self or group? cultural effects of training on self-efficacy and performance

Self or Group? Cultural Effects of Training on Self-Efficacy and PerformanceAuthor(s): P. Christopher EarleySource: Administrative Science Quarterly, Vol. 39, No. 1 (Mar., 1994), pp. 89-117Published by: Sage Publications, Inc. on behalf of the Johnson Graduate School of Management,Cornell UniversityStable URL: http://www.jstor.org/stable/2393495 .

Accessed: 20/06/2014 08:38

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

Sage Publications, Inc. and Johnson Graduate School of Management, Cornell University are collaboratingwith JSTOR to digitize, preserve and extend access to Administrative Science Quarterly.

http://www.jstor.org

This content downloaded from 31.51.20.235 on Fri, 20 Jun 2014 08:38:48 AMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=sage

http://www.jstor.org/action/showPublisher?publisherCode=cjohn

http://www.jstor.org/action/showPublisher?publisherCode=cjohn

http://www.jstor.org/stable/2393495?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


Self or Group? Cultural Effects of Training on Self-efficacy and Performance

P. Christopher Earley University of California, Irvine

? 1994 by Cornell University. 0001 -8392/94/3901-0089/$1 .00.

0

The author gratefully acknowledges the comments and suggestions of Elaine Mosakowski and Anne Tsui on an earlier draft of this paper. Requests for reprints and other correspondence should be addressed to the author at the Graduate School of Management, University of California at Irvine, Irvine, CA 92717. The author would like to thank Bao Ji Ming, Luo Xing-jian, Zou Qiming, and Zhang Yong-lin for their assistance in collecting data in the People's Republic of China as well as Susan Peterson for her assistance in collecting the data in the United States.

This paper examines the theoretical and empirical relationship of training and individualism-collectivism to self-efficacy (a person's estimate of his or her ability to perform a task) and performance in studies of managers from Hong Kong, the People's Republic of China, and the United States. A laboratory experiment and a six-month field experiment were used to test hypotheses predicting that for individualists, self-focused training would have a stronger impact on self-efficacy and performance than would group-focused training and, for collectivists, group-focused training would have a stronger impact on self-efficacy and performance than would individual-focused training. The results show consistent support for the hypotheses at both a cultural and an individual level of analysis. A general model of self-efficacy and culture in an organizational environment is discussed.'

While scholars have increasingly emphasized the important role in work performance of a person's cognitive estimate of his or her capability to perform a given task, or his or her self-efficacy (e.g., Bandura, 1986; Wood and Bandura, 1989), scant attention has been paid to how self-efficacy functions across national and cultural work contexts (for exceptions, see Triandis, 1989; Erez and Earley, 1993). In this paper, I study the underlying process through which cultural background influences how individual and group training affects self-efficacy and performance.

INTRODUCTION

Scholars have proposed several typologies of cultural dimensions that are useful for such a study. One such dimension is individualism and collectivism, or an individual's perceptions and attitudes toward him- or herself and others in social relationships (Kluckhohn and Strodtbeck, 1961; Triandis, 1989; Hofstede, 1991; Schwartz, 1993). In an individualistic culture, people look to their own actions to understand who they are, and these actions are relatively independent of others. In a collectivistic culture, people base their self-understanding on the reactions of important others around them. A worker from an individualistic culture strives to improve work performance because of the recognition he or she may receive, whereas a worker from a collectivistic culture seeks improvement because of the gains his or her group may receive (Wagner and Moch, 1986; Erez and Earley, 1993). Thus, people's self-concepts are regulated, in part, by their cultural orientation and values (Epstein, 1973; Rokeach, 1973).

The Role of Individualism-Collectivism in Shaping Self-efficacy

Bandura (1986: 391) posited that self-efficacy influences performance primarily through increasing a person's effort and persistence. An individual with high self-efficacy works harder and longer than an individual with low self-efficacy (Wood and Bandura, 1989). One way that self-efficacy is shaped is through social influence. Verbal coaching and information that a person receives about performance norms, future expectations, and past performance all

89/Administrative Science Quarterly, 39 (1994): 89-1 17



influence self-efficacy (Bandura, 1986) by persuading him or her that a given performance level is attainable. Garland and Adkinson (1987) found that self-efficacy was increased by simply telling subjects during the training before a task, "You can do it." Meyer and Gellatly (1988) found that subjects who were presented with normative information before a task on performance levels achieved by other subjects changed their levels of self-efficacy. These studies demonstrate that information, such as task training, a person receives shapes self-efficacy through a variety of influences. For instance, normative information may make cognitively salient certain performance levels over others through a priming or attributional effect (Garland, 1985; Gist and Mitchell, 1992). A person's confidence may be boosted by verbal coaching based on his or her relation to the coach (Hinrichs, 1976). Another effect of normative information on people's efficacy is due to framing and anchoring influences (Bazerman, 1990; Earley and Erez, 1991). What remains unstudied is where people look to get this information and how this might be related to people's cultural backgrounds.

Bandura (1986) suggested that self-efficacy is, in part, socially constructed and that such construction may differ as a function of national culture. Just as our culture teaches us what ideals to hold and what beliefs to endorse (Rokeach, 1973), it plays a role in how we construct our self-efficacy. Several researchers (Triandis, 1989; Markus and Kitayama, 1991; Erez and Earley, 1993) have argued that individualists and collectivists, categorized by the cultures from which they come, differentially sample their social environment. Triandis (1989) used Baumeister's (1986) distinction among the private, public, and collective selves, in which the public self refers to the self using generalized others, the private self refers to using personal reference points, and the collective self refers to using a specific reference group, or in-group, in an assessment of the self. He argued that the likelihood of sampling a particular self is related to cultural background, such that, for example, in families in which a child is urged to act independently, the private self is likely to be accessed when the child faces new challenges. Consistent with Markus and Kitayama (1991) and Triandis (1989), Erez and Earley (1993) suggested that individualists use privately referenced information (e.g., their own performance) in establishing their self-efficacy, collectivists use in-group-referenced information (e.g., the in-group's performance), and that, other aspects of culture being comparable, both individualists and collectivists sample the public self with equal frequency.

By extending this logic to training in a cultural context, I argue that individualism and collectivism partly determine a person's use of the information provided during training and, hence, self-efficacy and task performance. For the individualist, training that emphasizes personal capability (the private self) will tend to be sampled and used. For the collectivist, whose focus is on the collective self, training that emphasizes in-group capability will tend to be sampled and used:

Hypothesis la (Hia): People from a collectivistic culture who are provided with group-focused training will have higher self-efficacy

90/ASQ, March 1994



Self or Group?

and perform better than people from a collectivistic culture who are provided with individual-focused training. Hypothesis lb (Hib): People from an individualistic culture who are provided with individual-focused training will have higher self-efficacy and perform better than people from an individualistic culture who are provided with group-focused training.

Memory structures, knowledge, and experiences stored schematically are not solely accessed through a single self; rather, people more easily incorporate information that is provided when it is consistent with their culturally dominant self. As Triandis (1989) suggested, whether they are individualists or collectivists, people sample from all three selves, with the amounts varying by cultural background. This implies that individualists provided with group-focused training or collectivists provided with individual-focused training do not ignore the information they receive; they use it to different degrees in assessing their self-efficacy, provided that it is relevant to a given task. While training consistent with a person's cultural background will be more effective than inconsistent training, training that is inconsistent will still be sampled, and it will provide some benefits. Indirect support for this point is evident in the training literature, which has shown that people respond to both individual and group-based methods (Hinrichs, 1976). Thus, I also hypothesize,

Hypothesis ic (Hic): Regardless of a person's cultural background, either type of training will increase self-efficacy more than no training at all.

Training information and culture may also jointly influence performance. Based on the literature discussed by Bandura and his colleagues (Bandura, 1986; Wood and Bandura, 1989), the logical causal chain is that culture and training influence self-efficacy and effort which, in turn, influence performance. The relationship of self-efficacy and effort to task performance is well documented; people with high self-efficacy work harder and outperform people with low self-efficacy (Gist and Mitchell, 1992). A simplified version of the model described by Gist and Mitchell (1992) illustrates this chain. They argued that self-efficacy and its consequences, such as effort, mediate the influence of experience (e.g., verbal persuasion) on performance. Thus, I further hypothesize,

Hypothesis 2 (H2): Self-efficacy and effort will mediate the interactive effects of training and culture on performance.

