Gender bias in schooling: the case for Bhutan

Abstract

Differing parental considerations for girls and boys in households are a primary cause of the gender gap in school enrolment and educational attainment in developing countries, particularly in Sub-Saharan Africa and South Asia. While a number of studies have focused on the inequality of educational opportunities in South Asia, little is known about Bhutan. This study uses recent household expenditure data from the Bhutan Living Standard Survey to evaluate the gender gap in the allocation of resources for schooling. The findings, based on cross-sectional as well as household fixed-effect approaches, suggest that girls are less likely to enrol in school but are not allocated fewer resources once they are enrolled.

Keywords:

• education inequality
• gender bias
• two-part model

JEL classification:

• J16
• C23
• C25
• I24

Acknowledgement

We thank Lionel Page and Benno Torgler for helpful comments and suggestions. All remaining errors are ours.

Notes

1. Based on the budgetary outlay of education and health sectors in the fiscal year of 2010–2011 (National Budget Report, Ministry of Finance).

2. The adult literacy rate of the country as estimated from the Population and Housing Census (2005) reveals a 20% lower level for females suggesting the existence of gender-biased preferences for educational investment.

3. See, for example, Kingdon (2005), Aslam and Kingdon (2008), Himaz (2010) and Zimmermann (2012). While Zimmermann (2012), Kingdon (2005), and Aslam and Kingdon (2008) confirm the existence of a preference for spending on educating boys in India and Pakistan, respectively, the reverse is found in Sri Lanka where both the preference and the allocation of educational expense favour girls (Himaz 2010).

4. For instance, a pro-male bias may exist in the decision of households to spend on the child's education, while households may be pro-female in the amount spent, indicating an opposite bias in the second stage of the decision-making process of resource allocation by households.

5. It must be noted that this strategy only controls for a limited form of endogeneity attributable to intra-household resource allocation patterns. Addressing the concerns for other sources of potential endogeneity constitutes an important avenue for future research but is outside the scope of the current paper.

6. Further details of the survey design can be found in the NSB (2007).

7. Parents of disadvantaged children may incur zero expenses due to fee waivers. Given that less than 2% of the children in our sample fall in this category, the log transformation has minimal effect on the sample size.

8. The US$ equivalent of our sample average schooling expenditure is $36 (approx.). It amounts to roughly 7% of private consumption expenditure as reported by the NSB (2011). We omit observations that report more than Nu. 15,000 (80% of consumption expenditure) annual schooling expenses.

9. On the contrary, Chernichovsky and Oey (1985) find a positive association between the number of school-going age children in a household and school enrolment in Botswana. They suggest, with more children available, only some are needed for farm work, freeing others to attend school.

10. According to Figures A1 and A2 in Appendix 1 showing kernel density estimates of school expenditures in levels and logs, respectively, educational expenditure seems to be lognormally distributed in our sample.

11. Since less than 2% of enrolled children incur zero expenses in our study, positive enrolment decision is treated as equivalent to incurring positive expenditures.

12. While the Hausman test (not reported here but available upon request) implies that fixed effects is ideal to estimate the enrolment decision, the probit model does not lend itself to a fixed-effect treatment (Baltagi 1995). An alternative is provided in Greene (2004), but due to its technical complexity, it is outside the scope of the current paper. Hence, we resort to a random-effect model in the first stage.

13. We find a relatively large number of live-in servants reside in households with college-educated heads. These servants are also more likely to be females when compared with the households with non-educated heads. For example, 80% of live-in servants in the 6–16 age group are females. Almost 8% of households with college-educated heads have live-in servants (of which 11% are females) as compared with only 0.5% for households with non-educated heads (of which only 0.6% are females). On dropping the live-in servants from the estimation sample, the negative marginal effect of college-educated heads falls to –0.025 and becomes insignificant.

14. Following a referee's comment, we examine the possibility of boys being sent to better quality schools that may not be reflected in higher expenditures. We test for significant differences in various measures of school quality across gender. These include private schooling, availability of books and supplies, teacher absenteeism, teacher competence, relevance of the teaching programme and student–teacher ratio. The t-statistics for mean differences are not rejected in any case.

15. Additionally, we estimate a truncated normal regression for the expenditure equation (Cragg 1971). Results are qualitatively same as the lognormal model.

16. In the absence of interactions of the female dummy with the age groups, the gender gap is –9% and significant at the 1% level. These results are not presented here for space considerations but are available from the authors upon request.

17. In keeping with the results reported for the fixed-effect models in columns (l) and (3), we include estimates for the interaction terms only for the random-effect model in column (2).