Psychological Collectivism

Journal of Applied Psychology Copyright 2006 by the American Psychological Association 2006, Vol. 91, No. 4, 884 – 899 0021-9010/06/$12.00 DOI: 10.1037/0021-9010.91.4.884 Psychological Collectivism: A Measurement Validation and Linkage to Group Member Performance Christine L. Jackson Jason A. Colquitt Purdue University University of Florida Michael J. Wesson Cindy P. Zapata-Phelan Texas A&M University University of Florida The 3 studies presented here introduce a new measure of the individual-difference form of collectivism. Psychological collectivism is conceptualized as a multidimensional construct with the following 5 facets: preference for in-groups, reliance on in-groups, concern for in-groups, acceptance of in-group norms, and prioritization of in-group goals. Study 1 developed and tested the new measure in a sample of consultants. Study 2 cross-validated the measure using an alumni sample of a Southeastern university, assessing its convergent validity with other collectivism measures. Study 3 linked scores on the measure to 4 dimensions of group member performance (task performance, citizenship behavior, counterproductive behavior, and withdrawal behavior) in a computer software firm and assessed discriminant validity using the Big Five. The results of the studies support the construct validity of the measure and illustrate the potential value of collectivism as a predictor of group member performance. Keywords: collectivism, individual differences, teams, groups, personality As the nature of work has changed, cooperation among employ- Nason, & Smith, 1999; Murphy, 1999; Perrewe & Spector, 2002). ees has become increasingly important (Hedge & Borman, 1995; Highly collective individuals see themselves as members of one or Mohrman & Cohen, 1995; Motowidlo & Schmit, 1999). Work is more in-groups, are primarily motivated by the norms of those increasingly scheduled around group-based projects rather than in-groups, prioritize the goals and well-being of those in-groups, individual job descriptions, with individual employees working in and emphasize their connectedness to other in-group members several different groups at any given time (Ilgen & Pulakos, 1999; (Triandis, 1995). Scholars have theorized that collectivism could Mohrman & Cohen, 1995). The performance evaluation and com- enhance one’s tendency to cooperate in group contexts (Chen, pensation systems that govern the employees in such groups often Chen, & Meindl, 1998; Cox, Lobel, & McLeod, 1991; Earley & build in cooperative components, with members appraised and Gibson, 1998; Wagner, 1995). Still, a fundamental question re- rewarded on the basis of shared achievements (e.g., Bartol & mains largely unexplored: Does collectivism predict members’ job Durham, 2000; DeMatteo, Eby, & Sundstrom, 1998). Such sys- performance in work group settings? tems may also supplement the traditional focus on “taskwork” behaviors by considering “teamwork” behaviors on the part of One reason for the lack of research linking collectivism to group individual group members (e.g., LePine, Hanson, Borman, & member job performance may be the reliability and validity of Motowidlo, 2000). As a result, the very conceptualization of job collectivism measures. In a recent review, Earley and Gibson performance has changed, as the concept of “doing a good job” has (1998) concluded, “What seems to be a safe, but pessimistic, become more dependent on the concept of “being a good group conclusion is that if further research is to be conducted using member.” individualism– collectivism, then continued efforts need to be made in improving its measurement” (p. 296; see also Kagitcibasi, As job performance takes on a more cooperative character, it 1994; Oyserman, Coon, & Kemmelmeier, 2002; Triandis, Chan, becomes necessary to identify employees with a propensity to be Bhawuk, Iwao, & Sinha, 1995). With that in mind, we report the effective group members. Reviews in the personality and teams results of three independent studies that (a) validate a new measure literatures have speculated that collectivism could take on a new of collectivism for use in work contexts and (b) link scores on that importance in the contemporary workplace (Kozlowski, Gully, measure to group member performance. The first study developed and tested the new measure in a sample of management consul- Christine L. Jackson, Department of Management, Krannert School of tants. The second study cross-validated the measure in a university Business, Purdue University; Jason A. Colquitt and Cindy P. Zapata- alumni sample and examined its convergent validity with other Phelan, Department of Management, University of Florida; Michael J. collectivism scales. The third study assessed the discriminant Wesson, Department of Management, Texas A&M University. validity of the new measure with established individual differences and tested the predictive validity of the measure by linking it to Correspondence concerning this article should be addressed to Christine supervisor ratings of group member job performance in a sample L. Jackson, Department of Management, Krannert School of Business, of computer software developers. Purdue University, 403 West State Street, West Lafayette, IN 47907-2076. E-mail: [email protected] 884

PSYCHOLOGICAL COLLECTIVISM 885 Collectivism: Origins and Measurement Families of Measures Much of the attention devoted to collectivism can be attributed Most of the major measures of collectivism can be categorized to Hofstede’s (1980) cross-cultural study that identified into three principal families that tend to share items across scale individualism– collectivism as one of four major cultural variables. versions. Appendixes A and B summarize these families, including Scholars have continued to examine individualism– collectivism as the articles that introduced and revised the collectivism scales, the a cultural variable, often using the country rankings provided by in-groups referenced by those scales, the labels for scale facets in Hofstede as proxies. Others have developed new scales using cases where the measure was multidimensional, and sample items. Hofstede’s variables as a basis for their research. For example, the We also provide the number of Web of Science citations from Global Leadership and Organizational Behavior Effectiveness 2000 to 2005 for each of the articles, to provide a sense of the (i.e., GLOBE) Project created scales to assess two forms of col- current visibility of the various scales. The first family, shown in lectivism (one referenced to the collective distribution of resources Appendix A, is the Triandis family, introduced in Hui’s (1988) and one referenced to the cohesion of organizations and families) dissertation, based on earlier work (Hui & Triandis, 1986). The with the intention of predicting organizational practices and leader first four versions of the Triandis scales included between five and attributes and behaviors (House, Javidan, Hanges, & Dorfman, seven facets (Hui, 1988; Triandis et al., 1986; Triandis, Bontempo, 2002). Villareal, Asai, & Lucca, 1988; Triandis, McCusker, & Hui, 1990), whereas the more recent versions collapsed those facets into two Other studies have operationalized individualism and collectiv- broader dimensions (Gelfand & Realo, 1999; Singelis, Triandis, ism as individual-difference variables. For example, Triandis and Bhawuk, & Gelfand, 1995; Triandis & Gelfand, 1998). The second colleagues proposed that both could be viewed as psychological family, summarized in Appendix B, is the Earley family, intro- dimensions that correspond to the constructs at the cultural level duced in Erez and Earley (1987) and revised and expanded in (Triandis, Leung, Villareal, & Clack, 1985). The authors suggested Earley (1993). Unlike the Triandis scales, Earley’s measures are that the individual version of collectivism be termed allocentrism intended to capture a unidimensional form of collectivism. Appen- and the individual version of individualism be termed idiocen- dix B also summarizes the Wagner family. The measure intro- trism. The value of this perspective can be seen in arguments by duced in Wagner and Moch (1986) includes three facets, reflecting Triandis and Suh (2002), who suggested that only around 60% of collective beliefs, values, and norms. Wagner (1995) introduced a the members of an individualistic culture will be idiocentric as different version that utilized many of the Triandis items. individuals and 60% in collective cultures will be allocentric. This view is consistent with a recent meta-analysis by Oyserman et al. Scales from all three of the major families have struggled with (2002), which found that between-culture differences in collectiv- issues of reliability. Hui’s (1988) own validation study yielded an ism were “neither as large nor as systematic as often perceived” (p. average reliability of .60 for his subscales. Studies using Singelis 40). Like Triandis et al. (1985), the studies reported here cast et al.’s (1995) scales have found reliabilities as low as .59 (Oishi, collectivism as an individual difference. However, we eschewed Schimmack, Diener, & Suh, 1998; Probst, Carnevale, & Triandis, the allocentrism label in favor of psychological collectivism, a less 1999; Robert, Probst, Martocchio, Drasgow, & Lawler, 2000; confusing term that has been used occasionally in the literature Triandis & Singelis, 1998). In a review article, Earley reported that (e.g., Hui, Triandis, & Yee, 1991). An individual-difference-based his measures had poor psychometric properties and recommended view of collectivism has become more common in organizational that other scales be used (Earley & Gibson, 1998). Some studies behavior, with several studies occurring within one culture (e.g., using Wagner and Moch’s (1986) scale have also failed to reach Colquitt, 2004; Colquitt, Noe, & Jackson, 2002; Cox et al., 1991; the .70 standard (Chen, Meindl, & Hunt, 1997; Eby & Dobbins, Dickson & Weaver, 1997; Eby & Dobbins, 1997; Moorman & 1997). Finally, Wagner’s (1995) scale has failed to reach the .70 Blakely, 1995; Pillai & Meindl, 1998; Van Dyne, Vandewalle, standard in some instances, with reliabilities as low as .58 for the Kostova, Latham, & Cummings, 2000) or using an individual- subscales (Ramamoorthy & Carroll, 1998). These reliability prob- difference approach in multiculture studies (e.g., Earley, 1994; lems were noted in Oyserman et al.’s (2002) review, which esti- Gibson, 1999; Gomez, Kirkman, & Shapiro, 2000; Kirkman & mated that half of the cross-cultural studies on collectivism have Shapiro, 2001). used scales with poor reliability. The studies reported here also follow an emerging consensus The collectivism scales, as a set, have also been criticized with in the literature that collectivism and individualism are not respect to content validity (Earley & Gibson, 1998; Fiske, 2002; polar opposites but rather orthogonal, independent constructs. Kagitcibasi, 1994; Oyserman et al., 2002; Triandis, 1995). For Oyserman et al.’s (2002) review summarized this position as example, Earley and Gibson (1998) argued, “Quite frankly, if one follows: simply observed the highly varied operationalizations of individualism– collectivism without reference to the underlying Although the assumption of IND being the conceptual opposite of construct, it might appear that these measures tap unrelated con- COL may be intuitively appealing, an accumulation of recent research structs” (p. 291). In commenting on the inconsistent results of suggests this simple approach does not sufficiently represent the Oyserman et al.’s (2002) meta-analytic review, Fiske (2002) impact of IND and COL on basic psychological processes. Instead, concluded, IND and COL are better understood as domain-specific, orthogonal constructs. (p. 8; see also Ho & Chiu, 1994; Triandis, 1995) This can only mean that the various scales are not measuring the same two constructs—in other words, the various scales that are intended to Given our focus on predicting the performance of group members, measure IND are actually measuring several different, still unidenti- the remainder of this article focuses solely on collectivism. fied attributes of the respondents and that scales designed to measure COL are measuring a variety of unknown attributes as well. (p. 80)

886 JACKSON, COLQUITT, WESSON, AND ZAPATA-PHELAN Concerns about content validity are underscored by the correla- classification scheme for collectivism that consisted of 18 different tions among the various scale facets. Studies using scales from the “component ideas” for the construct. Almost all of the components Triandis family have sometimes yielded near-zero or even nega- could be viewed as more specific or nuanced versions of the tive correlations among facets purported to represent collectivism Triandis attributes described above. Exceptions included economic (Hui & Yee, 1994). Moreover, when scale scores were used as and religious philosophies that are more relevant at the national indicators of a higher order collectivism construct, those scores culture level, along with the importance of privacy. Oyserman et had either low factor loadings or poor internal consistency (Chen al. (2002) presented a summary of all of the concepts included et al., 1997; Hui & Yee, 1994). Studies using scales from the within collectivism measures in their meta-analysis. Most of the Wagner family have also yielded small or near-zero correlations concepts could again be viewed as more specific or nuanced among facets (Moorman & Blakely, 1995; Ramamoorthy & Car- versions of the Triandis attributes. Exceptions included the impor- roll, 1998; Wagner, 1995; Wagner & Moch, 1986). Such results tance of advice and a focus on hierarchy. It should be noted that raise questions about whether the various facets named in the Oyserman et al. argued that many of these exceptions, including Appendixes are actually assessing the same underlying construct. privacy and advice issues, should more appropriately be concep- tualized as consequences of collectivism, not defining qualities of Study 1 the construct itself. In response to the concerns raised about collectivism measure- We therefore limited the content of our measure to the five ment (Earley & Gibson, 1998; Kagitcibasi, 1994; Oyserman et al., facets summarized in Table 1: preference, reliance, concern, norm 2002; Triandis et al., 1995), in Study 1 we developed and tested a acceptance, and goal priority. Our terminology and discussion of new collectivism measure to be used in organizational research. each facet in Table 1 are distilled from the key sources cited in the The content of our measure was initially based in Triandis’s table in an attempt to emphasize the commonalities across multiple various discussions of the defining attributes of collectivism (Tri- descriptions of the collectivism construct. A close inspection of the andis, 1989, 1995, 1996; Triandis & Bhawuk, 1997). Triandis measures in the Appendixes illustrates that the scales in the Tri- argued that collectivists (a) use groups as the units of analysis in andis family sample most, if not all, of the facets in Table 1 their social space, preferring to draw meaning from group mem- (though often with low reliability or near-zero correlations among berships; (b) emphasize a sense of collective responsibility and facets). The scales in the Earley family sample the preference, common fate; (c) feel great concern about what happens to in- reliance, and norm acceptance facets (though in a purportedly group members; (d) emphasize group norms as a driver of behav- unidimensional fashion). The scales in the Wagner family sample ior rather than individual attitudes; and (e) prioritize the goals of most, if not all, of the Table 1 concepts (though again with the in-group rather than personal goals. Triandis further suggested near-zero correlations among subscales). that these defining attributes should be moderately correlated— that the attributes should converge to capture the common core of We designed our measure around a second-order factor struc- collectivism (Triandis, 1995; Triandis & Bhawuk, 1997). ture, with the five facets serving as latent indicators of a higher order collectivism construct. This structure is consistent with Tri- We compared these attributes with descriptions of collectivism andis’s view that the defining attributes of collectivism should be in other reviews of the literature. Ho and Chiu (1994) derived a moderately correlated (Triandis, 1995; Triandis & Bhawuk, 1997) Table 1 Key Facets of the Collectivism Construct Facet Discussion Key sources Preference Reliance Collectivists emphasize relationships with in-group members and prefer to exist Triandis (1995) Concern within the bounds of the in-group. They are affiliative by nature and believe Ho & Chiu (1994) that collective efforts are superior to individual ones. Oyserman et al. (2002) Triandis (1989) Collectivists believe that one person’s responsibility is the responsibility of the Ho & Chiu (1994) entire in-group. This sense of collective responsibility makes them comfortable Oyserman et al. (2002) relying on other members of the in-group. Triandis (1989, 1995, Collectivists are motivated not by self-interest but by a concern for the well-being 1996) of the in-group and its members. Triandis & Bhawuk (1997) Ho & Chiu (1994) Norm acceptance Collectivists focus on the norms and rules of the in-group and comply with those Oyserman et al. (2002) norms and rules in order to foster harmony within the collective. Triandis (1989, 1995, Goal priority Collectivists’ actions are guided by the consideration of the in-group’s interests. 1996) Thus in-group goals take priority over individual goals, even if this causes the Triandis & Bhawuk (1997) in-group member to make certain sacrifices. Ho & Chiu (1994) Oyserman et al. (2002) Triandis (1989, 1995, 1996) Triandis & Bhawuk (1997) Ho & Chiu (1994) Oyserman et al. (2002)

PSYCHOLOGICAL COLLECTIVISM 887 and is also consistent with the conceptualization of most multidi- Method mensional constructs in organizational behavior (Law, Wong, & Mobley, 1998). We felt that this strategy would allow us to create Item Generation a more reliable collectivism scale because the five facets should be more internally consistent than the dimensions used in other scales. The first step in constructing the measure was to decide what in-group We also felt that this strategy would maximize both “bandwidth” to reference in the items. The choice of in-group is critical, as the very and “fidelity” (Cronbach, 1990). Bandwidth refers to the amount definition of collectivism is in-group based (Triandis, 1995). Indeed, of information and is increased by assessing a wide variety of Deaux and Reid (2000) suggested that “collectivism should be assessed in construct elements. Fidelity refers to the reliability of the infor- terms of specific group memberships rather than some general feelings mation and is decreased by assessing too many loosely related toward groups in general. We suggest that people may be collective in their concepts. Of this bandwidth–fidelity tradeoff, Triandis et al. attitudes toward some groups but not others” (pp. 181–182). The instruc- (1995) wrote, tions leading into our measure therefore referenced work groups as the in-group, as opposed to the friend, family, neighbor, and kin groups used It should be noted here that researchers in this area ask for an in past research. We felt that this would make the measure more relevant instrument that will measure the common core, and yet not be too to applied research in organizational behavior and industrial– long. However, because allocentrism and idiocentrism are very broad organizational psychology. constructs, their measurement with a few items is very difficult. Most short scales have low fidelity (Cronbach, 1990), and it is not surpris- The second step in constructing the measure was generating items that ing that in most past studies the Cronbach alphas have been unsatis- reflect each of the facets in Table 1. We generated 5 items for each facet factory (lower than .70). Selection of a narrow aspect of the constructs with the goal of eventually retaining the strongest 3 (the final, shortened can result in high alphas, e.g. patriotism can be measured with a .95 form of the measure is shown in Table 2). We then asked 10 subject matter alpha (Schmitz, 1992), but then one needs a measure for each collec- experts to blindly classify the 25 items into the five facets in Table 1. Items tive, which is not convenient since most researchers want a few items were correctly classified 95% of the time, and only 2 of the items were to measure the broader construct. (p. 464) misclassified by more than one expert. We should note that 2 of the preference items were similar to those used in past research (Earley, 1993; Our goal was therefore to create a small set of items to measure Erez & Earley, 1987; Wagner, 1995; Wagner & Moch, 1986). We then each of the five facets. The second-order factor structure of the piloted the 25 items in a sample of 540 undergraduates taking management measure could then balance fidelity (within the five latent indica- courses at a large Southeastern university to gauge item readability and tors) with bandwidth (within the higher order factor). Because our form an initial picture of the factor structure. five facets should be significantly correlated, our measure should possess adequate reliability at the scale level, not just the facet Sample level. Moreover, the second-order structure could provide re- searchers with a great degree of flexibility. When psychological Study 1 participants were 241 full-time employees of a multinational collectivism is not the central focus of a study, the general factor consulting firm. All participants in the study were at the same level can be used as a single variable. However, when psychological (consultant) within the organization, and all were based in the United collectivism is the central focus, the facets can be assessed sepa- States. The consultants’ projects centered on the installation and customi- rately in order to gain a more fine-grained understanding of col- zation of various software systems (e.g., Peoplesoft systems, SAP sys- lectivism effects. Oyserman et al. (2002) suggested that the use of tems). Owing to the anonymous nature of the responses to the survey, exact such a facet-level approach could improve the theoretical clarity gender distribution cannot be determined. However, the initial sample found in the literature. population was 65% male, with an average age of 27.5. Listwise deletion of missing data resulted in a usable sample of 235 participants. We should also note that all five of the facets in Table 1 appear relevant to work contexts in which cooperation is important. K. G. Consultants represent an interesting setting for examining cooperative Smith, Carroll, and Ashford (1995) reviewed a number of theo- phenomena. Their tasks are organized around projects rather than jobs, retical perspectives that can be used to explain cooperation. One with some projects taking only a few days and others taking several perspective is attraction theories, which argue that cooperation is months. Although the task requirements vary from project to project, fostered when individuals have a natural affinity for one another. cooperation among the consultants is always a necessity. Unfortunately, The preference and concern facets seem relevant to cooperation one barrier to cooperation is that projects are often staffed with consultants from this standpoint, as both should create a sense of attraction to who lack experience or familiarity with one another. Thus, individuals the group. Another perspective is power and conflict theories, must supply the ingredients for cooperation themselves, rather than relying which argue that cooperation is fostered when goals and tasks are on past history and experience to supply them. interdependent. The goal priority and reliance facets seem relevant here, as both should facilitate the development of interdependence Procedure within the group. A third perspective discussed by K. G. Smith et al. is modeling theories, which link cooperation to the develop- Data were collected through the use of an Internet survey. A total of 643 ment of shared norms for prosocial behavior. The norm acceptance usable e-mail addresses were obtained through the participating company. facet seems most relevant to this perspective. Thus, each of the These e-mails contained an offer to participate in the study, which could be facets seems relevant to cooperation as it is conceptualized in the accessed by following a link in the e-mail. Owing to the nature of the literature. survey, respondents were guaranteed complete anonymity in both their responses and their decision to participate. In responding to the 25 collec- tivism items, respondents were asked to think about the work groups to which they currently belong and have belonged in the past. All items were assessed using a 5-point Likert scale with anchors of 1 (strongly disagree) and 5 (strongly agree). A total of 241 consultants chose to participate, resulting in a response rate of 37%.

