Discrimination on online markets: Evidence from a field experiment

Last registered on March 18, 2021


Trial Information

General Information

Discrimination on online markets: Evidence from a field experiment
Initial registration date
March 17, 2021

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
March 18, 2021, 6:15 AM EDT

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



Primary Investigator


Other Primary Investigator(s)

PI Affiliation
Grenoble School of Management
PI Affiliation
Cergy Paris Université
PI Affiliation
SOFI (Stockholm University)
PI Affiliation

Additional Trial Information

On going
Start date
End date
Secondary IDs
We study the extent and type of ethnic and gender discrimination that exists on two of the most widely used "P2P" (person to person) online marketplaces in France: an online classified advertisements website and a ride sharing platform. The first phase involves data scraping in which we collect large amounts of data on real ads and rides that are publicly available through the platforms’ APIs. This will allow us to better understand the characteristics of the two markets and provide correlational evidence on the existence of discrimination and its magnitude in different submarkets. As is typical in the literature using French data, we will categorize the first names (in the case of classified ads) to proxy for the gender and/or minority status (supposed non-French origin) of the buyers and sellers, as it is illegal to categorize ethnic status in France. Using insights from this observational data, we will then implement a randomized controlled trial (RCT) in which we will create sets of fictitious profiles (both buyers and sellers of fictitious goods and both fictitious drivers and passengers) that are closely matched on all observable characteristics but differ on the supposed minority status or gender. This will allow us to provide causal evidence on whether discrimination exists in these online marketplaces. Furthermore, by experimentally manipulating the fictitious buyer and seller profiles on other dimensions and exploiting existing market heterogeneity, we aim to better understand the ways in which online markets can be designed in order to minimize the negative manifestations of discrimination.
External Link(s)

Registration Citation

Chapelle, Guillaume et al. 2021. "Discrimination on online markets: Evidence from a field experiment." AEA RCT Registry. March 18. https://doi.org/10.1257/rct.7019-1.0
Sponsors & Partners

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

Request Information
Experimental Details


Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
Supply side:
-number of inquiries by type (positive, negative, neither) fictitious and real goods
-any inquiry
-time to first inquiry, delay time in correspondence

Demand side:
-number of responses by type (positive, negative, neither) for inquiries to goods/rides
-any response
-time to first response, delay time in correspondence
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
For both demand and supply:
-Offered or bargained purchase amount(s): mean, best, worst, difference with listed price
-Context of potential trade and response characteristics: buyer/seller included name, direct contact information (telephone, address, email); type of communication channel: phone, email, online message system, text messages, face to face; pick-up and drop off locations and times for rides and exchanges.
-Scams. Following Doleac and Stein (2013) we will use text analysis to tag suspicious inquiries/responses that appear to be scams.
-Reactions to modality of transaction (secure payment): hesitant, positive, negative
- Request for further information + Direct buy
-Language used: number of words in response, politeness and insults or pejorative language
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
We create sets of advertisements for common goods and rides using fictitious profiles (both buyers and sellers of goods for the online classified ads and both passengers and drivers for ride sharing) that are randomly assigned minority and gender status through first names and computer-generated profile pictures. On the supply side, these adverts are then posted in the online marketplaces. On the demand side, these fictitious profiles are used to inquire about real goods in the online marketplaces.
Experimental Design Details
Online profiles will be randomly assigned a supposed ethnic origin (French or North-/sub–Saharan African origin), a supposed gender (male, female) or a neutral pseudonym in which the ethnicity, gender or both cannot be deduced by the counter party.

In addition, we will experimentally vary the following dimensions of heterogeneity for goods sold (supply side sellers):
-Availability of secure payment
-Completeness of information
-Politeness/cordiality/spelling errors/contact information

On the demand side (buyers)
-Prices (offered)
-Completeness of information
-Politeness/cordiality/spelling errors/contact information

We will also exploit real market level heterogeneity in:
-Categories of goods
-Underlying bias (local vote shares for extreme right political party)
-Underlying minority share in market and in geographic area
-Existing ethnic/gender composition of car (for ride sharing platform)
-Location: local crime rates, employment levels, segregation
Randomization Method
Randomization Unit
Individual advertisements and fictitious buyer profiles
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
3000 advertisements and 3000 fictitious buyer profiles
Sample size: planned number of observations
3000 advertisements and 3000 fictitious buyer profiles
Sample size (or number of clusters) by treatment arms
Advertisements and buyer profiles will be randomized majority, minority or anonymous status with 1/3 probability
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Institutional Review Boards (IRBs)

IRB Name
IRB Approval Date
IRB Approval Number
Analysis Plan

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

Request Information


Post Trial Information

Study Withdrawal

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

Request Information


Is the intervention completed?
Data Collection Complete
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials