Does Health Insurance Reduce Child Labour and Education Gaps? Evidence from Rwanda

Africa

Rwanda

Social Protection

Health insurance

R Strobl

2017

Strobl, Renate. “Does Health Insurance Reduce Child Labour and Education Gaps? Evidence from Rwanda.” The Journal of Development Studies 53.9 (2017): 1376–1395.
View publication

Rwanda EICV2 Integrated Household Living Conditions Survey 2005/06. Available at: http://www.statistics.gov.rw/datasource/33.

2017

Community-based health insurance (CBHI)

From 1999 (introduction of prepayment schemes) up to 2003 (national system).

Reinstatement of Mutual Health Insurance in Rwanda.

Target evaluation of the paper: child labor is defined as total hours worked in the last seven days, including both economic activities and household chores.

Context: In 1996, two years after the genocide and a transitory provision of free health care, user fees for health services and drugs were reinstated in Rwanda. Therefore, health service utilization declined, since it was no longer affordable for many poor households (Schneider & Diop, 2001). This led the government 1999 to introduce prepayment schemes on a community basis in three pilot districts. Encouraged by the successful test phase, the insurance form has been progressively scaled up since 2003 to become a national system. With strong governmental leadership, intensive promotion campaigns, and external assistance, the population proportion covered by CBHI grew rapidly from 7 percent in 2003 to 44 percent in 2005, and 73 percent in 2006. Since 2008, mutual health insurance has been officially mandatory for the population and covered 91 percent of Rwanda’s inhabitants in 2010 (Ministry of Health, 2013). To capture schooling outcomes, the authors use three different measures. First, grade completion, that is the number of years of school completed by the child, as the dependent variable. Second, the education gap measures the difference between the highest schooling grade completed and the expected schooling grade according to the child’s age. Third, total education expenditures, that is, how much the household had paid in the last 12 months for children aged between 7 and 15. For the labor outcomes, the authors measure the number of hours worked separated between boys and girls.

• Grade completion, education gap, total education expenditures
• Number of hours worked (separated by boys and girls).

Propensity score matching.

Study design/identification strategy: one concern concerning the exogeneity of the instrument might be that there are unobserved community characteristics that influence both the insurance status of households and child outcomes. In this scenario, the instrument would be correlated with the error term in the outcome equation and therefore be invalid. However, by proxying the community’s poverty, remoteness, service supply, and markets, the author controls for the most relevant community-specific factors in our regressions. As a second instrumental variable, the author constructs a dummy indicating the satisfaction of the household with the quality of health services provided at the district hospital. As treatments at the district hospital are included in the Mutuelles benefit package, their quality should matter for enrolment and therefore satisfy the relevance condition. Further, this subjective quality perception is unlikely correlated with decisions on children’s time allocation.
Sample: The survey used (see below) provides the cross-sectional socio-economic data on 6900 households and 34,785 individuals, including information on consumption, education, health, assets, and agricultural production. For the rural areas of Rwanda, the survey additionally provides community-level data on items such as infrastructure, and access to health and educational facilities for 440 communities spread over 28 of the total 30 districts of Rwanda. The authors restrict the sample to the rural regions of Rwanda since community data and important information for our evaluations were only collected for these areas.
Survey Instrument: For the evaluations, the authors use the nationally representative Rwanda EICV2 Integrated Household Living Conditions Survey 2005/06 (NISR, 2006).

Children from households enrolled in a Mutuelle scheme work significantly less than children from non-enrolled households. Also, health insurance coverage improves schooling outcomes significantly, which is expressed in higher grade completion and smaller delays in education

Child Labour

Households

National

Children from households enrolled in the insurance scheme worked significantly less than children from non-enrolled households.

Health insurance coverage improves schooling outcomes significantly, which is expressed in higher grade completion and smaller delays in education.

Limitations of our study arise due to its observational character. Firstly, for lack of experimental data, causal inference has to be drawn with methods that build on critical, not testable assumptions. The existence of unobserved heterogeneity and its biasing effect on impact estimates cannot be ruled out.