Is there a quantity-quality tradeoff as enrollments increase? Evidence from Tamil Nadu, India

\*/PS 1" POLICY RESEARCH WORKING PAPER 1768 Is There a Quantity-Quality Does greater access to education come at the Tradeoff as Enrollments expense of educational Increase? quality? What public policies can diminisn the decline in quality of ecducation when Evidence from Tamil Nadu, India public furnds are limited and enrollrnents are expanding? P. Duraisamy Estelle James Julia Lanze Jee-Peng Tan The World Bank Policy Research Department Poverty and Human Resources Division and Human Development Department Education Division H May 1997  [OLICY RESEARCH WORKING PAPFR 1768 Summary findings Developing countries have been quite successful at resour-es that exist have not always been efficiently expanding school enrollment, especially at lower levels. utilized. Most notably, the student-teacher ratio has risen But for any given level of educational efficiency, dramatically over the past decade. In addition, while increased enrollments require increased resources, in many new schools hlave been started, many of them do order to maintain qualitv. If those resources are not not have buildings, or have only meager buildings. The forthcominig, the increase in educatiotnal quantity may qualit' of ediucation, as mneastired by the pass rate on the come at the expense of educational quality. state" ide tenth grade examination, has suffered as a When public budgets are constrained, is there a result The study shows that the rise in the student- tradeoff between quantity and quality of education? If teacher ratio and the consequent diminution of the so, what public policies can diminish the decline in growth rate in examination passes has been greatest in quality? distri, ts with the fastest enrollment growth - evidence Duraisamy, James, Lane, and Tan find a negative effect of a quantity,-quality tradeoff. of expanded enrollments on school conditions and Districts with a high proportion of privately-nianaged learning, using a cross-district time series analysis of schools perform better. Policy changes such as greater Tamil Nadu, [ndia. A wide arrav of initiatives use o, private managemiient and finanice and greater local undertaken by the governmenit of Tamil Nadu has made discrr tion in publicly-managed schools imlight improve schools accessible and attractive to families. But the the situation. resources have not kept up with enrollnients and those This paper - a joint product of the Poverty and HuLrnan Resources Division, Policy Research Department, and the Education Division, Human Development Department-is part of a larger effort in the Bank to tinderstaid and disseminate the determinants of educational quality. Copies of the paper are available free fronm the World liank, 181 8 H Street NW, Washington, DC 204.33. Please contact Selina Khan, room N8-024, telephone 2(12-47.3 -3651, fax 202-522-1153, Internet address skhan8((0worldbank.org. May 1 997. (31 pages) The 1'olicy Research Working Paper Series dissemninates the findings of uwork in p)rogress to encouerage the exchange of ideas about dervelopment issues. An objectiv'e of the series is to get the findings out quwkly, een if the presentations are less thanfully polished. 'The papers carry the names of the authors ,ind should be tited ac, ordingly. '1 hef indings, interpretations, and conclusions expressed in this paper are entirely hose (fi the authors. 'Iicy do not necessaiily represeict the vieur of the World liank, its Fhxecutive Directors, or the countries the)s represent. 1'rodiucecd by the Plolicy RcKearc:!i Dissemination Center  Is There A Quantity-Quality Trade-Off As Enrollments Increase? Evidence from Tamil Nadu, India By P. Duraisamy University of Madras Estelle James World Bank Julia Lane American University Jee-Peng Tan World Bank Acknowledgement. The authors would like to thank C. Palanivelu and his staff at the Tamil Nadu Department of Education for facilitating our access to the data.   Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I. The Impact of Increasing Enrollments on School Conditions . . . . . . . 2 Teachers and Class Time . . . . . . . . . . . . . . . . . . . . . . . 4 Buildings and Other Facilities . . . . . . . . . . . . . . . . . . . 7 II. What Has Been the Impact on Examination Results? . . . . . . . . . . . 8 The Model and Data . . . . . . . . . . . . . . . . . . . . . . . . . 9 The Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 III. How Can Private Resources Augment Public Resources? . . . . . . . . 14 IV. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Appendix: Did State Policies Increase Enrollments? . . . . . . . . . . . 19 Figure 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Tables Table 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Table 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Table 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Table 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Table 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Table 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Appendix Table 1 . . . . . . . . . . . . . . . . . . . . . . . . . 29 Appendix Table 2 . . . . . . . . . . . . . . . . . . . . . . . . . 30 Appendix Table 3 . . . . . . . . . . . . . . . . . . . . . . . . . 