To test these hypotheses, two studies were conducted in a highly individualistic culture (United States) and two highly collectivistic cultures (Hong Kong and People's Republic of China). U.S. culture consists of a strong work ethic emphasizing individual achievement and reward, as well as a strong individual goal orientation (Hofstede, 1980; Triandis, 1988). Chinese society, by contrast, has been historically focused on social interests, collective action, and an emphasis on shared responsibility (Li, 1978; Hsu, 1985; Boisot and Child, 1988; Bond, 1988). The cultural heritage shared by Hong Kong and the People's Republic of China has reinforced a number of similarities across the two cultural environments, including an emphasis on in-group loyalty and willingness to put group interests ahead of

91 /ASQ, March 1994



self-interests. While the cultural and economic revolutions of the 1970s in the People's Republic of China have placed additional emphasis on equality, contribution to society and group welfare, and concern for interpersonal and work relationships (Lindsay, 1983; Hsu, 1985; Laaksonen, 1988; Earley, 1993), recent research using Hong Kong Chinese participants has demonstrated the strong, collective cultural norms that exist there (e.g., Hofstede, 1980; Bond, Leung, and Giacalone, 1985; Bond, 1988). Although Hong Kong and the People's Republic of China are not identical cultures, what is important for my studies is the relative position of Hong Kong and the People's Republic of China compared with the United States on the cultural value of individualism-collectivism. Chinese culture's collective orientation and high social interest (e.g., social integration, interpersonal responsiveness) suggest that Chinese workers in general are more responsive to group context than Americans, and they focus more on social interests, collective action, and shared responsibility.

Overview of Experiments

Two types of information were provided during training in both studies: information concerning a person's own capability to perform a task (individual-focused) and information concerning the capability of a person's in-group (reference group) to perform a task (group-focused). In the individual-focused training condition, participants were given information about their own actions and capability. In the group-focused condition, the participants were given information about their group's capability.

In the laboratory experiment, I focused on the role of training and culture on an individual's performance as a result of his or her self-efficacy and effort. The only difference between the training conditions was the focal point of the information (self versus group). Study 1 illustrates the direct impact of information type on performance through self-efficacy and effort.

I also wanted to examine the effect of an individual's work strategy, which Wood and Bandura (1989) noted often plays a strong role in determining performance. In Study 2, therefore, I conducted a field experiment to extend the training intervention to include task-strategy information in the form of job-related information concerning how to perform better. This was done both to enhance the mundane realism of the field experiment, given that training in a real-world context contains work-strategy information as well as performance expectations, and to expand the sophistication of the intervention in order to determine if self-efficacy will still have an effect when the job-training information includes task-strategy information.

STUDY 1: LABORATORY EXPERIMENT

Method

Participants. Two hundred and fifty-one managers (67 Hong Kong Chinese, 96 Chinese from the People's Republic of China, and 87 Americans) participated in the study on a voluntary basis. The Chinese participants were recruited

92/ASQ, March 1994



Self or Group?

from management training courses hosted by a university in southern China and a university in Hong Kong. All of the managers were natives of the countries in which they were attending their training. The American participants were recruited from a management training course that they were attending on human resources management (HRM). Participants from all three countries were employed in full-time management positions, and they were sponsored by their organizations for the course. A comparison of the three groups based on age, education level, gender, and company size demonstrated no differences among the groups. The mean age of participants was 32.3 years, modal education level was a bachelor's degree, and company size was between 5,001 and 10,000 employees. In addition, 20 participants from Hong Kong, 25 from the United States, and 24 from the People's Republic of China were women.

Design and task. The design used a cultural variable, individualism-collectivism, and three types of task training, no training, individual-focused training, or group-focused training. The purpose of the training manipulation was to determine whether or not people derive their efficacy expectations from different sources (individual level versus group level versus no training). In the no-training condition, managers were not given any training. In the individual-focused training condition, managers were given information specific to their own performance potential and actions. In the group-focused training condition, managers were given information specific to their group's performance potential and actions. Individualism-collectivism was measured as a continuous variable.

The experimental task was to generate alternative daily work schedules of employees based on a three-shift workplace and 30 employees having various schedule preferences. The managers were asked to generate as many alternative schedules as possible during a 30-minute period, using the constraints of employees' preferences for shifts and their availability. The managers were told that they had to conform to several rules in scheduling: First, they had to use employees' preferences and availability for shifts; second, no employee could serve on more than a single shift on a given day; and, third, the schedules could not repeat themselves. The task was chosen both because it consisted of an activity familiar to all participants in their normal work activities and because similar scheduling tasks have been used successfully in other task-performance studies (e.g., Erez, Earley, and Hulin, 1985). The schedules were scored as correct if all scheduling rules were followed. Scoring was done by two raters, who had a high interrater reliability (r = .97, p < .01). Sets of materials were prepared for the subjects in their native language. The procedure for developing and translating the materials used back-translation (Brislin, 1980); the text was simplified through the use of short sentences and focused on specific rather than general concepts. The back-translation was performed by two assistants to the experimenter who are bilingual, and the translated version of the task was examined by a Hong Kong Chinese colleague in order to ensure that it made sense for the Hong Kong sample.

93/ASQ, March 1994



Dependent measures. Performance was measured by the number of work schedules correctly completed by a participant during the 30-minute performance period.

Self-rated effort was measured with two items, using a 5-point scale, before subjects began the task: (1) "How hard are you going to try as you work on this task?" (1 = not at all hard and 5 = extremely hard) and (2) "How much effort do you intend to exert as you complete schedules?" (1 = no effort and 5 = a great deal of effort). These items were averaged for a composite effort score, and the items were significantly correlated (r = .82, p < .01).

Individualism-collectivism was assessed on a 5-point Likert scale (1 = strongly disagree and 5 = strongly agree), using the eight-item version of a questionnaire developed by Earley (1993), who adapted previous items in order to focus on the goal, task-performance, and in-group aspects of this cultural value, individualism-collectivism. Items included (1) "Employees like to work in a group rather than by themselves"; (2) "If a group is slowing me down, it is better to leave it and work alone"; (3) "To be superior, a man must stand alone"; (4) "One does better work working alone than in a group"; (5) "I would rather struggle through a personal problem by myself than discuss it with my friends"; (6) "An employee should accept the group's decision even when personally he or she has a different opinion"; (7) "Problem solving by groups gives better results than problem solving by individuals"; and (8) "The needs of people close to me should take priority over my personal needs." I chose to use this questionnaire because goals and performance are integral aspects of self-efficacy (Wood and Bandura, 1989) and because it has been used successfully in the countries studied in my research.

Responses to the scale were coded so that a high score indicated collectivistic values, and a low score indicated individualistic values; the reliability (Cronbach's alpha) of the scale was .73. A principal-components analysis demonstrated that the items loaded on a single factor having an eigenvalue of 4.89, accounting for 49 percent of the total variance; factor loadings ranged from .51 to .82.

To measure self-efficacy, subjects were asked to rate their self-efficacy for nine levels of possible performance- completing 5, 10, 15, 20, 25, 30, 35, 40, and 45 schedules-using a 1 00-point certainty scale, where 0 =

"certain the performance level cannot be achieved" and 100 = "certain the performance level can be achieved." For subsequent analyses, the responses to the scale were averaged for a composite self-efficacy score that had a reliability (Cronbach's alpha) of .75.

Procedure. The participants in all samples followed the same experimental procedure. The managers participated in the experiment during an HRM executive education course, which I taught, as a normal exercise during their regularly scheduled program. I introduced the exercise to the managers as an illustration of general management planning and work activities; they were asked if they were willing to participate in the exercise, and none refused. The managers were randomly assigned to one of the three training

94/ASQ, March 1994



Self or Group?

conditions, and they were given a packet containing various materials, including a questionnaire assessing general demographic information, individualism-collectivism, and a statement concerning their willingness to participate in the exercise. After completing the questionnaire, the managers read the task instructions and then worked on sample schedules for ten minutes, after which they were permitted to ask questions about the task. I then picked up these materials, handed out booklets containing the experimental task materials, and began the training intervention.

The managers were put into three separate areas based on the number at the top of their task booklets, which had been distributed on a random basis. In the no-training condition, participants were asked to read some general information concerning management practices (an interview with a CEO reported in Academy of Management Executive), which took approximately the same amount of time as the training intervention in the other two conditions. Pilot testing with this task demonstrated that people reading this article prior to working on the task did not become more fatigued than people simply instructed to begin immediately working on the task.

In the individual-focused training condition, I gave the participants a sheet containing several pieces of training information intended to bolster their individual self-efficacy expectations about performing the task. First, a formula was presented into which the managers put their practice trial performance, years of work experience, job level (based on a 3-point classification scheme), and years of education and then calculated a number that they compared with a categorization scheme of supplied values to extrapolate their performance across a 30-minute period. The categorization scheme was constructed so that all of the managers fell into the same category, although none of them was aware that the outcome was contrived. This was accomplished by supplying values for the high category that everyone would fall into. Second, the managers read three "manager performance profiles" and were told to choose the one that most closely resembled them. They were again referred to the categorization scheme to see what performance level they might expect to achieve. As with the formula, the profiles were constructed so that the managers would find themselves in the same category that was indicated by the formula. This was accomplished by constructing the profiles such that only one profile would fit all the managers. Third, the managers completed two items that captured their "management performance quotient": (1) "Based on your typical work performance, would you characterize yourself as an energetic and dedicated employee or someone who is quite distracted and uninvolved in your work?"; and (2) "Are you a performance-oriented person or someone who just completes the minimum requirements?", where 1 = yes and 0 = no. They were again referred to a categorization scheme to see what performance level they might expect to achieve. As with the formula, the profiles were constructed so that the managers would find themselves in the same category that was indicated by the

95/ASQ, March 1994



formula. In this category, a response of yes to either question placed them in a high-performance category.

