Beegle 2017.

Study characteristics
Methods	Study design: cRCT Study grouping: parallel group How were missing data handled? NR Randomisation ratio: 1:1:1:1:1 (according to table 1) Recruitment method: village headmen together with the village committees select HHs to participate. Specific recruitment methods NR Sample size justification and outcome used: NR Sampling method: sampling for the trial: of 28 districts with the PWP programme, 12 districts randomly sampled, stratified by geographic region. Of these 12 districts, 182 villages (EAs) that had been sampled for the national survey in 2010–2011 AND preselected for PWP were selected. These were then randomised to the treatment groups. Villages in the sampling frame were randomly assigned to 1 of 5 groups. Group 0 was control group of villages that were not included in PWP programme in 2012–2013 Season. Groups 1–4 participated in the PWP in the planting season (cycle 1 of PWP). These 4 groups varied in terms of timing of the second cycle of programme and the schedule of payments in both cycles. At HH level: they choose 10 HHs from the 16 survey HHs in the village to be offered the programme. Sampling for MSFA programme covered all districts of Malawi through a 2‐stage targeting approach. In the first stage, there was pro‐poor geographic targeting and in the second there was a combination of community‐based targeting and self‐selection of beneficiaries. The amount of funds given to a district was proportional to the district's population and to the poverty rates as well as other measures of vulnerability. District officials then targeted a subset of EPAs based on poverty and vulnerability criteria. Traditional authorities in the EPAs then allocated funds to a subset of selected GVH who each oversaw 3–10 villages. The GVH determined how many HHs participated in each village based on available funding; the GVH then worked with the village committees in each village to select participating HHs. Study aim or objective: to determine the impact of these programmes by estimating the effect of Malawi's large‐scale PWP, which operates under the MSAF to improve food security and increase the use of fertiliser and other agricultural inputs. The MASAF PWP has been operational since mid‐1990s and aimed to provide short‐term labour‐intensive activities to poor, able‐bodied HHs for the purpose of enhancing their food security, mainly through increased access to farm inputs during the planting period. Programme was designed to be interlinked with Malawi's large‐scale FISP through the implementation of the PWP in the planting months of the main agricultural season when the FISP distribution also occurs. The premise behind this is that the PWP facilitates poor, credit‐constrained HHs to access subsidised fertiliser. Study period: baseline: 2010–2011 (based on data collected during the national integrated HH survey. Endline: November 2013. Data come from 5 rounds of panel HH survey data. Unit of allocation or exposure: villages allocated to control or intervention. Within villages, HHs were randomly selected.
Participants	Baseline characteristics Intervention or exposure: NR Control: NR Overall: NR Inclusion criteria: project: poor, able‐bodies HHs. Exclusion criteria: NR Pretreatment: 2 analysis carried out: for 23 villages included in sample but that had not been included in the national HH survey; and for 159 villages included that had been part of the national HH survey. For the former, analyses are based on round 1 of data collection, when interventions had not started in all but 3 villages but knowledge of PWP implementation existed and could have influenced behaviour. For the latter, they explored the balance between treatment and control villages in terms of pretreatment covariates and outcomes, they used the IHS3 data from 2010–2011. Using the first round of follow‐up data, they found that HHs in the non‐IHS3 sample were better off than the IHS3 sample, with better educated HH heads, smaller HH sizes and fewer children aged < 14. However, there was imbalance in preprogramme food security at both the village and HH levels in the 159 villages for which IHS3 data were available. The IHS sample was well balanced for a range of non food‐security outcomes. Attrition per relevant group: NR Description of subgroups measured and reported: effects by geographic region Total number completed and analysed per relevant group: NR: only total number of observations reported. Total number enrolled per relevant group: 10 HHs in each village offered the programme Total number randomised per relevant group: group 0 (control): 38 communities; group 1 (cycle 1: planting season, cycle 2: harvest season; lump sum): 40 communities; group 2 (cycle 1: planting season, cycle 2: harvest season; split payment): 34 communities; group 3 (cycle 1: planting season, cycle 2: lean season; lump sum): 35 communities; group 4 (cycle 1: planting season, cycle 2: lean season; split payment): 35 communities.
