Upload
indah-sebayang
View
213
Download
0
Embed Size (px)
DESCRIPTION
Jurnal Penelitian
Citation preview
PERSONNEL PSYCHOLOGY2005, 58, 281342
CONSEQUENCES OF INDIVIDUALS FIT AT WORK:A META-ANALYSIS OF PERSONJOB,PERSONORGANIZATION, PERSONGROUP,AND PERSONSUPERVISOR FIT
AMY L. KRISTOF-BROWN, RYAN D. ZIMMERMAN, ERIN C. JOHNSONHenry B. Tippie College of Business
Department of Management & OrganizationsUniversity of Iowa
This meta-analysis investigated the relationships between personjob(PJ), personorganization (PO), persongroup, and personsupervisorfit with preentry (applicant attraction, job acceptance, intent to hire,job offer) and postentry individual-level criteria (attitudes, performance,withdrawal behaviors, strain, tenure). A search of published articles,conference presentations, dissertations, and working papers yielded 172usable studies with 836 effect sizes. Nearly all of the credibility intervalsdid not include 0, indicating the broad generalizability of the relation-ships across situations. Various ways in which fit was conceptualized andmeasured, as well as issues of study design, were examined as modera-tors to these relationships in studies of PJ and PO fit. Interrelationshipsbetween the various types of fit are also meta-analyzed. 25 studies usingpolynomial regression as an analytic technique are reviewed separately,because of their unique approach to assessing fit. Broad themes emerg-ing from the results are discussed to generate the implications for futureresearch on fit.
Theories of person-environment (PE) interaction have been prevalentin the management literature for almost 100 years (e.g., Ekehammer, 1974;Lewin, 1935; Murray, 1938; Parsons, 1909; Pervin, 1968), making it oneof the more venerable lines of psychological theorizing (Dawis, 1992). Itis against this interactionist backdrop that the concept of PE fit emerged.Described as a syndrome with many manifestations (Schneider, 2001,p. 142), PE fit is broadly defined as the compatibility between an individualand a work environment that occurs when their characteristics are wellmatched. Despite, or perhaps because of, the simplicity of this definition,several distinct types of fit have garnered attention. Much emphasis has
The authors are indebted to the encouragement & feedback provided by SAPS and FrankSchmidt.
Correspondence and requests for reprints should be addressed to Amy L. Kristof-Brown,University of Iowa, Henry B. Tippie College of Business, Department of Management& Organizations, 108 Pappajohn Business Building, Iowa City, IA 52242; [email protected] C 2005 BLACKWELL PUBLISHING, INC.
281
282 PERSONNEL PSYCHOLOGY
been placed on the match between peoples interests and those of othersin a vocation (e.g., Holland, 1985). However, other types of fit, such as anindividuals compatibility with his or her job, organization, work group,and supervisors have also emerged as important research domains.
Although research on these other types of fit has been prolific, rarelyhas it been synthesized to draw conclusions about the true impact of fiton individual-level outcomes. Some progress toward integration has beenmade. However, most reviews have been nonquantitative, not differenti-ated between various types of fit, or focused solely on single types of PEfit. Edwards (1991) provided a comprehensive qualitative review of PE fitstudies published between 1960 and 1989, but did not include a quantita-tive analysis of the strength of the reported relationships. Furthermore, hisreview did not distinguish between fit with the job, group, organization,or vocation, which have been shown to have unique relationships with at-titudes and behaviors (Cable & DeRue, 2002). Another qualitative reviewconducted almost 10 years ago by Kristof (1996) focused exclusively onthe personorganization fit literature. A portion of this literature was re-cently meta-analyzed but only for a limited range of criteria (job attitudes)and a small number of effect sizes (k = 21; Verquer, Beehr, & Wagner,2003). Furthermore, little attention has been paid to the more nascentarea of persongroup fit or to related topics such as personsupervisor fit.Given the qualitative focus, limited scope, and somewhat dated nature ofthese past summaries, a quantitative analysis of various types of PE fit iswarranted.
The biggest challenge to this type of analysis is the proliferation ofconceptualizations, measures, and analytic approaches that make fit anelusive construct (Judge & Ferris, 1992). Fit has alternatively been con-ceptualized as similarity, needsatisfaction, and demandability match.Further, it has been operationalized using a variety of content dimensions,including skills, needs, preferences, values, personality traits, goals, andattitudes. Strategies for measuring fit also vary widely, from directly askingindividuals to report their perceived fit to researchers indirectly assessingfit through explicit comparisons of separately rated P and E character-istics. When indirect assessments are used, E may be the subjective en-vironment as assessed by the individual or the objective environment asdetermined by other sources. Moreover, the environment can be concep-tualized as an entity with its own unique characteristics or as an aggregateof its members characteristics. Sampling strategies and the potential forcommon method bias also vary widely across studies, rendering inter-pretation of some results problematic. Although this diversity poses achallenge to integration, it may also produce systematic variations in re-ported fitoutcome relationships. Much of this meta-analysis, therefore, isconcerned with articulating theory-driven rationale for moderators of the
AMY L. KRISTOF-BROWN ET AL. 283
fitoutcome relationships and evaluating the empirical evidence regardingtheir impact.
Accordingly, the purpose of this study is to empirically summarize theexisting literature in four critical domains of PE fit: personjob, personorganization, persongroup, and personsupervisor fit. Specifically, weaddress the relationship between these types of fit and individual-levelpreentry outcomes (applicant attraction, job acceptance, intent to hire,job offer) and postentry consequences (attitudes, withdrawal behaviors,strain, performance, and tenure). Delving deeper into these meta-analyticfindings, we examine multiple moderators of fitoutcome relationshipsincluding various conceptualizations, operationalizations, and measure-ment approaches, as well as study design. In addition, because of the re-cent growth in studies using a polynomial regression approach (Edwards,1993, 1994) to study fit indirectly, we also review these studies. However,because the interpretation of the multiple correlations from polynomial re-gressions is ambiguous, we describe the general forms of the relationshipsdepicted in these studies, rather than meta-analyze them. This allows us tocomprehensively review the fit literature before drawing conclusions. Inthe discussion section we review the implications of our findings, proposea coherent agenda for future research, offer methodological guidance forfuture studies of fit, and present practical guidance for how to capitalizeon the benefits of fit.
Types of PersonEnvironment Fit
Based in the tradition of interactional psychology, the notion of peoplebeing differentially compatible with jobs, groups, organizations, and voca-tions is almost axiomatic. The concept of PE fit has been described as, sopervasive as to be one of, if not the dominant conceptual forces in the field(Schneider, 2001, p. 142). It is not surprising, therefore, that managementscholars have a sustained interest in the fit between individuals and theirwork environments. Perhaps in part because of this interest, research onfit continues to be one of the most eclectic domains in management. Webriefly review the primary types of fit to establish and articulate the criteriaused for inclusion and exclusion of studies in our analyses.
Person Environment Interactions
Interactional psychology includes numerous studies that assess PE fitas a statistical interaction between the person and environment. Person-ality may be investigated as a moderator of environmental forces, as inthe case of growth need strength moderating the job characteristicsjobsatisfaction relationship (e.g., Fried & Ferris, 1987; Loher, Noe, Moeller,
284 PERSONNEL PSYCHOLOGY
& Fitzgerald, 1985), or situations may be viewed as moderators of thepersonalityoutcome relationships (e.g., Colbert, Mount, Harter, Witt, &Barrick, 2004; Turban, Lau, Ngo, Chow, & Si, 2001). Despite their concep-tual appeal, moderator studies rarely conceptualize the person and environ-ment on commensurate dimensions. Without this standard it is impossibleto directly compare P and E values, a fundamental property of the PEfit theory (Edwards, Caplan, & Harrison, 1998; French, Rogers & Cobb,1974). Thus, we address only studies that measure P and E on commen-surate dimensions. Even when commensurate dimensions are used, thereare additional complications with P E interaction studies (e.g., Arnaud,Ambrose, & Schminke, 2002; Goodman & Syvantek, 1999; Judge & Bretz,1992; Lievens, Decaesteker, & Coetsier, 2001). First, multiple correlationsthat are typically reported as effect sizes include main effects as well asinteraction effects, rendering them inconsistent with other fit measures.Second, the interaction between person and environment variables doesnot reflect their proximity to one another (Edwards et al., 1998, p. 41), anecessary precondition for fit.
Person-Vocation Fit
PE fit research is generally characterized by matching individuals tovarious levels of their work environment (Judge & Ferris, 1992; Kristof,1996). The broadest of these levels is the vocation or occupation. Thisresearch on personvocation (PV) fit includes vocational choice theo-ries that propose matching people with careers that meet their interests(e.g., Holland, 1985; Parsons, 1909; Super, 1953), and the theory of workadjustment (Dawis & Lofquist, 1984; Lofquist & Dawis, 1969), which em-phasizes that adjustment and satisfaction are the result of employees needsbeing met by their occupational environment. Because this research hasbeen extensively reviewed by others (Assouline & Meir, 1987; Spokane,1985; Tranberg, Slane, & Ekeberg, 1993), we do not include it in ouranalyses.
