To evaluate the impact of the program, we will exploit the fact that beneficiaries are randomly selected through a public lottery. Around 1 to 3 days prior to the registration, an information campaign takes place to disseminate information on the job opportunities that would be available and how the selection process will be. Field staff register up to double the number of workers needed in each worksite, on a first-come-first-serve basis. Among the registered participants, a public lottery is conducted, at the health zone level, to select the beneficiaries and those that will be part of a waiting list in each worksite. In each health zone, there will be as many lotteries as worksites. Out of 100 participants randomly selected to participate in the project, between 20 and 30 are randomly selected to be part of a waiting list. If someone from the main list decides not to participate, his/her position is filled with the first position in the waiting list and so on. People from the waiting list are called to replace beneficiaries who drop out, never show up or, complete 3 consecutive days of absence without a valid justification.
We will use as a counterfactual those that participated in the lottery and are randomly assigned to the waiting list. The random selection of beneficiaries, based on the whole list of people registered, allow us to get rid of the selection bias associated to program participation and hence, allow us to have comparable groups that only differ in the likelihood of having participated in the public works. Prior to estimating the impact of the program, we will verify that treatment and control groups are comparable in key observable characteristics.