Finally, I told the managers to calculate an overall performance-potential score from the number of times that they categorized themselves in the "high-performance" category (a maximum of three times, using the three sections of the materials), and they were told to evaluate their potential based on a final categorization scheme at the end of the materials: low performance-i 5-25 schedules; average performance-25-35 schedules; or high performance-over 35 schedules. A post-hoc analysis of the managers' classification demonstrated that they all correctly classified themselves in the high-performance category, which demonstrates that the manipulation was successful.

In the group-focused training condition, I gave the managers an information sheet similar to the one used in the individual-focused condition except that the various items were adjusted to reflect the potential performance level of others whom they viewed as important to them. The instructions specifically directed them to think about four or five of their closest friends and/or family who worked. I chose to include referent members from both family and friends, rather than limit the categorization to coworkers, because previous research on group membership and collectivism has shown the importance of familial connections to in-group composition (e.g., Triandis, 1989). Managers were asked to write the initials of these people at the top of their sheet and to keep thinking of them as they worked through the items. The first item was a formula in which the managers put in the average number of years of work experience that their referent friends/family members had accumulated, average job level (based on a 3-point classification scheme), and average degree acquired. The managers then calculated a number that they compared with a categorization scheme of supplied values to extrapolate their group's performance capability across a 30-minute period. As in the individual-focused condition, the categorization scheme was constructed so that all of the managers fell into the same category, although none of them was aware that the outcome was contrived. Second, the managers read three "manager performance profiles" and were told to choose the one that most closely resembled their chosen family/friends. They were again referred to the categorization scheme to see what performance level they might expect to achieve. The profiles were constructed so that the managers would inevitably pick the same category as indicated by the formula. Third, the managers completed two items that captured their "management performance quotient": (1) "Based on your typical work performance, would you characterize your chosen family members/friends as energetic and dedicated or individuals who are quite distracted and uninvolved in their work?"; and (2) "Are your chosen family members/friends performance-oriented or people who just complete the minimum of what is required of them?" Again, these items were constructed so that the managers would answer consistently with one another, and they were directed to evaluate the responses with the supplied

96/ASQ, March 1994



1 I thank Elaine Mosakowski for her suggestions concerning this analysis of culture and individual differences characteristics.

Self or Group?

categorization scheme. Finally, the managers were told to calculate an overall performance potential score for these referent others from the number of times that they categorized their family/friends in the "high-performance" category (a maximum of three times, using the three sections of the materials), then to categorize the performance potential of their referent choices based on a final categorization scheme at the end of the materials: low performance- 15-25 schedules; average performance-25- 35 schedules; high performance-over 35 schedules. A post-hoc analysis of the classification used by the managers demonstrated that all of the managers correctly classified their referent group in the high-performance category, which demonstrates that the manipulation was effective. I then gave the managers a short questionnaire assessing their intended effort level and self-efficacy expectations.

In each group, after the questionnaire was completed and collected, the managers were instructed to begin working on the task with the booklet of materials provided. At the end of the 30-minute performance period, I collected the materials, debriefed the managers concerning the purpose of the experiment, and answered any remaining questions they had. Finally, I discussed the relationship of the experiment to processes of work motivation and performance in organizations across various cultural settings.

Analysis. My method of analysis consisted of regressing performance on the predictor variables (effort, efficacy, individualism-collectivism, training condition), followed by country of origin, which was captured using two dummy variables contrasting Hong Kong with other countries (dummy 1) and the United States with the other countries (dummy 2). Inasmuch as I was interested in the relationship of individualism-collectivism and its interaction with training as mediating variables, I analyzed the data using a regression model rather than using an ANOVA approach.

Given that the hypotheses concern the relationship of training condition and individualism-collectivism to performance, I created two predictor variables that capture the cultural-level and individual-level aspects of individualism-collectivism.1 The general logic of this procedure is to separate the "shared" (or cultural-level) aspect of the assessed values and beliefs (individualism and collectivism) from that aspect of the values and beliefs that has been uniquely shaped by an individual's experiences (an individual-level characteristic). Separating collectivism- individual from collectivism-group allows me to estimate the unique contribution of each level of this construct (shared value versus individual differences characteristic). The hypotheses concerning individualism and collectivism were therefore tested comparably at the cultural level and the individual level, using collectivism-group (cultural-level) and collectivism-individual (individual-level), with collectivism- individual functioning as a psychological variable and collectivism-group functioning as a cultural variable.

I calculated the mean individualism-collectivism score for each country and assigned this score to each participant

97/ASQ, March 1994



from that country (labeled collectivism-group). I then calculated a difference score by subtracting each person's individualism-collectivism mean score for the 8-item scale from his or her country score (labeled collectivism-individual) after removing that person's own score from the overall mean. Collectivism-group represents a country (or the "culture") score, while collectivism-individual represents an individual's deviation from the country score. Thus, the former variable represents the general culture shared within a given nation, and the latter variable represents an individual-differences characteristic.

The use of a difference score is not without problems. In particular, difference scores are subject to low reliability (Cohen and Cohen, 1975: 375-382), particularly when calculating differences using change scores, which attenuates the magnitude of the relationship between the deviation score and other variables. This unreliability becomes increasingly problematic as the component reliabilities of the variables used to calculate the difference score become increasingly intercorrelated and unreliable themselves. In my studies, the correlation between collectivism-group and the individualism-collectivism score was moderate (.50 and .31 for studies 1 and 2, respectively), and the component scale reliabilities were in excess of .80, which yields an acceptable estimated reliability (.68 and .72 for studies 1 and 2, respectively) for the differences scores (Cohen and Cohen, 1975: 64).

I aggregated the samples for the analysis and examined the residual influences of country of origin in predicting performance after accounting for the collectivism-individual and collectivism-group variables. The general logic of this approach is that individualism-collectivism (partitioned into collectivism-individual and collectivism-group components), rather than country of origin, drives the hypothesized interaction. After accounting for variance attributable to collectivism-individual, collectivism-group, training, collectivism-group x training, and collectivism-individual x training, the two dummy variables for country and their interaction with training should not explain additional variance in performance. Further, I would anticipate that both collectivism-individual and collectivism-group (and their interactions with training) would significantly predict performance both before and after entering the two dummy variables for country and their interaction with training.

Resu Its

The means and standard deviations for performance, effort, efficacy, and individualism-collectivism are presented in Table 1 for the three training conditions and three countries. The Pearson correlations for all the variables are in the Appendix.

To test the hypotheses that people from individualistic and collectivistic cultures would respond differentially to individual-focused and group-focused training (H1a, Hib, HMc), I regressed performance hierarchically on the demographic variables of age, education level, gender, and company size (step 1), training, collectivism-group and

98/ASQ, March 1994



Self or Group?

collectivism-individual (step 2), the training by collectivism-group, collectivism-individual interactions (step 3), the two country dummy variables (step 4), and the dummy variables by training interactions (step 5). The results of this analysis are presented in Table 2, and there appears to be clear support for the hypothesized interactions. After controlling for the main effects, the interaction terms were significantly related to performance (change in R2 = .08, p < .01).

Table 1

Descriptive Statistics for Variables across Training Conditions and Country of Origin*

United States (US) Hong Kong (HK) People's Republic of China (PRC) Training condition Training condition Training condition

Indi- Indi- Indi- No vidual- Group- No vidual- Group- No vidual- Group-

training focused focused training focused focused training focused focused

Performance Mean 21 .00a 37.52b 28.96c 21 .50a 34.96b 36.40b 22.34a 31.02c 36.68b S.D. 2.36 6.21 6.65 3.01 4.95 5.52 4.14 6.74 5.40

Effort Mean 2.00d 4.04e 3.1 5f 2.33d 3.47ef 385 e 2,00d 3.23f 4.23e S.D. 1.08 1.51 1.05 1.46 1.12 1.14 .90 1.37 1.28

Self-efficacy Mean 72.00g 90.88h 74.92g9' 70.41g 86.47hJi 92.95h 74.21g 81.45i 90.79h

S.D. 8.21 12.15 8.43 7.53 13.31 6.52 7.15 11.45 10.91

Individualism- collectivism Mean 2.14i 2.32i 1,79i 3.75k 2.96k 3.60k 3.56k 3.52k 3.59k S.D. 1.00 1.46 .84 1.15 .98 .94 .84 1.04 .90

* Means within each dependent variable having different superscripts are significantly different at p < .05.

