Abstract
This article reports two studies aimed at developing a theory-based multidimensional measure of workgroup emotional climate (WEC) and exploring its relationship with workgroup effectiveness. In Study 1, a four-factor theory of WEC is derived from a review of the climate and emotions literature, followed by the use of both qualitative and quantitative methods to operationalize the model and develop a survey measure. An initial sample of 396 workgroup members provided the data for the exploratory factor analysis and item response theory analysis. Confirmatory factor analysis (CFA) of data from another sample of 334 workgroup members further confirmed the proposed structure. In Study 2, multilevel CFA of data collected from 840 workgroup members from 148 workgroups provided construct, consensual, and discriminant validity. We also examined the relationship between WEC and workgroup effectiveness. The resulting four-factor, 16-item measure demonstrated robust psychometric properties, with acceptable levels of reliability and validity.
Introduction
The past three decades have seen a dramatic increase in the use of teams in organizations. Related to this, Elfenbein and Shirako (2006) argued that emotion is a particularly important concept for teams. The relevance and centrality of emotion to group life becomes evident when considering that many human emotions grow out of social interactions (Barsade & Gibson, 1998). Knowing how emotions affect behavior in groups therefore is useful for understanding and predicting workgroup behavior (Druskat & Wolff, 2001).
Following Barsade and Gibson (1998), we use the terms “affect” and “emotion” interchangeably in this article, as we conceive both as semantically similar terms for the general constellation of individuals’ feeling responses. We also use the words “team” and “workgroup” interchangeably in this article (Cohen & Bailey, 1997) as our interest is not in separating the study of the two. Groups vary in their degree of “groupness,” with some groups being more interdependent and integrated than others. Some authors have used the label “team” for groups that develop a high degree of “groupness” (e.g., Katzenbach & Smith, 1993). Barrick, Bradley, and Colbert (2007) suggested that a team with high interdependence is a real team, whereas a team with low interdependence is in fact a workgroup. The terms workgroup and teams are used in this article to refer to an organizationally formed unit of multiple individuals for the purpose of performing some organizationally relevant task functions. These individuals interact, exhibit task interdependence, possess one or more shared goals, and are embedded in a larger organizational setting (Kozlowski, Gully, Salas, & Cannon-Bowers, 1996).
While there is a prolific research literature on the question of what makes teams effective (Guzzo & Dickson, 1996), few scholars have taken the emotional aspect of workgroups into consideration when addressing this question (Pirola-Merlo, Härtel, Mann, & Hirst, 2002). However, at the organizational level, Carr, Schmidt, Ford, and DeShon’s (2003) meta-review of more than 50 studies concluded that the affective facet of organizational climate was more strongly correlated with organizational members’ psychological well-being and performance than the cognitive or instrumental elements of organizational climate. Consequently, recent attention has turned to the study of workgroup emotional climate (WEC) or the perceptions of affect and affective exchanges that typify a workgroup (cf. De Rivera, 1992; Gamero, González-Romá, & Peiró, 2008; Pirola-Merlo et al., 2002; Tse, Dasborough, & Ashkanasy, 2008). Such perceptions are considered to have important consequences for group members (Härtel, Gough, & Härtel, 2006, 2008).
Although there are a few studies aimed at investigating WEC (e.g., Härtel et al., 2006; Pirola-Merlo et al., 2002), a rigorous structure and measure of WEC is still lacking, limiting the development of the theory of WEC. For example, due to the lack of an established measure of WEC, Pirola-Merlo et al. (2002) used the Team Climate Inventory, a measure of climate for workgroup innovation, as a proxy to measure team affective climate in their study. Gamero et al. (2008) aggregated team members’ affective well-being to measure team affective climate. Tse et al. (2008) measured team affective climate using a five-item positive group perception scale. Menges, Walter, Vogel, and Bruch (2011) measured organizations’ positive affective climate using four positive affect items adapted from a job-related affective well-being scale. Härtel et al. (2006) developed a Workgroup Emotional Climate Scale (WECS), but this measure is unidimensional with five items and assesses only positive WEC. A further limitation of the existing body of group-level emotion studies is the sampling exclusively from Western cultural groups, meaning the applicability of these measures to other cultures remains untested (Elfenbein & Shirako, 2006).
The purpose of this article is to present the results of a study undertaken to develop a theory-based multidimensional measure of WEC and establish its predictive validity in China. Specifically, in Study 1, we articulate the theoretical model of WEC based on a review of the organizational climate and emotions literature and use both qualitative and quantitative methodological approaches to operationalize the model into a valid survey measure, referred to as the WECS, which comprises two dimensions, a valence (positive–negative) and an interpersonal dimension (ego-focused and other-focused). The advantages of this structure are that it both reflects the universal and social nature of the emotional climate in workgroups and allows for cross-cultural comparison of WEC among workgroups in different cultures. This point is elaborated in an article by Härtel and Liu (2012), which provides the theoretical basis to expect WECs to be more other-focused in Eastern cultures than Western cultures, and more ego-focused in Western cultures than Eastern cultures. In Study 2, we further test the validity of the measure and establish its predictive validity by exploring its relationship with workgroup effectiveness.
Theoretical Background
Climate: Level of analysis and facet
Climate has been studied at different levels, such as group climate and organizational climate (Härtel & Ashkanasy, 2011). James (1982) proposed a composition theory for climate, in which it was suggested that the unit of theory (i.e., the unit on which a theory is based) for climate is the individual, but that the aggregation of individual climate perceptions (i.e., psychological climate) can serve as a powerful explanatory tool of higher levels of analysis. Shared perceptual agreement at the individual level of analysis in climate research provides the meaning of the construct at a higher level (e.g., group level, organizational level) of analysis (Härtel & Ashkanasy, 2011). Accordingly, climate research is described as composition oriented (D. M. Rousseau, 1985).
Organizational climate has come to be viewed as a logical extension of psychological climate and is now generally defined as shared psychological meanings (i.e., shared psychological climates; James et al., 2008). Although emotional climate can be conceptualized at the organizational level (Ashkanasy & Humphrey, 2011; Yurtsever & De Rivera, 2010), some scholars (e.g., Anderson & West, 1998; Dansereau & Alutto, 1990) point out that it is unlikely that shared climates exist at the overarching level of the organization in its entirety, particularly where the organization is large, with many divisions and layers. Ashkanasy and Nicholson (2003) empirically examined climate in two Australian restaurant chains and found that climate varied across the restaurants within the organizations, but not across the organizations and concluded that climate is a team-level phenomenon.
Workgroup climate perceptions are the shared perceptions of employees about their own workgroup, and evidence suggests that they are important predictors of variability in performance within an organization as well as the role behavior of group members (Zohar & Luria, 2005). Individuals are generally more likely to identify with their proximal workgroup than with organizations, and moreover, shared patterns of understanding and norms of behavior are most likely to develop at this level, allowing the opportunity for a shared climate to evolve (Campion, Medsker, & Higgs, 1993). For these reasons, this study focuses on proximal workgroup climate and develops a measure of WEC specifically at the group level of analysis.
