The 42 targeted localities were randomized into one of two equally sized treatment groups, the early or late group, at a public lottery carried out by the Government of Nicaragua in July 2000. To improve the likelihood that the selection of localities in the experimental groups would be well balanced in terms of poverty levels, a marginality index was used to classify the 42 localities into seven strata of six localities each. From each stratum three localities were randomly selected as early treatment and three as late treatment.
A program census, was implemented in May 2000 and and the 21 early treatment localities received their first transfers in November 2000. They were eligible to receive three years’ worth of cash transfers and received the last transfer in late 2003. Households in the late treatment localities were informed that the program would start in their localities later. The 21 late treatment localities were phased-in at the beginning of 2003. They were also eligible to receive three years’ worth of cash transfers. At the end of 2005, all program benefits were discontinued and the program no longer operated in these municipalities.
(In 2002, the study selected 21 localities in nearby municipalities as a non-experimental control group interviewed in 2002 and 2004.)
Between November 2009 and November 2011, i.e., between 9 and 11 years after the start of the program for the early-treatment group, researchers conducted a long-term follow-up survey (the 2010 survey). In this survey, researchers included all households in the original short-term evaluation survey, as well as a sample of additional households who, according to the 2000 program census, had children of ages critical to the long-term evaluation. Specifically, researchers oversampled households with children born between January and June 1989 and as well as households that, according to the RPS administrative data, had children born during the six months after the start of the transfers in 2000. The 2000 program census is used to obtain baseline information. The target sample has a total of 1,330 households from the early-treatment group and 1,379 households from the late-treatment group. In 2010, data was collected using an expanded household survey instrument, including new modules on labor market history and economic activities. In addition, a separate, individual-level instrument was designed to measure individual cognition, achievement levels and reproductive health outcomes of each child and young adult born after January 1, 1988. Respondents who could not be found in their original locations were tracked to new locations in Nicaragua and Costa Rica. Attrition ranges between 6 and 22% (depending on the cohort and the outcome).
(The target sample in the comparison group was 757 households.)
The main focus of the analysis of the 10-year differential effects on outcomes for young adults (study 1 and 2) is on the cohort aged 9–12 years at the start of the program in 2000. For this cohort, the combination of program eligibility, age at exposure, pre-program educational patterns, and age-of-menarche (for girls) suggest the CCT could have different effects in the early versus late treatment groups.
The analysis of the 10-year differential effects on anthropometrics and cognition (study 3), focuses on boys exposed to the program in utero and very early in life in the early treatment group, i.e. those born in the first year after the start of the transfers.
All analyses are carried out on an intent-to-treat (ITT) basis.