Madrassa Trap? Poverty, Preference, and Muslim Religious Schooling in Indonesia

Abstract

This paper examines the possibility of a low-income trap due to parental choice of madrassas, the Islamic religious schools, for children's education, which in turn may result in intergenerational persistence in madrassa education, i.e., a madrassa trap. A common narrative is that the poor households in Muslim developing countries send children to madrassas where they are taught conservative ideas. Using data from Indonesia Family Life Survey (IFLS) 2007, this paper estimates the effects of household income on children's madrassa enrollment. We use rainfall variation as an instrument for household income to correct for biases due to unobserved ability and preference heterogeneity and measurement error. The estimated income effect shows that doubling the average level of per capita household expenditure reduces children's probability of madrassa enrollment by about 34 percentage points. The estimated negative income effect is robust to various specifications and holds when the Conley et al. (2012) approach is used to relax the exclusion restriction imposed on rainfall variance. To understand if returns to education vary between madrassa and general schools, we take advantage of econometric techniques that tackle biases in the absence of valid external instruments. The OLS estimates of returns to education show that the average earnings of junior general school graduates are about 26.5 percent higher than that of junior madrassa graduates. We exploit selection on observables as a guide to selection on unobservables and use the recent approach developed by Oster (2016) in an extension of the Altonji et al. (2005) methodology. The relatively low returns to Madrassa education remain intact even when selection on unobservables is twice as large as the selection on observables. The estimates from heteroskedasticity based identification due to Klein and Vella (2009b) also suggest a similar pattern. The combination of the negative income effect on madrassa enrollment and lower returns to madrassa education indicates that there is a low-income trap due to the choice of madrassa for children's education. This also leads to a madrassa trap where intergenerational persistence in madrassa education operates through a low-income trap.

Keywords: Education, instrumental variable, returns to education, religion, madrassa, low income trap