Baird 2013.

Study characteristics
Methods	Study design: cRCT How were missing data handled: for the analysis of the postintervention follow‐up data, the study authors stated: "We conducted all analysis among those who were successfully interviewed in Round 4, which maximises sample size for the estimation of longer‐term impacts." Randomisation ratio: 1:1 Recruitment method: after the random selection of EAs and individuals into the treatment group, the local NGO implemented the cash transfers held meetings in each treatment EA between December 2007 and January 2008 to invite the selected individuals into the programme. At these meetings, the programme beneficiary and her parents/guardians were offered the monthly transfer amounts. This consisted of a transfer to the parents, a transfer directly to the girl and payment of school fees for girls attending secondary school. Transfer amounts to parents were varied randomly across EAs at USD 4, USD 6, USD 8 and USD 10 per month, so that each parent within an EA received the same offer. Within each EA, there was a lottery to determine the transfer amount to the programme beneficiaries, which was USD 1, USD 2, USD 3, USD 4 or USD 5 per month. The lottery was held publicly to ensure that the process was transparent. Secondary school fees were paid in full directly to schools. Sample size justification and outcome used: NR Sampling method: stratified random sampling. Zomba district contains 550 EAs (Zomba city 50; rural traditional authorities 500). 176 EAs were selected from 3 different geographical strata: urban (Zomba city, 29 EAs), near rural (16 km from Zomba city, 119 EAs) and far rural (≥ 16 km from Zomba city, 28 EAs). Of the 50 EAs in Zomba city, 21 were excluded on the basis of advice from local experts who deemed these areas too affluent for the proposed intervention. In each of the 2 rural strata, with the exception of 1 TA that was unsafe for field work, the study EAs were randomly selected. In the 176 sampled EAs, each dwelling was visited to take a census of all never‐married girls aged 13–22 years (sample frame). Girls were grouped according to those enrolled in school at baseline (baseline schoolgirls), or those not enrolled in school at baseline (baseline dropouts). In the cohort of baseline schoolgirls, a subset of eligible girls was randomly selected for the study. The sampling percentages for this cohort differed by geographical stratum and age group and was 14–45% in urban areas and 70–100% in rural areas. All the girls who were not enrolled in school at baseline (baseline dropouts) were sampled to participate in the study This sampling procedure yielded a baseline study sample of 4051 girls of whom 3796 (94%) were enrolled and completed a baseline interview at the end of 2007. Of these, 889 were baseline dropouts and 2907 were baseline schoolgirls. Study aim or objective: to examine whether a cash transfer programme targeted at adolescent girls in Malawi helped empower its recipients in the short‐run, i.e. during and immediately after the 2‐year intervention. Study period: 2007–2012 Unit of allocation or exposure: females aged 13–22 years
Participants	Baseline characteristics Overall group (n = 2706): NR Intervention or exposure group (n = 1211) Age (years): baseline school girls, means: CCT group 14.9 (SD 1.8); UCT group 15.4 (SD 1.9); girls not attending school at baseline: 16.8 (SD 2.4) Place of residence: urban HH, means: baseline school girls: CCT group 0.478 (SD 0.5); UCT group 0.418 (SD 0.494); girls not attending school at baseline: 0.129 (SD 0.335) Sex: female‐headed HH: baseline school girls, n (%): CCT group 63 (26); UCT group 78 (24); girls not attending school at baseline, n (%): 90 (39) Ethnicity and language: NR Occupation: N/A Education: highest grade attended: baseline school girls, means: CCT group 7.1 (SD 1.7); UCT group 7.9 (SD 1.6); girls not attending school at baseline, means: 5.8 (SD 2.9) SES: HH size: baseline school girls, means: CCT group 6.341 (SD 2.1); UCT group 6.7 (SD 2.1); Girls not attending school at baseline, means: 6.1 (SD 2.6); electricity in dwelling: baseline school girls, n (%): CCT group 31 (28); UCT group 49 (24); girls not attending school at baseline, n (%): 24 (11); piped water in dwelling: baseline school girls, n (%): CCT group 48 (41); UCT group 135 (60); Girls not attending school at baseline, n (%): 63 (25) Social capital: NR Nutritional status: NR Morbidities: NR Concomitant or previous care: NR Control group (n = 1495) Age: years: Baseline school girls, means: 15.3 (SD 1.9); girls not attending school at baseline: 17.6 (SD 2.2) Place of residence: urban HH, means: Baseline school girls: 0.35 (SD 0.48); girls not attending school at baseline: 0.18 (SD 0.39) Sex: female‐headed HH: baseline school girls, n (%): 275 (32); girls not attending school at baseline, n (%): 93 (42) Ethnicity and language: NR Occupation: N/A Education: highest grade attended: baseline school girls, means: 7.6 (SD 1.6); girls not attending school at baseline, means: 6.2 (SD 2.9) SES: HH size: baseline school girls, means: 6.375 (SD 2.262); girls not attending school at baseline, means: 6.12 (SD 2.388); electricity in dwelling: baseline school girls, n (%): 86 (20); girls not attending school at baseline, n (%): 16 (7); piped water in dwelling: Baseline school girls, n (%): 277 (47); girls not attending school at baseline, n (%): 64 (29) Social capital: NR Nutritional status: NR Morbidities: NR Concomitant or previous care: NR Inclusion criteria: never‐married girls aged 13–22 years enrolled in school at baseline (baseline schoolgirls), and those not enrolled in school at baseline (baseline dropouts) in selected 176 EAs. Exclusion criteria: NR Baseline imbalance: in the girls who attended school at baseline, those in the CCT group were younger than those in the UCT group, and, therefore, also a lower grade attained. Attrition per relevant group: reported at postintervention follow‐up (2 years after intervention): baseline school girls: control group: 10.7%; intervention group: NR. Baseline dropouts: control group: 15.7%; intervention group: NR. Reported at second postintervention follow‐up (4 years after intervention): Baseline school girls: control group: 12.5%; intervention group: NR. Baseline dropouts: control group: 15.7%; intervention group: NR Description of subgroups measured and reported: NR Total number completed and analysed per relevant group: NR Total number enrolled per relevant group: baseline school girls: CCT intervention group: 685 girls from 27 intervention EAs; UCT intervention group: 526 girls from 46 intervention EAs. Control group: 1495 girls from 88 control EAs. Baseline dropouts: CCT intervention group: 436 girls from 88 intervention EAs. Control group: 453 girls from 88 control EAs Total number randomised per relevant group: intervention group: 88 EAs; control group: 88 EAs
