Private Schooling and achievement in India: A new Educational - - PDF document

1 First Draft

Private Schooling and achievement in India: A new Educational Landscape?

Suvarna Pande1 and Amaresh Dubey2

• 1. Introduction

The Indian Education system has undergone massive structural change in the past few

• years. The trajectory and mechanism of effect of private school system in India

remains a sparsely understood and an under researched field. Increased enrollment has not meant similar results in learning outcomes especially in the Indian context where private schools are often unrecognized, unregulated with poorly trained

• teachers. Without a thorough analysis it cannot be concluded that they are of higher

quality than government schools. The World Bank in its Development Report 2018 discusses the global education and learning crisis and the associated future development costs. The wasted development

• pportunities and widened social gaps have been highlighted as a major public policy

failure in the recent times. With multiaxial lines of inequalities in India the resulting disadvantage is even greater- young students struggling with poverty, gender and social discrimination reach adulthood without even the most basic skills. Any sustainable economic transformation and catch-up is not possible if basic human skill and capital is left behind. Each additional year of schooling has the potential to raise individual’s earnings by 8 to 10 percent. The effect has been reported to be even greater for disadvantaged sections in particular women. In spite of these expected benefits, glaring gaps in cognitive capabilities have been found for India (India ranks second after Malawi in a list of 12 countries wherein a grade two student could not read a single word of a short text. India also tops the list of seven countries in which a grade two student could not perform two-digit subtraction). It is in this backdrop that the paper demands paramount attention. Numerous policy interventions in the field have resulted in greater quantity numbers, while quality focus has been left behind. Public education, despite massive central and quasi-central fund support is perceived as inferior and inefficient. The scenario has changed

1 Corresponding author: Ph. D. Student, Centre for the Study of Regional Development, Jawaharlal

Nehru University, New Delhi – 110 067, India. e-mail: suvarnapande2008@gmail.com

2 Professor, Centre for the Study of Regional Development, Jawaharlal Nehru University, New Delhi –

110 067, India. amareshdubey@mail.jnu.ac.in

2 drastically especially after the 1990s when growth picked up and private schools were hailed as substitutes and the “better off” solution to the education panacea. However, during the same period efforts at improving primary education were intensified even further leading to three ambitious and far-reaching initiatives, the District Primary Education Programme (DPEP) introduced in the mid-1990s, the Sarva Shiksha Abhiyan (SSA) launched in 2000 following the MDG declaration and the Mid-day Meal Program (MDM) universalized in 2001(Datta Gupta, Dubey et al. 2017). DPEP and SSA were ambitious programs primarily focused on large-scale infrastructure funded by World Bank and Central Government Cess. These interventions though succeeded in improving enrollments, did not lead to perceptible and documented increments in quality aspects. In the same period, numerous private institutions – aided and unaided mushroomed across the country as a way out for the populace struggling with government institutional apathy but understood the importance of education for growth. The paper is an attempt to explore effectiveness of private education in India. The work considers the quality debate surrounding the private schooling system. It empirically explores the following considerations: parental choices that propel children from certain backgrounds into certain types of schools (Hanushek 1997) and whether private schools in India, often unrecognized and unregulated were necessarily better than government schools in terms of learning outcomes. There has been relatively limited research on these important public policy issues for the Indian scenario due to lack of appropriate extensive data on the issue. The India Human Development Survey (IHDS-I; 2005) and IHDS-II (2011-12), a joint exercise conducted by the National Council for Applied Economic Research (NCAER) and University of Maryland makes it possible to link factors private school growth and quality variables. The paper takes off from and expands on similar work by Desai et al., 2008, where the authors use data from the first round of Indian Human Development Survey (2005) to describe public and private schools, factors driving parents in selecting private schools and association with student performance. Here, we use the findings from IHDS- II (2011-12) to document and reason out changes between the two periods and therefore “effectiveness” of private school enrollment. In particular, in addition to studying factors guiding school choices, we also seek to ascertain whether enrollment actually results in higher learning outcomes and whether those gains are concentrated among particular sections of the population. The paper progresses with a descriptive survey of international studies on public and private schooling, concluded policy considerations from the limited available Indian studies in section two. It is followed by a brief description of the India Human Development survey – I and II and the methodology used in the third section. These theoretical sections give way to statistical descriptions of the nature of school systems, characteristics of public and private schools and social and economic backgrounds of students who attend these schools in the fourth section. The next section examines

3 impact of private school enrollment on child’s outcome and is subsequently followed by unearthing the characteristics of children who benefit most from private school

• enrollment. The paper finishes with discussion on changes observed between the two

rounds, the policy conclusions and description of possible implications for the future.

• 2. Literature on Public and Private Schools

Comparative past experiences of currently leading world economies bring out the centrality of the issue at hand in a manner that doesn’t need much proof. The European and American cases show how government initiatives are necessary to f8acilitate and sustain long-term economic and social progress. Beginning Nineteenth century, the Asian countries learnt their lessons from the erstwhile developed nations and began recognizing the transformative role of school education. Japan, the first one to do so in Asia, its authoritative yet constructive focus on literacy allowed it to deliver exponential progress. In the following years, South Korea, Taiwan, Singapore, Hong Kong and China followed similar routes and firmly focused on state delivered basic education. The transformative role of public education in their development experience has been rightly emphasized as one of the most important factors behind their successes. “But that process was greatly helped by the achievements of these countries in public education. Widespread participation in a global economy would have been hard to accomplish if people could not read or write.” (Drèze and Sen 2013). This recognition of indispensable role of state supported education in any mature civil society resulted in massive expansion of publicly provided education(Meyer 1977). However, with increased enrollment rates, exasperation over learning outcomes from an already strained public education system exacerbated and private education was hailed as the more efficient alternative. The increased disappointment led to steady growth in private enrollments. The debate over public vs. private education systems can be triangulated into three aspects – a) international school effects debate; b) the Indian experience on quality of public and private schools and c) policy alternatives under consideration.

• 1. School effects debate in an International Context

The debate on school effects in the United States began with the Coleman report in 1966.The report concluded that it was not only school level characteristics but also family level household factors that determined what students learned. In fact the report specifically argued for placing greater importance on household factors to positively affect learning outcomes. The early focus on home environment and parental effects was an upshot in education research and an important policy takeaway from a field that gave primacy to role of school level inputs. Literature from developing countries also concluded weak to negligible relationship between school inputs and child outcomes (Hanushek 1995, Banerjee, Cole et al. 2007). Comparative

4 studies for low developing countries have concluded, however that school effects maybe far more important. (Heyneman and Loxley 1983) Once the importance of school characteristics is accepted, the next logical step is to understand the impact of nature of school on learning outcomes. While it may be easy to argue the role of social capital in higher learning student performances of American Children in catholic schools as compared to those enrolled in public schools. (Coleman, Hoffer et al. 1982). Similar simplistic analysis may be difficult in case of developing countries where reasons for public/ private school considerations may not be borne out of social capital but instead arise out of frustration with quality of public schooling and efficiency issues. (Glewwe and Patrinos 1999) While early studies have found better performance of private school children than those from public schools (Jimenez, Lockheed et al. 1991, Jimenez and Lockheed 1995). It is difficult to peter out reasons for greater effectiveness of private schools, that is, whether the higher educational outcome is a result of private schooling or the higher educational outcome is the result of characteristics of parents and home that may determine the kind of children joining the school itself. (Hanushek 1997). The effect of parents’ socioeconomic status and value placed on importance of education may lead to biased conclusion between private schools and educational outcomes. Randomly assigning children to public and private schools and comparing their learning outcomes may correct for this selection bias. Voucher experiments in Colombia and Chile provide interesting examples. Colombia began experimenting with school vouchers in 1991 and provided vouchers to students entering Grade 6 by randomly assigned lottery. This allows for a comparison of lottery winners and losers and the comparison indicates that the winners have lower dropout rate and somewhat higher tests scores than losers (King, Bettinger et al. 2002). However, while random assignment controls for the endogeneity of school choice, it is difficult to use this experiment to conclude that private schooling increases educational attainment. Since students were at a risk of losing vouchers for poor performance, participation in voucher program may increase student motivation to work hard. The effect of better school inputs may be inseparable from the effect of higher student motivation (Bettinger, 2005). Similar voucher experiment in Chile for Grade 4 children showed slight (Five studies) to little (Four studies) difference between the private and municipal school students (Out of total ten studies). In fact the latter performed better in one of the studies (Bellei, 2008). The review concludes, therefore that private school admissions are selective poorly performing student can be easily expelled, so the slight advantage in scores for private school students could easily be due to selectivity.

• 2. Research on Public and Private Schools in India

In recent years the past neglect on Indian school education system has been partially

• addressed. There has been steady increase in terms of school enrollment – both for

5 girls and boys. The Right to Education Act (2010) along with all- India Sarva Shiksha Abhiyan greatly improved school facilities across the country. Nevertheless the system continues to be plagued with two major deficiencies – first limitation of coverage and secondly, poor standards of the education offered and received (Drèze and Sen 2013). In comparison with extensive research in other countries, the field remains in infancy in India. Studies have noted that government schools are more expensive than private schools with lower accountability. The problem of teacher absenteeism, reluctance to teach is acute and pervasive in the sector (Drèze and Sen 2013). Studies found considerable absenteeism among schoolteachers in rural areas (ranging from 11 to 25 percent) and found that private school teachers are 2–4 percentage points more likely to be present in school than government schoolteachers (Kremer, Chaudhury et al. 2005). The irony is greater since studies also found government schools to be far more expensive than private schools. The difference explained by lower teacher salaries in private schools as compared to government schools. The Pay Commission scales have resulted in relatively high salaries of schoolteachers with little evidence to suggest that higher salaries are of particular help in raising teaching standards (Muralidharan 2012). Higher salaries may be instrumental in expanding the pool of applicants by transforming the job into plum posts that may attract candidates with requisite qualifications- even if they do not have interest in the profession per se. The bigger downside is the social distance (Drèze and Sen 2013) that is created between teachers and parents as a result of the higher pay fixation. The increasing gap has implications

• n those whose wages may not be determined by Pay Commission and may hamper

mutual cooperation between teachers and parents from most rural families, which could be very important for success of school education. Study in Delhi found that on average, the full-time teachers teaching Grade 4 in government schools earned Rs 10,071 per month compared to Rs 3,627 in private recognized schools and Rs 1,360 in private unrecognized schools (Tooley and Dixon 2005). Research on educational standards is fairly limited for the Indian case. The available information records higher performance on the part of students from private schools than from government schools. The overall levels however, remain depressingly

• abysmal. The Pratham ASER Survey, 2011, a large all India representative survey of

schoolchildren in rural areas found that only 58 percent of children enrolled in classes 3 to 5 could read class 1 text and less than half 47 percent were able to accurately perform simple two-digit subtraction (Pratham, 2011). Similar results are shown by a study in Delhi slums (Tooley and Dixon 2005). The picture becomes starker when international comparisons are made with the limited information available. When two

• f the better-schooled states in the country – Tamil Nadu and Himachal Pradesh were

tested for reading abilities of 15 year old students in the PISA plus survey in 2009, both the states figured in the bottom three of the list among total 74 countries. (Walker 2011). These studies while giving a general idea of the massive public policy problem are not exhaustive to the extent of controlling and studying the

6 socioeconomic differences among the children studying in government and private schools.

• 3. Private Schooling and Public Policy Debates

Privatization and public- private partnerships have been suggested as elixir for school education problems by many commentators. This is in part a result of increasing dissatisfaction with the quality of public schooling and the assumption that private school system is somehow more efficient than the school system. Increased private intervention and public private partnerships have been advocated for in form of state provision of vouchers for private schools in India and elsewhere. (Kelkar 2006, Muralidharan 2006, Chakrabarti and Peterson 2008, Panagariya 2008). The case for public private partnership in early education depends on the following assumptions- efficiency and cost effectiveness without trading off quality of education, government financing of privately delivered education may address some of the social inequalities in access to private institutions and increased government funding of private schooling and education would not damage public education. While private schools offer an alternative, they cannot become an organic system by themselves to substitute the public system, especially if the transformative role played by the latter in international development experience is considered. (Drèze and Sen 2013). First there is the question of affordability, since a large part of fees are linked to the profitability of the private schools. Large sections of children from disadvantaged backgrounds are biased against as a result of these high fees. Even if the affordability problem is removed (by voucher system or private aided schools) information asymmetry and imperfect competition at the school authority level pose significant challenges. Recent examples of private schools in Indian education offer examples of both dedicated successes and money grabbing enterprises with a reputation, for offerings not matched by performance (Drèze and Sen 2013). The greatest penalty that increased reliance on private schools imposes is the gradual shift of children and parents away from state schools who may contribute the most in making those schools accountable and more responsible. Research on school and neighborhood effects suggests that the social and economic composition of student population in schools has an impact on school functioning(Jencks and Mayer 1990) and accountability as well as attitudes and aspirations of peers (Pong 1998, Roscigno 2000, Goddard 2003). Therefore dependence on private schools can exacerbate the problems in public schooling system by providing a way out for the more prosperous vocal families who suffer low quality education in state sector (Hirschman 1970). Further, parents who are able and willing to send their children to private schools generally tend to be highly educated themselves and value educational attainment. Therefore it may be difficult to say whether it is private school enrollment that causes

• bserved differences in skill.

The fact that private schools in India do not seem to be doing much better than

7 government schools in terms of average pupil achievements(Drèze and Sen 2013) calls for a careful evaluation of the proposal. Evidence needs to be carefully looked into before taking a stand purely on idealistic basis. The educational conditions on ground need to be carefully modeled to provide a possible answer to whether private schools can be effectively utilized as viable alternative to public education. We aim to answer some of these questions with this paper.

• 3. Data and Methodological issues

3.1 Data Description India Human Development Survey I (2004-5) and II (2011-12) form part of a collaborative research program between researchers from the National Council of Applied Economic Research (NCAER) and University of Maryland. The goal of this program was to document changes in daily lives of Indian households in a society undergoing rapid transition. In documenting the way they live, work, educate their children, care for their aged parents, and deal with ill health, the researchers seek to infuse the development discourse with the lived experiences of ordinary people. Bulk

• f the funding for IHDS-II was provided by the U.S. National Institute of Child

Health and Human Development, with supplementary funding by The Ford

• Foundation. IHDS- I was funded by a grant from the U.S. National Institute of Health

and builds on a prior survey by NCAER. The first round–IHDS-I conducted in 2005 composed of a nationally representative survey of 41,554 urban and rural households. It covered all states and union territories

• f India –with the exception of Andaman/Nicobar and Lakshadweep. The households

were spread across 33 states and union territories, 384 districts, 1503 villages and 971 urban blocks located in 276 towns and cities. The size of the urban sample was 14,544. India Human Development Survey-II (IHDS-II), the second round of the survey in 2011-12 was also nationally representative, multi-topic survey of 42,152 households in 1,503 villages and 971 urban neighborhoods across India. These data were mostly re-interviews of households (85 per cent) interviewed for IHDS-I in 2004-05. Two one-hour interviews in each household covered topics concerning health, education, employment, economic status, marriage, fertility, gender relations, social capital, village infrastructure, wage levels, and panchayat composition. A major innovation of this survey was to conduct short assessments of reading, writing, and arithmetic skills for children aged 8–11 years. Conducting educational assessment in developing countries particularly India is difficult for a variety of reasons: children’s abilities vary tremendously and an instrument must capture children at both ends of the distribution; tests must be translated into many different languages with similar difficulty levels; the instrument must be simple and intuitive so that interviewers can administer it easily and it would not frighten children who are not used to standardized tests.

8 To aid in smooth administration of learning tests and experienced inputs, the survey team was able to work with Pratham, a non-governmental organization that has worked in the field of elementary education for many years. They have developed simple assessment tools to measure the effectiveness of their training programs and have administered these tools to over 250,000 children in their nationwide survey reported in the Annual Status of Education Report 2005 (Pratham, 2005). This test was included in the IHDS and allows us to measure whether a child is not able to read at all, or is able to read letters, words, sentences, paragraphs, or stories. Simple addition, subtraction, multiplication, and division problems were also developed. The English version of the test is reproduced in the appendix 3. Interviewers were trained extensively by Pratham volunteers using specially developed films so that they could differentiate between a child’s shyness and inability to read. They were also taught how to develop rapport with children. Tests were developed in twelve Indian languages as well as English, and children were asked to take the test in whichever language they were most comfortable in. In all the IHDS sample consists of 14702 (IHDS-II) and 17, 117 children (IHDS-I) aged 8–11 years. Reading and arithmetic tests were administered to 11693 almost 80 percent (IHDS-II) and 72 percent (IHDS-I) of the children aged 8–11 years. Children may not be tested for two reasons: (a) interviewers were explicitly instructed to obtain parental consent as well as assent from children for testing and were asked not to pressurize children who were reluctant and (b) since the household survey was the main focus of this study, the administration of the reading and arithmetic skills was left to the end. We suspect that household fatigue as well as interviewer fatigue may have played a role in missing skill testing. Appendix table A-1 in appendix 2 shows the proportion of children tested by a variety of household and background factors. The results suggest that children who are currently not enrolled are the least likely to be tested. Beyond this, while there is a mild difference in test completion rate between different social and economic groups, this difference is not large. There is little difference in test completion for children in private and government schools. While instruments for test completion are difficult to find, a Heckman selectivity correction relying on probit-linear regression combination was not statistically significant nor did it change any other coefficients substantially. The test data we have available to us were quite unique, particularly since they were combined with a wealth of household and contextual characteristics. Children are classified according to their ability to read in one of the five categories:

• 1. Cannot read at all.
• 2. Can read letters but not form words.
• 3. Can put letters together to read words but not read whole sentences.

9

• 4. Can read a short paragraph for 2–3 sentences but not fluent enough to read a whole

page.

• 5. Can read a one-page short story.

Excluding cases with missing data on independent variables as well as non-enrolled students, the analytic sample for reading skills consists of 11,702 children for the first round and 11693 for the second round. Children’s mathematical skills are classified in four categories:

• 1. Cannot read numbers above 10.
• 2. Can read numbers between 10 and 99 but not able to do more complex number

manipulation.

• 3. Can subtract a two-digit number from another.
• 4. Can divide a number between 100 and 999 by another number between 1 and 9.

We focused on 2-digit numbers to avoid calculations on fingertips and to get a better estimate of true understanding of subtraction and division. Also, given the Indian system of expecting children to memorize multiplication tables from 1 to 20, we chose to test children on division rather than multiplication skills. Excluding cases with missing data on independent variables as well as non-enrolled students, the analytic sample for reading skills consists of 11,655 and 11644 children for each of the rounds respectively. In addition to the household module, the survey also included a primary school module where the interviewers were asked to conduct a school facilities survey for

• ne public and one private primary school in each village and urban block. When

more than one facility was available in each block/ village, interviewers were asked to select the facility that was predominantly used by the residents. The school facilities survey provides an interesting description of the schooling climate in India. However, given the differential use of private and public schooling in different parts of India, the results from this survey should be treated as being indicative of the schooling climate around different parts of India rather than providing a representative sample

• f primary schools. However, this survey provides us with some interesting

exclusions restrictions to handle the endogeneity of choice of private schools. 3.2 Methodology The primary goal of this paper is to examine the relationship and analyze the changes between enrollment in private schools and academic skills for children aged 8–11 years for the two rounds of surveys in 2005 and 2011-12. In view of some of the

10 methodological considerations outlined above, we rely on a variety of techniques to

• btain a sense of the magnitude of this effect. Specifically, we examine the impact of

private school enrollment on children’s verbal and mathematical skills using ordinary least squares regression, Heckman control function method based on exclusion restrictions (Heckman and Navarro-Lozano 2004), and family fixed-effects models. Triangulation based on these three methods allows us to develop a range of estimates for the impact of private school enrollment on children’s skills. The Heckman control function method assumes that the underlying model is: Yi =𝛾𝑌! + 𝜀𝑎! + 𝜁! Yi being the child’s score on reading and arithmetic tests, Zi indicates private school enrollment and Xi controls for host of background characteristics : state of residence, urban/rural residence, caste/ tribe religious background of the parents, child’s age, sex, highest level of education of parents, household size, log of annual household income and households score on index of possession of a variety of consumer

• durables. The switching regression was determined through Wi as instruments that

affect private school enrollment: presence of private school in the village. Whether English was taught early on, presence of cook and households social networks. Zi is taken to be a result of unobservable latent variables: 𝑎! = 𝛿𝑋

! + 𝜈!

Therefore the decision to send a child to private school or not is made according to the rule: 𝑎! = 1 𝑗𝑔 𝑎! > 0, 0 𝑗𝑔 𝑎! ≤ 0} These equations were estimated in STATA using the TREATREG routine with full maximum likelihood. Instruments used in identifying the selection equation are discussed along with the characteristics of private and government schools in India

• below. Due to the reliance on probit-linear combination, the dependent variables—

reading and arithmetic skills—are assumed to be continuous variables for this analysis. Since results from this method are highly sensitive to the choice of exclusion restrictions (Stolzenberg and Relles 1997), we supplement this analysis with a highly restrictive family fixed-effects model. Impact of private schooling on children is riddled with concerns about the fact that families that choose private schools are different from those that choose government schools and any observed relationship between private schooling and child outcomes could be due to these unobserved

• factors. One way of addressing this is to compare the achievements of children within

the same family based on whether they go to private school or not, that is, adding a dummy variable per household. We supplement the analysis using Heckman control

11 function method with family fixed-effects models to give us another estimate of school effect.

• 4. Growth of Private School enrollment in India

The Indian education system is composed of variety of schools across spectrum and the line of demarcation may not be easy to navigate. While government sector is easily identifiable through schools run by Central, state and local governments, the private sector consists three types of schools. 1) Schools that receive government grant-in-aid but are privately run. 2) Schools that receive little government funding but are recognized based on certain criterion outlined by the government and must follow certain regulations. 3) Schools that are unrecognized and might not meet the criteria (such as infrastructure or teacher salaries) needed for recognition. Aided schools are private schools that that receive grants-in-aid resembled private schools in early decades following Independence. They received money from the government but teachers were directly hired and paid by the schools. Since the 1970s, these teachers receive their salary directly from the state and are recruited by a government appointed commission but their routine operations are governed by the private management (Chakrabarti and Peterson 2008). Therefore, in cost and teacher qualification requirements they resemble government schools but retain a private character management style. Private recognized schools are generally located in urban areas, have better infrastructure and tend to be larger. They are often run by non-profit management. At the other end of the spectrum are the unrecognized private schools frequently run like homegrown businesses in an ad-hoc fashion, sometimes in the back of teachers home. The private school enrollment in India has been rapidly rising from 20- 24 percent of rural students being reportedly enrolled in private schools in 2005 to about 30.5 percent in 2016 (Pratham, 2016). Focus on primary education by the Indian government for the past many decades has achieved massive improvements in literacy and enrollment rate. Continued emphasis has resulted in raising the literacy rates to about 74 percent from about 18 percent in 1951 (Census, 2011). Similarly, near universal enrollment rates (96.9 percent, 2016) have been achieved from a meager average of about 30 percent in 1951. The dismal state of education inherited at the time of independence has therefore been partially corrected by according priority to the sector. Successive five year plans have emphasized the importance of investing in primary schooling with a plethora of government programs (Govinda 2002). Official data from the Seventh All India Survey of Education show that the share of private schools in primary enrollment is about 6 percent in rural areas and about 29 percent in

12 urban areas. However, there are good reasons to believe that this is a substantial underestimate (Kingdon 2007). Official statistics do not usually collect data on unrecognized schools and consequently tend to underestimate the size of the private sector (Kingdon 2007). The 1993–94 household survey by NCAER (Shariff 1999) found that about 10 percent of the primary school students in rural India were in private school while the comparable figures from the Sixth All India Survey by National Council for Educational Research and Training (NCERT) conducted in 1993 recorded only about 3 percent in private unaided schools. The 2002 Seventh All India Educational Survey conducted by the NCERT found 5.8 percent enrollment in private (unaided) schools in rural areas and 28.8 percent in urban areas. If aided private schools are included, this number swells to 9 and 45 percent in rural and urban areas respectively. However, household based surveys—which include both recognized and unrecognized schools document a higher prevalence. Consequently, the Annual Status of Education Report (ASER) survey conducted by Pratham in 2005 (Pratham, 2005) and confined to rural areas, found that private school enrollment for rural children was nearly 20 percent. Successive ASER reports (2006 to 2016) on rural areas by Pratham have documented an increasing trend towards private school enrolment for the country as a whole. The pan national survey noted a rise in enrollment from about 19.66 percent in 2006 to about 28 percent in 2011 to about 30 percent in 2016.

Source: Authors calculation based on Annual Status of Education Reports 2006 – 2016

The India Human Development Survey 2005 and 2011-12 (IHDS-I and IHDS- II) documents similar enrollments. Table 1 shows that at the all India Level, about 68 percent of children were enrolled in government schools with 42 percent and 76 percent of the urban and rural students respectively in government schools in 2005. The numbers show massive changes in the second round with a jump in total private enrollment to 40 percent and a secular fall government enrollment to 61 percent. The rise in private enrollment is across spatial difference- that is, in both rural and urban areas combined enrollment in aided and unaided private schools, madrasas and

0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 2006 2007 2008 2009 2010 2011 2012 2013 2014 2016

Enrollment in Private Schools 2006 - 2016

Pvt

13 convents rose from 24 percent and 58 percent to 30 percent and 62 percent respectively, among children of age 6- 14 years. Private schools has been used a broad umbrella term for aided and unaided schools as parents may often not know the exact management of the schools their children

• attend. Moreover, the only difference between private aided and private recognized

may be the source of payment of teachers – the former being paid by the government and the latter by the management. (Chopra and Jeffery 2005). As Figure 1 indicates for round one of the survey, private enrollment rose for higher standards and the proportion of private schools was substantial for primary levels. When similar calculation is done for IHDS-II in Figure 1b interesting changes are

• bserved, enrollment picks up for lower primary levels falls for middle school and

again marginally picks up for higher levels. Overall private enrollment continued to remain high for both the periods. High enrollment values for the second round may point to documented belief among parents about benefits of early private schooling and an implicit assumption that private schools may offer better quality of education as compared to government schools at that age. This choice though, comes at a cost, Figure 2 and Figure 2b show average educational costs for private and public schools by current standard. The average primary student in a private school paid Rs. 600 in fees and Rs. 600 in expenses for book uniforms and transportation in the first round, the costs have fallen for the second round to an

Table 1. Distribution of Type of Schools Attended for Enrolled Children 6-14

Source: Authors calculation based on India Human Development Survey (IHDS – I and IHDS- II) 2005, 2012.

about Rs. 360 in fees and Rs. 235 other expenditures. The spending nevertheless remains high as compared to government schools Rs20 and Rs. 200 (Round-I) and Rs. 17, Rs 60 (IHDS-II) for each of the heads. Further only 17 percent and 19 percent government school students reported to take private tutoring for each round compared to 27 percent and 29 percent for their counterparts in private schools.

