WPS6897 Policy Research Working Paper 6897 Private School Participation in Pakistan Quynh Nguyen Dhushyanth Raju The World Bank South Asia Region Education Unit May 2014 Policy Research Working Paper 6897 Abstract : Private schooling is an important feature of the participation among children varies largely from one educational landscape in Pakistan and is increasingly a household to another, rather than within households, topic of public and government discourse. This study and to a greater extent than does government school uses multiple rounds of national household sample participation. The spatial patterns of private school surveys to examine the extent and nature of private supply are often strongly correlated with the spatial school participation at the primary and secondary levels patterns of private school participation. In the 2000s, in Pakistan. Today, one-fifth of children—or one-third private school participation rates grew in Punjab, of all students—go to private school in Pakistan. Private Sindh, and Khyber Pakhtunkhwa provinces and across school students tend to come from urban, wealthier, and socioeconomic subgroups, contributing in particular to more educated households than do government school the growth in overall school participation rates for boys, students and especially out-of-school children. Important children from urban households, and children from differences exist across Pakistan’s four provinces with households in the highest wealth quintile. Nevertheless, respect to the characteristics of private school students the composition of private school students has become relative to government school students, as well as in the less unequal over time. This trend has been driven mainly composition of private school students. Private schooling by Punjab province, which has seen declines in the shares is highly concentrated, with a few districts (situated of private school students from urban households and mainly in northern Punjab province) accounting for households in the highest wealth quintile. most of the private school students. Private school This paper is a product of the Education Unit, South Asia Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at nguyen4@ worldbank.org or draju2@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Private School Participation in Pakistan Quynh Nguyen World Bank Dhushyanth Raju ∗ World Bank We thank Amit Dar and Huma Ali Waheed for their encouragement of and support for this study; Andaleeb Alam and Mahesh Dahal for their research assistance; and Donald Baum, Seo Yeon Hong, and Shinsaku Nomura for their comments and suggestions. We gratefully acknowledge the Australian Department of Foreign Affairs and Trade (DFAT) and the World Bank for their financial support for the study. ∗ Nguyen: qnguyen4@worldbank.org. Raju: draju2@worldbank.org. 1. Introduction Private schooling in Pakistan has received growing and widespread attention in recent years. Researchers are studying it. The popular media at the local, national, and international level is reporting on it. Both the country’s government, at different tiers of the administration, as well as those international development agencies that provide Pakistan with financial and technical assistance in the education sector are grappling with the phenomenon’s implications when it comes to education policy, development and reform, and service delivery. Pakistan’s private school system has a long history, with its roots in the pre-independence era. In the early decades after independence, the system remained modest, being largely comprised of schools supported by nongovernmental organizations (NGOs), both religious (madrassas, missionary schools) and secular. These schools primarily catered to high-income families residing in major cities. Starting in the 1990s, there was a boom in private schools, leading to a dramatic structural transformation in the school system at large. 1 This transformation is still underway, as the private school system continues to proliferate, expand in reach, and change in composition. Using school census data from 1999/2000, Andrabi, Das, and Khwaja (2008) found that the majority of Pakistan’s roughly 36,000 private schools were established in the 1990s and were at the primary level (up to grade 5). The rural share of private schools established in each year was at least as large as the urban share. Furthermore, the vast majority of private schools established in the 1980s and 90s reported that they were for-profit. Using school census data from 2007/08, I-SAPS (2010) determined that the number of private schools has since doubled to 70,000, with particularly strong growth in schools at the middle and high levels in both rural and urban areas. Using multiple rounds of household sample survey data, Andrabi et al. also found that the private school share of enrollment rose markedly over the 1990s for both rich and poor households and urban and rural households, and rose more in the provinces of Punjab and Khyber Pakhtunkhwa (KP) than in Sindh and Balochistan. Over this same period, the government school system—the dominant provider of schooling in terms of the number of 1 The boom in private schools and private school participation is likely driven by multiple factors. One often-stated factor is poor service delivery in government schools, but the claim is yet to be empirically substantiated. Andrabi, Das, and Khwaja (2013) find that past expansion of government secondary schools for girls is one driver of the expansion of low-cost private schools. They argue that the pathway is secondary school educated women taking up employment as teachers in low-cost private schools at low, market-competitive wage rates. 2 institutions and share of enrollment—has seen its position steadily erode, particularly in urban areas and in the rural parts of Punjab and KP provinces. This has occurred despite the fact that government schools are ostensibly free for the user, while private schools typically charge fees. Today, the private school system is largely composed of institutions that are for-profit, fee-based, secular, autonomous, unregulated in practice, and which lack direct government support. In other words, they are purely private entities whose school service delivery decisions are dictated by market-competitive forces. A large segment of the private school system is also highly affordable. School fees are generally low enough that poor households manage to pay them. For example, Andrabi et al. (2008) find that average tuition fees constitute around 2 percent of the average household income in both rural and urban areas. Private schools are affordable due to their low operating costs, a main component of which is labor. Private schools tend to be staffed by young, unmarried women with low levels of education and little or no formal training in teaching. Private school teachers are also paid substantially less on average than government school teachers, even after accounting for differences in the characteristics of teachers between the two school types (Andrabi et al. 2008). In this study, we use recent rounds of household sample survey data that are national in coverage and representative at a low-administrative level—at the district level—to provide both a panoramic and a high-resolution profile of private school participation at the primary and secondary levels in Pakistan. 2 We specifically examine the extent and nature of the phenomenon by disaggregating the data in multiple ways to reveal patterns across (1) selected socioeconomic subpopulations, (2) administrative divisions/spatial units (country, province, and district), and (3) children within households. We also compare the extent of private school participation in Pakistan to that of India, as the two countries have a shared history of political, social, and economic development prior to achieving their respective independence in 1947. This study is descriptive. The nature and contents of the data (observational data, basic information on schooling) do not allow us, for example, to examine what factors encourage or inhibit private school participation or which benefits—human capital and other—might be obtained by children, families, and communities from private school participation. Existing research finds that private schooling is associated with higher student achievement (Alderman et 2 Pakistan has five administrative tiers: federal, province, district, tehsil/taluka, and union council. In 2010/11, the year of our most recent survey data, there were 113 districts in Pakistan’s four provinces. 3 al. 2001; Das et al. 2006; Aslam 2003, 2009; Andrabi et al. 2010; Andrabi et al. 2011) and labor market earnings (Asadullah 2009) in Pakistan. Among these studies, Andrabi et al. (2010) identify the causal effect of private schooling on student achievement by using standardized test score data for primary-grade students in selected villages of Punjab and instrumenting for the child’s private school participation by using the household’s distance to private school relative to the distance to government school, conditional on the distance of the household from the village center. Andrabi et al. (2010) additionally find that average student achievement in private schools is 0.8 to 1 standard deviations higher than in government schools. Using the same instrumental variables strategy, the study authors also find a causal effect of private school participation on student civic values, as measured through a standardized civic knowledge and disposition test. This study comes closest to previous work by Andrabi et al. (2008), in that we use household survey data to examine private school participation in the 2000s, updating the findings of Andrabi et al. for the 1990s. Our study also extends the previous work by extracting more information from the survey data on, for example, whether and to what extent private school participation differs spatially (as measured at the district level) as well as among children across and within households. At the same time, our study is more limited than the previous work in that we do not examine the characteristics of private schools and the private school participation decision at the village level (which Andrabi et al. do in their study by using school and population census data). I-SAPS (2010) has, however, provided some updated work on the characteristics of private schools using school census data collected over the 2000s. Our examination of current private school participation using household survey data from 2010/11 provides six main findings. First, the extent of school participation for children in the 6 to 10 and 11 to 15 age groups is large: about one-fifth of children go to private school in Pakistan, which translates into roughly one-third of all students, given the sizeable share of the country’s children that do not go to school at all. Pakistan’s national and provincial levels of private school participation do not, however, stand out when compared, for example, to corresponding private school participation rates in India and its states. Second, as expected, private school students tend to come from urban, wealthier, and more educated households than do government school students, and especially out-of-school children. Third, aside from differences in private school participation rates among provinces, there are, at times, qualitative 4 differences in the characteristics of private school students compared to government school students from one province to another. The composition of private school students also differs across provinces, with the sharpest distinctions arising between Punjab and KP, on one side, and Sindh and Balochistan, on the other. The differences in the composition of private school students between KP and Sindh are particularly interesting given that these two provinces have comparable private school participation rates. Fourth, private schooling is highly concentrated in Pakistan, with over 50 percent of private school students residing in 10 out of the country’s 113 districts. These 10 districts tend to be more urban and wealthier, and most of them are situated in northern Punjab. Fifth, most of the variation in school participation among children is due to the variation in school participation among children across households rather than within households. This pattern is even more pronounced in relation to private school participation than government school participation. Sixth, the spatial patterns in private school participation across provinces, districts, and rural vs. urban areas frequently overlap to a high degree with the spatial patterns in private school supply, obtained using separate school census data. Our examination of the evolution of private school participation over the 2000s using household survey data from 1998/99 onwards provides three main findings. First, private school participation rates grew markedly in Punjab, KP, and Sindh. Private school participation rates also grew markedly in all selected socioeconomic subgroups. Second, the growth in private school participation rates contributed more to the growth in overall school participation rates for boys, children from urban households, and children from households in the highest wealth quintile (which are the traditionally advantaged subgroups) than for other socioeconomic subgroups. Third, the growth in private school participation was nevertheless equalizing in nature, particularly in Punjab, where the shares of private school students from households in the highest wealth quintile and urban households fell. 3 The remainder of the paper is organized as follows. Section 2 describes the data and key variables. Section 3 discusses private school participation rates at the country level, and across provinces and selected socioeconomic subgroups. Section 4 compares private school participation rates in Pakistan’s provinces to those in India’s states. Section 5 compares the 3 Although appearing to be contradictory, the two findings are mutually possible. The first finding pertains to the extent of private school participation in subgroup x, while the second finding pertains to the extent of subgroup x in private school participation, where subgroup x is a minority subgroup in the population. 5 socioeconomic characteristics of private school students to those of government school students and out-of-school children, and considers the composition of private school students across provinces. Section 6 discusses the distribution of private school students across districts. Section 7 discusses the distribution of private school participation among children within the same household. Section 8 discusses how private school participation rates and the composition of private school students have evolved over the 2000s. Section 9 discusses associations between the spatial distribution of private schools and key spatial patterns in private school participation. Section 10 summarizes our main findings. 2. The data and variables The data for this study come from national household sample surveys administered by the Pakistan Bureau of Statistics (PBS). 4 The surveys are Living Standard Measurement Surveys and Core Welfare Indicator Questionnaires adapted to the Pakistan context. In constructing the current picture, we use data from the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey, the latest available survey for which primary data have been publicly released by PBS at the time of writing this paper. 5 The 2010/11 PSLM survey is representative at the district level and interviewed 75,979 households in 5,368 Primary Sampling Units (PSUs). 6 In constructing the picture over the 2000s, we mainly use data from the 1998/99 Pakistan Integrated Household Survey (PIHS) and the 2004/05 PSLM survey as baseline data to estimate the change in private school participation over 1998/99–2010/11, a twelve-year period, and 2004/05–2010/11, a six-year period, respectively. The 1998/99 PIHS is representative at the province level and interviewed 14,820 households in 1,050 PSUs. The 2004/05 PSLM survey is representative at the district level and covers 73,424 households in 5,164 PSUs. We also use data from the 2001/02 PIHS and the 2005/06, 2006/07, 2007/08, and 2008/09 PSLM surveys to track more finely the evolution of private school participation rates over the 2000s by province and by 4 The Pakistan Bureau of Statistics (PBS) is the country's main agency tasked with collecting and compiling statistical information on socioeconomic features of the economy. The PBS was created in 2011 through a merger of the Federal Bureau of Statistics and other federal-level statistical agencies. Prior to 2011, the PIHS and PSLM surveys were administered by the Federal Bureau of Statistics. 5 PBS released the 2011/12 PSLM survey report in June 2013; the release of the primary data will follow. 6 Rural PSUs are villages. Urban PSUs are blocks of cities or towns, where each block is composed of 200–250 households (PSLM survey reports, Federal Bureau of Statistics, Government of Pakistan). 6 selected subgroup (females vs. males, urban vs. rural). However we do not use these additional data to examine the nature of the evolution of private school participation at the same depth as with the 1998/99 PIHS and 2004/05 and 2010/11 PSLM survey data. All the surveys cover the four provinces and the Islamabad Capital Territory (ICT). ICT accounted for less than 1 percent of the population of Pakistan in 2012. 7 Given its relatively small size, we exclude ICT from our analysis and only examine private school participation in the four provinces. 8 For ease of exposition, we refer to the four provinces taken together as the country. The household sample surveys ask a small set of basic questions about education. Individuals age four and above are asked their current schooling status. Those who report being currently enrolled in school are asked in which grade or level they are and in which type of school. Regarding school type, the response options include government, private, and a few others (Masjid school, Deeni Madrassa, NGO/Trust school, and Non-Formal Basic Education (NFBE) community school). Thus, given these response options, the choice of “private” is likely to largely reflect for-profit, fee-based, secular private schools, although there may be some errors of inclusion if parents are not able to distinguish between private and NGO, Trust, and NFBE community schools in their response to the survey. In the 2010/11 survey, only 1.5 percent and 0.4 percent of children in the four to 18 age group were reported to be in Masjid schools/Deeni Madrassas and NGO/Trust/NFBE community schools, respectively. For the findings presented in Section 3, children are defined as students if they are reported to be in grade one or higher in any type of school. We disaggregate students by three types of school: (1) private, (2) government, and (3) other (Masjid school, Deeni Madrassa, NGO/Trust, NBFE community school). Non-students (or out-of-school children) are disaggregated into two types: (1) those never in school, based on their response that they never attended school or that the highest grade attended was katchi (preschool); and (2) those who 7 The denominator for the percentage is the population in the four provinces and ICT. The percentage is calculated from population projection data from the National Institute of Population Studies (NIPS). 8 Excluding ICT from the analysis does not influence the patterns we observe in private school participation in Pakistan. In addition, given its size, and albeit territories and provinces are at the same administrative tier, it seems inappropriate to include ICT as a separate unit in any analysis where we compare patterns in private school participation across provinces. Using the 2010/11 PSLM survey, we estimate total school participation and private school participation rates for the six to 10 age group of 89 percent and 33 percent respectively for ICT, which are higher than in each of the provinces. In any analysis of schooling, what might be more appropriate is to compare ICT to the provincial capitals or other major urban centers in the country. 7 dropped out, based on their response that they are currently not in school and the highest grade they attended was grade one or higher. 9 For the findings presented in Section 4 and later, children are defined as students only if they are reported to be in grade one or higher in either private or government school, and we disaggregate students into these two types of schools only. We examine private school participation for children in two age groups: six to 10 and 11 to 15. The age groups correspond to the official ages for primary schooling (grades 1 to 5) and secondary schooling (grades 6 to 10), respectively (Government of Pakistan, Ministry of Education 2009). In the 2010/11 survey, there were 76,806 children in the six to 10 age group in 42,606 households, and 61,623 children in the 11 to 15 age group in 37,620 households. 10 The private school participation rate for a given age group is defined as the share of children in that age group that is in private school. The private school share of enrollment for a given age group is defined as the share of students in that age group that is in private school. The characteristics of children we examine comprise of (1) gender, (2) age, (3) household location in terms of urban/rural and district, (4) household wealth measured by household asset index quintiles, (5) the completed education level of the household head (to which we loosely refer as the “education level of the household”), (6) total household size, and (7) the number of school- age children in the household (see Table 1 for the manner in which variables were defined and constructed). All statistics are estimated accounting for survey sampling weights and, where relevant, clustering at the PSU level. 3. The extent of private school participation In the context of Pakistan, the extent of private schooling—both in absolute terms and relative to the extent of other types of schooling—has to be referenced against the extent of schooling in general. The backdrop is one in which a large share of children simply do not go to school. 11 The 9 For those having dropped out of school, the surveys do not ask what type of school the individual last went to. Thus, we cannot examine whether the rates of exit from schoo l differ by school type. 10 There are children in the six to 10 age group that are in katchi (preschool). In 2010/11, 8.4 percent of children in the six to 10 age group were reported to be in katchi. For the purposes of our analysis, these children are assigned out-of-school status as, in practice, katchi typically serves as institutional childcare rather than formal, structured preschool education. 11 We can rule out the lack of school availability as a primary general explanation for low school participation in Pakistan. Using the 2010/11 survey data, we find that 84 percent of households with children in the six to 10 age group reside within fifteen minutes of the nearest primary school, but less than two-thirds of them send all of their children to school. Government schooling is also free, apart from nominal monthly contributions to funds that may 8 level of school participation in Pakistan is low relative to that of other countries in South Asia, but also in relation to other countries at its per-capita income level. Moreover, the country is likely to fall substantially short of the 2015 United Nations Millennium Development Goal of universal primary education. Distribution of children across schooling statuses at the country level Figure 1 depicts the distribution of children in age groups six to 10 (Panel A) and 11 to 15 (Panel B) in 2010/11 across five schooling statuses for the country as a whole. The schooling statuses are (1) in private school, (2) in government school, (3) in other types of schools, (4) never went to school, and (5) dropped out of school. At the country level, about one-third of children in the six to 10 age group are not in school. Specifically, 31 percent of children have never gone to school, while a negligible percentage has dropped out. Forty-five percent of children are in government school, while most of the remaining children—22 percent—are in private school. Given the sizeable share of children that are not in school, the private school participation rate of 22 percent translates into a private school share of enrollment of 32 percent. The picture remains roughly the same for children in the 11 to 15 age group. One-third of these children are not in school. Specifically, 12 percent of children have dropped out, whereas 22 percent have never gone to school. Forty-six percent are in government school. Eighteen percent are in private school, which is a few percentage points lower than the corresponding rate for the six to 10 age group. Again, given the sizeable share of children that are not in school, the private school participation rate of 18 percent translates into a private school share of enrollment of 27 percent. Private school participation rates across provinces be operated by schools. These estimates tell us that there are factors other than pure distance to school––to which we know households are sensitive (see, e.g., World Bank 2005)––and the direct cost of schooling that are behind the sizeable shortfall in school participation. The primary cause is likely related to the attributes and output of available schools (i.e., the features of service delivery), which raise costs and reduce benefits for households. The documented low level of student achievement would be key among them, making schooling an unwise investment decision in a standard economic decisionmaking framework, leading especially poor households to opt out of schooling (for all children or selectively for some children), given that their choice set in terms of school options may be particularly inferior. Indeed, over two-thirds of out-of-school children in the six to 10 age group come from households in the bottom two wealth quintiles. 9 Figure 1 also depicts the distribution of children in age groups six to 10 (Panel A) and 11 to 15 (Panel B) in 2010/11 across five schooling statuses, separately by province. For the six to 10 age group, Punjab has the highest private school participation rate at 27 percent, followed, in descending order, by Sindh (18 percent), KP (16 percent), and, trailing by a large distance, Balochistan (3 percent). The government school participation rate does not differ across provinces to the same extent; the rates vary between 44 and 49 percent. The lower private school participation rates in Sindh, KP, and Balochistan relative to Punjab’s are accompanied by higher out-of-school rates in these provinces, which imply that the relative difference in the private school share of enrollment between these provinces and Punjab is smaller. The patterns across provinces for the six to 10 age group are qualitatively similar for the 11 to 15 age group; province rankings in terms of the private school participation rate are the same as noted above, and the relative difference between provinces in the private school share of enrollment is smaller than the relative difference between provinces in private school participation rates. While the private school participation rate is lower for the 11 to 15 age group relative to the 6 to 10 age group in Punjab (21 percent vs. 27 percent), the rates across the two age groups are roughly equivalent in each of the other provinces. Private school participation rates across socioeconomic subgroups Figure 2 depicts the distribution of children in age groups six to 10 (Panel A) and 11 to 15 (Panel B) in 2010/11 in the five schooling statuses, separately by (1) location (urban vs. rural), (2) gender, and (3) household wealth (lowest, middle, and highest quintiles). Private school participation rates are substantially lower in rural areas than in urban areas; for example, for the six to 10 age group, it is 13 percent vs. 43 percent. In contrast, government school participation rates exhibit the opposite pattern: the rate is markedly higher in rural areas for the six to 10 age group (50 percent vs. 35 percent) and marginally higher for the 11 to 15 age group (48 percent vs. 44 percent). For both age groups, private school participation rates are slightly lower (by two to three percentage points) for girls relative to boys. The size of the female disadvantage in private school participation rates contrasts with the much larger female disadvantage observed in government school participation rates. For example, for the six to 10 age group, the female-male gap in government school participation rates is –8 percentage points. Exploring further, the gender gap 10 in private school participation rates remains similarly small when we separately examine urban and rural children and children from poor (lowest wealth quintile) and nonpoor households. In contrast, the gender gap in government school participation rates is largely a result of the corresponding gender gap among rural and poor children. 12 For both age groups, private school participation rates increase with household wealth quintiles. For example, for the six to 10 age group, the private school participation rate is 4 percent in the lowest wealth quintile, 20 percent in the middle quintile, and 57 percent in the highest quintile. In contrast, for both age groups, government school participation rates display an inverted-U shape in relation to household wealth, peaking for the middle quintile. In the lowest quintile, the out-of-school rate markedly exceeds the government school participation rate. In the highest quintile, the private school participation rate markedly exceeds the government school participation rate for the six to 10 age group and marginally exceeds it for the 11 to 15 age group. These patterns are consistent with the likelihood of school participation increasing with household income and households with higher incomes purchasing higher- quality schooling, which tends to be supplied by the private market (Andrabi et al. 2008). 4. The extent of private school participation: Pakistan vs. India While private school participation rates in Pakistan are significant in absolute terms, how do they compare to those in other parts of South Asia? We compare Pakistan to India (rather than to additional countries in South Asia) for two main reasons. First, the two countries shared a common administrative and institutional history under British rule until 1947. Second, we consider the common types of private schools in Pakistan to have sizeable counterparts in India than in other South Asian countries. 13 12 Certain features of government schools may discourage poor and rural parents from sending their children— especially their girls—to these schools. For example, private schools tend to be coeducational and staffed with female teachers (Andrabi et al. 2008, I-SAPS 2010). In contrast, a large share of government schools tends to be officially single-sex and staffed by female and male teachers accordingly. Where private or government schools are located within villages (on the periphery or centrally) also systematically differs (Andrabi et al. 2008). Poor and rural parents may be sensitive to these features, among others, contributing to the divergent results in the gender gap in school participation between the two school systems. 13 In South Asia, Bangladesh, India, Pakistan, and Sri Lanka were formerly under British rule. The representative private school in Bangladesh receives financial support from the government, which makes it different from the representative private school in Pakistan (Asadullah 2009). In Sri Lanka, virtually all private schools were nationalized in the early 1960s, and the private school system remains contained (Maurer 2012). See, for example, Dahal and Nguyen (2014) for a description of the types and prevalence of private schools in South Asian countries. 11 Comparing private school participation rates between provinces and states Instead of comparing private school participation rates at the national level between the two countries, given that the household survey data we use are representative at a lower level, we compare the rates for provinces in Pakistan in 2010/11 to those for India’s states in 2009/10 (using the 2009/10 National Sample Survey (NSS)). 14 Provinces and states can be viewed as equivalent administrative units and constitute a key level of government administration, with primary responsibility for education policy formulation and implementation. Comparing rates at the level of the province and state provides a more spatially-refined picture of observed differences in rates between the two countries. Figure 3 depicts private school participation rates for age groups six to 10 (Panel A) and 11 to 15 (Panel B), by province in Pakistan and by state in India. The states and provinces are organized in increasing order by rate. We define private schools in India as either unaided or aided. Aided schools receive financial support from the government for school salaries and nonsalary expenditures. 15 The green bars depict the rates for the provinces, whereas the stacked blue-red bars—blue for aided and red for unaided—depict the rates for the states. The patterned bars with black borders depict the rates at the country level for Pakistan (green) and for India (blue-red). Similarly, Figure 4 depicts the private school participation rates for age groups six to 10 (Panel A) and 11 to 15 (Panel B), by province in Pakistan and state in India. However, here, we restrict the definition of private schools in India to unaided private schools only. The green bars depict the rates for the provinces and the blue bars the rates for the states, while the patterned bars with black borders depict the rates at the country level for Pakistan and India. We prefer the comparison in Figure 4 because unaided private schools in India are likely to be largely equivalent to the typical types of private schools found in Pakistan. 16 14 The data for India come from the Employment and Unemployment schedule of the National Sample Survey (NSS) 66th round, which is a state-representative household sample survey of 100,957 households conducted over the period from July 2009 to June 2010. Union Territories in India are excluded from the comparison. 15 Definitions for aided and unaided private schools are obtained from the NSS manual. 16 While comparing private school participation rates in Pakistan to unaided private school participation rates in India brings us closer to comparing apples to apples, differences in the school system environments between the two countries may work to drive a wedge in the comparison. The presence of an aided private school system in India may bias the size of the unaided private school system (i.e., its size may be different than if only—or largely—an unaided private school system existed in India). The size of the unaided private school system may be larger than otherwise, as, for example, unaided private schools may enter the education market seeking to become aided in the 12 When we compare private school participation rates for the provinces in Pakistan to those for states in India, where private schools in India consist of both unaided and aided schools, the rates for the provinces largely lie in the bottom half of the distribution of rates for states. Specifically, the rate for Balochistan is at the bottom of the distribution of rates for states in India, the rates for KP and Sindh are roughly in the second quartile of the distribution, and the rate for Punjab is near the middle of the distribution. Aggregating up to the country level, the rate for Pakistan is lower than that for India. These patterns hold for both age groups. Restricting the definition of private schools in India to unaided private schools only, the rates for provinces in Pakistan climb the ranks in the distribution of rates for states in India, and now are roughly between the bottom of the second quartile and the top of the third quartile of the distribution. In addition, aggregating up to the country level, the relative positions of the two countries reverse, with Pakistan’s rate higher than that of India. Again, these patterns hold for both age groups. 17 Comparing urban and rural private school participation rates between provinces and states We also examine the distribution of private school participation rates between India’s states and Pakistan’s provinces separately for urban and rural areas. The same patterns noted above hold when we compare the countries using urban and rural rates at the province and state level. However, relative to rural rates, the urban rates for Pakistan’s provinces move up the ranks more when we shift from examining them in the distribution of urban private school participation rates to examining them in the distribution of urban unaided private school participation rates for India’s states. At the country level, rural and urban private school participation rates for Pakistan are below the corresponding private school participation rates for India. The rates become roughly identical between the two countries when the rural private school participation rate in Pakistan is future. Conversely, the size of the unaided private school system may be smaller than otherwise, as, for example, aided private schools may create market conditions that hinder the entry, survival, and growth of unaided private schools. Both effects may be at play simultaneously, among others, making the net bias on size theoretically ambiguous. 17 The above findings are based on comparing private school participation rates between the two countries. The overall school participation rate is about 20 percentage points lower in Pakistan than in India. Given this, if the comparison was based on the private school shares of enrollment, the extent of private school participation in Pakistan would outstrip that of India, in both urban and rural areas. 13 compared to the rural unaided private school participation rate for India. The relative positions of the two countries switch and the gap in rates widens between the two countries when the urban private school participation rate for Pakistan is compared to the urban unaided school participation rate in India. These findings hold for both age groups. 5. Characteristics of private school students and correlates of private school participation Differences between private school students and other groups at the country level Table 2 reports estimated means and proportions for selected child and household characteristics for private school students, as well as the difference in these means and proportions from those of government school students and out-of-school children, separately for age groups six to 10 (Columns 1–3) and 11 to 15 (Columns 4 to 6), in the country as a whole. We find that private school students are more likely than out-of-school children to be male and to come from urban, wealthier, and more educated households. Private school students also come from on average smaller households and households with smaller numbers of children than do out-of-school children. The same patterns hold when we compare private school students to government school students. The one exception is gender: private school students are more likely to be female than are government school students. The size of the differences between private school students and government school students is generally smaller than between private school students and out-of- school children. These findings apply to both age groups. Given that late entry into school is common and that the likelihood of schooling increases with age over primary school ages, for the six to 10 age group, the mean age of private school students is higher than for out-of-school children. Conversely, given that the likelihood of exiting school (dropping out) increases with age over secondary school ages, for the 11 to 15 age group, the mean age of private school students is lower than for out-of-school children. For both age groups, the mean age for private school students is lower than for government school students. Differences between private school students and other groups across provinces We also compare the characteristics of private school students to government school students and out-of-school children, separately by province (see Tables A1 to A4 in the Appendix). While the pattern of differences between private school students and out-of-school children that we 14 found at the country level is reflected in each of the provinces, the same does not hold true for the pattern of differences between private school students and government school students. Country-level findings that the mean age of private school students is lower than for government school students and that private school students are more likely to be female than government school students are only consistently reflected in Punjab and Sindh, respectively. The country- level finding that private school students are less likely to come from less-educated households (no schooling, grades 1 to 5) and more likely to come from more educated households (grades 6 to 8, 9 to 10, 11+) than are government school students is reflected much less sharply in some provinces. For both age groups, the share of private school students from households with a “mid” level of education (grades 6 to 8) is no different than for government school students in Punjab, KP, and Balochistan. Finally, the country-level finding that, on average, private school students come from smaller households than do government school students is only consistently reflected in Punjab and Sindh. 18 Differences in the composition of private school students across provinces Tables 3 and 4 present estimated means and proportions of selected characteristics of private school students in age groups six to 10 and 11 to 15, respectively, in each of the four provinces, and compare the differences in these means and proportions between private school students in each of the provinces. Private school students are more likely to be female in Punjab and Sindh than in Balochistan and KP, and private school students are much more likely to come from rural households in Punjab and KP than in Sindh and Balochistan. Sindh is a particularly extreme case: only 10 percent or less of private school students in the six to 10 and 11 to 15 age groups come from rural households. Private school students in Punjab are more likely to come from households in the lower wealth quintiles than in each of the other provinces. Balochistan is 18 Although we do not provide tables with the estimated results in the paper, we also fit multinomial probit regression models via maximum likelihood to the data to examine child and household correlates of the conditional likelihood of (1) being a government school student or (2) being an out-of-school child relative to the base status of (3) being a private school student. The regressions are run separately by age group, for the country as a whole as well as for each of the provinces. We find several cases of weakening or absence of statistical significance in the conditional associations relative to the unconditional differences. We suspect that this may partly be due to the presence of multicollinearity in the multiple regressions, given the types of covariates we use, which would increase the likelihood that we fail to reject a false null hypothesis (Type-II error). We also find a few conditional associations pick up significance vis-à-vis the unconditional differences: for example, we are able to pick up more frequently that private school students are younger than government school students across the provinces. 15 considerably more top-heavy than the other provinces: close to 90 percent of private school students in the 6 to 10 and 11 to 15 age groups in the province come from households in the highest wealth quintile. Private school students in Punjab and KP are more likely to come from less-educated households than in Sindh and Balochistan. On one end, private school students in Sindh come from smaller households than in each of the other provinces; on the other end, private school students in KP come from larger households than in each of the other provinces. To be sure, many of our findings on the pattern of inter-province differences in the composition of private school students apply to government school students as well. The inter- province differences in the composition of private school students are however much larger than the inter-province differences in the composition of government school students with respect to certain characteristics, such as location (urban vs. rural), household wealth, and household head’s education level. 6. The distribution of private school students across districts In Section 3, we examined the differences in private school participation rates across provinces and between urban and rural areas and found that the private school participation rate is highest in Punjab and higher in urban than rural areas. In this section, we further explore the spatial distribution of private school participation by measuring the distribution of private school students across districts, which is the lowest level of representativeness of our survey data. In examining the share of private school students at the district level, we find that private school participation is highly concentrated in Pakistan. Ten districts (out of the 113 districts in the four provinces) account for over 50 percent of private school students in the six to 10 and 11 to 15 age groups in the country. Table 5 reports summary statistics on the socioeconomic characteristics of these “top-ten” districts (referred to as the top-ten group), and compares them to the remaining districts as a whole (referred to as the non-top-ten group). For both age groups, private school participation is substantially overrepresented in the top-ten group: the group’s collective share of the total private school student population is roughly double its collective share of the total child population in the country. Consequently, private school participation rates are higher in the top-ten group relative to the non-top-ten group for both age groups. In contrast, government school participation rates are lower in the top-ten group relative to the non-top-ten group for both age groups. In terms of socioeconomic characteristics, the top-ten group is more 16 urban and wealthier (measured by the mean value of the household asset index) than the non-top- ten group. District-level information on the percent of private school students, the private school participation rate, the percent of the child population, and other selected socioeconomic characteristics are provided in Tables A5 and A6 in the Appendix for age groups six to 10 and 11 to 15, respectively. While it may not necessarily be the case, the systematic differences in the socioeconomic characteristics of districts between the top-ten group and the non-top-ten group are accompanied by similar systematic differences in the socioeconomic characteristics of private school students between the top-ten and non-top-ten groups. Table 6 reports estimated means and proportions for selected characteristics of private school students in the top-ten group, and the difference in these means and proportions from those of private school students in the non-top-ten group, separately for age groups six to 10 (Columns 1–2) and 11 to 15 (Columns 3–4). For both age groups, private school students in the top-ten group (1) are more likely to be female, (2) are more likely to come from urban, wealthier, and more educated households, and (3) come from smaller households than their counterparts in the non-top-ten group. For both age groups, the mean age of private school students is however not statistically different between the top-ten and non-top- ten groups. The districts in the top-ten group are themselves spatially concentrated. Apart from Karachi and Peshawar (which are in Sindh and KP, respectively), the remaining districts in the top-ten group are in Punjab. With the exception of Multan, the districts in the top-ten group in Punjab are largely clustered in the northeastern part of the province. Figures 5 and 6 depict the districts in Pakistan divided into three groups for private school students in age groups six to 10 and 11 to 15, respectively: (1) top-ten districts, (2) non-top-ten districts where the individual district shares of private school students are equal to or greater than 1 percent, and (3) non-top- ten districts where the individual district shares of private school students are less than 1 percent. In general, spatial patterns point to the predominance of districts in Punjab in accounting for the location of private school students: the first two groups (top-ten and ≥1 percent groups) are largely composed of districts from Punjab, while the third group (<1 percent group) is largely composed of districts from the other three provinces. 7. Distribution of private school participation among children within households 17 Thus far in the paper, we have described private schooling for all households with children in our age groups of interest, abstracting a child’s own schooling status from that of other children in her household. In the ensuing analysis, we restrict our attention to households with multiple children in the age groups of interest and examine the schooling decisions of these households for their children, specifically in terms of the extent of private schooling among children within households. 19 Decomposition 1: Between- and within-household breakdown of the variation in private school participation among children Table 7 presents standard analysis-of-variance estimates of the extent to which differences in school participation among children is attributable to differences among children across households (between-household variation) vs. differences among children within households (within-household variation), separately by school type (private vs. government) and by province, for age groups six to 10 (Panel A) and 11 to 15 (Panel B). 20 Estimations are performed on samples of households with at least two children in the relevant age group and at least one of them in school. For decomposing the variation in private school participation among children, the outcome variable is set equal to one if a child goes to private school, and to zero if otherwise. Likewise, for decomposing the variation in government school participation among children, the outcome variable is set equal to one if a child goes to government school, and to zero if otherwise. Private school participation is largely a phenomenon that varies from one household to the other rather than within households. At the country level, 82 percent and 79 percent of the variation in private school participation among children in the six to 10 and 11 to 15 age groups, respectively, is due to between-household variation (i.e., most parents choose to send all or none 19 The analysis does not strictly examine the distribution of private schooling among siblings, because the PSLM survey only provides information on the relation of household members to the household head. Thus, we cannot ascertain the sibling relations of children in the household that are not children of the household head. 20 The extent of total variation in school participation due to between-household variation is likely to be underestimated, as the extent due to within-household variation subsumes statistical noise. Elbers, Lanjouw, Mistiaen, and Ozler (2008) argue that between-group inequality for a certain decomposition should not be benchmarked against total inequality (which is equivalent to between-group inequality when all groups are simply individuals or households) but against a “maximum” between-group inequality that is derived when the number and relative sizes of groups for that decomposition are unchanged; this maximum between-group inequality would always be weakly smaller than total inequality. This, too, would imply that the extent of total variation in school participation due to between-household variation is likely to be underestimated. 