Ku, J. (2025) Sequential Search under Range-Dependent Attribute Weighting (Job Market Paper)
Abstract: Sequential search has typically been explored with single attribute objects and without recall, despite both empirical and experimental evidence for the latter. Neoclassical sequential search models rule out recall, which is defined as taking home an alternative that is not the one that most recently arrived when the decision maker stops searching. I design a new search experiment where objects have multiple attributes and recall is allowed. The model extends opposing behavioral models, Köszegi & Szeidl (2013) and Bushong, Rabin, and Schwartzstein (2021), into the sequential search environment. It allows for idiosyncratic and time varying valuations of multi-attribute objects, leading to predictions of recall in specific situations. Data indicates multiple types of searching agents within the subject pool (neoclassical, behavioral, and others), with recall used 39% of the time on average across all participants (similar to previous studies), with rates of recall of 33% for subjects that do not reject the neoclassical specification at a 5% level and 53% for relative thinking subjects (p<0.001). Structural estimates reject a neoclassical specification in favor of a behavioral specification for the representative agent (p<0.001).
Presented at:
BABEEW (Bay Area Behavioral and Experimental Economics Workshop) 2024. University of California, Davis.
2023 North American ESA Meeting (Economic Science Association). University of North Carolina at Charlotte.
Ku, J. (2025) Range-Dependent Attribute Weighting: An Experiment on Spreading and Consolidating (working paper)
Abstract: The range of an attribute's outcomes in a choice set can possibly change its relative importance. There are two prominent and conflicting theories of range-dependent attribute weighting, the focusing model of Köszegi and Szeidl (2013) and the relative thinking model of Bushong, Rabin, and Schwarzstein (2021); many researchers have tested the two models against each other by means of adding irrelevant alternatives/decoys. Extending the experiment in Somerville (2022) to an untested context, I derive predictions from these two theories and design and run an experiment in which subjects choose from pairs of bundles where the ``number" of advantages or disadvantages vary. No statistically significant effects were observed, likely due to the limited sample size. Results suggest that consolidating disadvantages might decrease the attractiveness of an option (p<0.01), aligning with one of the predictions of the Focusing model.
Presented at:
BABEEW (Bay Area Behavioral and Experimental Economics Workshop) 2023. San Jose State University.