Rural Urban All School Type IHDS-1 IHDS-II IHDS-I IHDS-II IHDS-I IHDS-II Government 76 70 42 39 68 61 Education Gurantee Scheme (EGS) 1 1 1 0.36 1 0.56 Government 75 69 41 38 67 60 Private 24 30 58 62 32 40 Private Aided 4 3 8 6 5 4 Private 17 26 45 53 24 34 Convent 1 0.5 3 2 2 1 Madrasa 1 0.8 1 1 1 0.58 Other 1 0.04 2 0.12 1 0.06 Sample Size 24949 23218 11776 10431 36725 33649

14 Characterstics of Public and Private Schools in India Parents generally list out two reasons for sending their children to private schools – the inferior qulaity of government schools- “Government schools are not good around here; teacher are often absent and do not work hard even when present” and the wish for their children to learn english – “ We want our children to learn english and the private schools are english medium or teach English earlier than government schools.” The parents observations had good empirical support in the first round, the second round statistics however bring out a changed picture in terms of reported facilites at government schools. This may be instrumental in steps to address some concerns regarding government school characterstics, though major concern still enjoy empirical support for the two rounds. Table 2 indicates, the school facilities survey in IHDS-I and II found about greater absenteesim on the day of the survey (12.4 percent in round –I and 17 percent in II ). Private school teachers were more likely to be prsent in school for both the rounds. It is interesting to note that Government Schools has higher trained teachers but lower teachers with college degree on its payroll. This may be a result of the fact that government school employment is based on formal training certificate. Government Schools show major improvements in terms of

Table 2. Characterstics of Private and Public Schools in India

Source: Authors calculation based on India Human Development Survey (IHDS – I and IHDS- II) 2005, 2012.

facilites like Toilets, Computer availibility, Desks and fans for children though they stilllag behind private institutions. Substantial differences between the two schooling systems in terms of provisioning of mid day meal. After Tamil Nadu introduced a successful mid-day meals program in its schools, the National Program of Nutritional Support to Primary Education was

Govt Govt IHDS 1 Private Private IHDS1

Percentage Of Teachers Present In School 82.93 87.6 87.45 89.4 Percentage Teachers Trained 87.96 85.9 52.93 43.8 Percentage Teachers College Degree 59.51 43.7 70.36 64.4 Percentage Students Present in School 85.4 86.9 90.08 91.9 Some Subjects Taught In Eng+ 32.79 26.8 64.63 51.1 Eng Instruction Begins In 1st Standard 83.79 53.2 97.34 88.2 No Classes Meeting Outside 0.20 0.7 0.08 0.3 No Of Mixed Grade Classrooms 0.93 0.9 0.96 0.6 Any Toilet Facility 91.10 60.9 93.55 78.3 Chairs Desk For All Students 37.94 29.2 71.77 63.5 Blackboard in all Classrooms 98.65 95.4 99.27 98.1 Computer Avlbl For Student Use 28.89 5.9 61.97 29.2 School Has Fans 55.12 28.4 75.07 63.3 Kitchen For Cooked Meals 74.50 41.3 12.9 10.8 Cook Employed By School 80.88 74.9 12.05 11.1 Any Teaching Material On Wall 89.08 77.3 85.75 78.9 Children’s work on the Work 85.66 67.6 83.6 73.9 Total 2248 2034 2011 1748

15 launched across India in 1995. The mid-day meals program (MDM) aims to increase primary school attendance, as well as improve the nutritional status of school

• children. Generally, the program serves the 6–11-year age group. However, some

upper-primary schools run the MDM program as well, and in recent union budgets, separate provision has been made for the upper-primary school also. Under the MDM scheme, cooked meals are to be served during the lunchtime in the school, with calorie value equivalent to 100 gm of wheat or rice per student per school day. In some places, a dry ration is provided to be carried home based on a certain minimum level of school attendance. The second round of IHDS reported a rise in mid day meal (MDM) provisioning for children up to Grade 5. From 60 percent in the first round to about 69 percent in the

• second. It is to the credit of the government that the MDM coverage has shown

stupendous improvement from about 80 percent to 98 percent in government schools while private schools still registered a measly increment of 4 percent from 8 percent to 12 percent. As has been pointed out that a well functioning MDM program may increase the likelihood that a child attends government school. One of the indicators

• f an active and running MDM program is the presence of cook in the school. (Dreze

and Goyal 2003). The more interesting case is narrowed gap between government and private schools in IHDS-II in terms of early instruction of English. Though private schools continue to lead in teaching English early on government schools have caught up massively from about 53 percent to 83 percent in the second round. While the percentage of children taught exclusively in English at government schools has shown modest increase from 2 percent to 3 percent, it is interesting to note the almost double rise to 42 percent from 26 percent in private schools. These figures are an indication that parental concerns and reasons for choosing to send their children to private schools are not unfounded claims. Private school fare better in terms of major school facility indicators, teacher absenteeism and English medium instruction. Government Schools seem to be closing in on some of these gaps in the second round and this may be an interesting avenue for future work on school choice studies on whether the tide is reversed back again in favor of public schools. Generally, private school students come in from selected subpopulation of higher socio economic standards. Their affluent backgrounds and higher priority to education if not controlled for is a potential source of bias for our work. Therefore, it will be important to control for this selectivity when examining the impact of private schools

• n student performance.

16 Characteristics of Private School Students Table 3 provides descriptive statistics for our sample, private school enrollment as well as children’s ability to read a simple paragraph and do basic two digit

• subtractions. As a result of sustained awareness and efforts, the school enrollment

rates increased to near universal rates to 96 percent (IHDS-II) from 92 percent (IHDS- I) for children aged 8 – 11 years; of these, about 41 percent are enrolled in private schools, a rise from the first round where about one third of the children were enrolled

Table 3. Sample Distribution, Private Schooling and Skill Levels by Background Characteristics

Proportion Proportion

• Prop. In
• Prop. In

Prop. Able to

• Prop. Able to
• Prop. Able

to

• Prop. Able to
• f Sample
• f Sample

(IHDS 1) Private School Private School (IHDS1 ) read a para. read a para.(IHDS 1) Subtract Subtract (IHDS 1 ) Gender Male 0.52 0.53 0.44 0.33 0.56 0.57 0.5 0.51 Female 0.48 0.47 0.36 0.29 0.53 0.54 0.46 0.46 Place of Residence Metropolitan 0.06 0.05 0.56 0.58 0.68 0.69 0.67 0.72 Other Urban 0.25 0.19 0.63 0.58 0.67 0.69 0.61 0.62 Developed Village 0.29 0.34 0.35 0.29 0.53 0.55 0.48 0.48 Less Developed Village 0.41 0.42 0.27 0.17 0.47 0.48 0.39 0.41 Household Income Quintile Poorest 0.23 0.18 0.24 0.16 0.45 0.45 0.36 0.38 Second 0.23 0.22 0.30 0.17 0.51 0.47 0.42 0.4 Third 0.21 0.22 0.38 0.26 0.56 0.51 0.5 0.45 Fourth 0.17 0.2 0.50 0.39 0.59 0.62 0.56 0.54 Affluent 0.15 0.18 0.68 0.59 0.7 0.73 0.68 0.69 Standard of Living Quintile (Asset Index) Poorest 0.26 0.2 0.16 0.1 0.34 0.34 0.26 0.29 Second 0.24 0.22 0.27 0.16 0.5 0.47 0.41 0.37 Third 0.2 0.24 0.39 0.27 0.6 0.54 0.54 0.49 Fourth 0.18 0.2 0.60 0.44 0.67 0.69 0.64 0.6 Affluent 0.12 0.15 0.81 0.69 0.82 0.81 0.78 0.78 Socio Religious Group Forward Caste 0.17 0.19 0.54 0.43 0.7 0.71 0.64 0.64 Other Backward Classes (OBC) 0.33 0.36 0.42 0.29 0.57 0.57 0.5 0.5 Dalit (Hindu, Sikh, Buddhist) 0.22 0.24 0.26 0.21 0.49 0.45 0.44 0.39 Adivasi (Any religion) 0.09 0.06 0.22 0.15 0.41 0.48 0.32 0.38 Muslim 0.16 0.13 0.41 0.38 0.47 0.46 0.41 0.42 Minority Religions 0.02 0.02 0.77 0.74 0.72 0.8 0.74 0.79

• Max. Adult Education in Household

Illiterate 0.2 0.24 0.20 0.16 0.32 0.37 0.27 0.31 1-4std 0.07 0.09 0.21 0.14 0.45 0.48 0.34 0.38 5-9std 0.36 0.35 0.32 0.26 0.53 0.55 0.45 0.47 10-11 std 0.14 0.14 0.47 0.45 0.64 0.66 0.58 0.61 High Sec & some College 0.11 0.08 0.58 0.53 0.69 0.72 0.65 0.66 College graduate 0.12 0.09 0.74 0.63 0.78 0.8 0.76 0.75

17 in private school. While enrollment may have risen across board inner biases remain or have spiked between the two rounds. In keeping with the preferential treatment and higher value

• f education placed on the boys in Indian families, boys are somewhat more likely to

be enrolled in private schools than girls. Private school enrollment seems clearly associated with higher income and education of the household. Interestingly, students in metro cities are less likely to enroll in private schools as students in smaller cities. This may be due to presence of better functioning and higher quality Central Government schools in major metropolitan areas. Caste and religion continue to be associated with private school enrollment. Forward castes and other minority groups such as Christians, Sikhs and Jains are far more likely to send their children to private schools than Dalits and Adivasis with Muslims and Other Backward Classes (OBC) falling in the middle. Similar discrepancy is

• bserved in their respective number for ability to read a paragraph or ability to

subtract with forwards castes and minority groups performing much better than the rest of the pyramid. State differences in private schools are interesting (Table 4). Private school enrollment in one of the high education states, Himachal Pradesh continues to remain low through the two periods while it persists to remain high Kerala, though it has fallen between the two periods. Higher private enrollment is also a feature of the North East. This may be due to the higher Christian population in the two states. They are more likely to be in convent schools. The lower rate over the two periods may indicate a shift due to better government schools in the state since the two states are also one of the better governed ones in the country. Uttar Pradesh, Rajasthan predictably have considerably higher private school enrollments. Some of these regional differences in private school enrollment may be associated with socioeconomic background of its residents but may also reflect some differences in state policies and historical factors.