Personjob Fit
Instead, we emphasize an area closely related to PV fit but defined morenarrowly as the relationship between a persons characteristics and thoseof the job or tasks that are performed at work. This is the domain of personjob (PJ) fit (e.g., Edwards, 1991; Kristof, 1996). Edwards (1991) outlinedtwo basic conceptualizations of the PJ fit. The first is the demands-abilitiesfit, in which employees knowledge, skills, and abilities are commensu-rate with what the job requires. The second form of PJ fit occurs when
AMY L. KRISTOF-BROWN ET AL. 285
employees needs, desires, or preferences are met by the jobs that theyperform. This type of fit, often labeled needssupplies or suppliesvaluesfit, has been the emphasis of various theories of adjustment, well-being, andsatisfaction (Caplan, 1983; French, Caplan, & Harrison, 1974; Harrison,1978; Locke, 1969; Porter, 1961, 1962).
Having restricted our definition of fit to the compatibility that resultsfrom personal and environmental characteristics being well matched, itwas necessary to exclude research on related but distinct topics. Studiesof underemployment (e.g., Feldman, Leana, & Bolino, 2002), overquali-fication (e.g., Johnson & Johnson, 1999), role overload (e.g., Netemeyer,Burton, & Johnston, 1995), and relative deprivation (Feldman et al., 2002)were not included because measures of these concepts typically assess thedegree to which there is misfit in only one direction. Studies of met ex-pectations were also not included (e.g., Bunderson, 2001). This decisionwas based on Schneiders reasoning that a persons expectations proba-bly represent some modification of his preferences based on information(1975, p. 460). Thus, they are likely to be tempered by reality or pastexperiences, rather than reflect the individuals characteristics, which arethe basis for fit.
PersonOrganization Fit
Research on personorganization (PO) fit, which addresses the com-patibility between people and entire organizations, is the second type of fitwe review. Beginning with Toms (1971) suggestion that individuals willbe most successful in organizations that share their personalities, researchhas emphasized individualorganizational similarity as the crux of PO fit.Some research has followed Toms (1971) operationalization of PO fit aspersonalityclimate congruence (e.g., Christiansen, Villanova, & Mikulay,1997; Ryan & Schmitt, 1996); however, Chatmans (1989) seminal the-ory of PO fit focused attention primarily on values. With the subsequentvalidation of the Organizational Culture Profile (OReilly, Chatman, &Caldwell, 1991), a values-based instrument, value congruence becamewidely accepted as the defining operationalization of PO fit (Kristof, 1996;Verquer et al., 2003). A lesser used but theoretically consistent operational-ization is PO goal congruence (e.g., Vancouver & Schmitt, 1991; Witt &Nye, 1992). In all cases, the emphasis is on the compatibility betweencommensurate individual and organizational characteristics.
Excluded from our analysis are studies that assess agreement on orga-nizational culture or climate. For example, studies of climate discrepancy(e.g., Joyce & Slocum, 1982) typically ask participants to describe theclimate in their organizations. Thus, they typically do not assess howwell the climate matches participants personal characteristics or meets
286 PERSONNEL PSYCHOLOGY
their needs, as much as the extent to which an individuals perceptionconcurs with others. Similarly, climate strength studies (e.g., Schneider,Salvaggio, & Subirats, 2002) emphasize within-group consistency inclimate perceptions, rather than individuals fit with the organizationalcharacteristics. Therefore, studies in both of these areas were not deemedrelevant to this analysis.
PersonGroup Fit
The third category of research we review is persongroup (PG) orpersonteam fit, which focuses on the interpersonal compatibility betweenindividuals and their work groups (Judge & Ferris, 1992; Kristof, 1996;Werbel & Gilliland, 1999). Of all types of fit, PG fit research is the mostnascent. Despite high levels of interest in coworker similarity on demo-graphic variables (e.g., Riordan, 2000), little research has emphasizedhow the psychological compatibility between coworkers influences indi-vidual outcomes in group settings. Only a handful of published studieshave examined the fit on characteristics such as goals (Kristof-Brown &Stevens, 2001; Witt, 1998) or values (Adkins, Ravlin, & Meglino, 1996;Becker, 1992; Good & Nelson, 1971). There are, however, several studiesthat have examined PG fit on personality traits (Barsade, Ward, Turner,& Sonnenfeld, 2000; Hobman, Bordia, & Gallois, 2003; Kristof-Brown,Barrick, & Stevens, in press; Slocombe & Bluedorn, 1999; Strauss,Barrick, & Connerley, 2001).
These studies are distinct from research on team similarity or ho-mogeneity (e.g., Barry & Stewart, 1997; Jehn, Northcraft, & Neale,1999) and studies that aggregate individuals fit to the unit level (e.g.,Harrison, Shaffer, & Bhaskar, 2002; Ostroff, 1993; Salvaggio, 2004).These aggregate-level studies predict unit-level outcomes, rather thanindividual-level criteria that we emphasize. Studies by Ostroff andRothhausen (1997) and Vancouver, Millsap, and Peters (1994) demon-strate that fitoutcome relationships differ when they are assessed at higherlevels of analysis, underscoring the need to differentiate aggregate-levelfit studies from others.
Research on relational demography is also excluded from our review.Although demography assesses a persons similarity to group members topredict individual-level outcomes, the emphasis is exclusively on demo-graphic variables (such as race, age, gender). Milliken and Martins (1996)and Harrison, Price, and Bell (1998) categorize demographic variables assurface-level or easily observable attributes and contrast them with otherdeep-level, underlying characteristics such as values or goals. Harrisonet al. (1998) demonstrated that there are meaningful differences between
AMY L. KRISTOF-BROWN ET AL. 287
similarity on these types of characteristics, and that over time it is fit ondeep-level characteristics that has the greatest impact on outcomes.
PersonSupervisor Fit
A final form of PE fit exists in the dyadic relationships between in-dividuals and others in their work environments. Although dyadic fitmay occur between coworkers (e.g., Antonioni & Park, 2001), appli-cants and recruiters (e.g., Graves & Powell, 1995), and mentors andproteges (e.g., Turban & Dougherty, 1994), by far the most well-researchedarea is the match between supervisors and subordinates (e.g., Adkins,Russel, & Werbel, 1994; Van Vianen, 2000). Given the limited numberof studies on other types of dyadic fit and the importance of supervisorsubordinate relationships on work outcomes (Griffeth, Hom, & Gaertner,2001), the final area we meta-analyze is personsupervisor (PS) fit. Stud-ies of leaderfollower value congruence (e.g., Colbert, 2004; Krishnan,2002), supervisorsubordinate personality similarity (Schaubroeck &Lam, 2002), and manageremployee goal congruence (e.g., Witt, 1998) areincluded in this category. In each case, the supervisors personal character-istics represented the environment. Studies in which supervisors reportedwork group or organizational characteristics (e.g., Becker, 1992; Becker,Billings, Eveleth, & Gilbert, 1996) were classified as either PG or PO fit,respectively.
Related to the PS fit domain is research on leader member exchange(LMX; e.g., Graen, 1976) and perceptual similarity (e.g., Wexley &Pulakos, 1983). Although each of these areas emphasizes the dyadic in-teractions between individuals, they are excluded from this review. In thecase of LMX, the emphasis is on the nature of the relationship that de-velops between leaders and followers not the match of their underlyingpsychological characteristics. Therefore, only LMX studies that also ex-amine PS fit are included in our analyses. Studies of perceptual similarity,that is, similarity between a subordinates description of the manager andthe managers self-description were also excluded, because they assessaccuracy of perceptions rather than fit.
Moderator Analyses
Even after these exclusions, there is still remarkable heterogeneity inthe way that fit has been conceptualized and measured. A thorough under-standing of fits influence on individual-level outcomes must account forthese differences. Next, we provide theory-driven rationale for concep-tual and methodological issues that we expect to moderate fitoutcome
288 PERSONNEL PSYCHOLOGY
relationships, with an emphasis on PJ and PO fit because of the numberof studies available on these topics.
Conceptualizations of Fit
Complementary versus supplementary. Muchinsky and Monahan(1987, pp. 268269) highlighted a troubling oversight in the PE fit lit-erature when they noted that although personenvironment congruencerefers to the degree of fit or match between the two sets of variables . . . whatexactly constitutes a fit or match is not totally clear. In response to thisambiguity, they proposed two distinct conceptualizations of PE fit: com-plementary and supplementary fit. For complementary fit, the basis fora good fit is the mutually offsetting pattern of relevant characteristics be-tween the person and the environment (Muchinsky & Monahan, 1987,pp. 272). Although they operationalized this type of fit strictly as individ-uals skills meeting environmental needs (demandsabilities fit), Kristof(1996) expanded the definition to include when individuals needs are metby environmental supplies (needssupplies fit). Thus, complementary fitoccurs when individuals characteristics fill a gap in the current environ-ment, or vice versa. The second major conceptualization of fit, supple-mentary fit, exists when the individual and the environment are similar.Whereas complementary conceptualizations of fit have dominated the PJfit literature, supplementary fit has been the focus of other types of fit.
