Evans 2014.

Study characteristics
Methods	Study design: cRCT Study grouping: parallel group How were missing data handled? Authors carried out both ETT and ITT regressions. However, in ETT analyses, treated HHs were those assigned to the intervention and actually received the intervention, whereas in an ITT analysis 'treated' HHs were those that were assigned to intervention, regardless of whether they actually received it. It appears as data not collected was excluded from the analysis. Randomisation ratio: 1:1; 40 intervention and 40 control communities Recruitment method: HHs were invited to enrol in the pilot. Enrolment of beneficiaries carried out in each community, with the enrolment process lasting 1–3 days, depending on total number of beneficiary HHs in the community. The enrolment team identified who would receive payments in each HH (usually the mother of the children in the HH if present), updated family information, linked children and the elderly with schools and health centres, provided an orientation session about the programme, and provided identity cards. Sample size justification and outcome used: once all communities were assigned into groups, power calculations identified the need to interview a mean of 25 HHs per community. With a total of 80 participating communities (40 treatment and 40 control) and a standardised effect size of 0.20, it was expected to need to interview 20 HHs per community to achieve 80% power. 25 HHs per community were then interviewed since not every HH would have vulnerable children: some few HHs would only have vulnerable elderly people. Calculation assumed 95% CIs for statistical significance and an intracluster correlation of 0.05. Sampling method: pilot study implemented in districts and communities targeted under TASAF I, which targeted the poorest and most vulnerable districts of Tanzania using a rigorous selection process. Regions were ranked using several indicators (poverty level, food insecurity, primary school gross enrolment ratio, access to safe water, access to health facilities, AIDS case rates and road accessibility). Districts were then prioritised within the regions using an index of relative poverty and deprivation constructed using data from Tanzania's 1992 Income and Expenditure Survey. Targeting done using screening forms designed to identify vulnerable children and elderly people based on specific criteria, which were defined by the communities themselves. The CMCs used these poverty indicators to identify the poorest (approximately) half of HHs in the community. Validation of the list of eligible HHs was done by the village assembly, allowing for community validation. They ranked HHs by priority. Random selection of control and treatment communities was done after identification of vulnerable HHs in all 80 communities. Study aim or objective: pilot project aiming to develop operational modalities for the community‐driven delivery of a CCT programme through a social fund operation; and test the effectiveness of the community‐based CCT model and ensure that lessons from the pilot informed government policy on support for vulnerable families. Study period: 31–34 months: January 2010 (when first payments were made) to October 2012 (endline survey). Unit of allocation or exposure: communities (with random selection of HHs within communities allocated to each intervention group)
Participants	Baseline characteristics Intervention or exposure Age: adults aged 18–59 years, n: 1.08; children, n: 1.69; elderly people, n: 1.91 Place of residence: villages in Tanzania Sex: NR Ethnicity and language: NR Occupation: HHs in agricultural self‐employment, %: about 72 Education: child ever in school (% children): 78.36; child (aged 6–17 years) now in school: 86.98; repeated a grade (% children in school); taken a national examination (% children in school): 98.09; own exercise books (% children in school): 94.48 SES, % HH: bank account: 1.6; borrowed part year: 19.3; improved roof: 33.0; improved floor: 3.0; improved toilet: 69.1; piped water: 30.2; electricity: 0.0 Social capital: contributed labour to a community development project, % HHs: 36.25. Can trust people in community, % respondents: 58.68. Can trust community leaders, % respondents: 80.87 Nutritional status: NR Morbidities: disabled, n in HH): 0.42; hospitalised, n in HH last month): 0.05; ill past month, % individuals: 31.3; taken medication, % individuals with health problem: 87.9; ill in past year, % individuals: 65.3 Concomitant or previous care: NR Control Age: adults aged 18–59 years, n: 1.04; children, n: 1.61; elderly people, n: 1.32 Place of residence: villages in Tanzania Sex: NR Ethnicity and language: NR Occupation: HHs in agricultural self‐employment, %: about 72% Education: child ever in school, %: 83.28; child (aged 6–17) now in school, %: 89.23; repeated a grade, % children in school; taken a national examination, % children: 98.10; own exercise books, % children in school: 95.56 SES, % HHs: bank account: 2.1; borrowed