Absence of conflict of interest.
Amarante, V., Ferrando, M., & Vigorito, A. (2013). Teenage school attendance and cash transfers: An impact evaluation of PANES. Economia, 61-93.
- This objective of this study was to examine the impact of a conditional cash transfer program, National Plan for Social Emergency Assistance (PANES), on child labor and school attendance for children 14-17 years old in Uruguay.
- The primary design was a regression discontinuity design (RDD) using households within a 2 percent range of the cutoff score. The authors also used a difference-in-differences (DID) approach. The study used administrative data collected before the launch of the program and household surveys implemented in two follow-up waves (at two months and 18 months after the program ended). The authors compared the differential effects of receiving the program on outcomes for those in the treatment group versus those in the comparison group.
- The study found no statistically significant relationships between the PANES program and child labor or schooling.
- The quality of causal evidence presented in this report under the DID design is low. The RDD does not receive a causal rating.
National Plan for Social Emergency Assistance (PANES)
Features of the Intervention
PANES was a government funded program that started in 2005 and lasted for 34 months. Uruguay is a middle income country with higher school attendance rates and lower child labor rates than other Latin America countries that have used conditional cash transfer programs to increase school enrollment and decrease child labor. The PANES program provided a monthly lump sum cash payment (equivalent to ~$56 USD) to eligible households, regardless of the number of children in the household. Households with pregnant women and children could also receive a food card. Minor program components included a workfare program, job training, adult educational interventions, and healthcare subsidies. By design, the cash transfers were conditional on children being enrolled in school and attending regular health check-ups, but in practice the conditions were not enforced and survey data indicated only 58 percent of the recipients were aware of the conditions. The targeted population was households in the bottom quintile of the national poverty line which represented about 8 percent of the population; but all low income households were invited to apply. Households with income below $50 USD were eligible and ultimately about 14 percent of the population received the program.
Features of the Study
The authors used a regression discontinuity and a difference-in-differences (DID) design. After determining eligibility, 188,671 households were given the baseline survey and assigned a score based on socioeconomic factors. Those with scores higher than the cutoff score (indicating higher predicted poverty) received the program. The study sample was restricted to households whose applications were evaluated between September 2005 and April 2006. The analyses are based on PANES administrative data (baseline) and household survey data (two waves of follow-up). The first follow-up occurred 18 months into the program and the second wave two months after the program ended. At baseline, 79% of the children in beneficiary households attended school and 11% participated in child labor. In non-beneficiary households, 79% of children attended school and 15% participated in child labor. For the main regression discontinuity design, the sample size was 3,000 households within two percent of the cutoff score. The sample included 2,000 eligible households and 1,000 ineligible households. At the initial follow-up survey, nonresponse was 30%, so households were replaced in the sample with households with the same or similar scores. At the second follow-up the response rate was 92%. The secondary approach was a difference in difference design comparing the differential in the treatment and comparison group after the intervention.
- Neither the RDD nor the DID designs found any statistically significant relationship between PANES and child labor in either the first or second wave of follow-up.
- Statistically significant relationships between PANES and child labor were not found either when examining age-divided or gender-divided subgroups.
Education (School participation/enrollment)
- Neither the RDD nor the DID designs found any statistically significant relationship between PANES and school attendance in either the first or second wave of follow-up.
- Statistically significant relationships between PANES and school attendance were not found either when examining age-divided or gender-divided subgroups.
Considerations for Interpreting the Findings
This study used a regression discontinuity design and as such was reviewed against CLEAR’s descriptive study evidence review guidelines and does not receive a causal rating. The study also employed a difference-in-difference approach. For studies using difference-in-differences designs, authors must demonstrate equivalent trends between treatment and control groups. In this case, because there was only one observed pre-treatment period, the authors must use a placebo method or adequately control for time-varying characteristics that might influence the outcomes of interest. In this study, authors did not include a control for time trends.
Causal Evidence Rating
The quality of causal evidence presented in this report under the difference-in-differences design is low because the authors did not account for trends in outcomes before the intervention or adequately control for time-varying characteristics that might influence the outcome. The regression discontinuity design does not receive a causal rating.
Amarante, V., Ferrando, M. & Vigorito, A. (2011). School attendance, child labor, and cash transfers. An impact evaluation of PANES. Retrieved from: https://ideas.repec.org/p/lvl/piercr/2011-22.html