31   Is There A Quantity-Quality Trade-Off As Enrollments Increase? Evidence from Tamil Nadu, India Expanding access to education, especially at lower levels, is a common objective of governments in developing countries, and it has met with considerable success over the last two decades (see, for example, UNESCO 1993; World Bank 1995) . But for any given level of efficiency, increased enrollments require increased resources, in order to maintain quality. Countries may not always take this into account in their educational plans. Does the increase in educational quantity come at the expense of quality? What public policies can raise efficiency or otherwise diminish this quantity-quality trade-off, in the face of strong constraints on public budgets? This paper explores these issues, using Tamil Nadu, India, as a case in point. We document the sustained and rapid increases in enrollments in Tamil Nadu since the 1970s, in part due to state government policies. However, the expansion has not been an unmitigated success. Because the education system has always depended heavily on public funding, and because these resources did not keep up with enrollments, conditions of schooling, particularly teacher-pupil ratios at the middle and secondary levels, declined sharply. We investigate whether this in turn has had a negative impact on academic performance, as measured by the pass rate on the tenth grade examination. The limited role that the private sector has been allowed to play has exacerbated the conflict between educational quantity and quality. In Tamil Nadu many schools--21 per cent at the primary-middle levels and 28 per cent at the secondary level--are "aided", that is, set up and run by private management but largely financed by the state, hence a public-private hybrid. While their teacher salaries are paid by the state, the managers are expected to provide the building and other school facilities. In exchange for public funding they give up the right to charge fees, but they retain other managerial rights, such as the right to select their teachers. Fewer than 1 per cent of schools are private, unaided and fee-financed. Therefore, as a secondary question, this paper 1  analyzes whether quality might be raised by expanding the number of privately- managed schools and by using private finance to loosen the resource constraint for the system as a whole. We utilize two sources of information: evidence based on aggregate time series data at the district level, drawn from (mostly unpublished) administrative records kept by Tamil Nadu's Department of Education; and evidence based on a field trip to 26 government and private schools in rural and urban areas around Madras and Madurai by three of the authors in July 1994.1 These visits offered an opportunity for discussion with school heads and teachers and for first-hand observation of actual school conditions. These diverse sources of information led us to conclude that Tamil Nadu was not willing or able to maintain a constant level of public per student funding as enrollments rose through the 1980's, so school conditions, and to some extent learning, fell. Part I asks, what were the consequences of increased enrollments for school conditions? Part II analyzes the impact on learning, as measured by the growth rate of passes on the 10th grade examination, and evaluates whether this effect differs across districts depending on their use of private management. Part III considers whether greater reliance on private finance might improve the situation. The Conclusion summarizes policy implications. I. The Impact of Increasing Enrollments on School Conditions During the past two decades the Tamil Nadu government has adopted numerous measures designed to increase enrollments: setting up a primary school in every village with more than 1,000 inhabitants (the threshold is now being lowered to 500 inhabitants); providing free mid-day meals 365 days a year to every child in primary and middle school; and giving free uniforms and books to every child in Although we had a general letter of introduction from the state government, these visits were not announced beforehand. The choice of schools to visit was random in the sense that we stopped at the schools we happened upon along our route; in some places we also asked local people for directions to particular types of schools (e.g. public or private aided or private unaided) we wished to see. No school denied us a visit. 2  the mid-day meal program.2 Some measures specifically encourage girls' schooling: special "marriage grants" to those who complete 8th grade; cash awards to headmasters who enroll and retain high proportions of girls from scheduled castes and scheduled tribes; hiring of large proportions of women teachers; and since 1.991, stipulating that only women can be hired as new teachers in primary schools. Considerable publicity has been given to the importance of education and literacy, backed up by awards to communities that achieve high enrollment and school completion rates (see Govt. of Tamil Nadu, India 1994 for details on the various initiatives). That the school system has expanded rapidly during the past two decades is apparent from the data in Table 1. Although the network of primary schools was already relatively developed in 1977 it continued to grow, adding another 2,700 schools by 1992. The gross enrollment ratio at the primary level rose from 93 percent in 1977 to 101 percent by 1992, suggesting nearly universal attendance among recent cohorts of primary school-age children. Coverage also took off after the mid-1970s at the middle school level, followed a decade later by a similar trend at the high school and higher secondary school levels. Thus between 1977 and 1992 age enrollment ratios rose from 57 percent to 96 percent at the middle school level and from 33 percent to 47 percent at the high school and higher secondary levels.3 The female share of enrollments increased at all levels. The 2 The mid-day meal program was initially intended only for children from low-income families. It is now open to everyone, but children from higher income households self-select themselves out of the program. Besides meeting the recipients' nutritional needs, the meals can be viewed as a payment to families to cover the opportunity cost of schooling. The uniforms provided do not always fit and are not always worn, but they do reduce the private cost of schooling to low-income families. 