Production of cognitive and life skills in public, private, and NGO schools in Pakistan.

Arif, G.M. ; Saqib, Najam us

The share of private and NGO schools in primary education has substantially increased over time, though the public sector is still a major player in this area. The present study analyses the factors determining the quality of education offered by the three types of schools and draws policy recommendations for improving primary education in Pakistan. The study compares learning achievement of Class 4 students enrolled in 50 public, private, and NGO schools located across six districts of Pakistan and in Azad Kashmir in terms of their scores in Mathematics, Urdu, and General Knowledge tests. The analysis shows that, practically, there is no gap between public and NGO schools in terms of the test scores of their students. However, a significant test score gap was found between the students enrolled in public and private schools. This gap was largely explained by family background and school-related variables, including teachers qualification and student/teacher ratio. However, the performance of private schools was not uniform across districts. In some districts public schools performed even better than private and NGO schools. The findings of this study highlight the need for improving the quality of education in public schools by recruiting more qualified teachers and improving overall supervision. Teacher training is the area where the public and private sectors can benefit by pooling their resources and expertise.

1. INTRODUCTION

Although primary education in Pakistan is dominated by the public sector, the role of private sector has gradually increased. The share of private schools in the total primary enrolment has jumped from only 14 percent in 1990-91 to 25 percent in 1998-99. (1) This increase has been observed in urban as well as in rural areas of the country, though private schools are more numerous in urban areas, where about half of the total primary-level students are enrolled in these schools. In urban Punjab, privately-managed schools account for more than half of the primary school enrolment. (2) The Government of Pakistan (GoP) aims at establishing schools in rural areas through private-public partnership [Pakistan (2001)]. However, the persistence of high levels of poverty, particularly in rural areas, may inhibit the expansion of private sector among low-income groups.

Besides coverage, the quality of education provided in public schools is a major concern in Pakistan. One of the arguments in favour of more private schools is that competition with private schools would improve the quality of public education. However, this may not be necessarily true. If the private schools attract more able students from public schools, the expected positive impact will be negated [Sander (1999)].

Quality of education, as measured by students' cognitive achievement, depends on personal attributes of the child, family background variables, school- and teacher-related inputs into the education process and the structure of the system itself. The most important family background variables are considered to be household income and parental education [Behrman, et al. (1999)]. These factors play an important role in improving the home learning environment and availability of educational materials ranging from a reading-lamp to dictionaries and encyclopedia. The quality of teachers is also very important in improving the quality of education. (3) School-related variables such as class size, school attendance, teacher-student ratio, availability of text/work books and assignment of homework also make a difference in the level of student achievement. Performance of students enrolled in small classes is likely to be better than that of the students enrolled in large-size classes, as former are more likely to receive personalised teacher attention [Bedi and Marshall (1999)]. Similarly, schools may also differ in terms of their management practices, which in turn may affect the educational outcomes.

Several studies have explored different dimensions of primary schooling in Pakistan. (4) However, only a few recent studies have been undertaken to investigate the role of the NGOs and private sector in this area. On the basis of a survey of private and NGO schools, Gazdar (1998) found that the latter were innovative and more efficient because of community participation, parents' trust in teachers, and proximity of the school to pupil's home. Zia (1999) assessed the performance of students from a Bunyad Centre, an NGO school, located in Sheikhupura district. According to his results, even the students who ranked low in the class, obtained more than 50 percent marks. Khan, et al. (1999) compared the quality of education provided by public, private, and NGO schools in all the provinces and concluded that NGO schools were performing better than public schools, whereas private schools ranked between the NGO and public schools.

As is evident from the brief review presented above, the studies of the comparative performance of the public, private, and NGO schools have serious limitations in terms of scope, coverage, and method of analysis. In particular, none of these studies has tried, systematically, to explore the difference in the educational outcomes produced by the three types of schools, while controlling for the four sets of factors (i.e., personal attributes of child, family background variables, school- and teacher-related inputs and school structure) that are known to influence these outcomes.

One of the major contributions of the present study is that it links students' cognitive achievement to their personal characteristics, family background, school-and teacher-related inputs into the educational process, as well as to the type of school in which they are enrolled. (5) This methodology allows us to extensively examine the key question of what determines students' success in school, while avoiding many of the shortcomings of the previous studies.

This paper is organised as follows. Section 2 gives a brief overview of the educational reforms in Pakistan, primary school enrolment, and the extent of involvement of the private sector in the field of education. Data sources and method of assessment of educational outcomes are discussed in Section 3, while the sample characteristics and results of the cognitive skills tests are reported in Section 4. Relative performance of students in different types of schools is analysed in Section 5. The final section concludes the paper with some policy recommendations.

2. EDUCATIONAL REFORMS, PRIMARY SCHOOLING, AND GROWTH OF THE PRIVATE SECTOR IN PAKISTAN

Educational policies introduced since Independence have assigned high priority to the universalisation of primary education (UPE). The first education conference held in 1947 laid down the target of free and compulsory primary education within a decade. Unfortunately, this target could not be achieved even half a century later. All the educational policies that followed the 1947 education conference simply postponed the achievement of UPE to a later date. For example, the policy introduced in 1959 made very comprehensive and ambitious recommendations, including character building through religious education, compulsory primary education for all within 15 years, and revision of curricula to impart technical and vocational education at schools. These recommendations were incorporated in the Second Five-year Plan (1960-65). However, this plan failed to achieve its targets related to primary and vocational education [Pakistan (1960, 1965)]. The 1969 Education Policy, which envisaged UPE by 1980, met a similar fate. The 1972 policy shifted the goal of universal primary education to 1986 for boys and 1992 for girls [Pakistan (1972)]. Keeping in line with the tradition so firmly established by the past educational policies, the 1979 and the 1992 education policies also shifted the goal of UPE to a still later date. Education Sector Reforms (ESR), an action plan to implement the 1998 education policy, has now set the target of 100 percent gross primary enrolment and 60 percent literacy rate by 2004 [Pakistan (2001a)].

Data show that the overall gross enrolment rate in the late 1990s was only 69 percent, while in the early 1980s it was only 44 percent (Table 1). The gap between the enrolment rates of males and females, though bridged a little during the last two decades, still leaves much to be desired. Even if we achieve the target of UPE by 2004, it will take some time before every person in the country will have primary-level education. (6) Before examining the reasons for not achieving the target of UPE in the past, a brief comment on the recent drop in the gross enrolment ratio from 75 percent in 1995-96 to 69 percent in 1998-99, as shown in Table I, seems to be appropriate here. This decline took place largely in rural areas; the gross enrolment in urban areas was quite impressive and there was no change in it during the 1990s, whereas rural areas witnessed a decline of about 7 percent between 1995-96 and 1998-99 (Appendix Table 1). The decline in enrolment reflects a serious regression. Although one can think of several reasons for this decline, the existing evidence points towards the rise in poverty in the 1990s as the main reason [Hussain (2003)]. The negative effect of poverty on primary school enrolment is found to be more pronounced in rural areas than in urban areas [Arif, et al. (1999)].

Now the question is why the target of UPE could not be achieved. Some valuable insight can be gained into the causes of this persistent failure by examining various Five-year Plans published by the Government of Pakistan, as each of them provides a critical review of the previous Plan. These Plans show that the physical targets necessary for providing UPE, particularly opening of new schools, could not be achieved mainly because of inadequate allocation of government resources. (7) The high birth rate, leading to constantly expanding numbers of potential pupils and the absence of appropriate managerial capacity needed for planning, implementation, and supervision of educational projects are some of the other major obstacles in achieving UPE. Among the demand-side factors that are responsible for the low enrolment rates in primary schools, poverty, parents' indifference, particularly towards girls' education, and irrelevance of school curricula to the skills demanded in the labour market appear to be the most important. Legislation for compulsory primary education proposed and mentioned in various official documents about 20 years ago to make up for the slack demand for primary schooling has not yet been enacted [Pakistan (1960, 1965, 1978, 1983, 1994, 2001)].