Climate has been studied by focusing on different facets such as climate for safety (Zohar & Luria, 2005), climate for innovation (Anderson & West, 1998), climate for service (De Jong, De Ruyter, & Lemmink, 2005), climate of diversity (Chen, Liu, & Portnoy, 2012), and climate of fear (Ashkanasy & Nicholson, 2003). Meanwhile, a number of studies have explored traditionally individual-level phenomenon as a group-level construct, such as justice climate (Li & Cropanzano, 2009) and empowerment climate (Siebert, Silver, & Randolph, 2004). Some scholars (e.g., D. M. Rousseau, 1988) argue that it is meaningless to apply the concept of climate without a particular referent (e.g., climate for change, climate for quality, climate for innovation, etc.) and advocate the study of “facet-specific climates.” Following in this vein, Ostroff (1993) argued that affective responses to relationships with co-workers and groups in an organization are an important dimension of organizational climate perceptions. De Rivera (1992) also identified an affective aspect to climate, defining the emotional climate of the nation-state as reflecting how individuals think the majority of others are feeling about a society’s current situation and how the people of a society emotionally relate to one another. Our focus in this study is on the emotional facet of workgroup climate referred hereon as WEC and defined as the shared perceptual agreement about the emotions shared and the emotional relationships in a workgroup (Härtel et al., 2006).
WEC
Workgroups have been conceptualized as social entities that, over time, develop a history of shared experiences or events (Härtel, Härtel, & Barney, 1998). According to affective events theory (AET; Weiss & Cropanzano, 1996), these experiences or events can elicit emotional reactions in the workgroup that have consequences for the attitudes and behaviors of the workgroup. Some scholars have also found that workgroup members develop shared emotions when they are performing their task (e.g., Barsade & Gibson, 1998). For example, Barsade and Gibson (1998) argued that workgroup members may share group emotions as they perform their task and that these emotions may result from a subtle but continuous transfer of affective states among members. Barsade (2002) also investigated group emotional contagion, the transfer of moods among people in a group, and its influence on workgroup dynamics. Similarly, research on the emotional labor of service workers illustrates how individuals can infect others with emotion to create particular affective climates that promote group or organizational objectives (Sutton, 1991).
According to Kelly and Barsade (2001), group members bring their individual-level emotional experiences, such as dispositional affect, moods, emotions, emotional intelligence, and sentiments, with them to a group interaction, and through a variety of explicit and implicit processes, these affective inputs are communicated to other group members and form the affective compositional group effects. The combination of the affective context (e.g., organizational emotion norms, local group norms, and the group’s emotional history), non-affective context, and the group’s affective composition leads to group emotion. In this study, we define WEC as the shared perceptions of emotions and emotional exchanges that typify a workgroup (cf. De Rivera, 1992), and such shared perceptions are considered to have important consequences for group members (Härtel et al., 2006).
Affective group tone is similar conceptually to WEC; however, they are not synonymous (Pirola-Merlo et al., 2002). Actually, the two constructs are conceptualized and operationalized differently. George (1990) defined group affective tone in terms of homogeneity in individual job affect, whereas WEC is defined in terms of team members’ shared perceptions of emotions and emotional exchanges that typify a workgroup. Consequently, group affective tone has always been operationalized from individuals’ ratings of their own positive and negative affect, whereas WEC has been operationalized by aggregating individuals’ ratings of the perceptions of the group’s emotions and emotional exchanges that typify a workgroup.
According to Chan’s (1998) five types of composition models, the composition models for group affective tone and WEC are different: Group affective tone is a direct consensus model, in which the construct conceptualized and operationalized at the lower level is functionally isomorphic to another form of the construct at the higher level, while WEC is a referent-shift consensus model, in which the lower level attributes being assessed for consensus are conceptually distinct though derived from the original individual-level construct. The critical difference between the two forms of composition is that in referent-shift consensus composition, there is a shift in the referent prior to consensus assessment, and it is the new referent (workgroup here) that is actually being combined to represent the higher level construct (Chan, 1998).
Dimensionality of WEC
On the classification of emotions, a number of researchers have tested a multidimensional structure underlying emotional states (Russell, 1980). These studies typically find two highly robust dimensions: positive versus negative evaluation (or pleasantness) and activation. Yet another replicable dimension often emerges: social-engagement-disengagement (Kitayama, Markus, & Matsumoto, 1995). In this study of WEC, we use the dimensions of valence and interpersonal to define it. We expand our discussion of these two dimensions next.
Valence (pleasantness)
An important dimension of subjective experience is that of valence: pleasure and displeasure (often referred to in classifications as positive and negative; Barrett & Russell, 1999). According to Russell (1991), hedonic quality is a universal aspect of affective experience. All languages appear to have words to distinguish between pleasure and displeasure (Wierzbicka, 1992), and valence is an organizing dimension of many emotion lexicons (Russell, 1983), though cultures appear to differ with regard to the preferred state on that dimension (Tsai, Knutson, & Fung, 2006). Watson and colleagues refined the valence concept, concluding that both state and trait affect are comprised of two distinct continua, rather than a single continuum ranging from negative to positive (Watson & Clark, 1984; Watson & Tellegen, 1985).
Interpersonal dimension
Emotions are not just private or personal bodily states; they also are social phenomena (e.g., De Rivera & Grinkis, 1986). As such, emotions can be differentiated as other-focused (emotions associated with interpersonal engagement) and ego-focused (emotions associated with private states and interpersonal disengagement). The ego-focused emotions such as anger, frustration, and pride have the individual’s internal attributes (his or her own needs, goals, desires, or abilities) as the primary referent, and thus foster and create independence (Markus & Kitayama, 1991). The other-focused emotions such as sympathy or love have another person, rather than one’s own internal attributes, as the primary referent and thus highlight and foster one’s interdependence (Markus & Kitayama, 1991). Other-focused emotions also focus people’s attention on social worth and belonging to self-other relationships (Mesquita, 2001). This dimension is also consistent with Ortony, Clore, and Collins’s (1988) cognitive structure of emotions, in which event-based emotional experiences are categorized as fortunes-of-others’ emotions, which represent the reactions to events that a person can have when the events are desirable or undesirable relative to the goals and interests of another person and fortunes-of-self emotions, which represent a person’s reactions to the consequences of events with respect to himself or herself.
There is considerable evidence of the validity and value of conceptualizing the social dimension of emotion (Kitayama & Markus, 1994). Many studies suggest that emotions vary in the degree to which they either engage or disengage one in social/interpersonal relationships (e.g., Elfenbein & Shirako, 2006). Evidence for the social or interpersonal dimension of emotion in both Western and non-Western countries is provided by many scholars (Kitayama et al., 1995; Markus & Kitayama, 1991; Mesquita, 2001; Triandis & Suh, 2002), although it is variously interpreted and named (e.g., “interpersonal relatedness,” Block, 1957; “trustful versus untrustful,” Dittman, 1972). De Rivera (1992) also classified the nation’s emotional climate using two dimensions: one relates to others, including fear (isolation), security (trust), hostility (polarization), solidarity (willingness to sacrifice), and the other relates to ideals, including dissatisfaction, satisfaction, despair, hope, depression, confidence, stability, and instability.