Interventions	Intervention characteristics Intervention or exposure Food access intervention category: income generation Intervention type: PWP Description: MASAF PWP has been operational since mid‐1990s and aims to provide short‐term labour‐intensive activities to poor, able‐bodied HHs for the purpose of enhancing their food security, mainly through increased access to farm inputs during planting period. Programme was designed to be interlinked with Malawi's FISP through implementation of PWP in planting months of the main agricultural season when the FISP distribution also occurs. The premise was that the PWP facilitates poor, credit‐constrained HHs to access subsidised fertiliser. Projects were mostly road rehabilitation or construction, with some forestation and irrigation projects. Wage rate was MWK 300/day (USD 0.92/day) for a total payment of MWK 3600 for a 12‐day wave. Cycle 1 of PWP was implemented during planting season (October–December) to align with the timing of the distribution of FISP. Cycle 2 of PWP was designed to take place after harvest in June and July. Duration of intervention period: November–December 2012 to November 2013 (1 year) Frequency: group 1: 4 payments in total (2 in cycle 1, 2 in cycle 2). Group 3: similar to group 1 only timing was different. Group 2 and 4: each payment in cycle 2 was split into 5 payments (so 12 payments in total). In 2012, as a response to a large currency devaluation, the programme was doubled in size and scaled up to cover about 500,000 HHs per year. Duration of project participation increased from 12 days to 48 days, split into 2 cycles of 24 days each; the cycles were further divided into 2 consecutive 12‐day waves, and payments were generally made within 1 or 2 weeks of the end of each wave." Number of study contacts: data came from 5 rounds of panel HH survey data. Basis for panel was the IHS3 fielded in 2010–2011 by Malawi's National Statistics Office. The 16 IHS3 HHs were interviewed in 4 additional rounds: before the public works projects started during planting season (November 2012) after the first cycle, preharvest (February 2013), after the lean season cycle, postharvest (April–May 2013) and finally after the completion of the 2012–2013 season (November 2013). Providers: MASAF PWP is a government programme but the study was implemented by research team. Payments in the study districts were facilitated by the research team for the purposes of the evaluation. This was intended to ensure that payments were made without delay, on specific schedules. Administrative payment records confirmed that there were no differences in time lag between work and payment across the districts. Delivery: groups 1–4 participated in the PWP in the planting season (cycle 1 of PWP). These 4 groups varied in terms of timing of the second cycle of the programme and the schedule of payments in both cycles. However, the analysis grouped all these intervention groups into 1 to assess overall effect of having a PWP programme in place. Payments in the study districts were facilitated by the research team for the evaluation. This was intended to ensure that payments were made without delay, on specific schedules. Administrative payment records confirmed that there were no differences in time lag between work and payment across the districts. In addition to the HH survey data, in terms of monitoring the intervention, administrative records included the dates and amounts of payments and the identities of recipients. These were used to confirm that beneficiaries received payments in accordance with the days they worked. Payments in the study districts were facilitated by the research team for the purposes of the evaluation, with physical delivery of the cash in conjunction with the district officials. The split‐payment variant slightly increased the cost of implementation. Epayments, which would entail a small marginal cost of delivery, were under consideration for future rounds of PWP. Co‐interventions: in 3 study districts, fertiliser subsidy coupon distribution took place between the first and second 12‐day waves of PWP activities, and, in the remaining 9 districts, fertiliser coupon distribution overlapped with PWP work and payment. The national fertiliser subsidy programme provided about half of HHs in the country with coupons that allowed 2 bags of fertiliser to be purchased for MWK 500 each. However, fertiliser coupons were more likely to be available to treated HHs in accordance with the designed linkage between PWP and the national fertiliser subsidy scheme. Resource requirements: NR Economic indicators: wage rate was MWK 300/day (USD 0.92/day) for a total payment of MWK 3600 for a 12‐day wave. The split‐payment variant slightly increased the cost of implementation. Control: no intervention
Outcomes	Food security: Food Security Score; Resilience Index; Principal Components Analysis Index Dietary diversity: FCS; number of food groups consumed in the last week for 7 main groups Dietary intake: log per capita food consumption for the last week; per AE calories of the food consumed
Identification	Sponsorship source: World Bank Research Committee, Knowledge for Change programme and GLM‐LIC (grant number C2‐RA2‐211). Country: Malawi Setting: poor and able‐bodied HHs in 3 regions of Malawi Author's name: Kathleen Beegle Email: kbeegle@worldbank.org (K Beegle); egalasso@worldbank.org (E Galasso); goldberg@econ.umd.edu (J Goldberg). Declarations of interest: NR Study or programme name and acronym: MASAF PWP (Malawi Social Action Fund's Public Works Programme) Type of record: journal article
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Low risk	Quote: "Villages were randomly assigned (by computer) to one of the four treatment groups or a control condition; households within treatment villages were randomly selected to be offered the program."
Allocation concealment (Selection bias)	Low risk	Allocation to treatment group was done at village level. Quote: "Villages were randomly assigned (by computer) to one of the four treatment groups or a control condition; households within treatment villages were randomly selected to be offered the program."
Baseline characteristics similar (Selection bias)	High risk	Allocation to treatment group was done at village level. Quote: "Villages were randomly assigned (by computer) to one of the four treatment groups or a control condition; households within treatment villages were randomly selected to be offered the program."
Baseline outcome measurements similar (Selection bias)	High risk	Even though randomisation was conducted by computer, there was imbalance in preprogramme food security at both the village and HH levels in the 159 villages for which IHS3 data were available. Unclear how well these were adjusted for in the analyses.
Blinding of participants and personnel (Performance bias)	Low risk	Blinding was not possible for this type of intervention; however, it was unlikely that lack of blinding would have influenced the delivery of the intended intervention.
Blinding of outcome assessment (Detection bias)	High risk	Outcome assessment was not blinded. Data were based on HH surveys, therefore, self‐reported data. Knowledge of intervention allocation could have biased responses.
Protection against contamination (Performance bias)	Low risk	Authors only reported contamination in relation to untreated HHs in villages selected to receive the PWP intervention. Allocation was by village and it was unlikely that the control group received the intervention.
Incomplete outcome data (Attrition bias)	Unclear risk	Number of villages were reported but not number of HHs and individuals assessed at the start and endline. It was unclear if there was any attrition or not.
Selective outcome reporting (Reporting bias)	Unclear risk	2 outcome measures were omitted due to space constraints. A composite measure was compiled and included.
Other bias	Unclear risk	Misclassification bias of exposure: low risk. Treated and control assigned externally. Measurement bias: unclear risk. Authors did not report which tools were used to collect data, who did it and whether they were validated. They only reported and defined the food security outcomes assessed. Incorrect analysis: low risk. Analyses adjusted for clustering. Recruitment bias: low risk. HHs in allocated villages were randomly selected to participate.