70388

THE EFFICIENCY OF PUBLIC EDUCATION IN UGANDA

MARCH 2008

This paper was written by Donald Winkler (Consultant) and Lars Sondergaard (AFTP2) with the very generous assistance of the Ministry of Education and Sports, Government of Uganda, and the helpful guidance of Harriet Nannyonjo (Task Team Leader). James Habyarimana was responsible for conducting a school survey to gather information on the sources and uses of funds at the school level. Maria Shkaratan did much of the analysis of EMIS data. The Planning Department, MoES, generously facilitated the provision of information and data for analysis. Several organizations provided very helpful comments on an earlier draft of the report, including MoES, UNICEF, GTZ, DCI, and Irish Aid.

1 TABLE OF CONTENTS

THE EFFICIENCY OF PUBLIC EDUCATION IN UGANDA

EXECUTIVE SUMMARY: KEY FINDINGS AND RECOMMENDATIONS

A. INTRODUCTION

B. AN EFFICIENCY FRAMEWORK

C. EXTERNAL EFFICIENCY

D. INTERNAL EFFICIENCY OF PRIMARY EDUCATION

E. EFFICIENCY OF PRIMARY TEACHER EDUCATION

F. INTERNAL EFFICIENCY OF SECONDARY LEVEL EDUCATION.

G. INTERNAL EFFICIENCY OF TERTIARY LEVEL EDUCATION

H. NEXT STEPS

I. SUMMARY AND CONCLUSION

J. REFERENCES

K. ANNEXES

1. Statistics 2. School Grants 3. Teacher Absenteeism 4. Formula Funding

2 ABBREVIATIONS

BOG Board of Governors CC Coordinating Centers CCT Coordinating Center Tutor DEO District Education Office DIS District Inspector of Schools EMIS Annual School Census of MoES ESA Education Standard Agency ESC Education Service Commission JSE Junior Secondary Education MoES Ministry of Education and Sports MoSP Ministry of Public Service NAPE National Assessment of Progress in Education NTC National Teachers College PETS Public Expenditure Tracking Survey PLE Primary Leaving Exam PSC Public Service Commission PTA Parent Teachers Association PTC Primary Training College SFG School Facilities Grant SMC School Management Committee SSE Senior Secondary Education TSC Teachers Service Commission UBOS Uganda Bureau of Statistics UNEB Uganda National Examinations Board UPE Universal Primary Education UPPET Universal Post Primary Education & Training

3 EXECUTIVE SUMMARY AND KEY ISSUES

This is a study of the efficiency of Uganda’s public education system. Since this type of study is relatively new for Ugandan education, the study begins by defining the basic concepts, terminology, and methods for analyzing efficiency.

The scope of the education sector—from pre-primary through university post- graduate; it’s magnitude—representing over 7 percent of GDP; the lack of basic financial and resource information for some sub-sectors; and the limited time and resources available for carrying out this study has resulted in some sub-sectors receiving more attention than others. In particular, given the lack of any recent cost or efficiency analysis of primary education, and the fact that this sub-sector absorbs almost two-thirds the government’s education expenditure, special emphasis was put on primary education. On the other hand, some sub-sectors—BTVET and teacher training, for example—have not received much attention, mainly due to the lack of data and the need to carry out original surveys to obtain the information that would have been required to analyze these relatively small sub-sectors. In addition, since recurrent expenditures represent more than 95 percent of the Government’s education budget, this study focuses on recurrent expenditures, although the future growth of enrollments and school infrastructure argues for a subsequent, separate analysis of the efficiency of development spending.

Uganda is very fortunate to have a large number of studies and evaluations that have been carried out in the education sector over the past decade. In addition, Uganda has rich data bases that provide much of the information required for the analysis of education efficiency—census data, household surveys, demographic and health surveys, service delivery surveys, and an Education Management Information System [EMIS] whose quality has been significantly improved in recent years. What Uganda lacks is the kind of finance, expenditure, and resource information required for analyzing efficiency. Hence, this study carried out a rapid unit cost survey of 180 public and private primary schools in six districts across three regions to provide this information1. While the survey is not nationally representative, it is as least representative of those six districts, which collectively reflect much of the nation.2

Issue: Cost and expenditure information is essential for monitoring the efficiency of Government spending. A nationally representative survey of expenditures, finance, resources, and outcomes at the primary and secondary school levels, including BTVET secondary level schools, would help provide the information needed for assessing efficiency. The MoES could contract a firm to carry out such a survey and provide cost and efficiency indicators.

1 A cost survey was recently carried out for tertiary education, and Shinyekwa (2006) recently did a cost survey of secondary schools, leaving the largest sub-sector—primary—as the one having the least cost and finance information. 2 While the sample was 180, survey data was collected on only 160 schools due to school holidays and some primary schools having been upgraded to secondary schools

4 FINDINGS AND RECOMMENDATIONS

This study has generated a large number of findings and recommendations, as well as a number of areas where additional information needs to be gathered in order to improve the analysis. The findings and recommendations are summarized below, followed by a discussion of the highest priority actions in the sector for improving efficiency.

This study documents the magnitude and extent of the leakage and misuse of educational resources. When possible, it identifies the principal causes of inefficiencies. However, in general, further research is needed in order to pinpoint causes and thus identify cost-effective solutions. For example, the study documents the problem of an inequitable and inefficient assignment of teachers across districts and schools. Determining the multiple reasons for poor deployment and developing programmatic and policy options for treating those reasons is beyond the scope of this work and requires a study of its own.

External Efficiency.

Uganda has done an admirable job of increasing access to primary education over the past decade. However, increased access has come at the expense of the quality of instruction. International evidence generally shows that improvements in quality—in terms of student knowledge--are more strongly related to economic growth than are improvements in access. Uganda needs to make a very serious effort to improve quality at all levels, while maintaining its impressive accomplishments with respect to coverage. Improvements in internal efficiency can help Uganda achieve both quality and quantity.

Issue: In terms of facilitating economic growth, improvements in quality—at all levels of education but especially in lower primary—are likely to have a high payoff. A careful assessment of the costs and benefits of raising quality at the lower primary level versus raising access at the post-primary level would help guide MoES resource allocation.

Internal Efficiency of Primary Education.

The internal efficiency of primary education is low. There are four principal sources of inefficiency. The first is the leakage of resources between the central government and the school through ghost teachers, misuse of UPE grants to district governments, etc. The second is the leakage of resources within the school, mainly attributable to high rates of student, teacher, and headmaster absenteeism. The third is the deployment of teachers both across and within districts, which appears to be unrelated to measures of need. The fourth is the allocation of resources within government schools, where class sizes are largest in the early grades and smallest in the later grades. While it is difficult to precisely quantify the overall magnitude of inefficiency, this study calculates that at least one-third of the expenditures on primary education are wasted

5 or used inefficiently. However, it’s important to note that several types of leakage— ghost teachers, UPE capitation grants, and teacher absenteeism--have all decreased over time.3

Teachers are the most valuable resource in improving educational outcomes. Uganda’s main efficiency problem is the poor utilization of its teaching staff. Three pieces of evidence to support this conclusion. First, over three-quarters of teachers are not in class teaching when unannounced school visits are conducted, and many of them are not even at work. Second, across districts, teachers are not deployed to the regions where there is greatest need for them. [see Figure D2 in the paper]. Third, within schools, teachers are not being assigned in such a way that class sizes across grades are the smallest possible: rather, the early grades [P1-P3] have large class sizes, and the the later grades [P4-P7] have much smaller class sizes [see Figure D9].

High levels of teacher, headmaster, and student absenteeism is the most important source of leakage at the school level. The magnitude of teacher absenteeism, in particular, is so large that reducing it should be a principal focus of Government efforts to improve efficiency in primary education. Since Government actions to reduce absenteeism are relatively recent, they may not have yet had much impact. [See annex 3 for analysis of the determinants of teacher absenteeism.] This report uses an internationally recognized methodology to measure absenteeism and presents an option of policy measures to reduce absenteeism.

Issue: In Uganda, a 20 percent reduction in teacher absenteeism alone would be the equivalent of hiring 5,000 more teachers (at a cost of Ush 12 bn). Policy measures to reduce teacher, headmaster, and student absenteeism could thus have a very high payoff. A careful assessment of the costs and benefits of specific policy measures would be useful to guide MoES policies to reduce absenteeism.

Government teachers are not deployed in sufficient numbers to the neediest districts, where their presence is likely to have the biggest impact on improving educational outcomes. Analyzing EMIS [Education Management Information System] data, the deployment of government-paid teachers across districts is perverse, with student teacher ratios in government schools being the highest in the poorest districts. [See Figure D3] . In addition, there is no relation between the current teacher deployment and measures of educational outcomes4.

Issue: Since the MoPS already has an explicit rule for the deployment of teachers across districts, it would be useful to know why the actual deployment differs. The district teacher service commissions should establish and implement a transparent, explicit rule for the deployment of teachers across schools within districts and establish procedures to regularly monitor deployment. Since much of the poor deployment of teachers appears to be linked to teacher transfers, 3 See the World Bank (2007b) Public Expenditure Review for Uganda for greater detail on changes in leakages over time. 4 There are questions as to the accuracy of EMIS data on teacher employment, so this analysis should be repeated using the Ministry of Public Service [MoPS] data base

6 consideration might be given to grant schools rights over teacher reassignments to prevent the movement of teachers from unpopular to popular schools without a replacement satisfactory to the school.

Issue: A more radical proposal would eliminate district deployment of teachers altogether. Each school would be given an annual formula-driven budget determined by student enrollments and special needs and would recruit as many teachers as allowed by that budget.5

Within schools, students at the lower primary level receive too few resources, which contributes to P3 children having low achievement levels and being poorly prepared for English only instruction beginning at P4. Government schools allocate fewer teachers (measured on a per student basis) to the early grades of primary school relative to the later grades. The large class size in the early grades constrains teachers in individualizing instruction. The low percent of total hours that teachers are actually present teaching in the classroom also contributes to low student achievement. Another factor contributing to low achievement, high repetition and high dropout is the low percentage of students entering primary school at the appropriate age.

Issue: MoES could create and enforce a norm that requires that class sizes be no larger in the early grades [P1 – P3] than in the later grades [P4 – P7]. In schools where there are a sufficient number of classrooms, this would require creating additional streams at the lower grades. In schools where there are an insufficient number of classrooms to create additional streams, this would require constructing new classrooms and/or introducing double shifts.

Issue: MoES might consider the following policy change: Put at least as high a priority and as much emphasis on learning achievement levels at P3 as on the percent of children passing the P7 school leaver’s examination, and create incentives [e.g., school merit pay, public recognition] that reward schools that show annual gains and better than expected performance on the P3 test.

Issue: School communities could play an important role in reducing teacher absenteeism from the classroom if there were transparent rules establishing clear expectations about teacher presence in the classroom. Absenteeism might be reduced if the MoES were to publicize norms around the number of hours that teachers should be in the classroom teaching each day, and engage communities to monitor compliance with the norms.

Issue: Strong incentives could reduce student absenteeism. MoES might consider establishing public information campaigns and incentives [e.g. school lunch] to encourage parents to enroll students at the proper age, and create incentives to schools and districts to adopt pro-active policies to enroll students in 5 This is the method Chile uses to fund its privately managed, publicly funded schools; Georgia and Armenia follow similar procedures to fund public schools. In most countries [but not Chile] a common salary scale and benefit structure applies to all teachers irrespective of where they are employed, and the central government maintains the pension scheme.

7 P1 at the appropriate age. Capitation based funding is one means of providing an incentive to schools to actively try to enroll students.

The lack of an effective inspection system at the district level combined with the limited powers of school management committees [SMCs] to hire and fire school personnel contribute to an almost complete lack of accountability by districts and schools to parents, the public, and the ministry for compliance with MoES norms and guidelines and for adequate educational performance.

Issue: There are several options MoES might consider adopting to strengthen local accountability. It might Increase the capacity of SMCs to develop school budgets, including the UPE grant, and monitor expenditures and involve the SMC in the annual performance evaluation of headmasters. It might also consier developing district level report cards which give district residents the information required to assess the performance of DEOs, as well as school level report cards that include the school budget, school outcomes, student and teacher absenteeism, and the level of resources that parents have a right to expect in their schools. Creating school-wide incentives for unexpectedly good performance that bring the community and teaching faculty together in pursuit of a common goal is also a good idea.

Efficiency of Primary Teacher Education.

Pre-service teacher training occurs in too many poorly resourced Primary Teacher Colleges [PTCs], while in-service training takes place in Coordinating Centers [CCs] dispersed throughout the country and affiliated with 23 core PTCs. The experience of other countries shows that using in-service training to produce qualified teachers is more cost-effective than traditional, pre-service training, and Uganda’s experience appears to be consistent with that finding. However, more careful analysis needs to be carried out to determine the cost-effectiveness of producing qualified teachers via the two modes of training. In addition, more analysis is required to determine what capacity is required for producing qualified teachers via traditional pre-service training, how many PTCs should be upgraded to provide higher quality and more cost-effective pre-service training, and what is the minimum size required for a PTC to be both effective and efficient.

Issue: MoES might consider carrying out a survey of PTCs and CCs to estimate the unit costs and the classroom effectiveness of qualified teachers produced by these two different modalities. It could determine the future demand for teachers, taking account of teacher attrition, demographic change, and policies concerning the pupil-teacher ratio and other factors that affect demand. On the basis of the demand and cost studies, MoES could determine how many PTCs should remain operating, and identify means of adequately resourcing both PTCs and CCs.

Internal Efficiency of Secondary Education.

8 The internal efficiency of public secondary education is low and unit costs are high. The reasons for low efficiency include low workloads, poor teacher deployment, and high teacher salaries. A significant portion of secondary school teachers are under- utilized. The reasons include an overly prescriptive curriculum, constraints on classroom space, and small schools in terms of student enrollment. In addition, the salaries of public secondary school teachers, and especially of public secondary school headmasters, are high relative to per capita GDP, high relative to primary school teachers, and high relative to salaries paid teachers in private secondary schools [see Table F2 and Figure F3]. If the Government carries through with its plans to significantly increase secondary school net enrollment rates, these high salaries may very well be unsustainable.