In our study, considering the characteristics of Chinese emotional experience and expression, we do not employ the dimension of arousal or activation into the structure of WEC, but rather the interpersonal dimension. There are two key reasons for doing so. First, as workgroups are settings that require much interpersonal interaction, it is necessary and appropriate to take this interpersonal dimension into account in structuring the construct of WEC. Second, the arousal–valence relationship varies more within individuals than between individuals and a primary explanation for this within-person variation is culture (Kuppens, Tuerlinckx, Russell, & Barrett, 2013). The cultural basis for the different structure between East and West is attributed to individualistic societies placing value on a strong positive emotional state (high positive arousal) and collectivistic societies placing value on low intensity emotions and a state of balance of negative and positive emotions (i.e., low arousal) that reflect harmony, balance, and serenity (Tsai et al., 2006). By replacing the arousal dimension of emotional climate with the interpersonal dimension of emotional climate, we avoid the problem of incomparability of valence–arousal structures across cultures, while reflecting the universal structure of affect that will enable cross-cultural comparison of WEC among workgroups in different cultures.
Based on the above-discussed classifications of emotion and emotional climate, we suggest that the valence (positive–negative) and the interpersonal dimension (the extent to which the person is engaged in or disengaged from an interpersonal relationship) can be jointly used to classify WEC. Consequently, in this study, the dimensionality of WEC could be classified as four specific dimensions: ego-focused and negative WEC (EN), ego-focused and positive WEC (EP), other-focused and negative WEC (ON), and other-focused and positive WEC (OP). This classification recognizes that some shared emotions in a workgroup are positive and ego-focused (e.g., pride, happy), some shared emotions are equally positive but other-focused (e.g., friendly feelings, feelings of closeness, feelings of respect), some negative shared emotions are ego-focused (e.g., unhappy, depressed, hopeless), and some negative shared emotions are other-focused (e.g., fear, feelings of hostility, jealousy).
Study 1: Item Generation and Scale Development
The purpose of Study 1 was to develop a four-dimensional measure of WEC based on the theory-based model we proposed, using item response theory (IRT) analysis to provide a detailed examination of the psychometric properties of the scale, and to confirm the dimensionality across the sample.
Method
Item generation
As WEC refers to the shared perceptions of emotions and emotional exchanges that typify a workgroup, in developing our measure of WEC we employed the reflective indicator model, viewing the measures as “reflections of” or “effects of” the underlying latent construct (cf. MacCallum & Browne, 1993).
In Step 1, an extensive review of the published measures of climate and affect was conducted. These measures were examined for their component subdimensions in relation to the posited four-factor emotional climate model. Only subscales or items appropriate to these factors were retained.
In Step 2, we conducted in-depth interviews with workgroup members and leaders to generate items to capture the construct. Ten workgroup leaders and 20 workgroup members working in airline profit management workgroups of a state-owned airline company were interviewed. The interviews were one-on-one and semi-structured. The time taken to conduct each interview was between 0.5 hr and 2 hr. Interviewees were given a definition and classification of WEC and then asked to describe the WEC according to their experience and observation. All interviews were recorded and transcribed verbatim. Following this, three PhD students majoring in organizational behavior (OB)/human resource management (HRM) sorted the results from the completed interview transcripts. Three types of items were deleted from the item pool based on the judgment of the authors. These were as follows: (a) overlapping or similar items suggested by different respondents, (b) items with unclear meaning, and (c) items that did not match the definition of WEC. Based on the above two steps, a list of 100 items was produced after coding and classifying.
In Step 3, focus groups were organized. The focus groups were composed of three scholars in the OB field, two workgroup leaders, four workgroup members, and the authors. The focus group participants discussed each item on the list and subsequently combined the same meaning items and deleted the items which were inappropriate, unimportant, or inaccurate. The focus group went through three rounds for screening all the items and sorting the items into the hypothesized four dimensions. Subsequent items were not addressed until there was agreement from a majority (75%) of participants in each round. The final list consisted of 60 items of WEC. The items were sorted into the hypothesized four dimensions with 15 items on OP, 12 items on EP, 19 items on ON, and 14 items on EN.
In Step 4, the 60-item, four-dimension original version of the WECS was piloted with 10 workgroup members working in profit management workgroups in an airline company to obtain reactions and comments on the measure. No data were collected as part of this pilot; the purpose was simply to evaluate the face validity and acceptability of the measure to respondents. Based on the results, we reworded and deleted some items, which led to a 58-item questionnaire. Respondents were asked to indicate the extent to which each statement was true of their workgroup on a 5-point scale ranging from 1 = not at all to 5 = completely.
Sample for item selection
The 58-item, four-factor original version of the WEC questionnaire was piloted on 396 workgroup members employed in government and schools in China. The first author hand-delivered surveys to participants during their non-peak work hours. She explained the objective of the research to participants and confidentiality was guaranteed. Participants completed the questionnaires on-site and returned them to the researcher immediately. Among the 396 workgroup members, 49% were female; average job tenure was 14.44 years; the age ranged from 20 to 58 years; 9.1% held a high school degree or below, 47.9% held a junior college degree, and 43% held a bachelor’s degree or above.
Results of item selection and factorial structure
The major purpose of the first study was to develop a psychometrically sound WEC questionnaire. As a result, the first task was to test the factorial structure of the instrument. All the measures in our study were presented in Chinese. The recommended translations and back-translation procedures (Brislin, 1980) were used to translate the items into English for preparation of the manuscript.
An exploratory factor analysis (EFA) of the 58 items was conducted using the maximum likelihood method with Promax rotation in SPSS 16.0. Promax is an oblique rotation technique, which allows factors to be correlated (Ford, MacCallum, & Tait, 1986). The Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy was .89, and the Bartlett test of sphericity was significant at p < .001, indicating the suitability of the data for factor analytic procedures.
