《社会心理学》课程教学资源（文献资料）attitude&stereotype——Identifying Organizational Identification as a Basis for Attitudes and Behaviors - A Meta-Analytic Review

1054 LEE.PARK.AND KOO c-mail list service such as the HR(hum n resou quately measuring the self-defining nature of identification.In ist and the( 18 arti Inclusion Criteria and Sample cluded is in the supplemental materia ly ex types of identi tion (e of the corelation were noni nt bec they wer samp Thus whe d ween orga 191.Pp.63-64).Fifty onindependent Third,the study hado ort the were interested in th relations bet al id h as ic perfo other tha eha 05: Fishbein, Fourth the study had 1006 we using is the Mae ed multiple melations from study if they and saks 0 inal art e by Ash nultiple correlation were Ma meo ation from on tudy.th ot viola ampl ind this app ident This sc includes ation.we t elation Gie the 1d- of th is the The funnel plot t that sel circles that d of sample and his or her the mean effect size,with smaller variability as the We present more formal publication bias aire:Cheney 1983)ho er the operationalization of iden Coding of Studies t Questi naire (Mo outcome,national

published in the area of organizational identification, and we also announced a call for unpublished manuscripts on organizational researchers’ e-mail list service such as the HR (human resources) Listserv and the OB (organizational behavior) Listserv. Those combined efforts yielded 341 published articles and 16 unpub￾lished manuscripts and dissertations. Inclusion Criteria and Sample To be included in the meta-analysis, a study had to satisfy the following four criteria. First, the study had to address an organi￾zational identification issue, not other types of identification is￾sues, such as occupational identification, professional identifica￾tion, and subgroup/team identification (e.g., Täuber & Sassenberg, 2012; Wann, Waddill, Polk, & Weaver, 2011). Second, the study had to be an empirical one that reported correlations between organizational identification and individual-level outcome vari￾ables. We thus excluded theory or review papers and empirical papers that used qualitative methods because they lacked the necessary information on the correlations between organizational identification and outcomes. Ninety-one articles were excluded for those two reasons. Third, the study had to report the correlation(s) between organizational identification and its attitudinal and/or behavioral outcome(s). Therefore, studies were excluded when they only include correlations between organizational identifica￾tion and its antecedents (e.g., organizational prestige, perceived organizational support) or correlations between organizational identification and its outcomes other than attitudinal or behavioral ones (e.g., emotion, burnout, turnover intention). In this process, 99 articles were excluded. Fourth, the study had to measure orga￾nizational identification using the scales that emphasize an indi￾vidual’s oneness perception or the self-defining nature of identi￾fication. A representative example is the Mael scale (Mael & Ashforth, 1992). This scale has long been known to measure the oneness perception of organizational identification, partly because of its association with the seminal article by Ashforth and Mael (1989), and thus has been the most widely used in the organiza￾tional identification literature (Haslam, 2004). The Mael scale includes items such as “When someone criticizes [name of orga￾nization], it feels like a personal insult,” and “When I talk about [name of organization], I usually say ‘we’ rather than ‘they.’” The scale developed by van Dick, Wagner, Stellmacher, and Christ (2004b) has also been widely used, as it contains the self-defining nature of identification. This scale includes items such as “I identify myself as a member of [name of organization],” and “Being a member of [name of organization] reflects my personal￾ity well.” Another exemplifying scale is the graphical Venn dia￾gram scale which measures organizational identification as the degree of overlap between two circles that denote a respondent’s own identity and his or her organization’s identity (e.g., Bergami & Bagozzi, 2000). The studies included in our analysis, in which organizational identification was measured using the scales other than the Mael scale, are listed