Theories of need fulfillment (e.g., Locke, 1976; Rice, McFarlin, Hunt,& Near, 1985) explain the primary mechanism by which complementaryneedssupplies fit influences attitudes. These theories share the commonproposition that people will experience more positive job attitudes whentheir needs are satisfied. Recently, it has been argued that supplementaryfit may also function through the process of need fulfillment (Van Vianen,2000). Theories such as Festingers theory of social comparison (1954),Heiders balanced state theory (Heider, 1958), and Byrnes similarity-attraction paradigm (1971) all suggest that people have a fundamentalneed for consensual validation of their perspectives, which can be met byinteracting with similar others. Therefore, achieving supplementary fit isone way to have personal needs met, but obtaining needssupplies fit ismore direct. Therefore, we expect that supplementary fit will have a some-what weaker relationship with job attitudes than does needssupplies fit.Consistent with this logic, demandsabilities fit should have the smallestrelationship with job attitudes because it emphasizes meeting environmen-tal, rather than individual needs.
We expect a different pattern of relationships for nonattitudinal cri-teria, such as performance, strain, or turnover. If a person does nothave the requisite abilities to meet situational demands, overall andtask performance are likely to suffer. Even contextual performance may
AMY L. KRISTOF-BROWN ET AL. 289
decline, if employees do not have adequate attentional resources (Kanfer& Ackerman, 1989) to engage in organizational citizenship behaviorsbecause they are being stretched by task demands. Strain should be highunder such conditions, and turnover may result because of consistent un-derperformance. This implies a stronger effect for demandsabilities fiton the aforementioned criteria. However, needssupplies and supplemen-tary fit are also likely to affect these outcomes. The theory of reasonedaction suggests that attitudes toward a given object will generally resultin behaviors that are consistent with those attitudes (Ajzen & Fishbein,1977; Fishbein & Ajzen, 1975). Thus, if individuals needs are being metat work, resulting positive attitudes may mitigate strain, facilitate higherperformance, and reduce turnover. Nonetheless, because proficiency mat-ters to outcomes of strain, performance, and turnover, we expect that theimpact of demandsabilities fit will be greater than the other conceptual-izations of fit. Studies that include multiple conceptualizations of fit (i.e.,assess needssupplies and demandsabilities fit) should produce strongereffects than those using single conceptualizations because they tap intomultiple mechanisms by which fit has an impact.
Content dimensions of PE fit. Further complicating research on PE fitis the variety of content dimensions used to operationalize fit. The decisionon what dimensions or characteristics to use is somewhat determined by thebroader conceptualization of fit being explored. For example, a demandsabilities fit is almost exclusively based on the content dimensions of indi-viduals KSAs and job or organizational demands. However, with grow-ing recognition of the importance of personality on work performance(Barrick & Mount, 1991), a few studies have expanded the demands-abilities fit to include personality traits and values (e.g., Edwards, 1996;Lauver & Kristof-Brown, 2001). Similarly, a needssupplies fit, which hastraditionally emphasized individuals needs and preferences, can also beconsidered to include value preferences, or what types of values a personwants in an organization. For example, the wording of the original Orga-nizational Culture Profile (OCP; OReilly, Chatman, & Caldwell, 1991)requests that respondents indicate what values they prefer or want in an or-ganization, rather than what their personal values actually are. These scoresare then compared to what they receive from the organization. Worded insuch a way, studies using the original OCP assess value-fulfillment whichcorresponds more directly to needssupplies complementary fit.
Studies of supplementary fit have traditionally used the broadest ar-ray of content dimensions. With roots in theories of interpersonal at-traction (Byrne, 1971) and the attractionselectionattrition framework(Schneider, 1987), supplementary fit requires that P and E are at similarlevels on commensurate dimensions. These dimensions may include val-ues, goals, personality traits, or attitudes. For example, Cable and Judge(1996) shortened the OCP and revised it to be more clearly a measure of
290 PERSONNEL PSYCHOLOGY
supplementary fit, by rephrasing the questions to What values are mostindicative of you? [emphasis added] and What values are most indicativeof the organization? Personality and goals have also been used to assesssupplementary fit in the domains of PO, PG, and PS fit (e.g., Adkins et al.,1994; Kristof-Brown & Stevens, 2001; Ryan & Schmit, 1996; Witt & Nye,1992).
Reviews of fit often contain heated debates about the relative meritsof these dimensions and how they should be measured (e.g., Edwards,1994; Kristof, 1996; Meglino, Ravlin, & Adkins, 1991). This has beenparticularly true in the domain of PO fit. Chatman (1991) argued for val-ues as the basis of PO fit because values are enduring characteristics ofindividuals and organizations. Yet, Ryan and Kristof-Brown (2003) sug-gested that personality traits are more stable, proximal to behavior, andvisible in others behavior than are values. They conclude, therefore, thatpersonality-based PO fit should have at least as strong as, if not a stronger,influence on individuals attitudes and behaviors than would values-basedfit (p. 266). An important consideration, however, is that although simi-larity on values is believed to be universally desirable (Meglino & Ravlin,1998), sometimes personality dissimilarity may be preferable (Carson,1969). For example, Kristof-Brown et al. (in press) demonstrate that peo-ple high on extraversion feel more attraction to their teams if the othermembers are more introverted, and vice versa. If similarity on some traits,but not others, is desirable, then overall personality congruence shouldhave weaker relationships with outcomes than does value congruence. Al-ternatively, for goals, PO congruence should always be preferable. Sharinggoals makes it more likely that individuals will receive support and rein-forcement for goal attainment. However, goals are less stable than values.Therefore, compared to other supplementary fit dimensions, we expect thatgoals will result in effect sizes greater than personality-based measures,but less than values-based measures. KSA-based studies of PO fit are rare,although some authors have included requisite KSAs into measures thatcombine multiple content dimensions to assess fit. For example, Bretzand Judge (1994) assessed fit along the dimensions of values, personality,needs, and KSAs. By including each of these dimensions, such combinedmeasures are likely to capture a more holistic assessment of fit. Therefore,it is expected that studies using combined measures will report strongereffect sizes than studies assessing just value congruence.
Measurement of Fit
Direct versus indirect measures. A potentially meaningful distinctionbetween types of fit studies is whether they assess fit directly or indirectly
AMY L. KRISTOF-BROWN ET AL. 291
(Kristof, 1996). Most scholars agree that perceived fit1 is defined by adirect assessment of compatibility (French et al., 1974; Kristof, 1996).Kristof (1996) distinguished this from actual fit, an inclusive term usedto describe measures in which researchers indirectly assess fit throughexplicit comparisons of separately rated P and E variables. French etal. (1974) further differentiated such explicit comparisons into subjec-tive fit, defined as the match between the person and environment as theyare perceived and reported by the person, and objective fit as the matchbetween the person as he or she really is and the environment as it ex-ists independently of the persons perception of it (French et al., 1974,p. 316). Over the years, the terms perceived and subjective fit have oftenbeen used interchangeably (e.g., Cable & DeRue, 2002; Judge & Cable,1997; Kristof, 1996). However, because the cognitive processes underly-ing each may differ, we believe it is important to distinguish between thesetypes of fit. Thus, we use the terms as follows: (a) perceived fit, when anindividual makes a direct assessment of the compatibility between P andE; (b) subjective fit, when fit is assessed indirectly through the compari-son of P and E variables reported by the same person; and (c) objectivefit, when fit is calculated indirectly through the comparison of P and Evariables as reported by different sources.
When an individual has a high degree of contact with reality (i.e., as-sesses the environment accurately) and an accurate self-assessment, thesethree types of fit should have similar relationships with criteria (Frenchet al., 1974). In practice, however, they are often only weakly related(Cable & Judge, 1997; Kristof-Brown & Stevens, 2001). This may be dueto individuals propensity to interpret environmental cues in ways that al-low them to maintain a positive self-concept (Endler & Magnussen, 1976;French et al., 1974). Both self-perception theory (Bem, 1967) and cogni-tive dissonance theory (Festinger, 1957) suggest that individuals are drivento maintain internally consistent perceptions. It would produce cognitivedissonance for an individual to appraise a work environment as providinga poor fit but still report a high level of satisfaction with that environ-ment. Therefore, it is likely that perceived, subjective, and objective fitdiffer not only in how they are measured but also in what they representconceptually.
Perceived fit allows the greatest level of cognitive manipulation be-cause the assessment is all done in the head of the respondents, allowingthem to apply their own weighting scheme to various aspects of the envi-ronment. This permits individual differences in importance or salience of
1Note that we define perceived and subjective fit to be consistent with French et al.s(1974) original use of the terms, but these labels are reversed in Verquer et al. (2003).
292 PERSONNEL PSYCHOLOGY
various dimensions to be captured in their ratings. Alternatively, indirectmeasures require separate ratings across specific P and E dimensions, andeach dimension is weighted equally. Indirect measures of subjective fitstill have P and E variables assessed by the same source, which does notfully negate consistency biases. Because both perceived and subjective fitare assessed by a single source, we expect that they will have similar ef-fects on outcomes, particularly attitudes, with perceived fit having slightlystronger effects because it allows a single individual to report a holisticassessment of fit and is more prone to consistency effects.
Compared to perceived and subjective fit, objective fit is a less prox-imal determinant of attitudes and behavior (Cable & DeRue, 2002). Yet,unless people are completely removed from reality, the objective environ-ment should have some influence on their perceptions of E. The objectiveenvironment may be made quite clear in experimental studies that use de-scriptions of the environment (e.g., Dineen, Ash, & Noe, 2002; Judge &Bretz, 1992; Kristof-Brown, Jansen, & Colbert, 2002). In addition, strongorganizational cultures and the shared perceptions of coworkers constitutesocially derived environments that have been found to have a meaningfulimpact on how people react to their work situations (e.g., Kozlowski &Klein, 2000; Salancik & Pfeffer, 1978). However, because this objectivereality must be filtered through individuals perceptions, we expect thatobjective fit will have the weakest relationships with most criteria. How-ever, this decrease should be less pronounced for performance criteriathan for attitudes because maximum performance should be limited by anobjective demandability mismatch.
