Haushofer 2013.

Study characteristics
Methods	Study design: RCT Study grouping: parallel How were missing data handled? missing data due to attrition excluded from analysis. Analyses were based on total 1372 HHs, which is the sum of HHs at baseline only. Randomisation ratio: 1:1 (503:505) Recruitment method: after HHs and members identified, in private conversation, members were asked questions about demographics, and informed that they had been chosen to receive a cash transfer of KES 25,200 (USD 404). The recipient was informed that this transfer was unconditional, that they were free to spend it however they chose, and that it was a one‐time transfer and would not be repeated. Sample size justification and outcome used: sample size 500 individuals in each of the treatment, control and pure control group was chosen based on a power calculation, which showed that a sample of 1000 individuals was sufficient to detect effect sizes of 0.2 SD for all treatment vs pure control HHs with 89% power. Different treatment groups within the treatment groups (male vs female recipient, lump‐sum vs monthly, large vs small transfers) could be compared with 60% power (from registry record). Sampling method: purposive sampling of villages and HHs followed by random selection of HHs into treatment or control groups. GiveDirectly selected poor HHs by identifying poor regions of Kenya according to census data. Region chosen was Rarieda, a peninsula in Lake Victoria west of Kisumu in Western Kenya. GiveDirectly identified target villages through a rough estimation of the population of villages and the proportion of HHs lacking a metal roof, which is GiveDirectly's targeting criterion. Identified 126 villages. 63 of these villages were randomly chosen to be treatment villages. Control villages were only surveyed at endline; in these villages, authors sampled 432 HHs referred to as 'pure control' HHs. In treatment villages, second stage of randomisation assigned 50% of HHs to treatment condition, and 50% to control condition. Process resulted in 503 treatment HHs and 505 control HHs in treatment villages at baseline. Note: numbers were different between reports. Study aim or objective: to assess the relative welfare impacts of 3 design features of UCTs: gender of transfer recipient; temporal structure of transfers (monthly vs lump‐sum); and magnitude of transfer. Study period: transfers between June 2011 and January 2013 Unit of allocation or exposure: HHs
Participants	Baseline characteristics Intervention or exposure Age: treatment effect vs control, coefficient: –1.15 (SE 0.86). Number of children vs control, coefficient: 0.04 (SE 1.12) Place of residence: NR Sex: NR Ethnicity and language: NR Occupation: wage labour primary income vs control, coefficient: 0.02 (SE 0.03). Own farm primary income vs control, coefficient: –0.02 (SE 0.03) Education: years of education completed (of respondent) vs control, coefficient: 0.27 (SE 0.18) SES: HH size vs control, coefficient: 0.02 (SE 0.13). Value of non‐land assets (USD) vs control, coefficient: –1.15 (SE 24.74) Social capital: NR Nutritional status: FSI vs control, coefficient: 0.00 (SE 0.06) Morbidities: Health Index vs control, coefficient: 0.03 (SE 0.06). Psychological well‐being Index vs control, coefficient: 0.03 (SE 0.05) Concomitant or previous care: NR Control Age: respondent, mean, years: 35.35 (SD 14.13). Number of children, mean: 2.88 (SD 1.91) Place of residence: NR Sex: NR Ethnicity and language: NR Occupation: wage labour primary income, mean: 0.25 (SD 0.43). Own farm primary income, mean: 0.37 (SD 0.48) Education: years of education completed (of respondent), mean: 8.56 (SD 2.95) SES: HH size, mean: 4.94 (SD 2.16). Value of non‐land assets (USD), mean: 383.36 (SD 374.15) Social capital: NR Nutritional status: FSI, mean: 0.00 (SD 1.00) Morbidities: Health Index, mean: 0.01 (SD 1.02). Psychological well‐being Index, mean: 0.00 (SD 1.00) Concomitant or previous care: NR Overall: NR Inclusion criteria: HHs lacking metal roof (indicator of poverty) Exclusion criteria: none reported Pretreatment: results were largely insignificant, suggesting that the treatment and control groups did not differ at baseline. Attrition per relevant group: overall: 68 (6.7%) (940/1008 surveyed at endline). Treatment: 32 (6.4%); LTFU (471/503 surveyed at endline); control: 36 (7.1%); LTFU (469/505 surveyed at endline) Description of subgroups measured and reported: male vs female recipients of transfers; monthly vs lump‐sum transfers; large vs small transfers Total number completed and analysed per relevant group: treatment: 471 HHs; control: 469 HHs Total number enrolled per relevant group: treatment: 503 HHs; control: 505 HHs Total number randomised per relevant group: treatment: 503 HHs; control: 505 HHs