18 Exclusion Restriction for Private School Enrollment The brief description of students in private schools as well as the literature cited earlier clearly suggest that private school enrollment is a choice variable and while we expect to control for observable family background factors such as education, income, and household size, these controls may be inadequate due to omitted variables as well

Table 4 . Private Schooling and Skill Levels by State Proportion in Proportion able To Private School Read A Paragraph Subtract All India IHDS-II 0.38 0.55 0.48 IHDS-I 0.31 0.55 0.49 Jammu and Kashmir IHDS-II 0.48 0.61 0.70 IHDS-I 0.46 0.41 0.61 Himachal Pradesh IHDS-II 0.28 0.74 0.67 IHDS-I 0.18 0.84 0.69 Utttarkhand IHDS-II 0.41 0.51 0.38 IHDS-I 0.34 0.63 0.47 Punjab IHDS-II 0.43 0.69 0.73 IHDS-I 0.52 0.67 0.73 Haryana IHDS-II 0.40 0.61 0.60 IHDS-I 0.33 0.66 0.63 Delhi IHDS-II 0.37 0.72 0.65 IHDS-I 0.31 0.77 0.72 Uttar Pradesh IHDS-II 0.56 0.48 0.35 IHDS-I 0.44 0.4 0.48 Bihar IHDS-II 0.19 0.37 0.37 IHDS-I 0.18 0.47 0.48 Jharkhand IHDS-II 0.40 0.46 0.40 IHDS-I 0.37 0.61 0.61 Rajasthan IHDS-II 0.47 0.58 0.48 IHDS-I 0.32 0.57 0.44 Chattisgarh IHDS-II 0.25 0.54 0.38 IHDS-I 0.19 0.62 0.37 Madhya Pardesh IHDS-II 0.29 0.52 0.33 IHDS-I 0.29 0.47 0.33 North East IHDS-II 0.42 0.55 0.70 IHDS-I 0.54 0.60 0.78 Assam IHDS-II 0.17 0.53 0.46 IHDS-I 0.09 0.75 0.46 West Bengal IHDS-II 0.16 0.66 0.61 IHDS-I 0.12 0.52 0.58 Orissa IHDS-II 0.17 0.57 0.50 IHDS-I 0.08 0.59 0.51 Gujarat IHDS-II 0.32 0.58 0.38 IHDS-I 0.20 0.65 0.43 Maharashtra/Goa IHDS-II 0.41 0.55 0.51 IHDS-I 0.29 0.66 0.54 Andhra Pradesh IHDS-II 0.43 0.43 0.61 IHDS-I 0.29 0.50 0.51 Karnataka IHDS-II 0.35 0.46 0.44 IHDS-I 0.27 0.53 0.55 Kerala IHDS-II 0.56 0.76 0.77 IHDS-I 0.61 0.82 0.60 Tamil Nadu IHDS-II 0.41 0.49 0.57 IHDS-I 0.42 0.80 0.72

19 as measurement error in some of the included variables. In order to estimate the Heckman control function discussed earlier, instead of relying simply on distributional assumptions, we rely on theoretically motivated exclusion variables that are expected to be associated with the decision to enroll in private school as well as private school admission but are not expected to be independently associated with educational outcomes. Availability of Private Schools Private school enrollment is dependent on a complex interplay of supply and demand. Social composition of an area, history, and state policies all play an important role in shaping the availability of private schools. Hence, availability of private schools is an important instrument for private school enrollment that has been used in the literature. We assume that in all urban areas private schools are available. Desirability of Public Schools Given the IHDS’s focus on school surveys, we also included a set of variables describing the characteristics of government schools in the village/ urban block as factors that may motivate parents to favor or not favor government schools. These include English medium instruction for some academic subjects, early introduction to English language, and presence of a cook in the government school as a marker for the draw of the mid-day meal program. Since school surveys for some localities were not conducted due to interviews taking place during weekends or holidays, a variable denoting missing school survey is included in the analysis. Parental Ability in Gaining Entrance in Private Schools Private school enrollment is not simply a function of parental preferences. In urban areas, admission into quality private schools can be a highly competitive process in which parents with broader social networks gain an edge over less connected parents. Consequently, we also control for two markers of family social networks, whether the household members know anyone working in the medical profession and whether they know anyone working for the government.3 These variables are described in Table 5.

3 Desai, S., Dubey, A., Vanneman, R., & Banerji, R. (2008). Private schooling in India: A new educational landscape. Brookings-NCAER India Policy Forum.

20

Table 5 : Sample Distribution, Private Schooling and Skill Levels by Instruments for Private Enrollment, IHDS-I and IHDS-II.

Over the two periods, availability of private primary school in the vicinity, cook in the local government school and early teaching of english in local govt school has shown vast improvement. Results for IHDS-II show that when private primary school is available in the locality, large number choose to opt for it even as their learning differentials are minor with government schools. This may point to the fact that mere presence of private school may not in itself gurantee higher learning per se. While switching regressions estimated with maximum likelihood are considered both unbiased and efficient, they are highly dependent on the validity of the exclusion criteria as well as their strength as predictors of private school. Table 6 shows the first stage regression with the exclusion variables listed above as predictors for IHDS-I and IHDS-II. The results show that with the exception of English medium instruction, each of the other variables is associated with private school enrollment and these relationships are statistically significant. Acquaintance with any government employee negatively impacts private school enrollment in our regression for IHDS-II, this seems reasonable as network with government employee may signal prevalence

• f expectation of higher quality from the institution and thus lower private school
• enrollment. Overall, the models are highly significant with 7 degrees of freedom.
• Prop. Of Sample

Prop in Private School Prop Able to read a para Prop able to subtract IHDS-I IHDS-II IHDS-I IHDS-II IHDS-I IHDS-II IHDS-I IHDS-II Know any medical personnel No 0.67 0.43 0.27 0.33 0.52 0.49 0.45 0.43 Yes 0.33 0.57 0.39 0.46 0.61 0.59 0.56 0.52 Know any Govt Workers No 0.68 0.73 0.26 0.37 0.51 0.52 0.45 0.44 Yes 0.32 0.27 0.41 0.5 0.64 0.63 0.58 0.58 Private Primary School in Village/town (all towns=yes) No 0.5 0.31 0.15 0.75 0.51 0.63 0.43 0.56 Yes 0.5 0.69 0.47 0.82 0.6 0.65 0.55 0.53 Local Govt. School has a Cook No 0.37 0.1 0.4 0.08 0.57 0.49 0.53 0.38 Yes 0.63 0.9 0.26 0.05 0.54 0.43 0.46 0.35 Local Govt. School teaches English in KG/Std 1 No 0.58 0.16 0.34 0.03 0.52 0.48 0.46 0.37 Yes 0.42 0.84 0.26 0.05 0.6 0.43 0.53 0.35 English as a Medium of Instr. In Local Govt. School No 0.83 0.73 0.31 0.05 0.54 0.43 0.47 0.34 Yes 0.17 0.27 0.28 0.07 0.63 0.45 0.56 0.41 School Survey Missing for Village/Block No 0.84 0.41 0.27 0.25 0.54 0.49 0.48 0.41 Yes 0.16 0.59 0.5 0.52 0.6 0.6 0.52 0.55

21

Table 6 Impact of Excluded Variables on Enrollment in Private Schools Results from the First Stage of Switching Regression Model IHDS-I and IHDS-II

• 5. Private School Enrollment and Child Outcomes

As the brief overview of literature presented above suggests, it is important to be cautious about drawing inferences based any perceived relationship between private school enrollment and children’s skill acquisition. Hence, in this section we first describe the basic relationship between private school enrollment and children’s performance on reading and arithmetic tests while controlling for observable characteristics of their households. Then we address the issue of endogeneity using a switching regression model in which school choice is captured by a set of theoretically motivated exclusion restrictions. Finally, we examine the impact of private school enrollment on child outcomes within a highly restrictive framework, family fixed-effects model. Figures 3 and 4 indicate basic differences in reading and arithmetic skills among children enrolled in government and private schools for the two periods. Results indicate that private school students have higher achievement on these tests. But over the two rounds the ability has shown a decrease or remained static. These differences are further analyzed by adding controls for parental socioeconomic background, place of residence, and children’s sex, age and current standard. In addition to private school enrollment, these regressions control for highest education level attained by any of the household adults, log of family income, a 30-item standard of living index consisting of owner- ship of various consumer durables (TV, refrigerator, telephone, car, cot, etc.) and quality of housing (toilet, piped water, etc.), household size, number of children under age 15, place of residence, state of residence, child’s sex, and age. Caste, ethnicity, and religion are particularly important to control for since they are linked to private school enrollment, especially enrollment in madrasas or convents, as well as having an independent impact on educational outcome. Controls for state of residence are also included in each