To overcome the above-mentioned obstacles, ESR, the most recent action plan for achieving UPE, has expressed the government's willingness to develop a partnership with the private sector and NGOs at the primary as well as other levels of education. For this purpose, a package of incentives has been proposed for the private sector. Among other things, it includes a rebate in income tax and provision of plots of land in residential schemes at subsidised prices [Pakistan (2001)].

To understand the importance of the private sector in education and its relationship with the public sector, it would be instructive to first look at its historical evolution in Pakistan. Private educational institutions existed in Pakistan even at the time of Independence in 1947, though they expanded rapidly in the 1960s, particularly in the urban areas. A major shift in the education policy occurred in 1972 when more than 3000 private schools and colleges were nationalised. As a result, the number of schools under government administration increased sharply, resulting in serious deterioration in the management, and resource availability. The standard of education in the government-run schools also declined accordingly, and the infrastructure wore out with the passage of time [Saqib (1998); Mahmood, et al. (1999)].

The 1979 educational reforms reversed the earlier policy and allowed private individuals and firms to establish schools. Initially, the number of private schools increased very slowly, and the significance of this development was not fully appreciated for some time. Perhaps this is why no reliable countrywide data became available on private schools until the publication of the first Pakistan Integrated Household Survey (PIHS) in 1991 and The Census of Private Educational Institutions in Pakistan in 2001. Some figures, however, are available for the Punjab, the most populous province of the country, which is home to more than half of its population. These figures indicate that in 1982 there were only 2 registered private primary schools in the province. This number increased to 268 in 1984, and to 451 in 1987, accounting for respectively 1.8 percent and 3.1 percent of the total primary school enrolment in the province [Punjab (1982, 1984, 1987)].

The decade of 1990s witnessed a mushroom growth of private schools in the country. The number of primary schools in the private sector increased from about 11000 in the early 1990s to approximately 15000 in 1999-2000, leading to a rapid decline in enrolment in the government schools as a percentage of the total primary school enrolment (Figure 1). This decline was observed in both urban and rural areas, although it was more pronounced in the urban areas, where the share of government schools in total primary school enrolment fell from 70 percent in 1991 to 53 percent in 1998-99. In other words, approximately half of the total enrolment in urban areas was in private schools. (8) Statistics given in Table 2 do not indicate a substantial gender differential in the school choice. However, substantial inter-province differentials do exist in the distribution of primary school enrolment between private and public schools. Private schools are more popular in the urban areas of Punjab and Sindh provinces, where they account for 53 and 44 percent of total primary school enrolments respectively (Appendix Table 2). Public sector still remains the main provider of primary education in the rural areas of all the four provinces and in the urban areas of the NWFP and Balochistan. In absolute numbers, enrolment in government schools has increased over time in all provinces and regions; it increased from about 11 million in 1991 to 20 million in 2000 (Table 3). The increase in total enrolment was observed for males as well as females. The share of females in total enrolment increased from about one-third in early 1980s to more than two-fifths in 1999-2000.

[FIGURE 1 OMITTED]

Income adds an important dimension to the distribution of primary school enrolment between public and private schools. As can be seen from Table 4, children belonging to low-income families are heavily concentrated in government schools: 91 and 68 percent of children from the 1st (the lowest) income quintile in rural and urban areas respectively were in government schools in 1998, while the corresponding percentages for the 5th (the highest) quintile were 73 and 16 respectively. The private sector has a presence at all income levels, but it is more pervasive among the rich urban families: 81 percent of the children in the highest income quintile in urban areas were attending private schools, while the corresponding figure for the lowest income quintile was 30 percent. There was a difference of 25 percentage points between these figures for the 4th and 5th quintiles.

In the urban areas of the four provinces of Pakistan, a decline in the percentage of pupils enrolled in government primary schools was observed in the 1990s in all quintiles, though it was more noticeable in the 5th quintile. In Punjab and Sindh provinces, this decline was just marginal in the 1st and 2nd quintiles. The case of the NWFP and Balochistan was different in this regard, where the proportion of pupils enrolled in government schools in the lowest quintiles increased over time, indicating heavy and increasing dependence of poor families on the public sector (Appendix Tables 3 and 4).

Overall, it appears that the role of public sector in basic education has been dwindling over time. The urban middle-income and rich families have gradually shifted from the public to the private school system, probably due to relatively better quality of education offered by the latter. However, it needs to be emphasised that most of the rural population as well as the urban poor still depend heavily on the public sector for primary schooling of their children. In the wake of rising poverty during the 1990s, this is not a good omen for the ESR initiative of establishing private-public partnership schools in rural areas. Persistence of high levels of poverty in rural areas may severely restrain the momentum of this initiative. It is noteworthy that in 1998-99, about 40 percent of the rural population was living below the poverty line [Jafri (1999); Qureshi and Arif (2001)].

3. DATA SOURCES AND METHODOLOGY

3.1. Assessing Cognitive Skills

The main objective of this study, as outlined in Section I, is to explore the differences in the educational outcomes produced by the public, private, and NGO schools. To accomplish this objective, the study has used a survey in which Class 4 students enrolled in these three types of schools were assessed in Mathematics, Urdu, and General Knowledge. The survey was carried out in 1999 in six districts of the country--Khushab and Multan in the Punjab, Khairpur in Sindh, Malakand in the NWFP, Turbat in Balochsitan, and Muzaffarabad in AJK. In each district, 3 to 4 NGO schools, 2 public and 2 private schools with at least one section of Class 4 students were selected. The total sample comprised 50 schools. All private and most of the NGO schools included in the sample enrolled both boys and girls, while public schools were either only for boys or only for girls. (9) In each sampled school, four separate questionnaires were administered to school heads, Class 4 teachers, all Class 4 students, and their parents. In total, 50 school heads and Class 4 teachers, and 965 students and their parents, were interviewed.

A few words about the organisation and the curricula followed by the three types of school are in order here. The NGO schools in Pakistan show wide variation in terms of their size and organisation. They are non-profit organisations and generally follow the curriculum of the public schools. However, a few NGOs have established education centres where the 5-year primary schooling is completed in 3 years. These schools were not included in the sample because in public and private schools' primary education is completed in 5 years, and this study has focused on Class 4 students, who are in Year 4 of their primary schooling. Private schools work for profit and follow either the national curriculum or a curriculum approved by foreign educational institutions. Public schools obviously follow the national curriculum and charge very little or no tuition fee. Most of the cost of operating public school system is borne by the public exchequer. Madrassahs in Pakistan are distinct from the public, private or NGO schools. They do not follow the national curriculum; rather they follow their own curriculum, emphasising Islamic studies. Madrassahs are not included in the sample because their curriculum is not comparable to that of other types of schools. The school sample on which this study is based is fairly representative of the public, private and NGO schools in Pakistan. (10)

The sampled students of Class 4 were almost evenly distributed across three types of schools: 37 percent students were attending public schools, 33 percent were enrolled in NGO schools, while 30 percent went to private schools. About 47 percent of the Class 4 students were females. This percentage was much higher (62 percent) in the case of the NGO schools (Table 5).

Students' cognitive achievement is usually measured by their performance on standardised tests. The score on these tests is considered to be the best available measure of a student's cognitive skills that would permit objective and transparent comparisons. (11) Three tests of cognitive achievement were given to all the Class 4 students in the sample. All tests were administered in Urdu language except for Sindh, where these tests were given in Sindhi language because most of the schools in this province used Sindhi as the medium of instruction. In schools with only one section for Class 4, all students in the class took the test; in schools with more than one section of pupils in the same class, one section was selected randomly.

The first question of all the three tests was relatively easy and its primary objective was a warm-up. Other questions varied in difficulty and coverage of subjects. The mathematics test included a wide variety of topics ranging from basic operations of addition, subtraction, multiplication, and division to more complex concepts of algebraic equations, sequence of numbers, fractions, conversion of kilograms into grams, and reading a bar chart. There were 18 questions in this test. The test of literacy consisted of 17 questions aimed at testing pupils' reading/comprehension capabilities and their ability to write sentences and to distinguish between the correct and incorrect use of tense in a sentence. There were 12 questions in the life skills and general knowledge test. These questions were designed to assess the student's civic sense, basic awareness of contemporary issues related to health, gender, and environment, and knowledge of the country and the people around him. (12) In short, the data set on which this study is based provides the kind of extensive information which is necessary for a meaningful and comprehensive analysis of the educational outcomes produced by the public, private and NGO schools.