Issue: Simplifying the secondary school curriculum by requiring fewer courses, and mandating that all teachers should be required to have the skills to teach at least two subject matters could reduce the unit costs of secondary schooling..

Isuse: The MoPS might consider contracting independent analysts to study the teacher labor market in Uganda to determine appropriate salary levels of secondary school teachers and headmasters and to propose options for reducing average salaries [e.g., protecting currently employed teachers and headmasters while employing new teachers using a new pay scale].

The distribution of Government secondary education expenditures, as proxied by teacher deployment across and within districts, appears to bear no discernible relationship with the measures of need and cost most commonly used in education funding formulas. Furthermore, this distribution shows no relationship between funding and proxies for educational outcomes. Prima facie, these findings provide evidence of an inefficient allocation of Government monies.

MoES: The Government might consider establishing a clear, transparent teacher deployment formula for secondary education. As with the UPE grant, the funding and deployment formula would be public information so each school would know its staffing entitlement, and schools and communities would have to give approval for teacher transfers which would reduce staffing.

9 FIGURE 1. AVERAGE PERCENT OF PRIVATE ENROLLMENT IN SELECTED SSA COUNTRIES

Eritrea

Congo

Botsw ana

Benin PE Senegal JSE Togo SSE Niger

Madagascar

Comoros

Uganda

0 10 20 30 40 50 60 % private enrollment

Source: UIS, World Bank; latest years available.

The role of the private sector in the finance and provision of secondary education in Uganda is critically important, including the sizeable fees paid by households to public secondary schools. As shown in Figure 1, Uganda is among the countries in Africa with the highest percentage of secondary school enrollments in private schools. In total, household expenditures on secondary education are triple those of Government. It is critical to protect and sustain household financing levels, most of which is provided by high income households, to permit the expansion of more heavily subsidized educational opportunities to lower income households. The present allocation of Government subsidies is not transparent and does not offer explicit incentives to private schools and households to sustain and increase private provision and finance.

Issue: The MoES might consider designing and implementing a transparent capitation grant to private institutions with clear, explicit incentives for sustaining household finance of secondary education [e.g., cost-sharing up to some desired level of expenditure]. Such a grant could include adjustments for differences in school location or type of school which may affect the school’s cost structure and, also, adjustments for the income levels of students [perhaps proxied by region or rural/urban location], to ensure equity. [In general, see La Rocque (2006) for further suggestions on strengthening public-private partnerships.]

Accountability by schools to either parents or the MoES is weak. School inspection is infrequent enough to be ineffective, thereby seriously weakening accountability to the MoES. The local BOGs and PTAs have unclear and sometimes competing roles and usually lack the capacity and information to effectively manage school budgets.

10 Issue: The BOGs and PTAs might be more effective if they were to have clearly delineated and strengthened roles, be given training to improve their governance capacity, and be provided the information they need to hold schools accountable, possibly in the form of school report cards that allow schools to assess their relative performance.

Internal Efficiency of Tertiary Education.

Higher [tertiary] education in Uganda is monitored and managed by three different organizations with the Ministry—the Teacher Education Department, BTVET, and Higher Education plus the newly created National Center for Higher Education [NCHE]. This division of management and oversight responsibilities is an obstacle to rational and efficient resource allocation within the sector.

Public higher education—especially the universities—is funded from both public and private sources. Among universities, the government funds only the most meritorious applicants, who also happen to mainly come from the highest income households in Uganda. At other tertiary institutions, Government funds specific inputs rather than students and allows those institutions very little management autonomy. Many public tertiary institutions could not survive without the revenue from student tuition fees. Both public and private institutions have difficulty accessing the credit markets to fund expansion, and the effects of UPE and UPPET will soon be felt at the tertiary level in the form of significantly increased demand.

Issue: Government might consider reforming government funding of universities by eliminating full-funding of government sponsored university students and replacing it with capitation grants and student loans targeted on financially needy students. Government funding of other tertiary institutions could be reformed by replacing input-based funding by capitation grants. Government might also support lines of credit to both public and private institutions to ensure they have adequate funding to support expansion.

Information for Decision-making.

Public education is one of Uganda’s largest industries. A private sector business with the number of employees and the budget of public education would typically have a sophisticated information system that would provide accurate and up to date information on organizational performance, expenditures, and efficiency on at least a weekly basis. In contrast, the public education system provides only some of this information and typically on an annual basis. Furthermore, the information provided to decision-makers is not always accurate. A good example is information on the single most important educational input—teachers. As shown in Figure 2, there are large discrepancies in the total number of payroll primary teachers as reported in the EMIS, as reported in the Education Sector’s Annual Performance Review (June 2006), and the records of the Ministry of Public Service [MoPS]. The EMIS data underestimate the number of total payroll teachers by about 10 percent per year relative to the MoPS data, which is thought

11 to be the more reliable . For an educational system to operate efficiently, it is absolutely essential that the managers of that system have reliable and up to date information on its key inputs.

12 FIGURE 2. HOW MANY TEACHERS ARE ON THE GOVERNMENT PAYROLL?

130,000 ) l l o r

y 125,000 a p

n o

s 120,000 r e h c a e

t 115,000

f o

r e b 110,000 m u n ( 105,000 02/03 03/04 04/05 05/06

Number of teachers on payroll (ESAPR 2005) Number of teachers on payroll (MPS, end-June figures) Number of teachers on payroll (EMIS 2005)

Issue: MoES could contract an external consultant to carry out an independent assessment of the quality and coverage of data provided by the MoES’s EMIS. The assessment could include revising the survey instrument and survey procedures to improve accuracy and to improve coverage, especially of non- public secondary schools and of school budgets and expenditures. To the extent possible include measures of school achievement like PLE and SLE results.

NEXT STEPS

The next steps in the analysis of education efficiency in Uganda are to fill in the gaps where cost and financing information is lacking, e.g., BTVET and Primary Teachers Colleges and school infrastructure, and to systematically develop and evaluate some of the recommendations made in this report in terms of their likely cost, impact, and administrative and political feasibility.

HIGH PRIORITY ISSUES

This study has identified a number of issues or problems of education efficiency and possible actions or policies which might be undertaken to correct those problems. Several issues appear to be of especially high priority, either due to the magnitude of the

13 underlying problem or because the proposed actions are so fundamental. These issues are summarized in the table below:

TABLE: HIGH PRIORITY ISSUES

SUB-SECTOR ISSUE/RECOMMENDATION Primary Reduce headmaster absenteeism. Increase teacher classroom time. Improve accountability arrangements. Rationalize teacher deployment. Prioritize grades 1-3. Secondary Reduce teaching costs. Ensure high quality. Facilitate privately-funded expansion. Tertiary Reform government finance. Stimulate privately-financed supply. MoES Strengthen information and analysis. Carry out specific efficiency studies.

14 A. INTRODUCTION

In 2004 the Ministry of Education and Sports [MoES] adopted the Education Sector Strategic Plan for 2004-2015. The previous Education Strategic Investment Plan [ESIP] for 1998-2003 had put top priority on getting children in school through implementation of the Universal Primary Education [UPE] program. Since this objective had been largely accomplished by 2004, this new plan set out new priorities focused on raising the quality and relevance of education and improving the efficiency and effectiveness of the education sector. The purpose of this paper is to contribute to the accomplishment of this latter priority—improving the efficiency and effectiveness of the education sector.

Through both Government and household expenditures, Uganda already allocates over 7 percent of its GDP to the education sector6. This exceeds the average 6 percent of GDP that OECD countries spend on education. Meanwhile, the coverage of post-primary education remains relatively low, and the quality of all levels of education needs improvement. To significantly increase post-primary enrollments at current unit cost levels would require an unsustainably large increase in education expenditures. While Government and household education expenditures may increase in future years, the Government’s goals to increase the coverage of post-primary and to improve the quality of all levels of education cannot be attained in the absence of improvements in efficiency in the use resources.

To fully analyze the efficiency of an education system is a very large task well beyond the limited resources and time available for this study. It was simply not possible to carry out an in-depth analysis of each sub-sector, and decisions had to be made as to how best to focus the study. Three factors guided these decisions: [a] the existence of recent analytic work on costs and efficiency; [b] the size or budget share of the sub- sector, and [c] the availability of the basic information required to carry out an efficiency analysis. First, in some cases—e.g., tertiary education—recent in-depth analytic work had already been undertaken, and the findings having to do with efficiency only had to be summarized and updated with recent information. Second, in other cases—e.g., primary education and secondary education—the mere size of the sub-sectors, which in the case of primary represents two-thirds of all government education expenditures, demanded careful attention. Third, in other areas—e.g., teacher training and secondary level BTVET—the information base is exceptionally weak, while in other areas—e.g., primary and general secondary—the information base is relatively rich as a result of the annual information census carried out by the EMIS. As a result of these considerations, this study put greatest emphasis on the primary and secondary sub-sectors and put least emphasis on BTVET and teacher training. There is an urgent need for much more intensive work on these latter sub-sectors, but that will require original data collection and the involvement of specialists from those sub-sectors.

6 The 7 percent figure includes both government and household spending. See Annex 1 for detailed statistics on government spending disaggregated by level of education.

15 Several years have passed since the last efficiency study of Ugandan education. For this reason, we start off this study by defining the basic concepts, terminology, and methods of analyzing education efficiency . The reader who is already familiar with the economic analysis of efficiency may wish to skip this section and proceed directly to the analysis sections.

Uganda is very fortunate to have a large number of studies and research that have been carried out in the education sector over the past decade. In addition, Uganda has rich data bases that provide much of the information required for the analysis of education efficiency—census data, household surveys, demographic and health surveys, service delivery surveys, and an Education Management Information System [EMIS] whose quality has been significantly improved in recent years. What Uganda lacks is an institutional survey that provides the kind of finance, expenditure, and resource information required for analyzing efficiency. Hence, this study carried out a rapid unit cost survey of 180 public and private primary schools in six districts across three regions to provide this information7. At the secondary level, a separate cost survey was carried out as part of the concurrent analysis of the cost implications of UPPET.8 And at the tertiary level, a similar cost survey had been recently carried out by the National Council for Higher Education [NCHE] with support from the Rockefeller Foundation.9 As noted above, similar surveys need to be carried out for secondary level BTVET schools and for Primary Teacher Colleges.

7 While the sample was 180, survey data was collected on only 160 schools due to school holidays and some primary schools having been upgraded to secondary schools. While the survey is not nationally representative, it is at least representative of these six districts, which collectively reflect much of the nation. 8 Shinyekwa (2006). 9 NCHE (2005).

16 B. AN EFFICIENCY FRAMEWORK

What is the meaning of efficiency in education?

Efficiency is measured by comparing education expenditures with education outcomes. Governments make expenditures at all levels of education, and two of the most basic efficiency questions are whether government is spending the appropriate amount on each level or type of education, and whether government is making the appropriate choices on quantity versus the quality of education. These questions are answered by looking at the success that the graduates of different levels of education have in the labor market relative to the costs of their education. This is called the external efficiency of the educational system.

It is important to have indicators of external efficiency not only to guide the allocation of government spending across levels of education, say pre-primary vs. upper secondary, but, also, across types of education. Thus, knowing the returns to general secondary vs. vocational secondary education would help guide a policy decision about whether or not to upgrade and expand vocational instruction.

A different type of efficiency question concerns the use of resources in producing the outcomes of education, which is called the internal efficiency of the educational system. Assessments of internal efficiency are typically done for a specific level of education, say primary education, and the simplest indicator of internal efficiency is the unit cost of producing one unit of educational output, which may be a student enrolled, a graduate of that level of education, or a student who has attained some minimum level of knowledge. Other things equal, an educational system which can produce a unit of output at lower unit cost than another educational system is said to be more efficient.

The economist’s emphasis on unit cost is frequently criticized by educators who mistakenly believe the economist is saying that improving efficiency is the same thing as reducing costs in general. This is not true. Indeed, increasing the unit cost of enrollment may actually reduce the unit cost of a graduate. The confusion between economists and educators results from a failure to carefully specify the unit of output.

What are some common indicators of efficiency in education?

External Efficiency. The most common indicators of external efficiency in education are estimates of the private and social rates of return to expenditures on education at the different levels or types [e.g., academic vs. vocational secondary] of education. Unfortunately, at this point there is no good country level indicator of the appropriate levels of access and quality of education.

The private rate of return shows the financial returns in terms of increase incomes accruing to individuals as a result of investing their own time and money in a given level of education. An individual who can attain a higher return to “investing” in,

17 say, secondary education than she could by making some alternative investment should make the decision to attend secondary school. To do otherwise would be an inefficient use of that individual’s own resources. Government policies—e.g., decisions to charge low tuition at the tertiary level—affect the private rate of return and, also, individual education decisions. Thus, the elimination of user fees in the PTCs should have increased the number of applicants for teacher training. Supply-side restrictions on enrollments may make it impossible for individuals to further their education even if their private rates of return are high. In addition, estimates of the private rate of return can influence cost recovery policies. Thus, if Kyambogo University finds the private rate of return to a bachelors degree in accounting and finance is very high, it will know it has the option of charging users a high percentage of their instructional costs and still be able to attract the number of students it wishes to enroll. On the other hand, if it finds the private rate of return to, say, a bachelors degree in arts is very low, it will know that it may not have the option of any cost recovery if it wishes to attract students and retain the program.

The social rate of return shows the financial returns to society resulting from investing society’s resources [i.e., government plus individual costs] by enrolling students in a given level or type of education. If calculated correctly, the social rate of return should guide government decisions about the supply and finance of education. Thus, a high social rate of return to tertiary education may influence the government to expand enrollment capacity this level.