To better understand the interactions, the means for performance were examined for high and low levels of individualism-collectivism, based on a median split within each country, across training conditions. These data were analyzed using one-way ANOVAs within each collectivism group (country), across the training conditions. The results demonstrate a significant main effect for training condition within each collectivism-group sample [F(2,83) = 74.63, F(2,64) = 72.74, F(2,93) = 45.66, for the United States, Hong Kong, and People's Republic of China, respectively]. A post-hoc least significant differences test demonstrates that performance was higher in the individual-focused than in the group-focused training condition in the United States, whereas the opposite was true for Hong Kong and the People's Republic of China. In all three samples, the no-training condition resulted in lower performance than either of the other training conditions. These results support Hypotheses 1a-1c. To further understand individualism-collectivism at the individual level, I examined performance within levels of individualism-collectivism (low, high) across the training conditions for each collectivism-group sample. These means, shown in Table 3, were analyzed using one-way ANOVAs parallel to those used in the analysis at the collectivism- group level. The overall effects are presented in Table 3, along with the results of post-hoc contrasts using the least-significance-difference test. The results demonstrate

99/ASQ, March 1994



Table 2

Hierarchical Regression Analysis of Performance for Study 1, Testing for Moderating Effect of Individualism-Collectivism

Step R2 AR2 Beta t (for beta)

Age 1 .03 .03 .11 1.75 Education .13 1.95 Company size .11 1.73 Gender .02 .31

Collectivism-individual 2 .45 .42 -.08 -1.72 Collectivism-group .10 2.05- Training condition .62 12.65w

Collectivism-individual x training 3 .53 .08 .43 5.76w Collectivism-group x training .33 1.95-

Country dummy 1 (HK vs. US, PRC) 4 .53 0 .04 .92 Country dummy 2 (US vs. HK, PRC) .00 .00

Dummy 1 x training 5 .53 0 -.03 -.38 Dummy 2 x training .01 .00

*p < .05; s-p < .01.

that within each collectivism-group (country), individualists responded more positively to individual-focused training than group-focused training, whereas collectivists responded more positively to group-focused training than individual- focused training-a pattern that is repeated for all three countries.

Table 3

Post-hoc Analyses of Performance Using Country (Cultural Group), Training Condition, and a Median Split for Individualism-Collectivism Values*

Training Condition

Individualism- No Individual- Group- F Country collectivism training focused focused (d.f.)

United Low 21.04 40.42 28.12 119.92- States (2.35) (2.87) (5.91) (2,66)

High 20.90 28.93 41.01 26.41- (2.51) (4.72) (5.79) (2,16)

Hong Kong Low 21.83 36.94 33.10 65.44- (2.75) (3.64) (4.17) (2,36)

High 21.16 29.33 39.70 59.28- (3.32) (3.72) (4.76) (2,25)

People's Low 24.00 35.94 32.89 21.39- Republic (5.40) (4.53) (4.03) (2,43) of China High 20.83 26.89 40.47 106.89-

(1.59) (5.40) (3.00) (2,47)

*p < .01. * Standard deviations are in parentheses. Post-hoc tests of performance conducted for

each country within each level of individualism-collectivism (low, high) demonstrated that each training mean significantly (p < .05) differed from each other mean within a given level (e.g., the no-training, individual, and group-focused training conditions differed pairwise from one another within the Hong Kong collectivism-group sample).

I then added a step to the regression reported in Table 2 to test for the mediating effect of the individualism-collectivism construct (collectivism-group, collectivism-individual) in the relationship of the country dummy variables to performance. I added in the country dummy variables and then added their interaction with the training condition in an additional step. I also conducted a parallel analysis in which the order of entry

100/ASQ, March 1994



Self or Group?

for training, collectivism-group and collectivism-individual, training x collectivism-group, collectivism-individual versus the country dummy variables and their interaction with training were reversed, so that I could compare the variance accounted for in performance by country before and after controlling for collectivism-group and collectivism-individual and their interactions with training. Finally, I examined the correlation between the training condition by country dummy variable interactions and performance. If a mediating approach is to be supported, the country dummy variables (and their interaction with training condition) should not be related significantly to performance after accounting for variance attributable to collectivism-group and collectivism-individual, training condition, and their interaction, but they should account for a significant amount of variance prior to controlling for these other variables. The results of this analysis support the assertion that collectivism-group and collectivism-individual and their interaction with training condition explain the country-of-origin influence. Prior to controlling for collectivism-group and collectivism-individual, training condition, and their interaction, the country dummy variables and their interaction with training condition accounted for 39 percent of the variance in performance, but after controlling for these other variables, the country dummy variables and their interaction with training accounted for no additional variance in performance. In addition, the country dummy variables and their interaction with training were significantly correlated with performance (r = .14, .12, .31, and .36 for dummy 1, dummy 2, dummy 1 x training, and dummy 2 x training, respectively; p < .05). Taken together, these results illustrate that collectivism-group and collectivism-individual and their interaction with training explain the influence of country-on performance in this study. To test hypothesis 2, that effort and self-efficacy would mediate the effect of the collectivism-group, collectivism-individual by training interactions on performance, mediated regression analysis was conducted. Performance was regressed on the demographic variables (step 1), effort and self-efficacy (step 2), followed by collectivism-group, collectivism-individual, and training condition (step 3), followed by collectivism-group, collectivism-individual x training (step 4). A second regression equation was constructed in which effort and self-efficacy were entered into the equation after entering the other variables, and I examined the Pearson correlations of effort and self-efficacy with performance. The results, presented in Table 4, show a strong mediating effect for effort and self-efficacy, although they do not completely mediate the relationship. After accounting for the variance in demographic variables (step 1) and effort and self-efficacy (step 2), collectivism-group and collectivism-individual, training condition, and the collectivism-group, collectivism-individual by training interactions (step 3) accounted for a significant amount of variance in performance (7 percent). The primary variables driving this significant relationship are training condition and collectivism-individual by training. Prior to entering effort and self-efficacy, collectivism-group and collectivism-individual,

1O1/ASQ, March 1994



training condition, and collectivism-group, collectivism- individual by training interactions were significantly related to performance, accounting for 42 percent of the variance. These results demonstrate that effort and self-efficacy partially mediate the relationship of the collectivism-group, collectivism-individual by training interactions on performance. In contrast, the hypothesized mediators (effort and self-efficacy) accounted for 66 percent of the variance in performance prior to controlling for collectivism-group and collectivism-individual, training condition, and collectivism-group, collectivism-individual by training condition, whereas effort and self-efficacy still accounted for 23 percent of the variance in performance after controlling for collectivism-group and collectivism-individual, training condition, and their interactions.

Table 4

Hierarchical Regression Analysis of Performance for Study 1, Testing for Mediating Effect of Individualism-Collectivism


Panel A

Age 1 .03 .03 .11 1.75 Education .13 1.95- Company size .11 1.73 Gender .02 .31

Effort 2 .69 .66 .42 9.83-- Self-efficacy .53 12.62--

Collectivism-individual 3 .74 .05 .01 .03 Collectivism-group .02 .58 Training condition .27 6.94--

Collectivism-individual x training 4 .76 .02 .21 3.88-- Collectivism-group x training - .12 - .79

Panel B

Age 1 .03 .03 .11 1.75 Education .13 1.95- Company size .11 1.73 Gender .02 .31

Collectivism-individual 2 .45 .42 -.08 - 1.72 Collectivism-group .10 2.05- Training condition .62 12.65--

Collectivism-individual x training 3 .53 .08 .43 5.76-- Collectivism-group x training .33 1.95-

Effort 4 .76 .23 .30 7.49-- Self-efficacy .43 10.25--

*p < .05; -p < .01.

A final set of analyses was conducted on self-efficacy and effort, using the least-significant-differences test (see superscripts in Table 1, above, for results). The results demonstrate that self-efficacy and effort were significantly higher in the individual-focused training condition than in the other training conditions for U.S. managers, whereas self-efficacy and effort were significantly higher in the group-focused training condition than in the other training conditions for managers from the People's Republic of China. Finally, self-efficacy and effort were significantly

102/ASQ, March 1994



Self or Group?

higher in the individual- and group-focused training conditions than in the no-training conditions.

Discussion

This study shows that efficacy training differentially shapes a person's performance depending on the relationship of the training method to his or her individualism-collectivism orientation. People who were high on collectivism-group and collectivism-individual (collectivists) responded best to group-focused training information, whereas people who were low in collectivism-group and collectivism-individual (individualists) responded best to individual-focused training information. In addition, training information that was incongruent with a person's collectivism-group and collectivism-individual orientation was still more effective in enhancing performance than no training at all.

