Date of Award

Summer 8-1-2017

Degree Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

James C. Cox

Second Advisor

Glenn W. Harrison

Third Advisor

Vjollca Sadiraj

Fourth Advisor

Rusty Tchernis


The dissertation looks at three topics that involve experimental economics methods or individual decision making under risk: how do people make educational decisions when facing the risk of drop out; which models best characterize individuals' decision processes under risk; how can physicians improve discharge decisions to reduce the risk of unplanned readmissions.

In the first chapter, I introduce the risk of dropout into Spence’s job market signaling model and test the modified model in the laboratory. I look at equilibria in the labor market and discuss the refinement based on the Cho-Kreps Intuitive Criterion. I derive the condition under which a separating equilibrium is the only perfect Bayesian equilibrium that survives the refinement and discuss the effects of workers' risk preferences on these equilibrium predictions. The data from lab experiments show that the market reaches the separating equilibrium more often when it is the only intuitive equilibrium. I also observe that, when the share of the low-ability type in the worker population decreases, or the cost to pursue a degree increases, the size of the wage premium for having the degree generally decreases. In the experiments, I use binary lottery tasks to elicit subjects' risk preferences to explain their strategies in the signaling games, and the analyses partially confirm the prediction that more risk-averse individuals pursue a higher degree less frequently in the presence of dropout risks. In the second chapter, as part of a joint project with Dr. Glenn W. Harrison and Dr. Rusty Tchernis, we apply the Bayesian econometric method to estimation of individual preferences under risk. We estimate a mixture model of Expected Utility Theory and Cumulative Prospect Theory using both simulated and observed binary lottery choices. We develop Markov Chain Monte Carlo algorithms to sample from the posterior distribution of parameters in the mixture model and compare the performances of different algorithms. The algorithms generally recover the true parameters used in the simulation, although some algorithms outperformed others in terms of efficiency. We also apply the algorithms to estimation using actual choice data. We find that 56.5% of the subjects can be characterized as consistent with Expected Utility Theory and 43.5% with Cumulative Prospect Theory. We find modest risk aversion among Expected Utility maximizers, and overweighting on the probabilities of extreme outcomes with very mild loss aversion among Cumulative Prospective Utility maximizers.

In the third chapter, coauthored with Dr. Ira L. Leeds, Dr. Vjollca Sadiraj, Dr. James C. Cox, Dr. Timothy M. Pawlik, Dr. Kurt E. Schnier and Dr. John F. Sweeney, we sought to define the association between information used for hospital discharge and patients' subsequent risk of unplanned readmission. De-identified data for patients from a tertiary academic medical center's surgical services were analyzed using a time-to-event model to identify criteria that statistically explained the timing of discharges. The data were subsequently used to develop a prediction model of unplanned hospital readmissions. Comparison of discharge behaviors versus the predictive readmission model suggested important discordance with certain clinical measures not being accounted for to optimize discharges. We suggest that decision-support tools for discharge may utilize variables that are not routinely considered by healthcare providers.