The decision of how many factors to retain in EFA is a critical component of EFA, and there are several methods developed for this decision, such as Kaiser or mineigen greater than 1 criterion (K1; Kaiser, 1960), Cattell’s (1966) scree test, parallel analysis (PA; Horn, 1965), and Velicer’s (1976) minimum average partial method (MAP). PA is regarded as the most accurate method for determining the number of factors to retain, followed closely by MAP, with K1 being extremely inaccurate and tending toward over factoring (e.g., Velicer, Eaton, & Fava, 2000; Zwick & Velicer, 1986). Thus, we conducted PA and MAP analysis in conjunction with the scree tests for our factor retention decision. We conducted PA with SPSS 16.0 following Hayton, Allen, and Scarpello’s (2004) tutorial and indicated a five-factor solution. The application of both the MAP analysis conducted with SPSS 16.0 following Connor’ guide and the scree test (Cattell, 1966) indicated a four-factor solution. Considering our theory and PA’s slight tendency to over factor by 1 (e.g., Zwick & Velicer, 1986), we set upon a four-factor solution. A further analysis was computed limiting the number of factors to four. After running the subsequent EFA that was limited to a four-factor solution, we selected the items based on two criteria: items loaded on their respective factor with loadings greater than .45 and cross-loadings below .30. Sixteen items remained and the four-factor solution accounted for 50.50% of the total variance. Four items were retained each for the dimensions of EN, EP, OP, and ON. As we assumed correlation among the factors, sums of squared loadings cannot be added to obtain a total variance. The factor loadings, cross-loadings, and eigenvalues, which come from the EFA carried out with the selected 16 items, are presented in Table 1.
Exploratory Factor Analysis of Workgroup Emotional Climate.
Note. Total variance accounted for 50.50%. Extraction method: Principal axis factoring. Rotation method: Promax with Kaiser normalization.
To summarize this four-factor solution, it is apparent that the underlying simple structure displays a fairly unambiguous pattern of item loadings, mostly in line with the postulated model of WEC. Table 2 presents means, standard deviations, intercorrelations, and Cronbach’s alpha coefficients for the four factors.
Descriptive Statistics, Reliabilities, and Correlations among the Four Factors (N = 396).
Note. EN = ego-focused and negative workgroup emotional climate; OP = other-focused and positive workgroup emotional climate; ON = other-focused and negative workgroup emotional climate; EP = ego-focused and positive workgroup emotional climate. Cronbach’s α reliability estimates are in parentheses along the diagonal.
p < .01 (two-tailed tests).
An IRT Analysis of the WECS Items
Although we have presented some information concerning WECS measurement properties (e.g., factor structure, reliability), our analysis relies on psychometric analyses grounded in classical test theory (CTT). As modern measurement theory, IRT has certain advantages over CTT (Hambleton, Swaminathan, & Rogers, 1991). For example, CTT cannot provide certain scale- and item-level information obtainable by IRT analyses (Embretson & Reise, 2000). Thus, to ascertain whether the WECS we just developed has a high and evenly distributed degree of measurement precision, we conducted an IRT analysis of the WECS to determine whether the WECS is adequate from an IRT perspective.
All IRT analyses were conducted with Multilog Version 6.0 (Thissen, 1991), a program designed to estimate a wide variety of item response models. The item parameters of Samejima’s (1969, 1996) GRM (Graded Response Model) were estimated for each 5-point Likert-type item used in this study. As the testable assumption of the GRM is that item covariation arises predominantly from a single underlying dimension, we estimated GRM item parameters for each of the 4 items within each subscale of the WECS (four subscales with 16 items in total). The GRM is a potentially useful item response model when item response options can be conceptualized as ordered categories (e.g., with Likert-type rating scales). In the GRM, each item has a single discrimination (α) value for all response options. Thus, an item rated on a 1-to-5 scale has four response dichotomies: (a) Category 1 versus Categories 2, 3, 4, and 5; (b) Categories 1 and 2 versus Categories 3, 4, and 5; (c) Categories 1, 2, and 3 versus Categories 4 and 5; and (d) Categories 1, 2, 3, and 4 versus Category 5. For the hypothetical 5-point item, we can generate four intercorrelation coefficients (ICCs). The response option difficulty represents a between-option “threshold” parameter (β). In the GRM, each item has a single discrimination (α) value for all response options and total test information of each subscale is obtained by summing the item information functions. Similarly, the conditional standard error of measurement for a given trait value equals the inverse square root of the information level at that theta value. As shown in Table 3, we estimated five parameters for each WECS item: one item-discrimination value and four item-difficulty or between-category threshold values. All the discrimination α values are above the .50 measurement criteria (Embretson & Reise, 2000; Fraley, Waller, & Brennan, 2000). Thus, the IRT analysis provided an in-depth examination of the psychometric properties of the WECS, and we can draw the conclusion that the WECS has good psychometric properties in terms of IRT estimates.
Item Response Theory Item Parameter Estimates for the 16-Item Workgroup Emotional Climate Scale.
Note. Items are sorted by their discrimination (α) values.
Cross-validation sample for factor structure assessment
Next, the reduced-length WECS was cross validated in a new data collection effort. There were two reasons for us to do this. First, Stanton, Sinar, Balzer, and Smith (2002) suggested that it is critical to field the new reduced-length scale without the discarded items, because item responses are highly dependent on the surrounding context within the survey instrument (e.g., Schwarz, 1999); Second, our goal is to ensure that the four-factor structure of the 16 WEC items can be generalized to other samples.
Sample
The sample consisted of 334 workgroup members working in teams of various kinds of firms. Participants were recruited employing a snowball technique, using the social networks of the researchers. Data collectors directly explained the objective of the research to participants and confidentiality was guaranteed. Respondents who agreed to participate completed the questionnaires online and the data came back to the researcher through the online data collection program. In all, 16.9% of the workgroup members belonged to the marketing workgroups, 51.1% belonged to the R&D workgroups, and 21.6% belonged to the service workgroups. The gender split was 52.9% female and 47.1% male; the age range was from 20 to 58; the average job tenure was 8.85 years; 6.3% held a high school degree or below, 18.0% of the sample held a junior college degree, 39.9% held a bachelor’s degree, 34.5% held a master’s degree, and 1.2% held a doctoral degree.
Results of the cross-validation sample
We conducted a confirmatory factor analysis (CFA) of the WEC items using LISREL 8.51 (Jöreskog & Sörbom, 1993). Table 4 presents the fit statistics associated with the four-factor baseline model, along with the six alternative models. We employed the maximum likelihood estimation method and tested the models using a variance–covariance matrix as input into the program. As shown in Table 4, the baseline four-factor model fitted the data well; the model χ2 of the CFA was χ 2 (98, N = 334) = 215.21. The standardized root mean square error of approximation (RMSEA), standardized root mean residual (SRMR), comparative fit index (CFI), goodness of fit index (GFI), incremental fit index (IFI), and non-normed fit index (NNFI) were .06, .05, .95, .93, .95, and .94 respectively. All the factor loadings were statistically significant (p < .01) and ranged from .55 to .83.
Comparison of Measurement Models.
Note. For all χ2, p < .01. RMSEA = root mean square error of approximation; SRMR = standardized root mean residual; CFI = comparative fit index; GFI = goodness-of-fit index; IFI = incremental fit index; NNFI = non-normed fit index; Δχ 2 = change in chi-square between the alternative model (Model A1, Model A2, Model B, Model C1, Model C2, Model D) and the baseline model; Δdf = change in degrees of freedom between alternative model (Model A1, Model A2, Model B, Model C1, Model C2, Model D) and the baseline model.
p < .05.
