Working Papers

Hendrik Rommeswinkel and Takashi Ui. Alpha divergence representations of dynamically consistent belief selection. 2026.

PDF

Alpha Divergence Representations of Dynamically Consistent Belief Selection

Abstract

We provide a normative foundation for the class of alpha divergences in a parsimonious model of belief selection. If a decision maker is to update beliefs dynamically consistently --- in the sense that additional information cannot reverse preferences over distributions that already agreed on the relevant states --- has convex preferences, and satisfies a separability condition across subsets of states, then she ought to choose her beliefs by minimizing the alpha divergence from a unique subjective reference distribution to the chosen distribution. We show that this axiomatically justifies a wide range of statistical estimators and practices.

PDF Slides

Yukihiro Funaki, Carol Ma, and Hendrik Rommeswinkel. Decision times in the luce model: an experiment. 2025.

PDF

Decision Times in the Luce Model: An Experiment

Abstract

We test a new model of decision times based on the decision probabilities of the Luce model of stochastic choice. The model extends the so-called Hick's law of reaction times. According to the model, expected decision times are represented by the Tsallis entropy of the decision probabilities. To test the prediction, we conduct a choice experiment with a value task under risk and time pressure and collect decision times for over 37 000 decisions from household item lotteries. For each subject in our experiment we find a statistically significant rank correlation between Tsallis entropy and decision time. With a mean absolute rank correlation of 0.32, the results are also economically significant, suggesting decision times can be used to detect ``difficult'' decisions in which the value difference of alternatives is small.

PDF Slides

Hendrik Rommeswinkel. Preference for verifiability. 2024.

PDF arXiv

Preference for Verifiability

Abstract

Decision makers may face situations in which they cannot observe the outcomes that result from their actions. In such decisions, motivations other than the expected utility of outcomes may play a role. The present paper axiomatically characterizes a decision model in which the decision maker cares about whether it can be ex post verified that a good outcome has been achieved. Preferences over acts uniquely characterize a set of events that the decision maker expects to be able to verify in case they occur. The decision maker chooses the act that maximizes the expected utility across verifiable events of the worst possible outcome that may have occurred. For example, a firm choosing between different carbon emission reduction technologies may find some technologies to leave ex post more uncertainty about the level of emission reduction than other technologies. The firm may care about proving to its stakeholders that a certain amount of carbon reduction has been achieved and may employ privately obtained evidence to do so. It may choose in expectation less efficient technologies if the achieved carbon reduction is better verifiable using the expected future evidence.

PDF Slides

arXiv

Min-Feng Lu and Hendrik Rommeswinkel. Preference for flexibility from incomplete resolution of uncertainty. 2023.

PDF

Preference for Flexibility from Incomplete Resolution of Uncertainty

Abstract

We analyze information and flexibility preferences in a Savage framework. Preference for flexibility is commonly modeled with subjective states that may be completely unrelated to the objective state space. We instead adopt a framework in which preference for flexibility is due to incomplete resolution of uncertainty from the objective state space before making a choice from a menu. We characterize both state-dependent and state-independent representations of (indirect) expected utility maximization. In both cases, the decision relevant hidden information that the decision maker expects to obtain prior to a choice from the menu is uniquely identified.

PDF Slides

Tzu-Ying Cheng, Hendrik Rommeswinkel, Guan-Yuan Wang, and Tsai-Shan Yang. Measuring consumer freedom. 2021.

PDF

Measuring Consumer Freedom

Abstract

This paper axiomatizes an index of consumer freedom of choice and applies it to U.S. consumer data from 2004 to 2017. The R{\'e}nyi (1961) entropy arises naturally as an index of the residual of welfare that remains after accounting for total consumption, the price level, and inequality. We show that this index can be interpreted as the freedom of choice of a representative consumer allocating a dollar of expenditure to goods. When applied to data, we find that the freedom of choice index was stagnant between 2004 and 2009 but substantially increased thereafter. This increase is driven by an increase in the total number of goods consumed as well as a more even distribution of expenditure across goods.

PDF Slides

Hendrik Rommeswinkel and Hao Wu. Monetary policy behind the veil of ignorance. 2020.

PDF

Monetary Policy Behind the Veil of Ignorance

Abstract

We analyze the problem of the choice of a central bank constitution. We model the decision problem as a choice behind a veil of ignorance in which the policy maker only receives information about predicted behavior of agents under different policies. We derive a policy criterion that only relies on this information and is robust to misspecification of the utilitarian welfare function. We compare price stabilization, output stabilization, and inflation targeting in a standard new Keynesian model with Calvo price staggering. Under our policy criterion, the policy maker perceives a tradeoff between output stabilization and price stabilization. The reason is that in the absence of knowledge about a utilitarian welfare function, stabilizing the natural level of output is not normatively desirable. Despite these findings, we show that for numerical approximations of parameters the utilitarian results of the standard new Keynesian model are suprisingly robust — the optimal policy under our criterion places a strong emphasis on price stability.

PDF

Hendrik Rommeswinkel. Measuring freedom in games. 2019.

PDF

Measuring Freedom in Games

Abstract

The paper axiomatizes a measure of freedom for game theoretic settings. The central idea of the measure is that freedom is increasing in the degree to which an agent's outcomes are determined by the agent's preferences. The measure is characterized by rational, nonconsequentialist preferences of an impartial observer over games endowed with the observer's beliefs over actions. The measure generalizes several measures from the opportunity set based freedom literature to situations where agents interact. This allows freedom to be measured in general economic models and thus derive policy recommendation based on the freedom instead of the welfare of agents. To illustrate this, optimal libertarian income tax progression policies are derived in a production economy with heterogeneous agents.

PDF Slides

Claudia Neri and Hendrik Rommeswinkel. Decision rights: freedom, power, interference. 2019.

PDF

Decision Rights: Freedom, Power, Interference

Abstract

We propose a general model of decision-rights allocation and choice, formulated in the context of a dynamic psychological game. Decision rights are valued not only according to the value of the outcomes, but also according to the procedure by which outcomes are achieved. We introduce freedom, power, and interference as such procedural motivations. In a novel laboratory experiment, we separately measure freedom, power, and interference preferences. Interference aversion best explains participants’ behavior. Most participants value decision rights not because they enjoy having freedom of choice or power over others, but because they dislike letting others interfere in their outcomes.

PDF

Hendrik Rommeswinkel. Procedural mixture sets. 2019. (R&R at Theoretical Economics)

PDF
Award arXiv

Procedural Mixture Sets

Abstract

The paper characterizes the Shannon (1984) and Tsallis (1988) entropies in a standard framework of decision theory, mixture sets. Procedural mixture sets are introduced as a variant of mixture sets in which it is not necessarily true that a mixture of two identical elements yields the same element. This allows the process of mixing itself to have an intrinsic value. The paper proves the surprising result that simply imposing the standard axioms of von Neumann-Morgenstern on preferences on a procedural mixture set yields the entropy as a representation of procedural value. An application to the relation between choice probabilities and decision times in decision processes elucidates the difficulty of extending the drift-diffusion model to multi-alternative choice.

Awards: https://sites.google.com/site/rudconference/jaffray-lectures

PDF Award

arXiv