Coef. Zvalue IHDS-I IHDS-II IHDS-I IHDS-II Know anyone in Medical Profession 0.24** 0.172*** 5.6 5.56 Know anyone n Government 0.27**

• 0.1061***

6.61

• 3.05

Private schools available in Village 0.92** 0.5654*** 21.69 5.06 Cook in a local govt school

• 0.08*
• 1.064***
• 1.88
• 15.45

Early English in a local govt school

• 0.08*
• 0.8771 ***
• 1.94
• 13.04

Instructions in English in local govt school 0.07

• 0.0343

1.56

• 0.39

Missing school schedule 0.34**

• 0.2902 ***

5.29

• 7.47

Constant

• 1.18
• 1.906***
• 19.65
• 7.24
• No. of Cases

11667 11369

22 regression although not presented in the tables. In Model I, the basic OLS model for IHDS-II, students reading and arithmetic skills are regressed on a set of independent variables including enrollment in private school. As expected, parental education, household income and index measuring standard of living are positively related with student performance on the skill tests. Urban residence was found to be positively related to private school enrollment for the first round, it isn’t the case in the second round. This may be a result of documented rise in private school in the interior ‘rural’ parts of the country. These schools may be private by classification but since they are unregulated and mostly unaided, quality differentials with government schools may be negligible. Standard of living—a marker of long-term economic status—is consistently statistically significant; log of household income is also negatively significant. This dichotomy maybe a result of the year on year fluctuations in income imposing constraints on learning. Permanent income, indicated by standard of living indicates has impacts well being in the long term and is as expected positively significant. While it is reasonable to see skills increase with age and current standard, the coefficient on sex is surprising. Holding age and current standard constant, girls have lower performance on both reading and arithmetic tests in round-I and lower arithmetic performance in round II, possibly due to greater demands of household chores compete with time spent doing homework. In international studies, girls generally perform slightly above boys in verbal tests and slightly below boys in mathematical tests. Enrollment in a private school is positively related with higher performance on both verbal and mathematical skills in both the rounds. While the coefficient for verbal skills is slightly larger, it is important to remember that the skill levels range from 0 to 4 for the verbal skill and from 0 to 3 for mathematical skills. The second model corrects for the endogeneity of school choice by using a Heckman type correction, in which the binary choice of attending private school or not is modeled with the set of exclusions restrictions described above. The results from this endogenous switching regime are presented in Model 2. The first stage probit model (presented in table 6) suggests that our instruments are highly correlated with private school enrollment in round-II as well. The second stage regression includes the effect

• f private school enrollment on reading and arithmetic skills, correcting for the biases

introduced due to endogeneity of school choice. The Wald test for independence of regressions was not statistically significant suggesting a possibility that the selection equation and achievement equation are unrelated cannot be ruled out for the first round results. For the second round, the regression coefficient for private school from the uncorrected model for reading skill is about 0.19 while in the model correcting for endogeneity it is about 0.47. That is, the omission of endogenous nature of school choice introduces not a very large bias in the regression estimate for the first round- the regression coefficient for private schools is 0.39 for the uncorrected model for reading skills and 0.36 for the model corrected for endogeneity. The difference for

23 arithmetic skills is similar in magnitude for round-I, 0.28 vs 0.22 and for round – II the difference is much less than as compared to reading skills – 0.21 and 0.22. Results from any models relying on instrumental variables are only as good as the instruments themselves. Hence, we compare these results with those from a strongly restrictive model—family-level fixed-effects model. Here we assume that all family influences such as desire for education and parental encouragement are shared by all children in the family. Children differ mainly in their personal characteristics such as gender, age, standard, and private school enrollment. These family-level fixed-effects models continue to suggest that private school enrollment is consistently related to higher performance and the magnitude of these coefficients is similar to those

• btained from the switching regression in round-I with minor differences for round-II.

These results from both rounds suggest three things:

• 1. Private School enrollment is associated with higher child outcomes, even after

controlling for variety of family factors.

• 2. While the coefficients are not noticeable different from each other for the first

round, the same cannot be said for the second round. The coefficient for the model controlling for endogeneity is vastly different albeit in the same direction for reading skills and exhibits minor differences for arithmetic skills. Certain biases that may have affected our results need to be discussed at length for better interpretation of the numbers here. These may confound the association between private school enrollment and children’s educational outcome in a manner that would make simple analysis of the relationship erroneous, if ignored. Within Family Choices Parents when faced with spending scarce resources on children’s education may choose to send an academically gifted child to private school. Hence, in within-family fixed-effects models, any association between private schools and child scores may be due to children’s ability rather than their school. The only way of addressing this would be via longitudinal data in which one would try to examine the differential growth in educational achievement between children in private and government schools, holding their initial talent constant. This may be particularly important because studies have also found that at times educational innovations or programs have a large initial impact, with declines in magnitude over time (Banerjee et. al, 2007) Differential Value Placed on Education between Families Some families value education more than others and may be more likely to invest in by sending children to private school and ensuring that they do their homework. While we have tried to control for these differences using switching regression, some

24

• f the variables in the model such as having greater access to social networks may not

be fully exogenous. In particular, household with greater social connections may have a greater ability to get their children into private schools (as we argue), and at the same time, may have greater returns to education in the form of better access to jobs. Differential Demand for Education across Communities Some of our exclusion restrictions rely on village level access to private schools and characteristics of public schools. It is possible that communities may differ in their demand for schools and certain types of education such as early instruction in

• English. Hence, it may be higher demand for high quality education that may lead to

better outcomes rather than access to private schools. While this seems a more remote possibility—it is difficult for parents and communities to change government school curriculum and ensure early English instruction—it is not impossible. However, we have used a variety of techniques and excluded variables with the expectation that while each may retain some sources of bias, together they provide us with a rough indication of whether private school enrollment might be associated with higher performance or not. Our results suggest remarkable similarity of effects across the three models. It is possible that some of these effects are overestimated; particularly, the within-family fixed effect may decline if children’s ability is taken into account. However, if the results we present suggest an upper bound for the impact of private school education, the estimated effects are no more than one-third to

• ne-fourth of a standard deviation. As we discuss below, in comparison to inter-state

differences in educational outcomes, these are modest effect

Table 7a: Impact of Private School Enrollment on Reading and Arithmetic Skills, IHDS-II

Reading Skills Arithmetic Skills 1 2 3 1 2 3 Basic OLS Switching Regression Family Fixed Efect Basic OLS Switching Regression Family Fixed Effect Residence (metro omitted)

• ther urban

0.0483 0.0398

• 0.0584
• 0.0597*

developed village

• 0.0492
• 0.0458
• 0.10007**
• 0.0995 **

less developed village 0.0091 0.1128

• 0.0990 **
• 0.0987**

Socio Religious Group(forward caste omitted) OBC

• 0.01226
• 0.01338
• 0.0143
• 0.0144

Dalit

• 0.1589***
• 0.1495 ***
• 0.0632 **
• 0.0618**

Adivasi

• 0.2200 ***
• 0.2122 ***
• 0.1604 ***
• 0.1592***

Muslim

• 0.1516 ***
• 0.1525***
• 0.07836 **
• 0.0784**

Other Minority religion

• 0.1412 *
• 0.1524 *

0.0268 0.02516 Maximum Household Education(none omitted) 1-4 standard 0.2145*** 0.2156*** 0.01822 0.0184 5- 9 standard 0.2649*** .2613*** 0.1298*** 0.1292*** 10-11 standard 0.3599 *** .3513*** 0.2128*** 0.2115*** High Secondary & some college 0.4848*** .4716*** 0.3106*** 0.3086*** College graduate 0.5519*** .5331*** 0.3986*** 0.3958*** Log of household income

• 0.0358 **
• 0.003821 **

0.02646** 0.0260** Score on Std. of living scale 0.1453*** .1348*** 0.1028*** 0.1012***

• No. of persons in Household
• 0.0069 *
• 0.0055
• 0.01244 **
• 0.01223**

26

Current standard 0.2760 *** 0.2826*** 0.3413*** 0.1487*** 0.1497*** 0.2035*** Age of the child 0.0516*** 0.04685*** 0.00749 0.04990*** 0.04915*** 0.0144* In private school 0.1892*** .4695*** .3622*** 0.2108*** 0.2247*** .2505*** Constant

• 0.534 *
• 0.1293
• 0.8077***
• 0.5645***
• 0.5603***
• 0.6447***

R squared 0.2202 0.1795 0.224 0.1603 Chi- Square(42 d.f) 5016 6182 Observations 11369 11369 11369 11324 11324 11324

27 Table 7b: Impact of Private School Enrollment on Reading and Arithmetic Skills, IHDS-I

Reading Skills Arithmetic Skills 1 2 3 1 2 3 Basic OLS Switching Regression Family Fixed Efect Basic OLS Switching Regression Family Fixed Effect Residence (metro omitted)

• ther urban

0.163*** 0.161*** 0.0112** 0.108** developed village 0.179*** 0.171** 0.092* 0.078 less developed village 0.176*** 0.167** 0.101** 0.082 Socio Religious Group(forward caste omitted) OBC

