Discrimination against LGTBI, Migrants and Women in the real State Markets of Ecuador

1) We randomize a minority condition (being LGTBI, woman, or a migrant) to one applicant across nine pairs of fake applicants to fictitious rental properties in the cities of Quito and Guayaquil. The two candidates applying to each of the nine properties were observationally equivalent, but, because of that random assignment, one of them had the minority condition.
2) In addition, we randomized half of the real state agents to receive a "pop-up" message showing them a "vignette' in the form of a pop-up message. The message told the recruiters that minorities have to overcome disadvantages to perform similarly to non-minorities.

2022-10-25

2022-12-15

To measure discrimination we count on three outcomes based on differences in how the real state agents treat the minority vs. the non-minority applicants in the experiment, namely: 1) the rate at which they prefer a minority applicant from a nonminority one; 2) their rating of the "fit" of each applicant for the property, and; 3) what the agents suggest candidates should pay for their rentals.

Time employed reviewing applications
Ratings of the factors influencing the evaluation of the candidates

We will conduct a field experiment to estimate the extent of discrimination toward LGTBI, migrants, and women in the Ecuadorian real estate market. We will hire 600-750 real estate agents to evaluate ten applications online to real estate properties (each application consisting of two families). We will employ agents in Quito and Guayaquil. They should select the family more suitable for each property, rate the suitability of all applicants for the property, and suggest a leasing amount each family should pay, given their risk level. This methodology fits within the class of vignette and factorial survey studies.

We will run the experiment will be conducted between October 2022 and February 2023. We will pay the real estate agents with gift cards, and they will be recruited as part of a research study where we will consult their professional services.

All participants will evaluate the same set of ten pairs of candidates associated with unique properties. In nine properties, the pair of applicants will include one from a minority background (either LGTBI, woman, or migrant). We designed each applicant pair for the minority attribute in one of the applicants to be conspicuous. Applicants in each pair are qualitative equivalent in all characteristics except for their minority condition. We will task the real estate agents to complete four activities: first, select the best applicant for the property; second, evaluate their suitability for the property on a scale from 1 to 10; third, suggest a leasing amount; and; 4) rate on a scale from 1 to 10 different aspects that could have influenced their choice, such as the probability that the applicant complies with the contract in terms of time; the likelihood that the applicant complies with the agreement in terms of payments; the chance of renewing the lease, and; the applicant's occupation.

We will interview the recruiters when we hire them to obtain information about their sociodemographic characteristics and cognitive and socio-emotional abilities. We will collect data about the interaction of agents with the online platform in terms of the time employed to review applications and study patterns of association between those times and personal characteristics. We will use that data to identify potential channels and understand whether some or all of those factors increase the agents' propensity to discriminate.

Properties and families were carefully crafted to replicate market conditions based on findings from a preliminary qualitative study. We conducted interviews with stakeholders to understand the functioning of the real estate market in both cities and focus groups. The characteristic of the properties to be displayed in the platform include the total rent value, a general description of the property (area, age, and stratum), and some other attributes such as the number of bedrooms, bathrooms, parking spaces, balcony, etc. We also included a list of secondary characteristics of the properties, including concierge service, common areas, etc., information about the neighborhood, and property photographs.

We designed 20 applicants' pairs of profiles in each city: 10 for Quito and 10 for Guayaquil. In each city, there will be three duplets where gender is the minority condition, three migrants, and three where LGTBI is the minority condition. There will also be one placebo duplet.

Before starting the eighth round of the experiment, we will randomize half of the agents with a pop-up message. The message will tell the agents that minorities must overcome disadvantages to perform similarly to non-minorities. Then the agents will have to finish the remaining experiment trials (i.e., trials 8, 9, and 10). We will measure the impact of that message on the outcomes of interest using experimental and nonexperimental methods.

In both cases (randomization of the minority condition and the pop-up message detailed above, randomization is done automatically by the computer algorithm within the online platform designed for the experiment.

Randomization will take place at the level of the property being evaluated.

No

600-750 clusters (each cluster is a series of responses given by a real state agent)

6000-7500 (600-750 recruiters x 10 answers each)

First field experiment: measuring discrimination
1) To detect discrimination against LGTBI: 2000-2500 answers (1000-1250 equally to treatment and control) nested in 200-250 real state agents; 2) To detect discrimination against migrants: 2000-2500 answers (1000-1250 equally to treatment and control) nested in 200-250 real state agents; 3) To detect discrimination against women: 2000-2500 answers (1000-1250 equally to treatment and control) nested in 200-250 real state agents;

Second field experiment: measuring the impact of a behavioral intervention
1) To detect impact of pop-up message on discrimination against LGTBI: 2000-2500 answers (1000-1250 equally to treatment and control) nested in 200-250 real state agents;