From protection to production: productive impacts of the Malawi Social Cash Transfer scheme

Sub-Saharan Africa

Malawi

Social Protection

Unconditional Cash Transfers (UCT)

K Covarrubias; B Davis; P Winters

2012

Covarrubias, K., B. Davis, and P. Winters. 2012. “From Protection to Production: Productive Impacts of the Malawi Social Cash Transfer Scheme.” Journal of Development Effectiveness, 4 (1), 50-77.
View publication

-

2012

The Malawi Social Cash Transfer (SCT) Scheme

2006-2007

The SCT program aims to reduce poverty, hunger and starvation, and improve school enrolment and attendance and the health and nutrition of children among the poorest 10 per cent of households in Malawi.

For the study households were deemed eligible if they belonged to the poorest 10 percent of households in Malawi. Within each village, eligibility was supposed to be checked by so-called Community Social Protection Committees against national criteria. However, in practice selection was often based on local criteria leading to variation in the targeting procedures across villages.

The SCT provides unconditional cash transfers ranging in value from US$4 per month for a household with one eligible member to US$13 per month for households with four or more eligible members. In addition, the program offers a school attendance bonus ranging from US$1.30 per month for primary school age children to US$2.60 per month for secondary school age children (average monthly per-capita income in the initial target district was US$7.80). Funded by the Global Fund through the National AIDS Commission and overseen by the Ministry of Gender, Children and Community Development, the program is expected to serve 300,000 households with 910,000 children by 2015.

• Education: school enrollment and days of school missed in previous month
• Children’s work: participation in a range of activities including household chores, childcare, work on family farm or non-farm business, non-household work for income, and domestic work outside the household (both paid and unpaid)
• Children’s work: for each of these activities the authors also report hours worked.

Randomized control trial combined with propensity score matching

Study design: As part of the study a one year pilot was designed in which 8 groups of villages (containing 23 villages in total) were randomized into treatment (4 groups) and control (4 groups).
Estimation procedures: The authors use two different estimation procedures: difference in differences regressions (including a vector of baseline controls) and difference in differences propensity score matching.
Data: Baseline data collection began prior to treatment in March 2007, after which the treatment group received the first transfer. Follow-up data were collected in September 2007 and again in April 2008, at which point the control households also began receiving transfers. In total 819 households were interviewed in 2007. but 33 treatment and seven control households (8.1% and 2.7% of the original treatment and control sample, respectively) were lost over the two follow-up rounds and are not included in this analysis. After excluding the attrition households, linking the three rounds of data yields 365 treatment and 386 control households with complete questionnaires over the 2007–2008 period. These 751 households contain 1876 children below 18 years of age, among which 1090 belonged to treatment households and 786 to control households.

Baseline: At baseline over 60 per cent of children were involved in childcare for household members in the week preceding the survey and close to one-fifth were responsible for caring for elderly household members. Further, more than 20 per cent were also involved in the household’s farm or non-farm business work. Roughly three quarters of children were attending school. Education: The difference in difference regressions report a 4 percentage point increase in school enrollment as a result of the intervention, statistically significant at the 10% level. However, this effect is not robust as the propensity score matching point estimate is negative and not statistically significant. There is evidence of a reduction in the number of school days missed per month by approximately 0.7. This effect is similar and significant both in the difference in difference regressions and the propensity score matching estimates. Children’s work: Involvement in domestic work outside the household, paid and unpaid, falls by 7 per cent as a result of the intervention. The number of hours worked on non-household domestic work falls by approximately a quarter of an hour. These effects are similar and significant both in the difference in difference regressions and the propensity score matching estimates. The SCT, however, does not appear to have lead to increased leisure time for children. Instead, children’s participation in household chores increased by 8 to 14 percentage points. Moreover, the SCT lead to more child participation in family farm/non-farm business activities. On average, children spend an additional 0.16 hours per week on these activities (robust across DD and PSM).

Child Labour

Households

National

The program limits child labour outside the home while increasing child involvement in household farm activities. Involvement in domestic work outside the household, paid and unpaid, falls by 7 per cent as a result of the intervention. The number of hours worked on non-household domestic work falls by approximately a quarter of an hour. The SCT, however, does not appear to have lead to increased leisure time for children. Instead, children’s participation in household chores increased by 8 to 14 percentage points. Moreover, the SCT lead to more child participation in family farm/non-farm business activities. On average, children spend an additional 0.16 hours per week on these activities.

The program generates agricultural asset investments and reduces adult participation in low skilled labour.

The authors examine whether their results are heterogeneous along several dimensions including age. Unfortunately, there appears to be an error in the main table presenting heterogeneous results by age of the child, as the difference in differences regression estimates are the same for children aged 13-17 and 4-12. Moreover, the authors do not indicate whether point estimates found for older and younger children are statistically different from each other. Some important issues are not explained with sufficient detail in the paper. It is not clear whether standard errors are clustered at the level at which the randomization took place (groups of villages). It is not clear how the authors deal with “hours worked”, which is a censored outcome variable. It is not clear whether the authors measure school enrollment or school attendance (the tables say attendance while the text mentions enrollment). It is not clear whether and how the schooling bonus is conditional.