in Table 1. In some organizational identification scales (e.g., the Organizational Identification Ques￾tionnaire; Cheney, 1983), however, the operationalization of iden￾tification is similar to that of commitment, and thus they are not clearly distinguishable from organizational commitment scales such as the Organizational Commitment Questionnaire (Mowday, Steers, & Porter, 1979) and the Affective Commitment Scale (Allen & Meyer, 1990). Accordingly, we excluded studies using the identification scales similar to commitment scales, thus not adequately measuring the self-defining nature of identification. In this process, 18 articles were excluded from our meta-analysis. Table 1 also presents the list of those excluded studies. After applying all the inclusion criteria, we obtained an initial data set of 149 organizational identification– outcome correlations from 114 studies in 86 articles. The Appendix provides a summary of the studies and samples used in the meta-analysis, and the complete list of articles that were considered but ultimately ex￾cluded is in the online supplemental material. To calculate the overall correlation, we coded all possible zero￾order correlations from each study. Of the coded correlations, some of the correlations were nonindependent because they were computed from the same sample. Thus, when correlations were based on multiple measures of the same criterion in the same sample, such as intrinsic satisfaction and extrinsic satisfaction in Becker’s (1992) study, we combined the multiple measures into a composite using the composite formulas (Ghiselli, Campbell, & Zedeck, 1981, pp. 163–164). Fifty nonindependent correlations were combined for this reason. Yet, many studies still yielded more than one relevant correla￾tion because we were interested in the relations between organi￾zational identification and several different types of individual outcomes, such as job satisfaction, affective organizational com￾mitment, and in-role performance. In such cases, adopting the approach used in the previous meta-analysis studies (e.g., Albar￾racín et al., 2005; Albarracín, Johnson, Fishbein, & Muellerleile, 2001; Baas, De Dreu, & Nijstad, 2008; Durantini, Albarracín, Mitchell, Earl, & Gillette, 2006), we allowed more than one correlation per study to be included in our final sample; that is, even after combining correlations for reasons of independence, we still used multiple correlations from one study if they concerned different types of outcomes—job satisfaction and in-role perfor￾mance in Ashforth, Sluss, and Saks’ (2007) study, for example. Although multiple correlations were used from one study, we ensured that, in the analysis of the correlation between organiza￾tional identification and each specific outcome, we use only one correlation from one study, thus not violating the sample indepen￾dence assumption (Cooper & Hedges, 1994). However, in the meta-analysis of the overall correlation between organizational identification and all outcomes, this approach could still violate the independence assumption. Hence, when we analyze this overall correlation, we used one correlation (i.e., the composite correla￾tion) per study. Table 2 shows the stem-and-leaf display of the 114 independent correlations, and Figure 3 shows the funnel plot of the correlations. The funnel plot provides initial evidence that selec￾tion or publication bias is unlikely to be present because the distribution of sample is symmetrical and the form of a funnel centers on the mean effect size, with smaller variability as the sample size increases. We present more formal publication bias tests below. Coding of Studies The studies were coded independently by two coders for corre￾lation, sample size, reliability estimates, type of organizational identification outcome, national culture, and study characteristics (e.g., data structure, publication status). The initial intercoder This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 1054 LEE, PARK, AND KOO