Study Design
Assessing the environment. In cases of objective fit, Van Vianen(2000) implied another potential moderator when she differentiated be-tween the fit of a person with an organization and the fit of a personwith the people in the organization. Although Schneiders (1987) ASAframework purports that the people make the place, group effects the-ory (e.g., Blau, 1960; Merton & Kitt, 1950) and theories of organizationalculture (e.g., Levinthal & March, 1993; Schein, 1992) suggest that higher-level units are more than the aggregate of attributes of their members.Thus, it may be overly simplistic to assume that the arithmetic mean ofthe people in a unit can capture the extent of culture or climate. Be-cause aggregate personal characteristics reflect current members only,whereas culture and climate reflect the totality of a units history, we ex-pect weaker relationships between fit and criteria in studies using aggregatemembers personal characteristics, rather than perceptions of the overallunit.
AMY L. KRISTOF-BROWN ET AL. 293
Sampling strategy. Recent results have demonstrated that meta-analytic estimates are not stable when single-unit (e.g., job, group,organization) samples and multi-unit samples are combined (Gully,Incalcaterra, Joshi, & Beaubien, 2002; Ostroff & Harrison, 1999; Roth,Bevier, Bobko, Switzer, & Tyler, 2001). Although PG and PS studies ne-cessitate multi-unit samples to obtain adequate sample size, PJ and POstudies are often conducted for one job or within one organization. There-fore, single- versus multi-unit samples are a potential moderator for thesetypes of fit. Because multi-unit samples permit a broader range of environ-ments to be represented, there is reason to expect that fit relationships willbe stronger in multi-unit samples. However, Chatman (1989) noted that fitwill be most influential in strong situations. Therefore, when a single-unitenvironment is strong, relationships should be greater than in multi-unitsamples, but the reverse is true for single-unit samples with weak envi-ronments. Because of these competing rationales, we are unable to predictthe direction that this moderator will take.
Common method bias. Finally, it is important to address concernsthat have been raised regarding the heavy reliance on single-source report-ing of fitoutcome relationships (Edwards, 1993; Van Vianen, in press).Despite general awareness that common rater bias can inflate reportedeffect sizes, particularly for attitudeattitude relationships (Podsakoff,MacKenzie, Lee, & Podsakoff, 2003), PE fit studies often include predic-tor and outcome variables reported by the same person. This is particularlyproblematic for studies that examine the relationship between perceivedor subjective fit and attitudes. Because common rater bias can produceartifactual covariance due to consistency biases and illusory correlations(Podsakoff et al., 2003), we anticipate that the strongest effect sizes willbe reported for studies in which fit and the criteria are rated by the sameindividual, followed by those using a rater for two of the three potentialvariables (P, E, and criteria), and finally by those using separate ratings ofeverything.
One approach to reducing the impact of common rater bias, whichis primarily a concern in fitattitude studies, is to temporally separatethe measurement of the predictor and criterion. Because a time delayshould reduce consistency and illusory correlations between constructs(Podsakoff et al., 2003), we expect larger effect sizes between fit andattitudes measured concurrently versus at different points in time.
Comparing Types of Fit
It is worth mentioning why we do not treat type of fit (PJ, PO, PG,PS) as a moderator. Because each type of fit represents compatibility withdifferent aspects of the environment, and has been shown to have unique
294 PERSONNEL PSYCHOLOGY
impacts on outcomes (Cable & DeRue, 2002; Kristof-Brown, 2000), weanalyze each separately. However, we do anticipate differences in the rela-tive magnitude of each type with various criteria. Research has shown thatjob-related constructs are most strongly associated with attitudes aboutthe job, whereas organization-related constructs are more closely relatedto organizational attitudes (Shore & Martin, 1989). In the domain of fit,job satisfaction should be most strongly associated with PJ fit, organiza-tional commitment with PO fit, satisfaction with coworkers with PG fit,and satisfaction with supervisor with PS fit. Other criteria, such as intentto quit, tenure, and turnover are likely to be influenced by multiple typesof fit because they represent attitudes or behaviors relevant to the totalwork experience (Barber, 1998; Mitchell, Holtom, Lee, Sablynski, & Erez,2001). The same is expected for organizational attraction because appli-cants generally get recruited based on elements of the job and organizationsimultaneously. However, intent to hire is expected to be more related toPJ than PO fit because the emphasis of most selection techniques is toassess whether the applicant has the requisite job-related skills (Bowen,Ledford, & Nathan, 1991).
Whereas task performance is highly dependent on skill-based job pro-ficiency, contextual performance has volition and predisposition as itsmajor source of variation and is not likely to be role-prescribed (Borman& Motowidlo, 1997). It is more directed toward helping and cooperatingwith others, supporting the organizations mission, and putting in extraeffort when necessary. Therefore, PJ fit should have the greatest impacton task performance, whereas contextual performance is more likely tobe influenced by PO or PG fit. Because overall performance contains bothtask and contextual elements, all types of fit should be relevant.
Although we treat the various types of fit separately, it is informativeto examine the relationships between them. Kristof (1996) called for anincrease in research that simultaneously examines multiple types of fit,and recent studies show a trend in this direction (e.g., Cable & Derue,2002; Saks & Ashforth, 2002). Because of their emphasis on the people inthe environment, and on needsupplies and supplementary fit conceptual-izations, we expect that PO, PG, and PS fit will be more strongly related toeach other than to PJ fit. This should particularly be true for cases of PJ fitthat reflect a demandsabilities perspective, rather than a needssuppliesperspective. Moreover, because relationships with coworkers and supervi-sors are involved in everyday work experiences, we anticipated that PS andPG fit would be more strongly related than either one with PO fit, whichmay be tempered by the closer relationships with colleagues. Finally, weexpect that all relationships between types of fit will be ameliorated whena direct versus indirect measurement strategy is used because of the in-creased likelihood of consistency effects in measures of perceived fit.
AMY L. KRISTOF-BROWN ET AL. 295
Polynomial Regression
One of the most dramatic changes to studies of PE fit in recent yearswas the introduction of polynomial regression as an alternative to ProfileSimilarity Indices (PSIs) for assessing fit indirectly. PSIs include differ-ence scores, a category of indices including (D, |D|, D1, D2, and Maha-lanobis D), base fit measures on the sum of differences between P and Eprofile elements (Edwards, 1993). They also include profile correlations,which assess similarity in the rank ordering of elements within P andE profiles (Edwards, 1993). Despite the prevalence of these approachesand their ability to capture the holistic nature of fit, they engenderedcriticism around issues of conceptual ambiguity, discarded information,insensitivity to the source of differences, and overly restrictive constraints(Cronbach, 1958; Edwards, 1993; Nunally, 1962).
In response to these concerns, Edwards (1993, 1994, 1996) suggestedusing polynomial regression to assess PE fit. Fundamentally, this approachavoids collapsing person and environment measures into a single score thatcaptures fit. Instead, both P and E, and associated higher-order terms(P2, P E, and E2) are included as predictors. Because the P and E and thedependent variables are all unique, fit relationships are depicted in three-dimensional surface plots. Although effect sizes can be calculated fromthe multiple correlations of these studies, these include main and quadraticeffects not present in direct or PSI measures of fit, rendering a meta-analysis of their results ambiguous. To determine whether a fit relationshipis supported, characteristics of the surface plot must be examined. Resultsoften depict more complicated relationships than are determinable withPSIs (e.g., Edwards, 1993, 1996). Because of their increased usage, as wellas the potentially useful insights that they provide regarding the nature offit, we include a qualitative summary of their findings in our review.
Method
Data Collection
To identify studies for inclusion in the meta-analyses, we searchedseveral social science databases and national conference programs usingthe following key words: fit, congruence, similarity, interaction, personenvironment, personorganization, personjob, personculture, persongroup, personteam, personsupervisor, and supervisorsubordinate.Specifically, we searched the American Psychological AssociationsPsycINFO (18872003), ABI Inform (19712003), and DissertationAbstracts International (18612003) databases for relevant studies. Toobtain unpublished or in-press research we searched the Society for
296 PERSONNEL PSYCHOLOGY
Industrial and Organizational Psychology (19952004) and Academy ofManagement (19942004) annual conference programs using the searchterms. We also conducted a manual search of our files and reviewedthe reference lists of other summaries (Edwards, 1991; Kristof, 1996;Verquer et al., 2003) to identify relevant articles missed in the comput-erized search. Finally, we contacted 24 prominent fit researchers andasked them to share their emerging or unpublished research relatingto fit.
Using these search procedures, we identified over 750 total studies,which we then screened to determine their relevance. We excluded nonem-pirical studies (e.g., review articles, conceptual articles) and articles on top-ics previously described as outside the purview of this review. Additionalexclusions were studies conducted in nonwork contexts (e.g., Burnett,Vaughan, & Moody, 1997; Luke, Roberts, & Rappaport, 1993; OReilly& Chatman, 1986) and those examining fit between people and nationalcultures (e.g., Harrison, Shaffer, & Bhaskar, 2002; Van Vianen, De Pater,Kristof-Brown, & Johnson, 2004).