To test this model of self-efficacy, effort, and performance further, a field experiment was done, in which U.S. and Chinese (People's Republic of China only) service representatives were trained using individual and group-focused training methods in the context of ongoing job training by the companies participating in the study. While this second study represents a conceptual replication of Study 1, it also included a longitudinal assessment of several key variables over a six-month period.

STUDY 2: FIELD EXPERIMENT

Method

Participants. One hundred and eight service representatives (62 American and 46 Chinese) from similar communication companies in the United States and People's Republic of China participated in the experiment. As in Study 1, I attempted to match the samples, although the constraints of the field settings precluded a perfect match. The participants were compared on a number of characteristics, including age, gender, years of service with their company, and education. A comparison of the two groups based on age, education level, gender, and job tenure demonstrated no differences between the groups. The mean age of participants was 28.6 years, modal education level was a bachelor's degree, and mean job tenure was 2-5 years. In addition, 19 participants from the U.S. and 20 from the People's Republic of China were women.

Work sites. The U.S. company is located in the Midwest, has over 20,000 employees, and has a centralized structure. It produces telecommunications equipment and other diversified products and provides communication services. The Chinese company is located in the northeastern part of mainland China, has over 20,000 employees, and it has a centralized structure. It produces telecommunications equipment and provides some communication services. Both companies produce directly for the market and act as contractors for the government. The U.S. company, however, has a more geographically dispersed operation than the Chinese company.

103/ASQ, March 1994



Both companies use a simple performance monitoring by the representative's supervisor that I could use to assess the training interventions. In both companies, the supervisor assesses a representative's performance every three months from observations made while accompanying him or her during a normal workday several times during the performance period. At both companies, it is the practice of the employees' supervisors to evaluate their subordinates every three months using a single-page assessment of performance and for them to do a more complete evaluation, with narrative descriptions, every twelve months after an in-depth discussion with the employee about his or her work objectives and self-reported performance. The companies evaluated a group's performance every three months by pooling the evaluations of individual members of the group as well as assessing the overall synergy of the group based on members' ability to coordinate and cooperate.

Dependent measures. Before the training intervention, I obtained a baseline performance for each of the employees from personnel files by using their overall performance rating from the prior three-month period. The variables of performance, self-efficacy, and self-rated effort were assessed at two times: three months (Time 1) and six months (Time 2) after the training intervention.

Performance was measured using each company's standard appraisal system. Because the training interventions had the potential to affect employees' actual work performance, I kept the performance evaluations confidential. I chose to use the standard evaluation provided by a service representative's supervisor, which consists of an interview and field observations, so as to avoid introducing additional biases into the study. To assure comparability across national samples, two general performance items rated on a 9-point scale were selected from their evaluation instruments: (1) "Overall, how would you judge the quality of this employee's work performance?" and (2) "Please evaluate the service provided by this employee" (1 = extremely poor, 5 = average, and 9 = outstanding). The correlation of these items was .89 (p < .01).

Self-rated effort was measured using two items: (1) "How hard are you going to try as you work on your job?" (1 = not at all hard and 5 = extremely hard); and (2) "How much effort do you intend to exert as you perform your job duties as a service representative?" (1 = no effort and 5 = a great deal of effort). These items were averaged for a composite effort score, and the items were significantly correlated (r = .73, p < .01).

Individualism-collectivism, age, education, gender, and job tenure were assessed prior to the experimental interventions using a single-page questionnaire presented to each participant by a personnel representative. As with Study 1, all materials were back-translated into Chinese with the help of the personnel department representative.

104/ASQ, March 1994



Self or Group?

Individualism-collectivism was assessed using the scale described in Study 1. Responses to the scales were coded so that a high score indicated collectivistic values and a low score indicated individualistic values; the reliability (Cronbach's alpha) of the scale was .71. A principal- components analysis demonstrated that the items loaded on a single factor having an eigenvalue of 5.12, accounting for 51 percent of the total variance (factor loadings ranged from .48 to .85).

To measure self-efficacy, participants were asked to rate their self-efficacy for seven levels of overall performance ratings-achieving a 3,4,5,6,7,8, and 9 for himself or herself-using a 100-point certainty scale on which 0 = "certain the performance level cannot be achieved" and 100 = "certain the performance level can be achieved." For subsequent analyses, the responses to the scale were averaged for a composite self-efficacy score having a reliability (Cronbach's alpha) of .77. The Pearson correlations for the variables in Study 2 are in the Appendix.

Procedure. After making an initial contact with the companies in each of the countries, I worked with each company's personnel representative to determine the best job category from which to obtain participants for the training intervention. The participants chosen for this study were service representatives whose task it is to contact corporate customers and provide them with technical assistance for their equipment. A great deal of this contact is made in person, and it involves meeting with corporate customers and servicing their existing equipment, following up initial calls or visits, maintaining ongoing relationships, and providing a general service contact for the companies. People in this job were chosen for the study because the job is quite similar in content across the two countries despite such obvious differences as political and economic systems.

The next step was to obtain employees' consent to participate in the training effort. The program was introduced to the employees as a personal improvement seminar for service representatives. As confidentiality concerning performance outcomes associated with the program was assured, no one refused to participate. A total of 120 employees were initially contacted for the study, but approximately 10 percent failed to complete the study because they were transferred, left their job, or their data were incomplete, resulting in a final sample of 108 service representatives. Within each country, the service representatives were randomly assigned to one of the two training conditions, the only constraint being that individuals from the same work unit were put into the same training condition to avoid cross-condition contamination and so that no employee was at a disadvantage compared with another member of his or her work unit. This constraint did not seem to introduce a strong bias to the random assignment procedure, however, as in both companies, only a maximum of five employees came from the same work unit.

The procedure used in the study was similar in each country, and a personnel representative working with me conducted

105/ASQ, March 1994



the training sessions. In the individual-focused training condition, the service representatives were brought in for their training in a large group meeting, and the training intervention was provided by the personnel representative. The training consisted of normative training information as well as specific tactical information about how to perform their work better. The training provided employees with performance information based on their own prior performance with the company (normative), and they were given a booklet containing information about the job of service representative and how one might improve job performance from a service and quality perspective (tactical or strategic). In addition, a personnel representative discussed materials presented in a three-hour training session that included a general lecture and a series of dyadic interactions (employee with personnel representative) to clarify material presented in the packet. All of the discussion focused on the employee him- or herself and how the employee's past performance might be used to generate future successes.

In the group-focused training condition, the service representatives were also brought in for their training in a large group and the training intervention was also provided by the personnel representative. As in the individual-focused training, the intervention consisted of normative information, prior performance, and specific tactical information, but all of it was adapted so that an employee's work group was now the unit of consideration. Prior performance thus consisted of telling employees how their respective work units (and fellow service representatives) had performed and how their unit might perform better as a group. By focusing this information on the reference group, performance was framed in terms of the group's past performance and future capability. The tactical information concerning how to perform a job more effectively (e.g., improving client satisfaction through an open approach to assessing the problems with the phone system) was posed in terms of how the work unit might perform better. Thus, the content of the tactics that might be used was held constant across the two training conditions, while the reference point (individual versus group) changed.

After completing the training intervention, the employees returned to their normal work activities. The companies' normal evaluation procedures (reporting employee performance at the individual and group level) were used in assessing each representative's performance, and a follow-up training session identical to the first one was conducted by the personnel representative after the first evaluation occurred (at the end of three months). The training was conducted a second time because the personnel departments wanted to reinforce the training intervention. Finally, at the end of the second performance period (after six months), the personnel representative and I brought the employees in for a general debriefing on the various training procedures used in the study, and they were provided with the alternate training approach, not used in their experimental condition. Additionally, the employees

106/ASQ, March 1994



Self or Group?

were assured that their performance during this training assessment would not be used in their general personnel records.

Resu Its

Descriptive statistics. Table 5 presents the means and standard deviations for performance, effort, efficacy, and individualism and collectivism (collectivism-group and collectivism-individual) across the two training conditions (coded as 0, 1 for the individual- and group-training conditions, respectively), and countries (coded 0, 1 for the United States and the People's Republic of China, respectively).