3 Our data on age enrollment ratios are based on actual grades in which students are enrolled divided by size of population in the relevant age group. However, our data on number of students and teachers are based on category of school in which students are enrolled. Schools in Tamil Nadu are categorized according to the highest grade that they offer. Since some middle schools began to offer higher grades during this period, they became recategorized as high schools, and since many students in middle school grades attended schools categorized as high schools, the growth in high schools and high school students appears much larger than the growth in age enrollment ratios at the high school level, and vice versa at the middle school level. The age enrollment data give the better picture of enrollment coverage for each group of potential students. See Table 1 for 3  Appendix presents evidence on the degree to which this enrollment expansion was due to state policies versus exogenous family circumstances. The focus of this stuidy is the degree to which these enrollment increases were accompanied by worsening conditions of schooling, and the consequences of this deterioration, as public spending on education, particularly on teaching resources and physical facilities, did not keep up with the growth in student numbers. Teachers and class time While enrollments at the primary and middle school levels expanded by 35 percent between 1977 and 1992, the number of teachers increased only 4 percent in the same period. Inevitably, the ratio of pupils to teachers rose--from 36 in 1977 to 47 in 1992 (Table 2). By 1992, pupil-teacher ratios in primary and middle schools across districts ranged between 39 and 54. At the high school and higher secondary levels the pattern of growth in numbers of pupils and teachers was even more striking (Table 3). Between 1977 and 1992, enrollments rose by 139 percent, reflecting substantial increases in conitinuation rates to secondary education, while the number of teachers grew only 44 per cent. Consequently, the pupil-teacher ratio rose from 25 to 42, an increase of 66 percent. On average, for every 1 percentage point increase in number of students since 1977, there was only a 0.11 percentage point increase in number of teachers at the primary and middle school levels and a 0.32 percentage point increase at the high school and higher secondary levels. At all levels, pupil-teacher ratios in Tamil Nadu are among the highest in the world today (see UNESCO 1993 for data on other countries). The rising pupil-teacher ratio has been exacerbated by the differential rate of enrollment growth across districts and the fact that teacher allocation has not responded to these differences, so districts with the fastest growing enrollments also had the fastest rising pupil-teacher ratios (Table 3) . For example, enrollments in primary and middle schools increased 44 per cent in Chengalpattu district between 1977 and 1992, compared with 25 per cent in Salem. further explanation. 4  But the number of teachers increased at almost the same rate in these 2 districts--5 and 4 per cent, respectively. As a result, the pupil-teacher ratio rose 37 per cent in the former, ending up at 51, while it rose only 19 in the latter, ending up at 45. Between 1977 and 1992, secondary school enrollments more than tripled in Dharmapuri but less than doubled in Kanniyakumari, while teachers grew at the same rate--about 54 per cent--in both districts. As a result. the pupil-teacher ratio more than doubled in Dharmapuri, to 50, but it grew only 25 per cent, to 32, in Kanniyakumari. In this paper we exploit these district differences, which become the basis for the regression analysis that is described in the next section. One reason for the slow growth in number of teachers is the fact that the salary scale for government-paid teachers is far above the market rate (in some private unaided schools that we visited, teachers were paid only one-quarter as much as government-paid teachers), and that cost-of-living adjustments are built into the salary scale. Thus, most of the increase in available public funding has gone to a relatively constant number of teachers in the form of salaries and salary increases, rather than to a larger number of teachers. High pupil-teacher ratios mean that students attend classes in very large groups. In some districts visited by the authors, groups of 60 to 100 students are commonplace. In secondary schools teaching resources are especially constrained because teachers work only 24 periods (18 hours) a week, while students are present for 34 periods. Thus, the nominal pupil-teacher ratio overstates the effective class size by about 40 percent. For example, if there is one teacher for every 50 pupils, the average class size would be 70, not 50, because each teacher only teaches for 70 percent of the full day. The sparseness of teaching resources per student is compounded by other factors that reduce effective teaching time, such as: 0 The slowness to fill teaching vacancies when they occur due to retirement or transfer. Headmasters in public schools do not have the authority to hire but must wait for the Recruitment Board in Madras (the state capital) to send replacements. This process can take a year or longer, 5  particularly during periods of budget stringency at the state level. many of the schools we visited had at least one vacant position. 