THE EFFECTS OF ORGANIZATIONAL IDENTIFICATION 105S Table1 Inclusion Decisions for the Studies Not Using the Mael Scale Organizational identification scale type Study Measure sourc Nedeveloped scales ince the Mael Samples I to 7 t201 2007 de L2010) Hofacker (2011 Salesiaiartooganizaionadcoammimean henev (1983 bel (1983) 00 201 Tyagi w ha199399 199 Not a ailab o in-role performanc cture. pt0a2000. hen's 0D1 Disa s were d into extra- ole per the stud ntually resulted in an intercoder agreemen mal iob de mple we oded varia ance.customer es of organizational identificatio nance citize voice chavior.and and affective orga commitment.Beho ral ou Appendix ucts of cal re in-role performance and extra-role performance using several dif- fected by the self-serving bias (e.gBorman.99:Spector,1994)

agreement rates were high, ranging from 98% to 100% for the coding of job involvement, job satisfaction, affective organiza￾tional commitment, extra-role performance, national culture, data structure, and publication status. The initial intercoder agreement was somewhat lower for the coding of in-role performance (Co￾hen’s .92, p  .001). Disagreements between raters were resolved by discussing coding criteria and further examination of the studies, which eventually resulted in an intercoder agreement rate of 100% for all coded information. Outcomes of organizational identification. Two categories were used to code the outcomes of organizational identification: attitudinal and behavioral outcomes. Attitudinal outcomes were coded using three subcategories: job involvement, job satisfaction, and affective organizational commitment. Behavioral outcomes were coded using two subcategories: in-role performance and extra-role performance. Some studies measured the constructs of in-role performance and extra-role performance using several dif￾ferent variable names. In such cases, we classified the variables into in-role and extra-role behaviors based on their definitions; performance behaviors were classified into in-role performance when they were required by formal job descriptions, directly serving the goals of the organization (van Knippenberg, 2000), while performance behaviors were classified into extra-role per￾formance when they concerned discretionary actions beyond for￾mal job descriptions (Podsakoff et al., 2000). For example, we coded variables such as productivity, job performance, customer￾oriented service behavior, and work effort into in-role perfor￾mance, and coded variables such as organizational citizenship behavior, helping behavior, voice behavior, and safety perfor￾mance into extra-role performance. We report all of our classifi￾cation information in the Appendix. In addition, we coded the rater of performance variables because several researchers noted that self-reported measures can be af￾fected by the self-serving bias (e.g., Borman, 1991; Spector, 1994). Table 1 Inclusion Decisions for the Studies Not Using the Mael Scale Organizational identification scale type Study Measure source Inclusion decision Newly developed scales since the Mael scale Gümüs et al. (2012) van Dick et al. (2004b) Included Michel et al. (2010) van Dick et al. (2004b) Included van Dick et al. (2004a) van Dick et al. (2004b) Included van Dick et al. (2006) Samples 1 to 7 van Dick et al. (2004b) Included van Dick et al. (2007) van Dick et al. (2004b) Included van Dick et al. (2008) van Dick et al. (2004b) Included Smidts et al. (2001) Smidts et al. (2001) Included Walumbwa et al. (2009) Smidts et al. (2001) Included Walumbwa et al. (2011) Smidts et al. (2001) Included Zhao et al. (2014) Smidts et al. (2001) Included Edwards & Peccei (2010) Edwards & Peccei (2007) Included Fuchs & Edwards (2012) Edwards & Peccei (2007) Included Peters et al. (2010) Doosje et al. (1995) Included Richter et al. (2006) Doosje et al. (1995) Included Norman et al. (2010) Avey et al. (2008) Included Christ et al. (2003) Christ et al. (2003) Included Johnson et al. (2012) Johnson et al. (2012) Included Stoner & Gallagher (2011) Stoner, Perrewé, & Hofacker (2011) Included Amiot et al. (2006) Terry & Hogg (1996) Included Hassan (2010) Tyler & Blader (2000) Included Graphical scales Bartel (2001) Bergami & Bagozzi (2000) Included Korschun et al. (2011) Bergami & Bagozzi (2000) Included Mayfield & Taber (2010) Shamir & Kark (2004) Included Wolfe (2007) Shamir & Kark (2004) Included Scales similar to organizational commitment scales Balfour & Wechsler (1991) Cheney (1983) Excluded Gautam et al. (2004) Cheney (1983) Excluded Ishii (2012) Cheney (1983) Excluded Johnson et al. (1996) Cheney (1983) Excluded Sass & Canary (1991) Cheney (1983) Excluded Scott et al. (1999) Cheney (1983) Excluded van Dick et al. (2006) Sample 10 Cheney (1983) Excluded Wolf (2009) Cheney (1983) Excluded Gould & Werbel (1983) Patchen (1970) Excluded Popoola (2005) Patchen (1970) Excluded Rotondi (1975) Patchen (1970) Excluded Jetten et al. (2002) Allen & Meyer (1996) Excluded Olkkonen & Lipponen (2006) Allen & Meyer (1990) Excluded O’Reilly & Chatman (1986) O’Reilly & Chatman (1986) Excluded Leavitt et al. (2011) O’Reilly & Chatman (1986) Excluded Kolodinsky et al. (2008) Efraty et al. (1991) Excluded Millward & Brewerton (1999) Millward & Brewerton (1999) Excluded Tyagi & Wotruba (1993) Not available Excluded This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. THE EFFECTS OF ORGANIZATIONAL IDENTIFICATION 1055

1056 LEE.PARK.AND KOO 心 and Leaf Display Correlations (r) organizational identification literature,we coded Stem Leaf 8 Meta-Analysis Procedure 23 elations 456 之k的 utcomes into self-reported "reported,and lata-ba &Russell.2000).Carpenter.Berry. itystimate based o the estim s from other studies in our Schm 4).The artifac of al id 82 and s of the cri ation (OCB-O).folowin the ob satist mmance:and mean 79 and standard deviation 09 geneity amons Canada,Ger amp all an Aus nd Japan.to in the set of rue popuati ifving the location d to each study the 2010)three and the United Study characteristics.We coded two aspects of study for addit hory mo ructure.for example ethod variance (Podsakoff,MacKenzie,Lee,&Podsakoff omes),and panel data (correlations between average org onal identi 24 6 publication of studies that have statisti ally significant find ings,has been an important issue for psychological science for a Figure3.Funnel plot of independent (

We distinguished in-role performance outcomes into self-reported, other-reported, and data-based. Similarly, we distinguished extra￾role performance outcomes into self-reported and other-reported (Allen, Barnard, Rush, & Russell, 2000; Carpenter, Berry, & Houston, 2014). No study in our sample measured extra-role performance based on objective data. Moreover, we also distin￾guished extra-role performance by target: organizational citizen￾ship behavior toward individuals (OCB-I) and toward the organi￾zation (OCB-O), following the previous notion of the distinctive characteristics of the two (Podsakoff, Whiting, Podsakoff, & Blume, 2009; Williams & Anderson, 1991). National culture. We coded national culture using the values provided by Hofstede, Hofstede, and Minkov (2010), as de Wit, Greer, and Jehn (2012) did in their meta-analysis. We first iden￾tified the geographical location in which the study was conducted. Example countries included in our study are the United States, Canada, Germany, China, France, the Netherlands, the Philippines, Thailand, Singapore, Pakistan, Australia, and Japan, to list a few (see the Appendix for all of the country information). After iden￾tifying the geographical locations, we assigned to each study the associated values of Hofstede et al.’s (2010) three cultural value dimensions: individualism/collectivism, long-term/short-term ori￾entation, and uncertainty-avoidance. For example, the United States was coded 91 for individualism, 29 for long-term￾orientation, and 46 for uncertainty-avoidance. The same dimen￾sions were coded 20, 118, and 30, respectively, for studies con￾ducted in China. Study characteristics. We coded two aspects of study char￾acteristics for additional exploratory moderator analyses. First, several researchers (e.g., Judge, Thoresen, Bono, & Patton, 2001) have noted that data structure, for example cross-sectional versus longitudinal, can influence correlations; cross-sectional data may present higher correlations than longitudinal data because of com￾mon method variance (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). Thus, we coded data structure into three types: cross￾sectional data (concurrent measures of organizational identifica￾tion and its outcomes), longitudinal data (time separation between the measurement of organizational identification and its out￾comes), and panel data (correlations between average organiza￾tional identification across times and average outcomes across times). Second, the issue of publication bias, defined as the selec￾tive publication of studies that have statistically significant find￾ings, has been an important issue for