Table 5

Descriptive Statistics for Variables across Training Conditions and Country of Origin*

People's Republic United States of China

Individual- Group- Individual- Group- Variable focused focused focused focused

Performance Baseline

Mean 5.06a 5.11a 4.88a 5.27a S.D. (.82) (.80) (.90) (.98)

Time 1 Mean 6.66b 6.74b 5.88c 7.21 b

S.D. (1.56) (1.24) (1.45) (1.41)

Time 2 Mean 6.94d 6.54d 5.75e 7.18 d S.D. (1.45) (1.28) (1.39) (1.61)

Effort Time 1

Mean 4.03f 3.429 3.639 4.1 2f S.D. (.78) (.69) (.71) (.78)

Effort Time 2

Mean 4.06 3.58 3.59' 4.03 S.D. (.71) (.84) (.81) (.85)

Self-efficacy Time 1

Mean 85.13i 79.92k 78.91k 86.45i S.D. (11.35) (11.80) (12.32) (9.89)

Self-efficacy Time 2

Mean 86.151 80.31 m 79.79m 86.81' S.D. (10.95) (11.05) (11.87) (10.68)

I ndividualism-collectivism Mean 2.72n 2.58n 3.580 3.270 S.D. (1.22) (1.26) (.97) (1.20)

* Means within each dependent variable having different superscripts are significantly different at p < .05.

As in Study 1, the method of analysis consisted of regressing performance on the predictor variables (effort, self-efficacy, collectivism-group and collectivism-individual, and training condition). Again, I partitioned individualism- collectivism into the country-level (collectivism-group) and

107/ASQ, March 1994



individual-level (collectivism-individual) components. In this two-country sample of Study 2, this means that collectivism-group and country are both dichotomous variables and, therefore, completely redundant, in contrast to Study 1, which had three countries and the partitioned collectivism-group variable assumed three values. Thus, the method of analysis used in Study 1, examining residual variance attributable to country, could not be used in Study 2. Instead, I tested the hypotheses using collectivism-group and collectivism-individual.

To test for the mediating role of training condition by individualism and collectivism interaction in the relationship of country by training interaction to performance, I conducted a regression using the procedure outlined in Earley (1989). I regressed performance at Time 1 and Time 2 hierarchically on the demographic variables of age, education level, gender, company size, and baseline performance (step 1), training, individualism-collectivism (step 2), training by individualism-collectivism interaction (step 3), and country and country-by-training interaction (step 4). The results of this analysis (available from the author) demonstrate support for the mediating effect. As with Study 1, it appears that individualism-collectivism captures the effect of country and training on performance.

Tests of hypotheses. To test the hypotheses that training condition and collectivism-group, collectivism-individual interact in predicting performance (H1a, H1b), performance at Time 1 and Time 2 was regressed hierarchically on the demographic variables of age, education level, gender, company size, and baseline performance (step 1), training, collectivism-group and collectivism-individual (step 2), and the training by collectivism-group, collectivism-individual interactions (step 3). The results of this analysis are presented in Table 6, panels A and B, and there appears to be clear support for the hypothesized interactions. After controlling for the main effects, the interaction terms accounted for a significant influence on performance (change-R2 = .35 at Time 1, .38 at Time 2, p < .01).

To understand better the nature of the interactions, the means for performance were examined for high and low levels of individualism-collectivism (based on a median split within each country) across training conditions. These data were analyzed using one-way ANOVAs within each collectivism-group (hence, within each country) across the training conditions. The results, shown in Table 7, demonstrate significant main effects for training condition for the People's Republic of China [F(1,53) = 19.13, 26.02, for Time 1 and Time 2, respectively] but no significant effects for the United States. These results demonstrate that training was differentially effective in the People's Republic of China (group-focused being superior to individual-focused) but not in the United States. These results support hypothesis la but not hypothesis lb.

To understand better individualism-collectivism at the individual level, I examined performance within levels of individualism-collectivism (low, high) across the training

1 08/ASQ, March 1994



Self or Group?

Table 6

Hierarchical Regression Analysis of Performance for Study 2, Testing for Moderating Effects of Individualism-Collectivism


Panel A: Performance, Time 1

Age 1 .16 .16 .04 .38 Education -.17 -1.88 Tenure .21 2.27- Gender .24 2.60w Baseline performance .14 1.45 Collectivism-individual 2 .21 .06 -.08 -.81 Collectivism-group - .01 Training condition .21 2.25- Collectivism-individual x training 3 .56 .35 .88 8.47" Collectivism-group x training 1.51 2.50

Panel B: Performance, Time 2

Age 1 .13 .13 .01 .14 Education -.10 -1.05 Tenure .17 1.78 Gender .21 2.28- Baseline performance .19 1.92 Collectivism-individual 2 .17 .04 - .02 - .23 Collectivism-group -.09 -.92 Training condition .18 1.87 Collectivism-individual x training 3 .55 .38 .89 8.50" Collectivism-group x training 2.07 3.38--

*p < .05; s-p < .01.

Table 7

Post-hoc Analyses of Performance (Time 1 and Time 2) Using Country, Median Split of Individualism-Collectivism, and Training for Study 2*

Individualism- Training Condition F

Performance collectivism Individual Group (d.f.)

Time 1 United States Low 8.07 6.10 25.15

(.92) (.99) (1,22) High 5.55 7.44 21.69-

(.92) (1.13) (1,25)

People's Republic Low 6.58 6.42 .11 of China (1.38) (1.26) (1,29)

High 5.17 8.28 66.98- (1.19) (.72) (1,24)

Time 2 United States Low 8.14 5.90 38.85

(.86) (.88) (1,22) High 6.00 7.67 14.88

(1.08) (1.00) (1,25)

People's Republic Low 6.42 6.32 .04 of China (1.24) (1.24) (1,29)

High 5.08 8.36 68.87 (1.24) (.75) (1,24)

*p < .01. * F tests for independent samples within each level of country and level of

individualism-collectivism; standard deviations are in parentheses.

109/ASQ, March 1994



conditions for each collectivism-group sample. These means were analyzed using one-way ANOVAs parallel to those used in the analysis at the collectivism-group level I just discussed. The results, shown in Table 7, demonstrate that within each collectivism-group (country), individualists responded more positively to individual-focused training than group-focused training, whereas collectivists responded more positively to group-focused training than individual-focused training. The single exception to this pattern is for the Chinese participants, who were low in individualism-collectivism. For these people, performance did not significantly differ as a function of training condition. Thus, the analyses at the individual level generally support hypotheses la and lb.

To test the hypothesis (H2) that effort and self-efficacy would mediate the effect of the collectivism-group, collectivism-individual by training interactions on performance, I conducted mediated regression analysis. Performance was regressed on the demographic variables (step 1), effort and self-efficacy (step 2), collectivism-group, collectivism-individual, and training condition (step 3), and collectivism-group, collectivism-individual x training (step 4). A second regression equation was constructed in which effort and self-efficacy were entered into the equation after entering the other variables. In addition, I examined the Pearson correlations of effort and self-efficacy with performance. The results of the analysis are presented in Table 8 (panels A and B for Time 1; panels C and D for Time 2). The analyses demonstrate a strong mediating effect for effort and self-efficacy, although training continued to have a modest, direct effect on performance. After accounting for the variance in demographic variables (step 1) and effort and self-efficacy (step 2), collectivism-group and collectivism- individual, training condition, and the collectivism-group, collectivism-individual by training interactions (step 3) accounted for a significant amount of variance in performance (4 percent). The variable driving this significant relationship is training condition. Prior to entering effort and self-efficacy, collectivism-group and collectivism-individual, training condition, and collectivism-group, collectivism- individual by training interactions were significantly related to performance, accounting for 35 percent of the variance. These results demonstrate that effort and self-efficacy partially mediate the relationship of the collectivism-group, collectivism-individual by training interactions on performance. In contrast, the hypothesized mediators (effort and self-efficacy) accounted for 47 percent of the variance in performance prior to controlling for collectivism-group and collectivism-individual, training condition, and collectivism-group, collectivism-individual by training condition, and effort and self-efficacy still accounted for 12 percent of the variance in performance after controlling for collectivism-group and collectivism-individual, training condition, and their interactions.

A final set of analyses were conducted on self-efficacy and effort using the least-significant-differences test. The results reported in Table 5 show that self-efficacy and effort were

11O/ASQ, March 1994



Self or Group?