0 The lack of substitute teachers to cover for absent teachers. When a teacher is absent, the class (of 50-100 students) is merged with another class or sits by it:self. Teachers are eligible for 14 days of "casual leave" annually, which is 6.5 percent of the school year. In addition, a 3-month paid maternity leave is allowed, as well as 18 months of sick leave over a teacher's career (which teachers usually take toward the end if they haven't already used it up earlier). Thus, one can expect a career teacher who works for 30 years, has 3 children and takes full sick leave and "casual leave" (as most do) to be absent an average of 34 days a year, or 16 percent of total school time. At most schools visited by the authors, the rate of absenteeism plus vacancies ranged between 10 and 20 percent. * The participation of teachers in non-teachingr activities. Teachers are expected to perform numerous other duties which cut into teaching time. For example, they are expected to help collect data for the census, assist in elections, and show up for in-service training. Other non-teaching activities cause teachers to arrive late or depart early, and additional tir.ie is lost to teacher strikes and political agitation. In 1993, for example, public school teachers were on strike for 2 months, and schools simply closed during that time. Privately-managed aided schools seem to have an advantage in retaining and utilizing the services of their teachers. First of all, their managers choose their teachers and can select people they judge to be conscientious, rather than simply accepting teachers who are assigned by madras; at least in principle, they can also fire ineffective teachers. Second, when a teacher leaves, the vacancy can be filled immediately at the discretion of the school manager, even though the state pays the salary. Thus, aided schools are less likely to have vacant positions than are public schools. Third, absent teachers can be replaced using resources provided and at salary scales determined by the school's management; 6  in our field visits we observed fewer missing teachers and classes without teachers in aided schools. And fourth, teachers in aided schools are likely to spend less time in out-of-school activities.4 In sum, the teaching input per student has been declining, even that small input is missing over 20 percent of the time, but less so in privately- than in publicly-managed schools, and political economy forces that keep public school teacher salaries relatively high have added to the problem. It is hardly surprising that rote learning is emphasized under these circumstances, both in the classroom and in the examinations. Given the on-going debate about the importance of resources versus pedagogical processes, it is worth noting that to some extent the latter is constrained by the former, making pedagogical improvement difficult in Tamil Nadu. Buildings and other facilities The government of Tamil Nadu has been very successful in starting new schools, so that practically every village has a primary school and easy access to a middle and secondary school as well. But many of the schools have no buildings, which means that lessons are held outdoors, on the ground, and are suspended whenever it rains, perhaps 5 percent of the time. This is yet another way in which teaching time is lost. Even in schools with buildings, the typical arrangement consists of one large unfurnished classroom, sometimes divided into several sections by partial partitions, each section accommodating 50 to 100 students. Tables and chairs are lacking and most students sit on the dirt floor for their lessons. Most schools we visited did not have lavatories. On average, 40 percent of Tamil Nadu's primary schools in the rural areas have no building or only one room, and nearly a quarter of the classes lack a blackboard (Table 4). Many of these ore-room schoolhouses have only one teacher. It is hardly surprising, given these difficult conditions, that the high rate of teacher absenteeism is more than matched by the students: in some of the 4 In most of the aided schools visited by the authors, conditions and discipline were visibly better than those in government schools. All the private unaided fee-charging schools visited were equipped with tables and chairs and classes contained fewer than 40 pupils. 7  schools visited by the authors, students absenteeism ranged as high as 25-50 percent. In this sense, the enrollment data overstate the actual number of students days, by 35-100 percent. It is perhaps more surprising that students remained enrolled in school despite these poor and declining conditions (see Hanushek and Lavy 1994 for a study of enrollments in Egypt, a case where students were found to have dropped out because of poor quality education.) II. What Has Been the Impact on Examination Results? To what extent has the decline in school conditions adversely affected student learning? Can policies be adopted that would minimize these adverse effects, despite the tight public budget constraint? Ideally we would examine these issues using individual-level time series data on achievement scores and student characteristics.5 Such data are unavailable in Tamil Nadu, however, so we rely on time series district-level data, using the results on the state-wide tenth grade public examinations as a proxy for learning outcomes.6 The many generic problems with the use of such an exam as an outcome indicator include the fact that it may measure rote learning rather than analytic skills, and grading standards may differ across districts and time periods. In this case examination papers are graded centrally, implying that the same standards were applied across all districts. It is possible that grading standards have changed