There are two serious problems with estimates of rates of return that argue for using them with caution to guide public policy. First, the estimates are calculated based on what individuals with different education levels have earned in the past, which may not necessarily be a good predictor of future earnings. Thus, while the social rate of return to secondary education in Uganda is relatively low, evidence from surveys of national competitiveness suggest that future returns may be considerably higher. Also, these estimates ignore the non-monetary, social benefits of additional education, such as better health and nutrition, better child-rearing, reduced poverty, and scientific advancement.

Finally, we lack information on the rate of return in which many countries are most interested these days—the rate of return to investing in quality improvements in the classroom. Raising quality, and thus student learning or achievement, in most cases requires additional resources. In principal, one could calculate the increased income that results from the additional learning resulting from additional investments. In practice these are extremely difficult calculations to make, but the little evidence we have suggests the returns may be very high indeed10. Many parents appear to know this and make sizeable private investments to raise their child’s learning and future educational and career prospects. .

Internal Efficiency. An analysis of internal efficiency of public education spending attempts to answer the questions: [1] Can educational outcomes be increased

10 Hanushek [2007].

18 without raising the current level of resources or funding? and [2] Can expenditures be reduced without adversely affecting the current level of educational outcomes?

There are several ways to proceed in answering these questions. This study uses three types of analysis to assess internal efficiency of public education in Uganda. The first is to identify ways in which resources may leak out of the system. Leakages like ghost teachers, misuse or diversion of education funds, and teacher absenteeism all reduce the size of the total resources available for the provision of education. Reducing leakages increases the available resources for delivering education more effectively and, thus, increases internal efficiency.

A second type of analysis concerns the allocation of education budgets across inputs. The question here is whether or not an education system is spending its monies wisely. For example, systems which fail to allocate sufficient funds for building and equipment maintenance, or systems which fail to provide textbooks to students may be inefficient because they could be producing a higher level of output without increasing overall spending. Or, it may be the case that output levels could be increased with a given budget if the delivery process were altered, e.g., using multi-grade instead of single grade instruction, or using distance learning instead of traditional classroom construction for teacher in-service training.

A third type of analysis concerns how public funds leverage private sector finance of education. For example, when government wishes to expand enrollments by 10,000 at the secondary level it can either directly provide public schools, or it can offer financial incentives to private schools to increase their enrollments by 10,000. If the private financial incentives cost less than direct public provision, the government is leveraging its funding. In some cases where private schools have excess capacity, the government may have to offer only small financial incentives to achieve its desired result. In other cases, such as offering financial incentives in remote rural areas where no private schools already exist, it may not be possible for government to leverage its funding.

Public funds can also leverage private finance through selective targeting of subsidies. Imagine a university with a fixed budget provided by the Government. That university has three options in charging tuition fees for students. First, it could charge no tuition fee whatsoever. Second, it could charge a uniform tuition fee to all students. Third, it could charge a tuition fee that varies with the student’s ability to pay. If the goal is to maximize university enrollment with the Government’s fixed budget, the third option would be the best one.11

Inputs, Outputs, and Outcomes. Since efficiency is defined as the relationship between inputs [expenditures] and outputs [a student enrolled, a student graduating, etc], it is necessary to carefully measure both the inputs or costs and the outputs of the education system. Figure B1 gives a visual representation of how these variables are used in an efficiency study. Funding from different sources purchases tangible inputs

11 Universities frequently charge a uniform tuition “price” and then provide scholarships (i.e., subsidies) based on student financial need.

19 FIGURE B1: INPUTS, OUTPUTS, AND OUTCOMES OF EDUCATION

MoES leadership, guidance,

s

t regulations and u

p non-financial n

i Government Donor funds Household support l

a funds i and others contributions Other factors c

n influencing a n i

F attendance (e.g. socio-economic s

t status, health,

u etc) p n s I t u

p Textbooks n i

e

l and other

b Teachers Classrooms i

g inputs n

a purchased T e t a s i t Number of

d Number of enrolled u

e enrolled children

p children in t in private schools

m government schools u r o e

t n I

Number of Number of graduates from graduates from government s private schools t schools u p t u O

Number of graduates Number of graduates from government from private schools schools with the with the desired level desired level of of proficiency proficiency s e Educated citizenry m participating in civil o

c society. Educated t workers with the right u set of skills to increase O competitiveness of the economy

20 which are then used to enroll students [an intermediate output] in order to produce an educated child [output] who then contributes to society [outcome], including economic competitiveness.

Perhaps the most difficult task in studying internal efficiency in primary education is specifying the desired output or outcome12. Table B1 lists several, alternative output measures in order of complexity. Some of these measures must be used with great care. Almost no one would advocate reducing unit costs of enrolling students if that would also reduce quality. On the other hand, if we observe two regions of the country, and one has lower unit costs and higher achievement than the other, it may very well be more efficient. Also, if low unit costs translate into lower quality, higher repetition, and higher dropout rates, the result may be higher unit costs of attaining literacy or graduation. Thus what appears to be efficient, or low cost, in terms of enrollment is actually inefficient, or high cost, in terms of output. Perhaps the most useful efficiency indicator is the unit cost of graduates having demonstrated some minimum competency in core subjects. Indeed, government may be able to reduce the unit costs of this measure by increasing the unit costs of enrollment in order to raise quality. The differences in these indicators can be large. For example, in Ugandan primary education for 2005/06, the unit cost per student enrolled is Ush 50,534, the unit cost per primary school graduate is Ush 923,833, and the unit cost per primary school graduate achieving some minimum level of knowledge is Ush 4,506,500.13

TABLE B1: MEASURES OF EDUCATION EFFICIENCY

Output/Outcome Measure/Indicator Enrollment Cost per student enrolled Increased Enrollment Cost per additional student enrolled Literacy Cost per student completing grade 4 Graduate Cost per primary school graduate Achievement Cost per student achieving at least some minimum level of knowledge in a specified grade Learning, or increased Cost per additional knowledge or achievement achievement gained Graduate with Cost per primary school graduate

12 In general, output refers to the concrete results and products which contribute to educational outcomes, such as, an enrolled child, a child graduating from school, a child graduating with some defined competency. An output is different from an outcome, the societal or ultimate goal of your education policy): for instance: build a competitive workforce; ensure that you have the foundation of a democratic society; empower citizens, etc. In some instances, the distinction between output and outcome is unclear. For instance, “graduating a student with a specified level of proficiency” can be thought of as both an output and an outcome. 13 Calculated from fiscal data and the annual MoES Statistical Education Abstract. We estimate the unit cost of enrollment by dividing recurrent government expenditure by total enrollment in government schools. We estimate the cost of graduating a student by dividing recurrent government by the number of graduates (in that particular year) from government schools. Finally, we estimate unit cost of graduating a student with a desired level of proficiency by dividing recurrent government spending by our estimate of the number of graduates that obtain the desired level of proficiency.

21 Minimum Achievement demonstrating some minimum achievement level

Table B1 suggests another unit cost distinction that is extremely important for policy decisions, and that is the difference between average and marginal costs. Perhaps this distinction is best seen by looking at UPE. Uganda has had exceptional success in enrolling students in primary school, attaining a net enrollment rate [NER] of about 92 percent by 2006. If we add up all the costs of providing primary education and divide by the number of students, we obtain a unit cost of primary school enrollment of UShs. 50,534. If we were to devise a program to enroll the remaining 8 percent of students in primary school, the unit cost for those additional students is likely to be considerably higher, say 70 thousand, because the students not in school are in especially remote areas or have learning difficulties which require specialized teachers. In addition, if parents don’t sufficiently value education to send their kids to school, it may be necessary to provide incentives—e.g., free school meals or small scholarships—to parents. In this hypothetical example, the unit cost of increased enrollment (i.e., marginal cost of enrollment) is 70 thousand.

Table B2 below calculates the three most important unit costs in primary education: the cost of enrolling a student, the cost of graduating a student, and the cost of graduating a student with a desired level of proficiency in literacy and numeracy.14 In a perfectly efficient system, there would be no drop-outs and no students repeating any classes. Moreover, every student passing through the system would be able to show proficiency in literacy and numeracy. Thus, in a school system with seven grades to complete, it would cost exactly 7 times the unit cost of enrolling a student to graduate a student with the desired level of proficiency

In Uganda’s case, the unit cost of a graduate is more than twice what is should be in a perfectly efficient system. In particular, it costs approximately $27 to enroll a child. Therefore, the aspirational target is that the cost of graduating a child with proficiency in literacy and numeracy should be $189 (7 x $27). In reality (as shown below), the cost is $2,424. Similarly, the cost of graduating a student from primary seven (ignoring the quality aspect) is $497

14 We estimate the unit cost of enrollment by dividing recurrent government expenditure by total enrollment in government schools. We estimate the cost of graduating a student by dividing recurrent government by the number of graduates (in that particular year) from government schools. Finally, we estimate unit cost of graduating a student with a desired level of proficiency by dividing recurrent government spending by our estimate of the number of graduates that obtain the desired level of proficiency.

22 TABLE B1: MAIN INDICATORS TO MEASURE INTERNAL EFFICIENCY OF GOVERNMENT RECURRENT SPENDING ON PRIMARY Estimates Based on Government Recurrent Expenditure (in 2005/06 constant prices) Intermediate Measure/Indicator 2000/01 2005/06 Outputs, Outputs (Ush) (Ush) and Outcomes Enrollment Cost per primary student enrolled 40,470 50,534 ($23) ($27) Graduate Cost per primary school graduate 817,944 923,833 ($464) ($497) Graduate with Cost per primary school graduate 2,370,853 4,506,500 Minimum demonstrating some minimum ($1,345) ($2,424) Achievement achievement level Source: Authors’ estimates based on Statistical Education Abstracts (various issues) and fiscal data

23 C. EXTERNAL EFFICIENCY

As noted above, indicators of external efficiency are important as guides for Government investment in education. Governments need to make critical decisions about how much to invest in each level of education, each type of education [e.g., different degree programs or different modalities], and in quality and access.

How much is Uganda investing in education?

As a society, Uganda invests over seven percent of GDP in formal primary, secondary, and tertiary education, excluding the income foregone by students. As shown below, in aggregate this investment is funded almost equally by government and private households. However, at the primary level government bears the larger financing burden, and at the secondary level households bear the larger financing burden. This in part reflects the fact that many students, in both public and primary secondary schools, pay relatively high user fees.

TABLE C1. UGANDA EDUCATION EXPENDITURES AS SHARE OF GDP

Level/Source Government/Donor Household Total Primary 2.27 1.32 3.59 Secondary 0.63 1.88 2.51 Tertiary 0.55 0.48 1.03 Total 3.45 3.68 7.13 . Source: Calculated on basis of UNHS 2006, Liang (2004) and UBS 2000 DHS

Is public education “free” in Uganda?

In 1996 Uganda adopted a policy of Universal Free Primary Education by abolishing school fees. As shown in Table C2, school fees are far lower at the primary level than other levels of education, but at 9,006 UShs, the average school fee paid in government schools according to the latest household survey, they are far above zero15. However, for most students school fees are very low, with the median fee collected in a rural public school being zero. In 2006, on average, students in public primary schools paid 4,892 UShs in school fees, and students in the bottom income quintile [in 2002] paid only UShs. Of course, school fees are only part of the financing burden facing families. On average, households with students in primary school pay as much for uniforms, transportation and school supplies as they do for school fees.

15 See Annex 1 for more details on the use of household survey data to calculate household contributions to education spending. As shown in the annex, there are significant differences in fees paid in government schools, especially by urban/rural location.

24 TABLE C2. AVERAGE PER PUPIL HOUSEHOLD EXPENDITURES ON EDUCATION BY LEVEL OF EDUCATION, 2006 [UShs]

School Fees Other School Total Expenditures

Primary, All Schools 14,254 14,240 28,494 Government 9,006 15,930 24,936 Non-Government 259,336 167,453 426,789

Secondary, All 161,432 76,512 237,944 Schools 1/ Government 270,123 158,449 428,572 Non-Government 259,336 167,453 426,789

1/ Includes boarding schools Source: Calculated from UNHS 2006.

Is Uganda allocating its publicly financed education expenditures appropriately across levels of education?

Government should make its decisions about allocating spending across levels of education by looking at estimated social rates of return, which compare the incremental income gains of that level of education to the social costs [government plus household costs] of providing that education. The social rates of return estimated using the 2000 household survey data are given in Table C3 and show high returns to all levels of education, especially primary. As noted earlier, these estimated rates of return reflect past, as opposed to future, labor market conditions. However, the results do support arguments for continuing and perhaps increasing expenditures at the primary level.

TABLE C3. RETURNS TO EDUCATION IN UGANDA, 2000

Level Private Return Social Return Primary 30.2 23.7 Secondary 11.5 10.5 Tertiary 24.2 13.4 Source: Appleton (2001) and Liang (2004) calculated from 2000 UNHS.

Absent from Table C3 is evidence on the returns to pre-primary education. International evidence on early childhood development programs suggests these returns can be very high, especially for children from disadvantaged backgrounds. Investing in the health and school preparedness of pre-school children can reduce delayed entry to primary school and increase the likelihood of success in school.

25

Is Uganda allocating its publicly financed education expenditures appropriately between quantity and quality?

Uganda has achieved a high net enrollment rate at the primary level16. However, there is evidence that the quality of primary education is low although it has improved since its low point after the introduction of UPE. As shown in the following figure, Uganda is in the middle of the African countries participating in SACMEQ with respect to both academic performance and cost-effectiveness. However, its performance is almost identical to that of South Africa, which scored at the bottom of all countries internationally participating in the most recent TIMSS. This suggests that the quality of Uganda’s education, too, lies far below that of international comparators outside of Africa.

