Risk Preferences Beyond Expected Utility Theory: Theoretical and Experimental Approaches

This thesis contributes to the development of theoretical and experimental approaches that extend the analysis of risk preferences beyond the classical concept of expected utility theory. The thesis contains four main chapters.The first two chapters are concerned with extensions and applications of the most promising descriptive theories of behavior under risk. In these models it is assumed that the assessment of alternatives partly depends on a comparison with a precisely specified reference point. The remaining two chapters propose experimental methods that are to a considerable degree independent of a specific model. They can therefore be applied in cases in which the correct model of behavior is uncertain or in which several subpopulations are best described by different models.Chapter 1 analyzes how higher-order risk preferences, such as prudence and temperance, form in models of reference dependence. In these models the reference point is shaped by recent expectations. General results are derived, which show that the pattern of higher-order risk preferences differs fundamentally in models of reference dependence compared to classical models of expected utility theory. It is shown that some empirical puzzles concerning precautionary saving and insurance demand can be explained by reference-dependent preferences. Chapter 2 discusses an extension of reference-dependent preferences to anticipatory utility. A new model of multi-attribute utility in this context is developed and compared to the most prominent alternative. The new model has the advantage that no restriction on the separability of attributes is necessary. Chapter 3 proposes a new method to experimentally measure the intensity of risk aversion in laboratory experiments. This method is based on the well-established theoretical concepts of mean-preserving spreads and mean-utility-preserving spreads. It is thus applicable in a wide range of models and does not depend on restrictive assumptions. The proposed method is experimentally compared to the most widely employed classical method, which requires that the assumptions of expected utility theory are met. It is shown that the methodological concerns are empirically relevant when a classification of subjects concerning their intensity of risk aversion is conducted. Chapter 4 describes an experimental measurement of higher-order risk preferences. In order to assess whether risk preferences differ for decisions over gains and losses the experiment was designed to allow for real monetary losses of subjects. It is shown that subjects are on average risk averse, prudent, and temperate. Risk preferences of different order are highly correlated. Risk preferences seem not to differ substantially for decisions in gains and in losses.