Education as a legacy

Pre-Registration Worksheet Study Name: Education as a Legacy in Northern Nigeria Have any data been collected for this study yet? ☒ (a) No data have been collected ☐ (b) Some data have been collected but not analyzed ☐ (c) Some data have been collected and analyzed Hypothesis: What’s the main question being asked, or hypothesis being tested? Hypotheses Among parents in Northern Nigeria, a treatment which frames education as a legacy which can be bequeathed to one’s grandchildren, compared to a control which suggests the benefits of nutrition to the immediate family, will: Increase intentions to devote resources to girls’ education (self-report assessed immediately). This will include reports of a. intention to increase savings for girls’ education vs other expenditure b. intention to increase financial allocation to daughters’ education c. intention to increase investment in daughters’ education d. greater importance accorded to daughters’ education Give a brief (1-2 sentence) overview of the methods. We will randomly assign 600 parents who live in the environs of Abuja (especially in communities where there are schools for children yet enrolment rates for girls are lower than those of boys) to either the treatment (education as a legacy) condition or the control (value of nutrition) condition. We will thereafter assess effects on investment in daughters’ education. Independent variables: Describe the conditions (for an experimental study) or predictor variables (for a correlational study). Participants who are parents will randomly be assigned to one of two between-subjects conditions: A control condition consisting of 3 testimonials and 1 recall question highlighting the benefits of good nutrition for children in their education. A treatment condition consisting of 3 testimonials and 5 open-ended, saying-is-believing questions suggesting that education of their daughters could help these girls become better parents themselves and bequeath intergenerational benefits to all their grandchildren. Dependent variables: Describe the key dependent variable(s) specifying how they will be measured. Measure 1: Scholarship (Allocation task 1, total must equal N500,000) Sometimes the Federal Capital Territory puts on a city-wide scholarship lottery. Imagine that you have 2 children, and you have no money set aside for their education, so you enter the lottery for this scholarship. You have two children: ● Version A ○ A boy who is 12 & a girl who is 11 (Counterbalancing version B) ● Version B ○ A girl who is 12 & a boy who is 11 (Counterbalancing version A) The scholarship lottery is worth N500,000. Imagine you win the lottery. And you can put this money in an education savings account for each child. How much would you put out of the N500,000 in each child’s education savings account? Version A Amount to the boy who is 12: N___________ Amount to the girl who is 11: N___________ Version B Amount to the girl who is 12: N___________ Amount to the boy who is 11: N___________ We will analyze the amount allocated to the girl (0-500,000). Measure 2: Advice to cousin (Allocation task 2, total must equal N70,000) Imagine that you have a younger cousin who lives in your community. His business is doing well, and he finds himself making an additional N70,0000 per month. Imagine that this cousin wants your advice on how to spend that N70,000 each month. There are 3 options he is considering. I’d like to know how much you would advise him to put aside each month for each option. You could advise him to allocate all of the money to one option, none of the money to one option, or something in between to each option. N___ Save for future school expenses for his 12-year old daughter N___ Save for a medium sized generator N___ Save for the daughter’s wedding ceremony in the future We will analyze the amount allocated to the daughter’s education (0-70,000). Exploratory: We will assess the difference score between savings for the daughter’s school expenses and savings for the daughter’s wedding ceremony. Measure 3: Importance of education (3 items per gender on a 4 point scale from 1=Not at all to 4=Very) How important is/was it to you for your son/daughter to finish senior secondary school? How important is it to you for son/daughter to START post-secondary school? How important is it to you for son/daughter to FINISH post-secondary school? In this measure, we will “pipe-in” the name of the parent’s oldest son/daughter who is still in school and participants will respond in terms of this individual person. We will conduct analyses on the daughter’s score alone for each item separately. If there are similar patterns across the items, we will collapse them into one score by averaging them. Exploratory: We will conduct mixed-model analyses with condition as a between subjects factor and items as within subjects factors, assessing the interaction with gender and with education level, as designated by the item. Measure 4: Willingness to pay measure (3 items per gender on a 4 point scale from 1=Definitely no to 4=Definitely yes) Two of your children (oldest son and daughter still in school) get admission around the same time into a good senior secondary school that is not far from your community. Here is the first situation. The school offers a complete scholarship. How likely would you be to send your son/daughter to this school? Here is the second situation. The school offers a partial scholarship. How likely would you be to send your son/daughter to this school? Here is the third situation. The school offers no scholarship and all fees must be paid by you. How likely would you be to send your son/daughter to this school? Here, we will once again “pipe-in” the name of the parent’s oldest son/daughter who is still in school and parents will respond in terms of this individual child. We will conduct analyses on the daughter’s score alone for each item separately. If there are similar patterns across the items, we will collapse them into one score by averaging them. Exploratory: We will conduct mixed-model analyses with condition as a between subjects factor and items as within subjects factors, assessing the interaction with gender and with scholarship level, as designated by the item. Exploratory: If participants answer in a consistent, ordered way across scholarship levels (i.e. respond subsequent answers after a ‘no’ are also rated as a ‘no’), then we will consider conducting a survival analysis in which we code responses into a binary variable of Y/N. Analyses: Describe what analyses (e.g., t-test, repeated-measures ANOVA) you will use to test your main hypotheses. Continuous measures We will use linear regression for continuous measures, regressing on treatment condition. We will also conduct mixed-model analyses with condition as a between subjects factor for certain outcome measures as described above. Likert type scales Although the Likert type scales are on an ordinal scale, we will treat them like numeric data and analyze them using linear regression. Where necessary, we may also treat them as ordinal data and analyze them using ordinal logistic regression. Nominal measures These are binary variables which will be analyzed using logistic regression, regressing on condition. Covariates For counterbalanced measures, we will add a covariate for order if it is significantly associated with the outcome. In all analyses, we will control for household income standardized for the number of children, number of children, primary caregiver’s level of education, and the education level of the eldest child if each is significantly associated with primary outcome measures. In the case that sociodemographic measures are imbalanced across treatment conditions at the 5% level, these will be controlled for in the analysis. More analyses. Are there any secondary analyses you plan to conduct? (e.g., order or gender effects) Other analyses could explore how the following variables moderate the effect of the treatment: Respondent’s religion (Muslim and Christian/Other), Age of respondents Educational level of first daughter/son Number of children Sample. Where and from whom will data be collected? How will you decide when to stop collecting data (e.g., target sample size based on power analysis, set amount of time)? If you plan to look at the data using sequential analysis, describe that here. To achieve power of 0.8 and detect a minimum effect size of Cohen’s d = 0.23 for continuous outcomes and of Cohen’s h = .16 for binary outcomes for computing treatment effects , the target sample size is 300 for each condition or 600 in total. We have identified a list of 6 communities in and around Abuja with both free and paid schools for children yet the enrolment rates for girls are lower than those of boys. We will visit these communities in person to recruit participants (aged between 18 and 45) who have at least one child. We plan to stop once 600 surveys are completed. Pre-registration written by (initials): AD Pre-registration reviewed by (initials): CT