|
Field
Abstract
|
Before
In this project I study how individuals' body mass affects access to credit in Uganda. Participants are loan officers from a random subsample of banks and micro financing institutions from Kampala, Mbarara and Gulu. Loan officers are a key step in the assessment of creditworthiness: They screen prospective clients screen prospective clients and refer them to the second stage of the loan approval process; for selected loan profiles, they are also be responsible for the entire loan approval process. In either context, their evaluation is a key step in obtaining access to credit. Hence, the measures of ease of access to credit in the experiment are: a creditworthiness rating; a referral of the applicant to the next step of the loan approval process (binary) or the binary choice of granting the loan or not; conditional on a referral, an assessment of the applicant's likelihood to receive the desired loan at the end of the process; conditional on a not referral, the choice of advising to apply for a smaller loan and an assessment of the applicant's likelihood to receive a smaller loan.
The design builds on Kessler et al. (NBER WP 2019) experiment. After answering a set of questions including demographics, financial literacy and banks characteristics, loan officers are shown 6 hypothetical loan applications which vary by applicants' characteristics and loan profiles. Most notably, each applicant's picture is randomly selected between either a group of underweight/norm weight or a group of overweight/obese pictures. Moreover, each application can either include information on wealth or not. Answers are incentivised by matching loan officers' preferences with real applicants' characteristics and providing referrals to prospective clients.
This set up allows me to study whether (1) body mass affects perceived creditworthiness and referral probability, and (2) whether reducing asymmetric information on income changes the relevance of body mass in determining the decision. As the gender of the applicants is randomly assigned, I will also investigate heterogeneity by gender.
|
After
In this project I study how individuals' body mass affects access to credit in Uganda. Participants are loan officers from a random subsample of banks and micro financing institutions from Kampala. Loan officers are a key step in the assessment of creditworthiness: They screen prospective clients screen prospective clients and refer them to the second stage of the loan approval process; for selected loan profiles, they are also be responsible for the entire loan approval process. In either context, meeting with them and their evaluation is a key step in obtaining access to credit.
In the experiment, I randomise body mass of hypothetical applicants, and evaluate how this affects access to credit. My measure of access to credit are 1) the choice of meeting the hypothetical applicant; 2) likelihood of refer the applicant to the next step of the loan application process 3) the expected probability of the applicant to receive the requested loan.
I exploit a design which builds on Bartoz et al. (2016) and Kessler et al. (WP 2019) insights, which allows to test for discrimination and pin down the mechanism at play. Treatment is assigned within subject. In a first treatment arm, I provide no information on the applicant, except picture and age. In a second treatment arm, I allow the loan officer to learn about the applicants' income and occupation. In the third treatment arm, I provide all loan officers the additional information.
After answering a set of questions including demographics, financial literacy and banks characteristics, loan officers are shown 45 hypothetical loan applications which vary by applicants' characteristics and loan profiles. The order of the treatment is fixed, but the applications are randomly assigned to treatment group.
This set up allows me to study whether (1) body mass affects perceived creditworthiness and access to credit, (2) whether reducing asymmetric information on income changes the relevance of body mass in determining the decision, and (3) at which part of the decision making process does the discrimination bites.
|
|
Field
Trial Start Date
|
Before
September 23, 2019
|
After
October 23, 2019
|
|
Field
Last Published
|
Before
August 27, 2019 04:56 AM
|
After
October 11, 2019 10:12 AM
|
|
Field
Intervention Start Date
|
Before
September 23, 2019
|
After
October 23, 2019
|
|
Field
Primary Outcomes (End Points)
|
Before
The key dependent variables are (1) A 0-10 Likert scale creditworthiness rating (2A) 0-1 choice of referring the applicant to the next step of the process or, when loan officer has full discretionary power, (2B) 0-1 choice of granting the requested loan. Conditional on (2A) == 1, (3) 0-1 ex-ante assessed applicants' likelihood of receiving the requested loan, at the end of the process. Conditional on (2A) == 0, (3A) choice of advising to apply for a smaller loan (with referral). Conditional on (2B) ==0 (3B) choice of advise to apply for a smaller loan (with approval). Conditional on (2A) ==1, ex-ante assessed applicants' likelihood of receiving a half-sized loan, at the end of the process.
The main explanatory variable of interest is body mass of the picture randomly assigned to the application. The second explanatory variable of interest is wealth of the applicant.
The main statistics monitored are the difference in outcome (1) and outcome (2) for applications which are randomly assigned a high body mass picture vs applications which are randomly assigned a low body mass picture. I assess the variation in this difference when the applicant wealth is revealed, and when applicant's wealth is not disclosed. Furthermore, as pre-experimental qualitative research suggested potential differences in return to body mass by gender, I will also investigate heterogeneity in these basic statistics by gender.
In addition to computing basic statistics, I will also identify parameters as in Kessler et al. (2019) by running regressions and interacting parameters in those regressions with the treatment variables (BMI/ Wealth).
The set of regressions that I will perform are the following. On the set of applications without income information, I will estimate for each main outcome:
(*) Y =α x BM +u,
where the error terms are clustered at the enumerator level. To this basic equation, I will then add sequentially 1) loan profiles fixed effect; 2) Bank Tier and Location FE; 3) controls on the applicant: Age, Gender, Occupation, Attractiveness; 4) controls on the loan officer: Age, Gender, Experience, Financial literacy.
As a second step, I will estimate the same model but allowing for heterogeneity by gender of the applicant.
On the set of applications with income information, I will estimate for each main outcome:
(**) Y =α x BM+γ×Wealth+β×Wealth*BM +u,
where the error terms are clustered at the enumerator level. To this basic equation, I will then add sequentially 1) loan profiles fixed effect; 2) Bank Tier and Location FE; 3) controls on the applicant: Age, Gender, Occupation, Attractiveness; 4) controls on the loan officer: Age, Gender, Experience, Financial literacy. As a second step, I will estimate the same model but allowing for heterogeneity by gender of the applicant.
On the full set of application, I will estimate for each main outcome:
(***) Y =α x BM+γ×WealthInformation+β×WealthInformation*BM +u,
where the error terms are clustered at the enumerator level. To this basic equation, I will then add sequentially 1) loan profiles fixed effect; 2) Bank Tier and Location FE; 3) controls on the applicant: Age, Gender, Occupation, Attractiveness; 4) controls on the loan officer: Age, Gender, Experience, Financial literacy. As a second step, I will estimate the same model but allowing for heterogeneity by gender of the applicant.
|
After
Creditworthiness; probability of accessing credit.
|
|
Field
Primary Outcomes (Explanation)
|
Before
BM: 0/1 dummy, 1 if picture was modified to show higher body mass; 0 if picture was modified to show lower body mass. BM will also be defined continuously based on assessment by similar population via survey.
Wealth: Low/High dummy (Low: Revenue/month: Ush. 1-1.5 million and Profits/month: Ush. 400-600'000 per month; High: Revenue/month: Ush. 3-4 million; Profit/month: Ush. 1-1.5 million)
WealthInformation: 0/1 dummy if wealth information is present.
Financial literacy: FCA of answers to Financial Literacy Section in Questionnaire.
Attractiveness: Assessed by similar population via survey, on a 0-10 Likert scale (Mobius and Rosenblat, 2006).
|
After
(1) A Yes/No choice of meeting the applicant to discuss and verify information; (2) a 1-6 Likert scale assessment of the probability of referring the person to the next step of the application process (3) a 1-6 Likert scale assessment of the applicants' probability of receiving the requested loan.
|
|
Field
Experimental Design (Public)
|
Before
The design builds on Kessler et al. (NBER WP 2019). After answering a brief questionnaire, loan officers are asked to evaluate 6 hypothetical loan applications, a set. The application in a set are grouped into two loan profiles, which vary by amount requested and repayment rate. The first 3 applications are assigned to the first loan profile (Ush. 1 million, with 6 months repayment); the second 3 applications are assigned to the second loan profile (Ush. 5 million, with 12 months repayment). Notably, for selected Tier 1 banks, who do not lend as low as Ush. 1 million, the first loan profile entails Ush. 10 million, with 12 months repayment. Within each loan profile, the applications vary by wealth category. The first application has no wealth information; the second application is "low wealth"; the third application is "high wealth". I define as cell the intersection of a loan profile and a wealth category. Within each cell, the applicants' occupation, appearance and body mass are cross-randomised, with the caveat that - to avoid suspicion - the same individual appearance cannot repeat within each set with different body mass or occupation.
Using this procedure, I build 23 sets and I randomly select 21 of those and assign each of them an enumerator. Each enumerator is randomly assigned to a set of branches to interview. The set of branches is obtain by randomly selecting 150 between branches and headquarters in Kampala, 28 in Mbarara and 24 in Gulu. For each branch, 1 to 3 loan officers are interviewed. The procedure is as follows: The enumerator asks who are the people that directly in touch with with (potential) borrowing customers. If there are more than 3, the enumerator randomly selects 3 of them to be interviewed. If there are less than 3, everyone is interviewed. From this randomisation technique follows that all the loan officers in a branch will rate the same set of applications. This is again to make sure that no loan officers happens to observe the same individual with different body mass or occupations.
Each loan officers is shown each of the 6 application singularly. First, the applications with no wealth information are shown. Then, the remaining applications are shown in random order. For each application, the loan officer answers the following questions: (1) A 0-10 Likert scale creditworthiness rating (2A) 0-1 choice of referring the applicant to the next step of the process or, when loan officer has full discretionary power, (2B) 0-1 choice of granting the requested loan. Conditional on (2A) == 1, (3) 0-1 ex-ante assessed applicants' likelihood of receiving the requested loan, at the end of the process. Conditional on (2A) == 0, (3A) choice of advising to apply for a smaller loan (with referral). Conditional on (2B) ==0 (3B) choice of advise to apply for a smaller loan (with approval). Conditional on (2A) ==1, ex-ante assessed applicants' likelihood of receiving a half-sized loan, at the end of the process.
When answering the questions, the loan officers are informed that the applicants are hypothetical. Honest answers are incentivised by informing the loan officers that their answers will be used to provide referrals to prospective clients whose characteristics match with their preferences. The matching is performed using a machine learning algorithm, following Kessler et al. (2019).