Against this baseline four-factor model, we tested six alternative models. Model A1 was a two-factor model where the WEC items of EN merged with ON to form a single factor of negative emotional climate and the items of EP merged with OP to form another single factor of positive emotional climate. Model A2 was another two-factor model where the items of EN merged with EP to form a single factor of ego-focused emotional climate and ON merged with OP to form another single factor of other-focused emotional climate. Model B was a one-factor model, with all items loading onto a single factor. The CFA results, shown in Table 4, suggested that the baseline four-factor model was superior to Models A1, A2, and B.
Furthermore, we fit the data to higher order models to compare. First, in Model C1, we fit the data to a higher order model with the two first-order factors of negative emotional climate and positive emotional climate, which are derived from the four second-order factors EN and ON, and EP and OP, respectively. This model fitted the data well with χ 2 (99, N = 334) = 217.22 and a standardized RMSEA, SRMR, CFI, GFI, IFI, NNFI of .06, .05, .95, .92, .95, .94, respectively. These results, as well as the chi-square difference test comparing the fit of the baseline model χ 2 Difference(1, N = 334) = 2.01, p > .05, indicate that Model C1 is indeed a more parsimonious model than the baseline four-factor model. All the factor loadings in Model C1 were statistically significant (p < .01) and ranged from .56 to .83. The correlation among the two first-order factors of negative emotional climate and positive emotional climate was −.56. To compare them, we also conducted a higher order Model C2 with the two first-order factors of ego-focused emotional climate and other-focused emotional climate, which are derived from the four second-order factors EN and EP, and ON and OP, respectively. Moreover, we also fit the data to a higher order model with the four second-order factors. These CFA results, shown in Table 4, suggest that the baseline model was superior to Model C2 and Model D.
According to the CFA result, the choice was made for two reasons to use Model C1, the higher order two-factor model with the two first-order factors of negative emotional climate and positive emotional climate. First, the valence of emotion is the most popular classification of emotions and hedonic quality is a universal aspect of affective experience (Russell, 1991). It thus makes sense to classify the dimensions of WEC based on the higher order of positive and negative emotional climate and then break them down into other-focused and ego-focused emotional climate, respectively. Second, the higher order model is more parsimonious than the four-factor model suggested by the CFA results. Overall, these results provide good support for the initial construct validity of this new measure.
Study 2: Scale Validation and the Relationship Between WEC and Workgroup Effectiveness
The purpose of Study 2 was to examine the validity of the newly developed measure and establish its predictive validity. First, we demonstrate that the factor structure of WEC is consistent at both the workgroup individual level and the workgroup level using multilevel CFA. Second, we demonstrate the consensual and discriminant validity of WECS to ensure that emotional climate can be aggregated at the group level, and to ensure that shared emotional climates do exist within workgroups. Finally, we establish its predictive validity and explore the relationship between WEC and workgroup effectiveness.
Theoretical Background and Hypotheses
Some scholars report findings showing that groups frequently converge in the tendencies of individuals to experience particular affective states and that such convergence generally has consequences for teams (e.g., Kelly & Barsade, 2001; Totterdell, 2000). According to Fredrickson’s (2001) broaden-and-build theory, positive emotions broaden individuals’ scope of attention, cognition, and action, and build enduring physical, intellectual, social, and psychological resources. Rhee (2007) suggested that broaden-and-build interactions in groups take three forms: cognitive broadening through building on each other’s ideas, morale-building communication, and active building of social resources, further justifying the relationship between positive group emotions and group effectiveness. At the organizational level, Patterson, Warr, and West (2004) found that a positive organizational climate was correlated with company productivity. Likewise, Dawson, González-Romá, Davis, and West (2008) found that a climate of positive employee well-being was related to overall organizational performance. Menges et al. (2011) found that organizations’ positive affective climate was positively associated with overall employee productivity and aggregate task performance behavior. We suggest that at the group level, positive emotional climate will similarly broaden the range of ideas discussed within groups and build enduring social resources among group members, and positive emotional climate will be positively related to workgroup performance. Correspondingly, the following hypotheses are proposed:
Organizational citizenship behavior (OCB) is defined as “individual behaviors that are discretionary, not directly or explicitly recognized by the formal reward system, and that in the aggregate promotes the effective functioning of the organization” (Organ, 1988, p. 4). OCBs have been studied mostly at individual level, showing the relevance for organizational outcomes, including productivity and well-being. With an increasing focus on teamwork, the importance of group organizational citizenship behavior (GOCB) within groups is growing, and there have been studies published that focused on group-or team-level OCB (i.e., Ehrhart, 2004; Ehrhart & Naumann, 2004). Besides the link between affect with performance, research shows that positive affect is also related to cooperativeness (e.g., Veitch & Griffitt, 1976) and helping behavior (e.g., Isen & Baron, 1991). Organizational research also demonstrates that positive affect is related to prosocial behavior (Lee & Allen, 2002). Isen and Baron (1991) suggested that positive feelings might be conducive to the development of an overall tone or “culture” of helpfulness within the organization. Thus, the positive emotional climate of workgroups may facilitate group OCB. Menges et al. (2011) found that organizations’ positive affective climate was positively associated with aggregate OCB. Correspondingly, the following hypotheses are proposed:
Conflict is central to team effectiveness because conflict is a natural part of the process that makes team decision-making so effective in the first place (Amason, Thompson, Hochwarter, & Harrison, 1995). There are mainly two kinds of conflict in teams. According to Jehn (1992), relational conflict includes personal and affective components such as friction, tension, and dislike among members within the group. Task conflict involves “differences in viewpoints and opinions pertaining to the task” (p. 11). These two types of conflict are closely related and so often occur together (De Dreu & Weingart, 2003). Thus, it is difficult to have one without also getting the other. Researchers have speculated that task and relational conflicts occur together because they share common antecedents (Amason & Sapienza, 1997; Jehn, 1995). Researchers have sought to identify that the team antecedents of conflict include characteristics of the team, such as size, composition, and diversity. We propose that negative WEC could also be a source of team conflict. Scholars have related both kinds of conflict to employees’ affect. Studies point out that relationship conflict is positively associated with group members’ negative emotions, such as stress and anxiety (Jehn & Mannix, 2001) and employees’ job tension (Medina, Munduate, Dorado, Martínez, & Guerra, 2005). There is some evidence showing that high levels of task conflict positively related to tension and unhappiness (e.g., Amason & Sapienza, 1997; Jehn, 1995; Jehn & Mannix, 2001). Team members in negative team emotional climate tend to experience more relationship and task conflict. Correspondingly, the following hypotheses are proposed:
Sample
Twelve organizations in mainland China served as research sites for this study. The twelve organizations were located in urban, mainly in the city of Beijing, Chongqing, Hangzhou, and Handan. To get better collaboration from the target organizations, we first approached the CEO or human resource (HR) professionals in each organization; we used multiple channels (telephone, emails, and personal meetings) to explain the purpose of the research and assure them of the confidentiality of the responses. Once they agreed to participate, we mailed or personally delivered the questionnaires to the organization and asked the HR professionals to help with data collection. Workgroups were selected based on five criteria: (a) the number of team members is more than 3 and preferably between 5 and 15; (b) the team members have worked together for more than 1 month to ensure that the WEC has formed through team members’ past work experience; (c) one formal team leader is well recognized by team members; (d) the interaction of the team members is frequent defined as interacting with each other for more than half of their working time; and (e) virtual-team and long distance communication teams are excluded. We also eliminated groups with fewer than three respondents (Kostopoulos & Bozionelos, 2011). The minimum criterion for inclusion in the study was for 50% of group members to have provided complete responses (V. Rousseau & Aubé, 2010). The final total response rate of teams was 85% and the average response rate was 75% within teams. These response rates were deemed comparable with that achieved in similar studies on teams (e.g., Haas, 2010; V. Rousseau & Aubé, 2010). As the CEOs or HR directors were very cooperative, almost all the samples we identified participated in our survey, except for some uncontrollable reasons, such as team members were away on sick leave or a business trip. The participating team members all fully filled in the questionnaires. For this reason, we believe that the missing data in our study were missing completely at random. Completed member and leader questionnaires collected from each team were put into one envelope and labeled with a team identification number.
