Experimental Design
We will measure impact of Idara Taleem-o-Agahi (ITA)’s remedial learning programs on female education and empowerment through a village-clustered randomized evaluation. The programs consist of a 6-month program (for girls with some schooling) and a 3-month program (for girls with no previous schooling). In the treatment villages, all households with eligible girls will be offered the opportunity to enroll their adolescent girls in either the six-month or the three-month functional literacy program, depending upon the girl's prior schooling record. We will maintain status quo in the control villages.
Our study will take place in Bahawalpur and Muzaffargarh districts in Punjab. In both districts, we take the universe of villages from administrative sources, and restrict to those villages that have a minimum number of households, and those that have never before received the program. Among villages that meet these eligibility criteria, we have done pair-wise matching by district and village size; within each pair, one village is randomly assigned to either receive ITA’s programs or not.
ITA is only able to implement their program if villages meet certain criteria, namely sufficient community support and interest, as well as a location to hold the lessons. We refer to this phase as the “mobilization” phase. Therefore, we randomly ordered all pairs so as to create a “priority” list in the event a treatment village failed mobilization,a new treatment village could take its place and the corresponding pair could be added to the sample. This priority ordering was stratified by geographic area.
In Muzaffargarh, we conducted a baseline survey in 102 villages, consisting of 51 pairs split between treatment and control villages. In each village, we conducted a listing exercise and then surveyed approximately 25 households with an eligible (never enrolled or dropout) adolescent female, giving us a sample of roughly 2,500 households. Following our baseline survey, ITA entered treatment villages and conducted a mobilization exercise. After the mobilization phase, only 26 of the 51 treatment villages that we conducted baseline in were able to have the program launched. We used the priority list and then found replacement treatment villages so as to arrive at the total amount of 35 program villages. Therefore, we only have a baseline survey in 26 out of 35 treatment villages, or 25 girls per village*26 villages*2(corresponding control village) = 1,300 adolescents with a baseline survey. We plan to conduct an endline survey of all of these same adolescents at the conclusion of the program. Pending budget, we may also conduct endline visits in both treatment and control villages at which we lack a baseline.
In Bahawalpur we were not able to begin data collection prior to the implementation of the program and thus do not have a true baseline survey for this district. In Bahawalpur treatment villages, we implemented a household-level randomization of the 3 month program, conducting a “midline” survey after the completion of one wave of the program and before the beginning of the second wave of the program (cohorts 1 and 2, respectively). We obtained from the implementing partner the names of all girls who were eligible for the programs and whose households expressed interest during a community mobilization activity; there were 55 eligible girls from each village on average in this data. Then, we randomized households into two cohorts (Cohort 1 or Cohort 2). Randomization was done at the household-level so that all girls of the same household could participate at the same time. We plan to survey approximately 55 girls after Cohort 1 concludes and prior to the start of Cohort 2 giving us a tentative sample of 55*34 = 1,870 for the midline data collection. We will also survey the ITA teacher in each treatment village to collect data on attendance. Pending budget, we will conduct a listing exercise and an endline in the 70 districts of Bahawalpur (half treatment and half control).
Therefore, across the two districts we will ideally collect endline data on 30 households with an eligible adolescent female in the 70 pairs of villages, half treatment and half control. This brings the total sample size for the endline to 140*30 = 4200 households. This will be the target sample size for the endline.