3.2. Educational Production Function

The studies of educational production function in developing countries have focused mainly on four types of variables: child characteristics, family characteristics, teacher characteristics and school characteristics [Glewwe and Jacobey (1993); Bedi and Marshel (1999); Mizala and Romaguera (2000)]. The performance of students may also vary across the geographical locations. To analyse the differences in the educational performance of the students enrolled in the three school types, namely public, private, and NGO, the following equation has been estimated:

L4 = f(C, F, T, S, G) (13)

Where,

LA = student achievement in school (total test score)

C = characteristics of the students

F= characteristics of the students' families

T= characteristics of the teachers in school

S = characteristics of the school

G = dummy variables for geographical location (districts).

The dependent variable is the total test scores in Mathematics, Urdu, and General Knowledge. The equation includes four child characteristics: age, sex, preschool attendance, and having homework regularly. Family characteristics included in the equation are: mother's education, household income, and number of school-going siblings. These variables can play an important role in shaping home learning environment. Variables used to capture the quality and type of the school attended by the child are student/teacher ratio and school type. Teachers' gender and education are the teacher characteristics used in this study. To determine the effect of geographical location on students' performance, district dummies have also been included in the equation. Operational definitions of these variables are as follows:

 School Type--Public school is the reference category
 and two dummies are used for the NGO and
 private schools.
 Child Characteristics--Age of the child in years.
 --Sex (male = 1, female=0).
 --Pre-school attendance = 1, 0 otherwise.
 --Child gets homework = 1, 0 otherwise.
Household Characteristics--Mother's primary and higher level of
 education=1, 0 otherwise.
 --Household income: 'low-income household'
 is the reference category. The remaining
 two categories for which separate dummy
 variables have been used are 'middle
 income households', referring to the
 middle 50 percent households, and 'high
 income households', referring to the top
 20 percent households.
 --Number of school-going children refers
 to the number of children in a household
 who were enrolled in any school at the
 time of school survey.
 Teachers/Class--Sex (male = 1, female=0).
 Characteristics--Education: (BA/MA = 1, 0 otherwise).
 --Student/teacher ratio refers to the total
 number of students in a school divided by
 the total number of teachers.

4. RESULTS

4.1. Sample Characteristics

Table 6 sets out data on some selected characteristics of children, their households, and teachers that can influence their educational outcomes, controlling for the type of school in which they were enrolled at the time of the survey. In public schools, usually, children enter at the age of 5 but private and NGO schools may have different criteria to admit children in different classes. If the sampled students were enrolled in school at the age of 5, they should be 8 to 9 years old in Class 4. However, the majority of these students was 9 to 13 years old at the time of the survey, and there was no major difference in mean age of children enrolled in the public, private, and NGO schools. Although the possibility of misreporting of age cannot he ruled out, class repeaters may have influenced the mean age of the students. About a quarter of the sampled children reported that they had been in primary school for four or more years. Private and NGO schools may also have been able to motivate older children to enrol in school.

Table 6 also presents data on the selected characteristics of teachers associated with different types of schools. The proportion of female teachers was substantially higher in NGO schools as compared to both public and private schools. In terms of qualification, teachers employed in private schools were more qualified than their counterparts in public and NGO schools. In NGO schools, about half of Class 4 teachers had a qualification of intermediate level or lower, while the corresponding figures for public and private schools were 32 percent and only 3 percent respectively. Compared to 30 percent in public and 33 percent in NGO schools, about 56 percent teachers in private schools had a bachelor's or higher degree. On the other hand, teachers in public schools were more experienced than the teachers in private and NGO schools (not shown in Table 6). With respect to school characteristics such as class size and student/teacher ratio, private schools appear to be better equipped than public schools. Mean values of these variables are lower for NGO schools as compared to the other two types of schools.

4.2. Standardised Performance Tests

Table 7 shows the test results of students, where the figures correspond to the percentage of correct answers given by the students. In general, the performance of students enrolled in private schools was better than that of their counterparts in the public and NGO schools. There was no substantial difference between the NGO and public schools in this respect. On average, private school students obtained 75 percent marks in the three cognitive skills tests compared to 66 and 64 percent marks respectively for the students enrolled in the NGO and public schools. The same pattern was observed in the separate scores of the three tests, though the gap between the private and public schools in terms of the marks obtained in Urdu and General Knowledge tests was much wider.

The range of marks obtained by the sampled students, as shown in Figure 2, highlights two more features of the assessment results. First, the number of students who obtained more than 80 percent marks was the highest in the private schools (46 percent), followed by the students of the NGO schools (31 percent), in the public schools, only 25 percent of the students obtained more than 80 percent marks. This pattern also persisted in all three subjects separately (Appendix Table 5). Secondly, in the case of students who obtained less than 40 percent marks, there was no major difference between the students of the public and NGO schools. However, only a tiny proportion of private school students obtained less than 40 percent marks. These results are consistent with two earlier studies, one by Gazdar (1998) and the other by Zia (1999). However, these results differ from the findings of Khan, et al. (1999). While the present study shows that the performance of the NGO schools was almost the same as that of the public schools and definitely lower than that of the private schools, Khan, et al. conclude that the performance of the NGO schools was better than that of both the public and private schools. The source of this divergence appears to be choice of the NGO school made by Khan, et al. (1999). Their primary focus was large NGOs that operated a multiple school system; only 7 of the 43 schools included in their sample were operating a single school. The NGOs that operate multiple schools are generally large, well-funded organisations offering high-quality education. However, very few such NGOs are active in the field of education in Pakistan. Therefore, Khan, et al. (1999) seem to have compared the best of the NGO schools with the other two types of schools.

[FIGURE 2 OMITTED]

Care must be taken in interpreting the test results, as comparison of scores across different school types may be affected by selection factors. Complementary studies and data are needed to analyse the issue of selection in Pakistan's educational system in greater depth. Mizala and Romaguera (2000) examined the heterogeneity of Chilean students in different types of schools by comparing dispersion of test scores across schools. Following their methodology, the average, maximum, and minimum values of the standard deviation of the test scores of students enrolled in the surveyed schools is reported in Table 8. There were four private and NGO schools that had a highly homogeneous student body as indicated by low values of the standard deviation of test scores. However, the public, private, and NGO schools with a highly heterogeneous student body were also found. In general, looking at the gap between the maximum and minimum values of standard deviation, it can be concluded that most homogeneous schools are private, followed by the public and then the NGO schools. When only the schools with relatively high test-scores (70 percent or above) are considered, the order between the three types of schools is reversed. This is evident from the last three rows of Table 8. It means that high-scoring private schools (>70 percent) are less homogeneous than other schools. There are some NGO and public schools with a highly homogeneous student body that obtained high scores, suggesting that educational performance of some public schools is so good that they attract good students. This has been the case at least in two sampled districts, Turbat and Muzaffarabad. (This will be further discussed in the next section).

5. RELATIVE PERFORMANCE OF STUDENTS: A MULTIVARIATE ANALYSIS

The relative influence of different variables on the performance of students is assessed in this section by the multivariate regression technique. The equation specified in Section 3.2 provides the basis for this assessment. Four models are estimated. Dependent variable in Model 1 is the total score obtained in the three subjects, namely, General Knowledge, Urdu, and Mathematics. Separate test scores on the three subjects are used as the dependent variables in the remaining three models. The explanatory variables are the same in all the models. Estimation results are presented in Table 9. The value of adjusted [R.sup.2] of Model 1 (total test score) shows that 55 percent of variation in the performance of students is caused by the independent variables, indicating a good fit to the cross-section data.

Results of all four models show that students enrolled in the private schools performed significantly better than their counterparts in the public schools. In these models, there was no significant difference between the performance of the public and NGO schools. To see the overall effect of school type (public, private, and NGO) on the performance of students, a specification test was also performed. All three categories of school type (public, private, and NGO) were dropped from model 1 (Table 10). The F-value was calculated to test the null hypothesis that there was no effect of school type on students' performance, which was rejected at 1 percent level of significance (Table 11). Thus the school system has an independent effect on the performance of children on cognitive and life-skills tests. The school-based management practised in private schools appears to be superior to the centralised management system of public schools, which is controlled by provincial or federal authorities. It is possible to achieve substantial gains in terms of improved efficiency by enabling public schools to adopt the management practices of private schools.