18 of their children to private school instead of sending some of their children to private school). In comparison, at the country level, relatively lower shares of the variation in government school participation among children—specifically, 66 percent and 60 percent for the six to 10 and 11 to 15 age groups, respectively—are due to between-household variation (i.e., the percentage of parents that send all or none of their children to private school exceeds the percentage of parents that do the same in relation to government schooling). These findings are qualitatively similar across provinces and age groups. The difference in the percentage due to between-household variation between private school participation and government school participation is smallest in Punjab and largest in Balochistan. In Balochistan, the percentage of total variation in private school participation for the six to 10 age group due to between-household variation is 88 percent, while the corresponding statistic with respect to government school participation is 44 percent, which represents a minority share. Decomposition 2: Breakdown of households by the extent of private school participation among children within households We also examine the distribution of households in terms of the extent of private school participation among in-school children. Table 8 presents estimated shares from decomposing households with multiple children and at least one child in school into three mutually-exclusive groups based on the extent of private school participation among children that are in school, separately for age groups six to 10 (Panel A) and 11 to 15 (Panel B) and by province. The three groups are (1) all in-school children in the relevant age group go to private school, (2) some in- school children in the relevant age group go to private school (and the other children go to government school), and (3) none of the in-school children in the relevant age group go to private school (all of the in-school children go to government school). The three groups are denoted by type A (A for all), type S (S for some) and type N (N for none), respectively. This alternative decomposition basically reproduces the earlier finding that private school participation varies mainly among households. When households with multiple children send at least one child to school, they tend to send more than one child to school. Examining the six to 10 age group, 25 percent, 5 percent, and 70 percent of households are type-A, type-S, and type-N, respectively. The same pattern of the relative shares of household types holds for the 11 to 15 age group and in each of the provinces. The distribution of households by type varies across 19 provinces, particularly between Punjab and Balochistan. For example, for the six to 10 age group, 31 percent and 7 percent of households are type-A and type-S in Punjab, respectively; the corresponding statistics for Balochistan are 4 percent and 1 percent, respectively. Differences among households in types A, S, and N Table 9 reports estimated means and proportions for selected household-level characteristics for the three types of households in Pakistan, separately for age groups six to 10 (Columns 1–3) and 11 to 15 (Columns 4–6). In moving from type-A to type-S to type-N, households tend to become progressively more rural, more poor (lowest wealth quintile), less rich (highest wealth quintile), more poorest educated (no schooling), and less highest educated (secondary schooling and higher). These patterns apply to both age groups. Although we do not provide the statistics in the paper, the country-level findings are also generally reflected in each of the provinces. The pattern noted above is broadly consistent with the pattern of change in the socioeconomic characteristics of children when we shift from private school students to government school students as noted in Section 5. This similarity underscores the predominant role of household-level differences in driving child-level differences across schooling statuses. Correlates of private school participation within households Finally, we examine whether the age and gender of the child are associated with private school participation when we examine the conditional relationship between the two variables within households. Table 10 reports parameter estimates for age and gender by estimating private school participation regressions via Ordinary Least Squares (Limited Probability Model), first accounting for differences in household location, wealth, household head’s education, size, and numbers of children in different age groups, and second with household-fixed effects. These regressions are run for age groups six to 10 (Panel A) and 11 to 15 (Panel B), separately, both for the country as a whole and by province. The outcome variable is set equal to one if the child goes to private school, and to zero if otherwise. Note that, under this definition, zero denotes both government schooling and out-of-school status in the outcome variable. At the country level, accounting for differences in household-level covariates, girls in both age groups tend to have a lower likelihood of private school participation and older children in the six to 10 (11 to 15) age group tend to have a higher (lower) likelihood of private school 20 participation. The same patterns remain when the associations are identified by looking among children within their households only. 21 Examining the associations separately by province, the findings at the country level related to the conditional female disadvantage in private school participation are reflected in Balochistan, KP, and Punjab. The size of the conditional female disadvantage in private school participation is largest in KP. 22 Depending on the age group and province, the percent of total variation in private school participation explained by the regressions rises from 10 to 37 percent when we include household-level covariates, and 55 to 80 percent when we include household-fixed effects, suggesting that a substantial portion of the variation in private school participation is explained by factors (both observed and unobserved) that vary at the household level and higher. This finding is consistent with what we discovered earlier from the decompositions of the extent and pattern of variation in private school participation among children within households. 8. Evolution of private school participation rates over the 2000s In this section, we turn to an exploration of how the extent and nature of private school participation has evolved in the 2000s. Table 11 presents the change in overall school participation rates and the change in private school participation rates (both in percentage point terms), as well as the contribution of the change in private school participation rates to the change in overall school participation rates (constructed as a ratio and expressed in percent terms) over the twelve-year period from 1998/99 to 2010/11. The statistics are estimated for the country, by province (Panel A), and by socioeconomic subgroup (Panel B), for age groups six to 10 (Columns 1–3) and 11 to 15 (Columns 4–6). Note two measurement-related points. First, we refer to the absolute percentage point change in rates as “growth.” Second, the growth is in net terms, as there are flows both into and out of (private) school participation status at any given point in time. 21 Our finding that the significant female disadvantage in private school participation continues to hold when we examine the relationship between gender and private school participation within households updates and confirms Aslam’s (2009) finding of a female disadvantage in private school participation within households using national household sample survey data from 2001/02. 22 We also ran regressions with household-fixed effects where the outcome variable was set equal to one if the child goes to private school and to zero if the child goes to government school, and found a similar pattern of a conditional female disadvantage in private school participation in Balochistan, KP, and Punjab. The size of the conditional female disadvantage was particularly large for both age groups in KP and for the 11 to 15 age group in Balochistan. 21 The table is accompanied by figures in the Appendix (Figures A1 to A4) that plot the change in overall vs. private school participation rates over the period 1998/99 to 2010/11 using eight rounds of national household sample survey data (PIHS and PSLM surveys), separately for the country as a whole, by province, and by selected socioeconomic subgroup (female, male, rural, urban). In the figures, actual rates are denoted by hollow circles. The trend lines are estimated via a locally-weighted least-squares smoother. Growth in private school participation rates We first examine the growth in overall and private school participation rates over the period 1998/99 to 2010/11 at the country level and for each province. At the country level, overall school participation rates grew by 17 and 14 percentage points for the six to 10 and 11 to 15 age groups, respectively. Over the same period, private school participation rates grew by 9 percentage points for both age groups. In KP, Punjab, and Sindh, overall and private school participation rates grew markedly. In Balochistan, while the overall school participation rate for the six to 10 age group grew significantly (12 percentage points), the corresponding rate for the 11 to 15 age group grew much less (4 percentage points). In addition, private school participation rates in Balochistan were virtually stagnant (1 percentage point) for both age groups. At the country level, depending on the age group, the growth in private school participation rates contributed equally—or more so than the growth in government school participation rates—to the growth in overall school participation rates over the period. At the province level, the growth in private school participation rates accounts for the majority of the growth in overall school participation rates in Punjab for both age groups, in Sindh for the 11 to 15 age group, and in KP for the six to 10 age group. In Balochistan, Punjab, and Sindh, the contribution of the growth in the private school participation rate to the growth in the overall school participation rate is higher for the 11 to 15 age group than for the six to 10 age group. Next, we examine the growth in rates over the period 1998/99 to 2010/11 by selected socioeconomic subgroups. Except for households in the highest wealth quintile (where overall school participation rates were relatively high to begin with), overall school participation rates grew by 10 to 20 percentage points for all subgroups, with higher growth for rural relative to urban households, girls relative to boys, and households in the middle wealth quintile relative to those in the lowest and highest wealth quintiles. 22 All socioeconomic subgroups also saw a significant increase in private school participation rates. However, in contrast to the finding for overall school participation rates, private school participation rates grew more for boys, urban households, and households in the highest wealth quintile—subgroups which are traditionally more socioeconomically advantaged. In the case of urban households and households in the highest wealth quintile, depending on the age group, the growth in the private school participation rates accounts for almost all or more than the growth in school participation rates, suggesting net gains to the private school system from students shifting from government to private schooling. Finally, the contribution of the growth in the private school participation rate to the growth in the overall school participation rate is roughly the same or larger across socioeconomic subgroups for the 11 to 15 age group relative to the six to 10 age group. Change in the composition of private school students Given the significant growth in private school participation rates in certain parts of the country and across the selected socioeconomic subgroups, we examine whether the composition of private school students has systematically changed over the period. Table 12 reports estimated means and proportions of selected characteristics of private school students, the changes in means and proportions over the twelve-year period from 1998/99 to 2010/11, as well as the changes in means and proportions in the last half of the period, from 2004/05 to 2010/11, separately for age groups six to 10 (Columns 1–3) and 11 to 15 (Columns 4–6). For both the six to 10 and 11 to 15 age groups, the share of private school students from rural households rose, while the share from households in the highest wealth quintile (rich households) fell. Although we found earlier that the private school participation rates grew more for urban than for rural households and more for rich than for nonrich households, urban and rich households represent a minority of the total household population. As a result, the growth in private school participation rates among rural and nonrich households was sufficient to lead to a less unequal composition of private school students. We also find that the share of private school students from households with the lowest level of education fell, while the share from households with the highest level of education rose, with both changes occurring in the later 2000s. These findings are probably attributable to some extent to the increasing education level of households in general over the period, with changes 23 concentrated at the low and high ends of the education attainment range. In addition, we find that the average number of members and number of children in the households to which private school students belong declined. This may be explained to some extent by declining household fertility rates in Pakistan in general. We do not find a change in the share of private school students that are female. All of these findings hold for both age groups. Finally, we examine the ways in which the socioeconomic characteristics of private school students have changed, separately by province (see Tables A7 to A11 in the Appendix). For Balochistan, we only estimate changes for the period 2004/05 to 2010/11, as the 1998/99 household survey is only representative at the province level and does not provide sufficient observations to obtain reliable estimates for the subgroups. The patterns of change at the country level are mainly reflected in Punjab. They are, however, not consistently observed in other provinces, where the changes are at times smaller and not statistically significant. Contrary to the finding at the country level of no change in the female share of private school students, for the 11 to 15 age group, the corresponding share rose in Sindh, whereas it fell in Balochistan (in the period 2004/05 to 2010/11). 9. The role of private school supply The private (government) participation rate reflects the equilibrium point between the levels of private (government) schooling demanded and supplied. Using data from the 2005 National Education Census (NEC), a survey conducted by the Pakistan Ministry of Education and the former Federal Bureau of Statistics, which attempted to capture some basic information on all government and private schools in the country, we examine whether patterns in the spatial variation in school supply by school type may be related to the patterns in the spatial variation of school participation rates (equilibrium values) by school type. 23 Before turning to the findings, we note that both market and policy explanations are potentially behind the observed spatial distribution of private and government schools. For example, the Pakistan government has had a longstanding policy of expanding school availability by constructing government schools across registered communities that meet 23 While there is an incompatibility in time between the school supply information and the school participation information (2005 vs. 2011), we check the sensitivity of our findings by comparing the school supply patterns from the 2005 NEC data against the school participation patterns from the 2004/05 PSLM survey data, and find that they are qualitatively similar. 24 minimal population level requirements and where land is donated by the community. The government also assigns centrally-recruited teachers through a system of transfers and postings to run the schools. In contrast, where private schools choose to locate is largely dictated by market forces, in both factor and product (provision of schooling) markets, which biases location decisions towards urban areas and more developed rural communities (Andrabi et al. 2008). The private school regulations in effect (in de jure terms) do not explicitly constrain where private schools can locate, although specific stipulations in the regulations related to, for example, infrastructure, space, amenities, and tuition and fees may influence where private schools choose to locate. We documented earlier that (1) private school participation rates and the shares of households with all or some in-school children in private school are highest for both age groups in Punjab, followed, in decreasing order, by Sindh, KP, and Balochistan; (2) the private school participation rate is lower for the 11 to 15 age group than for the six to 10 age group in Punjab but not in the other provinces; and (3) the private school participation rate is much lower in rural than urban areas. In contrast, government school participation rates differ much less between provinces for both age groups, and, depending on the age group, the rates are higher or roughly equal between rural and urban areas. We also found that the distribution of private school students was highly skewed across districts (and disproportionately so relative to the distribution of children across districts). This begs the question of whether the spatial pattern of private school supply is associated with these spatial patterns in private school participation across provinces, districts, and rural vs. urban areas. We discuss each possible bivariate association in turn. Private school supply across provinces: Punjab has the highest share of private schools with primary grades at 69 percent, followed, in descending order, by Sindh (18 percent), KP (12 percent), and Balochistan (2 percent). These shares of private schools roughly match the population shares across provinces. The distribution of private schools with secondary grades across provinces is similar to that for private schools with primary grades, although the number of private schools with secondary schools is about two-thirds of the number of private schools with primary grades. Thus, the spatial distribution in private school supply across provinces, measured by the numbers of schools, is consistent with the spatial distribution in private school 25 participation rates and shares of households with private school students across provinces. In line with the pattern of more comparable government school participation rates across provinces for both age groups, the spatial distributions of government schools with primary and secondary grades are less skewed than the corresponding spatial distributions for private schools. Primary-secondary private school supply across provinces: The ratio of private schools with secondary grades to private schools with primary grades by province is highest in Punjab and Sindh (7:10), followed, in descending order, by KP (3:5) and Balochistan (1:2). Given this pattern, we discount provincial differences in the size of this ratio as an important explanation behind the finding of a lower private school participation rate for the 11 to 15 age group relative to the six to 10 age group in Punjab and the absence of such differences between the two age groups in the other three provinces. Private school supply between urban and rural areas: The urban-rural ratio of private schools with primary grades is 3:2, while the corresponding statistic for government schools is 1:9. One- third of the country’s population resides in urban areas. Thus, consistent with the observed pattern of rural-urban difference in school participation rates by school type, private schools are disproportionately concentrated in urban areas, whereas government schools are disproportionately concentrated in rural areas. Private school supply across districts: We examine the bivariate association between district- level numbers of private schools with primary (secondary) grades and district-level private school participation rates for the six to 10 (11 to 15) age group. Private school sizes may systematically differ across districts, which can distort the picture that emerges from using the number of private schools as an indicator of private school supply. Given this, we also examine the bivariate association between district-level numbers of private school students in primary (secondary) grades captured in the 2005 NEC (which we use as a measure of school size- adjusted private school supply) and district-level private school participation rates for the six to 10 (11 to 15) age group. The associations are always positive. That is, there are more private schools or higher private school enrollment in districts with higher private school participation rates. We examine 26 the same associations between government school supply and government school participation and find no discernible relationship across districts. 