Irrational updating based on ethnicity and gender

Last registered on June 24, 2020

View Trial History

Pre-Trial

Trial Information

General Information

Title

Irrational updating based on ethnicity and gender

RCT ID

AEARCTR-0006041

Initial registration date

June 23, 2020

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

June 24, 2020, 11:08 AM EDT

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

Pol Campos-Mercade

Affiliation

Lund University

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

Friederike Mengel

PI Affiliation

University of Essex

Contact Investigator

Additional Trial Information

Status

In development

Start date

2020-06-24

End date

2021-12-31

Keywords

Gender, Labor

Additional Keywords

JEL code(s)

Secondary IDs

Abstract

In the absence of complete information about individual characteristics, employers use informative signals to form beliefs about the ability of a worker. In this experiment, we test 1) whether people are Bayesian updaters when considering the ability of others and 2) whether such updating depends on the identity group of the worker. If there is different updating based on a worker’s identity, we will test whether this leads to different hiring rates for different identities.

External Link(s)

Registration Citation

Citation

Campos-Mercade, Pol and Friederike Mengel. 2020. "Irrational updating based on ethnicity and gender." AEA RCT Registry. June 24. https://doi.org/10.1257/rct.6041-1.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

Intervention Start Date

2020-06-24

Intervention End Date

2020-06-25

Primary Outcomes

Primary Outcomes (end points)

Belief that a worker performed in the top half after obtaining the information signals

Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)

Whether a worker is hired

Secondary Outcomes (explanation)

Experimental Design

N/A

Experimental Design Details

We will test our hypotheses using an online experiment with a sample of the British population:
- Each subject has to assess 4 workers. The names of these workers are perceived as 1) white male, 2) white female, 3) south-asian male, and 4) black male. (These perceptions were pre-tested.) Each subject sees one of 4 possible names for each of these subgroups. The order in which they assess each of these workers is random.
- For each worker, the subject observes 1) the name and 2) a CV consisting of age, gender, and region where the worker lives.
- The subject assesses the probability that this person performed in the top half of a group of 20 workers in an IQ test. Subjects are told the names of these 20 workers.
- The subject then receives three signals on whether the performance of this worker was in the top half of his/her group or not. If the worker is a top (bottom) half performer, they receive a positive signal with 70% (30%) probability and a negative one with 30% (70%) probability.
- After receiving each of the signals, subjects are asked again about the probability that the worker is a top performer.
- Finally, subjects can decide whether they hire the worker or not. Hiring the worker implies that the subject pays a sum of money to the worker. If the subject hires a worker who scored in the top half of his/her group, then the subject gets a higher payment.
The main treatment variation is the ethnicity and gender of the worker that is being evaluated. We will use “white male” as the control group, and “white female”, “south-asian male”, and “black male” as the treatment groups.

Randomization Method

Randomization done by Qualtrics.

Randomization Unit

Subjects

Was the treatment clustered?

Yes

Experiment Characteristics

Sample size: planned number of clusters

1020 subjects

Sample size: planned number of observations

4080 observations

Sample size (or number of clusters) by treatment arms

1020 subjects (the four treatment arms are within subject)

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

Our aim is to have 90% power to detect an effect of 2 percentage points (at the 5% level) for our main test, which we deem as an economically significant effect. We performed a power analysis bootstrapping data from a pilot of 188 subjects, concluding that we need about 1020 observations to detect an effect of this size.

Supporting Documents and Materials

IRB