We evaluate the large scale pilot of an innovative and major welfare intervention in Colombia, which combines homes visits by trained social workers to households in extreme poverty with preferential access to social programs. We use a randomized control trial and a very rich dataset collected as part of the evaluation to identify program impacts on the knowledge and take-up of social programs and the labor supply of targeted households. We find no consistent impact of the program on these outcomes, possibly because the way the pilot was implemented resulted in very light treatment in terms of home visits. Importantly, administrative data indicates that the program has been rolled out nationally in a very similar fashion, suggesting that this major national program is likely to fail in making a significant contribution to reducing extreme poverty. We suggest that the program should undergo substantial reforms, which in turn should be evaluated.
Abramovsky, Laura et al. 2017. "Promoting Social Inclusion of the Extreme Poor in Colombia." AEA RCT Registry. May 24. https://doi.org/10.1257/rct.1976-1.0.
Conditional cash transfer program take-up, Expanded knowledge of existing social programs, Labor market outcomes, Welfare improvements, Health improvements, Financial inclusion, Housing conditions
Primary Outcomes (explanation)
Labor market outcomes are measured through participation rate, employment rate (and type of employment), unemployment rate, hours worked, employment earnings and tenure; Health improvements are measured by nutrition and growth, vaccination take-up, and use of health services.
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
The evaluation employed an experimental design to ensure that individuals in treatment and control groups were comparable along observable and unobservable dimensions. Random assignment to treatment and control groups followed a structured process. First, the population of eligible families within participating municipalities were identified in early 2008. Second, each participating municipality was divided into several neighborhoods, or 'barrios'. Third, between September 2008 and April 2009, each neighborhood was randomly assigned to one of four groups or cohorts. The program was rolled out to cohorts sequentially, so that the treatment began at different times across different neighborhoods. Given random assignment to cohorts, prior to the roll-out of the program the characteristics of households across neighborhoods should be identical on average.
Experimental Design Details
Randomization Method
By computer
Randomization Unit
Neighborhoods
Was the treatment clustered?
Yes
Sample size: planned number of clusters
1280 neighborhoods
Sample size: planned number of observations
5918 households
Sample size (or number of clusters) by treatment arms
551 neighborhoods in the control group control
534 neighborhoods in the classic treatment group
195 neighborhoods in the intensive treatment group
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Laura Abramovsky, George Stoye, Orazio Attanasio, Kai Barron, and Pedro Carneiro. "Challenges to Promoting Social Inclusion of the Extreme Poor: Evidence from a Large-Scale Experiment in Colombia," EconomÃa: Spring 2016.
Abramovsky, Laura, Orazio Attanasio, Kai Barron, Pedro Carneiro, and George Stoye. "Challenges to Promoting Social Inclusion of the Extreme Poor: Evidence from a Large Scale Experiment in Colombia." IFS Working Paper W14/33, November 2014.