Correlation in Voting

Last registered on November 15, 2024

View Trial History

Pre-Trial

Trial Information

General Information

Title

Correlation in Voting

RCT ID

AEARCTR-0014792

Initial registration date

November 08, 2024

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published

November 15, 2024, 1:42 PM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Country

United Kingdom of Great Britain and Northern Ireland

Region

Primary Investigator

Name

Friederike Mengel

Affiliation

Uni Essex

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

Javier Rivas Ruiz

PI Affiliation

University of Bath

Contact Investigator

PI Name

Qingchao Zeng

PI Affiliation

University of Essex

Contact Investigator

Additional Trial Information

Status

In development

Start date

2024-11-08

End date

2024-12-24

Keywords

Behavior, Electoral

Additional Keywords

JEL code(s)

C92, D71

Secondary IDs

Prior work

This trial does not extend or rely on any prior RCTs.

Abstract

In large elections where there is correlated information and the possibility, however small, of misinformation (i.e. that information is non-informative), information aggregation fails if voters have a non-zero, potentially negligible, idiosyncratic bias towards either party. The mechanism in play is that if a voter is pivotal, then the posterior on information is that the most likely event is that information is non-informative, regardless of how its prior probability might have been. Thus, a voter is better of voting following the bias even if it is arbitrarily small. Our experiments are designed to understand the empirical relevance of this phenomenon and the underlying mechanism.

External Link(s)

Registration Citation

Citation

Mengel, Friederike, Javier Rivas Ruiz and Qingchao Zeng. 2024. "Correlation in Voting." AEA RCT Registry. November 15. https://doi.org/10.1257/rct.14792-1.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

Group size (N=3, N=15);
Strategy Method or not;

Within subjects we vary signal accuracy and the distribution of biases.

Intervention (Hidden)

Intervention Start Date

2024-11-08

Intervention End Date

2024-12-08

Primary Outcomes

Primary Outcomes (end points)

(1) Share of efficient elections.
(2) Share of signals that are truthfully revealed.

Primary Outcomes (explanation)

(1) We call an election efficient if the candidate elected matches the state.
(2) We say a signal is truthfully revealed if the voter votes according to the signal that they received.

Secondary Outcomes

Secondary Outcomes (end points)

(1) Share of signals that are truthfully revealed conditionally on whether signal and bias are aligned.
(2) Welfare

Secondary Outcomes (explanation)

(1) We are particularly interested in how voters react when their signal and bias are not aligned.
(2) As a measure of welfare we will consider average payoffs across all voters.

Experimental Design

Our treatment design consists of a 2x2 between subjects design. In all treatments, participants are informed about their bias and the distribution of biases in their group of voters. They then receive a signal and vote. Participants make four decisions and do not receive feedback about others’ behaviour during the experiment. Treatments differ in the group size (N=3 or N=15) and in whether the strategy method is used to elicit the vote or not. Within subjects we also vary the signal accuracy and the distribution of biases.

Experimental Design Details

Randomization Method

Participants sign up to surveys online. They can only sign up to one survey and which survey they are offered is randomly determined. Within subjects randomization is done online within the survey.

Randomization Unit

individual

Was the treatment clustered?

Yes

Experiment Characteristics

Sample size: planned number of clusters

1080 individuals

Sample size: planned number of observations

4320 decisions.

Sample size (or number of clusters) by treatment arms

270 individuals

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Theoretically in our baseline case N=3 (conditional on the biases and signal accuracy 0.9) we should expect all voters to truthfully reveal their signals. By contrast with N=15 only 5 people should truthfully reveal their signal and 10 voters will vote according to their bias. As for half of these 10 voters their bias will align with the signal (in expectation) we will not be able to identify whether they vote according to the bias or according to the signal. In expectation we should see 5 people, i.e. 1/3 of the voters, vote against their signal and in line with their bias. We use this effect size as a basis for power calculation. The final sample size is informed by power calculation as well as by the fact that we plan to do heterogeneity analysis and that actual effect sizes might be smaller than theoretically predicted. We decided on a final sample size of 270 participants per (between subject) treatment.

Supporting Documents and Materials

IRB