Research sample
We sent surveys to members of the customer teams of a major office equipment and business services supplier, as well as to their customers. Company activities include outsourcing print room processes, operation and maintenance of complete copier and printer systems, fleet management, managing electronic and physical archives, and scanning and mailroom activities. Given the nature of its business, which is largely paper based, the company acknowledges that the potential environmental impact of its products significantly constrains its activities and its license to operate. In other words, producing and servicing paper processing-based products appears likely to become a vulnerable business strategy. In its 2005 sustainability report, the company explicitly stated that employees should accept individual and collective responsibility and explore opportunities associated with sustainability. The company views itself as part of an integral chain and concludes that its sustainability is codetermined by the sustainability of its partners. It organizes so-called “tool-box” sessions on a regular basis to allow team members to share their experiences with environmentally related issues and solutions.
Customer contact employees have a wide array of products and services available that were developed taking into account the environment, such as toner recycling services, a range of soy-based ink cartridges, asset recovery services (product revisions aimed at more environmentally-friendly operation) and on-site energy and ozone-emission assessment. According to the results of a client survey, which are rendered in the firm’s sustainability report, environmental issues that will become increasingly important as purchasing criteria are an efficient use of paper and toner, recyclability of products [meaning waste reduction], reduced energy consumption and reduced emissions of ozone and fine dust.
During a new product introduction event at the company headquarters a group discussion with representatives from two of the company’s most important business segments; (1) professional printing companies and (2) architects. To identify whether environmental stewardship is an important theme, the central question was “why bother having a green supplier”. The response to this can be summarized by the following main conclusions:
-
A personal as well as company need to be environmentally conscious. Personal conviction, moral obligation, but also the fact that their own clients demand responsibility for the environment. Respondents indicated that public opinion and stakeholders require this, and for them too it was clear that a responsibility towards the environment is directly related to their license to operate.
-
You only are as green as your supply chain As a manufacturer is the beginning of the supply chain, they expect it to set the standard and this is likely to have an impact on the rest of the chain.
-
The reputation is on the front line They increasingly evaluate their relationship with their supplier in the light of environmental issues; “Eco-labelling and sustainability reports are easy to produce, it is actions that count”
-
The number of green stakeholders is growing It was indicated that clients themselves are increasingly confronted with activist groups inquiring about carbon-footprints, recycling of paper, and paperless information streams.
-
It’s not only about threats, but also opportunities This gets back to the fact that a good reputation sells in the market, but also that the pressing need for environmental responsibility is sparking collaborative, innovation projects. Two of the respondents mentioned as an example the recent asset recovery initiative that the company has launched where it takes responsibility (at a fee) to discard old equipment and take care of its recycling.
The firm employs approximately 23,000 people, 45% of whom deliver services and sell product parts, and after-sales support through a team organization. In addition to servicing products and managing the relationship with the customer, the teams recently received sales quota assignments. The quotas primarily refer to direct selling of equipment upgrades and product accessories, as well as cross- and up-selling of service contracts. Teams depend on both geographic location and segmentation (i.e., regional networks) and vary in size between 6 and 14 members. Multiple teams (commonly two or three) typically work in the same local network. The teams can be regarded a meaningful entity as they share a history, present and future (cf. Gully
2000). Despite the fact that employees work individually at customer sites, they frequently cooperate on complex issues and meet for training activities. They frequently consult each other and collectively use a dispatch system to divide the workload. Moreover, they collectively decide on the use of resources, budgets, performance measurement and hiring new team members. As Mathieu et al. (
2007, p. 897) state ‘the efforts of individuals are really the by-products of the contributions and coordination of many others from their team’.
As part of an annual, international, employee and customer research program, all 52 customer contact teams of the largest business unit received a special appendix of the survey for our environmental stewardship research project (408 employees). We collected data at t–1 and t (8 months later) and received 351 questionnaires (86.0%) from 37 teams at t–1 and 324 questionnaires (79.4%) from 34 teams at t. Therefore, we use 34 teams and 324 questionnaires for the analyses. For the customer portion of the survey, we randomly selected samples of 50 customers per team, again at t–1 and t. In total, we gathered 416 (16.0%) questionnaires at t–1 and 312 (12.0%) questionnaires at t for the analysis.
