Structural Analyses of Behavioral Errors: The Case of Risk and Time Preferences

This thesis revives the interest in behavioral errors by evaluating their application in the decision making under risk in the first essay and observing their association to the econometric modeling of decision heuristics in time preferences in the second essay. The first essay considers task complexity as the underlying source of randomness in the decision making under risk. We evaluate three heteroskedastic models that specify the disturbance’s standard deviation as a function of ad hoc measures of task complexity. Our contribution is to adapt a stochastic model from the consumer behavior discipline to binary choice tasks over lotteries, which is the most common experimental method to elicit risk preferences. Our empirical results emphasize the importance of accommodating the impact of task complexity on behavioral errors by rejecting homoskedasticity in favor of at least two heteroskedastic models. Our analyses suggest that the models’ statistical goodness-of-fit is contributed by their ability to capture broader risk behaviors at individual level, which is useful for distinguishing decision makers. The second essay critically evaluates the decision heuristic models, which have been claimed to have better accuracy in explaining the individual’s time preferences than the structural discounting models. Our contention is that the seemingly superior performance of the heuristic discounting models is irrelevant to their ability to capture particular aspects of discounting behaviors but more an artefact of their oversimplified econometrics modeling. Their specification exhibits a linear approximation of a finite mixture of two behavioral error stories, which is not comparable to the structural discounting specifications with the representative agent model. The heuristic specifications can then be used as a simple diagnostic device for choosing between behavioral error specifications for the structural models. Finally, we contribute to the economic literature by showing that neglecting utility curvature completely reverses inferences about the relative performance of heuristic and structural discounting models