The sample consisted of 151 workgroups with 846 workgroup members in 12 different organizations, including 15 R&D workgroups in a server hire and trusteeship company, 8 airline profit management workgroups, 7 technical service workgroups in a telecommunications company, 26 product line teams in an air pressure production plant, 28 product line teams in an engine production plant, 2 management teams in a bank, 7 management teams in an agribusiness organization, 3 marketing teams in a dairy foods company, 4 management teams in an insurance company, 8 marketing teams in a musical instrument production plant, 11 research workgroups in a university, 10 service workgroups in a regional government, and 22 management teams in a regional government. After we deleted the groups with less than three workgroup members (n = 3), 148 teams with 840 workgroup members remained in total. The average team size was 6.93 (SD = 3.24) workgroup members, with a minimum of 3 and a maximum of 16 workgroup members per team. The average team tenure was 13.34 years. Analyses indicated that 2.03% of the teams were in the forming stage, 2.70% of the teams were in the storming stage, 41.89% of the teams were in the norming stage, and 53.38% of the teams were in the performing stage.
Measures
All the measures in our study were presented in Chinese. The recommended translations and back-translation procedures (Brislin, 1980) were followed when translated Chinese version scales were used. Response options ranged from (1) “strongly disagree” to (5) “strongly agree.”
WEC
The 16-item version of the WECS developed in Study 1 was administered to workgroup members in Study 2. Response options ranged from (1) “strongly disagree” to (5) “strongly agree.” As in Study 1, the scale assesses four dimensions of WEC: EN, OP, ON, EP, and the reliability estimates (α) were .83, .70, .79, and .78, respectively.
Workgroup performance
Team leaders were asked to rate their team’s performance. As the teams in our study performed different tasks and were from different organizations, it was hard to evaluate their team performance with an objective measure. We employed a seven-item team performance scale developed from Wu (2005). Workgroup leaders were asked to assess their team’s performance from the aspects of achievement of the group goal, product and service, productivity, efficiency and the ability to creatively solve problems. Response options ranged from (1) “strongly disagree” to (5) “strongly agree.” A sample item was as follows: “The workgroup met its work goals.” The internal consistency reliability of this scale in our sample was .78.
Group OCB
We also asked the team leaders to assess group OCB using the five-item scale taken from Lam, Hui, and Law (1999) and Podsakoff, MacKenzie, Moorman, and Fetter (1990). Response options ranged from (1) “strongly disagree” to (5) “strongly agree.” A sample item was as follows: “Members of this workgroup are willing to help each other on organizationally relevant tasks.” The internal consistency reliability of this scale in our sample was .73.
Task and relationship conflicts
We used Jehn’s (1995) eight-item instrument to measure task and relationship conflicts. Team leaders were asked to describe on a 5-point Likert-type scale, anchored by 1 = none and 5 = a lot. A sample item for task conflict was as follows: “How frequently are there conflicts about ideas in your team?” A sample item for relational conflict was as follows: “How much tension is there among members in your team?” The coefficient alphas for the scales of task conflict and relationship conflict were both .87.
Consensual and discriminable climates
Concerns have been raised in the past about whether shared climates can be claimed to exist at the group or organization level (Patterson, West, & Payne, 1996). Accordingly, we sought to demonstrate agreement among individuals’ perceptions of climate prior to averaging climate. Roberts, Hulin, and Rousseau (1978) used a climate study (Drexler, 1977) to illustrate and conclude that “small within-organization variance relative to between-organization variance suggests that averaged perceptions of climate might be a useful concept.” (p. 86). Hence, to demonstrate that our WEC does exist at the group level instead of only as individual perceptions, we attempted to illustrate the homogeneity of within-group variance and the differences between groups. The average rwg of the four dimensions of WEC (EN, OP, ON, EP) were .73, .80, .75, and .77, respectively; all above the cutoff of .70 identified by James, Demaree, and Wolf (1993) as an acceptable level for within-group agreement. These results suggest that the measure is consistently tapping shared climate perceptions, rather than aggregating radically diverse individual perceptions. It is, however, also important to demonstrate differences between groups to determine the discriminant power of any climate instrument (D. M. Rousseau, 1988). We conducted one-way ANOVAs on individual scores of the variables, finding that the “between-group variance” were significant: the ANOVAs produced F ratios of the four dimensions of WEC (EN, OP, ON, EP) were 4.23, 2.70, 4.37, and 3.60, respectively, all greater than the acceptable cutoff of 1.00 suggested by Hays (1981) and the F values were statistically significant (p < .001) for all the dimensions’ average scores. The ICC(1) and ICC(2) values of the four dimensions of WEC (EN, OP, ON, EP) were .37, .23, .38, .32 and .76, .63, .77, .72, respectively. All the ICC(1) values were above .12 and all the ICC(2) values were above .50, which James (1982) suggested as the minimum acceptable levels. Based on the assumption that missing data were missing completely at random in our study, within-group agreement and reliability indices are not biased (Newman & Sin, 2009). Moreover, according to Newman and Sin (2009), as our WECS has 16 items, which far exceeds their threshold of 5 items, the effects of non-response bias on ICC and rwg should be at a minimum (below 5%). These results suggest that the WECS measures shared climate perceptions and distinguishes between different groups and thus possesses adequate discriminant as well as consensual validity.
Multilevel CFA
As we establish the factor structure of WECS at the group member level in Study 1, we followed Dyer, Hanges, and Hall’s (2005) five-step procedure to conduct the multilevel confirmatory factor analysis (MCFA) with LISREL 8.51 (Jöreskog & Sörbom, 1993) to further assure the factor structure of WEC at the group level. The results are shown in Table 5.