With respect to child characteristics, four variables, i.e., age, sex, pre-school attendance, and a dummy variable for having homework regularly, were included in the four models (Table 9). Age of the student had an independent and significant effect on students' performance in Model 1 (total test score), Model 2 (General Knowledge), and Model 4 (Mathematics). The higher the age of students, the greater the scores obtained in the tests. The effect of students' sex turned out to be statistically significant only on the test score of Mathematics, showing that boys were more likely to perform better in Mathematics than girls. Pre-school attendance did not have a significant impact on the test scores. However, the effect of homework was positive and statistically significant in all four models. Assigning homework to students (and probably its regular checking) helps in creating an environment conducive to home learning.

Table 9 also shows that household income was positively related to the test scores. Children belonging to middle-income households performed better than those hailing from low-income households. However, the performance of the children from high-income households was not significantly different from that of the base category in Models 1, 2, and 3. In Mathematics (Model 4), their performance was even worse than that of the students coming from low-income households.

Parents' education is expected to exert positive influence on the cognitive achievement of the children, as educated parents are likely to create an environment conducive to learning and are able to help children in their studies. In Models 1 and 3, mother's education (more than primary) was significant at 5 percent level of significance. (14) Model 1 also shows that the number of school-going children in the household had a positive and significant effect on the learning achievement. It is likely that more school-going children create a better learning environment in the household and help each other in studies. The specification test, based on the F-statistic, showed that the combined effect of family background variables (household income, mother's education, and number of school-going children) and child characteristics was statistically significant (Tables l0 and 11).

Teachers' sex had significant and negative effect on the learning achievements of students (Models 1-3), indicating that the test performance of students taught by female teachers was better than those who had male teachers. Hence, at least at the primary level, females appear to be more effective teachers. Results further show that teacher's bachelor-level or higher education had significant and positive effect on the learning achievements of students, implying that teachers with B.A.B.Sc. or higher degree were more productive than those with less education. It has been reported earlier that private schools in the sample had more teachers with a bachelor's or higher degree than the other two types of schools.

Table 9 shows that the pupil-teacher ratio turned out to be negative and significant in all models, indicating that students' performance was better at the schools that had more teachers for a given number of students. The F-values reported in Table 11 show that the overall effect of the class- and teacher-related variables on the performance of students was also statistically significant at I percent level.

To capture the effect of geographical variation in the production of cognitive and life skills, five district dummies were included in all four models, with Multan as the reference district. Results show that the performance of students in Muzaffarabad, Turbat, Malakand, and Khairpur districts was significantly better than that of the students living in the reference district. However, Urdu language turned out to be a relatively difficult subject for those living in Malakand, as indicated by a negative coefficient of the dummy for this district in Model 3. It may be noted that most people in Malakand area speak Pashto, a language markedly different from Urdu. The null hypothesis that geographical location (district) has no effect on the performance of students was also rejected at 1 percent level of significance (Table 11). It shows that the district-specific effects are strongly correlated with the performance of students. The reason for this phenomenon cannot be determined with certainty from the quantitative data available to us. However, the evidence gathered through focus group discussions may offer some clues.

Participants of the focus groups in three of the six districts, namely Turbat, Muzaffarabad, and Khairpur, opined that the competition created by some private schools had contributed to the improvement in the quality of education offered by some public schools. The distribution of average marks obtained by students in different subjects across the sampled districts also lends support to this view (Table 12). It shows that public school students in Turbat performed even better than private school students, while there was no substantial difference in the test performance between the public and NGO schools. Subject-wise data show that the performance of public school students was better in Mathematics and Urdu than that of the students enrolled in the NGO and private schools, though in the case of General Knowledge, the NGO school students performed better than the students enrolled in the public and private schools of Turbat. Moreover, in Turbat and Muzaffarabad districts, the student/teacher ratio was much lower in all types of schools as compared to this ratio in other sampled districts (not shown in Table 12). In these two districts, the role of the head and the teachers of some public schools in improving the quality of education was particularly appreciated during the focus group discussion. Finally, and more importantly, public, private, and NGO schools received considerable cooperation from the parents and the community. These elements were missing in the districts where test performance of students was relatively poor.

6. CONCLUSIONS AND POLICY IMPLICATIONS

A review of the educational reforms introduced since the independence of the country shows that a major shift in policy occurred in 1972 when private schools and colleges were nationalised on a large scale. The 1979 educational reforms reversed the earlier policy and allowed private individuals and firms to establish schools. As a result of this change, the following decades saw the private and NGO schools emerging as major players in the field of primary education. The rapid growth in the number of these schools was accompanied by a corresponding decline in the public school enrolment as a percentage of the total primary school enrolment.

This study compares the learning achievement of the students enrolled in the three types of school. Significant gap in test scores was observed between students enrolled in the public and private schools. However, the gap between the public and the NGO schools was small or non-existent, and substantially less than what was suggested by some previous studies. The NGO schools were most heterogeneous in terms of the performance of their students on the tests of cognitive achievement, but among the schools with the highest test scores, these schools turned out to be the most homogeneous. In some districts, the NGO schools did better than their public sector counterparts. In others districts, the public schools performed better than the private schools. The test performance of students was largely explained by the school type, individual and family background characteristics, school-related factors such as teachers' qualification, pupil/teacher ratio, and geographical location of the sampled schools.

The private sector is certainly playing a pivotal role in providing quality education. The competition created by some private schools has also contributed to the improvement of the quality of education in some public schools. It is a healthy sign. However, only better-off families can bear the cost of education in private schools. The justification for a large public sector in primary education is provided by the affordability it offers to low-income households. By examining the share of public schools in the total primary school enrolment, this study has shown that the lowest-income households largely send their children to public schools. Since they cannot afford private education, it is essential to improve the quality of education in public schools. The Government of Pakistan has recently announced that it will recruit only graduates as teachers in primary schools [Pakistan (2001)]; it would be a step in the right direction. Its implementation may improve the quality of education in public schools. The negative impact of high pupil/teacher ratio on the learning achievement suggests that more teachers may be recruited for large public schools.

Improvement in the supervision of public schools was the common theme of focus group discussions in all six districts. The school-based management practised in private schools appears to be superior to the public school management controlled by the provincial or federal authorities. Gains in public school efficiency may be achieved by enabling public schools to adopt the management practices of private schools.

Most of the policy debate on improving the state of primary education in Pakistan has revolved round generating more financial resources for education. Issues related to production efficiency and demand for schooling have received relatively little attention. An effective strategy for solving the problems of primary education in Pakistan should emphasise the need for efficient use of available resources as well as recognise the fact that teachers, households, and the community all play an important role in promoting good-quality primary education.

Finally, to launch a concerted effort for quality education, it is important to pinpoint the areas where the public and private sectors can co-operate. Teacher training is an important area where these sectors can benefit from pooling their resources and expertise. The public sector has not yet benefited substantively from short-term teacher-training courses offered by the private sector, while the NGOs and the private sector cannot afford to organise such training programmes on a large scale. The government should provide the funds needed to organise such courses at the district level for teachers from all three sectors. These courses would provide a forum for teachers from the three sectors to share their experiences as well as help improve teaching skills.

Appendices

Appendix Table 1
Gross Enrolment Rate at the Primary-level by Rural/Urban Areas
and Sex, 1991, 1995-96, 1996-97, and 1998-99A

Area/Sex 1991 1995-96 1996-97 1998-99

Rural Area
 Male 82 81 74 73
 Female 48 54 53 48
 Both Sexes 66 68 64 61
Urban Areas
 Male 97 95 95 93
 Female 87 90 91 90
 Both Sexes 92 92 93 92

Source: PIHS (1991, 1991-96, 1996-97, and 1998-99).

Note: The gross enrolment rate was calculated as the number of
children aged 5-9 years attending primary school divided by
the total number of children in this age group multiplied by 100.

