Tilman Slembeck - Current Research

Welcome!  I am glad that you are interested in my research.

Abstracts of Papers in Progress - © by Tilman Slembeck and Coauthors

Overview

• "Monty Hall‘s Three Doors: Does the Choice Anomaly Survive Communication and Competition?" - see abstract
  
  
• "Noisy Decision Makers - On Errors, Noise and Inconsistencies in Economic Behaviour" - see abstract

 
• "Learning Preferences Over Risky Prospects - Experimental Evidence From Three Countries" - see abstract

 
• "Reinforcement Learning in Ultimatum Bargaining - A Multi-Armed Bandit Experiment" - see abstract

 
• "The Curse of Private Information - Experimental Evidence" - see abstract

For abstracts of completed papers click here

Monty Hall‘s Three Doors: Does the Choice Anomaly Survive Communication and Competition?

In a recent paper in the American Economic Review, Daniel Friedman (1998) studies a choice anomaly (the so-called Three-Door-Anomaly) in individual decision making that appears to be robust with regard to several learning environments.Although convergence toward rational behavior is rather weak in his experiment, Friedman (p. 941) asserts that "every choice 'anomaly' can be greatly diminished or entirely eliminated in appropriately structured learning environments."
In a new experiment we find no evidence for this assertion in "institution-free" environments even when the task is repeated many times. However, there is strong evidence that social and economic institutions -such as communication and competition- are able to eliminate the Three-Door-Anomaly entirely.

Noisy Decision Makers On Errors, Noise and Inconsistencies in Economic Behaviour

Although economists generally agree that real people at least sometimes err and laboratory experiments provide much evidence for "erroneous" and inconsistent behaviour, noisy decision-making is rarely studied in detail. In most theories errors do not matter and in experiments they are often regarded as a annoying factor that causes noisy data. Some attempts have been made to model noisy behaviour in experiments but a general approach that also addresses the sources of noise is missing. This paper outlines a general taxonomy of human error by identifying three main types of noisy behaviour: task dependent noise, learning noise and white noise. These types are formalized in a general framework and their relevance is discussed in the light of experimental experimental data.

Learning Preferences Over Risky Prospects: Experimental Evidence From Three Countries

Several experimental studies on decision making under risk have presented evidence that subjects do not maximise expected utility, but are risk averse. Also, when the task is repeated, many choices are reversed so that preferences appear to be unstable. These findings are interpreted as resulting from "errors" (Hey and Orme, 1994). This paper asks whether preferences do settle and "errors" are reduced when appropriate conditions for learning are provided. Based on the Hey/Orme study, a new experiment explores the roles of three factors: feedback, representation of the task, and differences in expected values. Evidence suggests that preferences are much more stable when subjects are allowed to learn, and that choices are more in accordance with expected utility theory.

Reinforcement Learning in Ultimatum Bargaining: A Multi-Armed Bandit Experiment

Reinforcement learning is prominent among the non-cognitive approaches in the current literature on learning in economics. Some authors have argued that this type of learning is the key to the dynamics of ultimatum bargaining. Experimental evidence is mixed. This paper presents an experiment where the Ultimatum Game is modelled as a multi-armed bandit problem in order to take the non-cognitive nature of reinforcement learning literally, and to provide a direct test of this theory. [Data show that play does not converge to what is usually observed in experiments so that the reinforcement learning hypothesis can be rejected to account for ultimatum play by itself. Additional hypotheses are discussed.]

The Curse of Private Information: Experimental Evidence

In many situations private information is assumed to be advantageous for the informed party. This paper presents evidence for the case where private information is disadvantageous, i.e. subjects in a laboratory experiment are willing to bear costs for not receiving private information, and can sometimes coordinate on a Pareto-efficient equilibrium in order to overcome the curse. It is argued that the "curse of private information" is the common rationale of several well-known problems under asymmetric information, such as principal agent problems, moral hazard problems, Akerlof's lemon problem, and Buchanan's veil of ignorance, and that establishing Pareto-efficiency can require costly signalling or coordination in order to overcome the curse.

Go to  Top of this page, my Home Page, or see my Favourite Links for (experimental) economists.

Last update: 27 April 2000, page created by T.S. Sept. 1995