888 JACKSON, COLQUITT, WESSON, AND ZAPATA-PHELAN Table 2 Psychological Collectivism Measure Items Measure item Facet 1. I preferred to work in those groups rather than working alone. Preference 2. Working in those groups was better than working alone. Preference 3. I wanted to work with those groups as opposed to working alone. Preference 4. I felt comfortable counting on group members to do their part. Reliance 5. I was not bothered by the need to rely on group members. Reliance 6. I felt comfortable trusting group members to handle their tasks. Reliance 7. The health of those groups was important to me. Concern 8. I cared about the well-being of those groups. Concern 9. I was concerned about the needs of those groups. Concern 10. I followed the norms of those groups. Norm acceptance 11. I followed the procedures used by those groups. Norm acceptance 12. I accepted the rules of those groups. Norm acceptance 13. I cared more about the goals of those groups than my own goals. Goal priority 14. I emphasized the goals of those groups more than my individual goals. Goal priority 15. Group goals were more important to me than my personal goals. Goal priority Note. The instructions participants read were as follows: “Think about the work groups to which you currently belong, and have belonged to in the past. The items below ask about your relationship with, and thoughts about, those particular groups. Respond to the following questions, as honestly as possible, using the response scales provided. (1 ϭ Strongly Disagree to 5 ϭ Strongly Agree).” Results and Discussion reliance, .90 for concern, .90 for norm acceptance, and .87 for goal priority. A confirmatory factor analysis of the 25 items yielded a rea- sonable fit for a second-order structure with the five facets as latent Study 2 indicators of the higher order psychological collectivism factor: ␹2(270, N ϭ 235) ϭ 532.26, p Ͻ .001; ␹2/df ϭ 1.97; incremental The purpose of Study 2 was to cross-validate the shorter 15-item fit index (IFI) ϭ .94; comparative fit index (CFI) ϭ .94; root- version of our collectivism measure and to assess its convergent mean-square error of approximation (RMSEA) ϭ .06. Good model validity. Convergent validity refers to the extent to which alterna- fit is typically inferred when ␹2/df falls below 3 and IFI or CFI rise tive measures of the same construct share variance (D. T. Camp- above .90 (Kline, 1998). RMSEA is interpreted as follows: Greater bell & Fiske, 1959; Schwab, 1980). We tested the convergent than .10 is poor fit, .08 to .10 is mediocre fit, .05 to .08 is validity of our collectivism measure by correlating its scores with reasonable fit, and less than .05 is good fit (Browne & Cudeck, those derived from other measures of the construct. We chose a 1993). representative version of the scales in each of the families sum- marized in the Appendixes. Specifically, we included the most We used the results of the confirmatory factor analysis to select recent scale from the Triandis family, used by Triandis and Gel- the best 15 items from the larger set of 25. Reducing the length to fand (1998) and Gelfand and Realo (1999), adapted from Singelis 15 items made our measure more similar in length to the most et al. (1995). We also included the most recent scale from the commonly used collectivism measures, which range from 8 (Ear- Earley family, in the form of Earley (1993). Finally, we included ley, 1993; Gelfand & Realo, 1999; Triandis & Gelfand, 1998) to Wagner and Moch’s (1986) scale, chosen over Wagner (1995) 11 items (Wagner & Moch, 1986). However, the 15-item length because the latter includes items from the other two families. still devoted 3 items to each facet, allowing facet level analyses to be performed when desired. The best 15 items were chosen on the Method basis of factor loadings, skewness and kurtosis, residual values, and item content. Sample The fit of the 15-item model was strong: ␹2(85, N ϭ 235) ϭ Study 2 participants were 140 individuals working in a variety of 117.51, p Ͻ .05; ␹2/df ϭ 1.38; IFI ϭ .98; CFI ϭ .98; RMSEA ϭ organizations and industries. Of the participants, 115 were registered .04. As shown in Figure 1, each of the collectivism items loaded alumni of a large, Southeastern university. The remaining 25 were students highly on its intended facet, with loadings ranging from .64 to .95, in an executive weekend master of business administration (MBA) course with an average of .82. Moreover, the five collectivism facets at the same university. Participants were 68% male and averaged 40 years loaded strongly on the general factor: .56 for preference, .49 for of age. Listwise deletion of missing data resulted in a usable sample of 139 reliance, .61 for concern, .65 for norm acceptance, and .46 for goal participants. priority. All 10 of the interfacet correlations were statistically significant. Thus, the general factor seemed to be driving the five Procedure facets to an approximately equal degree. The psychological col- lectivism measure also possessed strong reliability, whether used A list of 325 names was randomly drawn from a database of 7,000 as an overall scale (␣ ϭ .84) or as five more specific facets. registered alumni who were known to be employed full time. We mailed Facet-level reliabilities were as follows: .86 for preference, .81 for

PSYCHOLOGICAL COLLECTIVISM 889 Figure 1. Confirmatory factor analysis results for Study 1. the alumni a survey along with a letter explaining that the purpose of the Convergent Validity study was to assess attitudes about working in groups. Participants were told that they were eligible for the study if their job contained a significant Convergent validity can be judged by examining the correlation group component, meaning they needed to collaborate with others to between our collectivism measure and the Triandis, Earley, and perform day-to-day tasks and achieve collective work goals. Participants Wagner measures, shown in Table 3. Our measure was strongly earned a $5 cash payment for their participation. A total of 115 surveys correlated with the Triandis measure (r ϭ .52), the Earley measure were received, indicating a response rate of 35%. In addition, 43 executive (r ϭ .50), and the Wagner measure (r ϭ .50). These correlations MBA students were given a copy of the survey during their weekend visit provide strong support for convergent validity given the different to the campus, with the same cash incentive offered. Twenty-five surveys in-groups referenced by the scales, the varying coverage of the five were received, for a response rate of 58%. On average, the participants collectivism facets in Table 1, the different item styles used (e.g., worked in groups of eight people when performing their job duties and attitudes, past behaviors, norms, beliefs), and the lower reliabilities reported high levels of task interdependence with their fellow group of some of the other measures (.29 for Earley, .68 for Wagner). members (M ϭ 4.24 out of 5.00 using the measure developed by Pearce & The convergent validity correlations for our measure were also Gregersen, 1991). All items were assessed using a 5-point Likert scale with stronger than the correlations for the other three, as the Triandis, anchors of 1 (strongly disagree) and 5 (strongly agree). Earley, and Wagner measures had intercorrelations ranging from .23 to .35. Results and Discussion The facet-level correlations in Table 3 can be used to provide a Cross-Validation of Collectivism Measure more fine-grained picture of convergent validity. With respect to the Triandis scale, preference and concern should correlate most The fit of the 15-item model was strong: ␹2(85, N ϭ 139) ϭ strongly with the horizontal facet, which assesses collectivism 101.91, p Ͻ .10; ␹2/df ϭ 1.20; IFI ϭ .99; CFI ϭ .99; RMSEA ϭ from an equal status perspective. Norm acceptance, on the other .04. Each of the items loaded highly on its intended facet, with hand, should correlate most strongly with vertical collectivism, loadings ranging from .56 to .95, with an average of .83. Four of which focuses on hierarchical contexts in which norms and rules the five facets loaded strongly on the general collectivism factor: feed down to group members. With respect to the Wagner scale, .68 for preference, .69 for reliance, .47 for concern, and .42 for preference and reliance should correlate most strongly with the goal priority. The norm acceptance facet had a weaker loading of values facet, which captures the value placed on working in .21, though it was still statistically significant. Nine of the 10 groups. Norm acceptance should correlate most strongly with interfacet correlations were statistically significant, with the ex- the beliefs facet, which captures the belief that members should ception being reliance and norm acceptance. The measure pos- do what the group thinks is best. Goal priority should correlate sessed strong reliability, whether used as an overall scale (␣ ϭ .85) most strongly with the norms dimension, which captures the or as five, more specific facets (␣s ranging from .78 to .93). need to sacrifice individual goals for the good of the group. The