Model Fit for a Priori Single- and Multilevel Models.
Note. For all χ2, p < .01. RMSEA= root-mean-square error of approximation; SRMR =standardized root-mean-residual; CFI = comparative fit index; IFI = incremental fit index; NNFI = non-normed fit index.
First, we performed a conventional CFA on the sample total covariance (ST). This model appeared generally plausible (see Table 5). Because the total covariance matrix contains both between- and within-group-level information, however, this analysis fails to incorporate the hierarchical nature of the data and is potentially misleading (Dyer et al., 2005). Following Dyer et al.’s (2005) second step, we also assessed the amount of between-group-level variation contained in the total matrix. Although we conducted the ICC(1) and ICC(2) for each variable earlier, we followed the procedure to calculate the Muthén’s (1994) ICC, which assumes random level effects, whereas other procedures (e.g., rwg) assume fixed-level effects. In our study, the Muthén’s ICCs are .43, .39, .33, .42, .50, and .40 for the variables OP, ON, EP, EN, and the second-order variables of negative emotional climate and positive emotional climate, respectively. Given our relatively high ICC values, all above .12 which James (1982) suggested as the minimum acceptable level, we concluded that there was sufficient between-group variation to statistically warrant the use of multilevel analyses. In the third step, we performed a factor analysis on the sample pooled-within covariance matrix (SPW), a single-level CFA model was tested using the covariance matrix (SPW) based on individual-level scores, adjusted for their respective group means. The model estimated using SPW shows a bit of improved fit compared with that estimated using ST. Fourth, we performed a factor analysis on the sample between-group covariance matrix (SB). The fit indices reported in Table 5 show that the Step 4 a priori model has poorer fit than the model from the previous step, but this is in part because of the substantially smaller sample size used in Step 4 (i.e., the number of 148 workgroups vs. the 840 workgroup members). Finally, in the fifth step, we performed a multilevel CFA matrix (ST). As shown in Table 5, the results from a test of the a priori model of Step 5 suggest adequate fit at the within- and between-group level. Parameter estimates from this model included factor loadings at both the within and between levels. In this step, χ 2 (192) = 450.06, p < .001, RMSEA = .06. We could not get other fit indices except the RMSEA when conducting the multilevel analysis using Lisrel 8.51. Despite the fact that the sample size is low for multilevel CFA (Muthén, 1994), the RMSEA is satisfactory. The overall conclusion is that the proposed factor structure of WEC does fit the data well and is consistent both at the workgroup individual level and the workgroup level.
We have argued for conceptual isomorphism of the aggregation of team member perceptions of shared team emotions and emotional exchanges to represent WEC. Our claim is supported by demonstrating the theoretical similarity of the constructs, demonstrating similarity in dimensions across the individual and group levels, and establishing measurement equivalence, which exceeds the minimum requirement identified by Tay, Woo, and Vermunt (2014). Overall, the results provide good support for the construct validity of this new measure.
Predictive validity: The relationship between WEC and workgroup effectiveness
Table 6 shows the descriptive statistics, reliabilities, and correlations among the four dimensions of WEC and workgroup effectiveness at team level.
Descriptive Statistics, Reliabilities, and Correlations Among Workgroup Emotional Climate and Workgroup Effectiveness (N = 148).
Note. EN = ego-focused and negative workgroup emotional climate; OP = other-focused and positive workgroup emotional climate; ON = other-focused and negative workgroup emotional climate; EP = ego-focused and positive workgroup emotional climate; GP = workgroup performance; GOCB = workgroup OCB; RC = relationship conflict; TC = task conflict; OCB = organizational citizenship behavior.
p < .05 (two-tailed). **p < .01 (two-tailed tests).
LISREL 8.51 was used to test our hypotheses of the effects of WEC on workgroup effectiveness. We simplified the structural model by reducing the number of indicators for workgroup performance, workgroup OCB, and task and relationship conflict to curtail the problem of having too many indicators (Bentler & Chou, 1987) with the method commonly used in management research (e.g., Aryee, Budhwar, & Chen, 2002).
We conducted SEM (Structural equation modeling) on WEC with its higher order model of negative and positive emotional climate and workgroup effectiveness, including workgroup performance, group OCB, task conflict, and relationship conflict. The fit of the structural models is shown in Figure 1. Statistics related to the overall fit of the SEM revealed that the fit was acceptable, the model χ2 was 972.22 (df = 398) and the standardized RMSEA, CFI, GFI, NNFI, and IFI were .06, .94, .89, .93, and .94, respectively. As shown in Figure 1, positive WEC was positively related to workgroup performance (.36, p < .05), which supported our Hypothesis 1; positive WEC was positively related to group OCB, which supported our Hypothesis 2 (.33, p < .05). However, negative WEC was positively related to group relationship conflict (.25, p < .01), which supported our Hypothesis 3; and negative WEC was positively related to group task conflict (.19, p < .05), thus our Hypothesis 4 was supported.
Structural model of positive and negative workgroup emotional climate and workgroup effectiveness (N = 148).
We further used the SEM to investigate the specific effect of the four dimensions of WEC on workgroup effectiveness. Statistics related to the overall fit of the SEM revealed that the fit was good, the model χ 2 was 1,779.819 (df = 409) and the standardized RMSEA, CFI, GFI, NNFI, and IFI were .07, .90, .89, .90, and .90, respectively. As shown in Figure 2, although both OP and EP were positively related to group performance, the relationships were not significant, thus our Hypothesis 1 was not supported; OP and EP were positively related to group OCB (.21, p < .05; .25, p < .01), which supported our Hypotheses 2a and 2b; EN was positively related to group relationship conflict (.39, p < .01), thus our Hypothesis 3a was supported; ON was positively related to task conflict (.18, p < .05), which supported our Hypothesis 4b. The study therefore provides initial evidence of the relationship between WEC and workgroup effectiveness.
Structural model of workgroup emotional climate’s four dimensions and workgroup effectiveness (N = 148).
Discussion
With the rousing research interest in emotions in organizations (e.g., Barsade, Brief, & Spataro, 2003), a few scholars have identified emotion as a particularly important concept for teams (Elfenbein & Shirako, 2006) and begun to conduct studies examining the effects WEC for teams (e.g., Härtel et al., 2006). Unfortunately, research in this important area is being hampered by the lack of a psychometrically sound yet practical WEC measure for workgroup management research. This article described the development of a measure of an important aspect of workgroup climate, namely WEC. We used the de-contextualization approach as WEC is etic, culturally invariant, and context-sensitive (Jia, You, & Du, 2012), which provided the opportunity to develop a universal measure of WEC and facilitate scholarly exchange of research findings with the Western literature (Farh, Cannella, & Lee, 2006). Evidence of the four-factor structure of the measure was provided, based both on EFAs conducted on a sample of workgroup members in government and CFAs conducted on a cross-validation sample of workgroup members in firms. We also used MCFA and justified that the factor structure operated at the workgroup level of analysis as well as the workgroup member level. Supplementing this construct validity evidence for WECS was evidence for the consensual and discriminant validity of the scale at the group level and its predictive validity. The findings provide initial evidence for the utility of WECS as a self-report measure of emotional climate within workgroups as well as the relationship between WEC and workgroup effectiveness.