10. Summary In this study, using multiple rounds of national household sample survey data, we examine the contemporaneous (2010/11) extent and nature of private school participation in Pakistan, at the country, province, and district levels. We also examine the extent and nature of the evolution of private school participation over the 2000s. Our examination of current private school participation provides six main findings. First, the extent of private school participation for children in the six to 10 and 11 to 15 age groups is significant: roughly one-fifth of children go to private school in Pakistan, which translates into roughly one-third of students (given the large share of children that do not go to school in the country). This extent of private school participation does not however stand out when, for example, private school participation rates in Pakistan and its provinces are compared to corresponding rates in India and its states. Second, as expected, private school students tend to come from urban, wealthier, and more educated households than do government school students and especially out-of-school children. Third, aside from differences in private school participation rates across provinces, there are, at times, differences across provinces in the characteristics of private school students compared to government school students. The composition of private school students also differs across provinces, with the sharpest distinctions between Punjab and KP on one side and Sindh and Balochistan on the other. Differences in the composition of private school students between KP and Sindh are particularly interesting given that these two provinces have comparable private school participation rates. Fourth, private schooling is highly concentrated in Pakistan, with over 50 percent of private school students residing in 10 out of 113 districts in the country. These 10 districts tend to be more urban and wealthier, and most of them are situated in northern Punjab. Fifth, most of the variation in school participation among children is due to variation in school participation among children across households rather than among children within households, and this pattern is much more pronounced in relation to private school participation than government school participation. Sixth, the spatial patterns in private school participation across provinces, districts, 27 and rural vs. urban areas frequently overlap to a high degree with the spatial patterns in private school supply, obtained using separate school census data. Our examination of the evolution of private school participation over the 2000s, using household survey data from 1998/99 onwards, provides three main findings. First, private school participation rates grew markedly in Punjab, KP, and Sindh. Private school participation rates also grew markedly in all selected socioeconomic subgroups. Second, the growth in private school participation rates contributed more to the growth in overall school participation rates for boys, children from urban households, and children from households in the highest wealth quintile than for children in other socioeconomic subgroups. Third, the growth in private school participation was nevertheless equalizing in nature, particularly in Punjab, where the shares of private school students from urban households and households in the highest wealth quintile fell. The collective evidence indicates the importance of the private schooling system in Pakistan, in terms of both its present level and growth over the recent past in school participation. Some provinces and territories have introduced legislation to regulate the operations and performance of private schools. 24 Ostensibly motivated by concerns regarding unfair, deceptive, or abusive acts and practices by private schools, the regulations reach deep into many aspects of school operations, circumscribing a number of critical school management decisions. The regulations cover registration, curriculum, academic standards, length of school year/days, and recordkeeping/reporting. They also cover tuition and fees, teacher employment terms (including pay), teacher qualifications, and availability and quality of facilities. 25 To date, the regulations do not appear to have been applied in a broad, systematic, and meaningful manner. There is, however, growing demand for new, more intrusive regulations as well as for stringent implementation and enforcement of existing regulations. If applied, these regulations— in particular those that relate to school fees and schooling inputs—may be counterproductive. The regulations can weaken the growth and general dynamism and performance of private 24 They include the Islamabad Capital Territory Private Educational Institutions (Registration and Regulation) Act 2013; Khyber Pakhtunkhwa (formerly North West Frontier Province) Registration and Functioning of Private Educational Institutions Ordinance 2001, and Amendment 2002; Punjab Private Educational Institutions (Promotions and Regulations) Ordinance 1984; and the Sindh Private Educational Institutions (Regulation and Control) Ordinance 2001, and Rules 2005. 25 Such regulations are typically absent in, for example, regulations of private schools by states in the US (U.S. Department of Education, Office of Innovation and Improvement 2009). 28 schools. An alternative recasting of legislation, which does not take the approach of “micro regulating” the sector, may be more effective in protecting consumers and staff of private schools while preserving fair and effective competition that promotes private school entry, growth, and performance. 29 References Alderman, Harold, Peter F. Orazem, and Elizabeth M. Paterno. 2001. “School Quality, School Cost, and the Public/Private School Choices of Low-Income Households in Pakistan.” Journal of Human Resources 36(2): 304–326. Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. 2013. “Students Today, Teachers Tomorrow: Identifying Constraints on the Provision of Education.” Journal of Public Economics 100: 1–14. Andrabi, Tahir, Jishnu Das, Asim Ijaz Khwaja, and Tristan Zajonc. 2011. “Do Value-Added Estimates Add Value? Accounting for Learning Dynamics.” American Economic Journal: Applied Economics 3: 29–54. Andrabi, Tahir, Natalie Bau, Jishnu Das, and Asim Ijaz Khwaja. 2010. Bad Public Schools are Public Bads: Civil Values and Test Scores in Public and Private Schools. Manuscript. Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. 2008. “A Dime a Day: The Possibilities and Limits of Private Schooling in Pakistan.” Comparative Education Review 52(3): 329– 355. Asadullah, M. Niaz. 2009. “Returns to Private and Public Education in Bangladesh and Pakistan: A Comparative Analysis.” Journal of Asian Economics 20: 77–86. Aslam, Monazza. 2009. “The Relative Effectiveness of Government and Private Schools in Pakistan: Are Girls Worse Off.” Education Economics 17(3): 329–354. ———. 2003. “The Determinants of Student Achievement in Government and Private Schools in Pakistan.” Pakistan Development Review 42(4): 841–876. Dahal, Mahesh, and Quynh Nguyen. 2014. Private (Non-State) Sector Engagement in the Provision of Educational Services at the Primary and Secondary Levels in South Asia: An Analytical Review of its Role in School Enrolment and Student Achievement. Manuscript. Washington, D.C.: World Bank. Das, Jishnu, Priyanka Pandey, and Tristan Zajonc. 2006. “Learning Levels and Gaps in Pakistan.” World Bank Policy Research Working Paper 4067. Washington, D.C.: World Bank. Elbers, Chris, Peter Lanjouw, Johan A. Mistiaen, and Berk Ozler. “Reinterpreting Between- Group Inequality.” Journal of Economic Inequality 6: 231–245. Institute of Social and Policy Sciences (I-SAPS). 2010. Private Sector Education in Pakistan: Mapping and Musing. Islamabad: I-SAPS. 30 Maurer, Markus. 2012. “Structural Elaboration of Technical and Vocational Education and Training Systems in Developing Countries: the Cases of Sri Lanka and Bangladesh.” Comparative Education 48(4): 487–503. Ministry of Education, Government of Pakistan. 2009. National Education Policy 2009. Islamabad: Government of Pakistan. Available at: http://planipolis.iiep.unesco.org/upload/Pakistan/Pakistan_National_education_policy_2009.pdf (Last Accessed: May 15, 2013). U.S. Department of Education, Office of Innovation and Improvement. 2009. State Regulation of Private Schools. Washington, D.C.: U.S. Department of Education. World Bank. 2005. Pakistan: Country Gender Assessment, Bridging the Gender Gap, Opportunities and Challenges. Washington, DC: World Bank. 31 Table 1. Variable definitions and construction No. Variable Definition Construction 1. Age Child’s age in completed As recorded in the survey. years 2. Female Child female dummy As recorded in the survey. (0=male; 1=female) 3. Rural Household rural dummy As recorded in the survey. (0=urban; 1=rural) 4. Household (hh) Household wealth Collapsing the dataset to the household level, asset index quintiles. Five categories; a province-specific normalized household quintiles First (lowest), second, third asset index was constructed via Principal (mid), fourth, and fifth Components Analysis, using household (highest). sampling weights. The components included whether the household owns the home, the number of rooms in the household’s home, whether the main source of lighting is electricity, whether the main source of fuel for cooking is gas/electricity, whether the main source of drinking water is piped water, whether the toilet facility is of a flush type, whether the household has a fridge, a computer, a TV, an air conditioner, and a music player. Households were then split into asset index quintiles. The quintile for the household was then assigned to all children in the six to 15 age group in the household. 5. Household (hh) Highest grade of education Highest grade of education completed was head’s highest completed. Five categories: constructed using information on the education No. highest grade ever completed if the schooling (0 grades); household head is not currently in school. If grades 1–5 (primary school the household head is currently in school, grades); grades 6–8 information on the current grade is used to (middle school grades); assign the individual the preceding grade for grades 9–10 (secondary this variable. Using this continuous variable, school grades); grades 11+ household heads are split into the five (higher secondary grades categories of highest education completed. and above). The household head’s category is then assigned to all children in the six to 15 age group in the household. 6. Household size Number of members in the The sum of all individuals on the household household roster. The value is assigned to all children in the six to 15 age group in the household. 7. Number of Number of child The sum of children in the given age group on children in the members in the household the household roster. The value is assigned to household in a in the given age group (6– all children in the six to 15 age group in the given age group 10, 11–15). household. 32 Panel A. 6–10 age group Panel B. 11–15 age group Figure 1. Distribution of children in different schooling statuses, by province, 2010/11 33 Panel A. 6–10 age group Panel B. 11–15 age group Figure 2. Distribution of children in different schooling statuses, by socioeconomic subgroup, 2010/11 34 Panel A. 6–10 age group Panel B. 11–15 age group Figure 3. Percent of children in private schools, by state/province 35 Panel A. 6–10 age group Panel B. 11–15 age group Figure 4. Percent of children in unaided private schools, by state/province 36 Table 2. Mean characteristics of private school students, Pakistan, 2010/11 Characteristic 6–10 age group 11–15 age group Private Diff. from Diff. from Private Diff. from Diff. from school govt. out-of- school govt. out-of- students school school students school school students children students children (1) (2) (3) (4) (5) (6) Age (in complete years) 8.09 –0.12*** 0.52*** 12.83 –0.05*** –0.51*** (1.37) (0.02) (0.02) (1.37) (0.02) (0.02) Female 0.45 0.01** –0.11*** 0.44 0.05*** –0.13*** (0.50) (0.01) (0.01) (0.50) (0.01) (0.01) Rural 0.45 –0.34*** –0.37*** 0.42 –0.28*** –0.38*** (0.50) (0.01) (0.01) (0.49) (0.01) (0.01) Lowest (first) hh asset index quintile 0.06 –0.19*** –0.37*** 0.04 –0.12*** –0.32*** (0.24) (0.01) (0.01) (0.20) (0.01) (0.01) Mid (third) hh asset index quintile 0.17 –0.05*** 0.01* 0.15 –0.08*** –0.04*** (0.38) (0.01) (0.01) (0.36) (0.01) (0.01) Highest (fifth) hh asset index quintile 0.37 0.28*** 0.32*** 0.44 0.28*** 0.39*** (0.48) (0.01) (0.01) (0.50) (0.01) (0.01) Hh head: highest ed.: no schooling 0.26 –0.19*** –0.39*** 0.24 –0.16*** –0.42*** (0.44) (0.01) (0.01) (0.43) (0.01) (0.01) Hh head: highest ed.: grades 1–5 0.15 –0.05*** –0.01* 0.13 –0.05*** –0.03*** (0.35) (0.01) (0.01) (0.34) (0.01) (0.01) Hh head: highest ed.: grades 6–8 0.14 0.02*** 0.06*** 0.13 0 0.07*** (0.35) (0.01) (0.00) (0.34) (0.01) (0.01) Hh head: highest ed.: grades 9–10 0.23 0.09*** 0.15*** 0.23 0.07*** 0.16*** (0.42) (0.01) (0.01) (0.42) (0.01) (0.01) Hh head: highest ed.: grades 11+ 0.23 0.13*** 0.19*** 0.26 0.14*** 0.22*** (0.42) (0.01) (0.01) (0.44) (0.01) (0.01) Hh size 7.76 –0.32*** –0.43*** 7.5 –0.51*** –0.68*** (3.46) (0.06) (0.06) (3.13) (0.06) (0.06) # of children ages 6 –10 years in hh 1.99 –0.19*** –0.29*** 1.1 –0.20*** –0.34*** (0.95) (0.02) (0.02) (1.07) (0.02) (0.02) # of children ages 11 –15 years in hh 0.92 –0.22*** –0.16*** 1.87 –0.09*** –0.11*** (1.00) (0.02) (0.02) (0.82) (0.02) (0.02) Notes: hh denotes household. Pakistan comprises of the four provinces only. Standard deviations are reported in parentheses in Columns (1) and (2). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6); they are estimated accounting for clustering at the PSU level. *** denotes p<0.01; ** p<0.05; and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 37 Table 3. Mean characteristics of private school students, 6–10 age group, by province, 2010/11 Characteristic P S KP B P–S P–KP P–B S–KP S–B KP–B (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Age (in complete years) 8.08 8.11 8.14 8.1 –0.03 –0.07* –0.02 –0.03 0.01 0.05 (1.37) (1.41) (1.33) (1.42) (0.03) (0.03) (0.09) (0.04) (0.10) (0.10) Female 0.45 0.46 0.38 0.39 –0.01 0.07*** 0.06* 0.08*** 0.07** –0.01 (0.50) (0.50) (0.49) (0.49) (0.01) (0.01) (0.03) (0.02) (0.04) (0.04) Rural 0.52 0.1 0.66 0.15 0.43*** –0.14*** 0.37*** –0.56*** –0.06 0.50*** (0.50) (0.29) (0.47) (0.36) (0.02) (0.03) (0.04) (0.03) (0.04) (0.05) Lowest (first) hh asset index quintile 0.08 0.01 0.03 0.02 0.07*** 0.05*** 0.06*** –0.02** 0 0.01 (0.27) (0.12) (0.17) (0.13) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Mid (third) hh asset index quintile 0.19 0.13 0.15 0.01 0.05*** 0.03** 0.18*** –0.02 0.13*** 0.15*** (0.39) (0.34) (0.36) (0.08) (0.01) (0.01) (0.01) (0.02) (0.01) (0.01) Highest (fifth) hh asset index quintile 0.31 0.5 0.47 0.87 –0.19*** –0.16*** –0.55*** 0.03 –0.36*** –0.39*** (0.46) (0.50) (0.50) (0.34) (0.02) (0.02) (0.03) (0.03) (0.04) (0.04) Hh head: highest ed.: no schooling 0.28 0.15 0.34 0.18 0.13*** –0.07*** 0.09*** –0.19*** –0.03 0.16*** (0.45) (0.35) (0.47) (0.39) (0.01) (0.02) (0.03) (0.02) (0.03) (0.04) Hh head: highest ed.: grades 1–5 0.16 0.12 0.09 0.13 0.05*** 0.07*** 0.03 0.02* –0.01 –0.04 (0.37) (0.32) (0.29) (0.34) (0.01) (0.01) (0.02) (0.01) (0.02) (0.02) Hh head: highest ed.: grades 6–8 0.16 0.1 0.11 0.09 0.05*** 0.04*** 0.06** –0.01 0.01 0.02 (0.36) (0.30) (0.32) (0.29) (0.01) (0.01) (0.03) (0.01) (0.03) (0.03) Hh head: highest ed.: grades 9–10 0.24 0.2 0.22 0.15 0.03*** 0.02 0.08** –0.02 0.05 0.06* (0.42) (0.40) (0.41) (0.36) (0.01) (0.02) (0.03) (0.02) (0.03) (0.04) Hh head: highest ed.: grades 11+ 0.17 0.43 0.24 0.44 –0.26*** –0.07*** –0.27*** 0.19*** –0.01 –0.20*** (0.38) (0.50) (0.43) (0.50) (0.02) (0.02) (0.05) (0.02) (0.05) (0.05) Hh size 7.66 7.24 9.27 8.16 0.42*** –1.61*** –0.5 –2.03*** –0.92** 1.11*** (3.23) (3.14) (4.74) (3.73) (0.13) (0.21) (0.37) (0.23) (0.38) (0.41) # of children ages 6 –10 years in hh 1.97 1.91 2.25 2.17 0.06 –0.28*** –0.20** –0.34*** –0.26*** 0.08 (0.92) (0.90) (1.16) (0.95) (0.04) (0.05) (0.09) (0.06) (0.09) (0.10) # of children ages 11 –15 years in hh 0.9 0.88 1.14 0.98 0.02 –0.24*** –0.08 –0.26*** –0.1 0.16* (0.99) (0.97) (1.09) (0.99) (0.03) (0.04) (0.08) (0.05) (0.09) (0.09) Notes: hh denotes household. P denotes Punjab, S Sindh, KP Khyber Pakhtunkhwa, and B Balochistan. Standard deviations are reported in parentheses in Columns (1)–(4). Standard errors are reported in parentheses in Columns (5)–(10); they are estimated accounting for clustering at the PSU level. *** denotes p<0.01; ** p<0.05; and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 38 Table 4. Mean characteristics of private school students, 11–15 age group, by province, 2010/11 Characteristic P S KP B P–S P–KP P–B S–KP S–B KP–B (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Age (in complete years) 12.79 12.92 12.87 13 –0.13*** –0.08** –0.21** 0.06 –0.08 –0.13 (1.36) (1.38) (1.39) (1.32) (0.04) (0.04) (0.08) (0.05) (0.09) (0.09) Female 0.46 0.47 0.31 0.25 0 0.16*** 0.21*** 0.16*** 0.21*** 0.05 (0.50) (0.50) (0.46) (0.44) (0.02) (0.02) (0.04) (0.02) (0.04) (0.04) Rural 0.52 0.05 0.64 0.21 0.47*** –0.12*** 0.32*** –0.59*** –0.15*** 0.43*** (0.50) (0.23) (0.48) (0.41) (0.02) (0.03) (0.06) (0.03) (0.05) (0.06) Lowest (first) hh asset index quintile 0.06 0.01 0.03 0.01 0.05*** 0.03*** 0.05*** –0.02*** 0 0.02** (0.23) (0.08) (0.17) (0.10) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Mid (third) hh asset index quintile 0.18 0.10 0.14 0.03 0.08*** 0.04** 0.14*** –0.04** 0.07*** 0.11*** (0.38) (0.30) (0.35) (0.18) (0.01) (0.02) (0.02) (0.02) (0.02) (0.02) Highest (fifth) hh asset index quintile 0.37 0.57 0.52 0.89 –0.20*** –0.15*** –0.53*** 0.05 –0.33*** –0.38*** (0.48) (0.50) (0.50) (0.31) (0.03) (0.03) (0.03) (0.03) (0.04) (0.04) Hh head: highest ed.: no schooling 0.28 0.12 0.31 0.18 0.17*** –0.02 0.13** –0.19*** –0.06 0.13*** (0.45) (0.32) (0.46) (0.38) (0.01) (0.02) (0.04) (0.02) (0.04) (0.04) Hh head: highest ed.: grades 1–5 0.16 0.09 0.09 0.06 0.07*** 0.07*** 0.10*** 0.01 0.03 0.03 (0.37) (0.29) (0.28) (0.24) (0.01) (0.01) (0.02) (0.01) (0.02) (0.02) Hh head: highest ed.: grades 6–8 0.14 0.13 0.11 0.1 0.01 0.04*** 0.04 0.03* 0.03 0 (0.35) (0.34) (0.31) (0.30) (0.01) (0.01) (0.03) (0.02) (0.03) (0.03) Hh head: highest ed.: grades 9–10 0.24 0.21 0.22 0.16 0.03* 0.02 0.08** –0.01 0.05 0.06 (0.43) (0.41) (0.41) (0.36) (0.02) (0.02) (0.04) (0.02) (0.04) (0.04) Hh head: highest ed.: grades 11+ 0.18 0.45 0.29 0.5 –0.27*** –0.11*** –0.33*** 0.16*** –0.05 –0.22*** (0.38) (0.50) (0.45) (0.50) (0.02) (0.02) (0.06) (0.03) (0.06) (0.06) Hh size 7.48 6.98 8.63 7.86 0.51*** –1.14*** –0.38 –1.65*** –0.88*** 0.76** (2.96) (2.85) (4.10) (2.94) (0.13) (0.19) (0.30) (0.21) (0.31) (0.34) # of children ages 6 –10 years in hh 1.1 0.95 1.34 1.29 0.15*** –0.23*** –0.19* –0.38*** –0.34*** 0.05 (1.05) (1.00) (1.28) (1.06) (0.04) (0.05) (0.11) (0.06) (0.11) (0.12) # of children ages 11 –15 years in hh 1.88 1.8 1.97 1.94 0.09*** –0.08** –0.06 –0.17*** –0.15** 0.02 (0.82) (0.77) (0.89) (0.75) (0.03) (0.04) (0.07) (0.04) (0.07) (0.07) Notes: hh denotes household. P denotes Punjab, S Sindh, KP Khyber Pakhtunkhwa, and B Balochistan. Standard deviations are reported in parentheses in Columns (1)–(4). Standard errors are reported in parentheses in Columns (5)–(10); they are estimated accounting for clustering at the PSU level. *** denotes p<0.01; ** p<0.05; and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 39 Table 5. Characteristics of top-ten group vs. non-top-ten group Non-top-ten Indicator Top-ten group group Group share of private school students, 6–10 age group (%) 51 49 Group share of private school students, 11–15 age group (%) 57 43 Group share of total population, 6–10 age group (%) 25 75 Group share of total population, 11–15 age group (%) 29 71 Private school participation rate in group, 6–10 age group (%) 44 14 Private school participation rate in group, 11–15 age group (%) 36 11 Govt. school participation rate in group, 6–10 age group (%) 32 50 Govt. school participation rate in group, 11–15 age group (%) 41 49 Urban share in group (%) 62 21 Mean household asset index in group 0.72 –0.19 Notes: The top-ten group comprises of Karachi, Lahore, Gujranwala, Faisalabad, Sialkot, Rawalpindi, Multan, Sheikhupura, Gujrat, and Peshawar. The non-top-ten group comprises of the remaining 103 districts. The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 40 Table 6. Characteristics of private school students, by age group and top-ten group vs. non- top-ten group, 2010/11 6–10 age group 11–15 age group Top-ten Difference Top-ten Difference group from the group from the Characteristic non-top-ten non-top-ten group group (1) (2) (3) (4) Age (in complete years) 8.11 0.04 12.85 0.03 (1.38) (0.02) (1.37) (0.03) Female 0.47 0.04*** 0.48 0.07*** (0.50) (0.01) (0.50) (0.01) Rural 0.31 –0.28*** 0.28 –0.33*** (0.46) (0.02) (0.45) (0.03) Lowest (first) hh asset index quintile 0.02 –0.08*** 0.02 –0.05*** (0.15) (0.01) (0.13) (0.01) Mid (third) hh asset index quintile 0.13 –0.08*** 0.12 –0.08*** (0.34) (0.01) (0.33) (0.01) Highest (fifth) hh asset index quintile 0.46 0.18*** 0.53 0.21*** (0.50) (0.02) (0.50) (0.02) Hh head: highest ed.: no schooling 0.23 –0.06*** 0.21 –0.07*** (0.42) (0.01) (0.41) (0.01) Hh head: highest ed.: grades 1–5 0.13 –0.04*** 0.12 –0.03*** (0.33) (0.01) (0.33) (0.01) Hh head: highest ed.: grades 6–8 0.15 0.01 0.14 0.02** (0.35) (0.01) (0.35) (0.01) Hh head: highest ed.: grades 9–10 0.24 0.03*** 0.24 0.03** (0.43) (0.01) (0.43) (0.01) Hh head: highest ed.: grades 11+ 0.26 0.05*** 0.28 0.06*** (0.44) (0.01) (0.45) (0.02) Hh size 7.45 –0.62*** 7.19 –0.73*** (3.29) (0.11) (2.98) (0.11) # of children in the 6 –10 age group in hh 1.93 –0.13*** 0.99 –0.24*** (0.93) (0.03) (1.01) (0.04) # of children in the 11 –15 age group in hh 0.88 –0.08*** 1.84 –0.07** (0.99) (0.03) (0.81) (0.03) Notes: The top-ten group comprises of Karachi, Lahore, Gujranwala, Faisalabad, Sialkot, Rawalpindi, Multan, Sheikhupura, Gujrat, and Peshawar. The non-top-ten group comprises of the remaining 103 districts. hh denotes household. Standard deviations are presented in parentheses in Columns (1) and (3). Standard errors are presented in parenthesis in Columns (2) and (4). The standard errors are estimated accounting for clustering at the PSU level. *** denotes p<0.01, ** p<0.05, and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 41 Figure 5. Distribution of private school students, 6–10 age group, Pakistan, 2010/11 42 Figure 6. Distribution of private school students, 11–15 age group, Pakistan, 2010/11 43 Table 7. Decomposition of the variation in school participation, by school type, 2010/11 Households with multiple children and at least one child in school in each age group Percent of total variation in private Percent of total variation in Province school participation government school participation Between- Within-household Between- Within-household household household (1) (2) (3) (4) Panel A. 6–10 age group Pakistan 82 18 66 34 Punjab 81 19 70 30 Sindh 86 14 66 34 Khyber Pakhtunkhwa 77 23 57 43 Balochistan 88 12 44 56 Panel B. 11–15 age group Pakistan 79 21 60 40 Punjab 75 25 60 40 Sindh 89 11 64 36 Khyber Pakhtunkhwa 79 21 55 45 Balochistan 83 17 43 57 Notes: Pakistan comprises of the four provinces only. The sample for Panel A is households with multiple children in the 6–10 age group; the sample for Panel B is households with multiple children in the 11–15 age group. The estimated shares attributable to within-household variation in (private/government) school participation also include noise and, thus, are likely to be overestimates of the actual shares of within-household variation in (private/government) school participation. In each row, the estimated shares in columns (1) and (2) sum to 100%. In each row, the estimated shares in columns (3) and (4) sum to 100%. The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 44 Table 8. Distribution of households in terms of the extent of private schooling across in-school children within households, 2010/11 Households with multiple children and with at least one in school Mean number Mean percent Percent of households with of children in of children in All in-school Some in- No in-school Province household household in children in school children in school private school children in private school private school (1) (2) (3) (4) (5) Panel A. 6–10 age group Pakistan 2.5 82 25 5 70 Punjab 2.4 83 31 7 62 Sindh 2.5 81 20 3 77 Khyber Pakhtunkhwa 2.7 77 20 5 75 Balochistan 2.4 79 4 1 95 Panel B. 11–15 age group Pakistan 2.3 82 18 10 72 Punjab 2.3 83 20 13 67 Sindh 2.3 81 20 5 75 Khyber Pakhtunkhwa 2.4 80 13 9 78 Balochistan 2.3 75 4 2 94 Notes: Pakistan comprises of the four provinces only. The sample for Panel A is households with multiple children in the 6–10 age group and at least one of them in school; the sample for Panel B is households with multiple children in the 11–15 age group and at least one of them in school. In each row, the percentages in Columns (3)–(5) sum to 100%. The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 45 Table 9. Mean characteristics of households in groups in terms of the extent of private schooling across in-school children within households, Pakistan, 2010/11 Households with multiple children and with at least one in school In-school children, 6–10 age group In-school children, 11–15 age group Characteristic All in private Some in None in All in private Some in None in school private school private school school private school private school (1) (2) (3) (4) (5) (6) Rural 0.46 0.66 0.81 0.43 0.61 0.73 (0.50) (0.47) (0.40) (0.49) (0.49) (0.44) Lowest (first) hh asset index quintile 0.06 0.09 0.25 0.04 0.04 0.16 (0.24) (0.28) (0.43) (0.20) (0.20) (0.37) Mid (third) hh asset index quintile 0.18 0.24 0.24 0.17 0.24 0.25 (0.38) (0.43) (0.43) (0.37) (0.43) (0.43) Highest (fifth) hh asset index quintile 0.38 0.26 0.08 0.43 0.31 0.13 (0.48) (0.44) (0.27) (0.49) (0.46) (0.33) Hh head: highest ed.: no schooling 0.29 0.37 0.48 0.28 0.32 0.45 (0.45) (0.48) (0.50) (0.45) (0.46) (0.50) Hh head: highest ed.: grades 1–5 0.15 0.2 0.19 0.15 0.14 0.19 (0.35) (0.40) (0.39) (0.35) (0.35) (0.39) Hh head: highest ed.: grades 6–8 0.15 0.13 0.11 0.13 0.16 0.12 (0.35) (0.34) (0.31) (0.33) (0.37) (0.33) Hh head: highest ed.: grades 9–10 0.21 0.2 0.12 0.21 0.22 0.14 (0.41) (0.40) (0.33) (0.41) (0.41) (0.35) Hh head: highest ed.: grades 11+ 0.21 0.11 0.09 0.23 0.16 0.1 (0.41) (0.31) (0.29) (0.42) (0.37) (0.29) Hh size 9.19 11.02 9.19 8.76 9.79 9.31 (4.21) (4.95) (3.57) (3.68) (4.49) (3.64) # of children ages 6–10 years in hh 2.41 2.82 2.51 1.3 1.47 1.55 (0.76) (1.01) (0.81) (1.17) (1.35) (1.23) # of children ages 11–15 years in hh 1.02 1.35 1.29 2.27 2.47 2.33 (1.06) (1.14) (1.06) (0.56) (0.71) (0.62) Share of children in 6–10 (11–15) age 0.84 0.94 0.8 0.87 0.97 0.78 group that are in school (0.23) (0.14) (0.25) (0.22) (0.09) (0.25) Notes: hh denotes household. Standard deviations are provided in parentheses. The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 46 Table 10. Parameter estimates from private school participation regressions, 2010/11 Households with multiple children and at least one child in school Variable Pakistan Punjab Sindh KP Balochistan (1) (2) (2) (3) (4) (5) (6) (7) (8) (9) Panel A. Children, 6–10 age group Age (in complete years) 0.02*** 0.02*** 0.02*** 0.02*** 0.02*** 0.02*** 0.02*** 0.03*** 0.00 0.00 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Female –0.02*** –0.02*** –0.03*** –0.02*** –0.01 –0.00 –0.05*** –0.05*** –0.01* –0.01** (0.00) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Household-level covariates Yes No Yes No Yes No Yes No Yes No Household dummies No Yes No Yes No Yes No Yes No Yes R-squared 0.20 0.70 0.17 0.67 0.37 0.77 0.20 0.63 0.14 0.80 Number of children 46,864 46,864 16,818 16,818 12,198 12,198 9,459 9,459 8,389 8,389 Panel B. Children, 11–15 age group Age (in complete years) –0.02*** –0.02*** –0.02*** –0.02*** –0.02*** –0.01*** –0.01*** –0.01** –0.00 –0.00** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Female –0.01** –0.03*** 0.00 –0.02 –0.01 –0.01 –0.09*** –0.10*** –0.03*** –0.02*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Household-level covariates Yes No Yes No Yes No Yes No Yes No Household dummies No Yes No Yes No Yes No Yes No Yes R-squared 0.15 0.63 0.10 0.55 0.31 0.81 0.20 0.65 0.12 0.69 Number of children 33,246 33,246 13,476 13,476 7,695 7,695 7,290 7,290 4,785 4,785 Notes: Pakistan comprises of the four provinces only. KP denotes Khyber Pakhtunkhwa. Standard errors are reported in parentheses. The standard errors are estimated accounting for clustering at the PSU level. *** denotes p<0.01; ** p<0.05; and * p<0.10 (two-tailed significance tests). Household-level covariates comprise of household location (urban/rural), wealth (in asset index quintiles), household head’s highest education, household size, and number of children in different age groups. The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 47 Table 11. Evolution of overall and private school participation rates, by age group, 1998/99–2010/11 6–10 age group 11–15 age group Private Private ∆ in school ∆ in private school ∆ in school ∆ in private school Area/group PR school PR share of ∆ PR school PR share of ∆ (in ppts) (in ppts) in school (in ppts) (in ppts) in school PR (in %) PR (in %) (1) (2) (3) (4) (5) (6) Panel A. Country/province Pakistan 16.9 8.8 51.8 14.1 8.9 63.4 Punjab 18.4 11.2 61.1 15.3 10.3 67.2 Sindh 15.5 5.9 37.9 11.5 8.3 72.1 KP 16.6 8.5 51.0 17.0 7.8 45.9 Balochistan 12.2 0.8 6.8 4.4 0.8 17.2 Panel B. Socioeconomic subgroup Female 18.5 8.4 45.5 16.6 9.2 55.5 Male 15.1 9.0 59.9 11.0 8.6 77.7 Rural 18.4 7.3 39.7 15.1 7.2 47.7 Urban 12.0 11.2 93.3 11.1 12.1 108.8 Lowest quintile 13.0 3.0 22.6 11.0 2.2 20.4 Mid quintile 14.9 7.9 52.9 14.0 7.3 52.0 Highest quintile 6.9 11.0 160.4 5.3 16.3 309.3 Notes: ppts denotes percentage points; PR denotes participation rate; and KP is Khyber Pakhtunkhwa. Pakistan comprises of the four provinces only. The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey and the 1998/99 Pakistan Integrated Household Survey (PIHS). All statistics are estimated accounting for survey sampling weights. 48 Table 12. Mean characteristics of private school students, Pakistan, 1998/99, 2004/05, and 2010/11 6–10 age group 11–15 age group 2010/11 Diff. from Diff. from 2010/11 Diff. from Diff. from Characteristic 2004/05 1998/99 2004/05 1998/99 (1) (2) (3) (4) (5) (6) Age (in complete years) 7.915 0.023 0.023 12.832 0.064*** 0.234*** (1.411) (0.018) (0.033) (1.370) (0.024) (0.050) Female 0.444 –0.002 –0.013 0.444 –0.008 0.014 (0.497) (0.007) (0.012) (0.497) (0.011) –0.021 Rural 0.450 0.002 0.091** 0.423 0.006 0.115*** (0.498) (0.033) (0.041) (0.494) (0.036) (0.044) Lowest (first) hh asset index quintile 0.070 0.022*** 0.018* 0.045 0.011** 0.006 (0.256) (0.007) (0.010) (0.208) (0.005) (0.010) Mid (third) hh asset index quintile 0.199 0.018* 0.009 0.172 0.025** 0.023 (0.399) (0.010) (0.016) (0.378) (0.011) (0.018) Highest (fifth) hh asset index quintile 0.329 –0.056*** –0.109*** 0.408 –0.048** –0.094*** (0.470) (0.016) (0.025) (0.491) (0.020) (0.030) Hh head: highest ed.: no schooling 0.264 –0.049*** –0.024 0.245 –0.036*** –0.025 (0.441) (0.012) (0.020) (0.430) (0.014) (0.022) Hh head: highest ed.: grades 1–5 0.148 –0.009 –0.028* 0.135 –0.001 –0.028* (0.355) (0.008) (0.014) (0.342) (0.009) (0.017) Hh head: highest ed.: grades 6–8 0.222 0.013 0.011 0.228 0.002 –0.012 (0.416) (0.008) (0.015) (0.420) (0.010) (0.020) Hh head: highest ed.: grades 9–10 0.142 –0.002 0.007 0.134 –0.006 0.029** (0.349) (0.007) (0.012) (0.341) (0.009) (0.014) Hh head: highest ed.: grades 11+ 0.224 0.047*** 0.034 0.258 0.041** 0.036 (0.417) (0.012) (0.023) (0.437) (0.017) (0.024) Hh size 7.773 –1.958*** –0.563*** 7.499 –1.649*** –0.813*** (3.479) (0.139) (0.161) (3.128) (0.134) (0.227) # of children 6–10 age group in hh 1.997 –0.303*** –0.088** 1.094 –0.299*** –0.251*** (0.952) (0.036) (0.044) (1.072) (0.037) (0.054) # of children 11–15 age group in hh 0.918 –0.239*** –0.155*** 1.874 –0.211*** –0.120** (1.005) (0.026) (0.040) (0.823) (0.028) (0.050) Notes: hh denotes household. Pakistan comprises of the four provinces only. Standard deviations are reported in parentheses in Columns (1) and (4). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6); they are estimated accounting for clustering at the PSU level. *** denotes p<0.01; ** p<0.05; and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 and 2004/05 Pakistan Social and Living Standards Measurement (PSLM) surveys and the 1998/99 Pakistan Integrated Household Survey (PIHS). All statistics are estimated accounting for survey sampling weights. 