Skoufias 2013.

Study characteristics
Methods	Study design: cRCT Study grouping: parallel How were missing data handled? Missing data were excluded from analysis. Sample of HHs used for the analysis of consumption was what remained after dropping HHs with food consumption < 1 percentile and > 99 percentile of the food distribution in the sample. An additional 802 HHs were excluded from the analysis because of missing or incomplete data (Leroy 2010). Analyses provide an estimate of ITT effect of in‐kind and cash transfers. Randomisation ratio: 1:1:1:1 Recruitment method: NR Sample size justification and outcome used: sample size was calculated so that statistical tests had the power to detect statistically significant and biologically relevant differences in several nutritional and economic variables. Specifically, the calculations of the sample size prior to the baseline survey were based on 60 communities per treatment group, a power of 80%, and a minimum detectable difference in food per capita consumption between each treatment and control group of 17.8%. The final sample consisted of 33 HHs per community and around 52 communities per treatment group (ICC 0.220) (Skoufias 2013). Sampling method: 2‐stage random sampling. A random sample of 208 rural communities was drawn from a pool of communities within 8 of the poorest states in the Southeast region. Within each community, 33 HHs were selected using systematic random sampling. After baseline data collection, the 208 selected communities (6687 HHs) were randomly assigned to 1 of 4 study groups: food basket without education (52 communities, 1657 HHs), food basket with education (52 communities, 1680 HHs), cash transfer with education (53 communities, 1687 communities) or control (51 communities, 1663 HHs). Due to partial contamination of the original evaluation design the analysis pools both in‐kind/food basket groups. Study aim or objective: to examine the impacts of cash and in‐kind transfers on HH welfare as measured by food and total consumption, poverty and labour supply (Skoufias 2008; 2013). To estimate the programme's impact on HH energy and macro‐ and micronutrient consumption and to evaluate whether the cash and in‐kind transfers had a differential effect on these outcomes (LeRoy 2010). Study period: delivery of the PAL benefits began in June 2004 and the mean time of exposure to the availability of the programme transfers was 14 months. Unit of allocation or exposure: communities
Participants	Baseline characteristics Intervention (cash transfer) Age: NR Place of residence: NR Sex: NR Ethnicity and language: NR Occupation: NR Education: NR SES: mean total HH expenditure (per AE per month), n = 1492: MXN 551.6 (SD 336.5) Social capital: NR Nutritional status: mean dietary intake (per AE per day), n = 1492: total energy (kcal) 2776.8 (SD 1172.2); energy from fruits and vegetables (kcal): 79.7 (SD 70.5); energy from cereals and legumes (kcal): 1723.7 (SD 916.5); energy from animal source food (kcal): 261.7 (SD 214.0); energy from processed food (kcal): 664 (SD 344.1) Morbidities: NR Concomitant or previous care: NR Control Age: NR Place of residence: NR Sex: NR Ethnicity and language: NR Occupation: NR Education: NR SES: mean total HH expenditure (per AE per month), n = 1384: MXN 571.3 (SD 352.3) Social capital: NR Nutritional status: mean dietary intake (per AE per day), n = 1384: total energy (kcal): 2702 (SD 1140.3); energy from fruits and vegetables (kcal): 81.8 (SD 73.6); energy from cereals and legumes (kcal): 1625.3 (SD 838.9); energy from animal source food (kcal): 278.1 (SD 229.6) Morbidities: NR Concomitant or previous care: NR Overall Age: NR Place of residence: NR Sex: NR Ethnicity and language: NR Occupation: NR Education: at baseline, about 40% of the HHs heads had completed primary school. SES: families lived in homes with a mean of < 3 rooms. The mean number of AE per HH was about 4. Social capital: NR Nutritional status: NR Morbidities: NR Concomitant or previous care: NR Inclusion criteria: HHs eligible for PAL (localities had to meet some requirements such as having a population of < 2500, having a high or very high marginality and being accessible (not > 2.5 km from a road), and close enough (not > 2.5 km) to a DICONSA store (Mexican Government's agency that distributes the supply of food). Exclusion criteria: NR Pretreatment: the occasional significance of some variables indicated that the random assignment did not manage to balance totally the sample across the treatment and control groups, especially with respect to HH consumption (Table A.1). However, DID analysis accounted for this imbalance. In Leroy 2010, baseline nutrient intakes in cash transfer and control groups were similar. Attrition per relevant group: intervention group (cash transfers with education): 195/1687 (11.6%); control group: 279/1663 (16.8%). Total attrition: 864/6687 (13%) HHs. Reasons for attrition not provided, except for 1 cluster (33 HHs) in the control group and another in food basket without education group, that refused to participate in the follow‐up survey. Description of subgroups measured and reported: NR Total number completed and analysed per relevant group: full data were thus available for 5823 HHs (food basket without education: 51 communities, 1447 HHs; food basket with education: 52 communities, 1500 HHs; cash transfer with education: 53 communities, 1492 HHs; control: 50 communities, 1384 HHs) (Leroy 2010). Total number enrolled per relevant group: the data use were based on a longitudinal sample of 5851 HHs in 206 poor rural localities from 6 southern Mexican states (Chiapas, Guerrero, Oaxaca, Quintana Roo, Tabasco and Veracruz). Numbers per group NR. Total number randomised per relevant group: after baseline data collection, the 208 selected communities (6687 HHs) were randomly assigned to 1 of 4 study groups: food basket without education (52 communities, 1657 HHs), food basket with education (52 communities, 1680 HHs), cash transfer with education (53 communities, 1687 HHs) or control (51 communities, 1663 HHs) (Leroy 2010).