though time, thereby raising or depressing the number of passes; however, this effect would be picked up by the time trend included in the regression analysis and the cross- district regression analysis presented below should remain valid. Bashir (1994) is a rare example of a study using achievement data for individual students in Tamil Nadu. Her study focused on cross-sectional differences in achievement outcomes between public and private schools rather than on changes in achievement over time. Kingdon (1996) also uses individual student data to compare standardised achievement differentials between students of private and public schools in Uttar Pradesh. Students also sit for state-wide examinations at the end of the twelfth grade. Because the available data for these examinations refer to a relatively short time series of four years, we focused only on the data for the tenth grade examination. 8  Many recent studies have shown that expenditures per student are not a good predictor of school performance (Hanushek 1986 and 1994), so we might expect the deteriorating school conditions in Tamil Nadu to exert only a negligible adverse on examination results. But few of the studies pertain to situations where resource levels are very low and where changes have been as dramatic as those observed in Tamil Nadu. Between 1977 and 1992 the number of candidates for the tenth grade examinations grew by 73 percent for the state as a whole, from a base of nearly a quarter million candidates (Table 5). During the same period the number of passes rose by 118 percent, lifting the pass rate from 54 percent in 1977 to 69 percent in 1992. Much of the improving trend in pass rates reflects the above average performance of the increasing proportion of female candidates. These aggregate patterns would seem to suggest that the deteriorating conditions of schools in the state did not hold back examination performance. But it is also possible that the effect of deteriorating school conditions was negative and offset by other factors such as secular improvements in parental education and nutritional status of the children. In other words, had school conditions not deteriorated, examination outcomes might have been better still. We use cross-district regression analysis below to estimate how much better performance would have been. The model and data We model growth in tenth grade examination passes as a function of growth in candidates, changes in schooling conditions as proxied by the pupil-teacher ratio, the prevalence of privately-managed schools in each district, as well as a time trend to proxy systematic changes over time in, for example, social norms on schooling, parental education levels, children's nutritional status, labor market incentives and grading standards on the examinations. In an alternative specification we replace the private school variable with district dummies to capture all systematic differences across districts, including differences in the distribution of schools by management type. Data limitations precluded expansion of the regression models beyond this set of regressors. 9  Our data pertain to time series information for 15 districts during 1977 to 1992 culled from administrative records kept by the Government of Tamil Nadu.' We transform each variable (enrollments, teachers, candidates and passes) into an index based on its 1977 level, to facilitate comparisons over time across districts which vary widely in population size. nD focal point is the impact of the pupil-teacher ratio on educational performance. We note that each cohort's examination performance reflects the cumulative effects of past schooling conditions. Because students were exposed to middle school conditions for three years but high school conditions for only one year before taking the examination, we use the pupil-teacher ratio for middle school (grades 6-8), appropriately lagged, for each year of examination data. For the reasons elaborated earlier we expect this variable to have a negative effect on learning. As an indicator of the prevalence of private management, we used the percentage of private aided schools in the district. We expect this to have a positive effect, in light of the greater flexibility that private schools have in Tamil Nadu (see above), as well as findings from recent studies suggesting that privately-managed schools achieve greater efficiency or academic value-added than publicly-managed schools (Jimenez and Lockheed 1995; James, King and Suriyadi.,1996). This positive measured effect may occur because decision-making in privately-managed schools is less bureaucratized, more localized, closer to the locus of information about school processes (a true value-added), but it may The time series refer to the following years: 1977, 1980, 1983, 1986, 1988, 1989, 1990, 1992. Data for the more recent years refer to 23 districts, some of which existed previously as combined districts. Because the data for earlier years are available only for 15 districts, we amalgamated the data for the newly created districts to produce a complete time series for 15 districts for our analysis. The available data show the number of students and teachers in primary, middle, high and higher secondary schools. Because many schools offer the full range of grades, from kindergarten to higher secondary grades, the data do not correspond exactly to a breakdown by grade or cycle of education. For example, a child attending the primary wing of a high school is counted as a high school student in this data set. To minimize the effect of this problem in our calculation of student/teacher ratios, we use the same classification system for teachers. The data contain information on the composition of schools by type of management, as well as the number of candidates taking and passing the state-wide public examinations at the end of the tenth grade. 