49 APPENDIX Table A1. Mean characteristics of private school students, Punjab, 2010/11 6–10 age group 11–15 age group Private Diff. Diff. Private Diff. Diff. school from from out- school from from out- Characteristic students govt. of-school students govt. of-school school children school children students students (1) (2) (3) (4) (5) (6) Age (in complete years) 8.08 –0.16*** 0.56*** 12.79 –0.11*** –0.59*** (1.37) (0.02) (0.03) (1.36) (0.02) (0.03) Female 0.45 –0.02** –0.08*** 0.46 0.02** –0.07*** (0.50) (0.01) (0.01) (0.50) (0.01) (0.01) Rural 0.52 –0.28*** –0.30*** 0.52 –0.18*** –0.28*** (0.50) (0.01) (0.01) (0.50) (0.02) (0.02) Lowest (first) hh asset index quintile 0.08 –0.19*** –0.41*** 0.06 –0.11*** –0.33*** (0.27) (0.01) (0.01) (0.23) (0.01) (0.01) Mid (third) hh asset index quintile 0.19 –0.01 0.07*** 0.18 –0.04*** 0.02* (0.39) (0.01) (0.01) (0.38) (0.01) (0.01) Highest (fifth) hh asset index quintile 0.31 0.23*** 0.26*** 0.37 0.21*** 0.32*** (0.46) (0.01) (0.01) (0.48) (0.01) (0.01) Hh head: highest ed.: no schooling 0.28 –0.19*** –0.38*** 0.28 –0.12*** –0.38*** (0.45) (0.01) (0.01) (0.45) (0.01) (0.01) Hh head: highest ed.: grades 1–5 0.16 –0.03*** 0.01 0.16 –0.03*** –0.01 (0.37) (0.01) (0.01) (0.37) (0.01) (0.01) Hh head: highest ed.: grades 6–8 0.16 0.01 0.07*** 0.14 –0.01* 0.06*** (0.36) (0.01) (0.01) (0.35) (0.01) (0.01) Hh head: highest ed.: grades 9–10 0.24 0.09*** 0.16*** 0.24 0.06*** 0.17*** (0.42) (0.01) (0.01) (0.43) (0.01) (0.01) Hh head: highest ed.: grades 11+ 0.17 0.12*** 0.14*** 0.18 0.10*** 0.16*** (0.38) (0.01) (0.01) (0.38) (0.01) (0.01) Hh size 7.66 –0.16** –0.25*** 7.48 –0.24*** –0.40*** (3.23) (0.07) (0.08) (2.96) (0.07) (0.08) # of children ages 6 –10 years in hh 1.97 –0.14*** –0.24*** 1.1 –0.06** –0.21*** (0.92) (0.02) (0.02) (1.05) (0.03) (0.03) # of children ages 11 –15 years in hh 0.9 –0.22*** –0.13*** 1.88 –0.05** –0.06*** (0.99) (0.02) (0.02) (0.82) (0.02) (0.02) Notes: hh denotes household. Standard deviations are reported in parentheses in Columns (1) and (4). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6). The standard errors are estimated accounting for clustering at the PSU level. *** denotes p<0.01, ** p<0.05, and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 50 Table A2. Mean characteristics of private school students, Sindh, 2010/11 6–10 age group 11–15 age group Private Diff. Diff. Private Diff. from Diff. school from from out- school govt. from out- Characteristic students govt. of-school students school of-school school children students children students (1) (2) (3) (4) (5) (6) Age (in complete years) 8.11 –0.02 0.41*** 12.92 0.05 –0.40*** (1.41) (0.03) (0.04) (1.38) (0.04) (0.04) Female 0.46 0.07*** –0.09*** 0.47 0.13*** –0.11*** (0.50) (0.01) (0.01) (0.50) (0.02) (0.02) Rural 0.1 –0.62*** –0.66*** 0.05 –0.51*** –0.68*** (0.29) (0.02) (0.02) (0.23) (0.02) (0.02) Lowest (first) hh asset index quintile 0.01 –0.22*** –0.40*** 0.01 –0.14*** –0.34*** (0.12) (0.01) (0.02) (0.07) (0.01) (0.01) Mid (third) hh asset index quintile 0.13 –0.14*** –0.05*** 0.1 –0.18*** –0.12*** (0.34) (0.01) (0.01) (0.30) (0.01) (0.01) Highest (fifth) hh asset index quintile 0.5 0.45*** 0.46*** 0.57 0.44*** 0.54*** (0.50) (0.02) (0.02) (0.50) (0.02) (0.02) Hh head: highest ed.: no schooling 0.15 –0.20*** –0.45*** 0.12 –0.18*** –0.48*** (0.35) (0.01) (0.01) (0.32) (0.01) (0.01) Hh head: highest ed.: grades 1–5 0.12 –0.13*** –0.08*** 0.09 –0.13*** –0.12*** (0.32) (0.01) (0.01) (0.29) (0.01) (0.01) Hh head: highest ed.: grades 6–8 0.1 0.03*** 0.04*** 0.13 0.04*** 0.08*** (0.30) (0.01) (0.01) (0.34) (0.01) (0.01) Hh head: highest ed.: grades 9–10 0.2 0.07*** 0.13*** 0.21 0.05*** 0.14*** (0.40) (0.01) (0.01) (0.41) (0.02) (0.01) Hh head: highest ed.: grades 11+ 0.43 0.23*** 0.36*** 0.45 0.22*** 0.39*** (0.50) (0.02) (0.02) (0.50) (0.02) (0.02) Hh size 7.24 –0.92*** –1.01*** 6.98 –1.26*** –1.42*** (3.14) (0.12) (0.13) (2.85) (0.12) (0.11) # of children ages 6 –10 years in hh 1.91 –0.34*** –0.43*** 0.95 –0.48*** –0.59*** (0.90) (0.03) (0.04) (1.00) (0.04) (0.04) # of children ages 11 –15 years in hh 0.88 –0.24*** –0.22*** 1.8 –0.17*** –0.18*** (0.97) (0.03) (0.03) (0.77) (0.03) (0.03) Notes: hh denotes household. Standard deviations are reported in parentheses in Columns (1) and (4). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6). The standard errors are estimated accounting for clustering at the PSU level. *** denotes p<0.01, ** p<0.05, and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 51 Table A3. Mean characteristics of private school students in Khyber Pakhtunkhwa, 2010/11 6–10 age group 11–15 age group Private Diff. Diff. Private Diff. from Diff. school from from out- school govt. from out- Characteristic students govt. of-school students school of-school school children students children students (1) (2) (3) (4) (5) (6) Age (in complete years) 8.14 –0.15*** 0.73*** 12.87 –0.04 –0.44*** (1.33) (0.04) (0.04) (1.39) (0.04) (0.04) Female 0.38 –0.03** –0.20*** 0.31 –0.06*** –0.41*** (0.49) (0.01) (0.02) (0.46) (0.02) (0.02) Rural 0.66 –0.23*** –0.23*** 0.64 –0.22*** –0.25*** (0.47) (0.02) (0.02) (0.48) (0.02) (0.03) Lowest (first) hh asset index quintile 0.03 –0.20*** –0.29*** 0.03 –0.16*** –0.29*** (0.17) (0.01) (0.02) (0.17) (0.01) (0.02) Mid (third) hh asset index quintile 0.15 –0.07*** –0.05*** 0.14 –0.08*** –0.08*** (0.36) (0.02) (0.02) (0.35) (0.01) (0.02) Highest (fifth) hh asset index quintile 0.47 0.37*** 0.38*** 0.52 0.36*** 0.44*** (0.50) (0.02) (0.02) (0.50) (0.02) (0.02) Hh head: highest ed.: no schooling 0.34 –0.22*** –0.33*** 0.31 –0.23*** –0.43*** (0.47) (0.02) (0.02) (0.46) (0.02) (0.02) Hh head: highest ed.: grades 1–5 0.09 –0.03** –0.01 0.09 –0.03** –0.01 (0.29) (0.01) (0.01) (0.28) (0.01) (0.01) Hh head: highest ed.: grades 6–8 0.11 –0.01 0.03*** 0.11 –0.02 0.03*** (0.32) (0.01) (0.01) (0.31) (0.01) (0.01) Hh head: highest ed.: grades 9–10 0.22 0.09*** 0.13*** 0.22 0.08*** 0.15*** (0.41) (0.01) (0.01) (0.41) (0.02) (0.02) Hh head: highest ed.: grades 11+ 0.24 0.16*** 0.19*** 0.29 0.19*** 0.26*** (0.43) (0.02) (0.01) (0.45) (0.02) (0.02) Hh size 9.27 0.18 0.09 8.63 –0.17 –0.39** (4.74) (0.21) (0.21) (4.10) (0.19) (0.20) # of children ages 6 –10 years in hh 2.25 –0.08 –0.13** 1.34 –0.16*** –0.30*** (1.16) (0.06) (0.05) (1.28) (0.05) (0.06) # of children ages 11 –15 years in hh 1.14 –0.18*** –0.13*** 1.97 –0.07** –0.12*** (1.09) (0.05) (0.04) (0.89) (0.03) (0.04) Notes: hh denotes household. Standard deviations are reported in parentheses in Columns (1) and (4). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6). The standard errors are estimated accounting for clustering at the PSU level. *** denotes p<0.01, ** p<0.05, and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 52 Table A4. Mean characteristics of private school students, Balochistan, 2010/11 6–10 age group 11–15 age group Private Diff. Diff. Private Diff. from Diff. school from from out- school govt. from out- Characteristic students govt. of-school students school of-school school children students children students (1) (2) (3) (4) (5) (6) Age (in complete years) 8.1 –0.1 0.35*** 13 0.22** –0.23*** (1.42) (0.09) (0.10) (1.32) (0.09) (0.09) Female 0.39 0.07* –0.22*** 0.25 0.03 –0.33*** (0.49) (0.04) (0.04) (0.44) (0.04) (0.04) Rural 0.15 –0.57*** –0.73*** 0.21 –0.47*** –0.67*** (0.36) (0.05) (0.04) (0.41) (0.05) (0.05) Lowest (first) hh asset index quintile 0.02 –0.16*** –0.28*** 0.01 –0.13*** –0.25*** (0.13) (0.02) (0.02) (0.10) (0.02) (0.02) Mid (third) hh asset index quintile 0.01 –0.18*** –0.18*** 0.03 –0.13*** –0.18*** (0.08) (0.01) (0.01) (0.18) (0.02) (0.02) Highest (fifth) hh asset index quintile 0.87 0.66*** 0.80*** 0.89 0.62*** 0.81*** (0.34) (0.04) (0.03) (0.31) (0.03) (0.03) Hh head: highest ed.: no schooling 0.18 –0.23*** –0.53*** 0.18 –0.26*** –0.58*** (0.39) (0.04) (0.04) (0.38) (0.04) (0.04) Hh head: highest ed.: grades 1–5 0.13 –0.07*** –0.02 0.06 –0.12*** –0.06*** (0.34) (0.02) (0.02) (0.24) (0.02) (0.02) Hh head: highest ed.: grades 6–8 0.09 –0.01 0.05* 0.1 0 0.06** (0.29) (0.03) (0.03) (0.30) (0.03) (0.03) Hh head: highest ed.: grades 9–10 0.15 0.03 0.10*** 0.16 0.04 0.11*** (0.36) (0.03) (0.03) (0.36) (0.04) (0.04) Hh head: highest ed.: grades 11+ 0.44 0.27*** 0.40*** 0.5 0.34*** 0.47*** (0.50) (0.05) (0.05) (0.50) (0.05) (0.05) Hh size 8.16 0.79** 0.66* 7.86 0.18 0.07 (3.73) (0.36) (0.35) (2.94) (0.29) (0.29) # of children ages 6 –10 years in hh 2.17 0.03 –0.09 1.29 –0.24** –0.19* (0.95) (0.09) (0.09) (1.06) (0.11) (0.11) # of children ages 11 –15 years in hh 0.98 0.01 –0.05 1.94 0.02 –0.07 (0.99) (0.09) (0.08) (0.75) (0.07) (0.06) Notes: hh denotes household. Standard deviations are reported in parentheses in Columns (1) and (4). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6). The standard errors are estimated accounting for clustering at the PSU level. *** denotes p<0.01, ** p<0.05, and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 53 Table A5. The distribution of private school students across districts, 6–10 age group, Pakistan, 2010/11 Percent of Private Govt. Urban rate Percent of private participation participation in district District child Province District school rate in rate in x's average hh. population students in district x district x population asset index in district x district x (in %) (in %) (in %) Sindh Karachi 13.20 54.04 17.71 5.25 96.26 1.22 Punjab Lahore 8.52 49.48 30.22 3.70 83.37 1.04 Punjab Gujranwala 5.84 54.96 25.85 2.28 52.08 0.50 Punjab Faisalabad 5.35 33.41 40.87 3.45 45.35 0.38 Punjab Sialkot 3.90 47.27 35.21 1.79 24.54 0.44 Punjab Rawalpindi 3.83 47.05 40.69 1.75 49.18 0.76 Punjab Multan 2.95 29.07 37.96 2.18 38.21 0.09 Punjab Sheikhupura 2.71 40.57 35.27 1.44 36.70 0.25 Punjab Gujrat 2.50 39.57 43.28 1.36 26.98 0.42 KP Peshawar 2.38 29.87 35.55 1.71 52.75 0.70 Punjab Kasur 2.20 27.84 45.18 1.70 25.80 –0.05 Punjab Rahim Yar Khan 2.13 17.08 35.60 2.69 21.84 –0.24 Punjab Narowal 2.11 38.80 46.55 1.17 14.17 –0.07 Punjab Sargodha 1.94 24.97 53.32 1.68 25.82 0.05 Punjab Vehari 1.84 23.13 43.16 1.71 18.19 –0.24 Sindh Hyderabad 1.77 35.47 42.36 1.07 80.54 0.83 Punjab Muzaffargarh 1.65 14.37 40.46 2.46 14.21 –0.59 Punjab Jhang 1.51 23.38 43.18 1.38 23.27 –0.40 Punjab Nankana Sahib 1.46 36.38 39.96 0.86 20.50 –0.01 Punjab Toba Tek Singh 1.39 29.16 52.82 1.03 20.86 0.28 Punjab Khanewal 1.39 20.09 52.83 1.49 18.78 –0.21 Punjab Bahawalpur 1.38 14.01 36.19 2.12 30.64 –0.21 Punjab Bahawalnagar 1.32 16.67 46.11 1.70 20.52 –0.17 Punjab Okara 1.29 17.03 57.84 1.63 14.89 –0.17 Punjab Sahiwal 1.10 17.09 56.82 1.38 15.98 0.00 Punjab D. G. Khan 1.04 12.38 44.74 1.81 12.31 –0.60 Punjab Chakwal 1.02 34.86 57.20 0.63 14.65 0.39 KP Manshera 0.90 20.75 47.00 0.93 6.61 –0.06 Punjab Attock 0.88 26.81 57.01 0.71 21.93 0.16 KP Mardan 0.87 15.49 52.75 1.21 21.01 0.00 Punjab Jehlum 0.85 31.33 56.89 0.59 25.85 0.45 KP Abbottabad 0.82 28.48 49.83 0.62 14.58 0.25 Punjab Mandi Bahauddin 0.82 24.27 59.32 0.72 15.30 0.12 KP Swat 0.82 17.18 40.89 1.02 13.64 0.00 Punjab Lodhran 0.79 18.05 41.14 0.94 13.53 –0.29 Punjab Layyah 0.79 15.61 55.50 1.08 15.84 –0.56 Punjab Pakpattan 0.75 14.33 55.97 1.13 15.33 –0.37 KP Charsada 0.75 18.02 48.79 0.89 17.15 0.05 KP Nowshera 0.73 21.78 52.76 0.72 24.44 0.14 Sindh Dadu 0.69 12.06 60.66 1.24 21.37 –0.03 Sindh Khairpur 0.61 8.28 59.91 1.58 27.72 –0.37 KP Swabi 0.59 16.76 58.46 0.76 18.34 –0.01 Punjab Hafizabad 0.54 20.24 59.88 0.57 30.78 0.12 KP Haripur 0.53 25.86 58.47 0.44 13.20 0.40 Sindh Sukkur 0.52 13.72 52.20 0.81 46.13 0.06 Punjab Khushab 0.50 18.23 58.58 0.58 27.38 0.16 Balochistan Quetta 0.49 17.65 61.04 0.59 78.74 1.00 Punjab Chiniot 0.46 16.58 45.98 0.60 28.79 –0.18 Punjab Mianwali 0.45 13.46 67.34 0.72 21.03 –0.14 KP D. I. Khan 0.45 10.47 36.98 0.92 13.13 –0.25 54 Table A5. The distribution of private school students across districts, 6–10 age group, Pakistan, 2010/11 Percent of Private Govt. Urban rate Percent of private participation participation in district District child Province District school rate in rate in x's average hh. population students in district x district x population asset index in district x district x (in %) (in %) (in %) Sindh Sanghar 0.39 6.62 54.39 1.27 26.37 –0.19 Sindh Ghotki 0.39 7.84 47.32 1.07 15.03 –0.36 Sindh Mirpurkhas 0.39 9.81 51.06 0.86 32.74 –0.07 Sindh Larkana 0.33 7.42 54.15 0.96 39.57 0.03 Punjab Bhakhar 0.33 7.28 54.84 0.98 15.67 –0.47 KP Batagram 0.31 20.35 45.66 0.33 0.00 0.00 KP Kohat 0.31 14.82 51.00 0.46 28.26 0.18 Sindh Naushahro Firoze 0.31 7.60 54.07 0.88 19.77 –0.22 Sindh Kambar Shahdadkot 0.26 5.56 46.07 1.00 19.25 –0.22 KP Karak 0.25 12.37 57.27 0.43 6.00 –0.36 Sindh Kashmore 0.21 4.61 47.94 0.99 19.06 –0.45 Sindh Shikarpur 0.21 4.75 50.02 0.94 21.86 –0.35 Punjab Rajanpur 0.20 3.43 49.58 1.26 12.34 –0.73 Sindh Jaccobabad 0.19 4.88 46.85 0.84 21.32 –0.54 KP Bannu 0.19 5.81 57.82 0.69 4.16 0.05 Sindh S. Benazirabad 0.18 4.65 52.85 0.82 31.42 –0.15 KP Malakand 0.17 12.44 57.18 0.30 9.52 0.02 KP Shangla 0.17 8.23 32.86 0.45 0.00 –0.23 KP Chitral 0.16 14.85 63.27 0.24 10.57 –0.04 KP Hangu 0.16 13.02 44.17 0.26 21.72 0.01 KP Bonair 0.14 5.91 62.76 0.53 0.00 –0.32 Sindh Badin 0.13 2.75 46.34 1.05 15.69 –0.74 Sindh Jamshoro 0.13 5.94 48.44 0.47 24.74 –0.21 Sindh Tando Allah Yar 0.12 6.53 44.55 0.39 30.83 –0.25 KP Lower Dir 0.12 3.85 64.17 0.65 6.05 0.00 KP Tank 0.09 8.85 40.73 0.23 10.66 –0.26 Sindh Thatta 0.09 2.17 37.72 0.92 14.37 –0.74 Sindh Umerkot 0.09 2.79 63.51 0.68 18.15 –0.53 KP Upper Dir 0.08 2.57 71.87 0.66 3.70 –0.28 Sindh Matiari 0.08 4.59 53.83 0.36 22.39 –0.15 Balochistan Jafarabad 0.07 2.92 39.20 0.54 21.13 –0.65 Sindh Tharparkar 0.07 1.24 59.32 1.17 3.77 –1.01 Balochistan Sibbi 0.06 14.68 63.14 0.09 49.12 0.70 Balochistan Lasbilla 0.04 2.78 30.94 0.32 30.67 –0.71 KP Lakki Marwat 0.04 2.11 54.95 0.42 10.90 –0.25 Balochistan Ketch/Turb 0.04 1.45 54.97 0.58 17.06 –0.82 Sindh T. M. Khan 0.03 1.96 38.92 0.34 17.29 –0.62 Balochistan Qillah Abdullah 0.03 3.02 59.16 0.19 15.61 0.10 Balochistan Qillah Saifullah 0.02 3.02 36.19 0.11 10.58 –0.75 Balochistan Gwadar 0.01 1.19 71.19 0.23 46.60 –0.39 Balochistan Nushki 0.01 2.14 42.21 0.12 20.62 –0.62 KP Kohistan 0.01 0.54 27.49 0.44 0.00 –0.95 Balochistan Chagi 0.01 1.57 45.32 0.14 9.50 –0.95 Balochistan Washuk 0.01 2.05 69.88 0.09 0.00 –1.10 Balochistan Panjgur 0.01 0.50 65.40 0.37 6.82 –0.79 Balochistan Zhob 0.01 1.05 37.84 0.17 17.26 –0.58 Balochistan Khuzdar 0.01 0.31 74.28 0.52 27.13 –0.50 Balochistan Lorali 0.01 0.55 23.76 0.27 15.01 –0.73 Balochistan Nasirabad 0.01 0.42 41.15 0.34 12.24 –0.76 Balochistan Kalat 0.01 0.55 74.30 0.24 13.57 –0.56 55 Table A5. The distribution of private school students across districts, 6–10 age group, Pakistan, 2010/11 Percent of Private Govt. Urban rate Percent of private participation participation in district District child Province District school rate in rate in x's average hh. population students in district x district x population asset index in district x district x (in %) (in %) (in %) Balochistan Mastung 0.00 0.75 81.17 0.11 21.96 –0.24 Balochistan Bolan/Kacc 0.00 0.26 67.82 0.23 15.69 –0.62 Balochistan Musakhel 0.00 0.07 19.51 0.14 0.00 –0.80 Balochistan Awaran 0.00 0.00 77.69 0.10 0.00 –1.05 Balochistan Barkhan 0.00 0.00 24.47 0.09 11.24 –0.68 Balochistan Dera Bugti 0.00 0.00 14.32 0.23 3.57 –1.13 Balochistan Harnai 0.00 0.00 61.65 0.09 0.00 0.06 Balochistan Jhal Magsi 0.00 0.00 78.28 0.12 4.21 –0.63 Balochistan Kharan 0.00 0.00 67.25 0.10 14.68 –0.80 Balochistan Kohlu 0.00 0.00 34.97 0.10 6.71 –0.82 Balochistan Pashin 0.00 0.00 68.73 0.16 8.42 0.21 Balochistan Sherani 0.00 0.00 63.36 0.08 0.00 –0.72 Balochistan Ziarat 0.00 0.00 59.74 0.03 11.11 –0.04 Notes: The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 56 Table A6. The distribution of private school students across districts, 11–15 age group, Pakistan, 2010/11 Percent of Private Govt. Urban rate Percent of private participation participation in district District child Province District school rate in rate in x's average hh. population students in district x district x population asset index in district x district x (in %) (in %) (in %) Sindh Karachi 18.62 52.33 26.56 6.46 96.26 1.22 Punjab Lahore 8.91 36.29 41.83 4.46 83.37 1.04 Punjab Gujranwala 5.59 40.43 40.43 2.51 52.08 0.50 Punjab Sialkot 4.75 37.05 45.73 2.33 24.54 0.44 Punjab Faisalabad 4.73 22.75 50.88 3.79 45.35 0.38 Punjab Rawalpindi 4.47 36.35 49.56 2.23 49.18 0.76 KP Peshawar 2.53 26.12 40.45 1.76 52.75 0.70 Punjab Gujrat 2.52 28.53 50.63 1.61 26.98 0.42 Punjab Sheikhupura 2.34 26.63 39.39 1.60 36.70 0.25 Punjab Multan 2.30 19.55 39.82 2.14 38.21 0.09 Punjab Kasur 2.20 21.02 48.39 1.90 25.80 –0.05 Punjab RahimYar Khan 1.95 14.52 37.29 2.45 21.84 –0.24 Punjab Narowal 1.91 28.75 52.65 1.21 14.17 –0.07 Punjab Vehari 1.63 16.07 42.81 1.84 18.19 –0.24 Punjab Sargodha 1.58 15.78 53.00 1.82 25.82 0.05 Sindh Hyderabad 1.42 24.20 46.16 1.07 80.54 0.83 Punjab Jhang 1.39 18.05 41.39 1.40 23.27 –0.40 Punjab Toba Tek Singh 1.32 19.33 54.91 1.24 20.86 0.28 KP Swat 1.27 19.89 48.05 1.16 13.64 0.00 KP Abbottabad 1.26 30.29 54.34 0.76 14.58 0.25 Punjab Muzaffargarh 1.20 10.02 38.26 2.18 14.21 –0.59 Punjab Okara 1.18 14.17 51.19 1.51 14.89 –0.17 Punjab Nankana Sahib 1.11 25.28 48.08 0.80 20.50 –0.01 Punjab Bahawalpur 1.06 9.92 44.81 1.95 30.64 –0.21 Punjab Khanewal 0.97 10.50 50.57 1.68 18.78 –0.21 Punjab Attock 0.94 18.33 57.83 0.94 21.93 0.16 KP Manshera 0.94 15.35 51.59 1.11 6.61 –0.06 Punjab Bahawalnagar 0.91 9.41 51.07 1.76 20.52 –0.17 Punjab Mandi Bahauddin 0.89 21.59 51.95 0.75 15.30 0.12 Punjab Sahiwal 0.84 11.36 55.33 1.35 15.98 0.00 Punjab Chakwal 0.78 19.38 69.15 0.73 14.65 0.39 Punjab Jehlum 0.74 19.75 64.29 0.68 25.85 0.45 KP Mardan 0.71 11.00 55.71 1.18 21.01 0.00 Punjab D. G. Khan 0.70 8.33 38.63 1.52 12.31 –0.60 Punjab Layyah 0.69 13.38 50.99 0.94 15.84 –0.56 KP Nowshera 0.68 15.79 50.55 0.78 24.44 0.14 Punjab Khushab 0.67 18.26 58.40 0.67 27.38 0.16 KP Charsada 0.63 13.60 50.59 0.85 17.15 0.05 KP Swabi 0.58 13.85 65.42 0.76 18.34 –0.01 Punjab Lodhran 0.55 10.40 40.06 0.97 13.53 –0.29 KP Haripur 0.55 18.16 71.65 0.55 13.20 0.40 Punjab Hafizabad 0.52 16.33 53.38 0.58 30.78 0.12 Balochistan Quetta 0.50 18.23 55.88 0.50 78.74 1.00 Punjab Mianwali 0.47 9.51 61.06 0.90 21.03 –0.14 Sindh Khairpur 0.46 6.88 51.72 1.23 27.72 –0.37 Sindh Sukkur 0.44 10.37 45.50 0.78 46.13 0.06 Sindh Ghotki 0.43 7.77 39.61 1.01 15.03 –0.36 Punjab Pakpattan 0.42 7.88 46.79 0.96 15.33 –0.37 Punjab Bhakhar 0.41 7.62 50.24 0.99 15.67 –0.47 KP D. I. Khan 0.37 7.42 40.51 0.90 13.13 –0.25 57 Table A6. The distribution of private school students across districts, 11–15 age group, Pakistan, 2010/11 Percent of Private Govt. Urban rate Percent of private participation participation in district District child Province District school rate in rate in x's average hh. population students in district x district x population asset index in district x district x (in %) (in %) (in %) Sindh