• 0.051
• 0.051
• 0.054*
• 0.055*

Dalit

• 0.222***
• 0.222***
• 0.222***
• 0.222***

Adivasi

• 0.104*
• 0.104*
• 0.124***
• 0.125***

Muslim

• 0.231 ***
• 0.231***
• 0.241***
• 0.242**

Other Minority religion

• 0.101
• 0.102
• 0.0602
• 0.062

Maximum Household Education(none omitted) 1-4 standard 0.147*** 0.147*** 0.037 0.0038 5- 9 standard 0.186*** .187*** 0.110*** 0.111*** 10-11 standard 0.338 *** .338*** 0.252*** 0.253*** High Secondary & some college 0.387*** .389*** 0.302*** 0.305*** College graduate 0.417*** .419*** 0.388*** 0.390*** Log of household income 0.001

• 0.001

0.006** 0.006** Score on Std. of living scale 0.034*** .035*** 0.031*** 0.031***

• No. of persons in Household
• 0.0237 *
• 0.024***
• 0.019***
• 0.019***
• No. of children <15 in household
• 0.00504
• 0.005

0.004 0.003

28

Female Child

• 0.100***
• 0.100***
• 0.07
• 0.157***
• 0.156***
• 0.179***

Current standard 0.341*** 0.341*** 0.229*** 0.247*** 0.247*** 0.183*** Age of the child 0.025*** 0.025*** 0.164*** 0.037*** 0.037*** 0.123* In private school 0.392*** .362** .307*** 0.280*** 0.221*** .224*** Constant 0.497** 0.513** 1.482*** 0.148*** 0.179*** 0.879*** R squared 0.337 .286 0.355 0.287 Chi- Square(42 d.f) 3954 4782 Observations 11667 11667 11667 11619 11619 11619

29

• 7. Which Children Benefit the Most from Private School

Enrollment?

The discussion about effects and extent of private school has occupied public policy debates in such a way that there has been little or no room to look into potential beneficiaries of private schooling. We begin our analysis by exploring interaction between parental economic status and choice of school to explain the mechanism through which private schools may influence child outcomes. Research from the United States has concluded disproportionate advantages for children from relatively disadvantaged backgrounds (Hoffer, Greeley et al. 1985). To examine this, we interacted private school enrollment with household standard of living in Model 2 from Table 7a, that is, Heckman switching Regression. The standard of living index is the same 30-item Household asset index used for the first round. The private school enrollment was interacted with this index for both the rounds while controlling for selection into private schools using the instruments discussed above.

Table 8: Interaction Effect of Standard of Living and Private School Enrollment on Children’s Reading and Arithmetic Skills, Likelihood of Being Praised and Being Beaten

The interaction terms are significant and negative in sign for both the rounds for reading levels. The results have been presented in table 8 and graphically represented in figures 6, 6b and 7, 7b which suggest that for both the two rounds benefits to private school children from lower economic strata are far greater than those for children from upper economic strata. At higher standard of living levels, the difference between private and government schools narrowed considerably. The narrowed gap may be a result of the fact that parents with greater means would choose to send their children to government schools only if it is of high quality. The attempt to understand mechanisms of differed experiences of students according to the kind of school attended is still in its nascent stage for developing countries. In the following analysis we attempt to provide some qualitative information on experiences of children in government and private schools. This part is more suggestive in nature since it may be difficult to determine the casual direction of

• association. Nonetheless the two rounds of the surveys provide us with the only data

here even associations can be explored and analyzed. The IHDS – II and I interviewed parents about the schooling experiences of upto two children in the household. Two variables are of particular interest to us here- a) Whether the parent reported that the child was praised in the month preceding the

Reading Arithmetic Praised Beaten IHDS-I IHDS-II IHDS-I IHDS-II IHDS-I IHDS-II IHDS-I IHDS-II Standard of Living 0.0043*** 0.0663** 0.035*** 0.0736*** 0.022*** 0.0446***

• 0.013**
• 0.0328***

Private School Enrollment 0654*** 0.2472*** 0.364*** 0.1752** 0.628*** 0.2108 ***

• 0.123

0.7193*** Private*standard of Living

• 0.023 ***
• 0.0082**
• 0.012***
• 0.0041
• 0.006

0.0033 0.016**

• 0.0240***

30 survey and b) whether the parent reported that the child was physically beaten or pinched in the month preceding the survey. Both these variables give us broad qualitative approximation of private and government schools. These qualitative factors are particularly relevant in the Indian context where constant comparisons and humiliation are fairly common. Moreover, positive reinforcements have been found to be important indictors of learning environment. Here we seek to explore the role social class plays in the differentiation between private and government schooling system. While private school students fared better than their government counterparts for both the variables in the first round - about 25 percent of the government school students were praised compared to 42 percent in private schools and about 29 percent of the government school students were beaten compared to 25 percent in private schools. Things were not so simple for the second round – about 35 percent of the government school students were praised compared to 56 percent of the private school and about 26 percent reported to have been beaten compared to 29 percent in private school. This may indicate an ambiguous nature of much touted higher quality private schools in the country. The higher percentage may be an indication of lower regulated and lesser accountable nature and methodology of teaching at these institutions. However, it is the interaction of school type with family’s standard of living that is of greatest interest. Figure 8 and 8b show the predicted probability of a child being praised by school type and parental economic status. This probability is calculated from a probit model that controls for selection factors as well as family background factors in Table 7a with the coefficients presented in Table 8. The results show that children from higher economic backgrounds were more likely to be praised. What may be interesting to note here is the difference in slopes for the first round (Figure 8) and the second round (Figure 8b) – from no considerable difference to distinguishable slopes in 2011-12. This may point to a greater bias in private schools as compared to government schools in terms of probability of being praised for children from higher economic backgrounds. However when it comes to the probability that the child was beaten or pinched things look a lot different even within and between the two periods. Figure 9 and Figure 9b graph the predicted probabilities for each category of the schools for the respective

• rounds. While results from the first round showed little difference in the likelihood of

physical punishment by parental economic status for children in private school, the same cannot be concluded for the second round. Both the rounds, however, made clear the strong negative relationship between economic status and punishment in government schools. Therefore children from poorer households are more likely to be punished in government schools. Moreover for the same asset holding a child in government school is far more probable to not be punished in a government school than in a private school.

31 Many studies have pointed to the pervasiveness of physical punishment in Indian Schools (Team 1999); estimates for IHDS-I suggested that nearly a quarter of children were physically punished in the prior month, the second round suggested no improvement in the scenario. In fact around 27 percent reported in affirmative when asked whether they received any form of physical punishment. Humiliation in classroom in any form may not be the most conducive environment for learning especially if such punishment is perceived as unfair. The resulting alienation may be even higher in cases where students come from poorer households. It is interesting to note here that what may was hinted as possibly greater egalitarianism among private school teachers surveyed in round –I doesn’t hold ground in the second round. Social class evidently plays a conclusive role in government schools- rich students were found to receive greater attention in terms of praise and lower physical punishment. The results therefore, suggest a need to focus on modeling and studying qualitative dimensions of classroom environment. Non-explicit aspects related to what happens in classrooms such as positive reinforcement and reduced discrimination against disadvantaged children may be equally important.

• 8. Lessons for Public Policy

As we reiterate and conclude modest and statistically significant improvements in reading and arithmetic skills of students in private schools for both rounds, special note needs to be made regarding concentrated benefits among disadvantaged students. However, one must be careful in prescribing private schooling as solution to India’s education and learning woes as a policy takeaway from these conclusions. The conclusions from both the surveys need to looked at from two vantage points – the empirical results based on our two datasets and theoretical issues raised in the literature. Table 9 provides overview of inter-state variation in reading skills across India based

• n Model 2 from Table 7a and 7b for respective rounds with state of residence and

private school interaction term added. Column 1 shows unadjusted differences along states; column 2 shows the predicted scores for students in private schools, holding family characteristics constant at all India means; column 3 shows the predicted scores in private and government schools. The values tabulated are calculated for both the rounds and sorted from lowest difference to highest difference in the first round to aid our broad comparison.

32

Table 9: Predicted Reading scores for Children in Private and Government Schools by State- IHDS-I and IHDS-II

Unadjusted Reading Score Adjusted Difference

Govt Private (Private-Govt.) IHDS-II IHDS-I IHDS-II IHDS-I IHDS-II IHDS-I IHDS-II IHDS-I

North East 2.55 2.57 2.48 2.78 2.73 2.49 0.25

• 0.29

Maharashtra/Goa 2.6 2.83 2.6 2.77 2.64 2.55 0.04

• 0.22

Tamil Nadu 2.43 3.17 2.36 2.03 2.55 1.84 0.19

• 0.19

Delhi 3 3.09 2.76 2.79 3.29 2.69 0.53

• 0.1

Haryana 2.75 2.88 2.35 2.73 3.29 2.65 0.94

• 0.08

West Bengal 2.91 2.45 2.88 2.83 3.27 2.91 0.39 0.08 Gujarat 2.58 2.79 2.44 2.62 3.12 2.76 0.68 0.14 Kerala 3.08 3.29 3.06 3.7 3.11 3.87 0.05 0.17 Chhattisgarh 2.45 2.81 2.19 2.91 3.26 3.1 1.07 0.19 Orissa 2.59 2.65 2.46 2.67 3.55 2.95 1.09 0.28 Karnataka 2.4 2.5 2.32 2.35 2.72 2.64 0.4 0.29 Himachal Pradesh 3.18 3.43 3.19 3.13 3.17 3.48

• 0.02

0.35 Rajasthan 2.67 2.52 2.2 2.43 3.21 2.89 1.01 0.46 Andhra Pradesh 2.26 2.4 2.1 2.21 2.43 2.68 0.33 0.47 Punjab 2.93 2.94 2.7 2.46 3.24 3 0.54 0.54 Jharkhand 2.28 2.58 2 2.73 2.71 3.27 0.71 0.54 Assam 2.54 2.84 2.41 2.97 3.11 3.52 0.95 0.55 Madhya Pradesh 2.45 2.1 2.16 2.36 3.2 2.99 1.45 0.63 Uttar Pradesh 2.27 2.02 1.75 2.03 2.66 2.72 0.37 0.69 Utttarakhand 2.37 2.74 2.29 2.53 2.48 3.24 0.62 0.71 Bihar 2.05 2.31 1.86 2.72 3.06 3.48 1.2 0.76 Jammu and Kashmir 2.69 2.37 2.24 2.03 2.94 2.85 0.7 0.82

The results show substantial interstate variation in the scores of both government and private school students. Controlling for parental characteristics, the national average reading score adjusted for government and private school students is 2.34 and 2.94 respectively with 0.6 as difference between the two. Government school students in states like Kerala, Himachal Pradesh, Maharashtra, West Bengal performed better or nearly as good as private school students in many other states. Within states, the performance has generally fallen between the two periods. This is in line with other analyses that conclude across the board falling learning levels. Even so, changes can be observed in the second round when compared to the first round. Private school students generally outperform than government students with a few exceptions.

