Multi-proxy experiments need a pool of suitable proxies. Our behavior-centered work addresses an important design question: how can researchers map a behavioral domain and select proxies systematically rather than by convenience?
Start with the domain you want to represent
Broad constructs such as environmentally sustainable consumption, healthy consumption, or prosocial behavior can be expressed through many different behaviors. The construct is the abstract concept, while the behavioral domain comprises the broader universe of behaviors through which it can be expressed. The proxies selected for an experiment represent a subset of that domain.
Importantly, behaviors within the same domain can differ substantially in their psychological characteristics. For example, sustainable behaviors can differ considerably in how much effort they are perceived to require, or the extent to which they are already socially expected.
Map the domain before sampling it
One way to make proxy selection more systematic is to treat behaviors themselves as objects of analysis. Independent respondents evaluate a broad range of behaviors on relevant psychological characteristics. In our study of sustainable consumption, for example, we collected ratings of 183 behaviors, ranging from recycling cans to reducing air travel. Each behavior was evaluated on a variety of characteristics such as how much effort it requires, how socially expected it is, and how strongly it is associated with moral judgments.
These ratings are aggregated to create a psychological profile for each behavior (i.e., a set of scores on the different characteristics). These profiles can then be analyzed to identify broader dimensions along which behaviors differ. Using exploratory factor analysis, we identified five broader dimensions that capture key differences between the 183 behaviors:
Choose proxies to match the question
Random Sampling
Randomly select proxies from a clearly defined set of behaviors when you want to test if the effect generalizes across that set.
Best for generalizing across behaviors.Purposive Sampling
Select proxies that cover a broad range of one particular behavioral characteristic while limiting its overlap with other characteristics.
Best for testing a prespecified proxy-level moderator.Matched pair
Choose two proxies that differ on a focal characteristic while remaining similar on other characteristics.
Best for a controlled two-proxy comparison.Explore the Behavioral Domain Atlas
To put this approach into practice, we developed the Behavioral Domain Atlas, an interactive tool that allows you to explore the psychological characteristics of 183 environmentally sustainable consumption behaviors and systematically select proxies that match your research goals.
Behavioral Domain Atlas →