Perception of body mass as signal of wealth: a survey experiment in Uganda.

2019-11-13

2020-02-29

First order beliefs on wealth; Beliefs on other beliefs on wealth;
Expectations of returns to body mass in access to credit.

First order beliefs:
How would you rate this person's wealth?
1-4 scale

Second order beliefs:
On average, how did other respondents rated the individual’s wealth? Please provide your best guess on a scale from 1 (not at all wealthy) to 4 (very wealthy).
1-4 scale

Expectations of returns to body mass in access to credit:

- Out of 10 loan officers who evaluated this application, how many would want the referral of a similar applicant?
- On average, how likely would loan officers be to approve this application? Please provide your best guess on a scale from 1 (not at all likely) to 4 (very likely).
- According to you, would it make sense for a person with these characteristics to apply for a loan?

First order beliefs on Beauty, Health, Self-control, Life expectancy and productivity.

Beliefs on others' beliefs on Attractiveness, Health, Self-control, Life expectancy, productivity.

Explicit beliefs on returns to high body mass.

Beliefs on returns to healthy and unhealthy behaviours.

Other collected outcomes in the post experimental survey are: 1) nutritional knowledge information; 2) anthropometric information; 3) willingness to borrow

First order beliefs:

Beauty: How would you rate this person's attractiveness? Please, provide your answer on a scale from 1 (not at all beautiful/handsome) to 4 (very handsome/beautiful).
1-4 scale
Health: How would you rate this person's health? Please, provide your answer on a scale from 1 (not at all healthy) to 4 (very healthy).
1-4 scale
Self control: How would you rate this person's ability to resist to temptation? Please, provide your answer on a scale from 1 (not at all able to resist) to 4 (very able to resist).
1-4 scale
Life expectancy: How would you rate this person's life expectancy? Please, provide your answer on a scale from 1 (not at all live a long life) to 4 (live a very long life).
1-4 scale
Productivity: How would you rate this person's ability to get things done? Please, provide your answer on a scale from 1 (not at all able ) to 4 (very able).
1-4 scale

Beliefs on other beliefs:

Health: On average, how did other respondents rated the individual’s health? Please provide your best guess on a scale from 1 (not at all healthy) to 4 (very healthy).
1-4 scale
Attractiveness: On average, how do you think other respondents rated the individual’s beauty? Please provide your best guess on a scale from 1 (not at all beautiful/handsome) to 4 (very beautiful/handsome).
1-4 scale
Life expectancy: On average, how do you think other respondents rated the individual’s life expectancy? Please provide your best guess on a scale from 1 (not at all long) to 4 (very long).
1-4 scale
Self control: On average, how do you think other respondents rated this individual’s ability to resist to temptation? Please provide your best guess on a scale from 1 (not at all able to resist) to 4 (very able to resist).
1-4 scale
Productivity: On average, how do you think other respondents rated this individual’s ability to get things done? Please provide your best guess on a scale from 1 (not at all able) to 4 (very able).
1-4 scale

Explicit beliefs on returns to high body mass in the dating market, job market, credit market and in terms of health outcomes.

- If a person moved from S. X to S.Y, would he/she be more or less likely to find a partner?

- If a person moved from S. X to S.Y, would he/she be more or less likely to find a job?

- If a person moved from S. X to S.Y, would he/she be more or less likely to obtain a loan?

- If a person moved from S. X to S.Y, would he/she be more or less likely to develop a disease like diabetes, high blood pressure or heart disease?

- If a person moved from S. X to S.Y, would there be any other situation where he/she might be affected by the weight gain?

(All the questions refer to a Silhouette scale.)

Beliefs on returns to healthy and unhealthy behaviours are measured as in
Pietro Biroli, Teodora Boneva, Akash Raja and Christopher Rauh "Parental Beliefs about Returns to Child Health Investments" working paper 2019.

Imagine that the 10 individuals all exercise for 60 minutes every day (e.g. walking, running, working in the field) while they are 30-65 years old. Out of these 10 individuals, how many do you think will:

E1.A… be overweight at age at age 65?
E1.B… have heart disease at age 65?
E1.C… are alive at age 65?

Imagine that the 10 individuals do not exercise but engage in other activities every day (e.g. going to the salon or watching TV) while they are 30-65 years old.

Out of these 10 individuals, how many do you think will…
E2.A… be overweight at age at age 65?
E2.B… have cardiovascular disease at age 65?
E2.C… be alive at age 65?

Imagine that the 10 individuals all eat 2 large traditional meals (1 breakfast, 1 other meal) per day while they are 30-65 years old.

Out of these 10 individuals, how many do you think will…
E1.A… be overweight at age at age 65?
E1.B… have cardiovascular disease at age 65?
E1.C… be alive at age 65?

Imagine that the 10 individuals all eat 2 large traditional meals (1 breakfast, 1 other meal) plus one additional large traditional meal per day and a snack, while they are 30-65 years old.

Out of these 10 individuals, how many do you think will…
E2.A… be overweight at age at age 65?
E2.B… have heart disease at age 65?
E2.C… be alive at age 65?

This is a survey experiment.

In the first part, respondents see a portrait randomly chosen among a set of photo-morphed and original pictures. Each photo-morphed picture is obtained by photo-morphing a portrait of an individual from Kampala, either increasing or decreasing body mass.