FIGURE C2. PRIMARY SCHOOL SPENDING AS A PERCENTAGE OF PER CAPITA GDP AND MATHEMATICS TEST SCORES ON SACMEQ

Primary school spending as percentage of per capita GDP and math test scores

30 g P n

i D Kenya 25 d G

n Namibia

a Lesotho e t i p 20 p s

a

l Seychelles c o Malaw i S.Africa Mauritius

r 15 o

e Sw aziland h p c Uganda f s 10

o y Zambia Botsw ana r a %

5 s m i a r

p 0 400 450 500 550 600 scores in m ath

Table C4 below provides further evidence that the quality of education in Uganda is low even by the country’s own standards. Less than half of P3 and P6 students attain even minimum competencies in reading and math.

16 Estimates of the primary level NER vary depending on the data source. Using EMIS data, the MoES calculates an NER in 2006 of 91.7 as reported in the 2006 ESSAPR, but using household survey data, the NER is estimated at 84.

26 TABLE C4. PERCENTAGES OF PUPILS ATTAINING MINIMUM COMPENTENCIES IN ENGLISH LITERACY AND NUMERACY

Subject/Grade 1999 2003 Literacy P3 18.2 34.3 Literacy P6 13.2 20.0 Numeracy P3 38.6 42.9 Numeracy P6 41.5 20.5

Source: NAPE

Figure C1 and Table C4, along with other data, provide evidence that the quality of primary education in Uganda is low. In addition, country cross-sectional evidence demonstrates that [a] there is a large positive relationship between quality improvements and economic growth and [b] there is almost no relationship between increases in access to education and economic growth.

In Figure C3 below, the economic growth rates of countries are plotted against their scores on international assessments, controlling for other factors which may affect economic growth. As can be seen, the slope of the line fitted to these observations is positive and steep, indicating that, controlling for access, higher test scores contribute significantly to economic growth.

In Figure C4 below, the economic growth rates of countries are plotted against their average years of education of the populace, again controlling for other factors that affect growth. The slope of the resulting line fitted to the observations is almost flat, and the estimated slope of the line is statistically insignificant, indicating that, controlling for test scores, higher years of educational attainment do not contribute significantly to economic growth.17

17 Hanushek and Woessman (2007).

27 FIGURE C3. ECONOMIC GROWTH AND ACHIEVEMENT TEST SCORES.

FIGURE C4. ECONOMIC GROWTH AND YEARS OF EDUCATION.

28 What determines the quality of education in Uganda?

There have been several studies of the causes of low student achievement in Uganda.18 Table C5 presents the results of one such study carried out by the Education Standards Agency [ESA]. The challenge is not knowing what to do but, rather, how to do it in a context of poorly trained teachers and budget contraints.

TABLE C5. SCHOOL FACTORS AFFECTING TEACHING AND LEARNING: QUALITATIVE ASSESSMENT, 2004

Variable Assessment Teacher Qualifications Lack of qualified teachers, especially in rural schools Teaching Methods Inadequate lesson preparation Class Size Overly large classes constrain teachers’ time for class supervision and marking Learning Materials Lack of basic materials, especially materials written in indigenous languages Teacher Absenteeism High absenteeism attributed to low commitment, poor school management, lack of accommodation, and low salaries Language of Instruction Use of mother tongue constrains supply of teachers Source: Education Standards Agency, Report on Monitoring Learning Achievement in Lower Primary, 2004

There is, in addition, a large international literature on the characteristics of effective schools, which helps provide a checklist that can be used to self-assess how to improve instructional quality. A study of Ugandan schools produced the characteristics listed in Table C6. This list highlights several variables that are analyzed in some detail in this study: teachers and headmasters are supervised regularly; textbooks are available and used; students [and teachers] attend school regularly; there is an emphasis on preparing children to read in the early grades; communities and parents are actively involved in the school and in their childrens’ education

18 See, for example, Nannyonjo (2006) and Hicks (2005).

29 TABLE C6. CHARACTERISTICS OF EFFECTIVE SCHOOLS AND EVIDENCE FROM RWENZORI DISTRICT

Characteristics of Effective Schools Evidence from Rwenzori District Head Teacher monitors and supervises Head Teacher monitoring and supervision teachers’ lesson plans and teaching notably better in high-performing schools Teachers prepare for teaching through Little variety in teaching methods across lesson plans & varied teaching methods schools, with most teachers prepared but seldom enriching lessons or encouraging student-centred work. Very little difference between trained and untrained teachers in preparation, variety of teaching methods, and use of books. Pupils attend regularly and participate in Homework is seldom required and class work and homework feedback to students is seldom given. Schools with more regular student attendance perform better. Teachers use instructional materials, All schools have some textbooks but there especially textbooks is little evidence of their being used and students are not permitted to carry books home. Half the schools have learning aids, but they are not often on display. Teachers frequently assess student work Schools rating “high” on pupil assessment and provide meaningful feedback and had high PLE results, but most schools remedial work show little evidence of written feedback to pupils. Reading and writing are explicitly taught in Reading and writing are not emphasized in the early grades, including use of reading the schools studied. Textbooks are rarely cards used, and students infrequently read in class. P3 students have extremely poor reading skills in general. The school and Head Teacher are External supervision is not found to be externally supervised at least three times related to high PLE results because per term supervision is not focused on teaching and learning. The community is involved in providing A community’s financial and material financial and in-kind support to the schools support of schools is associated with high and parental support to children PLE results. Community participation in school governance is generally very weak. Source: DCI (2004)

30 As noted above, one option for improving the quality of education in Uganda is to introduce stronger teacher incentives. Box C1 relates Chile’s program to reward teachers on the basis of student test performance. Box C2 presents Guinea’s program of school grants to promote quality through teacher training. Box C3 presents some of the evidence on the impacts of teacher incentives on quality.

BOX C1: TEACHER PERFORMANCE INCENTIVES IN CHILE Established in 1994, Chile’s National System of Performance Assessment (SNED) awards teacher incentive grants to schools based on an index of school excellence measures.

Objectives of School Grant. The SNED creates competition among schools to encourage teachers to improve their performance.

Design Features. Chile’s National System of Performance Assessment (SNED) program mandates that schools spend grants in the form of teacher incentive awards and teacher bonuses. The teacher incentive grants are conditional in that awarded school directors must use 90 percent of the grant for teacher bonuses based on hours worked. The school director is to allocate the residual 10 percent to “outstanding” teachers at his/her discretion to avoid the “free-rider” problem. Another design feature of the SNED program is that the teacher incentive grants are distributed through a competitive process. Schools are stratified within regions by socioeconomic status and other external factors that affect school performance. This ensures that the process is competitive among comparable establishments. Every two years, schools are ranked according to an index of school performance measures using the national System for Measuring Educational Quality (SIMCE) test as the basic criterion. Schools can win the teacher incentive grants repeatedly. Source: The Authors.

BOX C2.: SCHOOL GRANTS FOR TEACHERS IN GUINEA Since 1994, Guinea has been implementing a unique and promising World Bank funded program that integrates school improvement with professional development for teachers known as the Small Grants Staff Development and School Improvement Program (PPSE). PPSE is a conditional school grant program that engages primary school teachers to participate in the process of education quality improvement through competitive small grants of approximately $1000 that are awarded to school-based teams of teachers.

Objectives of School Grant. The overall objective of Guinea PPSE is to improve the quality and relevance of specific school inputs, which in this case are teachers. As a means of improving the quality of primary education, PPSE provides organizational support and the incentives necessary for teachers to assume primary responsibility for their own professional development and to determine what is most appropriate in their local context for improving teaching practices. Furthermore, this program seeks to give teachers greater professional autonomy to analyze teaching and learning problems at the classroom level, define the problems or issues to be addressed in a 1-year project, propose and implement solutions, then evaluate and report results.

Design Features. Diverging from the traditional top-down approach of in-service teacher training where central education authorities mandate workshop contents for large groups of teachers, PPSE allows teams of teachers to design professional development programs unique to their local context and compete for grant funds to implement their own programs. Teachers learn about PPSE grant competition through a series of workshops led typically by a

31 pedagogical advisor or a regular school teacher, who presents the program’s operational manual and proposal-writing guidelines. Interested teams of teacher then go through a two-cycle, highly structured competition. First, teacher teams determine the contents of their projects, prepare their own budget, and then submit preliminary proposals for their own professional development program to a prefectural jury, which is presided over by the prefectural director of education (DPE) and composed of retired teachers and local education leaders. Second, once promising proposals are selected, pre-selected teacher teams are invited to revise their proposals with help from the facilitators based on critical comments received from a prefectural jury, and then submit their final proposals to a regional jury who makes final decisions of which team will receive grants. The regional jury is presided over by the Regional Inspector of Education (IRE) and composed of local educational leaders. Selected teacher teams are granted full funding, provided with project implementation support from the project facilitator, and visited by an evaluator, who is typically a prefectural or regional jury member, three times throughout the 1-year project cycle. In addition, since PPSE also has performance incentives as one of its design features, teacher teams are given the option of renewing their grant if they show that their projects attained good results. School grant schemes can also offer incentive based on performance. Source: The Authors.

BOX C3. INTERNATIONAL EXPERIENCE WITH TEACHER INCENTIVES. Teacher incentives can be broadly defined to include instruments that affect: (a) who becomes a teacher, (b) how long they stay in the profession and (c) what they do in class. This broad definition of incentives encompasses “general incentives” such as salaries and benefits, as well as “targeted” incentives such as bonuses given to teachers for their performance or for undertaking special activities (e.g. teaching in remote schools). Incentives can be monetary and non-monetary (e.g. status or career stability).

International experience provides fairly robust evidence that general incentives do have an impact on teaching quality and supply. The level and profile of teacher salaries, both in absolute terms and relative to the salary of comparable workers, matter. Chile’s more-than-doubling of average teacher salaries in the past decade is associated with an increase in the quality of students entering teacher education programs. Similarly, the increased and more equitable distribution of resources resulting from FUNDEF (Fund for Maintenance and Development of the Fundamental Education and Valorization of Teaching) in Brazil led to improvements in student outcomes. In Latin America, low teacher salaries and a flat wage profile are major factors contributing to the poor preparedness of teachers. Individuals that choose to become teachers often are not strong students, are not interested in teaching as a career and do not have the appropriate characteristics to succeed as teachers.

In theory, targeted incentives can be argued to be a superior policy tool to improve teaching quality than across-the-board salary increases on the basis of both fiscal and efficiency considerations. However, there has been very little experience with applying performance-based incentives. Targeted incentive reforms, such as merit pay, are relatively rare and existing plans are often small-scale and short-lived. Various teacher- and school-targeted incentive programs were implemented in the United States. The evidence on these programs’ effects is inconclusive.

32 Other countries like Chile and Mexico implemented national performance-based teacher incentive systems. A review of these and other Latin American countries’ experiences with targeted teacher incentives found that although teachers generally respond to incentives, they do not always do so in the expected way. Design flaws in performance-based incentive reforms were likely behind their lack of uniform success. In addition, many of the gains in student outcomes attributed to targeted incentive reforms have been small or short-lived. Cambodia has a small program recently introduced to recognize best teachers. Three teachers in each province receive a one- time award ranging from R80,000 to R120,000 (USD20-30).

Targeted incentive programs rewarding teacher for undertaking special activities, such as working in difficult areas are far more common than performance-based incentives. Beyond financial incentives, several governments have introduced school-based management reforms giving local communities greater authority over schools, in the hopes of increasing teacher accountability and, as a result, student achievement. The general principle is that engaging communities in school matters makes teachers more accountable for what they do in class and also makes their work more appreciated, thus creating an incentive for teachers to work harder and better. A review of the evidence on school-based management reforms in Central America concludes that while the reforms have improved class size, teacher absenteeism, increased working hours and homework assigned, they did not have an effect on teaching practices.

References: McEwan and Santibañez (2004), Vegas (2005), and Villegas-Reimers (1998). Source: World Bank (2007c)

33 D. INTERNAL EFFICIENCY OF PRIMARY EDUCATION

In 1996 Uganda made a strong political commitment to UPE, and the Education Strategic Investment Plan [ESIP] of 1998-2003 provided the roadmap for meeting that goal. Uganda’s ambitious education reform as enshrined in ESIP went well beyond simply expanding coverage and included curriculum reform, increased provision of learning materials, use of local languages at the lower primary level, reduced procurement costs for textbooks and instructional materials, increased use of in-service training to enhance teacher qualifications, and creation of an Education Standards Agency [ESA] to improve the system of school inspection. Not all elements of the reform have been uniformly and fully implemented to date, although there have been impressive achievements. The implementation of ESIP and the accompanying SWAP arrangements have been evaluated elsewhere and are not the focus of this study.19

Uganda’s UPE policy was implemented rapidly, leading at least in the short run to larger class sizes, higher percentages of unqualified teachers, and fewer school supplies and materials to students. While the Government has attempted to systematically address each of these resource problems, there is still a gap between desired goals and reality on the ground. Policies and programs that are implemented rapidly often lead to waste, so it will not be surprising if this analysis finds inefficiency in how resources are used to deliver primary education.

As noted earlier, an assessment of internal efficiency has several elements, beginning with the identification of possible leakages of resources between the government and the school, leakages within the school itself, and proceeding to analysis of how resources that reach the school may not be productively used.

How much leakage is there between the release of funding by the central government and the receipt of resources by the primary school?

Based on expenditure and personnel audits and evaluations, the estimated leakage of recurrent expenditures between the Ministry of Finance and the schools is Shs. 20 bn , or 6 percent of total budgeted recurrent primary education expenditures.20 As summarized in Table D1, this includes UPE leakages, ghost teachers and MoES administrative waste [questionable expenditures] identified in agency audits. It does not include ghost non-teaching personnel, nor does it include administrative waste at the district level, as there are no available estimates for these items.