Our study contributes to research on WEC and provides the building blocks for group-level emotions studies. First, and most importantly, the WECS has a theoretical framework. Furthermore, our WEC measure is multidimensional rather than unidimensional. Based on previous research on the classification of emotion and emotional climate, we proposed that the valence (positive–negative) and interpersonal dimension of emotion can be jointly used to classify WEC and accordingly structure WEC on four dimensions. The advantages of this structure are that it both reflects the universal and social nature of the shared emotions in workgroups and allows for cross-cultural comparison of WEC among workgroups in different cultures (e.g., Härtel & Liu, 2012).
Second, considerations of group climate have often been restricted to applications of cognitive theories despite evidence from a range of researchers showing that group affect is quite important (e.g., George, 1990; Kelly & Barsade, 2001; Smith, Seger, & Mackie, 2007). Our study contributes to climate research by applying an affective perspective. Specifically, our study contributes to D. M. Rousseau’s (1988) call for the development of facet-specific measures of climate, in this case by providing an emotion-specific measure of climate. Moreover, as our WEC measure is multidimensional rather than unidimensional, our study combines and moves forward the research of climate on specific emotions, such as climate for safety (Zohar & Luria, 2005) or climate of fear (Ashkanasy & Nicholson, 2003). Finally, our study provides evidence of group-level emotion in China, a non-Western cultural country, which enriches the existing body of studies sampling exclusively from Western cultural groups (Elfenbein & Shirako, 2006).
Our study addressed the question of team effectiveness from the emotional aspect of workgroups and contributes to research on workgroup effectiveness. Many researchers have described the concept of happiness leading to success as the “holy grail” of industrial/organizational psychology (e.g., Fredrickson, 2001; Lyubomirsky, King, & Diener, 2005). Although we adopted cross-sectional research designs, our study provides preliminary support for this supposition in the form of WEC. Specifically, as we predicted, we found that a positive WEC is positively related to workgroup performance and group OCB; while negative WEC positively related to team relationship and task conflict, specifically, EN was positively related to group relationship conflict and ON was positively related to task conflict. Thus, our study provides initial evidence of the predictive validity of WECS.
Limitations and Future Research
There are also limitations in our study. First, our study was conducted in China, thus further empirical work in other cultural contexts would be welcome, to refine and test the measure of WEC for use as well as to examine cross-cultural differences in the effects of WEC on workgroup effectiveness. Specifically, because China has been characterized as a collectivist cultural country, it would be useful to attempt replicating our study in individualist countries and undertake a cross-cultural comparison of WEC in the future. Second, there are two general approaches to study group emotions (Smith et al., 2007). One approach is that researchers can study emotional responses to specific objects or events, and the other is that people can be asked to indicate the extent to which they feel a series of emotions in general. The latter research strategy was adopted in our study to get at general affective climate profiles rather than at responses to specific objects or events. As the two research approaches are complementary (Smith et al., 2007), we encourage continued research consideration of team’s immediate emotional reactions to specific objects or events as a complement to team’s general emotional climate profiles to uncover different (though related) aspects of the overall picture of team emotional life. Third, though we theoretically distinguish between affective tone and WEC, we do not test empirically, which needs to be further explored in the future study. Fourth, it is a single time period study, which limits our ability to make causal inferences about the relationships proposed. Longitudinal studies are required to explore the causal relationships between WEC and workgroup effectiveness.
In addition to addressing these limitations, future research is required on a number of key questions. How does WEC develop over time? What are the specific mechanisms through which emotional climate influences workgroup effectiveness? How do different workgroup members with different roles and status influence the WEC? When the members’ personal emotional experiences are different or similar with the WEC, how does this similarity or difference affect workgroups? How does the team leader’s leadership style and emotions influence the WEC? The way to answering these questions is opened in providing WECS as a reliable and valid measure of WEC. Moreover, although we set up WEC as a referent-shift composition in our article, climate strength is also a possibility to operationalize WEC (Cole, Bedeian, Hirschfeld, & Vogel, 2011) and is an emerging and useful construct (Kuenzi & Schminke, 2009), which remains to be studied in the future.
Practical Implications
A number of practical implications emerge from this work. First, as emotion is a particularly important concept for teams (Elfenbein & Shirako, 2006) with important consequences for workgroup effectiveness, organizations are advised to provide team leaders and members with training on the emotional aspects of workgroups. Team leaders need to see the value of WEC and pay more attention to the emotional needs of individuals and the emotional linkages and relationships within the workgroup. Second, the results of this study suggest that positive emotional climate is vital to workgroup effectiveness. More specifically, OP and EP were positively related to group OCB. Leadership practices that facilitate a positive emotional climate (the “PEC practices”) should be encouraged in organizations, such as frequently giving positive feedback, offering opportunities for advancement, and rewarding employees who take special initiative (Ozcelik, Langton, & Aldrich, 2008). Finally, the leader should create a supportive culture and offer support for toxin handlers, especially when the emotional climate turns negative. As Frost (2004) suggested, the toxin handlers in organizations will take the initiative to handle toxic emotions constructively with discreet but skillful interventions in the workgroup when emotional pain is generated in the workplace. These handlers might be the team leaders or team members who focus on the emotional needs of individuals and on the emotional linkages and relationships within the workgroup. They have the empathetic capacity to notice when the WEC turns negative and toxic and step into situations at work to dissipate or to buffer the toxins, so that those who are “in harms way” are rescued or protected and can get on with doing their organizational work (Frost, 2004). Toxin handling interventions also should be designed to redress a negative emotional climate and to help team members recover their equilibrium, and develop a positive emotional climate in the workgroup.
At a more pragmatic level, the 16-item WECS is an accessible and easily administered measure that has potential for use in settings such as organizational climate surveys, team building and development, and group development interventions. We hope that the combination of high validity and supportive initial research results will encourage the continued use of the WECS including additional work on its validation.
Footnotes
Acknowledgements
We would like to thank Remus Ilies, Brian Cooper, Rob J Hyndman, Gile Hirst, Tao Yang and Kwok Leung for their very helpful comments and suggestions.
Authors’ Note
An earlier version of the article was presented at the Sixth International Conference on Emotions and Worklife: Emonet VI (Fontainebleau, France).
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Natural Science Foundation of China (Grant numbers 71002003, 71372005), the Program for Young Excellent Talents, UIBE (University of International Business and Economics) (Grant number 2013YQ04), and the Beijing Higher Education Young Elite Teacher Project (YETP0893).