Interventions	Intervention characteristics Intervention (cash transfer) Food access intervention category: increase buying power Intervention type: UCT Description: in‐kind and cash transfers provided for poor rural HHs. The PAL programme offers nutrition and health education sessions (platicas), as well as participation in programme‐related logistic activities. However, attendance was not compulsory. HHs received either a cash transfer of MXN 150/month (equivalent to USD 14 at the time) or a monthly food basket with a cost to the programme of MXN 150. The size of the cash transfer and the amount of food was the same for all HHs regardless of family size or composition. The value of the food basket at local prices was 30% higher for consumers than the actual cost to the programme (Leroy 2010). Duration of intervention period: delivery of the PAL benefits began in June 2004 and the mean time of exposure to the availability of the programme transfers was 14 months. Frequency: every 2 months Number of study contacts: the control and the treatment groups were surveyed on 2 occasions 2 years apart: at baseline in October 2003 to April 2004, and at follow‐up in October–December 2005. Providers: Mexican government, specifically: benefits were distributed through DICONSA, the related federal programme which distributes non‐perishable foods and housekeeping goods throughout rural poor communities. Delivery: distribution of transfers was through DICONSA stores (Mexican government's agency that manages the supply of food). 75% of beneficiaries were women. The value of cash transfer in real terms was smaller than the in‐kind transfers. The cash transfers provided by the programme were implemented for those very isolated communities where DICONSA did not regularly reach. Co‐interventions: none reported Resource requirements: NR Economic indicators: cash transfers: USD 14/month. Other: NR Control: no intervention
Outcomes	Dietary diversity: proportion of children with MDD (consumed foods from ≥ 3–6 food groups) Diet intake: consumption of iron‐rich or iron‐fortified foods Anthropometry: BMI
Identification	Sponsorship source: Mexican government Country: Mexico Setting: poor rural HHs in Southern states of Mexico Author's name: Emmanuel Skoufias Email: eskoufias@worldbank.org Declarations of interest: no conflicts of interest (Leroy 2010). Study or programme name and acronym: Programa de Apoyo Alimentario (PAL) (food support programme) Type of record: journal article
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Low risk	Quote: "Localities were randomised into four groups using a simple randomisation algorithm" (Cunha 2014).
Allocation concealment (Selection bias)	Low risk	Allocation was at location level (clusters), carried out by the Ministry of Social Development after baseline data collection.
Baseline characteristics similar (Selection bias)	Low risk	Although there were some differences between characteristics at baseline, the DID analysis adjusted for these.
Baseline outcome measurements similar (Selection bias)	Low risk	Baseline nutrient intakes in both groups were similar (Leroy 2010). Although HH food expenditure was less in the intervention group at baseline (P < 0.1) (Addendum A, Skoufias 2013), the data were analysed with DID methods which adjusts for pre‐existing baseline differences.
Blinding of participants and personnel (Performance bias)	Low risk	Blinding was not possible due to the nature of the intervention. This was unlikely to influence behaviour of participants and personnel.
Blinding of outcome assessment (Detection bias)	Low risk	Quote: "To avoid potential interviewer bias, field workers were, to the extent possible, unaware of the group assignment."
Protection against contamination (Performance bias)	Unclear risk	Allocation was done at the community level; however, there was no information about whether communities in control group received either intervention. Cunha 2014 reported that 1 control HH reported receiving aid.
Incomplete outcome data (Attrition bias)	High risk	11.6% attrition in the intervention group and 16.8% in the control group. 1 cluster lost as 1 community in the control group refused to participate in the follow‐up survey (33 HHs). The study authors reported that HHs excluded from the analyses tended to live in smaller houses than those included (2.48 vs 2.77 rooms; P < 0.05). Nutrient consumption at baseline was higher in excluded HHs, but no details were reported.
Selective outcome reporting (Reporting bias)	Unclear risk	No study protocol available.
Other bias	Unclear risk	Misclassification bias: unclear risk. Receipt of cash transfers were self‐reported in 1 paper (Cunha 2014), but NR as such in other papers. Measurement bias: low risk. Trained field workers interviewed the homemaker at baseline and follow‐up (semi‐quantitative FFQ questionnaire of 61 food items consumed at home in the 7 days prior). Incorrect analysis: unclear risk. SEs were corrected for clustering of individuals at the village level in Skoufias 2013, but adjusting for clustering NR in Leroy 2010. Recruitment bias: low risk. Participants were recruited after allocation of locations (clusters), but were randomly sampled in each location. Seasonality bias: low risk. The month of data collection was included to adjust for the possible effect of seasonality on consumption.