10  also be an artifact of an unobserved selection process in which private schools attract students with higher ability and motivation and weed out the others. In this analysis, which uses district pass rates and district proportions of private schools, selection bias is minimized because it can only operate through attracting better students from other districts, or getting students to drop out of any school in the district before they become candidates. In Tamil Nadu the share of schools that are privately-managed varies widely across districts but has remained relatively stable within each district in the period under analysis (see Appendix Table 3). As a result the variable for private schools is almost collinear with the district dummies, and we could not use it together with the district dummies in the same regression. District dummies capture the influence of private management as well as that of other factors such as the increased number of students from disadvantaged backgrounds who reach the tenth grade, sit for the exam, and ultimately fail, possible differences in grading standards across districts (unlikely, as noted above), variation in social norms and so on. Note that to the extent that these effects are uniform across villages, they will be picked up by the time trend rather than the district dummies. Because the district dummies capture managerial capacity plus a wide range of other effects, they should have greater predictive power than managerial type alone, although with less specificity about the forces at work. We control for growth in the number of candidates over this period, which would normally have a strong positive effect on the growth in number of passes. Ideally we would have used the growth in number of ninth grade students, instead of carndidates for the examination, as the independent variable for predicting growth in examination passes. This would minimize the effect of selection bias associated with the likelihood that schools encourage only better prepared students to sit for the examinations, an effect that may differ across villages 11  and time. However, information on enrollment by grade was not available so number of candidates was used.8 In a separate equation we explain growth in number of candidates as a function of growth in enrollments at the high school and higher secondary levels, the prevalence of privately-managed schools (or, alternatively, district dummies), and a time trend. The rosults The regression results appear in Table 6. Not surprisingly, the growth in passes at the tenth grade examinations depends mainly on the growth in examination candidates: a 1 percentage point increase in candidates results in a 0.99 percentage point increase in examination passes (all relative to 1977). The growth of candidates, in turn, is a function of the growth in enrollments: a 1 percentage point increase in upper school enrollments (i.e. in high school and higher secondary school) produces a 0.72 percentage point increase in candidates. These results suggest that policies aimed at getting students into school and keeping them there will have the most powerful effect on the stock of examination passes. Most important for our purposes, pupil-teacher ratios do indeed exert a negative influence on examination passes: holding number of candidates constant, every additional pupil in the class held back the growth in passes by 1.2-1.5 percentage points, depending on specification. Between 1977 and 1992, the state- wide average middle school pupil-teacher ratio rose by 12 pupils, from 35 to 47. According to our regression estimates, the number of passes on the tenth grade examinations in 1992, relative to the number in 1977, would have been 15-18 percentage points higher if the 1977 pupil-teacher ratio had been maintained. Because students attending grade 9 are spread across institutions which are labelled as middle, high, and higher secondary schools, it is not possible to compare the number of candidates directly to the number of ninth grade students. The number of candidates grew much slower (73%) than enrollments at the high and higher secondary levels (139%) . This was partly due to the fact that enrollments categorized as secondary in fact included an increasing number of students who were actually sitting in middle school grades. It may also have been due partially to increased selectivity in continuation from middle to high school grades or in exam takers among those attending high school grades. The available data do not enable us to disentangle these explanations. Also see footnote 3. 12  That implies the percentage of candidates passing the examination in 1992 would have been 73 percent instead of 68 percent (see Figure 1). As noted above the pupil-teacher ratio is subject to substantial measurement errors, implying that its regression coefficient is likely to be biased downward in absolute value. In addition, if increased selection occurred in districts with high and rising pupil-teacher ratios, this too would have biased downward its coefficient. Offsetting\these two negative biases is the positive bias stemming from the possibility that districts with rapidly rising enrollments and pupil-teacher ratios are recruiting students from less privileged backgrounds, where accumulation of human capital in the household is lower. Unfortunately, we do not have the data that would be needed to test for or eliminate these possible biases. On balance, the available evidence implies that enrollment growth has had the primary impact of increasing the number of students who take and pass the