Among the employee sample, 48% are younger than 40 years, most are men (93%), and most have a technical background (83%). More than half of the customer contact employees possess extensive company experience (52% > 10 years) but have been with their current team for a relatively short time (58% < 3 years). In the customer sample, the majority again are men (79%). Most customers had developed a long-term (71% > 10 years) relationship with the company.
Measures
With the notable exceptions of Groesbeck (
2001) and Barbuto and Wheeler (
2006), who develop five-item scales for general group and company stewardship, respectively, no existing scales are tailored to our research domain. Therefore, we adapt our scale for environmental stewardship from general stewardship scales, on the basis of 11 comprehensive interviews with team members. Our operationalization is consistent with earlier studies on stewardship, that emphasize stewardship as social responsibility over self-interest (Block
1993). We borrow the scale for the autonomy construct (4 items) from Kirkman and Rosen (
1999). Our operationalization of the supervisory support scale (4 items), was inspired by the scales of Hyatt and Ruddy (
1997) and Campion et al. (
1993). All scale items by means of the employee survey on a 7-point scale, ranging from “strongly disagree” (1) to “strongly agree” (7). Operationalizations of these scales are provided in the
Appendix.
Furthermore, the variable team size and team tenure were included as controls. Team size reflects the number of employees that a team counts and team tenure denotes the number of years an employee is a member of the team. The latter variable consisted of six answer categories ranging from ‘<1 year’ (1) to ‘>5 years’ (6).
In addition, we evaluated the measurement properties of these measures at the individual employee level (t–1) by conducting a confirmatory factor analysis (CFA) with three latent variables (environmental stewardship, autonomy, and supervisory support). We assessed the distributional properties for the items used in the analysis and found that none exhibited excessive univariate skewness (g1 < 3) or univariate kurtosis (g2-3 < 10; Kline
2005). However, Mardia’s (g2, p) normalized estimate for multivariate kurtosis equals 35.90. As a consequence, we decided to employ robust maximum likelihood estimation in EQS 6.1 to obtain the estimates (Bentler
1995). Our analysis reveals a good fit to the data: χ
SB
2
(62) = 107.86,
p < .001, Tucker-Lewis index (TLI) = .95, confirmatory fit index (CFI) = .96, incremental fit index (IFI) = .96, and root mean squared error of approximation (RMSEA) = .048. These measures indicate unidimensionality. To assess the convergent validity of the measures, we determined whether the manifest variables load significantly and adequately in magnitude on the hypothesized latent variable (Anderson and Gerbing
1988). All (standardized) loadings are significant at
α = .05 with a mean of ≥ .70, and all of them exceed .6, as recommended by Bagozzi and Yi (
1988). The (standardized) loadings, pattern of the residuals, and Lagrange multiplier tests show that none of the items should be omitted from the analysis. We calculated the composite reliability (CR) and average variance extracted (AVE) for each measure and find that the CRs exceed the recommended cut-off value of .7 (autonomy [CR = .81], supervisory support [CR = .87], environmental stewardship [CR = .86 (t–1) and .92 (t)]) and that AVE exceeds the recommended cut-off value of .5 (Fornell and Larcker
1981). Finally, to assess discriminant validity, we compared the square root of the AVE with the (attenuated) correlations of the latent variables (Fornell and Larcker
1981). For each pair of the latent variables, the square root of the AVE exceeds the (attenuated) correlations between the latent variables, which indicates discriminant validity.
We operationalize within-team stewardship consensus using the standard deviation of team member perceptions of their team. We use the standard deviation instead of the r
WG(j) statistic, because it better reflects the (lack of) within-team consensus (Schneider et al.