Table 8

Hierarchical Regression Analysis of Performance for Study 2, Testing for Mediating Effect of Effort and Self-efficacy


Panel A: Performance, Time 1

Age 1 .16 .16 .04 .38 Education -.17 -1.88 Tenure .21 2.27- Gender .24 2.60"- Baseline performance .14 1.45 Effort (time 1) 2 .63 .47 - .03 - .38 Self-efficacy (time 1) .73 8.62-- Collectivism-individual 3 .67 .04 -.05 -.82 Collectivism-group - .03 - .47 Training condition .19 3.08-- Collectivism-individual x training 4 .68 .01 .29 1.81 Collectivism-group x training .28 .78

Panel B: Performance, Time 1

Age 1 .16 .16 .04 .38 Education -.17 -1.88 Company size .21 2.27- Gender .24 2.60-- Baseline performance .14 1.45 Collectivism-individual 2 .21 .05 -.08 -.81 Collectivism-group .01 .01 Training condition .20 2.12- Collectivism-individual x training 3 .56 .35 .88 8.47-- Collectivism-group x training 1.51 2.50-- Effort (time 1) 4 .68 .12 .47 .64 Self-efficacy (time 1) .53 4.60--

Panel C: Performance, Time 2

Age 1 .13 .13 .01 .14 Education -.10 -1.05 Tenure .17 1.78 Gender .21 2.28- Baseline performance .19 1.92 Effort (time 2) 2 .68 .55 .22 2.64-- Self-efficacy (time 2) .60 7.20-- Collectivism-individual 3 .72 .04 -.06 -1.05 Collectivism-group -.11 -1.90 Training condition .18 3.1 0- Collectivism-individual x training 4 .72 .01 .20 1.57 Collectivism-group x training .39 .74

Panel D: Performance, Time 2

Age 1 .13 .13 .01 .14 Education -.10 -1.05 Tenure .17 1.78 Gender .21 2.28- Baseline performance .19 1.92 Collectivism-individual 2 .17 .04 - .02 - .23 Collectivism-group - .09 - .91 Training condition .18 1.87 Collectivism-individual x training 3 .55 .38 .89 8.50-- Collectivism-group x training 2.07 3.38-- Effort (time 2) 4 .73 .18 .24 2.97"- Self-efficacy (time 2) .48 4.64"-

*p < .05; *p < .01.

111 /ASQ, March 1 994



significantly higher in the individual-focused training condition than in the group-focused training condition for U.S. managers, whereas self-efficacy and effort were significantly higher in the group-focused training condition than in the individual-focused training condition for managers from the People's Republic of China.

Discussion

People who were high in collectivism-group (Chinese service representatives) responded best to group-based training information, whereas people who were low in collectivism- group (American service representatives) responded similarly to individual-focused and group-focused training information. At the individual level of analysis, people high on individualism-collectivism responded better to group-focused than individual-focused training. American employees low on individualism-collectivism responded better to individual- focused than group-focused training, whereas Chinese employees low on individualism-collectivism responded similarly to the two forms of training. Regardless of nationality, people high on individualism-collectivism responded better to group-focused than individual-focused job training.

GENERAL DISCUSSION

This study focused on self-efficacy and its relationship to training. More specifically, I compared the impact of individual versus group-focused training on self-efficacy, effort, and task performance in two intercultural studies conducted across the cultural dimension of individualism and collectivism. The results show that an employee's cultural orientation influences his or her use of training information. For the collectivist, training focused on individual-level cues and information was less effective in enhancing his or her efficacy expectations, effort, and performance than was training based on group-level information. An individualist is best trained by targeting that employee's personal actions and potential. Finally, I found support for a mediating model of effort and self-efficacy in predicting the effects of individualism and collectivism, training, and their interaction on performance.

Perhaps the most significant finding from these studies is that individualism-collectivism is relevant in understanding how training influences self-efficacy. Although Bandura (1986) posited that efficacy is shaped through maturation and socialization experiences, little direct evidence exists connecting cultural values to self-efficacy. Consistent with the efficacy literature (e.g., Locke et al., 1984; Gist, 1987; Gist, Schwoerer, and Rosen, 1989), I found that a person's self-efficacy concerning work performance was influenced by task training. What was unclear prior to these studies is that cultural values moderate the impact of training on performance. I found that individualists performed best when exposed to training focused at an individual level, whereas collectivists performed best when exposed to training focused at a group level. Also, I found that training information provided at either a group or individual level was better than no training at all for everyone. This suggests




2

I would like to thank one of the reviewers for this suggestion.

Self or Group?

that self-efficacy is established and maintained through multiple sources, consistent with Triandis's (1989) probability-sampling argument concerning the public, private, and collective selves.

There is another possible interpretation of these findings.2 In the People's Republic of China, workers are often assigned to their work units (danwe,) shortly after receiving their formal education, and they remain in these units for a long time. These units play a significant role in a worker's life, both during and after normal work hours, and a worker's rewards are often tied to the work unit's achievements. In addition, alternative employment opportunities are unusual, despite economic reforms (Laaksonen, 1988), voluntary job turnover is low, and while employees may request reassignment from their company to another one, such reassignments are not typical. Workers' long-term rewards are thus linked closely to their danwei's successes. This is not at all characteristic of the American workplace. Thus, Chinese employees may have responded well to group-focused training because their danwei is the source of their rewards and successes, and managers were maximizing their potential rewards in the existing incentive structures. This suggests that people's specific knowledge of how their danwei relates to reward allocation was responsible for the observed interaction of training with individualism-collectivism. There are two limits to this explanation: First, the results from Study 1 for samples from Hong Kong and the People's Republic of China were similar even though the samples differed greatly from one another in organizational reward practices and economic systems; and, second, the results at the individual and cultural levels were similar. This suggests that the results are not merely an artifact of the different work and reward structures. Within each cultural group, individuals responded better to training congruent with their level of individualism- collectivism. If the findings from Mainland China were simply attributable to anticipated rewards from one's danwe', I would not expect to see the differences reported in Tables 3 and 7 for low and high levels of individualism-collectivism within each country. This alternative explanation merits further exploration, however, given that Study 1 revealed stronger differences as a function of training condition between managers from the People's Republic of China and those from Hong Kong.

In this paper, I introduced a methodological refinement to separate the cultural-level from the individual-level aspects of a cultural dimension. I separated the individualism- collectivism variable into two components, with the first representing a shared value, or mean level, representing "culture." This partitioning procedure is not merely the creation of a dummy-coded, country variable (except in the two-sample case), as it captures the relative distances among countries on a given cultural dimension. This procedure uses a specific cultural dimension (rather than using country as a surrogate for culture), and it helps us scale the relative magnitude of a sample's "culture." Theoretically, however, this approach raises a number of issues. For instance, this approach uses country of origin-as




a defining characteristic of a cultural grouping, defining a cultural group by a mean score for a value based on a respondent's nationality. In the definitions of culture often cited by researchers (Hofstede, 1991; Erez and Earley, 1993; Triandis, 1994), culture refers to "shared values and meaning systems." This suggests that researchers need to consider what constitutes a cultural grouping. For example, must members in the group be in direct contact with one another? Are there constraints on how "shared" values are shared? and Can a national designation capture a cultural grouping? Also, simply using a mean score to represent each cultural group may mask the variability that one might expect to observe when crossing from an individualistic to a collectivistic culture (Earley and Mosakowski, 1995). Future work should be directed at defining the boundaries and membership of a cultural group as well as the most effective way to represent a cultural group's "values." Triandis (1989) explicitly distinguished between the societal and the individual level of analysis. At the individual level, he described the counterparts of individualism and collectivism as idiocentrism and allocentrism, respectively. He argued that within any society, people vary in their beliefs about a cultural dimension, so that a member of a collectivistic culture may endorse individualistic values and beliefs. This distinction among levels is captured in Hofstede's discussion of an "ecological," or country-level analysis, as well as in the work of other researchers who deal with multiple levels of analysis (House and Rousseau, 1990). In my studies, the differences between the cultural and individual levels were neither consistent nor strong, suggesting that the dual assessment of individualism-collectivism tapped parallel constructs. While the individual-level assessment appeared to have the strongest relationship to self-efficacy and performance, this may simply reflect its proximity to the dependent variables (same level of analysis and measurement). The findings have a number of important implications for a managerial context. The most obvious is that training should be congruent with a person's cultural background as well as with individual experiences. This does not mean that a manager must rely on a person's cultural background alone; rather, it requires a manager to recognize intracultural variation as well. Within any given national boundary, there are many subcultures and many individual deviations within a given subculture. The managerial challenge arises from getting to know each employee's values and beliefs as they are shaped by culture and by individual experiences.

Another important implication of these studies that has been stated by others (e.g., Wagner and Moch, 1986; Boyacigiller, and Adler, 1991; Hofstede, 1991) is that our theories have significant cultural limits that must be understood. Although Bandura's self-efficacy idea was supported in these studies, the impact of training and self-efficacy was neither uniform nor simple. Self-efficacy is influenced by different sources of information that are more or less persuasive depending on a person's cultural values. These findings suggest that a cultural contingency approach is needed for subsequent research on self-efficacy.

114/ASQ, March 1994



Self or Group?

The results of these studies clearly show that organizational training that is culturally misdirected is also misguided. A collectivist's self-efficacy is based on information that he or she gets about a work group, whereas an individualist's self-efficacy comes from self-referenced cues. The implications of this research do not simply end with national or cultural sociotypes, however; individual variation within a social system leads to very different implications for the general approach that a manager adopts for training in an organization. This means that it is no longer meaningful to talk about comparative analyses and prescriptions for managerial actions, because this approach blurs the unique differences among individuals within a given culture. Likewise, ethnographically "rich" descriptions of cultural groups are somewhat misguided, since they also blur individual patterns by falsely assuming that the depth of inquiry supplants the shallowness of a more positivist analysis. The key point is that to understand managing in an intercultural context requires a depth of understanding at both the cultural and individual levels.

