More green than gray? Toward a sustainable overview of environmental spillover effects: A Bayesian meta-analysis

In response to climate change, interventions have been implemented to encourage sustainable behavior. Such interventions may not only promote the target behavior but also increase (positive spillover) or reduce (negative spillover) non-targeted outcomes. This pre-registered meta-analysis integrated the experimental research on environmental spillover to update a previous meta-analysis (Maki et al., 2019). Database searches in several languages supplemented by searches to retrieve unpublished literature yielded 63 aggregated effect sizes from 38 studies and 29 articles (N = 26,613 unique participants). A three-level Bayesian meta-analysis provided weak support for no spillover on intentions and strong support for no spillover on behaviors. If spillover was present, it would likely be small and positive for intentions, δ = 0.15, 95% CrI [-0.01, 0.31], but negligible for behaviors, δ = 0.01, 95% CrI [-0.13, 0.16]. Positive spillover was most likely when interventions were autonomy-supportive (very strong evidence), provided a rationale (moderate to strong evidence), did not use financial (dis)incentives (weak to strong evidence), and addressed normative (extreme evidence) or a combination of normative and personal gain goals (strong evidence). Spillover was similar across research settings (moderate evidence) and partly across samples (weak to moderate evidence), which may suggest generalizability. To set standards for robust spillover research, we developed the Power-Reporting-Open science (PRO) guidelines. The Bayesian approach allows for robust conclusions and continuous updating with new evidence. We hope that this supports future revisions toward a sustainable overview of robust and high-powered spillover studies that independent researchers can easily update. •Weak evidence for null spillover effect on sustainable intentions.•Strong evidence for null spillover effect on sustainable behaviors.•Intervention characteristics can promote more positive spillover.•PRO (power, reporting, open science) sets standards for robust spillover research.•Bayesian approach allows easy updating when new evidence becomes available.