2002; Zohar and Luria
2005), whereas the distribution underlying the r
WG(j) does not always reflect the response range accurately (Bliese
2000). Our operationalization of the between-team stewardship consensus parameter relies on Zohar and Luria’s (
2005) operational definition of climate variability. Specifically, we operationalize between-team stewardship consensus by taking the standard deviation of the group means of team stewardship for each local network.
We assess customer satisfaction with respect to the following attributes related to employee attitudes and behavior: competence, empathy, friendliness, helpfulness, accuracy, and attentiveness. Customer respondents rate the six items on five-point scales ranging from “very dissatisfied” (1) to “very satisfied” (5). The items of the customer satisfaction scale are provided in the
Appendix.
We also perform a CFA to assess the construct validity. For both t–1 and t, we find good fits to the data (t–1: χ
SB
2
(9) = 36.04, p < .001, TLI = .97, CFI = .98, IFI = .98, RMSEA = .057; t: χ
SB
2
(9) = 46.15, p < .001, TLI = .96, CFI = .98, IFI = .98, RMSEA = .061). To assess reliability, we calculate both CR and AVE; they exceed the recommended cut-off values (satisfaction CR = .89 (t–1) and .90 (t)). Finally, we obtain data about sales per customer from the company’s internal database (results per quarter in K Euro).
We evaluate the longitudinal invariance of the environmental stewardship and satisfaction measures using multisample CFA to analyze the equality of the sample variance/covariance matrices for t–1 and t (Σ
1 = Σ
2; Vandenberg and Lance
2000). The equality hypotheses for the variance/covariance matrices cannot be rejected for stewardship (χ
SB
2
(15) = 20.23,
p = .16, TLI = .99, CFI = .99, IFI = .99, RMSEA = .033) or satisfaction (χ
SB
2
(21) = 26.51,
p = .19, TLI = .99, CFI = .99, IFI = .99, RMSEA = .041).
We analyze the linkage between environmental stewardship and its consequences at the group level of analysis. From a conceptual point of view, satisfaction represents the outcome of synergetic work processes among team members, as reflected by outgroup-homogeneity theory, which states that people tend to observe other groups as more uniform than their own (Quattrone and Jones
1980). The implication for our research setting is that customers (as members of the external customer group) likely perceive the attitudes and behavior of one or a few frontline employees as the general feature of the team. Finally, because the company prioritizes a privacy policy, we cannot empirically match employee and customer evaluations or sales at the individual level of analysis. Therefore, we aggregate stewardship and satisfaction to the group level.
Justification for aggregation
We calculate the r
WG(j) statistic and intra-class correlation (ICC) coefficients for autonomy, environmental stewardship at t–1 and t, supervisory support, and satisfaction at t–1 and t to justify our data aggregation to the team level. The r
WG(j) coefficient, which indicates homogeneity in individual ratings within teams, results in high values for all variables (from .86 to .96). These findings demonstrate that individual ratings within groups are highly consistent (James et al.
1993). Whereas the r
WG(j) coefficient only takes into account differences among individuals within groups, the ICC (1) coefficient involves a ratio of between-group variance to total variance and thus captures both within- and between-group variation. The ICCs (1)
1 for all variables are significant (F-values,
p < .07), ranging from .05 to .25, which indicates that each variable possesses a sizable amount of between-group variance. We also calculate ICC (2), which more precisely assesses the impact of interdependence because it accounts for group size. Except for autonomy (ICC (2) = .34), the ICC (2) values for all variables are greater than .50, which represents convincing evidence that group means can be considered reliable, even if the ICC (1) values are relatively small (Bliese
2000).
Means, standard deviations, and individual-level correlations between the employee variables are presented in Table
1. In Table
2, group-level means, standard deviations, and (partial) correlations of employee variables and external outcomes are represented. Environmental stewardship (t) appears to have the highest correlations with customer satisfaction. Furthermore, the antecedent-satisfaction correlations are noticeably weaker when the effect of environmental stewardship (t) is accounted for, implying that environmental stewardship (t) mediates the antecedent-customer satisfaction relationships (cf. Baron and Kenny
1986). In relation to sales, the mediating role of environmental stewardship is less obvious.
