Experimental findings often rely on a single outcome or stimulus. Multi-proxy experiments deliberately vary these proxies instead of keeping them fixed. This allows researchers to examine not only whether an effect generalizes across proxies, but also how much it varies and what explains that variation.
Suppose you manipulate X and measure its effect on a broad construct such as sustainable behavior. A traditional experiment typically selects one or a small set of behavioral outcomes and keeps them fixed. A multi-proxy experiment instead uses a large pool of behavioral outcomes and randomly assigns participants to one of them.
You learn what happened for that particular operationalization.
Participants receive one proxy at random, allowing the experimental effect itself to vary across proxies.
The central value of the approach is not merely “more measures.” It changes what an experiment can tell you.
Assess the robustness of the experimental effect across many proxies rather than basing conclusions on a single proxy.
Observe heterogeneity across proxies. An effect can be stronger for some proxies, weaker for others, or even reverse direction.
Measure the characteristics of each proxy and test whether these characteristics explain why the experimental effect is stronger for some proxies than for others.
Multilevel approaches have long been used to model variation across stimuli. Multi-proxy experiments extend that logic to the operationalization of the dependent variable itself: the behavioral outcome proxy is sampled rather than fixed.
Are penalties for environmentally harmful choices always perceived as less fair than rewards for greener alternatives? Our recent research suggests that the answer depends in part on how common the greener behavior is.
In a multi-proxy experiment involving more than 2,400 participants and 81 different environmental behaviors, we examined how people judge the fairness of rewards versus penalties. Rewards were generally perceived as fairer, but their advantage became smaller as the greener behavior became more widely adopted.
The multi-proxy approach was key to uncovering this pattern. Rather than studying a single environmental choice, we systematically varied the behavior targeted by the policy. This allowed us to examine whether the overall effect generalized across behaviors, how much it varied, and whether characteristics of the behaviors could help explain that variation.
The findings illustrate how variation across proxies can reveal when experimental effects become stronger or weaker, offering insights that might remain hidden in experiments relying on only one or a few fixed behaviors.
Explore this and other multi-proxy studies →A multi-proxy experiment requires a pool of suitable proxies. Which proxies researchers select matters for what they can learn from an experiment, including whether an effect generalizes, how much it varies across proxies, and what might explain that variation. Different research goals may therefore call for different ways of selecting proxies.