The relative effectiveness of government and private schools in Pakistan: are girls worse off?

Abstract

Recent evidence from Pakistan points to significant pro‐male bias within households in the allocation of education expenditures. This raises two important questions. Is less spent on enrolled girls than boys through differential school‐type choice for the two sexes; for example, through a greater likelihood of sending boys to fee‐charging private schools? And, if indeed this is the case, are girls thereby condemned to lower quality schooling, on average, than boys? By asking these questions, this paper makes three contributions to the literature. Firstly, this is one of a very few studies in Pakistan to explore the question of the relative effectiveness of public and private schools despite there being an unprecedented expansion of fee‐charging private schools in the past two decades. Secondly, unlike existing papers that focus on primary schooling, this study looks at potential learning gaps by school type for students in their last year of middle school (Grade Eight), very near their transition to secondary schooling. Thirdly, it exploits unique, purposively‐collected data from government and private school students, and thus, in estimating achievement production functions, is able to control for a number of variables typically ‘unobserved’ by researchers. The findings reveal that boys are indeed more likely to be sent to private schools than girls within the household, so that differential school‐type choice is an important channel of differential treatment against girls. Private schools are also found to be of better quality – they are more effective than government schools in imparting mathematics and literacy skills. Girls lose out vis‐à‐vis boys in terms not only of lower within‐household educational expenditures, but also in terms of the quality of schooling accessed.

Keywords:

• privatisation
• school choice
• gender bias
• Pakistan

Acknowledgements

The author would like to thank Geeta Kingdon for the ideas that resulted in this paper. This paper has also substantially benefitted from comments by Harold Alderman, John Knight, Francis Teal, Stephen Bond, Marcel Fafchamps, Frances Stewart, Katy Graddy, Christopher Colclough, Niaz Asadullah, Justin Sandefur, Maria Lugo and two anonymous referees. The author also thanks the Gorman Student Workshop participants at Oxford for helpful suggestions. The author is grateful to the Rhodes Trust for the funding for the survey which generated the data used in this study. Any errors are the author’s own.

Notes

1. Figures from http://www.dfid.gov.uk/countries/asia/pakistan/mdgs.pdf

2. The 1992 figures are from Pakistan Education Statistics (1992–1993), and 2004 figures are from Pakistan Economic Survey (2004–2005).

3. While age‐at‐marriage is increasing, the mean marriage age for females is 17.9 years among the 20‐year‐old to 24‐year‐old married youth (Sultana 2005).

4. It is possible to appear for public examinations either as a ‘Regular’ or ‘Private’ candidate. Regular candidates’ admissions for examinations are sent through recognised schools. Private candidates, on the other hand, send in their examination admissions without any institutional backing. This system has arisen historically as many candidates wish to take public examinations and obtain secondary or higher secondary school certificates without formally attending schools. This is especially true for females.

5. Although the Federal Bureau of Statistics (FBS 2001) claims to have interviewed both registered and unregistered schools, whether all unregistered schools have been documented is arguable. Despite the fact that my survey and fieldwork was conducted after the census of educational institutions had already been undertaken, there was no comprehensive list of the un‐registered schools that had apparently been counted in the census.

6. Alderman, Orazem, and Paterno (2001), in their study of schooling choices of low‐income families in urban Lahore, find that private schools catering to the urban poor charge low fees. In our sample, average private school fees range from Rs. 116/month to Rs. 3766/month.

7. But even this high FBS number is likely to be greatly underestimated since the number of private schools that are unregistered is not fully accounted for.

8. The 2000 estimates are based on two sources. Private school figures are from the Census of private institutions of 2001, and the total number of schools and enrolments in primary, middle and high schools are from the Pakistan Statistical Yearbook (2001). The figures from 1983 form a rough comparison as those figures are presumably based on numbers of and enrolments in primary, middle and high schools, while the 2000 figures are based on primary, middle, high and higher secondary schools.

9. This analysis is restricted to the 97% of children who are enrolled in government and private schools only.

10. There were too few observations in the 20–24 years old age group and thus it has not been included.

11. Both OLS and probit models were estimated. Since LPM and probit models yielded almost identical results, we can be confident in the use of LPM in household fixed‐effects estimation.

12. Using the PIHS, we estimated earnings functions incorporating the standard education/experience variables and a dummy variable, PRIVATE (= 1 if individual was enrolled in a private school when of school‐going age, and = 0 otherwise). Three additional variables were included: READ = 1 if individual can ‘read in any language with understanding’, = 0 otherwise; WRITE = 1 if individual can ‘write in any language with understanding, = 0 otherwise’; and MATHS = 1 if individual can ‘solve simple (plus minus) sums, = 0 otherwise. As READ/WRITE/MATHS may be correlated with PRIVATE, the first specification introduces them independently to capture their effect, if any on earnings before introducing the school‐type dummy. The results are suppressed due to space constraints but the main findings are reported above.

13. One of the most robust findings in educational research is that a child’s educational attainment is consistently driven by family background (Chevalier, Dolton, and Levacic 2004, 1).

14. While the Ravens measure of ability is possibly endogenous in that it is not independent of the child achievement level, as long as there is some exogenous element in the Raven’s test, it does provide a valid measure of ability – and several studies have used it as such.

15. We asked the child the question ‘What is the highest level of education you wish to attain?’, with all possible educational levels (such as middle school or up to eighth grade, Matric or Grade 10, etc.) as answer choices.

16. In the government subsample, the computed χ² statistic (12) is 1.89; while in the private sample, the computed χ² (12) is 0.16. The critical χ² statistics at 12 degrees of freedom is 21.03 at the 5% level. We can accept the H_o that the difference in coefficients between Heckman and OLS is not systematic.

17. For a comprehensive study of gender differences in academic achievement in Kenya, see Appleton (1994).