Experimental Design
The study is divided into two key parts. One part relies on a survey to document the main assumptions made in the model, i.e., about victimization and social taxation. This part relies primarily on Vignettes and Likert scale questions asked to a large stratified random sample of urban and peri-urban dwellers in and around Abidjan, Côte d'Ivoire. The other part investigates the existence of costly behavior aimed at manipulating one's perceived socio-economic rank in a group. This part is implemented using various lab-in-the-field experiments.
Part A: Lab-in-the-field experiments
1. Experiment “Tombola”: we offer respondents the possibility to choose the prize they would like to receive if they win a raffle that will take place at the end of the data collection. Respondents have three choices: a basket of basic goods, a basket with luxury items, or a contribution to charitable organizations. All three choices have the same monetary value, and this is made clear to participants.
The treatment arms are as follows:
- PRIVATE: respondents record their choice on their own with the tablet, the enumerator doesn't see what they sign up for.
- ENUMERATOR-AWARE: respondents must tell the enumerator which item they would want.
- PUBLIC: respondents sign up on a sheet of paper that has one column per item. The sheets have been prepared in such a way that, for a subset, the first row contains the name of the respondent, and the following rows contain fictional names resembling those of people in their neighborhood; for another subset, all rows are blanked and the respondent is asked to write their name and their choice. In both cases, the respondent should expect that participants surveyed after her will be given the same sheet of paper and hence observe her choice.
We use this to test the following hypothesis: respondents are less likely to choose the “status” goods (luxury basket or contribution to charity) in the PRIVATE treatment compared to the two NON-PRIVATE treatments.
2. Experiment “Photo”: we randomly assign people to treatment conditions as follows:
- Control: we ask respondents their consent to be taken a picture for our records as part of the data collection audit procedures.
-Treatment: we ask respondents their consent to be taken a picture for our records and to be used by a local panel to select participants for a video collecting views on the way people in Abidjan enjoy themselves.
We offer the option to pay a time/effort cost to provide a nicer picture at a later time in case the respondent wants the opportunity to improve their appearance.
We use this to test the following hypothesis: respondents are willing to pay (in time/effort) to produce a photograph that make them look higher-rank in order to improve their chances of being selected for the documentary.
In addition, we test that having a second photograph taken makes the person look higher-rank. To this effect, after the survey, we invite a panel of individuals from Abidjan to rank all the photographs taken by the likely socio-economic status of the person in the picture. We then test whether the person is ranked higher socially in the second photograph than in the first. As control, we also ask the panel to rate the attractiveness of the applicant, to net out its possible correlation with reported SES rank. This analysis is disaggregated by gender to test the hypothesis that appearances are a more informative signal about socio-economic status for women than for men.
3. Experiment “Documentary”: We tell respondents that we are recruiting people to participate in two documentaries about Abidjan. We are seeking two types of people: (1) people who can talk about current hardship and challenges they are facing (“Babi* est dure”) (note: Babi is the affectionate nickname that inhabitants of Abidjan have for their city)); and (2) people who can talk about successes (“Abidjan: terre d’opportunités”). We ask participants in which documentary they’d be willing to participate, if any. The same treatment arms as the experiment Tombola are used to elicit this choice: PRIVATE, ENUMERATOR, or PUBLIC. After choosing the documentary they want to participate in, we offer participants the option to pay a time/effort cost to provide a nicer picture at a later time in case they want a chance to improve their appearance.
We use this to test the following hypotheses: respondents are more likely to choose the “hardship” documentary in the PRIVATE treatment compared to the NON-PRIVATE treatments; respondents who choose the documentary on successes are more likely to have a second picture taken, and this picture is more likely to be rated as signaling a higher rank. This analysis is disaggregated by gender to test the hypothesis that appearance is a more informative signal about socio-economic status for women than for men.
4. Experiment “Social preferences”: in this experiment we capture the preferences/beliefs about what others should receive as compensation through 2 questions. We tell respondents that we are organizing a documentary and we ask: (1) What compensation should we give people who participate in the documentary; a basket of necessities or a luxury item? and (2) Do you think we should let people choose by themselves or should we just choose for them?
We randomize respondents into two groups:
- SUCCESS: we tell them the people we need to compensate are giving testimonials of success.
- HARDSHIP: we tell them the people we need to compensate are giving testimonials of hardship.
We use this to test the following hypotheses: (a) respondents are less likely to choose the “status” goods (luxury basket) in the HARDSHIP treatment compared to the SUCCESS treatments; (b) respondents are less likely to say people should be allowed to choose for themselves in the HARDSHIP treatment compared to the SUCCESS treatments.
PART B: Vignettes
There are three sets of vignettes requiring the respondents to state the suitability (on a scale from 1 to 10) of a certain individual to the different scenarios (e.g., being selected for an interview, being suspected of a crime, being invited to a social gathering, etc.) depicted in the vignettes
1. The first set of vignettes consist of 6 different profiles (i.e., short CV) and 4 scenarios. Respondents are presented randomly 3 out of the 6 profiles in each scenario. The vignettes in this set will help identify beliefs about occupation and education.
2. In the second set of vignettes, respondents are presented photographs instead of profiles. The second set consists of 4 scenarios. Respondents are presented with 3 photographs (randomly) in each scenario. There is a pool of 18 photographs of 6 different people. Each person is photographed dressed in a low, medium, or high SES outfits and respondents are never presented the same person in different outfits.
3. The third set of vignettes consists of 3 scenarios and 10 different profiles. In addition to standard profile features such as those used in set 1, each profile indicates whether the profiled applicant chose (1) a free meal or (2) a keychain and a tote bag as compensation for coming to a job interview. The choice made by the person in the scenario is part of their profile. This vignette captures the extent to which consumption patterns may be used to make value judgements.
Heterogeneity
We will estimate heterogeneity in choices in the experiments by own socio-economic status, following our theoretical framework which highlights that incentives to manipulate appearances are heterogeneous across the SES distribution.
Covariates
Our survey will collect information on covariates that we expect to be important determinants of choices and our main specifications will control for those to increase precision. This includes, in particular, gender, age and education level.