Allures of Power Data

Description

The experiment code, data collected from subjects, and analysis files posted here were used in a paper with the following abstract: This paper investigates intrinsic preferences for power over others' outcomes using a laboratory experiment. I decompose the intrinsic value of power into three components: self-recognition (the power-holder's awareness of her own power), other-recognition (others' awareness of the power-holder's power), and having options over others' outcomes. Using a combination of experimental games, including a novel "Points Power Game" in which subjects can pay to control an outcome that has no material consequences for their partner, I find that a substantial majority (86%) of subjects exhibit behavior consistent with intrinsic preferences for power by at least one criterion. While intrinsic preferences for power are widespread, I find little evidence that self- or other-recognition increase willingness to pay for power. The estimates are precise enough to rule out moderate effects, although this may reflect limitations of the recognition manipulation rather than a true absence of effects. In contrast, I find suggestive evidence that subjects intrinsically value having options over others' outcomes: 49% of subjects purchased more power than needed to achieve their chosen outcome. Subjects appear to place greater intrinsic value on options to increase others' outcomes than on options to decrease them. These findings offer preliminary evidence that the mere possession of options over others carries value beyond what distributional preferences alone can explain, although alternative explanations cannot be fully ruled out. The experiment files is the z-tree file "experiment_code.ztt." The data is in the Excel spreadsheet "data.xlsx." The main analysis code is the Stata file "stata_analysis.do," which requires "common_round.dta" and "stata_setup.do" to run. The Bayesian analysis, kindly provided by an anonymous reviewer, is in the Python file "Bayesian.py."

Steps to reproduce

The experiment was conducted in-person on computers using the "experiment_code.ztt" file. Instructions were displayed on screen by the code. To replicate the analysis, first download "data.xlsx" and "common_rounds.dta." Then run the "stata_setup.do," "Bayesian.py" and "stata_analysis.do" files. "stata_setup.do" tests whether round order matters. It also is necessary to run "stata_analysis.do" because it sorts the rounds for each subject, and divides Binary Choice Game rounds into common rounds and Monetary-Choice-Game-related rounds (the file needs "common_rounds.dta" to identify the common rounds). "Bayesian.py" produces the Bayes factor for the effect of information treatment assignment on spending in the Monetary Game. "stata_analysis.do" contains code for the rest of the analysis.

Institutions

• University of Arizona

  AZ, Tucson

Categories

Licence