Integrating and expanding on the preliminary insights offered by Experiment 1, Experiment 2 investigated whether green skepticism and corporate credibility serially mediates the interactive effect of environmental claims and external disconfirming information on brand attitudes and purchase intentions.
Participants and Design
We recruited 210 Americans following the same procedures detailed in Study 1 (Mage = 33.61 years old, SD = 10.37 years old; 79 males, 131 females) and used a 2 (environmental claims: vague, specific) × 2 (external information: vague, specific) between-participants design.
Participants were randomly assigned to either specific or vague environmental claims from a fictional paper company, “Green Co. Paper.” The stimuli were a mix of text and pictures (see Appendix 1), with the text stating either that “We only use materials sourced from certified, sustainable suppliers: 100% recycled paper, energy efficient machinery, non-toxic chemicals” (specific) or that “We are an environmentally-friendly company” (vague). On the following screen, we presented participants with third party information: “Now imagine that you heard on the news that Green Co. Paper used materials sourced from non-certified, non-sustainable suppliers. This included non-recycled paper, energy inefficient machinery and toxic chemicals” (specific), or “Now imagine that you heard on the news that Green Co. Paper is not an environmentally friendly brand” (vague).
After exposure to the experimental stimulus, participants completed the same measures for purchase intentions (single item) and brand attitudes (α = 0.98) as in Experiment 1. They also completed 3-item, 7-point Likert scales measuring credibility toward the company (α = 0.97; Ohanian 1990) and skepticism toward the environmental claims (α = 0.92; Matthes and Wonneberger 2014). Manipulation checks and sociodemographic information were recorded last following the same procedures detailed in Experiment 1. Refer to Appendix 1 and 2 for more detail on both the stimuli and the measures employed.
Results: Mean Differences
For environmental claims, participants in the vague condition reported that the claims were more vague (Mvague = 3.33, SD = 1.83) compared to participants assigned to the specific condition (Mspecific = 4.91, SD = 1.57; F[1, 208] = 45.02, p < 0.001). For external information, the same pattern of results occurred (Mvague = 3.38, SD = 1.71 vs. Mspecific = 4.92, SD = 1.52; F[1, 208] = 47.33, p < 0.001).
A 2 × 2 ANOVA indicated that the two-way interaction was significant, F(1,206) = 7.35, p < 0.01. When environmental claims were specific, specific external information reduced purchase intentions (M = 2.19, SD = 1.02) compared to vague information (M = 3.36, SD = 1.52), F(1,105) = 20.12, p < 0.001. But when the environmental claims were vague, specific external information did not impact purchase intentions (M = 2.39, SD = 1.71) compared to vague external information (M = 2.52, SD = 1.30; p = 0.65).
The two-way interaction was significant for brand attitudes also, F(1, 206) = 7.98, p < 0.01. When environmental claims were specific, specific external information reduced brand attitudes (M = 2.10, SD = 1.35) compared to vague information (M = 3.40, SD = 1.59), F(1, 105) = 20.31, p < 0.001. But when the environmental claims were vague, specific external information did not influence purchase intentions (M = 2.02, SD = 1.22) compared to vague external information (M = 2.29, SD = 1.07; p = 0.22).
A 2 × 2 ANOVA indicated the interaction was significant for credibility as well, F(1,206) = 6.67, p < 0.02. When environmental claims were specific, specific external information reduced credibility (M = 1.97, SD = 1.20) compared to vague information (M = 3.04, SD = 1.53), F(1,105) = 16.07, p < 0.001. However, when environmental claims were vague, external information did not affect credibility whether it was specific (M = 1.82, SD = 1.03) or vague (M = 2.03, SD = 1.05; p = 0.31).