Appendix Table 2
Primary-level Enrolment in Government Schools as a Percentage of
Total Primary-level Enrolment--by Region and Province

 1995-9G PIHS

Region/Province Male Female Both

Urban Areas 57 58 57
 Punjab 55 55 55
 Sindh 54 58 56
 NWFP 73 72 72
 Balochistan 87 87 87
Rural Areas 97 91 91
 Punjab 87 91 89
 Sindh 97 94 96
 NWFP 92 92 92
 Balochistan 97 94 96
Overall 80 78 79
 Punjab 79 78 78
 Sindh 76 70 74
 NWFP 89 86 88
 Balochistan 95 93 94

 1996-97 PIHS

Region/Province Male Female Both

Urban Areas 54 59 56
 Punjab 47 53 50
 Sindh 59 63 60
 NWFP 70 72 71
 Balochistan 82 88 85
Rural Areas 89 91 90
 Punjab 87 90 88
 Sindh 98 97 97
 NWFP 89 91 89
 Balochistan 93 96 94
Overall 78 78 78
 Punjab 75 77 76
 Sindh 78 76 77
 NWFP 85 87 86
 Balochistan 91 94 92

 1998-99 PIHS

Region/Province Male Female Both

Urban Areas 51 55 53
 Punjab 43 50 47
 Sindh 54 60 56
 NWFP 72 65 69
 Balochistan 80 88 83
Rural Areas 87 85 86
 Punjab 83 80 81
 Sindh 95 97 96
 NWFP 87 91 88
 Balochistan 97 98 97
Overall 76 74 75
 Punjab 71 69 70
 Sindh 74 74 74
 NWFP 85 86 85
 Balochistan 95 96 95

Source: PIHS (1995-96, 1996-97, and 1998-99).

Appendix Table 3
Primary-level Enrolment in Government Schools as a Percentage of the
Total Primary-level Enrolment in Rural Areas by Province, Gender,
and Quintile, 199-96, 1996-97, and 1998-99

 1995-96

Quintile/Province Male Female Both

Punjab
 1st Quintile 92 96 94
 2nd Quintile 92 94 92
 3rd Quintile 88 93 90
 4th Quintile 86 9n 88
 5th Quintile 78 80 79
Sindh
 1st Quintile 94 97 95
 2nd Quintile 98 97 97
 3rd Quintile 99 98 c>9
 4th Quintile 99 96 98
 5th Quintile 94 80 89
NWFP
 1st Quintile 96 100 97
 2nd Quintile 93 92 93
 3rd Quintile 97 99 98
 4th Quintile 93 93 93
 5th Quintile 81 81 81
Balcohistan
 1st Quintile 100 100 100
 2nd Quintile 97 100 98
 3rd Quintile 98 94 96
 4th Quintile 96 95 95
 5th Quintile 93 83 89

 1996-97

Quintile/Province Male Female Both

Punjab
 1st Quintile 88 90 89
 2nd Quintile 92 89 91
 3rd Quintile 92 95 93
 4th Quintile 81 84 82
 5th Quintile 76 83 79
Sindh
 1st Quintile 99 100 99
 2nd Quintile 48 100 99
 3rd Quintile 96 93 95
 4th Quintile 98 100 99
 5th Quintile 97 91 95
NWFP
 1st Quintile 97 96 96
 2nd Quintile 93 93 93
 3rd Quintile 92 98 95
 4th Quintile 88 97 q2
 5th Quintile 70 70 70
Balcohistan
 1st Quintile 95 100 96
 2nd Quintile 97 100 98
 3rd Quintile 95 100 97
 4th Quintile 93 94 93
 5th Quintile 89 84 87

 1998-99

Quintile/Province Male Female Both

Punjab
 1st Quintile 90 87 89
 2nd Quintile 89 89 89
 3rd Quintile 84 81 92
 4th Quintile 78 721 75
 5th Quintile 64 65 64
Sindh
 1st Quintile 99 99 99
 2nd Quintile 99 99 99
 3rd Quintile 96 98 97
 4th Quintile 93 97 94
 5th Quintile 90 90 90
NWFP
 1st Quintile 97 99 98
 2nd Quintile 92 94 93
 3rd Quintile 95 95 95
 4th Quintile 82 95 87
 5th Quintile 69 77 72
Balcohistan
 1st Quintile 100 99 100
 2nd Quintile 98 100 98
 3rd Quintile 100 100 100
 4th Quintile 97 96 97
 5th Quintile 92 96 93

Source: PIHS (1995-96, 1996-97. 1998-99).

Note: (a) Enrolment in kachi is excluded.
(b) Quintiles are based on per capita consumption. The 1st quintile
contains individuals with the lowest consumption level, whereas
the 5th quintile contains individuals with the highest
consumption level.

Appendix Table 4
Primary-level Enrolment in Government Schools as a Percentage of
the Total Primary-level Enrolment in Urban Areas by Province,
Gender, and Quintile, 1995-96, 1996-97, and 1998-99

 1995-96

Quintile/Province Male Female Both

Punjab
 1st Quintile 71 66 69
 2nd Quintile 69 G4 66
 3rd Quintile 53 53 53
 4th Quintile 45 51 48
 5th Quintile 30 35 33
Sindh
 1st Quintile 75 84 79
 2nd Quintile 70 67 68
 3rd Quintile 52 61 57
 4th Quintile 45 39 42
 5th Quintile 11 18 14
NWFP
 1st Quintile 87 88 87
 2nd Quintile 90 90 90
 3rd Quintile 68 83 77
 4th Quintile 72 60 66
 5th Quintile 36 42 40
Balcohistan
 1st Quintile 88 98 92
 2nd Quintile 96 91 94
 3rd Quintile 91 94 92
 4th Quintile 86 81 84
 5th Quintile 61 74 68

 1996-97

Quintile/Province Male Female Both

Punjab
 1st Quintile 54 68 63
 2nd Quintile 59 62 61
 3rd Quintile 44 55 50
 4th Quintile 38 48 43
 5th Quintile 22 26 24
Sindh
 1st Quintile 87 86 86
 2nd Quintile 63 70 66
 3rd Quintile 44 64 53
 4th Quintile 49 43 46
 5th Quintile 22 20 22
NWFP
 1st Quintile 86 90 88
 2nd Quintile 85 79 82
 3rd Quintile 77 91 82
 4th Quintile 49 68 58
 5th Quintile 33 29 31
Balcohistan
 1st Quintile 94 94 94
 2nd Quintile 86 93 88
 3rd Quintile 89 92 90
 4th Quintile 70 83 76
 5th Quintile 63 74 68

 1998-99

Quintile/Province Male Female Both

Punjab
 1st Quintile 63 67 65
 2nd Quintile 46 55 50
 3rd Quintile 40 52 46
 4th Quintile 30 47 38
 5th Quintile 18 15 16
Sindh
 1st Quintile 76 82 79
 2nd Quintile 61 71 65
 3rd Quintile 53 51 52
 4th Quintile 36 47 42
 5th Quintile 13 13 13
NWFP
 1st Quintile 86 98 91
 2nd Quintile 85 83 85
 3rd Quintile 73 77 75
 4th Quintile 60 53 57
 5th Quintile 26 28 27
Balcohistan
 1st Quintile 93 93 93
 2nd Quintile 88 87 88
 3rd Quintile 79 95 85
 4th Quintile 75 85 79
 5th Quintile 58 75 67

Source: PIHS (1995-96, 1995-97, 1998-99).

Note: (a) Enrolment in kachi is excluded.
(b) Quintiles are based on per capita consumption.
The 1st quintile contains individuals with the lowest
consumption level, whereas the 5th quintile contains
individuals with the highest consumption level.

