Data and scripts for the "Central Tendency Bias in Belief Elicitation" by Paolo Crosetto, Antonio Filippin, Peter Katuščák and John Smith
This repository contains original data and all scripts to recreate figures and tables from the paper Central Tendency Bias in Belief Elicitation, forthcoming in the Journal of Economic Psychology.
In the paper we document that players instructed about a uniform distribution and having passed control questions about it still hold central-single peaked subjective beliefs. This result is robust and cannot be imputed to mistakes and misunderstandings, the chosen elicitation mechanism, risk aversion, hedging,nor inconsistency. We provide indicative evidence that such subject beliefs can be rationalized by well-defined subjective beliefs that differ from the objective truth.
The paper can be find here : <<add JoEcoPsy link when available>>.
The working paper version of the paper can be found here: <<add preprint link when availble>>
The same dataset is provided here in two versions.
- as a Stata .dta file,
ctb_final.dta. This file is the one used by both our scripts to generate the figures and tables, and contains variable labels in Stata format - as a .csv file,
ctb_final.csv. This file contains the same data and is given here for interoperability with other statistical packages.
We conduct an experiment where 379 subjects engage in a first-price auction. Subjects are told the distribution of their random-draw opponent’s strategy: a uniform distribution on a range of possible integer bids. The subjects then place a bid in the auction. Subsequently, subjects are given an incentivized elicitation of their beliefs of the random-draw opponent’s strategy and their beliefs of winning the auction. The random-draw opponent’s bid is determined by a physical draw of a token at the end of the experiment.
Subjects go through a series of control questions to ensure understanding of the instructions for the experiment and of the uniform distribution.
After submitting a bid in the auction, subjects report their beliefs of the distribution of their random-draw opponent’s strategy. The strategy space of the random-draw opponent is divided into 5 bins of equal size. Subjects allocate probability weights into the bins. An automatic checker verifies that these amounts correctly sum to 100. Screenshots of the elicitation procedure are provided in Appendix D.6 The maximum that subjects can earn on this task is 20 ECUs, where 1 ECU = 0.20 Euro.
This dataset contains all the data from our experiment. It oontains data from 379 subjects. Data include: demographics, the probabilities assigned to each bin, the auction treatments, and number of mistakes and time spent on the control questions.
The variables are described in detail in the variable_description.txt file.
To generate all figures, run Generate_figure.R in R. The figures are also stored in the /Figures folder in this repo.
To generate the raw data used in all tables, run Generate_tables.do in Stata13 or higher. Not all results have the exact same formatting as the final tables in the paper.