The 2 × 2 interaction was also significant for credibility, F(1,206) = 5.73, p < 0.02. When environmental claims were specific, specific external information raised skepticism (M = 5.90, SD = 1.07) compared to vague information (M = 4.84, SD = 1.42), F(1,105) = 18.65, p < 0.001. But when environmental claims were vague, external disconfirming information did not affect skepticism whether it was specific (M = 5.92, SD = 1.05) or vague (M = 5.62, SD = 1.01; p = 0.13).
Results: Structural Equation Modeling and Process Evidence
To test our hypothesized model while also accounting for measurement error, we used structural equation modeling (SEM) with maximum likelihood estimation (software package: AMOS 22). A confirmatory factor analysis also allowed to ensure that all variables included in the model, in particular green skepticism and corporate credibility, exhibited both convergent and discriminant validity. Table 2 provides evidence of convergent validity, as all items loaded over 0.70 on their factors, with an average variance extracted (AVE) for each factor greater than 0.50 (Bagozzi et al. 1991). Table 3 provides evidence of discriminant validity, as the AVE of each construct was greater than the shared variance between each pair of constructs (Fornell and Larcker 1981). The measurement model achieved good fit (χ2/df = 54.82/30 = 1.82; AGFI = 0.91; CFI = 0.99; RMSEA = 0.063; PNFI = 0.65).
Next, we tested the proposed structural model. To provide a more conservative test, we specified all the paths from the exogenous to the endogenous variables. The structural model achieved good fit (χ2/df = 79.81/49 = 1.63; AGFI = 0.90; CFI = 0.99; RMSEA = 0.055; PNFI = 0.61). Figure 3 summarizes the overall serial moderated mediation model results with standardized estimates. Table 4 also offers a more detailed account of the unstandardized estimates for the structural model.
In Hypothesis 3, we predicted that the interaction effect should first increase skepticism toward the environmental claims and then reduce credibility toward the company, with downstream negative effects on consumers brand evaluations and purchase intentions. We thus constructed bias-corrected confidence intervals with 10,000 bootstrapped samples (Hayes 2013) to test the significance of the indirect effect of environmental claims × external disconfirming information → green skepticism → corporate credibility → (a) brand attitudes; (b) purchase intentions. PROCESS model 6 was used as the modified bias-corrected bootstrapping protocol (Hayes 2013), including the interaction terms as the independent variable, and both environmental claims and external disconfirming information as covariates. For brand attitudes, the total indirect effect was negative and significant (B = − 0.18; 95% CI − 0.33 to − 0.04), and this was mainly due to the serial mediation passing through green skepticism and corporate credibility (B = − 0.12; 95% CI − 0.23 to − 0.02). The total indirect effect of the interaction was also significant when purchase intentions were the dependent variable in the serial mediation model (B = − 0.13; 95% CI − 0.25 to − 0.03).
Tests of Reverse Causality
Extant literature provides strong support for the contention that, for brands about which consumers do not hold pre-existing attitudes, consumers rely on available cues to inform their judgment (Gigerenzer and Gaissmaier 2011; Payne et al. 1991). As such, consumers form attributions of skepticism toward the claims, then use those attributions to inform overall perceptions of corporate credibility (Forehand and Grier 2003). Nevertheless, we tested for reverse causality, estimating a serial mediation model in which corporate credibility was an antecedent to skepticism toward the environmental claims. As we had expected, when considering corporate credibility as the first mediator and skepticism as the second mediator, the serial mediation path was not significant (B = − 0.01; 95% CI − 0.05 to 0.01).
Experiment 2 offers evidence that the levels of information specificity for both environmental claims and external disconfirming information interact with each other to affect skepticism toward the environmental claims and credibility about the source of the claims. H2A and H2B are thus supported. Moreover, our results provide evidence for a moderated serial mediation, such that the interactive effect is carried over by green skepticism first, and corporate credibility second. We also rule out the possibility of reverse causality, which is theoretically unlikely in a context in which consumers have no existing attitudes about the brand to guide their judgment and thus have to rely on the given information (in our case, the environmental claims). H3 is thus supported.