The continuous treatment effect of an antipoverty program on children’s educational attainment: Colombia’s Familias en Accion

South America

Colombia

Social Protection

Conditional Cash Transfers (CCT)

JM Villa

2018

Villa, Juan M. “The Continuous Treatment Effect of an Antipoverty Program on Children’s Educational Attainment: Colombia’s Familias en Accion.” Review of development economics 22.3 (2018): 1239–1262. Web.
View publication

Secondary data: Familias en Accion. Available at: https://microdata.worldbank.org/index.php/catalog/3405.

2018

Familias en Accion

2001-2010

The objective of the program is to combat poverty by delivering cash transfers to households with children in extreme poverty.

The targeting of Familias en Accion relies on the identification of extremely poor households. Households in extreme poverty are identified with a proxy means test called Sisben (Castaneda, 2005; DNP [National Planning Department, Colombia], 2010). Sisben provides a welfare score ranging from zero to 100, denoting zero as the poorest and 100 as the wealthiest household. Households with scores below 11 in urban areas and 17.5 points in rural areas were eligible for the program. In addition to being identified as an extremely poor household, households must have children between zero and 17 years of age.

From 2001 to 2010 the program delivered two types of CCTs in municipalities with a population below 100,000 inhabitants. The first type is the nutrition cash transfer worth U.S.$25 a month, conditional on health check-ups of young children between zero and six years of age. The nutrition transfer was paid per household and not per child; thus, both households with one child and those with several children between 0 and 6 years of age received U.S.$25 a month. The program also delivered an education transfer worth U.S.$7 and U.S.$14 a month for each child between 7 and 17 years of age attending at least 80 percent of their classes in primary and secondary education, respectively. A household could receive several education transfers depending on the number of children attending school. On average, a household received U.S.$30 per month assuming the conditions set out by the CCT were met.

• School attendance
• Years of education

Regression Analysis

Study design/Identification Strategy: the empirical methodology of the paper is based on the estimation of the effects of the participation length in Familias en Accion on school attendance and years of education. The continuous treatment, measured as the length of participation in the program, is not set as random at the point of the household’s enrollment. Thus, the continuous treatment may be contaminated by potential endogeneity. In particular, a selection bias may arise when certain households with observed characteristics stay within the program and do not drop out after several years of participation. The proposal is to use a continuous treatment effects model as specified by Hirano and Imbens

(2004), in whose work the continuous treatment is randomized based on observational data. How the program was implemented provides the main source of external variation analyzed in this paper, which is variation in continuous treatment. The continuous treatment, dependent on the duration of households’ participation in the program, is determined by the introduction of the program in each municipality and household conditions at the date of enrollment. Continuous treatment is considered in relative terms. It is referred to as the elapsed proportion of the potential length of participation. Familias en Accion enrolls households with children of different ages. The enrollment age or the school grade of the youngest child determines the potential length of participation of each household in the program. All the children enrolled at 7 years old face censorship on the right at the age of 17 or when they graduate from school. The school grade differentials at enrollment can complicate the comparison of, for example, a child enrolled in the first school grade with an observed and potential participation length of 11 years to another enrolled in the third school grade with the same elapsed length of participation but a potential participation length of 8 years. The potential complications that emerge from the age/grade differentials are tackled by calculating the elapsed proportion of the potential length of participation of the household in the program with children enrolled at 7 years old. The elapsed proportion is determined by the actual participation duration since household enrollment at the point of the post-program survey collection. This allows for the comparability of children at different grades at enrollment by avoiding censorship on the right and the reference to specific time units.
Sample: The dataset contains more than 1.4 million households with 1.9 million children of school age in the period 2000 to 2010.
Survey Instrument: The empirical approach of this paper relies on the use of the Sisben data. Two main modules are part of the survey. The first is based on observational characteristics of a participant’s household, such as wall and floor materials, the provision of utilities, and the holding of durable assets. The second module is based on interviews to determine household composition and

individual characteristics, such as gender, education, and labor status. The Sisben data used in this second context are composed of two rounds. The first consists of the pre-program information about eligible households. The pre-program data were collected during the program implementation stage between 2000 and 2009. Between 2009 and 2011 the Sisben data were recollected with a new survey at the national level.

The paper focuses on the continuous treatment effects on school registration and educational attainment of participants in the Familias en Accion, a CCT program in Colombia. Although initial results show a fuzzy relationship between the program outcomes and the participation length, the empirical results confirm the fact that longer exposure to the antipoverty program led to higher school registration rates, accumulation of years of education, and lower child labor participation levels.

Child Labour

Households

National

Longer exposure to an antipoverty program in Colombia led to lower child labour participation levels.

Longer exposure to the antipoverty program led to higher school registration rates and accumulation of years of education.

Although the final empirical analysis is based on the children, identifying households in the follow-up survey and the determination of the elapsed years of participation are mostly determined by household characteristics. Also, after merging pre-program and post-program Sisben databases, the attrition rate was 18.2 percent (relatively high).