Portfolio Choice Under Uncertainty

This thesis is about portfolio choice under ambiguity and risk. At its core
is an experiment and a simulation, both concerning portfolio choice. The
experiment is under ambiguity, in which the the probabilities of the states are
not known to the subjects. We tested two multiple prior preference theories
(MEU and _-MEU), both of which are fit significantly better than Expected
Utility (EU) for around one third of the subjects, and better than Mean-
Variance (MV) for the majority of the subjects. We also find that subjects
have heterogenous beliefs about ambiguity, but on average they do a good
job in guessing the true probabilities. The simulation is in the context of risk.
Our interest here is in the specification of the stochastic process underlying
our observations. The simulation led to a surprising result - the maximum
likelihood estimation may suggest the wrong specification.