part year: 18.2; improved roof: 37.2; improved floor: 8.7; improved toilet: 31; piped water: 31.6; electricity: 1.3 Social capital: contributed labour to a community development project, % HHs: 35.27. Can trust people in community, % respondents: 52.58. Can trust community leaders, % respondents: 80.07 Nutritional status: NR Morbidities: disabled, n in HH: 0.44; hospitalised, n in HH last month: 0.04; ill past month, % individuals: 29.5; taken medication, % individuals with health problem: 90.1; ill in past year, % individuals: 63.8% Concomitant or previous care: NR Overall Age: NR Place of residence: villages in Tanzania Sex: NR Ethnicity and language: NR Occupation: HHs in agricultural self‐employment, %: about 72% Education: NR SES: improved floor (concrete/wood/tiles), %: 6, mud floor, %: 94; almost always use pit latrine, %: 71; access to piped water, %: 31; lack access to electricity, %: 99.0. Social capital: trust in community leaders, %: 80 Nutritional status: NR Morbidities: HH members reported being ill in the last month, %: about 30. Concomitant or previous care: HHs receiving ≥ TZS 5000 from Government/TASAF, %: 3.8; from NGO/religious organisation, %: 4.9 Inclusion criteria: HHs with vulnerable children (1 parent or both parents deceased; abandoned children; having 1 or 2 chronically ill parents (e.g. HIV/AIDS); chronically ill children, despite having 2 parents alive. Vulnerable elderly people defined as: elderly with no carers, poor health, very poor. Communities in the selected 3 districts. Exclusion criteria: none specifically reported. Pretreatment: HHs in treatment communities were less likely to have houses with improved floors or electricity. Control communities had slightly more elderly people, HHs electricity and improved roofs, floors and toilets, children ever in school, children with own textbooks, than treatment communities. Treatment communities had slightly more acres farmed, children than missed school in the previous week and participants that could trust other people on the community, than control communities. Attrition per relevant group: total attrition: 13% at endline. Per group attrition NR. Description of subgroups measured and reported: women vs men (or girls vs boys). Poorest half vs the less poor half of HHs (on an asset index constructed using principal components analysis). HHs in Kibaha vs Bagamoyo vs Chamwino districts. Age groups: all ages, age 0–1 year; 0–2 years; 0–4 years; 0–18 years; 7–14 years; 15–18 years; ≥ 60 years. Total number completed and analysed per relevant group: 13% (n = 325) of the 2500 recruited HHs were LTFU at endline; therefore, 2175 were analysed. Numbers per group NT. Total number enrolled per relevant group: 1764 HHs and 6918 individual beneficiaries in total at baseline. Numbers per group NR. Total number randomised per relevant group: 40 villages in treatment group and 40 villages in control group.
Interventions	Intervention characteristics Intervention or exposure Food access intervention category: increasing buying power Intervention type: CCT Description: payments to beneficiary HHs are made bimonthly (every 2 months), USD 12–36 maximum depending on number of people in HH. USD 3 per month for orphans and vulnerable children aged ≤ 15 years (about 50% of food poverty line). Initially TZS 3600, but later revised to TZS 5100 to account for inflation. USD 6 per month for elderly people aged ≥ 60 years (100% of food poverty line). Initially this was TZS 7200 but was later revised to TZS 10,500. No HH received < USD 6 per month, and no HH received > USD 18 per month. Funds routed to communities through the local government authorities. Payments disbursed by TASAF to a bank account managed by the LGA, which disbursed the funds directly to the community‐managed accounts. If the local government was not qualified to receive capital development grants, TASAF disbursed the funds directly to the community‐managed accounts. CMCs were then responsible for making payments to beneficiary HHs. Education conditions: enrolment in primary school and individual attendance for children aged 7–15 years. Health conditions: visit to health facility for growth monitoring 6 times a year for children aged 0–5 years; vaccination and growth monitoring for children aged 0–2 years; yearly visit to health facility for routine check and orientation for elderly people (aged ≥ 60 years). A module on community score cards was used as part of the intervention itself to enhance the accountability and process monitoring of the CCT roll out. Duration of intervention period: 31–34 months: January 2010 to August–October 2012 Frequency: cash transfers every 2 months Number of study contacts: 3: baseline: January–May 2009; Midline: July–September 2011; endline: August–October 2012 Providers: Tanzania Social Action Fund, World Bank, community management communities, village assembly, village council, local government authorities Delivery: funds routed to communities through the local government authorities. In districts