All work-related outcomes were considered for inclusion, but in gen-eral, only those having five or more total correlations or a total samplesize of over 1,000 are reported here. A few exceptions are made formoderator analyses or criteria that are conceptually relevant to PG andPS fit, which have a small number of total studies. Preentry criteria in-cluded organizational attraction and job acceptance reported by appli-cants, and intent to hire and job offers from the organization. Postentryoutcome variables included job satisfaction, organizational satisfaction,coworker satisfaction, cohesion, supervisor satisfaction, LMX, organi-zational commitment, organizational identification, intentions to quit,trust in manager, politics, task performance, contextual performance,overall performance, withdrawal behaviors, strain, turnover, and con-current tenure. Not all dependent variables were available for all typesof fit.
For studies that assessed perceived fit, the correlations of these directmeasures with the criteria were used to represent the effect size. For studiesassessing subjective or objective fit using a PSI, correlations of the calcu-lated fit term and the criteria were generally reported. In a small number ofstudies, an F-statistic or t-statistic was converted to a correlation (Cohen &Cohen, 1983). After evaluating the studies based on the inclusion criteria,172 codable studies, producing 836 unique effect sizes, were available forour meta-analyses. Sixty-two studies with 225 effect sizes were included inthe PJ fit meta-analyses; 110 studies with 450 effect sizes were included inthe PO fit meta-analyses; 20 studies with 104 effect sizes were includedin the PG fit meta-analyses; and 17 studies with 57 effect sizes were
AMY L. KRISTOF-BROWN ET AL. 297
included in the PS fit meta-analyses.2 Twenty-seven studies provided 34correlations between various types of fit. Twenty-five additional studiesreported polynomial regression results, which are qualitatively reviewed.All studies included in the meta-analyses are indicated in the referencesection, as are the polynomial regression studies.
Data Coding
We coded the characteristics of each study on multiple dimensions.Specifically, we coded the following: type of fit (PJ, PG, PO, and PS),context (preentry vs. postentry), sample size, effect size, and reliabilityof measures. We also coded for potential moderators. First, we coded theconceptualization of fit: supplementary, complementary needssupplies,complementary demandsabilities or combined (any combination of theprevious three types). Second, we coded for the content dimension onwhich fit was assessed: values, personality, needs, goals, KSAs, combinedmeasures (i.e., including multiple content dimensions), or other (e.g., strat-egy). Third, we coded whether direct measures of perceived fit or indirectmeasures of objective or subjective fit were used. Within the subgroup ofstudies that used objective measures, we coded whether the environmentwas assessed as an independent entity or as the aggregation of memberscharacteristics. We coded all PJ and PO studies for whether the samplingstrategy was single unit or multi unit. Finally, to assess the impact of com-mon method bias we coded studies with performance criteria as using allcommon raters (ACR), partially common raters (PCR), or no commonrater (NCR), and studies with attitudinal criteria as collecting predictorsand criteria concurrently or separated in time.
Each article was coded independently by at least two of the authors.Coding discrepancies were resolved through discussion and by referringto the information provided in the original article and in the coding defini-tions. Given the ambiguities inherent in many of these studies, we foundthe use of multiple raters per article to be beneficial. After a prelimi-nary discussion of sample articles, average agreement calculated acrossall codes for a random subsample of 15% of the studies was 97%.
Meta-Analytic Procedures
We conducted the meta-analyses using the formulas from Hunterand Schmidt (2004), with the aid of the meta-analytic software program
2If fit was assessed on multiple facets of a content dimension (e.g., fit on four separatevalues), the correlations were averaged to create one effect size.
298 PERSONNEL PSYCHOLOGY
developed by Schmidt and Le (2004). The means and variances of the meta-analytic estimates were corrected for artifactual variance due to samplingerror and attenuation due to measurement error in the predictor and thecriterion. Because the purpose of this study was to provide a meaningfulestimate of the relationship between fit and various criteria, correcting forattenuation in both the predictor and the criterion was necessary. Relia-bilities for the predictor and criterion were not available in each study;therefore, we used artifact distributions to correct for measurement errorin both (Hunter & Schmidt, 2004). This was done by utilizing the in-formation contained in all studies that did report reliability estimates foreither the predictor or criteria variables in question. A single distributionof reliabilities was then created for each variable based on these estimates,and this distribution was used to correct the appropriate variable for un-reliability. Note that in order to best estimate the artifact distribution fora given variable, the same distribution was used for that variable acrossall meta-analyses. This was done to better reflect the true distribution ofreliabilities in the population, rather than just in the subset of studies fora given meta-analysis.
Because reliabilities are not available for objective criteria (tenure,turnover, withdrawal, job offer, and job acceptance), no correction formeasurement error was made for these variables. In addition, no correctionwas made for dichotomization in turnover because almost none of thestudies provided turnover rates, and this type of correction is not withoutcontroversy (Hunter & Schmidt, 2004; Williams, 1990). For studies thatused a difference score to calculate fit, the reliability of the differencescore was calculated when enough information was provided in the study(i.e., the reliability of the person scale and the environment scale, and thecorrelation between the two scales; Hunter & Cohen, 1974). Finally, thepopulation of interest and the population examined in each study were thesame and therefore no correction for range restriction was needed. Thiswas true even when fit was examined in a preentry setting, as the criteriaof interest in those settings were either job acceptance or job offer.
It is important to note that several of the studies measured fit in multipleways for the same sample and criteria. To ensure that the assumption ofindependent samples was not violated (Hunter & Schmidt, 2004), no morethan one effect size from a study was included in each meta-analysis. Ifthere were multiple effect sizes from a single study (e.g., correlationsfrom a perceived PJ measure and a subjective PJ fit measure with jobsatisfaction), then the average of these two unique effect sizes was usedin the analysis of the overall relationship (i.e., PJ fit job satisfaction).However, when specific effect sizes pertaining to the various moderatorswere available, these, rather than the aggregate correlations, were used inthe moderator analyses.
AMY L. KRISTOF-BROWN ET AL. 299
Moderator analyses were conducted to test the theoretically derivedpropositions articulated earlier, when there was a large enough k and Nto merit discussion. Given the small number of studies on PG and PSfit with the same criteria, no moderator analyses were possible for thesetypes of fit. A moderator would be indicated when two conditions aremet. First, there would be a large difference in the effect sizes betweenthe compared groups of studies. Second, the average variance in effectsizes within each of these groups should decrease (Hunter & Schmidt,2004). The 80% credibility intervals reported in each meta-analysis tableindicate the generalizability of the relationship across situations (Hunter& Schmidt, 2004). If the credibility interval does not include zero, therelationship is said to be generalizable. If the interval does not includezero, yet there is still considerable variance unaccounted for, the direc-tion of the relationship may still generalize (i.e., still be positive or nega-tive across situations), but its strength may vary depending on additionalmoderators.
Results
Tables 14 present the meta-analyses for the four types of fit, the crite-ria, and the moderator analyses. The first column indicates the fitcriterionrelationship examined, and any relevant subsamples due to moderators.The second and third columns provide the number of effect sizes (k) andtotal sample size (N), respectively, included in the meta-analytic estimate.In the fourth column, the sample size weighted mean observed correlationis presented. The lower and upper bounds of the 95% confidence inter-val of the observed correlations are shown in the fifth and sixth columns,respectively. The seventh column contains the estimated true score corre-lation (), followed by the standard deviation of this estimate in the eighthcolumn (SD ). The lower and upper bound of the 80% credibility inter-val are presented in the ninth and tenth columns, respectively. Finally, thepercentage of variance accounted for by the artifacts is contained in thelast column. Table 5 reports the meta-analytic results of the interrelation-ships between types of fit, as well as moderator analyses for direct versusindirect measures.
PersonJob Fit
As shown in Table 1, the first set of meta-analyses includes the cor-relations between PJ fit and outcomes. PJ fit has strong correlations withthe three primary attitudes studied in the fit literature: .56 with job sat-isfaction, .47 with organizational commitment, and .46 with intent toquit. PJ fit has a moderate relationship with the attitudes of coworker
300 PERSONNEL PSYCHOLOGY
TAB
LE1
Met
a-An
alys
isRe
sults
ofP
erso
n
Job
Fit
95%
CI80
%CV
kN
Avg
rLo
wer
Upp
er
SD
Low
erU
pper
%var
expl
aine
dJo
bsa
tisfa
ctio
n47
12,9
60.44
.20
.68
.56
.13
8.38
.73
22.2
1Co
mpl
emen
tary
:DA
123,
735
.32
.14
.50
.41
.08
5.30
.52
40.8
7Co
mpl
emen
tary
:NS
328,
726
.48
.19
.77
.61
.17
2.39
.83
15.5
6Co
mbi
ned
[DA
&N
S]
61,
036
.49
.20
.78
.62
.15
9.41
.82
22.6
5D
irect
:Per
ceiv
ed23
7,20
5.45
.23
.67
.58
.11
2.43
.72
28.3
4In
dire
ct24
4,90
8.44
.13
.75
.56
.18
6.32
.79
16.4
2Su
bjecti
ve
184,
377
.46
.19
.73
.59
.16
3.38
.80
18.1
3O
bjecti
ve
653
1.22
.05
.49
.28
.11
8.13
.43
54.6
7Si
ngle
job10
2,54
6.47
.20
.74
.59
.15
4.39
.79
19.4
7M
ultip
lejob
s37
10,4
14.43
.19
.67
.55
.13
2.38
.72
23.4
5Sa
me
time
369,
704
.44
.19
.69
.56
.14
1.38
.74
21.6
7D
iffer
entt
imes
102,
742
.45
.20
.70
.57
.14
2.39
.76
21.2
9O
rgan
izat
iona
lcom
mitm
ent
184,
073
.39
.10
.68
.47
.17
0.26
.69
16.7
6Co
mpl
emen
tary
:DA
487
3.25
.01
.49
.31
.12
0.15
.46
30.6
9Co
mpl
emen
tary
:NS
91,
833
.31
.06
.56
.37
.13
9.20
.55
25.7
3Co
mbi
ned
[DA
&N
S]
586
1.42
.05
.79
.51
.21
5.24
.79
13.3
2D
irect
:Per
ceiv
ed11
2,15
4.36
.03
.69
.44
.18
7.21
.68
16.0
5In
dire
ct8
2,01
1.40
.13
.67
.50
.15
6.30
.69
18.0
0Su
bjecti
ve
321
6.17
.01
.33
.21
0.21
.21
100.