Dadu 0.35 6.71 53.25 0.94 21.37 –0.03 Sindh Naushahro Firoze 0.32 7.28 51.43 0.81 19.77 –0.22 KP Kohat 0.32 12.34 51.23 0.47 28.26 0.18 KP Batagram 0.29 16.38 49.47 0.32 0.00 0.00 Punjab Chiniot 0.28 8.86 43.37 0.57 28.79 –0.18 KP Karak 0.27 10.79 60.08 0.45 6.00 –0.36 Sindh Sanghar 0.27 4.38 49.78 1.10 26.37 –0.19 KP Lower Dir 0.27 7.74 71.93 0.62 6.05 0.00 KP Chitral 0.26 14.39 68.67 0.32 10.57 –0.04 KP Bannu 0.23 6.48 53.90 0.65 4.16 0.05 Sindh Mirpurkhas 0.22 5.19 54.17 0.76 32.74 –0.07 KP Hangu 0.21 14.54 45.59 0.26 21.72 0.01 KP Shangla 0.18 9.12 37.23 0.37 0.00 –0.23 KP Malakand 0.17 8.62 71.51 0.36 9.52 0.02 Sindh Larkana 0.16 3.46 56.84 0.85 39.57 0.03 Sindh Kashmore 0.16 3.77 40.94 0.76 19.06 –0.45 KP Bonair 0.14 6.24 51.74 0.42 0.00 –0.32 Sindh Jamshoro 0.13 6.70 41.95 0.36 24.74 –0.21 Sindh Jaccobabad 0.13 3.61 39.55 0.67 21.32 –0.54 Sindh S. Benazirabad 0.13 3.46 43.52 0.68 31.42 –0.15 KP Lakki Marwat 0.13 5.25 48.67 0.45 10.90 –0.25 Sindh Kambar Shahdadkot 0.12 2.96 48.28 0.71 19.25 –0.22 Sindh Badin 0.12 2.47 38.92 0.85 15.69 –0.74 Sindh Thatta 0.11 2.17 33.67 0.92 14.37 –0.74 Sindh Shikarpur 0.10 2.26 46.56 0.81 21.86 –0.35 KP Tank 0.10 8.60 43.43 0.20 10.66 –0.26 Sindh Tando Allahyar 0.09 4.85 45.79 0.35 30.83 –0.25 Punjab Rajanpur 0.09 1.79 39.45 0.93 12.34 –0.73 KP Upper Dir 0.07 2.19 74.63 0.54 3.70 –0.28 Balochistan Jafarabad 0.06 2.75 23.99 0.41 21.13 –0.65 Sindh Maitari 0.06 3.07 40.85 0.33 22.39 –0.15 Sindh Umerkot 0.06 1.49 54.22 0.68 18.15 –0.53 Sindh Tharparkar 0.05 0.91 52.36 0.92 3.77 –1.01 Balochistan Sibbi 0.04 7.37 63.48 0.11 49.12 0.70 Sindh T. M. Khan 0.04 2.61 32.51 0.26 17.29 –0.62 Balochistan Lasbilla 0.04 2.21 40.32 0.29 30.67 –0.71 KP Kohistan 0.03 1.54 31.31 0.35 0.00 –0.95 Balochistan Qillah Abdullah 0.03 2.18 56.55 0.23 15.61 0.10 Balochistan Lorali 0.02 1.98 29.54 0.18 15.01 –0.73 Balochistan Ketch/Turbat 0.02 0.66 56.53 0.44 17.06 –0.82 Balochistan Chagi 0.01 1.87 37.35 0.11 9.50 –0.95 Balochistan Zhob 0.01 1.28 36.83 0.15 17.26 –0.58 Balochistan Khuzdar 0.01 0.51 59.03 0.36 27.13 –0.50 Balochistan Nasirabad 0.01 0.60 32.49 0.29 12.24 –0.76 Balochistan Pashin 0.01 1.18 72.00 0.14 8.42 0.21 Balochistan Qillah Saifullah 0.01 1.23 41.08 0.09 10.58 –0.75 Balochistan Mastung 0.01 0.96 70.97 0.10 21.96 –0.24 Balochistan Nushki 0.00 0.85 44.73 0.11 20.62 –0.62 Balochistan Kalat 0.00 0.38 57.38 0.21 13.57 –0.56 Balochistan Washuk 0.00 1.11 55.83 0.07 0.00 –1.10 58 Table A6. The distribution of private school students across districts, 11–15 age group, Pakistan, 2010/11 Percent of Private Govt. Urban rate Percent of private participation participation in district District child Province District school rate in rate in x's average hh. population students in district x district x population asset index in district x district x (in %) (in %) (in %) Balochistan Ziarat 0.00 2.49 59.23 0.03 11.11 –0.04 Balochistan Gwadar 0.00 0.39 69.05 0.16 46.60 –0.39 Balochistan Panjgur 0.00 0.19 50.64 0.20 6.82 –0.79 Balochistan Kharan 0.00 0.34 46.78 0.08 14.68 –0.80 Balochistan Kohlu 0.00 0.35 38.60 0.06 6.71 –0.82 Balochistan Awaran 0.00 0.00 69.45 0.08 0.00 –1.05 Balochistan Barkhan 0.00 0.00 14.91 0.12 11.24 –0.68 Balochistan Bolan/Kacc 0.00 0.00 50.32 0.20 15.69 –0.62 Balochistan Dera Bugti 0.00 0.00 6.64 0.22 3.57 –1.13 Balochistan Harnai 0.00 0.00 54.48 0.10 0.00 0.06 Balochistan Jhal Magsi 0.00 0.00 57.93 0.07 4.21 –0.63 Balochistan Musakhel 0.00 0.00 16.03 0.10 0.00 –0.80 Balochistan Sherani 0.00 0.00 50.82 0.07 0.00 –0.72 Notes: The statistics are estimated using the 2010/11 Pakistan Social and Living Standards Measurement (PSLM) survey. All statistics are estimated accounting for survey sampling weights. 59 Panel A. Pakistan, overall vs. private Panel B. Punjab, overall vs. private 80 80 70 70 60 60 Participation rate (in %) Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: School PR Observations: Private school PR Observations: School PR Observations: Private school PR Lowess: School PR Lowess: Private school PR Lowess: School PR Lowess: Private school PR Panel C. Sindh, overall vs. private Panel D. KP, overall vs. private 80 80 70 70 60 60 Participation rate (in %) Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: School PR Observations: Private school PR Observations: School PR Observations: Private school PR Lowess: School PR Lowess: Private school PR Lowess: School PR Lowess: Private school PR Panel E: Balochistan, overall vs. private Panel F. All provinces, private only 80 80 70 70 Private school participation rate (in %) 60 60 Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: School PR Observations: Private school PR Lowess: Punjab Lowess: Sindh Lowess: School PR Lowess: Private school PR Lowess: KP Lowess: Balochistan Figure A1. Evolution of (private) school participation rates, by province, 6–10 age group, 1998/99–2010/11 60 Panel A. Pakistan, overall vs. private Panel B. Punjab, overall vs. private 80 80 70 70 60 60 Participation rate (in %) Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: School PR Observations: Private school PR Observations: School PR Observations: Private school PR Lowess: School PR Lowess: Private school PR Lowess: School PR Lowess: Private school PR Panel C. Sindh, overall vs. private Panel D. KP, overall vs. private 80 80 70 70 60 60 Participation rate (in %) Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: School PR Observations: Private school PR Observations: School PR Observations: Private school PR Lowess: School PR Lowess: Private school PR Lowess: School PR Lowess: Private school PR Panel E: Balochistan, overall vs. private Panel F. All provinces, private only 80 80 70 70 Private school participation rate (in %) 60 60 Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: School PR Observations: Private school PR Lowess: Punjab Lowess: Sindh Lowess: School PR Lowess: Private school PR Lowess: KP Lowess: Balochistan Figure A2. Evolution of (private) school participation rates, by province, 11–15 age group, 1998/99–2010/11 61 Panel A. Females, overall vs. private Panel B. Males, overall vs. private 80 80 70 70 60 60 Participation rate (in %) Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: School PR Observations: Private school PR Observations: School PR Observations: Private school PR Lowess: School PR Lowess: Private school PR Lowess: School PR Lowess: Private school PR Panel C. Rural, overall vs. private Panel D. Urban, overall vs. private 80 80 70 70 60 60 Participation rate (in %) Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: School PR Observations: Private school PR Observations: School PR Observations: Private school PR Lowess: School PR Lowess: Private school PR Lowess: School PR Lowess: Private school PR Panel E: Female vs. male, private only Panel F. Rural vs. urban, private only 80 80 70 70 Private school participation rate (in %) Private school participation rate (in %) 60 60 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: Male Observations: Female Observations: Urban Observations: Rural Lowess: Male Lowess: Female Lowess: Urban Lowess: Rural Figure A3. Evolution of (private) school participation rates, by socioeconomic subgroups, 6–10 age group, 1998/99–2010/11 62 Panel A. Females, overall vs. private Panel B. Males, overall vs. private 80 80 70 70 60 60 Participation rate (in %) Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: SPR Observations: PSPR Observations: SPR Observations: PSPR Lowess: SPR Lowess: PSPR Lowess: SPR Lowess: PSPR Panel C. Rural, overall vs. private Panel D. Urban, overall vs. private 80 80 70 70 60 60 Participation rate (in %) Participation rate (in %) 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: SPR Observations: PSPR Observations: SPR Observations: PSPR Lowess: SPR Lowess: PSPR Lowess: SPR Lowess: PSPR Panel E: Female vs. male, private only Panel F. Rural vs. urban, private only 80 80 70 70 Private school participation rate (in %) Private school participation rate (in %) 60 60 50 50 40 40 30 30 20 20 10 10 0 0 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 1998/99 2001/02 2004/05 2005/06 2006/07 2007/08 2008/09 2010/11 Survey year Survey year Observations: Male Observations: Female Observations: Urban Observations: Rural Lowess: Male Lowess: Female Lowess: Urban Lowess: Rural Figure A4. Evolution of (private) school participation rates, by socioeconomic subgroup, 11–15 age group, 1998/99–2010/11 63 Table A7. Mean characteristics of private school students, Punjab, 1998/99, 2004/05, and 2010/11 Characteristic 6–10 age group 11–15 age group 2010/11 Diff. from Diff. from 2010/11 Diff. from Diff. from 2004/05 1998/99 2004/05 1998/99 (1) (2) (3) (4) (5) (6) Age (in complete years) 7.904 0.026 –0.008 12.790 0.067** 0.235*** (1.405) (0.022) (0.043) (1.363) (0.029) (0.068) Female 0.452 –0.010 –0.019 0.464 –0.029** –0.011 (0.498) (0.009) (0.015) (0.499) (0.013) (0.027) Rural 0.524 0.015 0.101** 0.523 0.024 0.167*** (0.499) (0.037) (0.051) (0.500) (0.041) (0.056) Lowest (first) hh asset index quintile 0.095 0.037*** 0.024* 0.064 0.019** 0.007 (0.293) (0.009) (0.014) (0.244) (0.007) (0.016) Mid (third) hh asset index quintile 0.219 0.016 0.009 0.199 0.018 0.030 (0.413) (0.012) (0.021) (0.400) (0.013) (0.024) Highest (fifth) hh asset index quintile 0.276 –0.062*** –0.119*** 0.343 –0.035* –0.115*** (0.447) (0.016) (0.031) (0.475) (0.021) (0.038) Hh head: highest ed.: no schooling 0.283 –0.052*** –0.026 0.281 –0.028* –0.006 (0.451) (0.014) (0.025) (0.450) (0.015) (0.028) Hh head: highest ed.: grades 1–5 0.165 –0.001 –0.035* 0.161 0.013 –0.035 (0.372) (0.010) (0.018) (0.368) (0.011) (0.023) Hh head: highest ed.: grades 6–8 0.231 0.006 0.033* 0.239 –0.004 –0.000 (0.421) (0.011) (0.019) (0.426) (0.013) (0.027) Hh head: highest ed.: grades 9–10 0.157 0.002 0.016 0.141 –0.014 0.029 (0.363) (0.009) (0.016) (0.348) (0.010) (0.019) Hh head: highest ed.: grades 11+ 0.164 0.046*** 0.012 0.178 0.034*** 0.012 (0.370) (0.010) (0.028) (0.383) (0.013) (0.026) Hh size 7.677 –1.666*** –0.665*** 7.475 –1.373*** –0.762** (3.250) (0.138) (0.196) (2.955) (0.130) (0.305) # of children 6–10 age group in hh 1.977 –0.262*** –0.102* 1.098 –0.263*** –0.254*** (0.918) (0.036) (0.056) (1.046) (0.038) (0.063) # of children 11–15 age group in hh 0.894 –0.212*** –0.169*** 1.885 –0.129*** –0.090 (0.993) (0.029) (0.048) (0.826) (0.029) (0.071) Notes: hh denotes household. Standard deviations are reported in parentheses in Columns (1) and (4). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6); they are estimated accounting for clustering at the PSU level. *** denotes p<0.01; ** p<0.05; and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 and 2004/05 Pakistan Social and Living Standards Measurement (PSLM) surveys and the 1998/99 Pakistan Integrated Household Survey. All statistics are estimated accounting for survey sampling weights. 64 Table A8. Mean characteristics of private school students, Sindh, 1998/99, 2004/05, and 2010/11 6–10 age group 11–15 age group 2010/11 Diff. from Diff. from 2010/11 Diff. from Diff. from Characteristic 2004/05 1998/99 2004/05 1998/99 (1) (2) (3) (4) (5) (6) Age (in complete years) 7.945 0.026 0.160*** 12.922 0.075 0.203** (1.448) (0.041) (0.055) (1.376) (0.049) (0.089) Female 0.460 0.005 0.012 0.467 0.032* 0.088*** (0.498) (0.016) (0.025) (0.499) (0.019) (0.034) Rural 0.091 0.017 0.050* 0.055 0.018 0.001 (0.288) (0.026) (0.025) (0.228) (0.016) (0.022) Lowest (first) hh asset index quintile 0.014 0.000 0.010** 0.007 0.003 0.003 (0.115) (0.008) (0.005) (0.082) (0.003) (0.004) Mid (third) hh asset index quintile 0.157 0.022 0.004 0.128 0.047** 0.011 (0.364) (0.021) (0.032) (0.334) (0.023) (0.034) Highest (fifth) hh asset index quintile 0.449 –0.054 –0.062 0.531 –0.096** 0.007 (0.497) (0.038) (0.052) (0.499) (0.043) (0.061) Hh head: highest ed.: no schooling 0.157 –0.035* –0.026 0.118 –0.051** –0.103*** (0.363) (0.021) (0.036) (0.323) (0.025) (0.039) Hh head: highest ed.: grades 1–5 0.120 –0.034 –0.011 0.094 –0.023 –0.013 (0.325) (0.021) (0.026) (0.291) (0.017) (0.025) Hh head: highest ed.: grades 6–8 0.202 0.035** –0.047 0.209 0.025 –0.053 (0.402) (0.016) (0.033) (0.406) (0.020) (0.034) Hh head: highest ed.: grades 9–10 0.108 –0.005 –0.020 0.133 0.028 0.053** (0.311) (0.014) (0.024) (0.340) (0.021) (0.027) Hh head: highest ed.: grades 11+ 0.414 0.039 0.103** 0.446 0.022 0.116** (0.493) (0.034) (0.049) (0.497) (0.044) (0.055) Hh size 7.273 –1.951*** –0.608** 6.980 –1.171*** –1.035*** (3.193) (0.392) (0.305) (2.854) (0.202) (0.379) # of children 6–10 age group in hh 1.918 –0.319*** –0.153* 0.954 –0.151** –0.255** (0.915) (0.116) (0.082) (0.999) (0.069) (0.121) # of children 11–15 age group in hh 0.878 –0.169*** –0.098 1.796 –0.250*** –0.197** (0.976) (0.054) (0.086) (0.775) (0.057) (0.080) Notes: hh denotes household. Standard deviations are reported in parentheses in Columns (1) and (4). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6); they are estimated accounting for clustering at the PSU level. *** denotes p<0.01; ** p<0.05; and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 and 2004/05 Pakistan Social and Living Standards Measurement (PSLM) surveys and the 1998/99 Pakistan Integrated Household Survey. All statistics are estimated accounting for survey sampling weights. 65 Table A9. Mean characteristics of private school students, KP, 1998/99, 2004/05, and 2010/11 6–10 age group 11–15 age group 2010/11 Diff. from Diff. from 2010/11 Diff. from Diff. from Characteristic 2004/05 1998/99 2004/05 1998/99 (1) (2) (3) (4) (5) (6) Age (in complete years) 7.922 0.003 –0.058 12.867 0.017 0.287*** (1.379) (0.043) (0.088) (1.387) (0.065) (0.102) Female 0.371 0.023 0.001 0.306 0.014 0.016 (0.483) (0.019) (0.036) (0.461) (0.025) (0.054) Rural 0.660 –0.017 0.030 0.643 –0.038 0.006 (0.474) (0.086) (0.098) (0.479) (0.090) (0.101) Lowest (first) hh asset index quintile 0.028 –0.027** 0.000 0.028 –0.012 0.000 (0.166) (0.011) (0.015) (0.164) (0.009) (0.014) Mid (third) hh asset index quintile 0.164 0.028 0.030 0.127 0.021 0.000 (0.371) (0.018) (0.034) (0.333) (0.018) (0.035) Highest (fifth) hh asset index quintile 0.410 –0.032 –0.164*** 0.481 –0.030 –0.197*** (0.492) (0.033) (0.055) (0.500) (0.041) (0.061) Hh head: highest ed.: no schooling 0.340 –0.034 –0.039 0.307 –0.031 0.005 (0.474) (0.027) (0.049) (0.461) (0.029) (0.053) Hh head: highest ed.: grades 1–5 0.093 –0.020 0.017 0.087 –0.022 –0.003 (0.291) (0.015) (0.023) (0.282) (0.018) (0.026) Hh head: highest ed.: grades 6–8 0.213 0.029 –0.022 0.216 0.002 0.005 (0.409) (0.018) (0.039) (0.412) (0.023) (0.046) Hh head: highest ed.: grades 9–10 0.116 –0.018 –0.002 0.105 –0.032 –0.016 (0.321) (0.019) (0.027) (0.307) (0.020) (0.037) Hh head: highest ed.: grades 11+ 0.238 0.043* 0.045 0.285 0.083*** 0.009 (0.426) (0.022) (0.035) (0.451) (0.028) (0.044) Hh size 9.210 –3.192*** –0.008 8.626 –3.412*** –0.606 (4.700) (0.417) (0.437) (4.100) (0.513) (0.532) # of children 6–10 age group in hh 2.250 –0.435*** 0.109 1.338 –0.627*** –0.250* (1.153) (0.109) (0.102) (1.281) (0.116) (0.130) # of children 11–15 age group in hh 1.130 –0.426*** –0.258** 1.968 –0.468*** –0.127 (1.094) (0.093) (0.108) (0.887) (0.101) (0.094) Notes: hh denotes household. Standard deviations are reported in parentheses in Columns (1) and (4). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6); they are estimated accounting for clustering at the PSU level. *** denotes p<0.01; ** p<0.05; and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 and 2004/05 Pakistan Social and Living Standards Measurement (PSLM) surveys and the 1998/99 Pakistan Integrated Household Survey (PIHS). All statistics are estimated accounting for survey sampling weights. 66 Table A10. Mean characteristics of private school students, Balochistan, 1998/99, 2004/05, and 2010/11 6–10 age group 11–15 age group 2010/11 Diff. from Diff. from 2010/11 Diff. from Diff. from Characteristic 2004/05 1998/99 2004/05 1998/99 (1) (2) (3) (4) (5) (6) Age (in complete years) 8.077 0.040 -- 12.995 0.202 -- (1.425) (0.165) -- (1.314) (0.138) -- Female 0.387 –0.061 -- 0.253 –0.106* -- (0.488) (0.063) -- (0.436) (0.060) -- Rural 0.151 –0.051 -- 0.214 –0.008 -- (0.358) (0.101) -- (0.411) (0.131) -- Lowest (first) hh asset index quintile 0.013 –0.004 -- 0.010 0.010 -- (0.113) (0.017) -- (0.098) (0.007) -- Mid (third) hh asset index quintile 0.015 0.002 -- 0.026 –0.010 -- (0.120) (0.010) -- (0.160) (0.024) -- Highest (fifth) hh asset index quintile 0.769 –0.056 -- 0.803 –0.063 -- (0.422) (0.058) -- (0.398) (0.058) -- Hh head: highest ed.: no schooling 0.189 0.036 -- 0.188 0.052 -- (0.392) (0.057) -- (0.391) (0.047) -- Hh head: highest ed.: grades 1–5 0.133 0.008 -- 0.062 –0.129** -- (0.340) (0.039) -- (0.241) (0.058) -- Hh head: highest ed.: grades 6–8 0.153 –0.151** -- 0.157 –0.112 -- (0.361) (0.067) -- (0.365) (0.082) -- Hh head: highest ed.: grades 9–10 0.093 0.015 -- 0.101 0.050 -- (0.291) (0.037) -- (0.302) (0.033) -- Hh head: highest ed.: grades 11+ 0.432 0.093 -- 0.492 0.139 -- (0.496) (0.089) -- (0.501) (0.094) -- Hh size 8.137 –3.643** -- 7.848 –4.371** -- (3.717) (1.555) -- (2.931) (1.842) -- # of children 6–10 age group in hh 2.160 –0.809 -- 1.296 –1.098 -- (0.953) (0.497) -- (1.061) (0.718) -- # of children 11–15 age group in hh 0.978 –0.825 -- 1.937 –0.900 -- (0.994) (0.518) -- (0.749) (0.587) -- Notes: hh denotes household. Standard deviations are reported in parentheses in Columns (1) and (4). Standard errors are reported in parentheses in Columns (2), (3), (5), and (6); they are estimated accounting for clustering at the PSU level. *** denotes p<0.01; ** p<0.05; and * p<0.10 (two-tailed significance tests). The statistics are estimated using the 2010/11 and 2004/05 Pakistan Social and Living Standards Measurement surveys. Statistics using the 1998/99 Pakistan Integrated Household Survey are not provided as the sample sizes in the socioeconomic subgroups in Balochistan are too small to obtain reliable estimates. All statistics are estimated accounting for survey sampling weights. 67