The design allows to investigate whether (1) body mass affects perceived creditworthiness and referral probability, and (2) whether reducing asymmetric information on income changes the relevance of body mass in determining the decision. It also allows to test whether the estimated relations are heterogenous by gender.
|
After
Loan officers are asked to evaluate 45 hypothetical loan applications, in three sets. Each set is a treatment arm. The order of the treatment arms is fixed, but each hypothetical application is randomly selected into one of each arm for each participant.
The applications are grouped into two loan profiles, which vary by amount requested and repayment rate. The first loan profile is a Ush. 1 million, with 6 months repayment; the second loan profile Ush. 5 million, with 12 months repayment. Notably, I exclude Tier 1 banks from my sample as they generally do not lend as low as Ush. 1 million.
Within each loan profile, the applications vary by characteristics; in each application, the applicants' age and residence are defined statically, while the applicants' occupation, income, portrait, body mass are cross-randomised. In particular, to avoid suspicion, portraits are randomized without replacement across the full set of applications each respondents receives.
Using this procedure, I build 45 loan applications and randomly assign each to a treatment arm. In the first treatment arm, respondents only see the picture, age and loan profile requested. In the second treatment arm, respondents only see the picture, age and loan profile requested but can choose to learn occupation and income of the applicant. In the third treatment arm, participants see all the information (including the picture) at once.
Then, the remaining applications are shown in random order. For each application, the loan officer answers the following questions:
[Main outcomes]
(1) A Yes/No choice of meeting the applicant to discuss and verify information; (2) a 1-6 Likert scale assessment of the probability of referring the person to the next step of the application process (3) a 1-6 Likert scale assessment of the applicants' probability of receiving the requested loan.
[Secondary outcomes]
(4) A 1-6 Likert scale assessment of the probability that the applicant makes productive use of the money.
(5) A 1-6 Likert scale assessment of the probability that the applicant will repay/collateral will be collected, conditional on the investment not being successful.
For loan profiles where the loan officer has full discretionally on the application decision, question (2) and (3) are substituted by question (6) a 1-6 Likert scale assessment of the probability of granting the requested loan.
The set of branches is obtained by randomly selecting 70 branches and headquarters in Kampala. For each branch, 1 to 3 loan officers are interviewed. The procedure is as follows: The enumerator asks who are the people that directly in touch with with (potential) borrowing customers. If there are more than 3, the enumerator randomly selects 3 of them to be interviewed. If there are less than 3, everyone is interviewed.
When answering the questions, the loan officers are informed that the applicants are hypothetical. Honest answers are incentivised by informing the loan officers that their answers will be used to provide referrals to prospective clients whose characteristics match with their preferences. The matching is performed using a machine learning algorithm, following Kessler et al. (2019).
The design allows to investigate whether (1) body mass affects perceived creditworthiness and referral probability, (2) whether reducing asymmetric information on income changes the relevance of body mass in determining the decision, (3) at which point of the decision making process does discrimination bites.
Relevant heterogeneity analysis include: 1) gender; 2) body mass of the loan officer; 3) financial literacy/experience; 4) perception of high body mass. If heterogeneity in the share of applicant invited is detected (either at the loan officers' level or at the institution level), an interesting heterogeneity analysis compares "lemon dropping" types to "cherry picking" ones. The design builds on Kessler et al. (NBER WP 2019) and Bartoz et al. (2016).
At the end of the experiment, loan officers answer a set of questions on financial literacy; perception of high body mass;
|
|
Field
Planned Number of Clusters
|
Before
21 enumerators.
|
After
200 loan officers
|
|
Field
Planned Number of Observations
|
Before
6 applications per loan officer; 1 to 3 loan officers per branch for a total of 202 branches. Planned number of observations ranges from 1'212 to 3636 applications.
|
After
45 applications per loan officer; 1 to 3 loan officers per branch for a total of 200 branches. Planned number of observations is around 6000 applications.
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
21 clusters in the no- wealth information treatment arm; 21 clusters in the wealth information treatment arm (within loan officer design).
|
After
This is not an RCT. Half of the applications by treatment arm will have high body mass portrait, and half will have a low body mass one.
|
|
Field
Power calculation: Minimum Detectable Effect Size for Main Outcomes
|
Before
|
After
In Kessler et al. (2019), the effect of gender on hiring interest for the STEM majors
subsample ranged from 0.399 to 0.516 sd., with a single treatment arm. Because the effect the
effect of body mass is presumably mediated by wealth, I target a smaller MDE. 3000
observations allow to detect a MDE of 0.1 sd in each treatment arm.
|
|
Field
Secondary Outcomes (End Points)
|
Before
|
After
financial reliability; financial ability;
|
|
Field
Secondary Outcomes (Explanation)
|
Before
|
After
Financial reliability: A 1-6 Likert scale assessment of the probability that the applicant makes productive use of the money.
Financial ability: A 1-6 Likert scale assessment of the probability that the applicant will repay/collateral will be collected, conditional on the investment not being successful.
|