Skin Color and Secondary Education in Ghana

Scholarships for 4-year of secondary school were randomly awarded to 682 students among 2,064 Ghanaian youths who had been admitted into Senior High School as of Fall 2008, but had not enrolled by January 2009. Scholarships were paid directly to the schools and covered the entire tuition fees and other school costs (e.g., exam fees).

2009-01-01

2012-07-01

We will test for heterogeneity in impacts of scholarships by skin color, on the following outcomes:
Educational outcomes: score on achievement test (2013); highest education grade completed;
Family formation outcomes: partner’s characteristics; age at first pregnancy;
Labor market outcomes: employment; earnings.

To investigate mechanisms, we will test for heterogeneity in impacts of scholarships by skin color, on the following outcomes:
- Attitudes and preferences: aspirations, willingness to express opinions, extroversion, conformism, expectations of returns from education;
- Parental and teacher investment
- Self-efficacy

Components of Parental and Teacher Investment Indices:
1) When I was in school, I could count on my teachers to help out when I had a problem;
2) When I was in school, my teachers thought that I was capable of doing the work;
3) I worried about getting picked on or teased at school;
4) I felt that kids treated me kindly at school or in social situations;
5) My parent/guardian was involved in my schooling when I was a child;
6) My parent/guardian criticized me or my ideas when I was a child;
7) My parent/guardian trusted my judgement when I was a child.

Self-efficacy Index:
1) I can solve difficult problems if I try hard enough;
2) I feel that I am a person of worth, at least on an equal level with others;
3) I can deal efficiently with unexpected events;
4) I have little control over the things that happen to me.

See AEA-RCT-00015

Our primary outcome data are obtained from the 2013 in-person survey and callback surveys conducted on a regular basis from 2016 to 2023 by Duflo, Dupas and Kremer (2023) and Duflo et al. (2024). We added two sets of questions in 2023 to collect information on parental and teacher investment and self-efficacy, which may be pathways for skin tone to yield disparities in outcomes. Each question is asked on a 5-point scale from Strongly Disagree to Strongly Agree, and we will convert these responses to binary variables (one if the respondent Strongly Agrees or Agrees). These newly added questions are listed above under "secondary outcomes (explanation)".

Not available

randomization done in office by a computer

Individual

No

N/A

Photographs were taken of participants at baseline (2008) and again at midline (2013) to help facilitate tracking for future surveys. In summer 2023, participants were contacted to obtain consent for these photographs to be used for analysis. Though 2,064 participants were in the original study, we only were able to reach 1,659 (81%) to ask their consent to use their photo for this study. Of those, only 1,578 consented (76% of the overall sample). We have skin tone scores for 1,243 students in 2008, 1,313 in 2013. For 1,042 students, we have scores for both 2008 and 2013. The gap between the consented number and the recorded number come from cases where we could not find the correct image for that participant. This study will focus on the sample of respondents who provided consent and whose picture could be found. If the consent rate is not balanced by treatment status, we will use entropy balancing to reweight the sample for analysis.

The initial sample was 1/3 T, 1/3 C. We have not yet matched the pictures to the treatment status so we do not know the breakdown for the subset that consented and for whom a picture is available.