890 JACKSON, COLQUITT, WESSON, AND ZAPATA-PHELAN Table 3 Convergent Validity Correlations Variable ␣ M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1. Psychological collectivism .85 3.46 0.49 — 2. Preference facet .93 3.17 0.86 .73* — 3. Reliance facet .85 3.29 0.86 .68* .43* — 4. Concern facet .89 3.96 0.65 .63* .36* .31* — 5. Norm acceptance facet .78 3.87 0.58 .40* .17* .03 .17* — 6. Goal priority facet .89 3.05 0.88 .65* .29* .23* .22* .18* — 7. Triandis collectivism .75 4.07 0.46 .52* .30* .28* .55* .25* .30* — 8. Horizontal facet .72 4.02 0.54 .57* .38* .36* .65* .15 .28* .81* — 9. Vertical facet .70 4.12 0.58 .29* .12 .10 .26* .26* .22* .83* .34* — 10. Earley collectivism .29 3.40 0.37 .50* .39* .46* .31* .06 .28* .23* .30* .10 — 11. Wagner collectivism .68 3.70 0.37 .50* .45* .32* .21* .31* .29* .35* .31* .27* .33* — 12. Beliefs facet .81 3.78 0.73 .11 .01 .03 .03 .26* .09 .35* .04 .12 .11 .58* — 13. Values facet .74 2.95 0.71 .68* .80* .49* .30* .16 .31* .10 .39* .12 .45* .56* Ϫ.07 — 14. Norms facet .79 4.09 0.46 .15 .07 .09 .06 .14 .14 .30* .15 .26* .06 .70* .12 .13 — Note. N ϭ 139. * p Ͻ .05, one-tailed. correlations in Table 3 match that expected pattern, with the criterion needed to foster declarative and procedural knowledge, as exception of the relation between goal priority and norms. We collective members prefer to work in groups and are concerned should note, however, that Wagner’s norms facet yielded very about group welfare. Collective members may also have more past few significant correlations across the table. In fact, the three experience working in groups, as past research has linked those Wagner facets were not correlated with one another (rs ranging two variables (Eby & Dobbins, 1997). In addition, the collectivism from Ϫ.07 to .13). facets may supply motivation in task contexts, as the tendency to prioritize group goals should foster a more intense and long-lasting Study 3 exertion of effort. We are aware of only one study that has linked a member’s collectivism levels to his or her task performance in an Having described the development and validation of our new actual group setting (Shaw, Duffy, & Stark, 2000). Although this measure of psychological collectivism, in Study 3 we seek to make study did not find a significant zero-order link between the two a more substantive contribution. As noted at the outset, the rela- variables, an interaction effect was uncovered in which the rela- tionship between collectivism and group member performance tionship between collectivism and task performance became more remains largely untested. Past research has linked collectivism to positive at higher levels of perceived task interdependence. We reward generosity (Gomez et al., 2000), preference for group- therefore predicted the following: based appraisals (Ramamoorthy & Carroll, 1998; Waldman, 1997), self-efficacy for teamwork (Eby & Dobbins, 1997), and Hypothesis 1: Psychological collectivism will be positively cooperative choices in a prisoner’s dilemma game (Cox et al., related to supervisor ratings of group member task 1991). We explored collectivism’s linkage with four dimensions of performance. job performance: (a) task performance—the proficiency with which members perform the activities recognized as part of their Citizenship Behavior group role (Borman & Motowidlo, 1993); (b) citizenship behav- iors— discretionary behaviors that lie outside of an employee’s job Citizenship behaviors form the core of what is more commonly description and may not be formally rewarded by the group (C. A. called “teamwork” (LePine et al., 2000), and such behaviors are Smith, Organ, & Near, 1983); (c) counterproductive behaviors— typically explained using a social exchange theory lens (Blau, intentional behaviors on the part of employees that are contrary to 1964). Specifically, Organ (1990) argued that a social exchange the group’s legitimate interests (Sackett & DeVore, 2001); and (d) relationship is a prerequisite for citizenship behaviors, given that withdrawal behaviors—a more passive response to dissatisfaction such behaviors may not be formally rewarded by the organization. characterized by psychological and physical avoidance of the Triandis and Bhawuk (1997) suggested that collectivists are more group and its work (Hulin, 1991). likely to adopt a social exchange mindset as they are predisposed to “play relationships by ear,” adopt a long-term time perspective, Task Performance give more than they receive, and ground their willingness to reciprocate in a deeply held concern for their group. Consistent According to J. P. Campbell, McCloy, Oppler, and Sager’s with this reasoning, past research has supported the linkage be- (1992) theory of performance, the most proximal individual pre- tween collectivism and individual citizenship behaviors in two dictors of task performance are motivation and declarative and studies (Moorman & Blakely, 1995; Van Dyne et al., 2000). We procedural knowledge, which are themselves predicted by person- therefore predicted the following: ality, interest variables, and past experience, among others. In group settings, the collectivism facets should supply the interest

PSYCHOLOGICAL COLLECTIVISM 891 Hypothesis 2: Psychological collectivism will be positively measure because its 214 citations since 2000 make it the most related to supervisor ratings of group member citizenship commonly used measure and because it was shown to have ac- behavior. ceptable reliability and interfacet correlations in Study 2. We also included the Big Five personality factors in Study 3 in order to Counterproductive Behavior assess discriminant validity, which provides evidence that mea- sures that purportedly assess different constructs in fact do so Robinson and Bennett’s (1997) model of workplace deviance (D. T. Campbell & Fiske, 1959; Schwab, 1980). argues that counterproductive behaviors are triggered by provoca- tions such as inequity and poor work conditions. These provoca- Method tions create two forms of motivation for engaging in counterpro- ductive behavior: instrumental (to resolve the disparity in the Sample exchange relationship) and expressive (to vent one’s anger). These motivations can be neutralized by a number of constraints, includ- Study 3 participants were 186 full-time employees of a computer soft- ing internalization of norms, bonds to a social system, and the ware firm. The organization is a supplier and developer of software potential for formal and internal sanctions. Collectivism should systems based in the southwest United States. The company splits its foster two of these constraints given that collective members are workforce into groups on the basis of clients or regions served. The sample more likely to internalize the norms of the group and possess consisted of programmers, software support, and client support, with all strong bonds to the social system (Ho & Chiu, 1994; Oyserman et jobs being fairly technical in nature. In all cases, participants worked in al., 2002; Triandis, 1989, 1995). Although we are not aware of any groups that were highly interdependent, requiring cooperation to meet the studies linking collectivism to the counterproductive behaviors of specific needs of their clients. Participants were 71% male and averaged individual group members, we predicted the following: 31.54 years of age and 3.35 years with the organization. Listwise deletion of missing data resulted in a usable sample of 178 participants. Hypothesis 3: Psychological collectivism will be negatively related to supervisor ratings of group member counterproduc- Procedure tive behavior. All measures were assessed through the use of an Internet-based survey. Withdrawal Behavior A total of 262 usable e-mail addresses were obtained from the participating organization. These e-mails contained an offer to participate in the study, Some forms of withdrawal are found within taxonomies of which could be accessed by following an Internet link within the e-mail. A counterproductive behavior, including poor attendance, tardiness, total of 186 of the employees chose to participate, resulting in a response and the withholding of effort (Robinson & Bennett, 1997; Sackett rate of 71%. Participants’ supervisors were also asked to fill out a survey & DeVore, 2001). We separated the two constructs in this study, assessing job performance in the context of work group tasks. In all cases, however, as withdrawal behaviors are more passive responses the supervisor who filled out this survey was the same person who relative to actions like sabotage, intentional breaking of group performed the employee’s formal performance appraisal (though it was rules, and verbal abuse of group members. Most models of with- stressed that these rankings did not feed into that system in any way). Of drawal view the psychological and physical disengagement from the 186 participants, complete supervisor data were obtained for 128, a work as a response to dissatisfaction (Hulin, 1991; Johns, 2001). response rate of 69%. Although past research has linked collectivism to member satis- faction in student project groups (Shaw et al., 2000), researchers Measures have argued that satisfaction alone is incapable of explaining a large portion of variance in withdrawal (Johns, 2001). Such be- All measures were assessed using a 5-point Likert scale with anchors of haviors also depend on contextual variables like job embedded- 1 (strongly disagree) and 5 (strongly agree). ness, which is characterized by strong ties to other members in a group’s social network (Mitchell, Holtom, Lee, Sablynski, & Erez, Task performance. Supervisors rated their employees using the 7-item 2001). Collectivists form stronger bonds with in-group members scale developed by Williams and Anderson (1991), with the wording and are more likely to make sacrifices for the good of the group changed to reflect work group duties and responsibilities. Supervisors were (Ho & Chiu, 1994; Oyserman et al., 2002; Triandis, 1995), making asked, “How much do you agree with the following statements: This withdrawal less likely. Although we are not aware of any studies employee _____,” with sample items including “adequately completes linking collectivism to the withdrawal of individual group mem- assigned work group duties,” “fulfills responsibilities specified by his/her bers, we hypothesized the following: work group,” “meets formal requirements of the work group,” and “fails to perform essential duties for the work group” (reverse scored). Hypothesis 4: Psychological collectivism will be negatively related to supervisor ratings of group member withdrawal Citizenship behavior. Supervisors rated their employees using the 16- behavior. item measure created by Lee and Allen (2002), with the wording changed to reflect work-group-directed citizenship. Sample items included “will- Linking collectivism to group member performance makes a ingly gives time to help other group members who have work-related substantive contribution to the literature while also testing the problems,” “shows genuine concern and courtesy toward other group predictive validity of our measure. We also compared our mea- members, even under the most trying business or personal situations,” sure’s predictive validity with that of the most recent scale from “attends functions that are not required but that help the work group’s the Triandis family: Triandis and Gelfand (1998). We chose this image,” “defends the work group when other employees criticize it,” and “offers ideas to improve the functioning of the work group.” Counterproductive behavior. Supervisors rated their employees using the nine-item scale developed by Robinson and O’Leary-Kelly (1998), with the wording changed to assess work-group-directed counterproductive behaviors. Supervisors were asked, “Within the last year, how often has this person _____ ?” Sample items included “said or did something to