Ghost teachers are those who appear inappropriately—for whatever reason—on the payroll. Weak payroll information systems and fraud are two possible reasons for the appearance of ghosts, who are normally identified during “payroll clean-up exercises” in which every teacher on the payroll is verified. A 1993 payroll exercise revealed that 20

19 For example, see Ward, Penny, and Read (2006). 20 See “Annual Budget Performance Report, 2005/06, MOFPED, p. 55), the latest Public Expenditure Tracking Survey (“UPE Capitation Grant Tracking Study, FY 2005/06, USAID), analysis of Accountant General’s detailed budget for the MoES, the Auditor General’s audit of the MoES, and the school survey undertaken as part of this report.

34 percent of primary teachers were ghosts. A 2003 audit undertaken by MoPS found 9 percent of teachers to be ghosts. A conservative estimate of 4 percent is used here. Hence, the loss is 4 percent of the teacher wage bill.

TABLE D1. EXPLAINING THE LOSS OF CENTRAL GOVERNMENT RESOURCES FROM THE CENTER TO THE PRIMARY SCHOOL

Percent Basis Amount Source for Loss Loss (Shs. Estimate Bn.) Expenditure 314 Reaching the School UPE Leakage 16% UPE Grants 5 PETS survey carried out in 2006. Ghost Teachers 4% Wage Bill 11 MoPS (2003)21 Central 334 Government estimates Government of budget releases in Expenditure 2006.

The single largest source of government to school leakage is the UPE grant. The most recent estimate of the percent of leakage is 16 percent.22 The leakage is 16 percent of total UPE grants, or Shs. 5 bn. Leakage has decreased over time as measured by several Public Expenditure Tracking Surveys [PETS]. In addition to the leakage, there is a two month delay from the time UPE grants are released by the Central Government to the time they arrive at schools; this, too, has decreased from the 5 month delay measured in 2001. The cost of this delay is not included in the leakage or waste reported in Table D1. As shown in Table D2, several grant management problems were identified and these include: [1] delays and uncertainty in funding, which make it difficult to plan and spend efficiently; [2] inadequate supervision of construction projects; [3] delays in receiving and damage to textbooks; and [4] failure to use textbooks by teachers and students in the classroom. In terms of financial loss these problems are of relatively minor importance. However, the impact of these losses on student learning may be of considerable importance. Given the very low ratio of books to students [about 1:3] in Uganda, loss and failure to use textbooks may have a considerable impact on student learning.

TABLE D2. GRANT MANAGEMENT PROBLEMS IDENTIFIED BY HEAD TEACHERS Type of Fund Management Problems UPE Delayed and irregular release of funds

21 This is a conservative estimate. In its efficiency study done for the MoPS, Price Waterhouse Coopers cites earlier studies estimating between 9.2 and 20 percent of teachers may be irregularly on the payroll. 22 USAID (2006).

35 SFG Contractors delay implementation of construction School heads not involved in management [turn-key projects] IMG Delayed supplies of textbooks Damaged materials due to storage Choice of materials do not match user’s interests Wages Not received and paid promptly into teachers’ bank accounts Source: Northern Uganda PETS (2006), p. 38.

Unfortunately, there are no clear benchmarks as to what level of leakage from government to school should be expected in a well-managed education system. Even though Uganda’s leakage is only 6 percent of primary education recurrent expenditures, this is equal to twice the total public expenditures on primary school instructional materials in Uganda. In any case, the MoES should undertake aggressive actions to reduce leakage and invest in monitoring and oversight up to that point where the marginal returns equal the marginal costs of monitoring.

Issue: The MoES should regularly monitor leakage of all types and set targets and develop strategies for reducing wastage. This may require strengthening the internal audit unit of the ministry..

How much leakage is there in the school itself?

According to the calculations in Table D1, about Shs. 314 bn in resources reach primary schools. Now, the question is how much of these resources actually reach the pupil. The most obvious leakage at the school level itself is absenteeism. When headmasters and teachers have unexcused absences, pupils fail to receive their services. And there is a serious problem of headmaster and teacher absenteeism in Uganda.

A study of teacher absenteeism carried out in 2004 found an average rate of teacher absenteeism of 27 percent in Uganda23. As shown in Table D3, this was considerably higher than most other countries which carried out similar surveys at the same time. Given the high estimate of teacher absenteeism, this study carried conducted unannounced school visits to 160 government and non-government schools in November 2006. The schools were randomly selected across three regions (Western, Eastern, and Central) and six districts.24 The 2006 and 2004 surveys are identical methodologically, making the results comparable.

TABLE D3. TEACHER ABSENTEEISM RATES, 2002-03

Country Absence rate (%) Bangladesh 15 Ecuador 14

23 Estimates of teacher absenteeism from unannounced visits are significantly higher than absenteeism recorded in official records in most countries. 24 Habyarimana (2007); see Annex 2.

36 India 25 Indonesia 19 Peru 11 Papua New Guinea 15 Uganda 27 Zambia 17

Sources: Chaudhury, Hammer, Kremer, Muralidharan, and Rogers 2004 for most countries; NRI and World Bank 2003 for Papua New Guinea; Habyarimana, Das, Dercon, and Krishnan 2003 for Zambia Note: Absent staff are fulltime teachers on current shift who were not found anywhere in the school at the time of an unannounced visit.

As shown in Figure D1, the 2006 survey finds teacher absenteeism has improved somewhat [possibly due to MoES initiatives] with 19 percent of teachers having unexcused absences from the school25. This translates into a loss of Ush 60 bn. of teacher’s time, almost double the government’s financial annual contribution to Makerere University. Absenteeism varies by teacher rank with 27 percent of head teachers, 16 percent of senior teachers, and 14 percent of junior teachers being absent.

The magnitude of leakage due to headmaster and teacher absenteeism present a powerful case for investing in actions to reduce absenteeism. Analysis of variations in teacher absenteeism across schools finds that strong parental involvement in the school, competition from a nearby non-government school, and the presence of useable teacher housing are just some of the factors that contribute to lower absenteeism rates26. While statistical analysis suggests policies that may deter absenteeism, Annex A demonstrates that with the exception of the India experiment described in Box D1, teacher incentive experiments have not been successful in reducing absenteeism. .

BOX D1. MONITORING TEACHER ABSENTEEISM IN INDIA Duflo and Hanna (2006) evaluate a randomized intervention in community schools in which an NGO provides cameras to teachers and institutes attendance-dependent remuneration. Teachers are expected to take pictures every morning, and teachers will be paid depending on the number of “full” days attended. The results of this intervention were surprisingly large. Teacher absence fell by about half from a high of 36% in comparison schools to 18% in program schools. The authors discuss the political-economy of this intervention and conclude that it is not a realistic option for national scale up. Source: Duflo and Hanna (2006).

There is strong evidence that teacher absenteeism can be reduced by monitoring. While in most countries, teachers are required to sign an attendance book, this data is rarely collected and analyzed or used to monitor average attendance for each teacher. Private schools report higher teacher attendance, even when they pay the same or less

25 This is considerably smaller than the 27% absenteeism rate estimated in the 2002/03 study. 26 See Annex A.

37 than government schools, partly because they have managers at each school who monitor attendance. In the Gambia, teachers in church schools receive the same pay as those in government schools, but they report much higher teacher attendance because the church is responsible for distributing the pay, and can withhold payment to teachers who have poor attendance. Also in the Gambia, teacher attendance is reported to have increased since the introduction of a system of cluster monitors, which ensures that each school is visited regularly.

Some writers have suggested that absenteeism could be improved by community monitoring. Clearly parents have a strong motivation to ensure that teachers attend, and their willingness to pay for low cost private schools is often explained by the belief that teacher attendance is higher there. To date it is difficult to find examples of successful mechanisms to build on this motivation.

Some teacher absenteeism is caused by the need to travel to collect pay. In Liberia and Zambia, rural schools can close for up to a week each month as teachers leave to collect their pay. In Lesotho and the Gambia the government has begun to pay teachers by electronic transfer to avoid this problem. However this sometimes makes it more difficult to withhold pay from teachers who have absconded, or moved to another school without permission. Where the banking system is not sufficiently developed to allow bank transfer, some countries like Malawi have arranged for district officials to travel to each school to deliver pay. This is expensive and time consuming, but provides an opportunity for district officials to check attendance in each school on a monthly basis.

FIGURE D1. WHERE TEACHERS ARE AT TIME OF ENUMERATOR’S VISIT.

Can't find teacher, In class, teaching, 19.2% 18.2%

Out of class, break, 17.6%

Out of class, in Administrative school, 34.2% work, 8.1%

With surveyor, In class, not 0.2% teacher, 2.4%

38 As shown in Table D4, the leakage of central government resources due to headmaster and teacher absenteeism reduce the expenditure reaching students to Shs. 254 bn, or 76 percent of government recurrent expenditures.

39 TABLE D4. EXPLAINING THE LOSS OF CENTRAL GOVERNMENT RESOURCES IN THE SCHOOL

Percent Amount Basis for Loss (bn. Ushs) Estimate Expenditure 314 Reaching the School Loss Due to 19% 53 Teacher Headmaster and unexcused Teacher absences from Absenteeism 2006 school survey. Expenditure 261 Reaching the Classroom Note: The loss due to absenteeism is 19% of the Ush 276 bn wage bill.

The picture becomes still more complicated if one considers student absenteeism. Student absenteeism is rarely measured accurately by administrative records. Even asking teachers to recall student absenteeism tends to understate the magnitude of the problem.27 Only direct observation yields an accurate answer.

TABLE D5. STUDENT ABSENTEEISM PATTERNS IN KABULASOKE AND NAKASEKE, 2004

DISTRICT P.3 P.6 Kabulasoke Boys 36.4% 31.5% Kabulasoke Girls 32.1% 29.5% Nakaseke Boys 25.1% 20.9% Nakaseke Girls 25.7% 15.7% Source: Education Standards Agency, Report on Monitoring Learning Achievement in Lower Primary, 2004

Table D5 shows the results of direct observation in two districts. Student absenteeism rates are between 16 and 36 percent. Other direct observation studies have also found high absenteeism rates.28 The causes of absenteeism are multiple. One study found the principal reasons to be illness, household work, and the low value put by parents on education.29

27 This was the procedure used in the 2000 SACMEQ study, for example. 28 See the Musisi (2006) study, which found pupil absent on average 23 days per year. 29 Musisi (2006).

40 Using statistics reported in observational studies, one can assume an average primary school student absenteeism rate of about 20 percent. This means that only 80 percent of the resources that reach the classroom in turn reach the student. In the case of Uganda, Shs. 52 bn are lost as the result of student absenteeism30. If one adds the losses associated with student absenteeism to the leakages given in Tables D1 and D4, total leakage between the central government and the pupil is on the order of Shs. 125 bn, or 37 percent of the government primary education expenditure..31

The MoES should [a] closely monitor leakages due to absenteeism and [b] adopt measures to attempt to reduce absenteeism. Improved school inspection, community and PTA monitoring, higher penalties for unexcused absenteeism, and bonuses for high attendance rates are all measures that could reduce teacher absenteeism. Student absenteeism could be reduced through stronger incentives to schools to ensure students attend class by, for example, tying UPE grants to average daily attendance, and stronger incentives to parents to send their children to school by, for example, providing free school lunches or cash transfers to households that regularly send children to school.32

The capacity to monitor teacher absenteeism and other leakages is dependent on better school inspection. Like Uganda, most countries have external supervisory unit, expected to visit schools to monitor and enhance quality. While these personnel were traditionally called inspectors, this role has now a wide variety of titles. These inspection/supervision staff are often based either at district offices or at a central headquarters. Frequency of inspection visits is usually severely curtailed by transport and logistical difficulties, and most schools are visited less frequently than once a year. Where transport is a major problem, the most isolated schools tend to be visited least frequently.

A few countries have managed to have more frequent supervision by having a highly decentralized inspection staff. In the Gambia a “cluster monitors” is allocated to every ten schools, and is expected to live a one of the schools, and travel by motorbike to visit each school every 2 weeks. There are indications that teacher absenteeism has been reduced following the introduction of these frequent external visits. In Eritrea, there is a cluster leader for every 80 teachers, with the expectation that each teacher can be observed twice a year.

.

30 If as a result of student absenteeism teachers spend larger amounts of time with the remaining students in the classroom, the Shs. 52 bn ia an overestimate of the true loss. But most evaluations of teaching in Uganda suggest that having marginally fewer students in the classroom does not affect traditional pedagogical practices. 31 These estimates do not include books which are sent to the school but are not used by the teacher or other losses of non-personnel inputs. 32 This is the principal behind Mexico’s Progresa and Brazil’s Bolsa Familia, both of which give small cash transfers to families that send their children to school. See Box D1.

41 BOX D2. MEXICO: FINANCIAL INCENTIVES TO ATTEND SCHOOL

42

How do high repetition and dropout rates affect efficiency in the production of primary school graduates?

If the goal is a primary school graduate, students who fail to graduate can be thought of as having “wasted” the resources expended on them33. Since the unit cost of an enrolled student is Ush. 50,534, a student who successfully went through primary school with no repetition would cost Ush. 353,738. In reality, one primary school graduate, including repetition and dropout, costs about Ush. 923,833, a difference of Ush. 570,095. In other words, the actual unit cost of a graduate is 2.6 times what would be the case if there were no repetition or dropout. The large “savings” from reducing repetition and dropout argue for aggressive efforts on the part of the MoES to improve the quality of instruction and to enforce automatic promotion.

One of the principal determinants of repetition and dropout is the age at which a child enters primary school. Students who enter late for their age face low probabilities of academic success. Unfortunately, Table D6 shows that 62 percent of students entering P1 are older than age 7, the normal entrance age. Programs to encourage on-time, first- time enrollment of children could thus have a large impact on repetition and dropout rates.

TABLE D6. DISTRIBUTION OF HOUSEHOLD MEMBERS BY CLASS AND AGE Class 5 6 7 8 9 10 11 12 13-24 Total P1 4.7 12.9 20.1 23.1 14.2 12.8 4.4 4.0 3.8 100 P2 1.0 3.5 8.8 17.6 17.2 20.9 9.1 10.9 11.0 100 P7 0.9 0.8 5.7 92.5 100 Source: UBOS, 2004 National Service Delivery Survey, p. 15

Is government funding allocated efficiently across districts?

