Jan Feld | Design Spaces & Sex Differences in Risk Attitudes
Design Spaces and Sex Differences in Risk Attitudes
Event details
Design Spaces and Sex Differences in Risk Attitudes
Many findings in the social sciences appear mixed or inconsistent. A key reason is that behavior is context-dependent, while most research designs capture only a narrow slice of possible contexts. I propose a four-step approach to move beyond this dilemma. First, map the complexity of reality using a design space that lays out all potentially relevant dimensions of heterogeneity. Second, populate this design space with estimates from existing datasets, making transparent both what we know and where the gaps are. Third, use machine learning to identify which dimensions best predict outcomes while limiting overfitting. Finally, validate these dimensions with experiments. I apply this approach to the question of whether men are more willing to take risks than women. My design space includes seven dimensions (such as incentives and culture), yielding 504 possible elements. Using data from 94 countries, I am able to populate 207 of them. Machine learning results suggest one dimension matters most: whether the risk situation is in the gain and mixed domains (where men are generally more risk-seeking) or in the loss domain (where women appear more risk-seeking). I am now working on the final step—designing experiments to validate these findings—and look forward to your feedback.
Expression of interest
About the speaker
Jan Feld is a Senior Lecturer in economics at Victoria University of Wellington. His research focuses on economics of education, labour economics, and meta-science.
Learn more about him on his personal webpage.