Public works are one of the most popularsafety net and employment policy instruments inthedeveloping world, despite limited evidence on theireffectiveness and optimal design features.This paperpresents results on contemporaneous and post-program impactsfrom a public worksintervention in Côte d'Ivoire. Theprogram provided 7 months of temporary employment inroadmaintenance to urban youths. Participants self-selected toapply for the public works jobs,which paid the formalminimum wage and were randomized among applicants.Randomizedsub-sets of beneficiaries also receivedcomplementary training on basic entrepreneurship orjobsearch skills. During the program, results show limitedcontemporaneous impacts of publicworks on the level ofemployment, but a shift in the composition of employmenttowards thebetter-paid public works wage jobs. A year afterthe end of the program, there are no lastingimpacts on thelevel or composition of employment, although positiveimpacts are observed onearnings through higher productivityin non-agricultural self-employment. Large heterogeneityinimpacts are found, particularly during the program. Resultsfrom machine learningtechniques suggest potential trade-offsbetween maximizing contemporaneous and postprogramimpacts.Traditional heterogeneity analysis shows that a range ofpractical targetingmechanisms perform as well as the machinelearning benchmark, leading to strongercontemporaneous andpost-program benefits without sharp trade-offs. Overall,departing fromself-targeting based on the formal minimumwage would lead to strong improvements inprogram cost-effectiveness.
PDF
download
878987797
028-85220240
