Gender discrimination in the mortgage market of Ecuador

Our study employs a lab-in-the-field experiment to investigate gender discrimination in Ecuador's mortgage market. We designed a platform featuring the credit applications of 16 mortgage applicants.
Analysts are tasked with evaluating the terms of credit applications on an online platform where the gender and credit characteristics of applicants are assigned randomly, resulting in an equal distribution of 8 women and 8 men. Out of the total 16 applications, analysts answer 8 questions for 12 of them, while the remaining 4 applications involve an additional 4 questions specifically related to renegotiation possibilities. The randomization of the applicant's gender occurs within the context of demographic and credit information.
Analysts are prompted to answer the following questions in the first 12 applications:
1. Is the approved credit rate appropriate?
• If "No," what rate would you recommend? (Numeric field %)
2. Does the applicant require a warrantor?
3. Is the loan-to-value ratio appropriate?
• If "No," what ratio would you recommend? (Numeric field %)
4. Is the loan term appropriate?
• If "No," recommend a term: 15, 20, 25 years
5. Is the repayment period appropriate?
6. Which amortization system would you recommend: German or French?
7. Are there other dimensions requiring adjustment based on risk?
• If "Yes," specify the dimension
8. Specify the criteria used for your evaluation.
For the last 4 applications, analysts consider an additional 4 renegotiation questions:
9. If this person asked to negotiate again by increasing the payment term by 5 years to have lower monthly payments, would you accept this new condition or not?
10. If this person asked to negotiate again by reducing the mortgage interest rate by 1 percentage point, would you accept this new condition or not?
11. If this person requested to negotiate again, reducing the initial amount requested by 5%, would you accept this new condition or not?
12. If this person requested to negotiate again by increasing the percentage of the property that is covered by the credit by 10%, would you accept this new condition or not?
Therefore, the intervention occurs at the applicant level, with the randomization of the gender. This intervention will measure gender discrimination in the mortgage market and the extent to which credit conditions change if women ask for renegotiations.

Our study employs a lab-in-the-field experiment to examine discrimination against women in Ecuador's mortgage market. This method integrates the control of laboratory experiments with the real-world relevance of field studies, allowing for the manipulation of variables and isolation of causal effects while enhancing external validity (Gneezy & Iman, 2017). Specifically, we apply this approach to measure gender discrimination in Ecuador's mortgage market.

2023-11-13

2024-01-22

Mortgage rate, Mortgage coverage, Terms in years, Request warrantor

We looked at four dependent variables:
1. "Mortgage rate" refers to the interest rate proposed by credit analysts to applicants for their mortgage.
2. "Mortgage coverage" represents the proportion, in percentage terms, of the total monetary value of the property that the mortgage is suggested by analysts to cover.
3. "Term in years" indicates the number of years for which analysts recommend repayment of the mortgage.
4. "Request warrantor" is a binary variable that takes the value 1 if the analyst requests a warrantor and 0 if they do not request one.
Our focus was on the applicants' gender, which we treated as an independent variable.
We used two different models to analyze the data:
The first model looked at the relationship between the gender of the applicant, the risk tolerance score of credit analysts, the selection order, and the outcomes without any conditions.
The second model added controls for applicants' sociodemographic characteristics. Fixed effects were added to variations in standard errors were considered at the analyst level.
We calculated the average differences in outcomes using these two models. The estimated models for the analysis follow the general equation number 1:
Y_itc=β_0+ β_1 X_it+ ε_itc (1)
where Y is the dependent variable, corresponding to synthetic applicant i, evaluated by credit analysts c; X is the independent variable of interest; and ε is the error term. The dependent variables, the main independent variable, and the control variables are included according to the specifications indicated in the estimates.

Loan officers will be contacted through a LinkedIn job post, which will explain the nature of the assignment. The job post will clarify that it is a temporary, paid, remote job that does not require leaving their current job. For each loan officer who finishes evaluating the 16 applicants, they will be paid the sum of USD 20. In addition, they will be asked to recommend three (3) loan officers possibly interested in participating. For each referral who completes the exercise, they will be paid an additional USD 10. Payment to loan officers is based on the project budget, response rate goals, and previous experience.
Registration module
Once the loan officer accepts the task, he/she interacts with a registration module that consists of four phases personal information collection, risk tolerance assessment, and socioemotional and cognitive tests. First, the loan officers will provide information on their personal, work, and academic characteristics, which will serve to use control variables in the regression analyses of the research.
Finally, information will be gathered on the socio-emotional and cognitive traits of loan officers. Regarding cognitive information, we implemented a Wonderlic test. For the socioemotional characteristics, we used the self-esteem test, following Rosenberg (1965), and a personality test based on Digman's (1990) or the OCEAN test. Once the registration is completed, the agent receives an e-mail with the credentials to access the selection module.
Selection module
The selection module begins by asking for the credentials to access the platform: an email and password. The credentials are sent automatically by email to every recruiter that finishes the registration process.
The platform presents a welcome message and a brief explanation of the exercise on the first screen. The loan officer is tasked with evaluating 16 pre-approved mortgage loan applications and determining the loan conditions for each applicant. To begin the review of applications, the loan officer must click on the blue button labeled 'Review applicants and applications.
Upon clicking, the screen displays the 16 credit applications, and the loan officer needs to click on the button located on the right side, labeled 'Review,' to access the details of each application.
Subsequently, the platform shows the personal identification information of the applicant selected, including name, surname, e-mail, nationality, identity card or passport, profession or occupation, and marital status. Also, the form displays the following information: socio-demographic data of the applicant, details of the property to be purchased, personal income, and, if applicable, the spouse's income, as well as the loan pre-approved.
Once the application review is complete, the loan officer proceeds to click on the blue button at the bottom of the platform to evaluate the applicant. Subsequently, a form appears with the following questions based on the information provided: i) Does the rate at which the credit has been approved seem appropriate to you? If it is not the most appropriate rate, what rate would you recommend for this client?; ii) Do you think this person requires a co-signer?; iii) Does the percentage of the property covered by the loan seem appropriate to you? If not, what percentage of the property do you suggest the loan covers?; iv) Does the agreed term to pay the loan seem appropriate to you? If not, what payment period would you grant to this person?; v) Does the amortization system seem appropriate to you? If not, what amortization system would you recommend?; vi) Is there any other credit dimension that you considered that is necessary to make changes given the applicant’s credit risk level?; vii) Additionally, please comment on the criteria you used to assess the applicants. Once they finish reviewing the application, they click on “Finish” and they repeat the same process until they complete the 16 applications.

The randomization in this experiment is done in the office by a computer, through the custom-made online platform.

The randomization unit is at the level of the application. The applicant's gender was randomized between the applicants' demographic information and credit information. Additionally, the order of the 16 applications is randomized. Of the total 16 applications, 8 are randomly assigned to women and the remaining to men.

No

There are no clusters in this experiment

The planned number of observations is 150 credit analysts which complies with the following criteria. Credit analysts for this study must have a minimum of two years' work experience in credit analysis or risk assessment and hold at least a university degree in a finance-related field. Eligible analysts will initially be contacted via LinkedIn; those expressing interest will be further communicated with via email.

There are no treatment arms in this experiment.