Appendix Table 5
Percentage Distribution of Class 4 Children by Subjects,
Marks Obtained, and Type of School

Subjects/ Range of Marks Obtained
School Type

 <40 40-99 60-79
Mathematics
 Public 28.8 29.9 20.4
 NGO 27.4 23.0 25.8
 Private 13.5 19.4 38.8
 All Schools 23.7 24.5 27.7
Urdu
 Public 9.2 31.8 27.9
 NGO 9.1 32.4 21.1
 Private 0.7 15.9 27.7
 All Schools 6.6 27.3 25.6
General Knowledge
 Public 14.0 12.3 25.7
 NGO 11.9 14.5 23.6
 Private 1.4 7.6 19.7
 All Schools 9.5 11.6 23.2

Subjects/ Range of Marks Obtained
School Type

 [greater than
 or equal to] 80 Total

Mathematics
 Public 20.9 100
 NGO 23.9 100
 Private 28.4 100
 All Schools 24.1 100
Urdu
 Public 31.0 100
 NGO 37.4 100
 Private 95.7 100
 All Schools 40.5 100
General Knowledge
 Public 48.0 100
 NGO 50.0 100
 Private 71.3 100
 All Schools 55.6 100

Source: AAPk survey (1999).

REFERENCES

Alderman, Harald, Jere R. Behrman, David R. Ross, and Richard Sabot (1996) Decomposing the Gender Gap in Cognitive Skills in a Poor Rural Economy. Journal of Human Resources 31:1, 229-254.

Alderman, Harald, Jere R. Behrman, Shahrukh R. Khan, David R. Ross, and Richard Sabot (1997) The Income Gap in Cognitive Skills in Rural Pakistan. Economic Development and Cultural Change 46:1, 96-122.

Arif, G. M., and Najam us Saqib (1999) Education for All: A Comparative Study of the Role of Public, Private and NGO Managed Schools in Pakistan's Primary Education. Islamabad, Project report submitted to ActionAid Pakistan.

Arif, G. M., Najam us Saqib, and G. M. Zahid (1999) Poverty, Gender, and Primary School, Enrolment in Pakistan. The Pakistan Development Review 38:4, 979-992.

Bedi, Arjun S., and Jeffery H. Marshall (1999) School Attendance and Student Achievement: Evidence from Rural Honduras. Economic Development and Cultural Change 47:3, 657-682.

Behrman, J. R., A. D. Foster, M. R. Rosenweig, and P. Vashishtha (1999) Women's Schooling, Home Teaching and Economic Growth. Journal of Political Economy 107:4, 682-714.

Burney, Nadeem A., and Mohammad Irfan (1991) Personal Characteristics, Supply of School, and Child School-enrolment in Pakistan. The Pakistan Development Review 30:1, 21-62.

Burney, Nadeem A., and Mohammad Irfan (1995) Determinants of Child School Enrolment: Evidence from LDCs Using Choice-Theoretic Approach. International Journal of Social Economics 22:1, 24-40.

Gazdar, Haris (1998) Universal Basic Education in Pakistan: A Commentary on Strategy and Results of A Survey. Sustainable Development Policy Institute, Islamabad. (Working Paper Series No. 39.)

Glewwe, P., and H. Jacoby (1993) Student Achievement and Schooling Choice in Low-Income Countries: Evidence from Ghana. Journal of Human Resources 29:3, 843-864.

Goldhaber, D. D., and D. J. Brewer (1997) Why Don't Schools and Teachers Seem to Matter Assessing the Impact of Unobservables on Educational Productivity. The Journal of Human Resources 32:3, 505-523.

Hussain, Akmal (2003) Pakistan: National Human Development Report (2003). Karachi: UNDP Oxford University Press.

Jaferi, S. M. Y. (1999) Assessing Poverty in Pakistan. In 'A Profile of Poverty' in Pakistan. Islamabad: Mahbub ul Haq Centre for Human Development.

Khan, Shahrukh Raft, and M. Irfan (1985) Rates of Returns to Education and the Determinants of Earnings in Pakistan. The Pakistan Development Review 24:3&4, 671-680.

Khan, Shahrukh Rafi, Sajid Kazmi, and Zainab Latif (1999) The State of Basic Education in Pakistan: A Qualitative, Comparative Institutional Analysis. Sustainable Development Policy Institute. (Working Paper Series No. 47.)

Mahmood, N., and G. M. Zahid (1992) Measuring the Education Gap in Primary and Secondary Schooling in Pakistan. The Pakistan Development Review 31:4, 729-740.

Mahmood, N., G. M. Zahid, and All Muhammad (1999) Educational Development in Pakistan: Trends, Issues, and Policy Concerns. Pakistan Institute of Development Economics, Islamabad. (Research Report No. 172.)

Mizala, Alejandra, and Pilar Romaguera (2000) School Performance and Choice: The Chilean Experience. The Journal of Human Resources 35:2, 392-417.

Pakistan, Government of (1960) The Second Five-year Plan (1960-65). Islamabad: Planning Commission.

Pakistan, Government of (1965) The Third Five-year Plan (1965-70). Islamabad: Planning Commission.

Pakistan, Government of (1972) The Education Policy, 1972-1980. Islamabad: Ministry of Education.

Pakistan Government of (1978) The Fifth Five-year Plan (1978-83). Islamabad: Planning Commission.

Pakistan Government of (1983) The Sixth Five-year Plan (1983-88). Islamabad: Planning Commission.

Pakistan Government of (1994) The Eighth Five-year Plan (1993-98). Islamabad: Planning Commission.

Pakistan Government of (2000) Pakistan Integrated Household Survey (PIHS) 1998-99. Islamabad: Federal Bureau of Statistics.

Pakistan Government of (2001) Education Sector Reforms: Action Plan 2001-2004. Islamabad: Ministry of Education.

Pakistan Government of (2001a) Census of Private Educational Institutions in Pakistan. 1999-2000. Islamabad: Statistics Division, Federal Bureau of Statistics.

Pakistan Government of (2001b) Economic Survey 2000-2001. Islamabad: Economic Advisor's Wing, Finance Division.

Punjab, Government of (1982) Educational Statistics in Punjab 1979-80. Lahore: Bureau of Education, Education Department.

Punjab, Government of (1984) Educational Statistics in Punjab 1981-82. Lahore: Bureau of Education, Education Department.

Punjab, Government of(1987) Educational Statistics (Schools) in Punjab 1984-85. Lahore: Bureau of Education, Education Department.

Qureshi, S. K., and G. M. Arif (2001) Profile of Poverty in Pakistan, 1998-99. Pakistan Institute of Development Economics, Islamabad. (MIMAP Technical Paper Series No. 5.)

Sander, W. (1999) Private Schools and Public School Achievement. The Journal of Human Resources 34:4, 697-709.

Saqib, Najam us (1998) A Critical Assessment of Free Public Schooling in Pakistan. The Pakistan Development Review 37:4, 955-976.

Sathar, Zeba A., and Cynthia B. Lloyd (1994) Who Gets Primary Schooling in Pakistan: Inequalities Among and Within Families. The Pakistan Development Review 33:2, 103-134.

Zia, Rukhsana (1999) Assessment of Learning at Primary Grades Class-I to Class-V. Bunyad/UNICEF. Occasional Paper, Lahore.

(1) This percentage excludes Kachi class. If this class is taken into account, the share of private schools in primary enrolment is reported to be 28 percent [Pakistan (2000)].

(2) The statistics make no distinction between the NGO and private schools and treat both types of schools as private organisations. A directory, of NGOs working in the field of education shows that about 400 NGOs are operating education programmes in all regions of the country.

(3) However, studies, including the present one, usually have to rely on imperfect measures of teaching ability such as teacher's degree attainment. Obviously, such measures do not account for many other important aspects of teaching quality in the process of imparting education. [Goldhaber and Brewer (1997) and Bedi and Marshall (1999)].

(4) See, for example, Khan and Irfan (1985): Burney and Irfan (1991, 1995). Sathar and Lloyd (1994); Alderman, et al. (1996, 1997), Mahmood (1999): Arif, et al. (1999); Arif and Saqib (1999).

(5) However, because of data limitations, some other factors that may also affect academic achievement, such as nutritional status of a child, conduct by teachers, and scores from the previous class examination, have not been included in the analysis.

(6) Unfortunately, Pakistan is included in those 28 countries that are in serious risk of not achieving the Education For All (EFA) goal. However, it is considered to be capable of achieving EFA goals if the international community helps the country fill the existing technical and financial gaps (Dawn, November 28, 2002, p. 18).