Table 1
Means, standard deviations, and correlations of individual-level variables
1. Team tenure (t–1) | 3.53 (1.83) | – | | | | |
2. Initial environmental.stewardship(t–1) | 5.50 (.88) | .03 | – | | | |
3. Autonomy (t–1) | 6.24 (.65) | .14** | .47*** | – | | |
4. Supervisory support (t–1) | 4.59 (1.22) | .03 | .28*** | .31*** | – | |
5. Environmental stewardship (t) | 5.60 (.81) | .11* | .40*** | .42*** | .38*** | – |
Table 2
Group-level means, standard deviations, and correlations of group-level variables
1. Customer satisfaction (t) | 4.17 (.16) | – | | | | .32 | −.03 | −.03 | −.02 | −.13 | | | |
2. Sales (t) | 2.44 (.60) | −.06 | – | | | −.20 | −.77** | .06 | .57** | .16 | | | |
3. Team tenure (t–1) | 3.52 (.59) | .12 | .33* | – | | | | | | | | | |
4. Team size (t–1) | 7.85 (3.43) | −.29* | .34* | .13 | – | | | | | | | | |
5. Customer satisfaction (t–1) | 4.18 (.16) | .40** | −.21 | .10 | −.44** | – | | | | | | | |
6. Sales (t–1) | 2.71 (.46) | −.10 | .76***** | .22 | .49** | −.26 | – | | | | | | |
7. Initial environ. stewardship (t–1) | 5.51 (.38) | 34* | .01 | −.31* | .09 | .03 | .02 | – | | | | | |
8. Autonomy (t–1) | 6.20 (.28) | .20 | .44** | .01 | .15 | −.18 | .20 | .67*** | – | | | | |
9. Supervisory support(t–1) | 4.54 (.65) | .08 | .12 | .02 | −.31* | .05 | −.03 | .38** | .30* | – | | | |
10. Within-team stewardship cons. (t-1) | .78 (.18) | .15 | .18 | .39** | .26 | −.13 | .13 | −.57*** | −.21 | −.37** | – | | |
11. Between-teams stewardship cons. (t-1) | .20 (.19) | −.12 | −.11 | .16 | −.28 | .08 | −.25 | −.51** | −.39** | −.01 | .38** | – | |
12. Environ. stewardship (t) | 5.56 (.33) | .37** | −.04 | −.08 | −.24 | .37** | −.18 | .68*** | .57*** | .48** | −.44** | −.35** | − |
Results of the analyses
We specify hierarchical linear regression models using MLwiN software (Rasbash et al.
2000) to estimate the lagged effects of the antecedent variables at t–1 on environmental stewardship at t. We initially include the control variables and the antecedents at the individual and group levels (direct consensus model 1). Next, we add interactions between within-team stewardship consensus and the antecedents to test the full model (dispersion model 2a). Finally, we test a competing model (dispersion model 2b) with interactions of between-team stewardship consensus and the antecedents. Our full model represents the following multilevel equation:
$$\begin{array}{*{20}c} {{\text{STEW}}_{{\left( {\text{t}} \right),{\text{ij}}}} = \gamma _{{00}} + \gamma _{{10}} {\text{TEN}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{ij}}}} + \gamma _{{{\text{20}}}} {\text{STEW}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{ij}}}} + \gamma _{{30}} {\text{AUT}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{ij}}}} + \gamma _{{40}} {\text{SUP}}_{{\left( {{\text{t}} - 1} \right),{\text{ij}}}} } \hfill \\ { + \gamma _{{50}} \left( {{\text{STEW}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{ij}}}} \times {\text{STEWCON}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} } \right)_{{{\text{ij}}}} + \gamma _{{60}} \left( {{\text{AUT}}_{{\left( {{\text{t}} - 1} \right),{\text{ij}}}} \times {\text{STEWCON}}_{{\left( {{\text{t}} - 1} \right),{\text{j}}}} } \right)_{{{\text{ij}}}} } \hfill \\ { + \gamma _{{70}} \left( {{\text{SUP}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{ij}}}} \times {\text{STEWCON}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} } \right)_{{{\text{ij}}}} + \gamma _{{01}} {\text{TEN}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} + \gamma _{{02}} {\text{TSIZE}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} } \hfill \\ { + \gamma _{{03}} {\text{STEW}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} + \gamma _{{04}} {\text{AUT}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} + \gamma _{{05}} {\text{SUP}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} + \gamma _{{06}} {\text{SERVQUAL}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} } \hfill \\ { + \gamma _{{07}} {\text{SALES}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} + \gamma _{{08}} {\text{STEWCON}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} + \gamma _{{09}} \left( {{\text{STEW}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} \times {\text{STEWCON}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} } \right)_{{\text{j}}} } \hfill \\ { + \gamma _{{010}} \left( {{\text{AUT}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} \times {\text{STEWCON}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} } \right)_{{\text{j}}} + \gamma _{{011}} \left( {{\text{SUP}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} \times {\text{STEWCON}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} } \right)_{{\text{j}}} } \hfill \\ { + \gamma _{{012}} \left( {{\text{SERVQUAL}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} \times {\text{STEWCON}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} } \right)_{{\text{j}}} + \gamma _{{013}} \left( {{\text{SALES}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} \times {\text{STEWCON}}_{{\left( {{\text{t}} - {\text{1}}} \right),{\text{j}}}} } \right)_{{\text{j}}} } \hfill \\ { + {\text{u}}_{{0{\text{j}}}} + {\text{u}}_{{1{\text{j}}}} + {\text{u}}_{{2{\text{j}}}} + {\text{u}}_{{3{\text{j}}}} + {\text{u}}_{{4{\text{j}}}} + {\text{e}}_{{{\text{ij}}}} ,} \hfill \\ \end{array} $$
(1)
where i refers to individuals; j indicates groups; STEW
(t) is the employee’s environmental stewardship appraisal at t–1; TEN
(t–1) and TSIZE
(t–1) refer to team tenure and team size at t–1, respectively; STEW
(t–1), AUT
(t–1), and SUP
(t–1) are environmental stewardship, autonomy, and supervisory support at t–1, respectively; and STEWCON
(t–1), SAT
(t–1), and SALES
(t–1) are stewardship consensus (either within- or between-teams), satisfaction, and sales at t–1, respectively. In addition, e
ij denotes the individual-level error term. The random terms u
qj (
q = 0, …, 4) reflect the unique variation of group j from the overall effect on the intercept (β
0j), after we partial out the effects of all group-level predictors. The coefficients β
0j, …, β
4j are random terms that may vary across teams.
We split the environmental stewardship, autonomy, and supervisory support variables according to group means and the individual scores of the employees to compare their group- and individual-level effects on environmental stewardship. To calculate the group-level coefficients, we use group means, but we derive the individual-level coefficients from the individual scores. The individual-level coefficients thus act as controls on the group-level coefficients (e.g., Vancouver et al.
1994); if the group-level coefficient of an antecedent remains significant after we include its individual-level coefficient, the coefficient explains additional variance in stewardship, beyond what its individual-level counterpart can explain. In this case, we can conclude that the antecedent has a specific group-level effect on the dependent variable.
To estimate our models with interaction terms, we employ a centring procedure that enables us to avoid multicollinearity between the main effects and the interaction variables. Therefore, we grand-mean-center the first-order variables first, and then develop the interaction terms (see Aiken and West
1991).
In Table
3, we present the findings of our multilevel analyses. All three models yield a higher R
2 at the group level than at the individual level, which indicates that the antecedents explain between-group variation of stewardship better than they do within-group variation. The findings further reveal that Model 2a does not provide a better fit (
χ
2 (6) = 11.145) than Model 1, whereas Model 2b yields a significantly better fit than Model 1 (
χ
2 (6) = 24.328) and displays substantially higher explanatory power than Model 2a. Specifically, Model 2b reveals positive individual-level effects of environmental stewardship, autonomy and supervisory support (t–1) on environmental stewardship (t,) in support of Hypotheses 1–3. At the group level and contrary to our expectations, we find a positive significant effect of satisfaction (t–1) on stewardship (t), so we fail to support Hypothesis 4. In line with our expectations, we also uncover a negative effect of sales (t–1) on environmental stewardship, in support of Hypothesis 5. The significant positive group-level effects of stewardship (t–1) and supervisory support (t–1) on stewardship (t) supports Hypotheses 6a and 6c. Conversely, autonomy (t–1) appears not to have a significant group-level effect on stewardship (t), so we cannot support Hypothesis 6b.