psychological science for at least three decades (e.g., Ferguson & Brannick, 2012; Rosenthal, 1979, 1995). To investigate whether publication bias exists in the organizational identification literature, we coded publication status into two categories: studies published in academic journals versus unpublished manuscripts and dissertations. Meta-Analysis Procedure All the correlations were first corrected for measurement error in the independent and dependent variables. We used the internal consistency coefficients reported in the respective study as the reliability estimates. When reliability estimates were available, we divided individual effect sizes by the square root of the reliability estimates of the two correlated variables, as recommended by Hunter and Schmidt (2004). Among the initial sample of 149 correlations, 12 correlations (8%) did not have the reliability estimate information for the organizational identification variable and 36 correlations (24%) did not have the reliability estimate information for the outcome variables. For studies that did not report a reliability estimate, we assigned them the average reliabil￾ity estimate based on the estimates from other studies in our sample (Hunter & Schmidt, 2004). The artifact distribution of the reliability of organizational identification was a mean of .82 with a standard deviation of .07. The artifact distributions of the crite￾rion variables’ reliabilities were mean .77 and standard deviation .11 for job involvement; mean .80 and standard deviation .09 for job satisfaction; mean .83 and standard deviation .05 for affective organizational commitment; mean .77 and standard deviation .11 for in-role performance; and mean .79 and standard deviation .09 for extra-role performance. We calculated meta-analytic correlations using a random effects model (Hunter & Schmidt, 2004) to consider heterogeneity among the studies. The random effects model assumes that sampling error plus variability in the population of the correlations (unique dif￾ferences in the set of true population correlations) caused the variability among the organizational identification– outcome cor￾relations. As an estimate of variability, we calculated 95% confi￾dence intervals and 80% credibility intervals. A 95% confidence interval excluding zero indicates that one can be 95% confident Table 2 Stem and Leaf Display of 114 Correlations (r) Stem Leaf .1 .0 5,4 .0 1,2,3,3,3,4,5,5,5,5,6,8,8,9,9,9 .1 1,2,3,4,5,5,6,8,8,8,8,9,9 .2 0,0,0,0,0,0,0,1,2,2,2,3,3,3,4,4,4,5,5,5,6,6,7,7,7,8,8,9,9,9,9,9 .3 0,0,1,1,1,1,2,2,3,3,3,3,3,4,4,4,5,6,6,6,6,6,6,7,7,8,8,9,9,9 .4 0,0,0,1,1,1,1,1,1,2,3,3,3,3,4,4,5,6,6,6,7,7,8,8,8,8,8,9,9 .5 0,0,0,0,1,1,1,2,2,2,3,3,3,4,5,6,6,7,7,8,8,8 .6 1,3,5,6,6 .7 3 .8 Figure 3. Funnel plot of 114 independent correlations (r). This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 1056 LEE, PARK, AND KOO

THE EFFECTS OF ORGANIZATIONAL IDENTIFICATION 1057 maximem erass truc correlation dimerent moor thus.a The second panel of Table 3presents the tests of the correlations onal ide s repored ar of-fit statistic showed that the correlati (0aD=17.96.p01 nable to a 35.95%C[32.39 ing two indices:the O-s i 95%C44.54 f Table 3 present the 003)The whether a statistically signif ficant level of variability exists in 200 =.64.95%C[55.72 o in observed effec The fourth set of resul in Table 3 reports the tests of the g The that the Compared ith the tatistic.the Fis known to he mes were signif ded (Bo aling of 9).However.the F should on 2002 the meta-analytic 95C36. 09 ince (p omes on deling (MASEM:Vi ran One 1995)to tes and in-role performance was ed n.t n and its out mes.we com obiective data .19.95%CI [.03.34D).The relation Lipsey and Wilson (2001). highe wh en-rated n (45.9s Results Relations Between Organizational Identification 103.p1).In additio we also examin and Outcomes performace yarCB-n OCB- that they were ot signficantly differe) vailable correlati The fifth panel of Table 3 shows how the organizational T-LI ding on d nd antly moderate the (02)-209.0 gani7a dence interval did o %C[37,.45.Hou ver.the showed arge ce:the 158713.P)This indicates that the We used MASEM (Viswesvaran &On 1995)to test how naybas mod nizational identification and outcomes. MASEM analysis.we first constructed the matrix of meta-