exam, but the concomitant decline in school conditions, as reflected in dramatically rising pupil-teacher ratios, has had a partially counteracting negative effect on student performance. Also as expected, the distribution of schools by management type affects performance across districts: a 1 percentage point rise in the proportion of aided schools increases the growth rate of passes by 0.22 percentage point. Thus, if we compare Periyar-Coimbatore, where 29.5 per cent of the secondary schools are privately-managed aided or unaided, with Pudukottai, where only 13.5 per cent are in this category, the expected number of examination passes in each year relative to 1977 would be 3.5 percentage points higher in Periyar, ceteris paribus. If this is true value-added, it suggests that some practices of aided schools, such as the right to select their own teachers (subject to state standards), to hire substitute teachers and to fill vacancies expeditiously, may contribute to their effectiveness and should also be extended to public schools. Over-all, our regression results suggest that there is a quantity-quality trade-off if public budgets are constrained, and policies aimed at improving efficiency through choice of managerial type will improve examination outcomes as enrollments increase. 13  III. How Can Private Resources Augment Public Resources? The decline in school conditions, with its adverse effects on achievement, arises in part because the education system in Tamil Nadu depends predominantly upon government funding and government funding has become increasingly scarce. Until recently, the state government did not permit augmentation by private resources. Fees were not allowed in public or aided schools and even voluntary contributions were prohibited. These restrictions are found in many countries in public and private subsidized schools, as a condition for government aid (James 1991) . In most cases the ostensible rationale for prohibiting fees and contributions is equity-- wealthier families or communities should not be allowed to secure better quality education for their children by paying. A related political economy rationale is that many low and middle-income families do not have a high effective demand (i.e. a willingness to pay) for incremental quality, do not want the market to push up the price of schooling for them, and do not want others to gain a competitive advantage in the labor market by attending schools that provide better quality through private finance. Their political clout results in the restrictions on fees and contributions (see James 1993, for a more rigorous cross-country analysis of the nolitical economy of government spending on education). At the same time, above-market salary scales have often been set for government-paid teachers, who are a well-organized politically articulate group. In Tamil Nadu, as we have seen, the result was a resource base that failed to keep pace with enrollment growth, with most of the monetary increase going to teachers in the form of higher wages. One way out of this situation is to increase public spending on education. But in many cases the inability to pay on the part of low income groups, and the unwillingness of high income groups to subsidize them, limits the political feasibility of this option. Another way out is to tap private resources and to allow these resources to be used to hire additional teachers through the private 14  market. This increases the total resources used for education and may also increase the efficiency with which these resources are transformed into academic outputs (James and King 1996, Jineniez and Paqueo 1996) . In recent years, perhaps because school conditions have n1ow sunk so low that middle class support has shifted, the government has indeed begun to institute such policies. For example, fee-charging schools are now permitted to open freely, and the number of publicly financed aided schools is no longer expanding. "Matriculation schools", which are self-financed English-medium schools, are by far the fastest- growing part of the education sector. Privately-financed unregistered schools have also been growing, especially at the primary level. These avoid the high costs of registration, which requires posting a bond. Their students switch to recognized matriculation schools at grade 6, when attending a recognized school becomes more important. The government has also introduced measures to mobilize private contributions in publicly financed schools, effectively lifting a previous ban on such contributions. Among the new policies: 3 Parent-teacher and mother-teacher organizations have been set up in each school, and donations of up to 25 rupees per family are encouraged. For a school with 1,000 students, the yield would be 25,000 rupees a year, enough to pay for 2 or 3 additional teachers. - Schools can collect donations from private benefactors under the "School Improvement Conference." Donors who contribute 1,000 rupees or more are honored as school patrons, and their contributions are put into bank accounts whose interest can be spent by the school. In-kind donations of furniture and materials are also encouraged. These two measures will increase the discretionary funds and consequently the managerial control of some public schools. * Using these funds, retired or unemployed teachers can be hired to substitute for permanent teachers on leave or for absent teachers, for "honoraria" that are only a small fraction of the regular pay scales. Unemployed teachers are encouraged to participate by giving them credit 15  for this experience when they apply for a permanent job. This is a way of overcoming the relatively high permanent salary scales for government-paid teachers.