892 JACKSON, COLQUITT, WESSON, AND ZAPATA-PHELAN purposely hurt a work group member,” “damaged property being used by Descriptive Statistics and Correlations the work group,” “deliberately bent or broke a group rule(s),” and “said rude things about work group members.” The means, standard deviations, reliabilities, and correlations for all Study 3 variables are shown in Table 4. Three of the four Withdrawal behavior. Supervisors also rated their employees with 10 outcome measures had acceptable reliabilities, with the alpha for items developed by Lehman and Simpson (1992). The item wording was withdrawal behavior falling below the .70 standard. The four out- again changed to reflect the work group context and was also altered to comes were also weakly to moderately correlated, with correlations allow for supervisor-reported rather than self-reported data. Supervisors ranging from Ϫ.08 to .30. Our measure of psychological collectivism were asked, “Within the last year, how often has this person _____ ?” was significantly correlated with Agreeableness, as suggested by Sample items included “been absent from work group activities,” “left Triandis (1997), but was uncorrelated with the other Big Five work early without permission,” “seemed to be daydreaming rather than factors. These results provide evidence of discriminant validity for working,” “spent time on personal matters rather than group duties,” and the psychological collectivism measure, relative to the Big Five. “talked about leaving current job or work group.” Tests of Hypotheses Psychological collectivism. The measure shown in Table 2 was again used. Hypotheses 1– 4 predicted significant relationships between psychological collectivism and the task, citizenship, counterpro- Triandis collectivism. As in Study 2, the scale used by Triandis and ductive, and withdrawal dimensions of group member job perfor- Gelfand (1998) and Gelfand and Realo (1999), adapted from Singelis et al. mance. As shown in the top row of Table 5, all four hypotheses (1995), was used. received support. Psychological collectivism explained 10% of the variance in task performance, 4% of the variance in citizenship Big Five. We measured these personality dimensions using the Big behavior, 10% of the variance in counterproductive behavior, and Five Inventory (John, Donahue, & Kentle, 1991). Sample items included “I 5% of the variance in withdrawal behavior. persevere until the task is finished” (Conscientiousness, 9 items), “I am helpful and unselfish with others” (Agreeableness, 9 items), “I am some- Table 5 decomposes the reasons for the psychological collec- times depressed or blue” (Neuroticism, 8 items), “I have an active imag- tivism effects by providing information on the relative importance ination” (Openness to Experience, 10 items), and “I am outgoing and of the five facets. In addition to the standardized regression sociable” (Extraversion, 8 items). weights, the table provides two alternative methods for judging the importance of correlated predictors: dominance analysis (Azen & Results and Discussion Budescu, 2003) and the epsilon statistic (Johnson, 2000). The general dominance statistic (denoted D in the table and computed Cross-Validation of Factor Structure using Dominance Analysis 4.30 by James LeBreton) is computed as the squared semipartial correlation averaged across all possible We conducted another second-order confirmatory factor analy- subset regression models. The values in the table are then rescaled sis to provide further cross-validation of the measure in Table 2. by dividing them by the total variance explained in the outcome to The fit of the model was again strong: ␹2(85, N ϭ 124) ϭ 148.06, provide an index of the average usefulness of each predictor. The p Ͻ .001; ␹2/df ϭ 1.74; IFI ϭ .95; CFI ϭ .95; RMSEA ϭ .06. epsilon statistic (denoted ␧ in the table and computed using SPSS Each of the collectivism items loaded highly on its intended factor, syntax by Jeff Johnson) is calculated by deriving uncorrelated with loadings ranging from .59 to .97. Moreover, the five collec- principal components from a set of predictors and combining two tivism facets loaded significantly on the general factor: .53 for sets of standardized regression weights: the outcome regressed on preference, .50 for reliance, .61 for concern, .52 for norm accep- the principal components and the original predictors regressed on tance, and .34 for goal priority. Nine of the 10 interfacet correla- the principal components. The resulting statistic provides an index tions were statistically significant, with the exception being pref- erence and goal priority. The measure again possessed strong reliability, whether used as an overall scale (␣ ϭ .82) or as five, more specific facets. Facet-level reliabilities were as follows: .89 for preference, .79 for reliance, .84 for concern, .82 for norm acceptance, and .83 for goal priority. Table 4 Correlations Between Study 3 Variables Variable ␣ M SD 1 2 3 45 6 78 9 10 11 1. Psychological collectivism .82 3.67 0.42 — 2. Task performance .78 4.37 0.54 .32* — 3. Citizenship behavior .94 4.41 0.63 .20* .20* — 4. Counterproductive behavior .86 1.44 0.48 Ϫ.31* Ϫ.18* Ϫ.30* — 5. Withdrawal behavior .50 1.65 0.32 Ϫ.23* Ϫ.26* Ϫ.08 .20* — 6. Triandis collectivism .76 3.46 0.58 .33* .16* .09 Ϫ.11 Ϫ.03 — 7. Conscientiousness .70 2.56 0.50 Ϫ.02 .10 .17* .05 .10 Ϫ.05 — 8. Agreeableness .77 2.68 0.58 .23* .15* .12 .05 Ϫ.07 .07 Ϫ.01 — 9. Neuroticism .76 2.72 0.61 Ϫ.01 Ϫ.06 Ϫ.06 Ϫ.02 .06 .00 Ϫ.03 .05 — 10. Openness to experience .81 2.59 0.62 Ϫ.07 Ϫ.05 Ϫ.01 Ϫ.04 Ϫ.14 .00 Ϫ.12 Ϫ.05 Ϫ.09 — 11. Extraversion .79 3.56 0.63 .11 Ϫ.01 Ϫ.04 Ϫ.04 .06 .22* .07 .11 Ϫ.15* .03 — Note. N ϭ 124. * p Ͻ .05, one-tailed.

PSYCHOLOGICAL COLLECTIVISM 893 Table 5 Relationships Between Psychological Collectivism and Group Member Performance Member task Member citizenship Member counterproductive Member withdrawal performance behavior behavior behavior Variable R2 ␤ D ␧ R2 ␤ D ␧ R2 ␤ D ␧ R2 ␤ D ␧ Psychological collectivism .10* .32* General factor .10* Ϫ.31* .05* Ϫ.23* .04* .20* Specific facet Preference .15* .18* 18.2 19.2 .09* Ϫ.09 3.9 2.9 .11* .01 2.6 2.4 .09* Ϫ.05 4.0 4.0 Reliance Concern .06 9.1 8.1 .23* 52.5 53.3 Ϫ.07 9.7 8.8 .03 1.2 1.5 Norm acceptance Goal priority Ϫ.09 2.1 1.4 .21* 34.9 35.8 Ϫ.13 24.1 25.3 Ϫ.14 11.1 11.2 .12 14.3 13.8 Ϫ.04 1.8 1.6 Ϫ.18* 40.0 40.8 Ϫ.16* 8.2 8.2 .29* 56.3 57.5 .05 7.0 6.5 Ϫ.14 23.6 22.7 Ϫ.24* 75.5 75.2 Note. N ϭ 124. * p Ͻ .05. of the proportionate contribution each predictor makes to total analysis, in which the incremental variance explained by each variance explained. A recent review suggested that dominance measure was compared using two separate regressions. As shown analysis and the epsilon statistic are the two preferred methods for in Table 6, scores on our collectivism measure explained incre- judging the relative importance of correlated predictors (LeBreton, mental variance in all four outcomes after controlling for scores on Ployhart, & Ladd, 2004). the Triandis measure, with ⌬R2 values ranging from .03 to .08. In contrast, scores on the Triandis measure failed to explain incre- The results in Table 5 reveal that the five facets explain about mental variance. 35% more variance in the outcomes than the higher order factor. Each of the five facets significantly contributes to at least one of Although we included the Big Five in Study 3 primarily to the outcomes. Preference is relatively important in predicting task gauge the discriminant validity of our collectivism measure, we performance, whereas reliance and concern contribute to the pre- felt it would be worthwhile to compare the predictive validity of diction of citizenship. Norm acceptance is important to the pre- psychological collectivism with that of the Big Five. Table 7 diction of counterproductive behavior and, to a lesser extent, summarizes the effects of collectivism and the Big Five in com- withdrawal. Finally, goal priority has a high degree of importance bination. The results reveal that collectivism has the highest rela- for two outcomes: task performance and withdrawal behavior. tive importance of the predictors, regardless of whether standard- These results illustrate that all five collectivism facets contributed ized regression coefficients, dominance analysis, or the epsilon to the predictive validity of the psychological collectivism factor. statistic are used to assess importance levels. We also compared the predictive validity of our collectivism General Discussion measure with that of the Triandis scale. Triandis collectivism was significantly correlated with member task performance (r ϭ .16) The three studies described in this article point to a number of but was not correlated with the other three outcomes (see Table 4). strengths for our psychological collectivism measure. The measure We compared the predictive validity levels using a usefulness Table 6 Usefulness Analysis With Triandis Collectivism Member task Member citizenship Member counterproductive Member withdrawal performance behavior behavior behavior Regression order R2 ⌬R2 ␤ R2 ⌬R2 ␤ R2 ⌬R2 ␤ R2 ⌬R2 ␤ 1. Triandis collectivism .02 .02 .16 Order 1 .01 .01 Ϫ.11 .00 .00 Ϫ.03 .10* .09* Ϫ.30* .06* .06* Ϫ.25* 2. Psychological collectivism .10* .08* .30* .01 .01 .09 .04† .03* .19* Order 2 1. Psychological collectivism .10* .10* .32* .04* .04* .20* .10* .10* Ϫ.31* .06* .06* Ϫ.23* 2. Triandis collectivism .10* .00 .05 .04† .00 .02 .10* .00 Ϫ.01 .06* .00 .05 Note. N ϭ 124. † p Ͻ .10. * p Ͻ .05.