Government primary education spending per pupil often varies across districts or schools due to differences in cost structure [especially, due to higher costs of delivering schooling in remote, sparsely populated regions] and differences in poverty levels [with education spending being higher to compensate for the effects of poverty]. Since the Ugandan EMIS does not capture all spending at the school level and since salaries represent approximately ninety percent of all central government funding of primary education, this analysis uses the personnel wage bill as a proxy for total government funding. Nationally, the mean wage bill per student is 39,259 UShs, but there is considerable variation across districts and schools. For example, the top spending quintile of schools has a mean wage bill per student of 50,526 UShs, almost double the figure [26,585 UShs] for the bottom quintile of schools in the country34. It’s appropriate 33 Alternatively, one could define the goal as completion of 4th grade, which often corresponds to literacy, or some other measure. 34 Calculations based on the 2004-05 EMIS.

43 to ask whether this variation in wage bill per student is the result of explicit MoES policy —perhaps to compensate for high costs in sparsely populated, rural areas—or is the result of policies and practices having nothing to do with the effective delivery of education to children.

FIGURE D2. GOVERNMENT WAGE BILL PER STUDENT ACROSS DISTRICTS

Government wage bill per student by district -2 s.d. -1 s.d. mean +1 s.d. +2 s.d. +3 s.d. +4 s.d. 5 2 0 2 s t c i r t 5 s i 1 d

f o

r e b 0 1 m u N 5 0 20000 30000 40000 50000 60000 70000 Government wage bill per student Source: EMIS (downloaded Oct 2006) and World Bank calculations

Variance in the wage bill per student across schools can be disaggregated into the variance that lies between districts and the variance that lies between schools within districts. Figure D2 illustrates the distribution of the wage bill per student across districts. Since the Central Government’s Ministry of Public Service [MoPS] directly deploys teachers in the country, this distribution is the direct result of its actions and subsequent teacher transfers.

Most education systems allocate teachers to schools based on the number of students enrolled in a school. If that were the case in Uganda, the distribution in Figure D2 would be much more tightly distributed around the mean. So what could explain the wide variation in wage bill per student.across districts? Perhaps teachers are allocated according to need as measured by poverty levels of districts. To investigate this possibility, the district wage bill per student is plotted against district poverty rates in Figure D3.

FIGURE D3. WAGE BILL PER STUDENT AND DISTRICT LEVEL POVERTY

44 80,000

) 70,000 d l i h c

60,000 d e l l o

r 50,000 n e

r e

p 40,000

l l i b 30,000 e g a w

t 20,000 ' v o g ( 10,000

0 0 20 40 60 80 100 (percent of population living in poverty)

As is readily seen, the wage bill per student across districts in Uganda is in fact inversely, not directly, related to need. Furthermore, regressing the wage bill per student against district poverty rates and the percent of district population that is rural [another common determinant of expenditures internationally] yields the finding that the percent rural population is not related to the wage bill, while the poverty rate is strongly negatively related to the wage bill. In short, primary education expenditures [as proxied by the teacher wage bill] in Uganda appear not to be distributed according to the types of variables often found in other countries.35

35 For example, in Chile a local government’s education expenditures are largely based on a capitation grant from the central government’s education ministry. Children from poor households or children in rural schools are weighted more heavily than other students such that average expenditures per pupil are somewhat higher in local governments having high percentages of children from poor households and high percentages of rural population.

45 FIGURE D4. MEAN CLASS SIZE BY DISTRICT, 2006.

Mean class size by district in public and non-public schools s s a

l 100 c

r

e 80 p

s t

n 60 e d u

t 40 s

f o

20 r e b

m 0 u

n Tororo Mayuge Luweero Mukono Kibaale Ntungamo public non-public

Figure D4 also looks at the distribution of central government funding, this time proxied by average class size, across the six districts used as the sample for the unit cost survey. Here, too, there is a large difference in average class size in public schools—but not private schools--across districts. Again, it is not clear what the rationale is behind the allocation algorithm that determines this distribution.

Of course, teacher deployment may be done on the basis of sound criteria not immediately evident to the analyst. If so, one would expect a positive relationship between the wage bill per student and measures of educational outcomes. However, as shown in Figure D5, there is also no obvious relationship between the wage bill and a crude measure of educational outcome [the ratio of P5 to P1 students]. For example, for the expenditure level Shs. 40,000 on the horizontal axis, one finds the measure of educational outcome to vary between 0.10 to more than 0.90. If government policy were to increase funding for districts of low educational outcome [a measure of need], one would expect to find higher expenditures for low outcomes and lower expenditures for high outcomes.

46 FIGURE D5: WAGE BILL PER STUDENT AND EDUCATIONAL OUTCOMES ACROSS DISTRICTS IN UGANDA.

1.00 )

s 0.90 t n e

d 0.80 u t s 0.70 1 P

o 0.60 t

s t

n 0.50 e d u

t 0.40 s

5 P

0.30 f o

o 0.20 i t a r

( 0.10 0.00 0 20,000 40,000 60,000 80,000 (gov't wage bill per enrolled child)

Is government funding allocated efficiently across schools within districts?

While the central government [MoPS] deploys teachers—and thus the wage bill— across districts, it is the district governments [district teacher service commissions] which deploy teachers within their districts, presumably following national guidelines.

Figures D6 and D7 show the distribution of school average government wage bills per student [i.e., excluding community funded teachers] in two districts. Both districts illustrate very large differences in per pupil allocations, which are unrelated to any stated objective allocation criteria.

47 FIGURE D6. GOVERNMENT WAGE BILL PER STUDENT ACROSS SCHOOLS WITHIN LUWERO DISTRICT

Government wage bill per student within a particular district Luwero district -2 s.d. -1 s.d. mean +1 s.d. +2 s.d. +3 s.d. +4 s.d. 0 6 0 5 s l o o 0 h 4 c s

f o

0 r 3 e b m u 0 2 N 0 1 0 0 20000 40000 60000 80000 Government wage bill per student Source: EMIS (downloaded Oct 2006) and World Bank calculations

FIGURE D7. GOVERNMENT WAGE BILL PER STUDENT ACROSS SCHOOLS WITHIN NTUNGAMO DISTRICT

Government wage bill per student within a particular district Ntungamo district -2 s.d. -1 s.d. mean +1 s.d. +2 s.d. +3 s.d. +4 s.d. 0 6 0 5 s l o o 0 h 4 c s

f o 0

r 3 e b m u 0 2 N 0 1 0 0 20000 40000 60000 80000 Government wage bill per student Source: EMIS (downloaded Oct 2006) and World Bank calculations

As with the central government’s allocation of wage expenditures across districts, it may be that there is a direct relationship between that allocation and unobserved

48 measures of need. If so, one would expect a positive relationship between the allocation of the wage bill and measures of educational outcomes. Figure D8 shows just such a scatter diagram. Once again there is no obvious relationship.

FIGURE D8. GOVERNMENT WAGE BILL PER STUDENT AND A CRUDE MEASURE OF SCHOOL OUTCOMES. 1 1 P

n i

d e l l o r 8 n . e

s t n e d u t s

6 . o t

7 P

n i

d e 4 l . l o r n e

s t n e d 2 . u t s

f o

o i t a 0 R

5000 10000 15000 20000 25000 Gov't wage bill per student

In conclusion, there is no obvious relationship between the distribution of education expenditures [i.e., the wage bill] across districts or across schools and the usual measures of need. In addition, there is no relationship between the distribution of education expenditures and crude measures of education outcomes—in Figure D8 there is no discernible pattern to the data. The distribution of teachers appears to be unrelated to educational criteria and is thus inefficient.

Are teachers used productively within schools?

FIGURE D9. MEAN CLASS SIZE BY GRADE IN PUBLIC AND NON-PUBLIC PRIMARY SCHOOLS

Mean class size by grade in public and non-public schools 100 s

s 90 a l c

80 r e

p 70

s t 60 n e

d 50 u t

s 40

f o

30 r e

b 20 m

u 10 n 0 grade 1 grade 2 grade 3 grade 4 grade 5 grade 6 grade 7 public non-public

49 A principal determinant of unit cost is class size, and we find higher class sizes in the lower than the upper grades for public schools. For private schools the differences across grades are relatively minor. In other words, children in the lower grades receive fewer resources than those in the higher grades in public schools36. Most educators would argue for the reverse: the lower grades should have smaller class sizes, and higher unit costs, than the higher grades within primary schools. Large class sizes in the early grades are more likely to contribute to higher repetition and dropout, especially among children from lower income homes.

FIGURE D10. RATIO OF ENROLLMENT IN P7 TO P1 AND STUDENT TEACHER RATIO FOR GRADES 1-3

Ratio of enrollment in P7 to enrollment in P1 and student /teacher ratio for grades 1 to 3 t n e 1.0 m l l 0.8 o

r y = -0.0056x + 0.5508 n 0.6 2 e

R = 0.2321 1 0.4 P /

7 0.2 P

o 0.0 i t a

r 0 20 40 60 80 100 student/teacher ratio for grades 1 to 3

As shown in Figure D10, there is a weak correlation between large class sizes in P1-P3 and the survival rate to P7. According to the regression line, having a student teacher ratio of 40 instead of 60 would result in a survival rate that was 0.11 points higher according to this simple regression line. However, the large dispersion around the line indicates that having a small student-teacher ratio is not a guarantee of high performance. Some schools with low student-teacher ratios perform poorly but almost no schools with high student-teacher ratios perform well.

The productivity of primary school teachers appears to be low in general. Average class sizes are considerably larger than student-teacher ratios. The discrepancy between class size and student-teacher ratio suggests that teachers have relatively light workloads. Figure D1 is consistent with this conclusion, showing that only 21 percent of teachers are physically in the classroom at the time of an unannounced survey.37 The use

36 While official education policy as stated in the ESSP is for smaller class sizes in P1 and P2, headteachers do not allocate teachers across grades this way. 37 This result may be influenced by the timing of the survey, which took place in November, at the end of the school year. An earlier study carried out in 2003 found 41 percent of teachers physically present in the classroom, although only 26 percent were actively teaching.

50 of teacher time both within the school and within the classroom needs further investigation, but the preliminary evidence suggests that teachers are underemployed in public primary schools.

Other than personnel, are expenditures allocated efficiently across inputs?

Research shows that having an adequate number of textbooks can be very productive, especially when teachers are not adequately trained. While the Government’s goal is to increase the ratio of books to students, at present these ratios [0.2 textbooks per pupil for P1-P3 and 0.33 textbooks per pupil for P4-P6] are far below accepted international norms. Research evidence tells us that the quality of instruction and the level of student learning could be improved by employing fewer teachers—especially given current low student-teacher ratios—and purchasing more textbooks and instructional materials, assuming the textbooks are in fact used by teachers and students.

Could the way resources are organized to deliver education be improved substantially [i.e., change the production technology]?

Public primary schools in Uganda are predominantly organized as single shifts with single grade classrooms. In principle, multigrade classrooms can be as effective as single grade classrooms, and they can be considerably more cost-effective, especially in rural areas where the number of pupils in a single grade is low. About one-fifth of primary schools in Uganda have less then the 7 teachers that would be required to teach all grades in a monograde setting. Also, the high dropout of students in many schools means that class sizes are often small in P6 and P7, meaning these grades should have highest priority for multigrade teaching. In 1998 the MoES initiated a multigrade pilot in two districts, Kalangala and Sembabule. An evaluation of this experience carried out in 2006 found that multigrade teaching can be as effective as monograde teaching, but success requires continued teacher training and management attention from headmasters.38

In areas where school facilities and/or the supply of teachers is constrained, double-shifting is an option for making more intensive use of school facilities while employing current teachers for a second shift at considerably less than double the wage of a single shift39. Double shifts have little impact on instructional quality, but some students may have difficulty traveling to or from schools when shifts end after sunset.

What is the status of school infrastructure? According to the 2004 National Service Delivery Survey, the physical facilities of primary schools are woefully inadequate, especially classrooms, teachers houses, and toilets. Poor infrastructure can pose health and security risks and thus adversely affect learning . Inadequate toilet facilities has been shown to contribute to dropout when girls reach puberty. Indeed, in the 2004 Survey 30.2% of respondents reported inadequate buildings as the most serious 38 Higgins, et.al. (2006). Effectiveness was judged on the basis of student attendance and retention and PLE results as well as classroom observations of multi-grade classrooms relative to a control group of single grade classrooms. 39 Smith (2007)

51 constraint to school performance, while 17.6 percent reported this as a serious constraint, more than any other item. While the Government has been addressing this problem, it has not always done so efficiently. The School Facilities Grant has served to decentralize construction activities to the district level, but SMCs and headmasters have very little voice and even less responsibility. There’s a need to do a systematic inventory of school infrastructure with the aim of identifying the highest priorities for investment.

TABLE D7. EDUCATIONAL FACILITIES: PERCENT RESPONDING AVAILABLE AND ADEQUATE Facility Available Adequate Class rooms 98.8 28.5 Teachers houses 51.8 8.3 Library 15.3 26.8 Laboratory 0.5 42.9 Workshop 1.3 33.3 Latrine/Toilets 97.6 30.9 Source: UBOS, 2004 National Service Delivery Survey, p. 19

How can incentives be used to improve school and student performance?

Educational researchers have begun to experiment with and rigorously evaluate how to use incentives to improve school and student performance. The incentives they evaluate are located at three main levels:

 Teacher level incentives such as prizes, remuneration bonuses, promotion and individual recognition. Outcomes such as average test scores are measured at the teacher level. This set of incentives can also involve negative interventions such as naming and shaming or even firing poorly performing teachers.

 School level incentivesl: high performing schools (measured at the level of the school, rather than teacher incentives as above) are given collective rewards. These can include private rewards to all teachers (sharing a collective pot of money or other prizes), or a local public good at the school such as a staff room, library , etc.