00O
bjecti
ve
51,
795
.43
.19
.67
.53
.13
2.36
.70
19.1
5Si
ngle
job7
1,98
7.43
.19
.67
.53
.13
4.36
.70
21.2
8M
ultip
lejob
s11
2,08
6.35
.02
.68
.43
.18
6.19
.66
16.7
1Sa
me
time
102,
438
.44
.19
.69
.54
.14
0.36
.72
21.4
6D
iffer
entt
imes
81,
635
.31
.02
.60
.38
.16
7.17
.59
19.2
5
AMY L. KRISTOF-BROWN ET AL. 301
Inte
ntto
quit
163,
849
.37
.59
.15
.46
.11
8.
61.
3029
.09
Com
plem
enta
ry:D
A6
982
.18
.42
.06
.23
.10
7.
36.
0943
.69
Com
plem
enta
ry:N
S8
2,81
6.
40.
64.
16.
50.12
6.
66.
3420
.98
Com
bine
d[D
A&
NS
]5
783
.46
.75
.17
.57
.16
7.
78.
3521
.45
Dire
ct:P
erce
ived
112,
065
.40
.65
.15
.49
.13
2.
66.
3228
.25
Indi
rect
61,
866
.33
.57
.09
.40
.13
4.
57.
2320
.74
Subje
ctive
326
7.
15.
50.20
.18
.18
2.
41.05
33.4
6O
bjecti
ve
31,
599
.36
.52
.20
.44
.07
6.
54.
3434
.92
Sing
lejob
41,
770
.35
.51
.19
.43
.07
7.
53.
3337
.27
Mul
tiple
jobs
122,
059
.39
.64
.14
.48
.14
0.
66.
3027
.63
Sam
etim
e7
2,38
7.
38.
50.
26.
47.04
4.
52.
4169
.26
Diff
eren
ttim
es8
1,13
1.
38.
73.
03.
47.19
5.
72.
2218
.99
Cow
ork
ersa
tisfa
ctio
n6
2,28
0.25
.13
.37
.32
.03
9.27
.37
75.9
8Su
perv
isors
atisf
actio
n4
1,98
7.28
.12
.44
.33
.07
8.23
.43
32.4
6O
rgan
izat
iona
lide
ntifi
catio
n8
1,61
9.29
.11
.47
.36
.06
3.28
.44
65.8
3O
vera
llpe
rform
ance
191,
938
.16
.19
.51
.20
.18
6.
04.44
29.9
5Co
mpl
emen
tary
:DA
854
7.10
.33
.53
.12
.22
8.
17.41
30.6
3Co
mpl
emen
tary
:NS
81,
558
.16
.00
.32
.20
.04
1.14
.25
81.9
2D
irect
:Per
ceiv
ed3
552
.18
.25
.61
.22
.24
9.
10.54
11.4
7In
dire
ct16
1,38
6.16
.17
.49
.19
.15
3.00
.39
42.7
8Su
bjecti
ve
1029
5.13
.48
.74
.16
.30
4.
23.55
36.1
7O
bjecti
ve
61,
091
.17
.01
.35
.20
.06
4.12
.29
66.5
9Si
ngle
Job
880
1.08
.17
.33
.10
.10
0.
03.23
60.3
0M
ultip
leJo
bs15
1,44
1.22
.11
.55
.27
.17
2.05
.49
32.2
5St
rain
103,
505
.23
.39
.07
.28
.07
8.
38.
1840
.82
Com
plem
enta
ry:D
A4
2,02
5.
25.
43.
07.
30.09
1.
42.
1926
.64
Com
plem
enta
ry:N
S6
1,64
2.
25.
43.
07.
31.08
1.
41.
2043
.98
Ten
ure
51,
484
.16
.08
.40
.18
.11
5.03
.32
23.4
2Tu
rnover
81,
496
.07
.25
.11
.08
.05
9.
16.
0164
.95
Org
aniz
atio
nala
ttrac
tion
48,
131
.40
.09
.71
.48
.18
3.25
.72
4.46
Inte
ntto
hire
41,
132
.59
.30
.88
.67
.15
6.47
.87
12.6
7
DA
=de
man
dsa
bilit
iesfi
t.N
S=
nee
dss
uppl
iesfi
t.
302 PERSONNEL PSYCHOLOGY
TABL
E2
Met
a-An
alys
isRe
sults
ofP
erso
nO
rgan
izat
ion
Fit
95%
CI80
%CV
kN
Avg
rLo
wer
Upp
er
SD
Low
erU
pper
%var
expl
aine
dJo
bsa
tisfa
ctio
n65
42,9
22.35
.08
.62
.44
.16
7.23
.65
9.42
w/o
Vanco
uver
and
Schm
itt(19
91)
6429
,534
.40
.13
.67
.50
.16
8.29
.72
11.8
4Su
pple
men
tary
4133
,767
.34
.07
.61
.43
.16
7.22
.64
8.27
w/o
Vanco
uver
and
Schm
itt(19
91)
4020
,379
.41
.14
.68
.52
.16
9.30
.73
11.1
8Co
mpl
emen
tary
:NS
309,
284
.37
.06
.68
.46
.18
6.22
.70
12.3
8Va
lues
4523
,211
.41
.16
.66
.51
.15
7.31
.71
12.6
2Pe
rson
ality
554
8.07
.01
.13
.08
0.08
.08
100.
00G
oals
314
,367
.24
.20
.28
.31
0.31
.31
100.
00M
ultid
imen
siona
l10
3,52
4.44
.19
.69
.55
.15
2.35
.74
16.3
5D
irect
:Per
ceiv
ed30
19,5
34.45
.23
.67
.56
.12
4.41
.72
17.7
0In
dire
ct:
3924
,339
.28
.04
.52
.35
.14
2.17
.53
12.1
2w
/oVa
nco
uver
and
Schm
itt(19
91)
3810
,951
.32
.01
.65
.40
.19
7.15
.65
11.6
5Su
bjecti
ve
248,
187
.37
.04
.70
.46
.19
8.21
.71
10.5
0O
bjecti
ve
1916
,706
.23
.11
.35
.29
.05
6.22
.36
39.3
3w
/oVa
nco
uver
and
Schm
itt(19
91)
183,
318
.19
.05
.43
.24
.12
1.08
.39
34.6
9R
atin
gsofo
rgan
izat
ion
asE
131,
780
.17
.10
.44
.22
.13
2.05
.39
38.3
2A
ggre
gate
mem
berp
rofil
esas
E6
14,7
50.24
.18
.30
.30
0.30
.30
100.
00w
/oVa
nco
uver
and
Schm
itt(19
91)
51,
362
.24
.08
.40
.30
.07
5.20
.39
49.4
3Si
ngle
org
aniz
atio
n28
9,44
6.34
.01
.67
.42
.19
8.17
.67
10.4
5M
ultip
leorg
aniz
atio
n37
34,1
06.36
.11
.61
.44
.15
5.25
.64
9.05
w/o
Vanco
uver
and
Schm
itt(19
91)
3620
,718
.43
.19
.67
.54
.13
5.36
.71
15.6
6Sa
me
time
4939
,012
.36
.09
.63
.45
.16
5.24
.66
8.74
w/o
Vanco
uver
and
Schm
itt(19
91)
4825
,624
.42
.15
.69
.52
.15
9.32
.73
12.2
3D
iffer
entt
imes
113,
084
.31
.02
.60
.39
.17
2.17
.61
15.3
1
AMY L. KRISTOF-BROWN ET AL. 303
Org
aniz
atio
nalc
om
mitm
ent
4436
,093
.42
.01
.85
.51
.26
0.18
.85
3.55
w/o
Vanco
uver
and
Schm
itt(19
91)
4322
,705
.54
.13
.95
.65
.23
9.34
.95
6.30
Supp
lem
enta
ry27
30,3
57.44
.01
.89
.53
.27
0.19
.88
2.99
w/o
Vanco
uver
and
Schm
itt(19
91)
2616
,969
.61
.28
.94
.74
.20
2.48
.99
8.03
Com
plem
enta
ry:N
S22
6,40
2.32
.03
.61
.39
.16
1.19
.60
15.4
9Va
lues
2818
,589
.56
.19
.93
.68
.22
7.39
.98
6.04
Mul
tidim
ensio
nal
102,
581
.49
.08
.90
.59
.24
8.27
.91
7.47
Dire
ct:P
erce
ived
2015
,275
.64
.35
.93
.77
.17
2.55
.99
10.8
9In
dire
ct:
2620
,965
.27
.07
.47
.32
.10
9.18
.46
14.7
2w
/oVa
nco
uver
and
Schm
itt(19
91)
257,
577
.33
.06
.60
.40
.15
8.19
.60
15.7
7Su
bjecti
ve
165,
886
.37
.12
.62
.44
.14
3.26
.62
16.1
5O
bjecti
ve
1215
,316
.23
.15
.31
.27
.03
9.22
.32
47.2
0w
/oVa
nco
uver
and
Schm
itt(19
91)
111,
928
.19
.05
.43
.23
.11
5.08
.38
37.4
3R
atin
gsofo
rgan
izat
ion
asE
880
6.13
.05
.31
.16
0.16
.16
100.
