As a robustness check to the results I find in Chapter4, we can also use the education rounds of the Na- tional Sample Survey to perform similar tests on a different source of data. The National Sample Survey conducted surveys on education in 1986-87, 1995-96, 2007-08, and 2014. Although the panel is not as clean and proximate to the passage of DPEP, it does allow us to test household level effects of DPEP on private enrollment, expenditure, distance to schools, and literacy.
For this, I use the four nationally representative education waves of the National Sample Survey (NSS) conducted in 1986-1987, 1995-1996, 2008, and 2014. The NSS education rounds are stratified by rural and urban areas for each district. Surveying is then further subdivided into four sub-rounds each lasting three months. The NSS oversamples some types of households and therefore provides sampling weights. All statistic and estimates are adjusted with these sampling weights.
The education sub-rounds of the NSS provide data on household level participation in education and out of pocket expenditure on education. Beginning with the 71st round of 2014, the NSS began to ask detailed questions about exit to the private sector, including the reasons households chose private over public education. Unfortunately there is only one round of this data and was administered too late to be useful. Along with detailed individual level data on educational attainment and expenditure, each round provides data on household consumption, expenditure and other household-level demographics.
First, I plot regression discontinuity plots to visualize the impact of DPEP around the DPEP threshold on the four variables of interest from the NSS data: distance to the nearest government school, the number of children in a private primary school, whether a household has at least one child in a private primary school, and the logged out of pocket expenditure on education.
The figures support the earlier results from Figure4.12at the household instead of district level. House- holds in DPEP districts were more likely to send their children to private schools and more households had at least one child in private schools than households in non-DPEP districts. There is no clear rela- tionship from the regression discontinuity plots between DPEP funding and out-of-pocket expenditure on education, although we should remember that these do not take into account the fuzzy nature of DPEP assignment. I account for the full implementation of DPEP next by fitting a fuzzy regression discontinuity
1 1.2 1.4 1.6 1.8 2
Distance to Government School
0 20 40 60 80 1991 Female Literacy (%) 0 .2 .4 .6 .8 1 Private School 0 20 40 60 80 1991 Female Literacy (%) 0 .2 .4 .6 .8
Any Private School (\%)
0 20 40 60 80 1991 Female Literacy (%) 3 4 5 6
Log Out of Pocket Expenditure (Rs.)
0 20 40 60 80
1991 Female Literacy (%)
Figure B.7:Discontinuity Plots Using National Sample Survey Data
The top left plot presents the effect of DPEP funding on a household’s self-reported distance to a government school. The top right plot presents the effect of DPEP funding on the number of children in a household in private primary education. The bottom left plot presents the effect of DPEP on whether a household sent any children to private primary schools. The bottom right plot presents the effect of DPEP funding on logged out of pocket expenditure on private education in 2010 Rupees. The literacy cutoff for receiving DPEP funds is 39.3 percent. The points show the average percent of villages with a government school in a district within a small bin of the literacy margin. The lines are the second-order local polynomial best-fit lines fit separately on each side of the cutoff. The plots are based on the procedure developed byCalonico, Cattaneo and Titiunik(2014b).
(1) (2) (3) (4) Distance to School Private School One Child in Private School Log Expenditure (Rs.)
Literacy Cutoff -0.33∗∗∗ -0.18∗∗∗ -0.11∗∗∗ -0.60∗∗∗
(0.03) (0.03) (0.02) (0.09)
Observations 40092 54161 54034 53603
* p<0.1, ** p<0.05, *** p<0.01. Robust standard errors in parentheses.
Notes: The dependent variable in column one is an ordinal variable that measures the distance to the nearest primary government school in four ranges: 1=Less than 1 km; 2=1 to 2 kms; 3=2 to 5 kms; 4=Greater than 5 kms. Column 2 is the number of children in the household in a private primary school. Column 3 is a dummy for whether the household has at least one child in private primary school, and column 4 is the logged out of pocket expenditure on primary education in 2010 Rupees.
Source: National Sample Survey, 52nd Round, Schedule 25.2 (1996).
Table B.5:Regression Discontinuity Estimates of DPEP
the nearest government school, the number of children in a private school, whether at least one child is in a private school, and the log out of pocket expenditure on education by the household. Within the NSS, the distance to the nearest primary school is coded as 1 if the household is less than 1 kilometer from a primary school, 2 if it is between 1-2 kilometers from the nearest primary school, 3 if it is between 2-5 kilometers to the nearest primary school, and 4 if it is more than 5 kilometers to the nearest primary school.
Column 1 suggests that, contrary to the findings in Chapter 4, households in districts that received DPEP were further from a government school than households in non-DPEP districts. This finding con- tradicts findings from the Economic Census, DISE, and Census of India that all suggest that the introduction of DPEP significant reduced the number of villages without a government school. The NSS data is, how- ever, self-reported, so households answering this question might not have known that a new government school opened near them. Regardless, this results merits further unpacking.
Column 2 suggests that households in DPEP districts were more likely to send their children to a private school than households in non-DPEP districts. The negative point estimate suggests that households to the right hand side of the DPEP implementation cutoff, or households that did not live in districts where DPEP was implemented, werelesslikely to send their children to private schools. Column 3, which fits the same model as column 2 but uses whetheranychildren were in private school instead of the number of children, suggests that households in DPEP districts were about eleven percent more likely to send their children to private schools that households in non-DPEP districts. Finally, column 4 suggests that households in DPEP districts spent more out of pocket on education than households in non-DPEP districts.
created exit to the private sector. Therefore results at the household and district level all suggest that DPEP had a substantial impact on exit to the private sector.
B.3.1 Discussion
The results using household-level data from the NSS support results from district level data from the Eco- nomic Census and DISE. Districts that received DPEP funding saw a greater level of exit to the private sector than districts that did not receive DPEP funding. The finding that households in DPEP districts lived farther from a government primary school merits further unpacking, but one potential explanation is that households were unaware of new government schools built near them. Finally, although the ef- fects on out of pock expenditure on education are small, this is also consistent with a growth inlow-cost
private school in DPEP districts that cater to low-income households. Together with the fact that more households in DPEP districts are sending their children to private schools, along with low growth in out of pocket expenditure on private education suggests that although households in DPEP districts are send- ing more children to private schools, they are not spending much more out of pocket for education. This suggests that the additional children being sent to private schools are not being sent to expensive private schools.