that the average true correlation is different from zero; thus, a maximum of 2.5% are larger than the upper bound of the interval and fewer than 2.5% are smaller than the lower bound. A credi￾bility interval indicates whether the correlations reported are gen￾eralizable to other samples. For example, an 80% credibility in￾terval excluding zero indicates that at least 80% of the correlations reported are different from zero. In addition, we conducted homo￾geneity analysis (Lipsey & Wilson, 2001), which tests whether it is reasonable to assume that all effect sizes are estimating the same population mean, using two indices: the Q-statistic, which indi￾cates the level of variance across study results relative to the sampling error variance (Hedges & Olkin, 1985), and the I-squared (I 2 ) statistic, another indicator of homogeneity statistic (Higgins, Thompson, Deeks, & Altman, 2003). The Q-statistic indicates whether a statistically significant level of variability exists in correlations across studies, by calculating the categorical model results for the between-groups goodness-of-fit statistic (QB) and the within-group goodness-of-fit statistic (QW; Field, 2001; Hedges & Olkin, 1985; Lipsey & Wilson, 2001). The I 2 indicates the ratio of true heterogeneity to total variation in observed effect sizes, which ranges from 0% to 100% with higher values indicat￾ing greater heterogeneity of effect sizes and higher likelihood of moderators. Compared with the Q-statistic, the I 2 is known to be less affected by the scaling of the measures or the number of the studies included (Borenstein, Hedges, Higgins, & Rothstein, 2009). However, the I 2 should also be interpreted with caution because it depends on the size of individual studies (Higgins & Thompson, 2002). Finally, we created a correlation matrix using the meta-analytic correlations and conducted fixed-effect meta-analytic structural equation modeling (MASEM; Viswesvaran & Ones, 1995) to test our hypothesized path models (in Figures 1 and 2). In addition, to test the moderating effect of national culture on the relations between organizational identification and its outcomes, we con￾ducted a random-effects weighted least square (WLS) regression analysis using the meta-regression SPSS syntax developed by Lipsey and Wilson (2001). Results Relations Between Organizational Identification and Outcomes Table 3 presents the analysis of the relations between organiza￾tional identification and its attitudinal/behavioral outcomes using the available correlations. The top panel of Table 3 reports the meta-analysis results using all available independent correlations (k 114; N 36,526). The average corrected correlation between organiza￾tional identification and all attitudinal/behavioral outcomes across all studies was positive ˆ .41), and the 95% confi￾dence interval did not include zero (95% CI [.37, .45]). How￾ever, the corrected correlation showed large variance; the ho￾mogeneity of effect size tests was significant across the analyses (QW 1587.13, p  .01). This indicates that the use of a random effects model is justifiable. It also indicates that moderators may be present for the associations between orga￾nizational identification and outcomes. The second panel of Table 3 presents the tests of the correlations between organizational identification and outcomes by outcome type: attitudinal outcomes and behavioral outcomes. The between￾groups goodness-of-fit statistic QB showed that the correlations between organizational identification and outcomes were signifi￾cantly different across outcome types (QB(1) 17.96, p  .01). Specifically, the results revealed that the size of the correlation between organizational identification and behavioral outcomes ˆ .35, 95% CI [.32, .39]) was smaller than that between organizational identification and attitudinal outcomes ˆ .49, 95% CI [.44, .54]). The third and fourth panels of Table 3 present the organizational identification correlations with detailed outcomes within each out￾come type. The third set of results in Table 3 shows that organi￾zational identification was significantly and positively related to job involvement ˆ .50, 95% CI [.39, .62]), job satisfaction ˆ .45, 95% CI [.40, .50]), and affective organizational commitment ˆ .64, 95% CI [.55, .72]). The fourth set of results in Table 3 reports the tests of the relations between