00A
ggre
gate
mem
berp
rofil
esas
E5
14,5
73.23
.13
.33
.28
.05
5.21
.35
21.2
2w
/oVa
nco
uver
and
Schm
itt(19
91)
41,
185
.20
.15
.55
.24
.20
4.
02.50
10.3
5Si
ngle
org
aniz
atio
n20
5,88
6.52
.11
.93
.63
.24
9.31
.95
6.86
Mul
tiple
org
aniz
atio
n24
30,2
07.40
.03
.83
.49
.25
5.16
.82
2.95
w/o
Vanco
uver
and
Schm
itt(19
91)
2316
,819
.54
.15
.93
.66
.23
6.35
.96
5.28
Sam
etim
e30
32,4
26.42
.01
.85
.51
.25
9.18
.85
3.18
w/o
Vanco
uver
and
Schm
itt(19
91)
2919
,038
.56
.19
.93
.68
.22
3.39
.97
6.26
Diff
eren
ttim
es14
3,71
0.36
.09
.81
.43
.26
8.09
.77
6.62
Inte
ntto
quit
4334
,276
.29
.56
.02
.35
.16
7.
57.
147.
24w
/oVa
nco
uver
and
Schm
itt(19
91)
4220
,888
.38
.60
.16
.47
.12
8.
64.
3117
.93
Supp
lem
enta
ry29
28,8
46.
29.
58.00
.36
.18
2.
59.
125.
35w
/oVa
nco
uver
and
Schm
itt(19
91)
2815
,458
.42
.64
.20
.51
.12
2.
66.
3517
.46
Com
plem
enta
ry:N
S23
6,80
2.
28.
52.
04.
34.12
8.
50.
1822
.55
Valu
es32
18,2
22.
38.
60.
16.
46.12
6.
62.
3015
.84
Mul
tidim
ensio
nal
102,
606
.39
.70
.08
.48
.18
5.
72.
2413
.19
Dire
ct:P
erce
ived
2415
,400
.43
.61
.25
.52
.09
4.
64.
4024
.58
304 PERSONNEL PSYCHOLOGYTA
BLE
2(co
ntinu
ed)
95%
CI80
%CV
kN
Avg
rLo
wer
Upp
er
SD
Low
erU
pper
%var
expl
aine
dIn
dire
ct:
2319
,497
.18
.34
.02
.22
.08
4.
33.
1120
.77
w/o
Vanco
uver
and
Schm
itt(19
91)
226,
109
.25
.47
.03
.31
.12
0.
46.
1527
.15
Subje
ctive
155,
036
.26
.48
.04
.32
.11
2.
46.
1825
.62
Obje
ctive
1114
,835
.15
.25
.05
.19
.05
0.
25.
1232
.66
w/o
Vanco
uver
and
Schm
itt(19
91)
101,
447
.18
.49
.13
.23
.16
7.
44.
0126
.64
Sing
leorg
aniz
atio
n20
4,56
4.
38.
69.
07.
46.18
5.
70.
2314
.48
Mul
tiple
org
aniz
atio
n24
29,8
60.
28.
55.
01.
34.15
9.
54.
135.
85w
/oVa
nco
uver
and
Schm
itt(19
91)
2316
,472
.38
.58
.18
.46
.11
4.
61.
3116
.67
Sam
eTi
me
3632
,729
.29
.56
.02
.35
.16
7.
56.
136.
66w
/oVa
nco
uver
and
Schm
itt(19
91)
3519
,341
.38
.60
.16
.46
.12
7.
62.
3015
.92
Diff
eren
ttim
es7
1,54
7.
39.
64.
14.
47.13
3.
64.
3025
.15
Org
aniz
atio
nals
atisf
actio
n4
997
.52
.23
.81
.65
.16
8.44
.87
18.4
6Co
work
ersa
tisfa
ctio
n10
2,50
9.31
.09
.53
.39
.11
1.25
.53
35.8
3D
irect
:Per
ceiv
ed5
2,06
3.34
.18
.50
.43
.06
9.35
.52
51.5
0In
dire
ct7
696
.15
.03
.33
.19
0.19
.19
100.
00Su
perv
isors
atisf
actio
n9
2,44
6.28
.04
.52
.33
.12
1.18
.49
24.6
7D
irect
:Per
ceiv
ed5
1,98
8.32
.12
.52
.38
.10
0.25
.51
25.5
6In
dire
ct4
458
.12
.10
.14
.14
0.14
.14
100.
00Tr
ust
inm
anag
er5
1,01
0.36
.26
.46
.43
0.43
.43
100.
00O
vera
llpe
rform
ance
225,
827
.05
.24
.34
.07
.16
2.
14.27
17.5
2D
irect
:Per
ceiv
ed7
2,38
6.10
.10
.30
.12
.09
5.00
.24
32.1
3In
dire
ct:
153,
441
.02
.31
.35
.03
.18
6.
21.26
15.7
8A
llco
mm
on
rate
r(AC
R)4
1,27
4.18
.02
.34
.22
.07
1.13
.31
47.9
1Pa
rtia
llyco
mm
on
rate
r(PC
R)9
3,70
0.
01.
25.23
.02
.14
5.
20.17
14.5
0N
oco
mm
on
rate
r(NC
R)9
853
.15
.16
.46
.18
.02
5.
02.38
38.4
0Ta
skpe
rform
ance
172,
860
.11
.11
.33
.13
.10
0.01
.26
45.8
7D
irect
:Per
ceiv
ed5
1,10
8.18
.02
.34
.22
.06
1.14
.30
63.2
3In
dire
ct:
121,
660
.04
.14
.22
.05
.02
2.02
.07
95.4
2
AMY L. KRISTOF-BROWN ET AL. 305
All
com
mon
rate
r(AC
R)3
389
.15
.09
.39
.18
.10
6.04
.31
48.9
6Pa
rtia
llyco
mm
on
rate
r(PC
R)8
1,57
7.12
.06
.30
.14
.06
9.06
.23
60.3
0N
oco
mm
on
rate
r(NC
R)9
1,07
8.03
.15
.21
.04
.02
7.00
.07
94.2
6Co
ntex
tual
perfo
rman
ce13
2,12
2.21
.04
.46
.27
.13
5.09
.44
33.4
2D
irect
:Per
ceiv
ed9
1,21
0.26
.16
.36
.32
0.32
.32
100.
00In
dire
ct5
994
.16
.19
.51
.20
.20
7.
06.46
15.2
0A
llco
mm
on
rate
r(AC
R)5
746
.36
.20
.52
.45
0.45
.45
100.
00Pa
rtia
llyco
mm
on
rate
r(PC
R)6
830
.21
.15
.27
.27
0.27
.27
100.
00N
oco
mm
on
rate
r(NC
R)3
629
.04
.02
.06
.06
0.06
.06
100.
00Te
nure
286,
036
.02
.18
.22
.03
.08
2.
08.13
45.7
5D
irect
:Per
ceiv
ed10
2,28
3.02
.14
.18
.03
.04
6.
03.08
71.2
8In
dire
ct22
4,48
4.02
.22
.26
.02
.10
1.
11.15
36.8
4Su
bjecti
ve
112,
882
.01
.23
.25
.01
.10
8.
13.15
028
.42
Obje
ctive
121,
744
.05
.17
.27
.06
.08
0.
05.16
55.9
1Tu
rnover
102,
157
.13
.38
.12
.14
.12
3.
30.01
26.8
2D
irect
:Per
ceiv
ed5
1,15
3.
15.
35.05
.16
.07
6.
26.
0647
.15
Indi
rect
51,
004
.11
.42
.20
.12
.15
8.
33.08
19.3
0W
ithdr
awal
(Abs
entee
ism)
71,
370
.05
.23
.13
.05
.06
0.
13.02
63.5
7St
rain
83,
605
.22
.47
.03
.27
.14
5.
46.
0813
.27
Dire
ct:P
erce
ived
62,
184
.28
.55
.01
.34
.15
4.
54.
1414
.10
Indi
rect
21,
421
.14
.20
.08
.17
0.
17.
