In the behavioral sciences it has long been a goal to identify variables that are systematically associated with people’s risk preferences. Yet, evidence concerning many “candidate correlates” (e.g., wealth, age, sex, education) has been mixed, because previous studies tended to focus on single variables (i.e., not taking into account many competing predictors) and to implement few operationalizations of risk preference. In our paper recently published in JPSP (see here for the PDF), we tackled this issue in a novel way: using specification curve analysis (SCA), we assembled all possible model specifications given the variables of our dataset, which resulted in over 1 million models (incl. simulation analyses). Thanks to our powerful sciCORE at the University of Basel, we could efficiently estimate these models using traditional OLS and Bayesian methods.

A key advantage of SCA is its possibility to visualize results of extensive modeling analyses transparently. The main findings indicated that a person’s sex and age have robust and consistent associations with people’s risk preferences, whereas other candidate correlates showed less consistent or no associations. The results also demonstrate the important role of construct operationalization when assessing people’s risk preferences: self-report measures picked up various associations with the proposed correlates, but behavioral measures largely failed to do so. In sum, we hope that our paper illustrates how exhaustive modeling analyses can provide conclusive answers to important theoretical issues in the behavioral sciences.

Frey, R., Richter, D., Schupp, J., Hertwig, R., & Mata, R. (2020). Identifying robust correlates of risk preference: A systematic approach using specification curve analysis. Journal of Personality and Social Psychology. doi:10.1037/pspp0000287 | PDF

First appeared on https://renatofrey.net/blog