“Less Is More”: Decision-Making under Increasingly Complex Ambiguity

Xueting Yang

Abstract: This paper examines whether increasingly complex decisions under ambiguity are better described by more complex behavioural models. Using an allocation experiment, we vary structural complexity through the number of possible outcomes and decision stages and compare Expected Utility (EU), αMaxMin (AMM), and Choquet Expected Utility (CEU). EU is the most frequently selected model according to AICc in all three ambiguous-lottery parts, although the strength of its relative advantage varies across tasks, with CEU remaining competitive in the most complex part. Estimated risk-attitude parameters exhibit a relatively stable central pattern across tasks, whereas ambiguity-attitude parameters vary more noticeably across task structures. Overall, the results show that explicitly incorporating ambiguity attitudes does not uniformly improve penalized model fit as structural complexity increases. Rather, the empirical value of more flexible ambiguity models depends on the structure of the choice environment. The findings highlight the importance of evaluating model complexity empirically rather than assuming that structurally more complex decisions necessarily require more flexible behavioural models.

JEL codes: D81, D91, C91

Keywords: Ambiguity, Complexity, Allocation Elicitation Method, Model Fitting, Experiments

Yang2: allresults.xls (MS Excel , 113kb)