1710
0.00
Org
aniz
atio
nala
ttrac
tion
119,
001
.38
.18
.58
.46
.10
4.33
.59
17.3
2D
irect
:Per
ceiv
ed10
8,79
7.52
.42
.62
.62
.04
1.57
.67
61.9
5In
dire
ct4
7,52
5.18
.16
.20
.22
0.22
.22
100.
00Jo
bac
cept
ance
41,
829
.22
.04
.40
.24
.08
9.13
.36
24.9
1In
tent
tohi
re9
2,51
8.53
.31
.75
.61
.11
6.47
.76
23.8
9D
irect
:Per
ceiv
ed8
2,40
2.61
.34
.88
.70
.14
6.52
.89
16.0
1In
dire
ct3
562
.16
.13
.45
.18
.15
3.
02.38
22.5
8Jo
boffe
r8
1,55
6.29
.14
.72
.32
.23
5.02
.63
9.16
Dire
ct:P
erce
ived
61,
178
.45
.16
.74
.50
.14
7.31
.69
18.2
9In
dire
ct6
918
.03
.19
.25
.03
.07
5.
06.13
58.3
3
306 PERSONNEL PSYCHOLOGY
TABLE 3Meta-Analysis Results of PersonGroup Fit
95% CI 80% CV% var
k N Avg r Lower Upper SD Lower Upper explainedJob satisfaction 9 1822 .24 .12 .36 .31 0 .31 .31 100.00Organizational 12 2306 .15 .05 .35 .19 .084 .08 .30 52.28
commitmentIntent to quit 6 943 .17 .33 .01 .22 .028 .25 .18 92.31Coworker 5 1210 .32 .16 .48 .42 .057 .34 .49 70.16
satisfactionSupervisor 4 1545 .23 .05 .41 .28 .091 .16 .39 31.28
satisfactionOverall 9 1946 .15 .03 .27 .19 0 .19 .19 100.00
performanceContextual 4 1015 .18 .00 .36 .23 .089 .12 .35 44.72
performancePolitical 5 979 .13 .23 .03 .17 0 .17 .17 100.00
perceptionsTenure 10 1429 .05 .13 .23 .06 .032 .02 .10 89.85Group 4 889 .22 .03 .47 .27 .140 .09 .45 24.60
cohesion
satisfaction (.32), supervisor satisfaction (.33), and organizational identi-fication (.36). It has a modest correlation with overall performance (.20)and is correlated somewhat more strongly (.28) with indicators of strain.Among the criteria of organizational longevity, tenure, and turnover, PJfit had correlations of .18 and .08, respectively. In the preentry con-text, PJ fit had strong correlations of .48 with organizational attractionand .67 with an organizations intent to hire. Almost all of the 80%credibility intervals for PJ fit did not include zero, with the exceptionof PJ fit and overall job performance (80% CV: .04 .44). There-fore, the generalizability of this relationship is less definite than theothers.
Based on the number of studies of each criterion available for moder-ator analysis, four moderators could be reliably examined. The first wasthe conceptualization of fit. Only five studies included a supplementaryconceptualization of PJ fit, and no more than three of these addressedthe same criterion. Therefore, a reliable comparison of supplementaryand complementary conceptualizations could not be conducted for PJ fit.Within the complementary category, however, separating the studies bydemandsabilities, needssupplies, or a combination of the two concep-tualizations explained more of the variance in the correlations in fourof the five cases where there were enough studies to perform moderator
AMY L. KRISTOF-BROWN ET AL. 307
TAB
LE4
Met
a-An
alys
isRe
sults
ofP
erso
n
Supe
rviso
rFit
95%
CI80
%CV
kN
Avg
rLo
wer
Upp
er
SD
Low
erU
pper
%var
expl
aine
dJo
bsa
tisfa
ctio
n5
1,19
9.35
.11
.59
.44
.11
2.30
.59
30.0
6O
rgan
izat
iona
lcom
mitm
ent
71,
346
.07
.15
.29
.09
.10
0.
04.21
42.5
4Pe
rcep
tions
ofp
oliti
cs5
853
.09
.21
.03
.12
0.
12.
1210
0.00
Supe
rviso
rsat
isfac
tion
591
8.39
.02
.80
.46
.22
8.17
.75
9.94
Lead
erm
embe
rex
chan
ge3
628
.37
.19
.55
.43
.07
6.34
.53
47.3
5Te
nu
re6
1,19
1.08
.10
.26
.09
.06
2.01
.17
60.4
3O
vera
llPe
rform
ance
143,
461
.15
.12
.42
.18
.15
3.
01.38
19.8
9D
irect
:Per
ceiv
ed8
1,17
9.25
.01
.49
.31
.11
4.16
.45
41.3
6In
dire
ct7
2,30
7.09
.09
.27
.11
.09
4.
01.23
33.1
2
308 PERSONNEL PSYCHOLOGY
TAB
LE5
Met
a-An
alys
isRe
sults
ofth
eIn
terr
elat
ions
hips
Betw
een
Type
sofF
it
95%
CI80
%CV
kN
Avg
rLo
wer
Upp
er
SD
Low
erU
pper
%var
expl
aine
dPO
PJ
1610
,239
.58
.38
.78
.72
.10
6.58
.85
30.0
5D
irect
:Per
ceiv
ed14
9,99
7.59
.43
.75
.74
.07
1.65
.83
49.5
3PJ
:Com
plem
enta
ry:D
A7
1,30
3.37
.02
.72
.45
.20
0.20
.71
16.0
8PJ
:Com
plem
enta
ry:N
S6
2,28
2.59
.47
.71
.73
.02
5.70
.76
89.8
7In
dire
ct:O
bjecti
ve
444
0.19
.12
.50
.24
.15
4.04
.43
35.7
0PO
PG
81,
607
.43
.10
.76
.54
.19
8.28
.79
16.0
7D
irect
:Per
ceiv
ed5
1,32
1.39
.25
.53
.48
.05
0.42
.55
70.9
3In
dire
ct:O
bjecti
ve
551
2.48
.13
1.00
.60
.37
1.12
1.00
8.02
POP
S4
976
.38
.03
.73
.46
.20
6.20
.72
11.3
6PJ
PG
377
5.38
.14
.62
.49
.13
3.32
.66
26.5
5PG
PS
371
3.30
.06
.54
.37
.13
2.20
.54
26.0
1
DA
=D
eman
dsA
bilit
iesfi
t.N
S=
Nee
dsS
kills
fit.
AMY L. KRISTOF-BROWN ET AL. 309
analyses.3 It should be noted that for PJ fit, conceptualization (demandsabilities vs. needssupplies) and content dimensions (KSAs vs. needs) arevirtually identical. Therefore, the analyses are relevant to both types ofmoderators.
Conceptualization of fit (or content dimension) acted as a modera-tor for job satisfaction, organizational commitment, intent to quit, andoverall performance but not for indicators of strain. As expected, the com-bined measure of PJ fit had a stronger correlation with attitudinal criteria,followed by needssupplies fit, and then demandsabilities fit. The dif-ferences in the relationships for each were job satisfaction (combined:.62; needssupplies: .61, demandsabilities: .41), organizational commit-ment (.51, .37, .31), and intent to quit (.57, .50, .23). For overallperformance only needssupplies and demandsabilities conceptualiza-tions could be compared, with the former being the stronger predictor(.20, .12). Although this does not support our prediction that demandsabilities fit would have a greater impact on performance, the effect sizesare more similar than for the attitudinal criteria. In contrast to the re-sults before the moderator analysis, the 80% credibility interval for the PJneedssuppliesoverall performance relationship did not include zero,supporting the generalizability of this particular relationship.
The second moderator examined was measurement strategydirectmeasures of perceived fit, indirect measures of subjective and objectivefit. Due to sample size constraints, the moderator analysis was conductedonly for the criteria of job satisfaction, organizational commitment, intentto quit, and overall performance. Like fit conceptualization, this analysisexplained more of the variance among the PJ fit correlations. We expectedthat the effects would be largest for direct measures of perceived fit, fol-lowed by indirect measures of subjective fit, and then objective fit. Thisprediction was supported for intent to quit (perceived: .49, subjective:.44, objective: .18) and overall performance (.22; .20; .16). Althoughthe small difference in effect sizes across these measurement strategiesfor the performance criterion is notable, partial support for our predictionwas found for job satisfaction, in which direct perceived (.58) and indirectsubjective (.59) effects were virtually identical, with indirect objective fit adistant third (.28), and for organizational commitment in which objectivefit was lowest (.21), but subjective fit (.53) was greater than perceived fit(.44).
3Because studies using a demandsabilities conceptualization of fit used KSAs as theexclusive content dimension, and those investigating needssupplies conceptualizationsused predominantly needs, we did not conduct a separate moderator analysis for contentdimensions.
310 PERSONNEL PSYCHOLOGY
The third moderator we examined was whether the sample containedonly one type of job (single-unit) or multiple jobs (multi-unit). We did notmake a prediction about the direction of this moderator, and the resultswere mixed. The single-job studies produced higher correlations for jobsatisfaction (single: .59, multiple: .55) and organizational commitment(single: .53, multiple: .43); whereas the multiple-job studies had largereffect sizes for intent to quit (single: .43, multiple: .48) and overallperformance (single: .10, multiple: .27).
Fourth, to examine the impact of common method bias on fit