 Household level incentivesl: Incentives located at the household typically aim at motivating students to increase the effort dedicated to learning. Prizes are announced at the pupil level and recipients typically receive scholarships. Other forms of household level incentives include empowering of parents/local communities to monitor teachers or manage schools directly.

52 Teacher level incentives work by directly raising the returns to teacher effort).40 For these to work, teachers must believe that the costs of increased effort are less than the expected gains in remuneration (or whatever the form of the performance-based reward). Western Kenya and India provide examplses of such interventions.

Teacher Incentives in Western Kenya. This study attempted to measure the impact of a randomized intervention in which high performing teachers would be rewarded with prizes at the end of the school year.41 Prizes were substantial and included bicycles, mattresses and other household durables. Performance was determined by the rank of the school in the district and the degree of improvement relative to the previous year’s exam results. The authors measured a variety of inputs before and after the intervention. While the authors find improvements in test scores in ‘treatment’ schools, this improvement is driven primarily by teachers using different teaching strategies – particularly teaching to the test. There is no significant different across treatment and comparison schools in teacher absenteeism (teacher absence rates are around 20% in these schools).

Monitor-less monitors in India. Another teacher incentives paper evaluates the impact of monitoring and performance-based remuneration on teacher absence.42 It evaluates a randomized intervention in community schools in which an NGO provides cameras to teachers and institutes attendance-dependent remuneration. Teachers are expected to take pictures every morning, and teachers are paid depending on the number of “full” days attended. Teachers in treatment schools received a base pay as well as performance-based (measured by ‘full days attended’) component. The results of this intervention were surprisingly large. Teacher absence fell by about half from a high of 36% in comparison schools to 18% in treatment schools. In addition, test scores are higher in treatment schools. The authors discuss the political-economy of this intervention and conclude that it is not a realistic option for national scale up.

School-based incentives rely on mobilizing peer monitoring/head teacher effort to support higher teacher effort levels. This India study outlines an innovative approach that evaluates group (or school-level incentives) vs teacher incentives43. This paper seeks to determine the impact of two broad interventions on rural primary schools in India: 1) a set of smart inputs (an additional volunteer teacher and cash block grants vs 2) group or individual teacher incentives (performance-based pay). These four interventions are evaluated in a randomized-control design in Andra Pradesh, India. The four treatment groups are: (i) individual-incentives schools, (ii) group-based incentives schools, (iii) schools that get an additional teacher, and (iv) schools that receive block grants (equivalent expenditure).

This study finds evidence in support of the incentives treatment arms. Learning outcomes are higher in incentives schools, but the authors do not find any differences in treatment impact between group vs individual incentives. They conclude that it is likely that the small size of the schools (three teachers on average) makes it easy for peer 40 See Jacobson (1989) for an early intervention. 41 See Glewwe, Illias and Kremer (2003). 42 See Duflo and Hanna (2006). 43 See Karthik and Sundararaman (2006).

53 monitoring to have the same effects as individual incentives. However, like the teacher incentives study in Kenya, teacher attendance is not affected. Instead, teachers choose to increase “cheap effort” – assigning more homework and practice tests rather than show up.

Household incentives. Finally incentives located at the household level typically work through their effects on student effort or mobilizing parental involvement in monitoring or the management of schools. Improvements in student effort can have a profound effect on the learning environment and greatly boost the satisfaction that teachers get from teaching. As such, teacher attendance and preparation can improve dramatically under the ‘collective mission’ to win merit-based scholarships. Western Kenya is an example of such an intervention.

In Western Kenya a randomized intervention was carried out in which a series of pupil scholarship possibilities are announced at the beginning of the school year44. About 200 girls are eligible for scholarships in 60 schools in two districts (about 15% of the eligible enrollment). The scholarships pay for tuition for the next two years and parents receive an unconditional cash transfer of $12 per recipient. The results of this intervention were quite dramatic. Performance of all pupils, including non-eligible scholarship recipients (boys), in treatment schools perform much better than control schools. In addition, teacher absenteeism falls by 6.5 percentage points in treatment schools. The study attributes this to an improvement in working conditions engendered by increases in pupil effort.

The other form of incentives-at-the-household-level includes empowering parents to monitor teachers. This ranges from establishing school management committees to outright hiring, firing and remuneration policy control. A number of studies have looked at the effects of greater community control in Latin America and show large effects on school performance.45)). When control was transferred to the community, so that parents could hire and fire teachers, teacher attendance and test scores went up.46 While the associations here are very strong, it is difficult to interpret these results as causal. An interesting analogy is the community monitoring study in Uganda that focuses on health care centers. Using report cards as the means to provide information to the community on the performance of health centers, communities that were randomized to receive this information have much better performing health care centers. Attendance of health center staff and quality of service were higher in report card communities.

Are the accountability mechanisms strong enough at the primary level?.

Education systems with strong accountability have transparent financial and resource flows, provide timely and accurate information to all stakeholders, and provide incentives, or consequences, for good and bad performance. Accountability in education is provided by two principal mechanisms in Uganda. The first is community

44 See Miguel, Kremer and Thornton (2005). 45 See see Jimenez and Sawada (1998) and King and Ozler (2001). 46 See Lewis (2005) for a recent review.

54 participation in and monitoring of school budget and performance through participation in School Management Committees. For the SMCs to play this role effectively they require active participation by parents and information provided to parents as to [a] the minimum requirements or standards for learning to take place, [b] the performance of their schools in terms of budget and school outcomes, and [c] the responsibility and capacity to hold headmasters and teachers responsible for poor performance.

Table D8 shows that SMCs and PTAs exist in almost all primary schools in the country, but they offer relatively limited opportunities for broad participation.l According to the 2004 UBOS National Service Delivery Survey 22.4% of PTA’s and SMC’s meet monthly, but 58.6% meet only once a term. Research in other countries demonstrates that one of the variables most strongly related to student performance is parental participation in the school.

55 TABLE D8 . SCHOOL GOVERNANCE, 2003 Frequency of Meetings Percent School Management Committee Exists 95.8% SMC Meets Once per Term 43.5% SMC Meets Twice per Term 47.8% SMC Meets Less than Once per Term 8.6% PTA Exists 70.8% PTA Meets Once per Term 37.5% PTA Meets Twice per Term 37.5% PTA Meets Less than Once per Term 25.0%

Source: Education Standards Agency, Report on Monitoring Learning Achievement in Lower Primary, 2004

The second mechanism for ensuring accountability is the inspection system. Each DEO is responsible for the inspection of schools within its district. However, there are few inspectors, and they seldom have the vehicles and fuel to visit schools47. The central government’s Education Standards Agency is in principle responsible for overall inspection of the school system but lacks the budget, manpower, and authority to carry through on this mandate.

In the absence of an effective inspection system, and it appears very unlikely that the funding will be forthcoming in the medium term to support an adequate inspection system, the best option open to the MoES may be to strengthen local governance, and give local governing bodies the information and the authority to perform effectively.

47 A 2007 survey by the ESA found that most districts had only three inspectors to cover as many as one thousand schools. For example, Wailiso district has three inspectors and 1001 primary schools, and Kampala has three inspectors and 985 primary schools. The number of schools per inspector also varies greatly across districts. For example, Kalangala district has two inspectors and only 25 primary schools.

56 E. EFFICIENCY OF PRIMARY TEACHER EDUCATION.

Uganda has forty-seven Primary Teacher Colleges [PTCs], forty-five of which are public and two of which are private. The PTCs use a standard curriculum, defined by Kyambogo University. Of the forty-five public PTCs, twenty-three are designated core PTCs. All PTCs provide residential training leading to the Primary Teaching Certificate [Grade III], which is the minimum required qualification. The core PTCs also offer a three-year, part-time in-service training program, which is delivered through 539 affiliated Coordinating Centers [CCs], each of which is staffed by a Coordinating Center Tutor [CCT], who is an employee of the corresponding PTC. On average, each core PTC staffs about 23 CCs, and each CC serves on average18 schools. In addition to this training program, the CCs offer in-service training and technical assistance to all staff of primary schools, including management training of head teachers.

Teacher Supply and Demand.

As of 2006, about 145 thousand teachers were employed in Uganda’s primary schools, including about 19 thousand privately employed and about 126 thousand employed by the TSC.. This number represents a large increase over the 89 thousand teachers employed at the introduction of UPE in 1997. To accommodate the demands of UPE, many uncertified, untrained teachers were employed, and at the same time pupil- teacher ratios increased dramatically. The response of the Ministry to the problem of unqualified teachers has been to put emphasis on in-service teacher training—the Teacher Development and Management System [TDMS]. However, as shown in Table E1, the number of unqualified teachers enrolling in the three year in-service program and passing the Grade III teacher’s examination is only one percent of the stock of teachers. The number of unqualified teachers achieving passes annually through the in-service program represents only six percent of the total number of unqualified teachers.48

As shown in Table E1, the number of pre-service teachers passing the Grade III examination is smaller than the number of teachers leaving the teaching profession, which means that even if all Grade III passes in fact enter teaching, unqualified teachers must be employed to fill the gap. Indeed, the percent of licensed teachers appears to be increasing49. Also, while the post-UPE enrollment bubble peaked between 2003-6, the Ministry policy to reduce student-teacher ratios and demographic trends suggest the demand for teachers will continue to grow, albeit at a more moderate pace than the past decade50. At the same time, the current level of output from the PTC pre-service and in- service programs is not sufficient to increase the overall level of teacher qualifications. If the goal is to increase the percent of qualified teachers, it’s clear that the annual number of Grade III passes needs to increase. This can be done by increasing the pass rate or by increasing enrollments in the pre-service and/or in-service programs.

48 MoES ESSAPR (2006), p. 41. 49 World Bank (2007) reports 89.7 percent of primary teachers are Grade III or higher, but the average district percentage of qualified teachers is as low as 55.1 percent (Nakapiripirit District in the Northeast). 50 Projected enrollments decrease from 2003-2006 and then increase again. Between 2006 and 2015 enrollments are projected to increase by 2.6 million, or 40 percent over the 2006 base.

57 TABLE E1. STOCK AND FLOW OF PRIMARY SCHOOL TEACHERS, 2004

Number Percent of Stock Stock of Teachers 147,291 100% Teachers at Grade III 93,831 64% Licensed Teachers 22,756 15% Transfers 11,760 8% Annual Departure from Teaching 6,843 5% Annual Pre-Service PTC Passing Test 5,746 4% Annual In-Service PTC Passing Test 1,334 1% Additional Teachers Required by 2015* 42,001 29% * Assuming constant PTR = 55:1

Efficiency of Teacher Education.

In total, the PTCs enroll almost 18 thousand students in the two-year residential pre- service training . The dropout rate between the first and second year of the program is about 12 percent, and the failure rate of PTC students taking the Grade III examinations is about 22 percent51. This high failure rate represents a significant wastage of resources that could be reduced by better selection of PTC pre-service students and improved evaluation and assessment of students throughout the two-year program.

Since no in-depth study of PTC expenditure and finance has been undertaken since the introduction of UPE, it’s difficult to assess the internal efficiency of the PTCs. However, unit costs appear to be high, especially for the pre-service program. Teacher education recurrent expenditures were UShs. 16.6 bn in FY 2005/06, or 2.8 percent of the total MoES recurrent budget and 6.5 percent of primary education salaries. The UShs. 16.6 bn. in recurrent expenditures compares with UShs. 8.4 bn in development expenditures for teacher education. Since 17,511 students were enrolled in PTCs in 2004/05, the approximate unit cost of an enrolled student was UShs. 950k, or almost twenty times the unit cost of primary education. This compares with the estimated unit cost of UShs. 243k calculated in the 1996 study. The approximate unit cost of a graduate [of both the residential and in-service programs] was about UShs. 484k in 1996 and about UShs. 2.3 mn in 2005/06. Data do not permit a separate calculation of the unit cost of a graduate of the pre-service and in-service programs, but a study of similar teacher training programs in Malawi found that the unit cost of pre-service programs is about triple that of in-service programs52. Given the fact that pre-service programs do not

51 ESSAPR (2006). The pass rate varies by year, ranging from 61 percent in 2002/03 to 85 percent in 2005/06. 52 Kunje and Lewin (2000).

58 produce higher Grade III examination pass rates, the cost-effectiveness of the CC in- service program is likely to be higher and thus argue for expansion of that program.53

A 2004 value for money audit of the CCs carried out by Ernst and Young for the Auditor General provides additional information on the internal efficiency of teacher in- service training. Among the problems identified are:

 Operational costs are under-budgeted;  Delays in funding required for course delivery;  Lack of funding for equipment maintenance and replacement;  Some supplies (e.g., bicycles) provided but not used.

Two design problems affecting the effectiveness, and thus the cost-effectiveness, of the CCs have also been identified.54 One problem is mission creep. Since the CCs are in close proximity to the schools and in regular contact with head teachers, teachers and communities, they are asked to take on a wide range of activities. As a result, the CCTs have long and heavy work schedules and sometimes take on tasks for which they are not qualified. Mission creep prevents the CCT from focusing on its core functions, including in-service training. Another problem is sometimes a lack of coordination and clarification of roles between the CCTs and the district inspectors. The CCTs report to the MoES through the PTCs, but given their proximity to the schools they also serve as the de facto first line of inspection, which is a district responsibility. When CCTs and district inspectors work harmoniously they can complement each other’s work, but there is no administrative requirement that they do so.

Need for Data and Analysis.

The relatively low quality, the high costs, and the lack of recent data and economic analysis on teacher education makes it difficult to document the magnitude of efficiency problems, to diagnose their causes, and to recommend options for improvement. The PTC Cost Effectiveness and Efficiency Study carried out by MoES in 1996 should be repeated to attempt to more rigorously answer the following questions:

1. What are the unit costs of producing a qualified teacher via [a] traditional, pre- service training and [b] in-service training provided through the CC? Given the multiple activities of the CC, answering this question will require careful cost accounting.55 Also, what would be the unit costs of each training modality if all inputs were provided in adequate amounts?