894 JACKSON, COLQUITT, WESSON, AND ZAPATA-PHELAN Table 7 Psychological Collectivism and the Big Five in Combination Member task Member citizenship Member counterproductive Member withdrawal performance behavior behavior behavior Predictor R2 ␤ D ␧ R2 ␤ D ␧ R2 ␤ D ␧ R2 ␤ D ␧ Conscientiousness .13* .10 7.7 5.4 .09† .18* 38.2 34.1 .12* .04 38.2 2.3 .10† .07 8.3 9.8 Agreeableness .09 11.1 11.2 .09 12.3 11.6 .13 12.3 6.8 Ϫ.04 3.5 3.6 Neuroticism Ϫ.08 3.8 4.1 Ϫ.07 4.9 4.9 Ϫ.04 4.9 0.5 .07 4.8 5.7 Openness to experience Ϫ.02 1.2 1.7 .02 0.1 0.1 Ϫ.05 0.1 1.0 Ϫ.14 21.7 17.9 Extraversion Ϫ.08 7.8 1.4 Ϫ.09 9.6 5.0 Ϫ.02 9.6 1.0 .10 16.8 6.5 Psychological collectivism .31* 76.2 76.2 .19* 44.6 44.4 Ϫ.34* 44.6 88.4 Ϫ.24* 61.7 56.6 Note. N ϭ 124. † p Ͻ .10. * p Ͻ .05. is focused solely on work groups as the in-group, making it more explain how and why collectivism predicts each of the outcomes. relevant to applied research in the organizational sciences. The Our results showed that each of the facets served as an “active measure was carefully constructed around the most frequently ingredient” for at least one performance dimension, with prefer- discussed facets of the collectivism construct, as described in key ence explaining task performance effects, reliance and concern reviews of the literature (Ho & Chiu, 1994; Oyserman et al., 2002; facets explaining citizenship effects, norm acceptance explaining Triandis, 1989, 1995, 1996; Triandis & Bhawuk, 1997). The counterproductive and withdrawal effects, and goal priority ex- second-order structure of the collectivism measure allowed us to plaining relationships with task performance and withdrawal. balance bandwidth and fidelity by capturing the broad spectrum of collectivism facets while measuring each one in an internally The contributions above should be interpreted in light of this consistent fashion. This is in contrast to previous collectivism article’s limitations. Our collectivism scale used self-report data scales, which tended to sample from widely different domains that from a common source. This limitation creates the possibility that did not always seem to be tapping the same construct. the higher order factor observed in our confirmatory factor anal- ysis results was (at least partially) an artifact of common method Taken together, three samples totaling 552 group members variance. Parker (1999) illustrated how common method variance supported the construct validity of our new measure on a number can inflate the influence of a higher order factor in his examination of fronts. The measure was shown to be reliable at both the overall of the hierarchical model of psychological climate. A more strin- scale level and the facet level, in contrast to other scales, whose gent test of whether a psychological collectivism factor underlies reliability sometimes falls below the .70 standard. In addition, the the five facets in our measure would require either multiple 15 items seemed to be valid indicators of the five facets, and the sources (e.g., peers, significant others) or multiple methods (e.g., five facets seemed to be valid indicators of the higher order factor interview, projective test). These steps would allow the model in (as evidenced by the significant second-order loadings and the Figure 1 to be tested in a multitrait–multimethod fashion. In significant interfacet correlations). Convergent validity was sup- addition, we should note that the adequate reliability found for our ported in the form of high correlations with representatives of all measure may be a function of the semantic similarities of the items three major scale families (Earley, 1993; Triandis & Gelfand, within each facet. However, it should be pointed out that the 1998; Wagner & Moch, 1986). Discriminant validity was sup- reliability may also be a function of the significant correlations ported by separating psychological collectivism from a measure of found between the five facets, in contrast to other scales that have the Big Five. used facets that are generally uncorrelated in empirical tests. Moreover, we should note that the withdrawal behavior measure The creation of a construct valid measure of collectivism al- used in Study 3 exhibited poor reliability. Lehman and Simpson’s lowed us to test a key substantive question: Do collective group (1992) withdrawal items were intended to be self-report in nature, members tend to be better performing group members? Study 3 and it may be that supervisors are unable to reliably evaluate subtle provides some preliminary support for that question, as scores on behaviors like moonlighting, pretending to be busy, or discussing psychological collectivism were associated with supervisor ratings nonwork topics. of four separate dimensions of group member performance. Col- lective members performed their group tasks better, contributed A final limitation of our study concerns the assessment of our more discretionary citizenship, and were less likely to engage in measure’s predictive validity. Specifically, we chose to pit our counterproductive or withdrawal behaviors. These results add to a measure against the Triandis and Gelfand (1998) measure, which small body of research that supports the performance benefits of is meant to be used in more diverse settings and is more general in collectivism with respect to individual group members (Moorman nature than our measure. Researchers generally agree that match- & Blakely, 1995; Shaw et al., 2000; Van Dyne et al., 2000). ing the specificity or breadth of predictors and criteria enhances validity (Hogan & Roberts, 1996; Schneider, Hough, & Dunnette, Our facet-level analyses shed further light on the performance 1996). The breadth of the Triandis and Gelfand (1998) measure implications of collectivism. Oyserman et al. (2002) suggested that may have hindered its ability to predict the specific dimensions of facet-level analyses of collectivism could bring more theoretical group member job performance assessed in our study. As such, the clarity to the literature by isolating the “active ingredients” in a deck may have been stacked in favor of our more narrowly focused given relationship (p. 41). In other words, the facets can help

PSYCHOLOGICAL COLLECTIVISM 895 measure, which was developed with work contexts in mind. It may get psychological collectivism when necessary. Whereas typical be that scales from the Triandis family will be more significant development programs may be geared toward teaching task strat- predictors of nonwork outcomes than our scale, because the Tri- egies that mimic high levels of conscientiousness or agreeableness, andis items are more rooted in family, friend, and personal con- other programs could be used to teach collective strategies. Such texts. Thus, although the work-focused nature of our scale may programs could also be incorporated into the team-building pro- prove to be an advantage in predicting job attitudes and behaviors, grams routinely used to indoctrinate new group members. other scales may be more predictive of outcomes rooted in other life domains. References Despite these limitations, our results suggest some avenues for Akers, R. L. (1973). Deviant behavior: A social learning approach. Bel- future research. First and foremost, research is needed to further mont, CA: Wadsworth. refine the measure introduced in this article, given that scale development is always an iterative process. Future studies should Azen, R., & Budescu, D. V. (2003). The dominance analysis approach for assess convergent validity with other collectivism scales, discrimi- comparing predictors in multiple regression. Psychological Methods, 8, nant validity with other individual differences, and predictive 129 –148. validity with other outcomes. With respect to the latter, we would particularly encourage researchers to begin connecting scores on Bartol, K. M., & Durham, C. C. (2000). Incentives: Theory and practice. In our scale to some of the other commonly examined collectivism C. L. Cooper & E. A. Locke (Eds.), Industrial and organizational outcomes, such as conflict resolution styles, communication styles, psychology: Linking theory with practice (pp. 1–33). Malden, MA: and reward preferences (Earley & Gibson, 1998). Future research Blackwell. should also continue to test the boundary conditions for collectiv- ism effects. Shaw et al. (2000) showed that collectivism effects Blau, P. M. (1964). Exchange and power in social life. New York: Wiley. became less positive as task interdependence decreased. We sus- Borman, W. C., & Motowidlo, S. J. (1993). Expanding the criterion pect that the predictive validity of our scale would decline if the context changed from the highly interdependent groups in our domain to include elements of contextual performance. In N. Schmitt, samples to the more casual work groups used in some organiza- W. C. Borman, & Associates (Eds.), Personnel selection in organiza- tions. A number of other variables could help capture such con- tions (pp. 71–98). San Francisco: Jossey-Bass. textual boundaries, including group life span and member Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model proximity. fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136 –162). Thousand Oaks, CA: Sage. Finally, given that the five facets were differentially related to Campbell, D. T., & Fiske, D. M. (1959). Convergent and discriminant the performance outcomes, an interesting avenue for future re- validation by the multitrait–multimethod matrix. Psychological Bulletin, search would be to explore the theoretical underpinnings of these 56, 81–105. facets. For example, Akers’s (1973) social learning approach to Campbell, J. P., McCloy, R. A., Oppler, S. H., & Sager, C. E. (1992). A deviant behaviors, which identifies key mechanisms that should theory of performance. In N. Schmitt & W. C. Borman (Eds.), New increase or decrease the likelihood of deviance, is one theoretical developments in selection and placement (pp. 35–70). San Francisco: framework that would be particularly helpful in explaining the Jossey-Bass. relationships found between the norm acceptance facet and coun- Chen, C. C., Chen, X., & Meindl, J. R. (1998). How can cooperation be terproductive and withdrawal behaviors. Possible intervening fostered? The cultural effects of individualism– collectivism. Academy mechanisms for these relationships include collectivists’ imitation of Management Review, 23, 285–304. of group members’ behavior as a means to comply with group Chen, C. C., Meindl, J. R., & Hunt, R. G. (1997). Testing the effects of norms and the value that collectivists place on complying with vertical and horizontal collectivism: A study of reward allocation pref- these norms. Social exchange theory (Blau, 1964), on the other erences in China. Journal of Cross-Cultural Psychology, 28, 44 –70. hand, would be useful in understanding the relationships between Colquitt, J. A. (2004). Does the justice of the one interact with the justice the concern and reliance facets and citizenship. Specifically, the of the many? Reactions to procedural justice in teams. Journal of nature of the exchange relationship that collectivists have with Applied Psychology, 89, 633– 646. their group could serve as an intervening mechanism of the facet- Colquitt, J. A., Noe, R. A., & Jackson, C. L. (2002). Justice in teams: level relationships. From a social exchange theory perspective, Antecedents and consequences of procedural justice climate. Personnel collectivists’ concern for and reliance on group members should Psychology, 55, 83–110. foster a social exchange relationship, which in turn has been Cox, T. H., Lobel, S. A., & McLeod, P. L. (1991). Effects of ethnic group argued to encourage more prosocial discretionary behaviors. cultural differences on cooperative and competitive behavior on a group task. Academy of Management Journal, 34, 827– 847. Our results also offer important practical implications. Given the Cronbach, L. J. (1990). Essentials of psychological testing. New York: linkage between psychological collectivism and group member HarperCollins. performance, organizations should strive to maximize collectivism Deaux, K., & Reid, A. (2000). Contemplating collectivism. In S. Stryker, using selection and placement or employee development systems. T. J. Owens, & R. W. White (Eds.), Self, identity, and social movements In jobs where cooperation is vital, psychological collectivism (pp. 172–190). Minneapolis: University of Minnesota Press. could be assessed during employment interviews or personality DeMatteo, J. S., Eby, L. T., & Sundstrom, E. (1998). Team-based rewards: tests. Such personality scores, which are routinely stored in the Current empirical evidence and directions for future research. In B. M. employee profiles of many human resource software systems, Staw & R. Sutton (Eds.), Research in organizational behavior (Vol. 20, could also be used for placement decisions when staffing work pp. 141–183). Greenwich, CT: JAI Press. groups. Alternatively, employee development programs could tar- Dickson, P. H., & Weaver, K. M. (1997). Environmental determinants and individual-level moderators of alliance use. Academy of Management Journal, 40, 404 – 425. Earley, P. C. (1993). East meets West meets Mideast: Further explorations of collectivistic and individualistic work groups. Academy of Manage- ment Journal, 36, 319 –348. Earley, P. C. (1994). Self or group? Cultural effects of training on self-

