Developing Youth Groups in Bangladesh to Support Peers at Risk of Early Marriage

The intervention aims to empower youth groups (including both girls and boys in the age range 16-25) to tackle the problem of female early marriage within their communities through paralegal training and mentoring and supervision by experienced paralegals. The aim of the intervention is to reduce female early marriage and school dropout, and improve agency of adolescent girls and young women in Bangladesh.

2023-06-01

2025-06-01

schooling, marriage and childbirth

School enrolment, marital status, childbirth and related information will be self-reported by the intended programme beneficiaries and their guardians during the baseline and endline surveys.

aspirations

Data collection on aspirations will be based on survey questions the achievements along various dimensions (e.g. education, employment marriage, children) that respondents believe it is important for a young woman to “get by” and “live well”.

We will identify 100 urban wards across four districts for inclusion in the study. 50 of these will be randomly assigned to the treatment arm and the remaining 50 will be assigned to the control arm.

Not available

The randomisation of treatment across the study neighbourhoods will be conducted by the Principal Investigator using computer software (STATA's built-in randomisation functions).

The unit of randomisation is an urban 'ward', the smallest administrative unit in urban Bangladesh.

Yes

The intervention will be conducted in low-income urban neighbourhoods, each covering an area of about 600 households, in 50 wards in four districts of Bangladesh: Dinajpur (15 wards), Dhaka (15 wards), Khulna (10 wards), Patuakhali (10 wards) (a ‘ward’ is the smallest administrative unit in urban areas of Bangladesh). 50 additional wards from the same districts will also be included in the study as controls.

The study sample will consist of 4,000 households (40 households per community x 100 communities). Using the stratified random sampling described above, we anticipate that the sample will include approximately 3,500 adolescent girls and 1,800 young women.

50 treatment clusters and 50 control clusters. Each cluster is a neighbourhood within an urban 'ward', the smallest administrative unit in urban Bangladesh.

We have conducted power calculations for the key outcome variables which are shown in Figure 2 in the Analysis Plan. The figure shows the required number of treatment clusters for various effect sizes for two key outcome variables (i) probability of marriage below 18 at the end of the intervention period – 2 years – for girls aged 13-17 years at baseline; (ii) total years of schooling at the end of the intervention period for girls aged 13-17 years at baseline. For these calculations we use alpha = 0.05, power = 0.80, equal number of clusters in treatment and control arms, and obtain the intra-cluster correlations and estimates of variance from the 2014 WiLCAS (rho = 0.053 for years of schooling, and rho = 0 for probability of marriage below 18; sd = 2.4 for years of schooling). In the figure, we also show the power calculations for two alternative scenarios: (a) rho = ICC in 2014 WiLCAS + 0.01; (b) rho = ICC in 2014 WiLCAS + 0.02. As the figures show, for all scenarios considered, 50 treatment clusters are sufficient to detect an effect size of a decrease in the rate of early marriage by 0.06% points or more; and an increase in schooling by 0.045 years or more, at the 5% significance level.