x

Please fill out this short user survey of only 3 questions in order to help us improve the site. We appreciate your feedback!
Costly information acquisition and uninformed voting
Last registered on August 14, 2020

Pre-Trial

Trial Information
General Information
Title
Costly information acquisition and uninformed voting
RCT ID
AEARCTR-0006224
Initial registration date
August 13, 2020
Last updated
August 14, 2020 9:56 AM EDT
Location(s)

This section is unavailable to the public. Use the button below to request access to this information.

Request Information
Primary Investigator
Affiliation
University of Adelaide
Other Primary Investigator(s)
PI Affiliation
University of New South Wales
PI Affiliation
University of Queensland
Additional Trial Information
Status
In development
Start date
2020-10-01
End date
2021-06-30
Secondary IDs
Australian Research Council DP190102629
Abstract
This project investigates if election and referendum outcomes are distorted due to some voters casting their vote despite being uninformed. A game-theoretic model with rational agents shows that the voters in a society who have the lowest cost of getting informed about candidates or alternatives in an election or referendum should acquire the information and vote. Citizens who have higher information acquisition cost should not get informed and abstain from voting (if voting is not compulsory). Anecdotal evidence suggests that many people vote without being properly informed (e.g. spike in google searches for “what is Brexit?” after the referendum in the UK). Uninformed voting reduces the likelihood that the actually preferred alternative in a referendum or their preferred candidate/party in a general election is chosen. Hence, uninformed voting is a threat to the efficiency of the democratic process.This study uses the Amazon Turk platform to investigate if the phenomenon of uninformed voting is sizable. Treatment variations are designed in order to identify the mechanisms causing the phenomenon and to evaluate their relative importance.
External Link(s)
Registration Citation
Citation
Bayer, Ralph-C, Marco Faravelli and Carlos Pimientas. 2020. "Costly information acquisition and uninformed voting." AEA RCT Registry. August 14. https://doi.org/10.1257/rct.6224-1.0.
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details
Interventions
Intervention(s)
We will use the computer-work platform Amazon Mechanical Turk to implement a voting game with costly information acquisition. Participants are grouped with other participants in an electorate and receive 2 US Dollars for reading the instructions and answering a few control questions. The electorate's task is to choose one out of two alternatives via majority voting. If the collective chooses the "correct" alternative then everybody receives 5 US Dollars. If the wrong alternative is chosen, then no payout occurs. Ex-ante both alternatives are equally likely to be correct. Participants are then asked if they would like to buy information. Buying information allows participants to update their beliefs such that they know which alternative is twice as likely to be correct than the other. Subjects know their own cost, which is drawn from a uniform distribution but only know the distribution the cost of for the others in the electorate are drawn from. After acquiring information (or not), participants are asked to either vote for one of the two alternatives or to abstain. Votes are tallied and payouts are realized. Note that in equilibirum nobody should ever vote without acquiring information.
We will vary this basic game in two dimensions: a) size of the electorate, b) framing of the situation and information provided. The size dimension allows us to investigate if the likelihood of being pivotal impacts the prevalence of uninformed voting. Varying the framing and the specific information provided allows us to differentiate between different psychological mechanisms that could lead to uninformed voting.
A secondary question is how size and framing impact the voting informed versus abstaining decision.
Over-all we are planning on a 24 treatments.
Intervention Start Date
2020-10-01
Intervention End Date
2020-10-20
Primary Outcomes
Primary Outcomes (end points)
Probability of voting uninformed as a function of information cost.
Primary Outcomes (explanation)
The primary outcome will be the predicted probability of voting uninformed conditional on information cost or each treatment. For this we will take observed behaviour (voting uninformed) and information cost drawn and estimate a linear probability model, a logit or probit model.
Secondary Outcomes
Secondary Outcomes (end points)
Predicted probability of abstaining conditional on information cost.
Secondary Outcomes (explanation)
For this we will take observed behaviour (abstaining from voting) and information cost drawn and estimate a linear probabiity model logit or probit model.
Experimental Design
Experimental Design
Our experimental design varies two dimensions:
a) the specific presentation of the task
b) the size of the electorate
By varying the presentations and specifics of the task we will be able to distinguish between different mechanisms that contribute to uninformed voting such as, bounded rationality, expressive voting and biased beliefs. By varying the size of the electorate we will be able to investigate how the likelihood to be pivotal influences the prevalence of uninformed voting. Interaction effects between mechanisms and the likelihood to be pivotal will also be identifiable.
Experimental Design Details
Not available
Randomization Method
We use fully automatic randomization via the survey tool Qualtrix. When a participant clicks on our listing in AmTurk, then Qualtrix automatically assigns the participant to a treatment at random.
Randomization Unit
Randomization on the individual level
Was the treatment clustered?
No
Experiment Characteristics
Sample size: planned number of clusters
4800 individuals
Sample size: planned number of observations
4800 individuals
Sample size (or number of clusters) by treatment arms
200 individuals per treatment. We have a 4x3x2 Design (24 treatments).
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
The minimum detectable effect size for marginal effect in logistic regressions depend on a lot of ex-ante unobservables. Hence, we report here the effect size would be able to detect with a likelyhood ratio test on raw proportions of participants voting uninformed. We would be able to detect a difference in the fraction of participants voting uninformed of 0.0681 with power 0.8 if we assume that about 5 percent of people vote uninformed in the the treatment where we expect less uninformed voting.
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
Human Research Ethics Committee of the University of Adelaide
IRB Approval Date
2020-08-12
IRB Approval Number
H-2020-155