Speaker: André Couder, Tilburg University
Title: Dynamic Optimal Consumption for an Agent with Reference-Dependent Preferences, a Study
Abstract: Empirical evidence challenges Life-Cycle theory in showing that agents smooth their consumption responses to shocks; a puzzle known as excess smoothness and excess sensitivity. To comply with these dynamics, early models have been complemented with behavioral features such as habit formation or hyperbolic discounting, causing excessive flexibility and loss of explanatory power.
The current work proposes a dynamic intra-period reference-dependence consumption model inspired by Prospect-Theory, that retains time additive preferences. In the model, the rational agent forms a stochastic consumption reference point for the current period before observing the price state in the economy. Importantly, the agent does not update her beliefs about the price of the representative good, which is stationarily distributed. This permits to separate the reference process of the intertemporal optimization.
The model captures smoothed out consumption responses to different types of wealth shocks; transitory, permanent, anticipated and unanticipated; by linking optimal consumption choices to the dynamic risk of binding constraints, which would push the reference point outside the feasible consumption set and evolve with income flow.
Although calibrations and empirical analysis remain to be done, early simulations hint that the model could reconcile Life-Cycle theory with the aforementioned empirical puzzles, without needing more ambitious behavioral assumptions.
Speaker: Wassim Zaazaa, Tilburg University
Title: Eliciting risk preferences: a comparative analysis of stochastic choice models
Abstract: I present a comparative analysis of five different stochastic choice models. These include four different versions of the random utility model (RUM) and the random parameter model (RPM). Using empirical data, I examine individual irrationality and assess how each model reacts to it. I argue that the `standard’ RUM (based on expected utility) performs worst in terms of providing reliable estimates. Determining the best model, however, depends on whether the user prioritizes reliability or likelihood fit. Furthermore, I find that the tendency for reflective thinking helps explain the extent to which individuals behave irrationally, with those demonstrating greater reflective thinking – and arguably higher cognitive ability – making more precise choices, in general.
In my simulation study, I find that all models systematically underestimate the true risk aversion level when the models’ i.i.d. assumption is violated. The assumption is violated by ruling out multiple switching behavior – a restriction that is often enforced in lottery choice tasks (see, e.g., Holt and Laury (2002), Andersen et al. (2008) and Garagnani (2023). The underestimation may lead to individuals appearing more risk-seeking than they truly are, leading to for example suboptimal (individual) investment decisions. I conclude by emphasizing the importance of carefully selecting lottery choice tasks for risk preference elicitation using stochastic choice models.