Impact of private secondary schooling on cognitive skills: evidence from India^†

† The authors may be contacted at [email protected] or [email protected]. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. All the errors remain of the authors.

ABSTRACT

We examine the effect of attending private secondary school on educational achievement, as measured by students' scores in a comprehensive standardized math test, in two Indian states: Orissa and Rajasthan. We use propensity score matching (PSM) to control for any systematic differences between students attending private secondary schools and public secondary schools, and assess the sensitivity of our estimates with respect to unobservables using the Rosenbaum bounds. We find that students in private schools in rural (urban) Rajasthan scored about 1.3 (0.4) standard deviation (SD) higher than their counterparts in the public schools. Importantly, the positive private school impact in rural (urban) Rajasthan survives a large (moderate) amount of positive selection on unobservables. We do not find statistically significant difference in urban Orissa, while a positive impact of 0.3 SD in rural Orissa is susceptible to small amount of positive selection on unobservables.

KEYWORDS:

• Public and private school
• secondary school
• propensity score matching
• India

JEL CLASSIFICATION:

• C21
• I21

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

† The authors may be contacted at [email protected] or [email protected]. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. All the errors remain of the authors.

1. The Annual Status of Education Report (ASER 2015) shows that the share of private schools in total elementary school enrolment in rural India increased from 16% to 31%, in the short period 2006–2014. According to National Sample Survey, 29.7%, 22.8%, and 25.5% students at the primary, middle, and secondary school levels in India attended private unaided schools in 2014. This is a considerable increase over 2007 when 20.3%, 17.3%, and 18.4% students at the primary, middle, and secondary school levels reported attending private schools (National Sample Survey Organization 2015, 2010). Several research papers also point out the rapid increase in the enrolment share of private elementary schools in India (Kingdon 2007, 2008; Muralidharan and Kremer 2008; Tooley and Dixon 2005).

2. Kin Bing Wu led the study in her capacity as lead education specialist for the World Bank. For further details on the survey design, see Wu et al. (2006, 2007). The same data has been used in Wu et al. (2006, 2007, 2009), World Bank (2009), and Das and Zajonc (2010).

3. In 2014, the per capita GDP of Rajasthan and Orissa was $1443 and $1150, respectively, while national per capita GDP was $1627 (source: Ministry of Statistics and Programme Implementation, India).

4. The test comprises 36 items which assess general mathematic content, knowledge of data representation and analysis, fraction and number sense, algebra, geometry, probability, and statistics (see Table A1, in the online supplemental data at 10.1080/09645292.2015.1110116). The items were selected from a sample of published items from the Third International Mathematics and Science Study (TIMSS) for grade 8. Although the original TIMSS populations were the eighth grade, in both the Indian states, test was implemented to students of grade 9. However, more difficult items from the TIMSS tests were chosen to adjust for the grade difference. The tests were shown to teachers, students, and state-level officials to ensure that they were within the curriculum.

5. Since, the survey asked questions to the students, it does not have information on household income.

6. A total of 185 and 396 students attending private-aided schools in Rajasthan and Orissa, respectively, were dropped from the sample. Seven students were further dropped from Rajasthan sample as they did not respond to any of the 36 questions.

7. Altonji, Elder, and Taber (2005b) evaluate the instrument used in two influential papers – Evans and Schwab (1995) and Neal (1997) – on the effect of Catholic school in the USA, and conclude that none of the candidate instruments is a useful source of the Catholic school effect, at least in the NELS:88 data set.

8. We used psmatch2 (Leuven and Sianesi 2003) in STATA to get PSM estimate.

9. Standard errors are clustered at the school level to allow for correlation among students attending the same school.

10. We also estimated the equation with additional controls such as attended pre-primary, hours spent doing math's homework, started grade at less than 5 years age, hours spent on household chores, works outside, individual private tutoring, group private tutoring, and hours per week private tutoring. However, all these variables were insignificant. Hence, we do not include those in our final model.

11. Altonji, Elder, and Taber (2005a) also provide an informal way to use information about selection on the observables as a guide to selection on the unobservables, applicable to continuous case. The idea is to assess how much selection on unobservables there must be, relative to the amount of selection on observables, to fully account for the positive association between private school attendance and achievement. Under the hypothesis of equality of selection (normalized) on observables and unobservables, Altonji, Elder, and Taber (2005a) derive an implied ratio. However, in our case the is very large, and as discussed in Altonji, Elder, and Taber (2005a), in case of very large , the implied ratio is not very informative.

12. We used kernel weighting with the epanechnikov kernel and a fixed bandwidth of 0.10. Confidence intervals are obtained using 100 bootstrap repetitions.

13. The matching is done with replacement. Abadie and Imbens (2006) imply that the use of four neighbors minimizes mean-squared error, although our results are largely insensitive to including between one and five nearest neighbors.