Table 3
Lagged multilevel regression analyses of antecedent–environmental stewardship relationships
Individual-level variables:
|
Team tenure | .119** | .115** | .127** | |
Initial environmental stewardship | .168** | .212** | .232** | H1 |
Autonomy | .222** | .181** | .166** | H2 |
Supervisory support | .233** | .233** | .241** | H3 |
Group-level variables:
|
Team tenure | −.078 | −.093 | .006 | |
Team size | .065 | .016 | .131 | |
Initial environmental stewardship | .151* | .030 | .249** | H6A |
Autonomy | .025 | −.001 | .061 | H6B |
Supervisory support | −.027 | −.009 | .171* | H6C |
Customer satisfaction | .150* | .116* | .235** | H4 |
Sales | −.152* | −.117* | −.291** | H5 |
Within-team stewardship consensus | | .069 | | |
Between-teams stewardship consensus | | | .377** | |
Cross-Level Interactions: Stewardship consensus × individual level variables
|
Within-team stewardship consensus × env. stewardship | | .047 | | |
Within-team stewardship consensus × autonomy | | .021 | | |
Within-team stewardship consensus × supervisory support | | −.035 | | |
Between-teams stewardship consensus × env. stewardship | | | −.052 | |
Between-teams stewardship consensus × autonomy | | | .019 | |
Between-teams stewardship consensus × supervisory support | | | .032 | |
Group-Level Interactions: Stewardship consensus × group level variables
|
Within-team stewardship consensus × env. stewardship | | −.129 | | H7A |
Within-team stewardship consensus × autonomy | | −.006 | | H7B |
Within-team stewardship consensus × supervisory support | | .047 | | H7C |
Within-team stewardship consensus × satisfaction | | .181** | | H7D |
Within-team stewardship consensus × sales | | .090 | | H7E |
Between-teams stewardship consensus × env. stewardship | | | .687** | H8A |
Between-teams stewardship consensus × autonomy | | | −.150 | H8B |
Between-teams stewardship consensus × supervisory support | | | −.275* | H8C |
Between-teams stewardship consensus × satisfaction | | | −.068 | H8D |
Between-teams stewardship consensus × sales | | | .175** | H8E |
Increase in model fit a: | χ2 (15) = 125.7** | χ2 (9) = 12.3 | χ2 (9) = 25.8** | |
Explained individual level variance (%) | 36.0% | 39.2% | 41.7% | |
Explained group-level variance (%) | 59.4% | 69.9% | 75.5% | |
To test Hypothesis 7, we construct interaction terms of within-team stewardship consensus and the antecedents (see Model 2a) and find a significant, positive interaction of within-team stewardship consensus and satisfaction. That is, when within-team stewardship consensus increases, the positive effect of satisfaction on team stewardship strengthens, in contrast with Hypothesis 7d. None of the interactions of within-team stewardship consensus with the other four antecedents demonstrates significance, which means that we do not find support for Hypotheses 7a–c or 7e.
