Max R. P. Grossmann

We present z-Tree unleashed, a novel approach and set of scripts to aid the implementation of computerized behavioral experiments outside the laboratory. z-Tree unleashed enables subjects to join the experiment using a web portal that requires no software apart from a web browser. Experimenters are likewise enabled to administer their experiments from anywhere in the world. Except for z-Tree itself, z-Tree unleashed is entirely based on free and open-source software. In this paper we give a high-level overview of z-Tree unleashed’s features and benefits and its design. We also show how to set up the server and demonstrate the steps required for conducting an entire experiment. We subsequently explain how to leverage the security and routing features of a virtual private network with z-Tree unleashed, enabling servers to securely run behind routers.

We study the relationship between information and paternalism. When is autonomy granted to an individual that is less or better informed, and if no autonomy is granted, what form will intervention take? We introduce the Estimation Game, a novel experimental design that varies the amount of ambiguity inherent in a binary lottery. Our analysis is concerned with the behavior of policymakers (“Choice Architects”). Based on the utilitarian standard, we introduce a simple majoritarian standard that relies on implementing full information choice counterfactuals. This standard is approximately utilitarian, can be falsified and is psychologically plausible. It is also a novel normative microfoundation of the median voter theorem in the case of two policies. We contrast Mill's early utilitarianism with his later classical liberalism and show that both are equivalent under full information and without externalities. We conduct experiments of the Estimation Game with a Choice Architect to examine the extent to which Chooser knowledge matters in terms of restrictions on the Chooser's freedom of choice. More information leads to fewer interventions on the extensive margin. However, which option is imposed is a matter of personal Choice Architect preference, not majoritarian consideration. We interpret these findings as a “moral license:” the act of helping a clueless individual induces Choice Architects to then impose their own tastes, even when adjusting for the false consensus bias that is prevalent in beliefs. We also provide a classification of experimental subjects. Only 5\% of Choice Architects intervene irrespective of the Chooser's information structure.

Large Language Models (LLMs) have the potential to profoundly transform and enrich experimental economic research. We propose a new software framework, "alter_ego", which makes it easy to design experiments between LLMs and to integrate LLMs into oTree-based web experiments with human subjects. Our toolkit is freely available at github.com/mrpg/ego. To illustrate, we run simple differently framed prisoner's dilemmas with interacting machines as well as with human-machine interaction.

The importance of deliberation and time to consider one's decision have been emphasized in fields ranging from healthcare to firearms regulation. However, little is known about waiting periods' political economy and their effects on other policy features. We examine these issues using a survey experiment. First, we investigate whether mandatory waiting periods lead to increased respect for others' decisions. We leverage an experimental design in which "Choice Architects" can make rules for future subjects to investigate the effect of deliberation on paternalism. Here, deliberation is a given. Do Choice Architects show increased respect for more considered decisions? Second, we consider Choice Architects’ endogenous choice of waiting periods. Simply put, does the possibility of giving someone else additional “time to think” reduce the willingness to impose hard-nosed restrictions? This project aims to elucidate determinants of public policy given non-traditional measures such as mandatory waiting periods.

This thesis focuses on two extensions to the stochastic volatility (SV) model: First, explanatory variables are introduced in the form of a low-frequency mixed-data sampling (MIDAS) term along with the daily autoregressive component of the SV model. Second, we consider a model in which an additional autoregressive term is introduced, but it evolves in a low-frequency manner compared to the daily term. Additional macro variables permit the identification of this supplementary term. Chapter 1 introduces the basic models and provides an overview of previous approaches. Chapter 2 presents the new models and discusses some of their properties. All SV models are estimated using the Kálmán filter; moreover, it is shown how volatility forecasts can be obtained. In chapter 3, we study the models using Monte Carlo simulations. In chapter 4, we apply the models to financial series. Finally, we summarize the results in chapter 5. Proofs, additional figures and tables, and the code of the reference implementation are provided.

See here for further information. The previous world record computation (from 2013, also by me) can be downloaded here.