Interventions	Intervention: CCT and UCT Food access intervention category: increase buying power Description: CCT group: payment received if girl attended school for 80% of days that school was in session during previous month. UCT group: payment received if girl attended cash transfer points. Cash transfers were split between guardian and girl. HH amount varied randomly (using computer‐generated random numbers) by EA, with monthly values of USD 4, USD 6, USD 8 or USD 10. Girl amount varied randomly, with monthly values of USD 1, USD 2, USD 3, USD 4 or USD 5, decided by drawing numbers from an envelope. The transfer amounts offered to the parents were randomised at the village level, and those offered to the girls were randomised at the individual level. Duration of intervention period: January 2008–December 2009 (coincided with 2008 and 2009 school years) Frequency: monthly cash transfers Number of study contacts: October 2007–January 2008 (baseline); October 2008–February 2009 (follow‐up during intervention period); February–June 2010 (postintervention follow‐up) and 2012 (postintervention follow‐up) Providers: 2 NGOs Delivery: local distribution points Co‐interventions: NR Resource requirements: NR Economic indicators: NR Control: no intervention
Outcomes	Dietary diversity: number of times respondents ate protein‐rich food in past 7 days Cognitive function and development: cognitive test scores (Raven's Coloured Progressive Matrices) Anxiety and depression: psychological distress (GHQ‐12 score)
Identification	Sponsorship source: Global Development Network, the Bill and Melinda Gates Foundation, NBER Africa Project, World Bank Research Support Budget Grant, 3ie Open Window (Round 2) as well as several trust funds at the World Bank: Knowledge for Change Trust Fund (TF090932), World Development Report 2007 Small Grants Fund (TF055926), Spanish Impact Evaluation Fund (TF092384) and Gender Action Plan Trust Fund (TF092029). Country: Malawi Setting: urban and rural HHs, Zomba district Author's name: Sarah J Baird Email: sbaird@gwu.edu; bozler@worldbank.org; ctmcintosh@ucsd.edu Declarations of interest: yes; no conflicts of interest Study or programme name and acronym: Schooling, Income, and Health Risks study (SIHR) Type of record: journal article, policy research working paper Trial registration: none reported Protocol availability: no
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Low risk	Clusters (EAs) randomly assigned to intervention or control groups with computer‐generated random numbers. Within intervention group, EAs were randomly assigned to 1 of 2 intervention groups (UCT or CCT groups). Girls in the intervention EAs were allocated to different cash transfers by drawing numbers from an envelope.
Allocation concealment (Selection bias)	Low risk	Allocation by EAs at start of study. Codes for randomisation of the EA into trial groups, and for random assignment of different cash transfer amounts, were written by 1 of the study investigators. The programme field manager then studied staff with EA identification numbers and individuals selected.
Baseline characteristics similar (Selection bias)	Low risk	For baseline school girls, those in CCT group were younger than those in UCT group, and, therefore, also attained a lower grade. The study authors controlled for this difference in their analysis of programme impacts.
Baseline outcome measurements similar (Selection bias)	Unclear risk	Relevant baseline outcome measurements, e.g. food consumption not presented by study authors.
Blinding of participants and personnel (Performance bias)	Low risk	Quote: "study participants were not masked to their assignment, but did not know what the comparison groups were because they were assigned at the enumeration area level."
Blinding of outcome assessment (Detection bias)	High risk	Unclear whether the outcome assessors were blinded but outcomes were self‐reported and likely to be influenced by lack of blinding.
Protection against contamination (Performance bias)	Low risk	To reduce the possibility of crossover from intervention group to control group, participants were assigned to trial groups on basis of random assignment of EAs that they lived in.
Incomplete outcome data (Attrition bias)	Low risk	Loss of clusters: low risk. Study authors stated that none of the EAs had complete LTFU. Attrition rates at second postintervention follow‐up were reported as 15.7% in control group of baseline dropouts. Study authors reported that girls in this group who received CCT and who were from urban areas, were more likely to be LTFU. In baseline school girls, attrition was 12.5%, with no differences in baseline characteristics of those who remained in the study, compared to those LTFU.
Selective outcome reporting (Reporting bias)	Low risk	Outcomes reported were in line with those prespecified in the trial registry preanalysis plan.
Other bias	Low risk	Misclassification bias: unclear. Recruitment bias (cRCT): low risk. A baseline survey was conducted among all eligible participants in all the EA before randomisation of EA. Measurement bias: low risk. Outcomes reported at baseline and follow‐up included the number of times the respondent ate protein‐rich food in the 7 days prior. Incorrect analysis: unlikely. Seasonality bias: unclear risk. Low risk of seasonality bias for follow‐up during the intervention period; however, it was unclear whether the study authors adjusted for seasonality in the postintervention follow‐up analysis.