Interventions	Intervention characteristics Intervention or exposure Food access intervention category: increase buying power Intervention type: UCT Description: monthly transfers: first instalment transferred on first of month following initial visit, and continued for 8 months thereafter. Lump‐sum transfers: a month was randomly chosen among the 9 months following the date of the initial visit. For receipt of transfer, recipients were provided with a SIM card by Kenya's largest mobile service provider, Safaricom, and asked to activate it and register for Safaricom's mobile money service M‐Pesa. HHs with both a primary female and primary male member stratified on recipient gender and randomly assigned the woman or the man to be the transfer recipient in an equal number of HHs. 258/503 treatment HHs were assigned to monthly group, and 245 to the lump‐sum group. Total amount of each type of transfer was KES 25,200 (USD 404). Amount included an initial transfer of KES 1200 (USD 19) to incentivise M‐Pesa registration, followed by either a lump‐sum payment of KES 24,000 (USD 384) lump‐sum group, or 9 monthly transfers of KES 2800 (USD 45) each in the monthly group. 137 HHs in the treatment group were randomly chosen and informed in January 2012 that they would receive an additional transfer of KES 70,000 (USD 1112), paid in 7 monthly instalments of KES 10,000 (USD 160), beginning in February 2012. Thus, the transfers previously assigned to these HHs, whether monthly or lump‐sum, were augmented by KES 10,000 from February 2012 to August 2012, and, therefore, the total transfer amount received by these HHs was KES 95,200 (USD 1525). The remaining 366 treatment HHs constituted the 'small' transfer group, and received transfers totalling KES 25,200 (USD 404) per HH. Duration of intervention period: 20 months. Transfers were made between June 2011 and January 2013. Frequency: lump sum or monthly transfers. 258/503 treatment HHs assigned monthly group, and 245 to lump‐sum group. Total amount of each type of transfer was KES 25,200 (USD 404), which included an initial transfer of KES 1200 (USD 19) to incentivise M‐Pesa registration, followed by either a lump‐sum payment of KES 24,000 (USD 384) in the lump‐sum group, or 9 monthly transfers of KES 2800 (USD 45) each in the monthly condition. In both the monthly and the lump‐sum groups, recipients received the initial transfer of KES 1200 immediately following the announcement visit by GiveDirectly. In the monthly group, recipients received the first transfer of KES 2800 on the first of the month following M‐Pesa registration, and the remaining 8 transfers of KES 2800 on the first of the 8 following months. In the lump‐sum group, recipients received the lump‐sum transfer of KES 24,000 on the first of a month chosen randomly among the 9 months following enrolment. Number of study contacts: 2 (baseline and endline) Providers: GiveDirectly NGO Delivery: for lump‐sum group, a small initial transfer of KES 1200 was sent on the first of the month following the initial GiveDirectly visit as an incentive for prompt registration. Registration had to occur in the name of the designated transfer recipient, rather than any other person. To facilitate easier communication with recipients and reliable transfer delivery, GiveDirectly offered to sell mobile phones to recipient HHs that did not own 1 (by reducing the future transfer by the cost of the phone). In a few additional cases, delays in registration occurred due to delays in obtaining an official identification card, which was a prerequisite for registering with M‐Pesa. Withdrawals and deposits could be made at any M‐Pesa agent (about 11,000 throughout Kenya). GiveDirectly reported that recipients typically withdrew the entire balance of the transfer upon receipt. Due primarily to registration issues with M‐Pesa, 18 treatment HHs had not received transfers at endline, thus, only 485 of the treatment HHs had received transfers. Co‐interventions: NR Resource requirements: GiveDirectly estimated the mean travel time from recipient HHs to the nearest M‐Pesa agent was 42 minutes. Economic indicators: GiveDirectly estimated the mean cost from recipient HHs to the nearest M‐Pesa agent at USD 0.64. Withdrawals incur costs between 27% for USD 2 withdrawals and 0.06% for USD 800 withdrawals, with a gradual decrease of the percentage for intermediate amounts. The sender also incurred costs for M‐Pesa transfers; according to GiveDirectly's estimates, the costs of transferring money to recipients in this was amount to 1.5% of the transfer amount for foreign exchange fees, and 1.6% for M‐Pesa fees. Together with 4.8% of transfers spent on recipient identification and staff costs, GiveDirectly estimated that 92.1% of the donations it received were transferred to recipients' M‐Pesa accounts. Control: no intervention