where the local government was certified compliant via Tanzania's Local Government Development Capital Grant programme, TASAF disbursed 5 payments to a bank account managed by the LGA, which disbursed the funds directly to the community‐managed accounts. If the local government was not qualified to receive capital development grants, TASAF disbursed the funds directly to the community‐managed accounts. The CMCs were then responsible for making payments to beneficiary HHs. Monitoring of conditions began after the first payment was disbursed to beneficiaries in January 2010, and then was done every 4 months. The monitoring process was conducted by TASAF and the CMCs, with support from schools, health centres and district staff. If beneficiaries failed to comply with the conditions, a warning was issued to them by the CMCs. If, after the next monitoring period (8 months after the first payment), beneficiaries still failed to comply with conditions, payments were reduced by 25% and a second warning was sent. After 2 warnings, beneficiaries who failed to comply were suspended indefinitely, but allowed to return to the programme after review and approval by the communities and TASAF. CMCs were responsible for monitoring and also visited beneficiary HHs regularly to keep abreast of any developments. HHs could also leave or be asked to leave the programme for the following reasons: if they chose to opt out, and had informed the CMC, if the HH no longer had an elderly person or a child age < 15 years who was in primary school, if HH members failed to comply with conditions after a warning had been issued 3 consecutive times for children and 2 consecutive times for elderly people. Co‐interventions: transfers from government/TASAF or from NGOs/religious organisation Resource requirements: 'staff' involved in delivering intervention: CMCs village council, village assembly. No other resources reported. Economic indicators: payments: per child USD 6; per elderly person USD 12; maximum payment USD 36; mean payment USD 1450 (Figure ES.2) Control: no intervention
Outcomes	Value of flour/rice purchased Anthropometry: weight; height; MUAC; HAZ; WAZ; WHZ; BMIZ Morbidity: proportion reported being ill in the past 4 weeks; number of days too ill for normal activities in the past 4 weeks
Identification	Sponsorship source: Japan Social Development Fund (JSDF); Trust Fund for Environmentally and Socially Sustainable Development (TFESSD); Spanish Impact Evaluation Fund (SIEF), International Initiative for Impact Evaluation (3ie), and the Consultative Group on International Agricultural Research (CGIAR) Research Program on PIM. Country: Tanzania Setting: communities in 3 poorest and most vulnerable districts (Bagamoyo, Chamwino and Kibaha) Author's name: David K Evans Email: devans2@worldbank.org; pubrights@worldbank.org Declarations of interest: NR Study or programme name and acronym: Community‐Based Conditional Cash Transfers in Tanzania Type of record: report
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Unclear risk	Authors mentioned that villages were randomly selected for intervention and control groups but did not describe any method of random sequence generation.
Allocation concealment (Selection bias)	High risk	No allocation concealment and HH selection was done after villages had been allocated to each intervention group. Unclear how this was done and whether knowledge of the group to which the village had been allocated influenced the process.
Baseline characteristics similar (Selection bias)	Low risk	Baseline differences between groups were reported and adjusted for in the difference‐in‐difference analysis.
Baseline outcome measurements similar (Selection bias)	Unclear risk	NR
Blinding of participants and personnel (Performance bias)	Low risk	No blinding but this was unlikely to affect participant and personnel behaviour.
Blinding of outcome assessment (Detection bias)	High risk	No blinding and some outcomes were self‐reported or subjective outcomes that could have been influenced by knowledge of treatment allocation.
Protection against contamination (Performance bias)	Low risk	Allocation to intervention group by village so there was no risk of contamination.
Incomplete outcome data (Attrition bias)	Low risk	Comment: overall, there were no data for 13% of HHs at baseline. Samples varied for different outcomes reported and it seemed that data were excluded from analysis. However, authors indicated that (quote) "Overall, these balanced rates of attrition across treatment and comparison suggest that the impact evaluation results are unlikely to be affected by attrition."
Selective outcome reporting (Reporting bias)	Unclear risk	No protocol available for this study/report.
Other bias	Unclear risk	Misclassification bias: unlikely. Measurement bias: unclear. Validated tool NR for measuring food consumption and it was only measured 3 times in an almost 3‐year period, which may be insufficient. Incorrect analysis: high. Authors adjusted for intracluster correlation.