Our research both develops the multi-proxy approach and applies it to substantive research questions. This includes methodological work on multi-proxy experimentation and behavioral domain mapping, alongside empirical studies that illustrate the approach and its value. Below, we highlight how each paper contributes to the development or application of the method, with links to publications, supplementary materials, and available data and code.
Methodological research
The value of multi-proxy experiments to study pro-environmental behavior
Introducing the multi-proxy approach. This paper presents the rationale for varying behavioral outcome proxies within an experiment. It explains how multi-proxy designs allow researchers to examine generalizability, estimate effect heterogeneity, and investigate proxy-level moderators.
Practical resources: The accompanying tutorial appendix provides a step-by-step guide to designing and analyzing multi-proxy experiments, including R and Mplus syntax and considerations for statistical power. Data, Mplus and R code are available in Research Box.
Mapping and Sampling Behavioral Domains: A Behavior-Centered Approach Applied to Environmentally Sustainable Consumption
Mapping and selecting behavioral proxies. This paper introduces a behavior-centered approach for identifying the psychological dimensions along which behaviors differ. Applied to 183 environmentally sustainable consumption behaviors, it identifies five broader dimensions and six descriptive subdomains. The resulting behavioral map can help researchers understand the scope of existing findings and select proxies systematically for future studies.
Practical resources: The Behavioral Domain Atlas allows researchers to explore behavioral characteristics and select proxies using random sampling, purposive sampling, or matched pairs.
Empirical applications
When pro-environmental behaviors are more widely adopted, penalties for non-adoption feel fairer
Prospective proxy-level moderation. In a preregistered multi-proxy experiment involving 2,401 participants and 81 pro-environmental choice pairs, rewards were generally perceived as fairer than penalties. However, their fairness advantage became smaller as the greener behavior was perceived as more widely adopted. The study illustrates how differences between behavioral proxies can reveal when experimental effects become stronger or weaker. It also demonstrates how multi-proxy research can build on earlier studies: data from a previous experiment (Study 2, Millet et al., 2022) were used to construct one of the measures of behavioral adoption, which served as a proxy-level moderator in the present study.
Practical resources: The appendix provides details of a Monte Carlo power simulation for the multi-proxy design. Preregistration, data and R code are available in Research Box.
The behavioral intervention “Positive Cueing”: Altering self-perception, increasing green awareness, or undermining the signaling value of costly green behavior?
Post-hoc proxy-level moderation. This study revisits the multi-proxy experiment reported in Study 2 of Millet et al. (2022). By combining the original experimental data with newly collected information about the identity signaling value of the behavioral proxies, it identifies a crossover interaction that was not apparent in the original analysis. The findings illustrate how adding proxy-level information to existing experimental data can generate new theoretical insights.
Practical resources: The appendices provide information on the identity signaling value of the 81 behavioral proxies and the combination of the original experimental data with additional proxy-level information. The preregistration, data and Mplus code are available in Research Box.
The limited impact of positive cueing on pro-environmental choices
Early multi-proxy application. This research investigates whether positive cueing increases pro-environmental choices across a broad set of behavioral outcomes. It provides an early application of the multi-proxy approach in Study 2, with these data subsequently used to investigate how the behavioral characteristic “identity signaling value” helps explain variation in the experimental effect (Millet & Weijters, 2023).
Practical resources: The appendices provide information on the validation and characteristics of the 81 behavioral proxies, as well as a power analysis based on the study results. Data are available in Research Box.