896 JACKSON, COLQUITT, WESSON, AND ZAPATA-PHELAN efficacy and performance. Administrative Science Quarterly, 39, 89 – Johnson, J. W. (2000). A heuristic method for estimating the relative 117. weight of predictor variables in multiple regression. Multivariate Be- Earley, P. C., & Gibson, C. B. (1998). Taking stock in our progress on havioral Research, 35, 1–19. individualism– collectivism: 100 years of solidarity and community. Journal of Management, 24, 265–304. Kagitcibasi, C. (1994). A critical appraisal of individualism and collectiv- Eby, L. T., & Dobbins, G. H. (1997). Collectivistic orientation in teams: ism: Toward a new formulation. In U. Kim, H. C. Triandis, C. Kagit- An individual and group-level analysis. Journal of Organizational Be- cibasi, S. C. Choi, & G. Yoon (Eds.), Individualism and collectivism: havior, 18, 275–295. Theory, method, and applications (pp. 52– 65). Thousand Oaks, CA: Erez, M., & Earley, P. C. (1987). Comparative analysis of goal-setting Sage. strategies across cultures. Journal of Applied Psychology, 72, 658 – 665. Fiske, A. P. (2002). Using individualism and collectivism to compare Kirkman, B. L., & Shapiro, D. L. (2001). The impact of cultural values on cultures—A critique of the validity and measurement of the constructs: job satisfaction and organizational commitment in self-managing work Comment on Oyserman et al. (2002). Psychological Bulletin, 128, teams: The mediating role of employee resistance. Academy of Manage- 78 – 88. ment Journal, 44, 557–569. Gelfand, M. J., & Realo, A. (1999). Individualism– collectivism and ac- countability in intergroup negotiations. Journal of Applied Psychology, Kline, R. B. (1998). Principles and practice of structural equation mod- 84, 721–736. eling. New York: Guilford Press. Gibson, C. B. (1999). Do they do what they believe they can? Group efficacy and group effectiveness across tasks and cultures. Academy of Kozlowski, S. W. J., Gully, S. M., Nason, E. R., & Smith, E. M. (1999). Management Journal, 42, 138 –152. Developing adaptive teams: A theory of compilation and performance Gomez, C., Kirkman, B. L., & Shapiro, D. L. (2000). The impact of across time and levels. In D. R. Ilgen & E. D. Pulakos (Eds.), The collectivism and in-group/out-group membership on the evaluation gen- changing nature of performance: Implications for staffing, personnel erosity of team members. Academy of Management Journal, 43, 1097– actions, and development (pp. 240 –292). San Francisco: Jossey-Bass. 1106. Hedge, J. W., & Borman, W. C. (1995). Changing conceptions and prac- Law, K. S., Wong, C., & Mobley, W. H. (1998). Toward a taxonomy of tices in performance appraisal. In A. Howard (Ed.), The changing nature multidimensional constructs. Academy of Management Review, 23, 741– of work (pp. 451– 482). San Francisco: Jossey-Bass. 755. Ho, D. Y., & Chiu, C. (1994). Component ideas of individualism, collec- tivism, and social organization. In U. Kim, H. C. Triandis, C. Kagitci- LeBreton, J. M., Ployhart, R. E., & Ladd, R. T. (2004). A Monte Carlo basi, S. C. Choi, & G. Yoon (Eds.), Individualism and collectivism: comparison of relative importance methodologies. Organizational Re- Theory, method, and applications (pp. 200 –210). Thousand Oaks, CA: search Methods, 7, 258 –282. Sage. Hofstede, G. (1980). Culture’s consequences: International differences in Lee, K., & Allen, N. J. (2002). Organizational citizenship behavior and work-related values. Beverly Hills, CA: Sage. workplace deviance: The role of affect and cognitions. Journal of Hogan, J., & Roberts, B. W. (1996). Issues and non-issues in the fidelity– Applied Psychology, 87, 131–142. bandwidth trade-off. Journal of Organizational Behavior, 17, 627– 637. House, R., Javidan, M., Hanges, P., & Dorfman, P. (2002). Understanding Lehman, W. E. K., & Simpson, D. D. (1992). Employee substance abuse cultures and implicit leadership theories across the globe: An introduc- and on-the-job behaviors. Journal of Applied Psychology, 77, 309 –321. tion to project GLOBE. Journal of World Business, 37, 3–10. Hui, C. H. (1988). Measurement of individualism– collectivism. Journal of LePine, J. A., Hanson, M. A., Borman, W. C., & Motowidlo, S. J. (2000). Research in Personality, 22, 17–36. Contextual performance and teamwork: Implications for staffing. In Hui, C. H., & Triandis, H. C. (1986). Individualism– collectivism: A study G. R. Ferris (Ed.), Research in personnel and human resources man- of cross-cultural researchers. Journal of Cross-Cultural Psychology, 17, agement (Vol. 19, pp. 53–90). New York: Elsevier Science. 225–248. Hui, C. H., Triandis, H. C., & Yee, C. (1991). Cultural differences in Mitchell, T. R., Holtom, B. C., Lee, T. W., Sablynski, C. J., & Erez, M. reward allocation: Is collectivism the explanation? British Journal of (2001). Why people stay: Using job embeddedness to predict voluntary Social Psychology, 30, 145–157. turnover. Academy of Management Journal, 44, 1102–1121. Hui, C. H., & Yee, C. (1994). The shortened Individualism–Collectivism Scale: Its relationship to demographic and work-related variables. Jour- Mohrman, S. A., & Cohen, S. G. (1995). When people get out of the box: nal of Research in Personality, 28, 409 – 424. New relationships, new systems. In A. Howard (Ed.), The changing Hulin, C. L. (1991). Adaptation, persistence, and commitment in organi- nature of work (pp. 365– 410). San Francisco: Jossey-Bass. zations. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of industrial and organizational psychology (Vol. 2, pp. 445–506). Palo Moorman, R. H., & Blakely, G. L. (1995). Individualism– collectivism as Alto, CA: Consulting Psychologists Press. an individual difference predictor of organizational citizenship behavior. Ilgen, D. R., & Pulakos, E. D. (1999). Introduction: Employee performance Journal of Organizational Behavior, 16, 127–142. in today’s organizations. In D. R. Ilgen & E. D. Pulakos (Eds.), The changing nature of performance: Implications for staffing, motivation, Motowidlo, S. J., & Schmit, M. J. (1999). Performance assessment in and development (pp. 1–18). San Francisco: Jossey-Bass. unique jobs. In D. R. Ilgen & E. D. Pulakos (Eds.), The changing nature John, O. P., Donahue, E. M., & Kentle, R. L. (1991). The Big Five of performance: Implications for staffing, motivation, and development Inventory. Berkeley: University of California. (pp. 154 –191). San Francisco: Jossey-Bass. Johns, G. (2001). The psychology of lateness, absenteeism, and turnover. In N. Anderson, D. S. Ones, H. K. Sinangil, & C. Viswesvaran (Eds.), Murphy, K. R. (1999). The challenge of staffing a postindustrial work- Handbook of industrial, work, and organizational psychology (Vol. 2, place. In D. R. Ilgen & E. D. Pulakos (Eds.), The changing nature of pp. 232–252). Thousand Oaks, CA: Sage. performance: Implications for staffing, motivation, and development (pp. 295–324). San Francisco: Jossey-Bass. Oishi, S., Schimmack, U., Diener, E., & Suh, E. M. (1998). The measure- ment of values and individualism– collectivism. Personality and Social Psychology Bulletin, 24, 1177–1189. Organ, D. W. (1990). The motivational basis of organizational citizenship behavior. In L. L. Cummings & B. M. Staw (Eds.), Research in orga- nizational behavior (Vol. 12, pp. 43–72). Greenwich, CT: JAI Press. Oyserman, D., Coon, H. M., & Kemmelmeier, M. (2002). Rethinking individualism and collectivism: Evaluation of theoretical assumptions and meta-analyses. Psychological Bulletin, 128, 3–72. Parker, C. P. (1999). A test of alternative hierarchical models of psycho- logical climate: PCg, satisfaction, or common method variance? Orga- nizational Research Methods, 2, 257–274. Pearce, J. L., & Gregersen, H. B. (1991). Task interdependence and