Outcomes	Proportion of HH expenditure on food: Total monthly HH food expenditure (cereals, tubers, meat/fish, dairy, fruit/vegetables, other food, food eaten out, Food security: FSI (based on weighted mean of measures of food security and hunger based on 17 outcome measures) Anthropometry: MUAC; height; weight Anxiety and depression: psychological well‐being index (standardised weighted mean of 6 psychological and neurobiological measures); log cortisol; CES‐D; Cohen PSS
Identification	Sponsorship source: NIH Grant R01AG039297 and Cogito Foundation Grant R‐116/10 to Johannes Haushofer. Country: Kenya Setting: poor rural villages in Kenya Comments: additional documentation: online Appendix. RCT ID (trial registry): AEARCTR‐0000019 (www.socialscienceregistry.org/trials/19) Author's name: Johannes Haushofer Email: joha@mit.edu Declarations of interest: NR either authors. Quote: "Shapiro is a co‐founder and former director of GiveDirectly, Inc. (2009–2012). This paper does not necessarily represent the views of GiveDirectly, Inc" Study or programme name and acronym: N/A Type of record: report
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Unclear risk	Randomisation method NR.
Allocation concealment (Selection bias)	Unclear risk	Concealment of allocation was not described, and HHs in the same cluster (village) were randomised to either receive the intervention or not.
Baseline characteristics similar (Selection bias)	Low risk	Quote: "The only significant difference between treatment and control households appears in income from self‐employment, where treatment households have a $33 PPP [purchasing power parity] lower income relative to the control mean of $85 PPP (39%) at baseline. This difference is significant at the 10% level, but does not survive FWER [familywise error rate] correction for multiple inference"
Baseline outcome measurements similar (Selection bias)	Low risk	Quote: "Online Appendix Table 35 shows only minor differences in the estimates of the treatment effects when baseline controls are included; none of the significant results become non significant or vice versa. Thus, baseline covariates do not affect our results strongly." "The only significant difference between treatment and control households appears in income from self‐employment, where treatment household shave a $33 PPP lower income relative to the control mean of $85 PPP (39%) at baseline. This difference is significant at the 10% level, but does not survive FWER correction for multiple inference."
Blinding of participants and personnel (Performance bias)	Low risk	No blinding. Unlikely to influence behaviour or experience of participants.
Blinding of outcome assessment (Detection bias)	High risk	No blinding, which may have affected self‐reported outcomes of participants who did not receive the cash transfers.
Protection against contamination (Performance bias)	Low risk	Quote: "First, we find no spillovers in consumption. This is surprising, given that we might have expected some informal insurance among households: in effect, the transfer is a temporary lottery gain, and theory predicts that households should have been sharing it with their insurance network."
Incomplete outcome data (Attrition bias)	Low risk	Few and balanced missing data – approximately 6% in each group. This was unlikely to introduce bias.
Selective outcome reporting (Reporting bias)	Low risk	No study protocol available but all outcomes outlined in registry were reported.
Other bias	Unclear risk	Misclassification bias: low. Measurement bias: unclear. Potential bias as information on dietary intake only captured at baseline and after 1 year. Incorrect analysis: unclear. Although there was a cluster randomisation, the analysis used for this review were at HH level and not cluster level. Comparison of intervention and control HHs in villages allocated to intervention group.