Document Type


Publication Date



This paper presents evidence on how children’s educational outcomes are affected by their family characteristics and their social background. We particularly focus on the school dropout rate of children by using data from the Bangladesh Demographic and Health Survey (BDHS) 2011. Recent research suggests that increasing dropout rates tend to raise inequality in income distribution which in turn can lead to greater school dropouts (Kearney and Levine, 2015; Heckman and LaFontaine, 2008). It is therefore important to understand the factors underlying school dropout rates.

In this paper, to make use of available data, we estimate the dropout in two different ways. The fact that many households have one or more children who stop going to school at some point allows us to construct for each family the school dropout percentage, a continuous variable, which can be easily estimated using OLS. We also set up a dichotomous variable indicating whether a household has one or more children who quit formal schooling.

Our results show that children from poor families with less educated parents experience a higher dropout rate with a lasting effect on educational outcomes. Father’s education plays a significant role in reducing dropouts whereas mother’s education is not found highly consequential. Households that have more than three children show greater dropouts. Larger households with more than five members also have fewer children going to school. Dividing households into four categories of wealth we find that the richer and the richest have the lowest dropouts. This indicates the need for policy interventions that prevent poorer households from pulling their children out of school. Many households do not have access to electricity or they are farther from a school with ease of transportation. Both these factors are associated with a higher dropout rate. On the other hand, the place of residence does not seem to matter since the coefficient for the ‘urban’ variable is statistically insignificant implying that urban households fare no better than those in the rural areas in keeping their children at school.

To see if the results of the OLS model described above change substantially when we estimate dropouts using a binary dropout variable through the probit model. But the probit results virtually mirror those from the OLS. The two models do not diverge much in sign or significance for most of our family or social characteristics such as urban-rural location and father’s education.

This paper contributes to the literature by studying an important aspect of human capital accumulation in Bangladesh. The results from our empirical exercise can suggest a way to assess the policies taken by the Government of Bangladesh. Over the last few years the school dropout rate has increased in the country. Any strategy that seeks to encourage building human capital must simultaneously address how to contain and reduce dropouts.