PSYCHOLOGICAL COLLECTIVISM 897 extrarole behavior: A test of the mediating effects of felt responsibility. psychology. In J. P. Forgas & J. M. Innes (Eds.), Recent advances in Journal of Applied Psychology, 76, 838 – 844. social psychology: An international perspective (pp. 491– 499). North Perrewe, P. L., & Spector, P. E. (2002). Personality research in the Holland, the Netherlands: Elsevier Science. organizational sciences. In G. R. Ferris (Ed.), Research in personnel and Triandis, H. C. (1995). Individualism and collectivism. Boulder, CO: human resource management (Vol. 21, pp. 1– 63). New York: Elsevier Westview Press. Science. Triandis, H. C. (1996). The psychological measurement of cultural syn- Pillai, R., & Meindl, J. R. (1998). Context and charisma: A “meso” level dromes. American Psychologist, 51, 407– 415. examination of the relationship of organic structure, collectivism, and Triandis, H. C. (1997). Cross-cultural perspectives on personality. In R. crisis to charismatic leadership. Journal of Management, 24, 643– 671. Hogan, J. Johnson, & S. Briggs (Eds.), Handbook of personality psy- Probst, T., Carnevale, P. J., & Triandis, H. C. (1999). Cultural values in chology (pp. 439 – 464). San Diego, CA: Academic Press. intergroup and single-group social dilemmas. Organizational Behavior Triandis, H. C., & Bhawuk, D. P. S. (1997). Culture theory and the and Human Decision Processes, 77, 171–191. meaning of relatedness. In P. C. Earley & M. Erez (Eds.), New perspec- Ramamoorthy, N., & Carroll, S. J. (1998). Individualism/collectivism tives on international industrial/organizational psychology (pp. 13–52). orientations and reactions toward alternative human resource manage- San Francisco: New Lexington Press. ment practices. Human Relations, 51, 571–588. Triandis, H. C., Bontempo, R., Betancourt, H., Bond, M., Leung, K., Robert, C., Probst, T. M., Martocchio, J. J., Drasgow, F., & Lawler, J. J. Brenes, A., et al. (1986). The measurement of the etic aspects of (2000). Empowerment and continuous improvement in the United individualism and collectivism across cultures. Australian Journal of States, Mexico, Poland, and India: Predicting fit on the basis of the Psychology, 38, 257–267. dimensions of power distance and individualism. Journal of Applied Triandis, H. C., Bontempo, R., Villareal, M. J., Asai, M., & Lucca, N. Psychology, 85, 643– 658. (1988). Individualism and collectivism: Cross-cultural perspectives on Robinson, S. L., & Bennett, R. J. (1997). Workplace deviance: Its nature, self– group relationships. Journal of Personality and Social Psychology, its causes, and its manifestations. In R. J. Lewicki, R. J. Bies, & B. H. 54, 323–338. Sheppard (Eds.), Research on negotiation in organizations (Vol. 6, pp. Triandis, H. C., Chan, D. K. S., Bhawuk, D. P. S., Iwao, S., & Sinha, 3–28). Greenwich, CT: JAI Press. J. B. P. (1995). Multimethod probes of allocentrism and idiocentrism. Robinson, S. L., & O’Leary-Kelly, A. M. (1998). Monkey see, monkey do: International Journal of Psychology, 30, 461– 480. The influence of work groups on the antisocial behavior of employees. Triandis, H. C., & Gelfand, M. J. (1998). Converging measurement of Academy of Management Journal, 41, 658 – 672. horizontal and vertical individualism and collectivism. Journal of Per- Sackett, P. R., & DeVore, C. J. (2001). Counterproductive behaviors at sonality and Social Psychology, 74, 118 –128. work. In N. Anderson, D. S. Ones, H. K. Sinangil, & C. Viswesvaran Triandis, H. C., Leung, K., Villareal, M. J., & Clack, F. L. (1985). (Eds.), Handbook of industrial, work, and organizational psychology Allocentric versus idiocentric tendencies: Convergent and discriminant (Vol. 1, pp. 145–151). Thousand Oaks, CA: Sage. validation. Journal of Research in Personality, 19, 395– 415. Schmitz, K. (1992). Collectivism and patriotism. Unpublished honors Triandis, H. C., McCusker, C., & Hui, C. H. (1990). Multimethod probes thesis, Department of Psychology, University of Illinois. of individualism and collectivism. Journal of Personality and Social Schneider, R. J., Hough, L. M., & Dunnette, M. D. (1996). Broadsided by Psychology, 59, 1006 –1020. broad traits: How to sink science in five dimensions or less. Journal of Triandis, H. C., & Singelis, T. M. (1998). Training to recognize individual Organizational Behavior, 17, 639 – 655. differences in collectivism and individualism within culture. Interna- Schwab, D. P. (1980). Construct validity in organizational behavior. In tional Journal of Intercultural Relations, 22, 35– 47. L. L. Cummings & B. M. Staw (Eds.), Research in organizational Triandis, H. C., & Suh, E. M. (2002). Cultural influences on personality. behavior (Vol. 2, pp. 3– 43). Greenwich, CT: JAI Press. Annual Review of Psychology, 53, 133–160. Shaw, J. D., Duffy, M. K., & Stark, E. M. (2000). Interdependence and Van Dyne, L., Vandewalle, D., Kostova, T., Latham, M. E., & Cummings, preference for group work: Main and congruence effects on the satis- L. L. (2000). Collectivism, propensity to trust, and self-esteem as pre- faction and performance of group members. Journal of Management, 26, dictors of organizational citizenship in a non-work setting. Journal of 259 –279. Organizational Behavior, 21, 3–23. Singelis, T. M., Triandis, H. C., Bhawuk, D., & Gelfand, M. J. (1995). Wagner, J. A., III. (1995). Studies of individualism– collectivism: Effects Horizontal and vertical dimensions of individualism and collectivism: A on cooperation in groups. Academy of Management Journal, 38, 152– theoretical and measurement refinement. Cross-Cultural Research: The 172. Journal of Comparative Social Science, 29, 240 –275. Wagner, J. A., III, & Moch, M. K. (1986). Individualism– collectivism: Smith, C. A., Organ, D. W., & Near, J. P. (1983). Organizational citizen- Concept and measure. Group and Organization Studies, 11, 280 –304. ship behavior: Its nature and antecedents. Journal of Applied Psychol- Waldman, D. A. (1997). Predictors of employee preferences for multirater ogy, 68, 653– 663. and group-based performance appraisal. Group and Organization Man- Smith, K. G., Carroll, S. J., & Ashford, S. J. (1995). Intra- and interorga- agement, 22, 264 –287. nizational cooperation: Toward a research agenda. Academy of Manage- Williams, L. J., & Anderson, S. E. (1991). Job satisfaction and organiza- ment Journal, 38, 7–23. tional commitment as predictors of organizational citizenship behaviors. Triandis, H. C. (1989). A strategy for cross-cultural research in social Journal of Management, 17, 601– 617. (Appendixes follow)

898 JACKSON, COLQUITT, WESSON, AND ZAPATA-PHELAN Appendix A The Triandis Family of Collectivism Measures Measure In-groups Dimension labels Dimension sample items Hui (1988): 63 items (82 Coworkers 1. Consideration of 1. The decision of where one is to work should be jointly citations last 5 years) Friends implications of one’s own made with one’s spouse, if one is married. Neighbors actions for others Triandis et al. (1986, Spouse 2. I would help if a colleague at work told me that he/she 1988, 1990): 17–29 Parents 2. Sharing of material resources needed money to pay utility bills. items (347 citations last Kin 5 years) 3. Sharing of nonmaterial 3. One needs to return a favor if a colleague lends a helping None resources hand. Singelis et al. (1995); Coworker Triandis & Gelfand Friend 4. Susceptibility to social 4. My good friends and I agree on the best places to shop. (1998); Gelfand & Neighbor influence Realo (1999): 8–16 Family 5. Each family has its own problems unique to itself. It does items (214 citations last Spouse 5. Self-presentation and face- not help to tell relatives about one’s problems. 5 years) Parent work Kin 6. The motto “sharing is both blessing and calamity” is still Note. R ϭ reverse scored. Nationality 6. Sharing of outcomes applicable even if one’s friend is clumsy, dumb, and causes Work a lot of trouble. 7. Feeling of involvement in group others’ lives 7. There is everything to gain and nothing to lose for Coworkers classmates to group themselves for study and discussion. Friends 1. Concern for in-group Neighbors 1. When my colleagues tell me personal things about Relatives 2. Interdependence themselves, we are drawn closer together. 3. Family integrity 2. I would help within my means, if a relative told me that s(he) is in financial difficulty. 4. Self-reliance (R) 5. Distance from in-groups (R) 3. One should live life independently of others as much as possible. (R) 1. Horizontal collectivism (includes concern for in- 4. One does better working alone than in a group. (R) group, distance from in- 5. My happiness is unrelated to the well-being of my group, interdependence, sociability) coworkers. (R) 1. The well-being of my co-workers is important to me. 2. Vertical collectivism (includes concern for in- I feel good when I cooperate with others. group, family integrity) To me, pleasure is spending time with others. 2. It is important to me that I respect the decisions made by my groups. Parents and children must stay together as much as possible. Family members should stick together, no matter what sacrifices are required.

PSYCHOLOGICAL COLLECTIVISM 899 Appendix B The Earley and Wagner Families of Collectivism Measures Measure In-group Dimension labels Dimension sample items Earley family Erez & Earley (1987): 4 Work group Unidimensional Only those who depend on themselves get ahead in life. (R) items (27 citations last Family One should live life independent of others as much as possible. (R) 5 years) Working with a group is better than working alone. In society, people are born into extended families or clans who Earley (1993): 8 items None Unidimensional (72 citations last 5 Work group protect them in shared necessity for loyalty. years) Friends One does better work working alone than in a group. (R) If a group is slowing me down, it is better to work alone. (R) Problem solving by groups gives better results than problem solving by individuals. An employee should accept the group’s decision even when personally he or she has a different opinion. Wagner family Wagner & Moch (1986): Work group 1. Beliefs 1. My work group is more productive when its members follow 11 items (24 citations Work group their own interests and concerns. (R) last 5 years) 2. Values 2. I prefer to work with others in my work group rather than to Wagner (1995): 20 items 3. Norms work alone. (72 citations last 5 years) 1. Self-reliance (R) 3. People in my group should recognize that they are not always 2. Competitive success (R) going to get what they want. 3. Value attached to 1. Only those who depend upon themselves get ahead in life. (R) working alone rather 2. Winning is everything. (R) than in a group (R) 3. Working with a group is better than working alone. 4. Subordination of personal needs to group 4. People in a group should realize that they sometimes are going interests to have to make sacrifices for the sake of the group as a whole. 5. Effects of personal pursuits on group 5. A group is more productive when its members do what they productivity want to do rather than what the group wants them to do. (R) Note. R ϭ reverse scored. Received December 20, 2004 Revision received July 28, 2005 Accepted August 10, 2005 Ⅲ Instructions to Authors For Instructions to Authors, please consult the January 2006 issue of the volume or visit www.apa.org/journals/apl and click on the “Instructions to authors” link in the Journal Info box on the right.