Next, we specify interactions between the antecedents and between-team stewardship consensus to test Hypothesis 8 (see Model 2b). We find a significant positive direct effect of between-team stewardship consensus on team stewardship and significant interactions with three antecedents. First, a significant positive group-level interaction exists for between-team stewardship consensus and initial stewardship on stewardship (t), which implies that when stewardship consensus between teams increases, the positive effect of initial stewardship on subsequent stewardship becomes stronger, in support of Hypothesis 8a. Second, and contrary to our expectations, we find a significant negative interaction of between-team stewardship consensus and group-level supervisory support. When between-team stewardship consensus increases, the positive impact of supervisory support on stewardship (t) is weaker, indicating no support for Hypothesis 8c. Third, our findings reveal a significant positive interaction of between-team stewardship consensus and sales (t–1), such that when between-team consensus increases, the negative effect of sales weakens, in support of Hypothesis 8e. We do not find significant interactions of between-team stewardship with group-level autonomy or satisfaction (t–1). Hence, we fail to find support for Hypotheses 8b and 8d.
Finally, we find that none of the specified cross-level interactions in Models 2a and 2b turn out to be significant. In relation to the control variables, all models feature only a positive individual-level effect of team tenure on environmental stewardship perceptions.
We use the data collected at t to estimate the group-level effects of team stewardship on its consequences (i.e., Hypotheses 9 and 10) through a multivariate regression model, formulated as a two-level hierarchical linear model, where level 1 reflects the dependent variables indexed by h = 1,…, m, and level 2 represents the teams j = 1,…, N. To formulate the multivariate regression model as a hierarchical linear model, we employ the dummy variables d
1 to d
m to reflect the dependent variables (i.e., satisfaction and sales). Dummy variable d
h equals 1 or 0, depending on whether the data line refers to the dependent variable Y
h or to the other dependent variable. Thus, the regression models for the m dependent variables can be integrated into a two-level hierarchical model, in which the variables (including the intercept) multiply with the dummy variables. This approach yields the following equation:
$$ {\text{Y}}_{{{\text{hj}}}} = \gamma _{{0{\text{h}}}} + \gamma _{{1{\text{h}}}} {\text{STEW}}_{{\text{j}}} + {\text{e}}_{{{\text{hj}}}} , $$
(2)
where Y
hj is the measurement of the hth variable for team j, and STEW is the team’s average environmental stewardship score at t. Our results in Table
4 reveal a significant positive effect of environmental stewardship on satisfaction, but no relationship exists with sales; hence, we fail to find support for Hypotheses 9 and 10.
Table 4
Multiresponse regression analysis of environmental stewardship–outcome relationship
Env. stewardship → Satisfaction | .371** | 13.8% | H9 |
Env. stewardship → Sales | −.043 | .2% | H10 |
Residual between-group (co)variance terms a: |
\( \sigma_h^2 \)= var (e
hj
), (h = 1) | .022 (.005) | | |
\( \sigma_h^2 \)= var (e
hj
), (h = 2) | .352 (.085) | | |
\( \sigma_{12} \)= cov (e1j
, e2j
) | −.004 (.015) | | |
Increase in model fit b: | | χ2 (3) = 5.215 | |
Additionally, we set up a system of equations which allowed us to test for the mediation of the antecedents of environmental stewardship(t) and the outcomes, sales(t) and customer satisfaction(t) (Iacobucci et al.
2007; MacKinnon
2008). Since the outcome variables, sales(t) and customer satisfaction(t) are only available at the team level, mediation can only be assessed using group-level variables. For sales(t) we found no significant influence of environmental stewardship(t) using a significance level of 0.05. Therefore, we may conclude that environmental stewardship(t) does not mediate the relationships between the antecedents and sales(t). However, for customers satisfaction our results suggest complete mediation for customer satisfaction(t–1), environmental stewardship(t–1) and autonomy(t), as the null hypothesis that direct effects of the antecedents are equal to zero could not be rejected using a Wald test (F(7,76) = 0.82,
p = 0.58; cf MacKinnon
2008). Using the bootstrap approach suggested by Shrout and Bolger (
2002) to determine the standard error of the indirect effects (Sobel test) we found that at a significance level of 0.05 the indirect effects of customer satisfaction(t–1) [
b = 0.12,
z = 1.78,
p = 0.04], environmental stewardship(t–1) [
b = 0.07,
z = 1.72,
p = 0.04] and autonomy(t) [
b = 0.08,
z = 1.82,
p = 0.03] are significant.