33 What is more important to note that not all private school advantage can be reasoned as result of much touted supremacy of the system. This is because the states with the greatest differences are the ones infamous for having inefficient governance systems. In particular , these states are known to have poorly functioning public institutions in addition to being som eof the poorer states in India. Therefore instead of a blind embrace of touted benefits of private schooling it may be essential to first understand why government schools function well in some states. Teacher absenteeism cannot be blamed as the only reason for the dismal performance. Kerala is an interesting case in point. The difference between government and private school student learning outcomes is low even as 56 percent of the students are in private schools. The observations and results suggest that it may be worthwhile examining the differences in classroom environment between the two. A better understanding of how parental and social class and networks operate (Dubey et al. Forthcoming) on the demand side of learning outcomes may help in better evaluation

• f merits and demerits of private education. Evaluation of public and private

education therefore needs to be comprehensively undertaken by conducting few experimental studies with diverse factors to get a fair sense of reasonable explanation for the development issue at hand.

34

Figures and Graphs

10 20 30 40 50 60 1 2 3 4 5 6 7 8 9 10 Percentage Current Standard Figure 1b: Enrollment in Private School by Current Standard, IHDS-II

35

1000 2000 3000 4000 5000 Education Costs 1 2 3 4 5 6 7 8 9 10 EQ4 2.6 Standard (years) <1=0 Govt School Private School

Figure 2b Total Education Cost by Standard for Public and Private School Students (Age 6-14) IHDS-II

36

14 17 22 19 27 12 15 24 22 27 5 11 18 22 47 5 10 16 20 47 10 20 30 40 50 Cannot Read Letters Words Paragraph Story

Figure 3: Distribution of Reading Skills by School Type

Government IHDS-II Government IHDS-I Private IHDS-I Private IHDS-II 20 40 27 13 21 37 24 17 10 26 31 34 8 30 35 26 10 20 30 40 50 No Numbers Numbers Subtraction Division

Figure 4: Distribution of Arithmetic Skills by School Type

Government IHDS-II Government IHDS-I Private IHDS-I Private IHDS-II

37 Figure 6b

.5 1 1.5 2 2.5 3 3.5 4 Reading Scores 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Total household assets, same as IHDS 2005

• Govt. School

Private School Predicted Reading Scores by Standard of Living for Government and Private Schools, IHDS-II

38 Figure 7b

.5 1 1.5 2 2.5 3 3.5 4 Arithmetic Scores 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Total household assets, same as IHDS 2005

• Govt. School

Private School

Predicted Arithmetic Scores by Standard of Living for Government and Private Schools, IHDS-II

39

Figure 8b. Probability of a Child Being Praised in the Last Month by Standard of Living .2 .4 .6 .8 Probability of Being Praised 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Total household assets, same as IHDS 2005

• Govt. School

Private School

Probability of a Child Being Praised in the Last Month by Standard of Living, IHDS- II

40 Figure 9b.

0 .05.1.15.2.25.3.35.4.45.5.55.6.65.7.75.8 Probability of Being Beaten 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Total household assets, same as IHDS 2005

• Govt. School

Private School Probability of a Child Being Beaten in the Last Month by Standard of Living, IHDS-II

41

References

Banerjee, A. V., S. Cole, E. Duflo and L. Linden (2007). "Remedying education: Evidence from two randomized experiments in India." The Quarterly Journal of Economics 122(3): 1235-1264. Chakrabarti, R. and P. E. Peterson (2008). School Choice International: Exploring Public-Private Partnerships, ERIC. Chopra, R. and P. Jeffery (2005). Educational regimes in contemporary India, Sage. Coleman, J. S., T. Hoffer and S. Kilgore (1982). High school achievement: Public, Catholic, and private schools compared, New York: Basic Books. Datta Gupta, N., A. Dubey and M. Simonsen (2017). "Rising school attendance in rural India: an evaluation of the effects of major educational reforms." Education Economics: 1-20. Dreze, J. and A. Goyal (2003). "Future of mid-day meals." Economic and political weekly: 4673-4683. Drèze, J. and A. Sen (2013). An uncertain glory: India and its contradictions, Princeton University Press. Glewwe, P. and H. A. Patrinos (1999). "The role of the private sector in education in Vietnam: Evidence from the Vietnam Living Standards Survey." World Development 27(5): 887-902. Goddard, R. D. (2003). "Relational networks, social trust, and norms: A social capital perspective on students’ chances of academic success." Educational evaluation and policy analysis 25(1): 59-74. Govinda, R. (2002). India education report, Oxford University Press New Delhi. Hanushek, E. A. (1995). "Interpreting recent research on schooling in developing countries." The world bank research observer 10(2): 227-246. Hanushek, E. A. (1997). "Assessing the effects of school resources on student performance: An update." Educational evaluation and policy analysis 19(2): 141- 164. Heckman, J. and S. Navarro-Lozano (2004). "Using matching, instrumental variables, and control functions to estimate economic choice models." The Review of Economics and Statistics 86(1): 30-57. Heyneman, S. P. and W. A. Loxley (1983). "The effect of primary-school quality on academic achievement across twenty-nine high-and low-income countries." American Journal of sociology 88(6): 1162-1194. Hirschman, A. O. (1970). Exit, voice, and loyalty: Responses to decline in firms,

• rganizations, and states, Harvard university press.

Hoffer, T., A. M. Greeley and J. S. Coleman (1985). "Achievement growth in public and Catholic schools." Sociology of Education: 74-97.

42 Jencks, C. and S. E. Mayer (1990). "The social consequences of growing up in a poor neighborhood." Inner-city poverty in the United States 111: 186. Jimenez, E. and M. E. Lockheed (1995). Public and private secondary education in developing countries: A comparative study, World Bank Publications. Jimenez, E., M. E. Lockheed and V. Paqueo (1991). "The relative efficiency of private and public schools in developing countries." The World Bank Research Observer 6(2): 205-218. Kelkar, V. (2006). "Let Every Parent be a Consumer." India Today(January 16). King, E., E. Bettinger, E. Bloom, J. Angrist and M. Kremer (2002). Vouchers for private schooling in colombia: Evidence from a randomized natural experiment, The Field Experiments Website. Kingdon, G. G. (2007). "The progress of school education in India." Oxford Review of Economic Policy 23(2): 168-195. Kremer, M., N. Chaudhury, F. H. Rogers, K. Muralidharan and J. Hammer (2005). "Teacher absence in India: A snapshot." Journal of the European Economic Association 3(2-3): 658-667. Meyer, J. W. (1977). "The effects of education as an institution." American journal

• f Sociology 83(1): 55-77.

Muralidharan, K. (2006). "Public-private partnerships for quality education in India." Unpublished draft. Muralidharan, K. (2012). "Long-term Effects of Teacher Performance Pay: Experimental Evidence from India." Society for Research on Educational Effectiveness. Panagariya, A. (2008). India: The emerging giant, Oxford University Press. Pong, S.-l. (1998). "The school compositional effect of single parenthood on 10th- grade achievement." Sociology of education: 23-42. Roscigno, V. J. (2000). "Family/school inequality and African-American/Hispanic achievement." Social Problems 47(2): 266-290. Shariff, A. (1999). India, Human Development Report: A Profile of Indian States in the 1990s, Oxford University Press, USA. Stolzenberg, R. M. and D. A. Relles (1997). "Tools for intuition about sample selection bias and its correction." American Sociological Review: 494-507. Team, P. (1999). Public report on basic education in India, Oxford University, New Delhi, IN. Tooley, J. and P. Dixon (2005). "An inspector calls: the regulation of ‘budget’private schools in Hyderabad, Andhra Pradesh, India." International Journal of Educational Development 25(3): 269-285. Tooley, J. and P. Dixon (2005). "Private Schools Serving the Poor: A Study from Delhi, India." EG West Centre, University of Newcastle upon Tyne, EG West Centre. Walker, M. (2011). "PISA 2009 Plus Results: Performance of 15-year-olds in reading, mathematics and science for 10 additional participants."