(7) During the Sixth Five-year Plan (1983-88) only 17,000 schools could be opened as against the target of 40,000 schools. Similarly, the target of opening up new primary schools tell short by about 35 percent during the next Plan period (1988-93). No wonder that only 31 million additional children got enrolled at the primary level, 1.5 million less than the Plan target [Pakistan (1994)].

(8) See footnote 2.

(9) However, two government schools in Muzaffarabad district were an exception. They provided primary education to both boys and girls.

(10) This survey was conducted by ActionAid Pakistan (AAPk), and further details on the survey design can be seen in Arif and Saqib (1999).

(11) However, there is also consensus that these tests are only a partial measure of the pupils' educational attainment [Mizala and Romaguera (2000)].

(12) In addition to the assessment of Class 4 students, six focus groups were also organised in the six sampled districts. The composition of all the groups was similar across the districts: 2 to 3 teachers, 1 or 2 parents, 2 education department administrators, 1 or 2 community-leaders, and 1 journalist. Although in these locus groups, a variety of education-related issues were discussed thoroughly, the main locus was on the quality of education in different types of schools and parents' interest in educational problems Discussions carried out in these locus groups generated useful information, particularly on the quality of education, that has been incorporated in this study.

(13) Mizala and Romaguera (2000) have argued that one of the problems with these estimations is that they are based on cross-section analysis, that is, gross values, whereas the dependent variable is the score achieved on the test in a given period. A better estimation procedure, as discussed in the literature, would be to use lagged test scores for the same group of students from a previous period as an independent variable. Such a specification would allow for differential achievement growth based on the previous score. Unfortunately, this type of information is not available.

(14) However, lather's education did not turn out to be significant and was dropped from the analysis.

G. M. Arif and Najam us Saqib are Chief of Research and Senior Research Economist respectively at the Pakistan Institute of Development Economics, Islamabad.

Authors' Note: We are thankful to Dr Faiz Bilquees, Joint Director, Dr Munir Ahmed, Chief of Research, both at PIDE, and to Dr Zafar Iqbal, Economist at the IMF, for their valuable comments on an earlier draft of the paper. We also wish to thank the anonymous referee for useful suggestions to improve the earlier draft of the paper. We are grateful to Mr Masood Ishfaq and Ms Nabeela Arshad, Computer Analysts at PIDE, for their valuable assistance in data analysis, and also to Mr Muhammad Sarwar for typing assistance.

Table 1
Gross Enrolment Rate at the Primary-level by Rural/Urban Area and Sex,
1982-83, 1987-88, 1991, 1995-96, 1996-97, and 1998-99

Area/Sex 1982-83 1987-88 1991 1995-96 1996-97 1998-99

Male 60 63 86 85 80 78
Female 30 34 59 64 64 59
Both Sexes 44 49 73 75 72 69

Source: Mahmood and Zahid (1992); PIHS (1991, 1995-96, 1996-97,
1998-99).

Table 2
Primary-level Enrolment in Government Schools as a Percentage of Total
Primary-level Enrolment by Gender and Rural/Urban Area,
1991, 1995-96, 1996-97, and 1998-99

 Pakistan Rural Urban
Year Male Female Male Female Male Female

1991 87 85 95 94 68 72
1995-96 80 78 90 91 57 58
1996-97 78 78 89 91 54 59
1998-99 76 74 87 85 51 55

Source: PIHS (1991, 1995-96, 1996-97,and 1998-99).

Note: Enrolment in Kachi is excluded. Kachi is a pre-primary class in
Pakistan.

Table 3 Number of Children Enrolled at Primary-level (1-5)
in Primary Schools, Pakistan

 Total Enrolment (000) Female as % of Total

1980 5474 32.6
1981 5741 33.0
1982 6179 32.5
1983 6860 31.7
1984 6828 33.0
1985 7094 33.3
1986 7639 33.1
1987 7959 33.6
1988 9254 33.5
1989 10400 32.1
1990 10837 33.9
1991 10736 34.6
1992 12726 36.1
1993 13288 38.0
1994 14264 39.5
1995 14527 39.3
1996 15395 40.0
1997 17063 41.0
1998 18731 41.8
1999-2000 20399 42.5

Source: Pakistan (2001a).

Table 4
Percentage Distribution of Children in Primary Schools by Income
Quintiles and Type of School for Urban and Rural Areas, 1998-99

 Government Private
Rural-urban Areas Income Quintile Schools Schools

Urban Areas 1st quintile 68 30
 2nd quintile 60 39
 3rd quintile 53 46
 4th quintile 42 56
 5th quintile 16 81
Rural Areas 1st quintile 91 7
 2nd quintile 91 7
 3rd quintile 88 11
 4th quintile 81 17
 5th quintile 73 26

 Other All
Rural-urban Areas Schools Schools

Urban Areas 2 100
 1 100
 1 100
 2 100
 3 100
Rural Areas 2 100
 2 100
 1 100
 2 100
 1 100

Source: Computed from the PIHS 1998-99 data set.

Table 5
Percentage Distribution of Students by Type of School and Gender

Types of School No. of Schools % All Students % Female Students

Government 13 37.1 46.1
NGO 23 33.0 61.6
Private 14 29.9 30.8
All 50 100.0 46.6
(No. of Students) -- (965)

Source: AAPk survey (1999).

Table 6
Selected Student, Family, Teacher, and School
Characteristics by Type of School

 Types of School
Selected Characteristics Public NGO

Mean Age (years) of Class 4 Students 10.2 10.5
Family Characteristics
 % Father Literate 69.2 65.7
 % Mother Literate 25.7 39.4
 Mean Household Monthly Income (Rs) 3271.0 4291.0
Teachers' Characteristics
 Share of Female Teachers (%) 44.4 73.6
 Qualification (%)
 Matriculation -- 9.1
 Intermediate 31.6 41.2
 B.A./B.Sc. 38.8 17.0
 M.A./M.Sc. 18.2 24.5
 Others 11.4 8.2
School Characteristics
 Mean Class Size 52 21
 Mean Student/Teacher Ratio 33 30

 Types of School
Selected Characteristics Private All

Mean Age (years) of Class 4 Students 10.0 10.3
Family Characteristics
 % Father Literate 86.1 73.1
 % Mother Literate 49.5 37.3
 Mean Household Monthly Income (Rs) 5506.0 4350.0
Teachers' Characteristics
 Share of Female Teachers (%) 45.0 54.2
 Qualification (%)
 Matriculation -- 3.0
 Intermediate 2.8 26.1
 B.A./B.Sc. 41.5 32.4
 M.A./M.Sc. 48.1 29.2
 Others 7.6 9.3
School Characteristics
 Mean Class Size 34 36
 Mean Student/Teacher Ratio 31 31

Source: AAPk survey (1999).

Table 7 Average Marks Obtained by Class 4 Students by Subject and
Type of School

Type of School Mathematics Urdu General Knowledge Total

Public 56.6 67.3 71.1 64.0
NGO 58.7 68.2 73.1 65.6
Private 66.2 79.5 83.0 75.1
All 60.2 71.3 75.3 67.9

Source: AAPk survey (1999).

Table 8 Standard Deviation of School Performance (Total Test Score)

 Minimum Maximum
School Type Value Value Average

All Sampled School
 Public 8 46 29.5
 NGO 3 46 30.2
 Private 15 46 34.5
High Scoring Schools (> 70%)
 Public 21 46 38.6
 NGO 23 46 39.1
 Private 18 46 37.4

Source: AAPk survey (1999).

