditions that indicate politics.LePine et al., 2005; Podsakoff et al., 2007), the Thus, it behooves managers to focus on social con-similar effects may be attributable to perceptions of text when attempting to understand employee atti-organizational politics serving as the dominant in- tudes and motivation. For example, managers maydicator of the unified hindrance stressor construct, reduce perceptions of organizational politics andthereby driving the effects of the construct. Thus, subsequent deficits in morale and motivation byfuture studies should explore whether (and when) providing clear feedback regarding which behav-it is appropriate to consider perceptions of organi- iors their organization desires (Rosen et al., 2006),zational politics separately from other hindrance by reducing incentives for employees to engage instressors. For example, if the goal of research is to political activities, or by aligning individual andpredict general work attitudes and behaviors, then organizational goals (see Witt, 1998). In extremeit may be prudent to conceptualize perceptions of situations, it may benefit an organization to targetorganizational politics and role stressors as part of key political players whose activities are especiallythe unified, general hindrance stressor construct. salient and damaging. If these individuals are notThis approach is akin to considering different willing to reduce their political activities, then theyforms of justice (viz., distributive, procedural, in- should be removed from the organization. Al-terpersonal, and informational) as indicators of an though extreme, such tactics may benefit employeeoverall justice evaluation (see Ambrose & health and performance in the long run.Schminke, 2006). However, if the goal of a study is Asking managers to monitor and reduce theirto understand more specifically how the interplay own political activities may not always be a realis-between these stressors (e.g., the effects of percep- tic solution, and firing employees for being tootions of organizational politics may work through political may carry some risk in today’s litigiousrole ambiguity, or role ambiguity may exacerbate society. Therefore, we recommend that human re-effects of such perceptions) relates to employee source departments actively create competencyoutcomes, then researchers may want to consider models (see Shippmann et al., 2000) that incorpo-these hindrance stressors separately to capture rate the goals of discouraging political activities