organizational identification and behavioral out￾comes. The QB indicates that the correlations between organiza￾tional identification and behavioral outcomes were significantly different across performance types (QB(1) 19.64, p  .01). Specifically, the correlation between organizational identification and in-role performance was significant and positive ˆ .27, 95% CI [.20, .34]) and was smaller than the correlation between organizational identification and extra-role performance ˆ .42, 95% CI [.36, .48]). We further classified both performance out￾comes based on the rater of performance. The relation between organizational identification and in-role performance was rela￾tively high when in-role performance was self-rated ˆ .33, 95% CI [.23, .38]) and low when in-role performance was rated based on objective data ˆ .19, 95% CI [.03, .34]). The relation between organizational identification and extra-role performance was higher when extra-role performance was measured using self-rated measures ˆ .48, 95% CI [.43, .52]) than when it was measured using other-rated measures ˆ .29, 95% CI [.18, .39]) and the difference between the two was statistically significant (QB(1) 10.38, p  .01). In addition, we also examined whether the relation between organizational identification and extra-role performance differed by target—OCB-I and OCB-O—and the results showed that they were not significantly different (QB(1) 2.02, ns). The fifth panel of Table 3 shows how the organizational identification– outcome correlations change depending on data structure—the study design used to collect the data. The results indicated that data structure did not significantly moderate the relations between organizational identification and its outcomes (QB(2) 2.09, ns). Positioning of Organizational Identification in General Attitude–Behavior Relations We used MASEM (Viswesvaran & Ones, 1995) to test how organizational identification functions in the nomological net￾work of attitudinal and behavioral variables. To conduct the MASEM analysis, we first constructed the matrix of meta￾This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. THE EFFECTS OF ORGANIZATIONAL IDENTIFICATION 1057

1058 LEE.PARK.AND KOO Variable kN7SD,方SD。95%C 80%CVPQ。 Q All outcon ndent correlation 1436.52634214121【37.4114.69193% 17.96 1587.1 oral ou 的出智治做监 13.97 tisfaction nalcommitment oral outcome 4113,87022132726【20.341[-.0660127% 19.64 e performance by rate 266 L18.66] 10.38- 9器”品88保刻傲绸 4 202 4胸器格名9路纷制 2.09 70 g站R4竖烟版 ations from independent samples:N=total sample size for all studies combined:=sample-size weiehted mean ower ithin group:OCB-Io lity in alytic com in this row may not be Is less 3.but we decided to repo this for informational purposes D<01 SEM tical,wecalculated the harmonic mean of the tion andecvo ults show that Model 1 did not fit the dat M analyses (Viswesvaran&Ones 995 .CF the arithn e it a uch less weight to al identification as a bas cker-Lewis inde TLI;also known as the non-normed fit index,Tucker&Lewis ttitude and general beh havior (Model 2 in Figure 2).When Mode han 90.an RMSEA less than or equal to .08.and an SRMR ally used in sem to com sted models estimated than1(Kline,2005 Reise,&Kim,2007 ison of tw ition ty in the attitude 673.42).As 2 fit the data

analytic correlations among all of the study variables (see Table 4). We then entered this meta-analytic correlation matrix into an SEM analysis using AMOS (Version 18; Arbuckle, 2009). As the sample sizes across the various cells of the matrix were not identical, we calculated the harmonic mean of the sample sizes and used it for the SEM analyses (Viswesvaran & Ones, 1995). The harmonic mean is a more conservative parameter estimate than the arithmetic mean because it assigns much less weight to large sample sizes. We used five model fit indices—chi-square (2 ), the comparative fit index (CFI), the Tucker-Lewis index (TLI; also known as the non-normed fit index, Tucker & Lewis, 1973), the root-mean-square error of approximation (RMSEA), and the standardized root-mean-square residual (SRMR)—to investigate the viability of the structural models. Desirable model fit is associated with a CFI greater than .90, a TLI greater than .90, an RMSEA less than or equal to .08, and an SRMR less than .10 (Kline, 2005). Table 5 presents the comparison of two alternative models