The Aarambha-cohort-II baseline evaluation adopted a pre-post research design to measure changes that can be attributed to the project interventions. The evaluation was guided by the longitudinal mixed-method approach, comprising of quantitative and qualitative data collection techniques. The quantitative survey comprised of the household survey with parents of the girls (400 sampled households) and the girls ‘survey including the ASER tool was conducted with 400 sampled OOS girls. The qualitative data collection comprised of Focus Group Discussion (FGD) and Key Informant Interview (KII) with various stakeholders.Aarambha project is funded by the UK AID flagship, Girls’ Education Challenge (GEC) program’s Leave No Girls Behind (LNGB) window. The project is being implemented by People in Need (PIN) Nepal and aims to mitigate the risk of early marriage among Out-of-School (OOS) adolescent girls in order to uplift their social status and help them lead healthy, safe, and educated lives. The project works in two districts of province-2, namely Bara and Rautahat. Over the period of four years, the PIN will work with four cohorts and aims to reach 8,500 young married and unmarried OOS girls, 17,000 of their family members, 4,000 in-school girls and 4,000 in-school boys, 400 newly elected local government officials, and community/religious leaders. The implementation capacity of the project is to work directly with approximately 2125 girls each year (i.e., evaluation for each year will have approximately 2125 MOOS girls). In line with the main objective, the project implemented Community Learning Center (CLC) classes targeting the marginalized OOS girls, to enhance their learning proficiency, eventually preparing their transition into formal schooling. Apart from the learning intervention, the project also is teaching girls life skills that include financial literacy, social skill, and family planning, training school teachers to create a safe learning environment, organizing gender-transformative workshops with in-school adolescents, and training MOOS girls’ families and the local government officials. For Cohort II, a pre-post research design was used unlike for Cohort I where a quasi-experimental research design was used to compare the findings between treatment and control groups. Although there is a difference in regards to population and research design in Cohort II, the findings from both cohorts are similar. Barriers such as poor household, unsupportive parents, repressive parental attitude existed in cohort II as well which refrained girls from accessing education.

The sample for baseline evaluation was calculated based on the sampling framework. After having the sampling frame, sample size was calculated based on GEC evaluation guideline, which suggested using minimum standards (p0=0.58, pa=0.50, Power=80%, Power=80%, Confidence interval= 95%, Margin of error = 0.05, Test=2-sided test). Stat.ubc.ca website suggested by FM was used to calculate the total sample. Adding a further attrition buffer of 30%, the final sample size was 395, rounding off to 400. Hence, 400 girls were selected as samples covering at least 50% of the CLCs so that clustering need not be applied for sample size calculation. Once the sample size was calculated, stratified random sampling was done to select the targeted OOS adolescent girls to draw out individuals for a baseline from the sampling frame. The sample for each sub-group was drawn from a total number of beneficiaries and was divided proportionately among each subgroup i.e OOS girls of age 10-14 and 15-19. Additionally, the project has also envisioned learning outcomes as one of the transitions pathways for OOS girls; since the project has not yet defined the proportion of girls who will transition, the sample size for transition was the same as the learning sample. The sample selected for the evaluation was fully representative. Representativeness of sampling was ensured considering the project’s marginalization framework, following inclusion criteria to select the primary beneficiaries. CLCs from each district were stratified based on their location. From the stratified list, two CLCs from each district were randomly selected for qualitative data collection. CLC was considered to be the primary sampling unit. A purposive sampling method was adopted to identify respondents in the communities in order to yield rich information on the status of girls’ education, early marriage, and other underlying contexts. Purposive sampling was also useful to ensure representativeness in the qualitative discussions. The methods used for data collection were focus group discussions (FGDs) and key-informant interviews (KIIs). A total of 12 FGDs and 18 KIIs were conducted with direct and indirect beneficiaries. OOS girls aged 10-14 and 15-19, along with CLC facilitators, parents, change champions, and social mobilizers were consulted. To triangulate the information gathered, parents and social mobilizers from a different location other than the sampled location were informally consulted. In order to gather overall community perspective, headteachers, were also interviewed in which the girls were more likely to enroll into were mapped and visited.