Regional Returns to Education, Child Labour and Schooling in India

Abstract

We offer evidence from India that higher regional returns to primary education not only increase the likelihood that boys and girls attend school but also decrease the likelihood that they work. These relationships hold only for the top three quintiles of the income distribution and mostly for children in the age group 10–14 years. The former result suggests that liquidity constraints may not allow poorer households to respond to the economic benefits of education. Policies that raise the economic benefits of education may increase human capital investments in households that do not rely on their children's incomes for survival. However, low schooling and high child labour will persist among credit constrained families unless these households are provided with the economic ability to respond to these benefits.

Acknowledgements

I thank Roger Betancourt, Deborah Minehart, Ginger Jin, two anonymous referees, and other seminar participants at the University of Maryland for their helpful comments.

Notes

1. Bhalotra and Heady (2003) find that the daughters of land-rich households are more likely to work than the daughters of land-poor households in both Ghana and Pakistan. They refer to this phenomenon as the ‘wealth paradox’.

2. There is conflicting evidence on the returns to education in developing countries. Some researchers find relatively high returns to education in developing countries while others find no consistent pattern between growth or development and the returns to education (Psacharopoulos, 1994; Bennell, 1996; Nielson and Westergard-Nielson, 1998; Denny et al., 2001).

3. For example, in the 55th Round (1999–2000) intra-district, intra-state and inter-state migrants constituted approximately 6.2 per cent, 5.0 per cent, and 4.0 per cent respectively of the sample households.

4. The all-India CPI for agricultural and industrial workers is used to adjust rural and urban wages respectively.

5. Until the year 2002, education was not compulsory in India. In December 2002, however, Article 21A was incorporated in the Indian Constitution, which stated that every child in the age group 6–14 years has the right to free and compulsory education. Since all our data is pre-2002, our empirical analysis is not affected by this right.

6. Even though all children attend school during five or at most six days of the week, these children report full intensity of attending school on seven days because they spend their free time engaged in homework or other school-related activities rather than in work or idleness.

7. The inclusion of the number of primary schools per 1000 population at the state level also provides a control for state level differences in availability of schools. An alternative method of controlling for state level differences in the access to and quality of schools (which is considerable in India) is to include state fixed effects. However, since we include region fixed effects we cannot also include state fixed effects because each state consists of several regions. Given that the main variables of interest (returns to education) are measured at the regional level, it seems more relevant to control for region fixed effects rather than state fixed effects, the reason being that the exclusion of region fixed effects may result in a spurious correlation between regional returns to education and participation in school or work.

8. Approximately 51 per cent of children in the data have both parents with less than primary education (see Table 4).

9. We also estimate the model using the teacher-pupil ratio instead of the number of primary schools in a state to use a measure of the quality rather than the quantity of schooling. Replacing the number of primary schools in a state with the teacher-pupil ratio in that state provides almost identical results.

10. Each of these Wald statistics is based on a Wald test where the null hypothesis is that the sum of these coefficients is zero. With 1 degree of freedom, the critical c2 is 6.63, 3.84, and 2.70 at the 1 per cent, 5 per cent, and 10 per cent levels of significance.

11. Since households where children work will have a higher household income and, therefore, higher expenditure, the potential endogeneity of monthly per capita expenditure will result in a negative correlation with schooling and a positive correlation with child labour.

12. We divide our sample into income quintiles based on reported monthly per capita expenditure for each household.

13. The omitted dummy is age 9 for the 5–9 age-group regression and age 14 for the 10–14 age-group regression.