REFERENCES

Bandura, Albert 1986 Social Foundations of

Thoughts and Action: A Social Cognitive Theory. Englewood Cliffs, NJ: Prentice-Hall.

Baumeister, Richard F. 1986 Identity: Cultural Change and

the Struggle for Self. New York: Oxford University Press.

Bazerman, Max H. 1990 Judgment in Managerial

Decision Making, 2d ed. New York: Wiley.

Boisot, Max, and John Child 1988 "The iron law of fiefs:

Bureaucratic failure and the problem of governance in the Chinese economic reforms." Administrative Science Quarterly, 33: 507-527.

Bond, Michael H. 1988 "Invitation to a wedding:

Chinese values and global economic growth." In P. Sinha and H. Kao (eds.), Social Values and Development: 197-209. New Delhi: Sage.

Bond, Michael H., K. Wan, Kwok Leung, and R. A. Giacalone 1985 "How are responses to verbal

insult related to cultural collectivism and power distance?" Journal of Cross-Cultural Psychology, 16: 111-127.

Boyacigiller, Nakiye, and Nancy J. Adler 1991 "The parochial dinosaur:

Organizational science in a global context." Academy of Management Review, 16: 262-290.

Brislin, Richard W. 1980 "Translation and content

analysis of oral and written materials." In H. C. Triandis and J. W. Berry (eds.), Handbook of Cross-Cultural Psychology: 398-444. Boston: Allyn and Bacon.

Cohen, Jacob, and Patricia Cohen 1975 Applied Multiple

Regression/Correlation Analysis for the Behavioral Sciences. Hillsdale, NJ: Wiley.

Earley, P. Christopher 1989 "Social loafing and

collectivism: A comparison of the United States and the People's Republic of China." Administrative Science Quarterly, 34: 565-581.

1993 "East meets West meets Mideast: Further explorations of collectivistic and individualistic work groups." Academy of Management Journal, 36: 319-348.

Earley, P. Christopher, and Miriam Erez 1991 "Time dependency effects of

goals and norms: An examination of alternative methods to influence performance." Journal of Applied Psychology, 76: 717-724.

Earley, P. Christopher, and Elaine Mosakowski 1995 "A framework for

understanding experimental research in an international and intercultural context." In B. J. Punnett and 0. Shenkar (eds.), Handbook of International Management Research. London: Blackwell (forthcoming).

Epstein, Seymour 1973 "The self-concept revisited, or

a theory of a theory." American Psychologist, 28: 408-416.

Erez, Miriam, and P. Christopher Earley 1993 Culture, Self-identity, and

Work. New York: Oxford University Press.

Erez, Miriam, P. Christopher Earley, and Charles L. Hulin 1985 "The impact of participation

on goal acceptance and performance: A two-step model." Academy of Management Journal, 28: 50-66.

Garland, Howard 1985 "A cognitive mediation theory

of task goals and human performance." Motivation and Emotion, 9: 345-367.

Garland, Howard, and John H. Adkinson 1987 "Standards, persuasion, and

performance." Group and Organization Studies, 12: 208-220.

115/ASQ, March 1994



Gist, Marilyn E. 1987 "Self-efficacy: Implications for

organizational behavior and human resource management." Academy of Management Review, 12: 472-485.

Gist, Marilyn E., and Terence R. Mitchell 1992 "Self-efficacy: A theoretical

analysis of its determinants and malleability." Academy of Management Review, 17: 183-211.

Gist, Marilyn E., Catherine Schwoerer, and Benson Rosen 1989 "Effects of alternative training

methods on self-efficacy and performance in computer software training." Journal of Applied Psychology, 74: 884-891.

Hinrichs, John R. 1976 "Personnel training." In M. D.

Dunnette (ed.), Handbook of Industrial and Organizational Psychology: 829-861. Chicago: Rand McNally.

Hofstede, Geert 1980 Culture's Consequences:

International Differences in Work Related Values. Newbury Park, CA: Sage.

1991 Culture and Organizations: Software of the Mind. London: McGraw-Hill.

House, Robert J., and Denise Rousseau 1990 "On the bifurcation of OB or if

it ain't meso it ain't OB." Working paper, the Wharton School, University of Pennsylvania.

Hsu, Francis L. K. 1985 "The self in cross-cultural

perspective." In A. Marsella, G. DeVos, and F. L. K. Hsu (eds.), Culture and Self: Asian and Western Perspectives: 24-55. New York: Tavistock.

Kluckhohn, Florence, and Frederick Strodtbeck 1961 Variations in Value Orientation.

Westport, CT: Greenwood Press.

Laaksonen, Oiva 1988 Management in China: During

and after Mao. Berlin: de Gruyter.

Li, D. J. 1978 The Ageless Chinese. New

York: Scribner's.

Lindsay, Cindy P. 1983 "China: Motivational systems

in flux." In Richard M. Steers and Lyman W. Porter (eds.), Motivation and Work Behavior: 623-634. New York: McGraw-Hill.

Locke, Edwin A., Elizabeth Frederick, Cynthia Lee, and Philip Bobko 1984 "Effects of self-efficacy,

goals, and task strategies on task performance." Journal of Applied Psychology, 69: 241-251.

Markus, Hazel, and S. Kitayama 1991 "Culture and the self:

Implications for cognition, emotion, and motivation." Psychological Review, 98: 224-253.

Meyer, John P., and Ian R. Gellatly 1988 "Perceived performance norm

as a mediator in the effect of assigned goal on personal goal and task performance." Journal of Applied Psychology, 73: 410-420.

Rokeach, M. 1973 The Nature of Human Values.

New York: Free Press.

Schwartz, Shalom H. 1993 "Cultural dimensions of

values: Toward an understanding of national differences." Unpublished paper, Department of Psychology, Hebrew University.

Triandis, Harry C. 1988 "Collectivism vs.

individualism: A reconceptualization of a basic concept in cross-cultural social psychology." In G. K. Verma and C. Bagley (eds.), Cross-Cultural Studies of Personality, Attitudes and Cognition: 60-95. New York: St. Martin's.

1989 "The self and social behavior in differing cultural contexts." Psychological Review, 96: 506-520.

1994 "Culture: Theoretical and methodological issues." In M. D. Dunnette and L. Hough (eds.), Handbook of Industrial and Organizational Psychology, 2d ed., vol. 4. Palo Alto, CA: Consulting Psychologists Press (forthcoming).

Wagner, John A., Ill, and Michael K. Moch 1986 "Individualism-collectivism:

Concept and measure." Group and Organization Studies, 11: 280-304.

Wood, Robert E., and Albert Bandura 1989 "Social cognitive theory of

organizational management." Academy of Management Review, 14: 361-384.

APPENDIX: Correlations

Table A.1

Pearson Correlations for Study 1

Variable 2 3 4 5 6 7 8 9 10

1. Age -.21* -.12 -.07 .07 .07 .10 -.03 .07 .05 2. Education - -.01 .06 .10 .09 .03 -.02 .04 -.21 3. Company size - .13- .09 .02 .11 .07 -.12 .07 4. Gender - .07 .15- .07 -.02 -.01 -.01 5. Performance - .70- .75- -.01 .63- .13- 6. Effort - .53- -.03 .48- .09 7. Self-efficacy - -.04 .46- .14- 8. Individualism-

collectivism - -.01 .48- 9. Training - .11

10. Country

* p < .05.

116/ASQ, March 1994



Self or Group?

Table A.2

Pearson Correlations for Study 2

Variable 2 3 4 5 6 7 8 9 10 11 12 13 14

1. Age -.10 -.15 .05 .23- .06 .05 -.12 -.07 .05 .04 - .24 -.08 -.02 2. Education - .12 -.02 -.17 -.18 -.12 -.06 -.06 -.13 -.10 .06 -.19 -.01 3. Tenure - -.01 .04 .19 .16 .13 .15 .16 .16 -.29- .03 -.12 4. Gender - .15 .26- .24- .15 .22- .11 .13 .02 -.05 -.13 5. Baseline performance - .22- .24 .14 .10 .18 .16 - .07 .13 .02 6. Performance (time 1) - .83- .51 .62- .76- .75- -.14 .24- -.01 7. Performance (time 2) - .52- .68- .76- .79- -.1 1 .19 -.09 8. Effort (time 1) - .62- .67- .65- .07 .01 .07 9. Effort (time 2) - .69- .72- .07 .01 -.02

10. Self-efficacy (time 1) - .90- -.08 .06 .01 11. Self-efficacy (time 2) - .03 .05 -.01 12. Individualism-collectivism - -.03 .300 13. Training - .21 14. Country

* p < .05.

117/ASQ, March 1994



self or group? cultural effects of training on self-efficacy and performance

Documents