Respondents provide their best guess, and expectations on others’ beliefs, on wealth, attractiveness, health, longevity, self-control and productivity of the individual as portrayed in the picture (see Survey tool Appendix). First order beliefs are not incentivised. Beliefs' on others beliefs are incentivised using pilot data. The experiment has two treatment arms: in 1) participants see a portrait and learn the age of the person. In 2) participants additionally learn that the person owns a car or a land title. The aim of the second treatment arm is to reduce asymmetric information in wealth, allowing to test for the signaling channel. Information treatment is randomly assigned at the individual level, to avoid lack of information to be interpreted as not-owning.

In the second part, individuals are shown a set of hypothetical loan applications and are asked to guess how the applications have been evaluated by a set of real loan officers currently employed in the greater Kampala area. The applications are created in the following way. In each application, the applicants' age, nationality and place of residence are defined statically, while the applicants' loan profile, collateral, reason for loan, occupation, income (profits and revenues), gender, portrait, body mass are cross-randomised. Body mass is randomized by assigning the high body mass or low body mass photomorphed version of the randomly selected portrait. To avoid suspicions, portraits are randomized without replacement across the full set of applications each respondents receives so that no picture is seen twice by the same respondent. Using the procedure detailed in the appendix, I first build 30 loan applications with all the cross randomized information, except body mass. Then I create the final set of 60 applications by creating two versions (a high body mass or low body mass) of each application. The 4 applications are randomly selected from the set of 60 applications, conditional on the picture in the application not being randomly selected in the first part. Answers are incentivised using the answers of real loan officers currently employed in Kampala (Ref. AEARCTR-0004528).

In the third part, respondents answer a set of questions including demographics, returns to healthy behaviours, nutritional knowledge, willingness to borrow and explicit beliefs on returns from body mass.

In the first part, the main explanatory variable of interest is body mass of the picture observed. The second explanatory variable of interest is whether information on the individuals' wealth was disclosed (car/land title ownership). Body mass is measured in two ways: 1) dichotomous variable which takes value 1 if the picture was photo-morphed to get a higher body mass, and 0 if the picture was photo-morphed to have a higher body mass. 2) a continuous measure obtained by having 8 individuals from Kampala rate each picture body mass by pointing at a 1-9 silhouette scale and averaging at the picture level.

The main statistics monitored are the difference in primary and secondary outcomes when participants are randomly assigned a high body mass vs a low body mass picture (dichotomous measure). I assess the variation in this difference when the portrayed individuals 's wealth is revealed, and when portrayed individuals' wealth is not disclosed.

In addition to computing basic statistics, I will also identify parameters estimating the following baseline specification separately by treatment arm: y_ijz = a0 + a1 BM_i + u_ijz; I will also estimate a long model: y_ijz = a0 + a1 BM_i + a2 CarInfo_i + a3 CarInfo_i * BM_i + u_ijz , where y_ijz is the rating of individual j of picture i on outcome z; BM_i takes value 1 if picture i was a higher body mass picture; CarInfo_i takes value 1 if information on car ownership for picture i. I will estimate both the simple regression, a regression which includes controls for age of picture i, and a regression which includes picture fixed effects. I also estimate the same regressions using the 1-9 scale.

Heterogeneity analysis of interest are: gender and race of the picture; gender of the respondent; body mass of the respondent (underweight vs overweight); age of the respondent (age strata), wealth level of the respondents (wealth strata).

In the second part, the main statistics monitored is the difference in the expected ease of access to credit (expected share of referral requests and expected perceived creditworthiness), as well as beliefs on returns to apply for a loan, when respondents are randomly assigned a high body mass vs a low body mass application. I will compare the expectations to the actual loan officers ratings, as collected in a related survey experiment (Ref. AEARCTR-0004528).

On top of computing basic statistics, I will also identify parameters estimating the comparable regression specification: y_ijz = a0 + a1 BM_i + u_ijz . To this baseline specification, I will sequentially add application fixed effects, respondent fixed effects and controls for attractiveness, as measured in the first part of the survey.

Computer

Within subject for the body mass randomisation; Individual level randomization for the information treatment.

Yes

500 individuals. Participants to the survey are individuals aged 20+ living in the greater Kampala area (Kampala, Mukono, Wakiso districts). The sample is stratified by gender, age and by wealth. Stratification by age is implemented by targeting three age groups: 20-30; 30-45; 45+. Stratification by wealth is obtained by ranking wards according to a poverty index based on dwelling's characteristics, share of households living on less than 2 meals per day, share of households which do not have a bank account.

Part 1: 2000 pictures (4 pictures per respondent) Part 2: 2000 applications (4 applications per respondent)

This is not an RCT.
In the first part (section Beliefs and Perception), for each picture shown, either the higher body mass or the lower body mass version is randomly selected. For the income information treatment, 250 participants per treatment arm.

In the second part (Section Ratings), for each picture shown, either the higher body mass or the lower body mass version is randomly selected.

In the third part (Section Returns), 250 respondents are assigned to the nutrition treatment and 250 are assigned to the exercise treatment.

Concerning Explicit beliefs, 1/3 of the sample is assigned a S3 to S5 weight change, 1/3 is assigned to a S5 to S7 weight change, 1/3 is assigned to a S7 to S9 weight change.

Exploratory analysis from Uganda and data from a small-scale experiment in Malawi suggest that higher body-mass index pictures are between 30 p.p. to 38 p.p. more likely to be perceived as rich, equivalent to 0.5 sd effect. Power calculations, assuming portrait fixed effects, suggest a sample of 500 (750) could detect a MDE of 0.4 sd (0.3 sd) for the primary outcome.