Table 9
Regression of the Effect of School Type, Child, Family, and Teacher
Characteristics and Districts (Dependent Variables = Test Scores
Total and Subject-wise)

 Model 1 Model 2
Correlates (all subjects) (GK (a))

Constant 14.258 *** 0.292 ***
School Type
 Govt. School (Reference Category) -- --
 NGO Schools (=1) -0.005 -0.01
 Private Schools (=1) 3.659 *** 0.082 ***
Child Characteristics
 Age (Years) 0.515 *** 0.021 ***
 Sex (Male=1, Female=0) 0.568 0.009
 Gets Homework (Yes=1, No=0) 5.645 *** 0.131 ***
 Pre-school Attendance (Yes=1, No=0) -0.758 -0.011
Family Characteristics
 Household Income
 Low Income (Reference Category) -- --
 Middle Income (=1) 1.310 ** 0.037 **
 High Income (=1) -0.313 0.022
 No. of School-going Children 0.376 *** 0.010 ***
 Mother's Education (Higher than 0.867 ** 0.007
 Primary=1, Otherwise=0)
Teacher/Class Characteristics
 Sex (Male=1, Female=0) -1.203 ** -0.067 ***
 Education (BA/MA=1, Otherwise=0) 2.426 *** 0.067 ***
 Student/Teacher Ratio -0.054 *** -0.001 ***
District Dummies
 Multan (Reference Category) -- --
 Khushab (-1) 1.009 0.01
 Muzatfarabad (-1) 13.888 *** 0.228 ***
 Malakand (-1) 2.385 *** 0.061 **
 Turbat (-1) 12.051 *** 0.192 ***
 Khairpur (-1) 2.539 *** 0.066 ***
 Adjusted [R.sup.2] 0.55 0.31
 F 66.662 24.837
 N 953 953

 Model 3 Model 4
Correlates (Urdu) (Mathematics)

Constant 0.395 *** 0.241 ***
School Type
 Govt. School (Reference Category) -- --
 NGO Schools (=1) -0.017 0.022
 Private Schools (=1) 0.087 *** 0.071 ***
Child Characteristics
 Age (Years) 0.007 0.010 **
 Sex (Male=1, Female=0) -0.007 0.033 ***
 Gets Homework (Yes=1, No=0) 0.134 *** 0.106 ***
 Pre-school Attendance (Yes=1, No=0) -0.02 -0.017
Family Characteristics
 Household Income
 Low Income (Reference Category) -- --
 Middle Income (=1) 0.051 *** 0.002
 High Income (=1) 0.026 -0.033 **
 No. of School-going Children 0.009 *** 0.007 **
 Mother's Education (Higher than 0.303 ** 0.015
 Primary=l, Otherwise=0)
Teacher/Class Characteristics
 Sex (Male=1, Female=0) -0.033 ** 0.005
 Education (BA/MA=1, Otherwise=0) 0.062 *** 0.035 **
 Student/Teacher Ratio -0.001 ** -0.002 ***
District Dummies
 Multan (Reference Category) -- --
 Khushab (-1) 0.001 0.050 ***
 Muzatfarabad (-1) 0.286 *** 0.362 ***
 Malakand (-1) -0.038 0.131 ***
 Turbat (-1) 0.241 *** 0.325 ***
 Khairpur (-1) 0.008 0.093 ***
 Adjusted [R.sup.2] 0.46 0.53
 F 45.347 61.125
 N 953 953

(a): GK=General Knowledge.

** Statistically significant at the 0.05 level.

*** Statistically significant at the 0.01 level.

Source: AAPK survey (1999).

Table 10
Regression of the Effect of School Type, Child Family, and Teacher
Characteristics and Districts (Dependent Variable: Total Test Score)

 Model 1 Model 2
Correlates (all subjects) (GK (a))

Constant 14.353 *** 27.142 ***
School Type
 Govt. School (Reference Category)
 NGO Schools 0.108
 Private Schools -- 3.760 ***
Child Characteristics
 Age 0.465 *** --
 Sex (Male=l, Female=0) 0.710 --
 Gets Homework (Yes=1, No=0) 6.270 *** --
 Pre-school Attendance (Yes=1, No=0) -0.588 --
Family Characteristics
 Household Income
 Low Income (Reference Category) -- --
 Middle Income 1.683 *** --
 High Income 0.539 --
 No. of School-going Children 0.355 *** --
 Mother's Education (Higher than 0.152 --
 Primary =1, Otherwise=0)
Teacher/Class Characteristics
 Sex (Male=1, Female=0) -0.65 -0.931
 Education (BA/MA =1, Otherwise=0) 3.449 *** 1.996 ***
 Student/Teacher Ratio -0.040 *** -0.058 ***
District Dummies
 Multan (Reference Category) -- --
 Khushab (=1) -0.265 0.289
 Muzaffarabad (=1) 13.615 *** 13.933 ***
 Malakand(=1) 1.458 2.504 ***
 Turbat(=1) 11.178 *** 13.165 ***
 Khairpur(=1) 1.732 ** 1.688 ***
 Adjusted [R.sup.2] 0.52 0.53
 F 65.959 109.055
 N 953 953

 Model 3 Model 4
Correlates (Urdu) (Mathematics)

Constant 13.423 *** 5.567 ***
School Type
 Govt. School (Reference Category) --
 NGO Schools 0.697 0.338
 Private Schools 4.166 *** 3.3752 ***
Child Characteristics
 Age 0.318 ** 1.332 ***
 Sex (Male=1, Female=0) 1.375 *** 0.022
 Gets Homework (Yes=1, No=0) 5.183 *** 7.709 ***
 Pre-school Attendance (Yes=1, No=0) -0.481 0.086
Family Characteristics
 Household Income
 Low Income (Reference Category) -- --
 Middle Income 1.714 *** 2.232 ***
 High Income 0.276 2.286 ***
 No. of School-going Children 0.385 *** 0.798 ***
 Mother's Education (Higher than 0.587 0.826
 Primary =1, Otherwise=0)
Teacher/Class Characteristics
 Sex (Male=1, Female=0) -- 1.200 **
 Education (BA/MA =1, Otherwise=0) -- 2.556 ***
 Student/Teacher Ratio -- -0.106 ***
District Dummies
 Multan (Reference Category) -- --
 Khushab (=1) 1.08 --
 Muzaffarabad (=1) 14.461 *** --
 Malakand(=1) 2.321 *** --
 Turbat(=1) 13.693 *** --
 Khairpur(=1) 4.063 *** --
 Adjusted [R.sup.2] 0.53 0.34
 F 71.757 38.245
 N 953 953

Source: AAPK survey (1999).

** Statistically significant at the 0.05 level.

*** Statistically significant at the 0.01 level.

Table 11 Hypotheses Testing

Null Hypotheses F-value F-critical Result

No Effect of School System 31.68 4.61 rejected
No Effect of Household Characteristics 7.47 2.51 rejected
No Effect of Teacher Characteristics 19.70 3.78 rejected
No Effect of Geographical Location 92.18 3.02 rejected

Source: AAPK survey (1999).

Note: All hypotheses were rejected at I percent level of significance.

Table 12
Average Marks Obtained by Class 4 Children by Subject, School,
and District

District School Type Math. Urdu

Multan Public 43.1 61.3
 NGO 44.3 56.9
 Private 47.1 72.6
 All 44.8 63.2
Khushab Public 45.3 57.1
 NGO 49.1 65.9
 Private 50.5 70.6
 All 47.0 61.3
Turbat Public 89.1 93.7
 NGO 85.7 89.9
 Private 80.5 91.6
 All 85.4 92.0
Khairpur Public 55.2 68.3
 NGO 51.3 61.5
 Private 66.9 73.4
 All 55.5 61.1
Malakand Public 45.2 53.5
 NGO 55.9 55.8
 Private 70.0 75.3
 All 83.7 94.2
Muzaffarabad Public 80.4 87.6
 NGO 87.9 96.0
 Private 82.3 97.5
 All 60.2 71.3

District School Type G. Knowledge All Tests

Multan Public 67.8 55.7
 NGO 59.4 52.6
 Private 74.2 63.0
 All 66.6 56.8
Khushab Public 56.5 52.3
 NGO 74.4 61.4
 Private 74.5 63.7
 All 63.6 56.3
Turbat Public 91.9 91.5
 NGO 93.3 89.1
 Private 90.5 87.0
 All 91.9 89.4
Khairpur Public 72.3 64.1
 NGO 69.7 59.5
 Private 80.9 72.7
 All 72.3 61.6
Malakand Public 66.7 53.4
 NGO 60.4 56.9
 Private 87.4 76.1
 All 91.3 89.4
Muzaffarabad Public 86.7 84.6
 NGO 95.8 92.8
 Private 90.4 89.8
 All 75.3 67.9

Source: AAPk survey (1999).