that position organizational identification differently in the attitude– behavior relations. First, we tested the model of organizational identification as a component of general attitude that treats orga￾nizational identification as constituting a general set of attitudes along with other types of attitudes (job involvement, job satisfac￾tion, and affective organizational commitment; Model 1 in Figure 1). The MASEM results show that Model 1 did not fit the data particularly well (2 [8] 647.42, CFI .95, TLI .90, RMSEA .11, SRMR .04). Next, we tested the model of organizational identification as a basis for general attitude and behavior that treats organizational identification not as a part of general attitude but as an independent predictor of both general attitude and general behavior (Model 2 in Figure 2). When Model 2 was estimated, all of the fit indices reached acceptable levels (2 [7] 344.21, CFI .97, TLI .94, RMSEA .08, SRMR .03). In addition, to further compare the two models, we utilized Akaike’s information criterion (AIC; Akaike, 1974), which is generally used in SEM to compare non-nested models estimated using the same data (Henson, Reise, & Kim, 2007; Kline, 2005). The AIC for Model 2 (372.21) was smaller than that for Model 1 (673.42). As a lower AIC value indicates a better fit, this result indicates that Model 2 fit the data better than Model 1. Table 3 Meta-Analysis of Bivariate Relations Between Organizational Identification and Outcome Variables Variable kNr SDr ˆ SD 95% CI 80% CV I 2 QB QW All outcomes All independent correlations 114 36,526 .34 .21 .41 .21 [.37, .45] [.14, .69] 93% 1587.13 Outcome type 17.96 150.78 Attitudinal outcomes 55 18,084 .41 .14 .49 .22 [.44, .54] [.25, .73] 0% 46.90 Behavioral outcomes 94 43,754 .29 .19 .35 .19 [.32, .39] [.11, .60] 10% 103.88 Attitudinal outcomes 13.97 55.71 Job involvement 6 3,939 .41 .12 .50 .15 [.39, .62] [.31, .69] 42% 8.68 Job satisfaction 37 11,216 .37 .12 .45 .18 [.40, .50] [.22, .68] 0% 32.25 Affective organizational commitment 12 2,929 .52 .14 .64 .17 [.55, .72] [.42, .86] 26% 14.77 Behavioral outcomes 19.64 94.29 In-role performance 41 13,870 .22 .13 .27 .26 [.20, .34] [.06, .60] 27% 39.78 In-role performance by rater 2.66 40.43 Self-rated 20 8,468 .27 .13 .33 .18 [.23, .38] [.13, .53] 14% 22.16 Other-rated 16 3,315 .18 .16 .23 .16 [.14, .31] [.03, .43] 5% 15.83 Data-based 5 2,087 .09 .13 .19 .18 [.03, .34] [.04, .42] 0% 2.44 Extra-role performance 53 14,459 .35 .15 .42 .22 [.36, .48] [.18, .66] 5% 54.52 Extra-role performance by rater 10.38 48.75 Self-rated 39 9,928 .38 .12 .48 .19 [.43, .52] [.27, .68] 0% 34.66 Other-rated 9 1,753 .24 .12 .29 .15 [.18, .39] [.08, .49] 21% 14.09 Extra-role performance by target 2.02 22.94 OCB-I 5 927 .23 .18 .27 .20 [.09, .45] [.01, .53] 0% 1.34 OCB-O 18 4,089 .34 .17 .42 .21 [.32, .51] [.15, .69] 21% 21.60 Study characteristics Data structure 2.09 149.41 Cross-sectional 138 59,419 .34 .12 .41 .23 [.38, .45] [.16, .67] 0% 140.33 Longitudinal 10 2,275 .26 .16 .32 .19 [.19, .44] [.05, .58] 0% 9.08 Panela 1 144 .36 .17 .39 .21 [.01, .79] [.13, .65] — — Publication status .70 149.67 Academic journal articles 140 44,188 .34 .12 .41 .24 [.38, .44] [.15, .67] 3% 143.56 Unpublished manuscripts/dissertations 9 2,225 .29 .18 .35 .21 [.22, .48] [.09, .61] 0% 6.12 Note. k number of correlations from independent samples; N total sample size for all studies combined; r sample-size weighted mean correlation; SDr sample-size weighted observed standard deviation of correlation; ˆ mean true-score correlation corrected for measurement error in both predictor and criterion measures; SD standard deviation of the true-score correlation corrected for measurement error in both predictor and criterion measures; 95% CI lower and upper limits of 95% confidence interval; 80% CV lower and upper limits of 80% credibility interval; I 2 homogeneity statistic; QB homogeneity statistic Q between groups; QW homogeneity statistic Q within group; OCB-I organizational citizenship behavior toward individuals; OCB-O organizational citizenship behavior toward the organization. a The meta-analytic correlation in this row may not be meaningful because k is less than 3, but we decided to report this for informational purposes. p  .01. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 1058 LEE, PARK, AND KOO