2. What is the contribution of training to the probability of passing the Grade III entrance examination, and what is the relative cost-effectiveness of pre-service and in-service training?

53 Again, the pass rate for in-service teachers varies by year, from 64 percent in 2002/03 to 94 percent in 2003/04. See World Bank (2007) for further information. 54 Burke (2002). 55 See Kunje and Lewin (2000) for a discussion of the difficulties involved.

59 3. Given likely future demand for new teachers, what is the best combination of pre- service and in-service training in terms of the percent of newly qualified teachers produced by each?

4. What is the minimum and the optimal size for a PTC to produce high quality teacher training? Given the answer to this question and the one above, how many PTCs should be supported by the Government? Which criteria should be used to select PTCs for closure or conversion to other uses?

5. Aside from training unqualified teachers, what is a feasible and desireable level of spending on in-service training expressed as a percentage of the total salary bill?

6. What is the expected future need for newly recruited teachers given teacher retirement and attrition rates, reduced student-teacher ratios, and growth in primary school enrollments? How may changes in teacher salaries and working conditions affect teacher retirement and attrition?

60 F. INTERNAL EFFICIENCY OF SECONDARY EDUCATION.

Secondary school enrollments are growing rapidly, and the Government’s commitment to universal secondary education [UPPET] indicates this growth will continue indefinitely. At the same time, the unit costs of secondary education are high— both in absolute terms and relative to per capita GDP. The combination of increasing enrollments and high unit costs yield future secondary level expenditures that are not sustainable. Unit costs will need to decrease if UPPET is to be come reality. Improvements in efficiency are thus critical to the success of UPPET. One option is to use distance education for the provision of secondary instruction in remote areas. Box F1 relates the experience of countries in Central America with this technology.

Why are Secondary School Enrollments Growing so Rapidly?

The success of the Uganda’s UPE program in increasing primary school enrollments has led to large increases in enrollments at the secondary level. As shown in Figure F.1, enrollments in the first year of secondary, S1, and total junior secondary enrollments, S1-S4, have increased dramatically in the past five years. The growth reflects a combination of growing numbers of P7 graduates and changes in EMIS data collection procedures to capture a higher percentage of private school enrollments56.

FIGURE F1: UGANDA SECONDARY SCHOOL ENROLMENTS.

S1 and S1-S4 Enrolments

700,000

600,000

500,000 S1 Total S1-S4 400,000

300,000

200,000

100,000

0

56 However, even with these improvements the EMIS does not capture all private schools in its survey, so Figure F1 understates total secondary enrollments. Since accurate secondary school enrollment data are reported in Uganda’s household survey, the EMIS should develop an algorithm to “correct” for non- reporting by private schools.

61 BOX F1. DISTANCE EDUCATION IN CENTRAL AMERICA.

62 The demand for secondary school is likely to increase even more rapidly in future years due to demographic growth, an increase in the transition rates from P6 to P7, and possible improvements in PLE results. Currently there are about 750,000 pupils in P6, of whom about 500,000 enroll in P7, of which 400,000 take the PLE and about 350,000 qualify for S1. In 2006 about 200,000 students entered S1.57 These numbers indicate the potential magnitude of growth in secondary school enrollments, if spaces were made available to all students successfully passing the PLE. Relatively small improvements in P6 to P7 transition rates and in PLE pass rates could easily double or triple secondary school enrollments. Future demographic growth and reduced dropout at the primary level can further increase demand.

FIGURE F2: HOUSEHOLD EXPENDITURE ON SECONDARY EDUCATION AS PERCENT OF TOTAL SECONDARY EDUCATION EXPENDITURE.

M a l i * 10 T a n z a n i a 15 M a u r i t a n i a 20 N i g e r * 23 T o g o 30 B e n i n 30 E t h i o p i a 30 S e n e g a l 31 C a m e r o o n 43 K e n i a * 57 R w a n d a * 59 U g a n d a 59 L e s o t h o 60 M a l a w i 68 C o n g o , D e m . R e p . * 84

0 10 20 30 40 50 60 70 80 90 % private expenditures

*Excludes public capital expenditure Sources: SEIA estimates, Education Sector Reviews (CSRs, 2001-2006), Lewin, 2005, Tanzania PER, World Bank 2003, World Bank, 2007.

At present, as can be seen in Figure F2, demand is tempered by the relatively high “price” of both public and private secondary education to those passing the PLE. The share of total costs paid by households is high in Uganda relative to most other countries of Sub-Saharan Africa.

Even if Government is able to pay the fees for financially needy students, families will still need to pay the other private costs of secondary schooling, which can be considerable, especially when children in rural areas need to pay boarding or

57 Lewin (2007).

63 transportation costs to access schools distant from their homes. If Government wishes to pursue a policy of universal secondary education, it would need to address not only the fees charged by schools but, also, the other private costs which negatively affect the demand for secondary school.58 The potential total costs of expanding secondary education coverage are very high and warrant a detailed examination of how unit costs might be reduced without adversely affecting the quality of instruction.

Are the high unit costs of secondary education sustainable?

TABLE F1: SECONDARY EDUCATION PUBLIC EXPENDITURE AS PERCENT OF GDP IN SSA Country Year % country year % Benin1* 1998 0.98 Niger 2001 0.59 Chad 2003 0.5 Niger 2002 0.58 Cameroon 2001 0.91 Rwanda* 2001 0.59 Cote d'Ivoire 1999 0.98 Senegal 2001 0.63 Kenya* 2003 1.61 Tanzania 2002 0.23 Ethiopia* 2001 0.35 Uganda 2005 0.63 Mali 2004 1.05 Zambia* 2000 0.51 Mozambique 1999 0.2 WEI av. 1999 0.21 OECD av. 1999 0.25

Sources: CSR, Africa Region, World Bank, 2001-2006; Kenya, World Bank 2004; Tanzania, World Bank, 2003; UIS 2002; WB World Development Indicators (GDP p.c.)

The unit costs of secondary education in Uganda, expressed relative to per capita GDP, is 0.63, as is shown in Table F1. This figure is about average for Sub-Saharan Africa, but is much higher than the average for the countries that comprise WEI [World Education Indicators] or OECD, countries which by and large have near universal access to secondary schooling.

Table F2 shows that the unit cost of lower secondary is five times that of primary education, whereas the unit cost of upper secondary is eight times that of primary education. These differences in costs are mainly driven by the ratio of students to teacher salaries relative to GDP per capita. As shown in Table F2 student-teacher ratios are lower at the secondary level, while salary levels are higher. The costs of secondary education in Uganda are high relative to what one finds in countries with high coverage at the secondary level. In countries with high secondary education coverage, the unit cost of secondary education is no more than double that of primary education, and the unit cost of secondary education is no more than 30 percent of GDP.

58 One means of doing this is to introduce cash transfers to poor households contingent on children enrolling in school, as Brazil has done with its Bolsa Escola and Mexico has done with its Progresa program.

64

TABLE F2. SECONDARY EDUCATION UNIT COSTS

Primary Lower Upper Secondary Secondary Pupil Teacher Ratio 50 19 15 Average Teacher Salary /GDP Per 3.8 6.9 9.4 Capita Non Teacher Salary/GDP Per Capita 0.5 2 3 Non-Salary Expenditure/GDP Per 0.5 2 3 Capita Teacher Wage Bill as Percent of 79% 63% 61% Total Recurrent Expenditure Total Unit Cost as Percent of GDP 10% 57% 103% Per Capita Unit Cost in Thousands of UShs 60 300 500 School age pop as % total pop 22% 11% 6% GER Government 100% 12% 2% GER Private 10% 8% 2% GER Total 110% 20% 4% Government Expenditure as Percent 2.27% 0.63% * of GDP Spent on This Level Household Expenditure as Percent of 1.32% 1.88% * GDP Spent on This Level Source: Adapted from Lewin (2007). * Included in Lower Secondary estimate.

The cost of secondary education varies by type of school, as shown in Table F3. Boarding school is almost three times as costly as day school. Schools in urban areas are more costly than those in rural areas. And public schools are more than twice as costly as private schools. In addition, as was shown in Table F2, unit costs also vary by level of schooling, with upper secondary costing almost twice as much as lower secondary.

TABLE F3. RELATIVE UNIT COSTS OF SECONDARY SCHOOLS, 2001.

Type of Boarding Day Rural Peri-Urban Urban Public Private School Unit Cost 2.86 1.00 1.64 1.67 1.94 2.13 0.95 Relative to Day School Source: Unit Cost Study 2001 as reported in Bennell and Sayed (2002).

As noted earlier, currently Government education spending is about 3.5 percent of GDP. As calculated by Lewin (2007), if the Government were successful in attaining 100 percent enrollment at the primary and lower secondary level and 50 percent

65 enrollment at the upper secondary level, it would need to spend closer to 13% of GDP on education, which is clearly not feasible.

The implication of the high ratio of secondary education unit costs to primary education unit costs and the infeasibility of expanding secondary education coverage at such high unit costs is that secondary education must become much more efficient. Secondary schools must begin educating far larger numbers of students with their current budgets, hopefully without reducing the quality of instruction. Thus, it is important to analyze why secondary education is so costly at present.

Why are secondary education unit costs so high?

The reasons for higher unit costs at the secondary level than the primary level are lower pupil teacher ratios, lower teaching loads, and higher teacher salaries. The lower pupil-teacher ratios and lower teaching loads are in turn partly a product of the secondary school curriculum59. The pupil-teacher ratio of 19 in lower secondary [S1-S4] and 15 in upper secondary [S5-S6] is lower than the Africa-wide average of 25 and far lower than the primary school average of 52. Clearly, efficiency could be improved through moderate increases in student-teacher ratios.

Secondary education unit costs are also high because teachers on average teach only 22.5 periods per week out of a fifty period week, according to a 2001 Teacher Utilization Study. The average teacher spends less than 15 hours per week in the classroom and in total works about 29 hours per week. According to the Teacher Utilization Study most under-utilized teachers teach only one subject, and 45 percent of teachers teach only one subject. Teacher productivity could be increased significantly by simplifying and reducing the secondary school curriculum and by requiring all teachers to teach at least two subjects60.

Another part of the story as to why secondary education is so costly is the constraint on classroom space. The MoES deploys teachers to schools based on the number of streams under the assumption that one stream should have no more than 45 students. However, the lack of classroom space may force the school to combine streams or sections in any particular grade and thus reduce the number of needed teachers. The result is that about 34 percent of secondary school teachers are under-utilized.

Finally, small schools may contribute to the high costs of secondary education. The average enrollment in all rural schools only 232, and that of all urban schools is 414.61 About 92 percent of private rural day schools have fewer than 50 students in the 6th form, compared to only 6 percent of public urban day schools. Private rural secondary schools have an average enrollment of only 160, while at 542 students urban boarding schools have the largest average enrollment. While private schools have fewer students, their custom of contracting teachers by the hour helps them control costs, in

59 See Bregman et.al. (2007). 60 The topic of secondary school curriculum reform is treated in depth in Bregman, et.al. (2007). 61 Bennell and Sayed (2002), Table 4.1.

66 contrast to Government schools which hire mainly full-time teachers, even for specialized courses.

What should be done about high teacher salaries?

Currently there are 15 thousand teachers on the Government payroll in Government sponsored schools plus approximately 4 thousand teachers employed by PTAs. In addition, there are almost 19 thousand teachers employed by private schools. As seen in Table F4, secondary school teacher salaries are almost double those of primary school teachers, and these figures do not include the salary supplements often funded from PTA contributions. At 6.9 times per capita GDP, junior secondary school teacher salaries are high by middle income country standards but, as shown in Figure F3, about average by African standards. The salaries of headmasters are especially high. Head teachers of junior secondary [O level] schools are usually at Grade U2 of the public service, which translates to a monthly salary of 921,989 UShs in 2006, which is 2.5 times the average public secondary school teacher salary.

TABLE F4. PUBLIC SECONDARY SCHOOL TEACHER SALARIES, 2006.

Primary Secondary Teachers Teachers Number of Teachers 126,000 15,078 Employed by TSC62 Number of Teachers NA 3,953 Employed by PTA Average Monthly 209,534 361,484 Salary Mode Grade Level U7U U5U Salary at Mode Grade 163,425 361,484 Level Total Wage Bill 279 billion 77 billion Source: Estimates based on EMIS and Unit Cost Survey, November 2006.

Public secondary school teachers are well paid in Uganda in part because they are well-educated relative to the general populace. Thirty percent of secondary school teachers have at least a first graduate degree, and almost all the remaining seventy percent of teachers have secondary school degrees.63 However, this is not the complete explanation since the 3,953 teachers funded exclusively out of PTA funds and the 18,576 teachers in private secondary school receive much lower salaries than government-funded teachers64. Since the wages paid privately-funded teachers [who outnumber the publicly-

62 Teachers in government schools only. In addition, there are 18,576 teachers in private secondary schools. 63 Annual School Census, 2000. 64 A survey of 74 schools in 2006 found the average PTA teacher salary to be 178,492 UShs in public day schools and 130,925 UShs in private schools, both of which are less than half the average government funded teacher salary. The 2002 Teacher Utilization Study found that both PTA teachers and private

67 funded teachers] are market-determined and since the supply of teachers to public secondary schools exceeds the demand, an argument can be made that publicly funded secondary school teachers are paid more than required by the labor market65. .

FIGURE F3. JSE AND SSE TEACHER SALARIES AS MULTIPLES OF PER CAPITA GDP.

14 12.1 12 10.2 9.8 10 9.2 .

c 8.8 . 8.5 p 7.9 P 7.7 D 8 7.3

G 7 / 6.9 6.8 6.8 y r

a 6.1 l 5.7 5.8 a

s 6

' 4.9

s 4.8 r e h c

a 4 e t

2

0

JSE SSE

The high salary levels of Government-funded teachers and headmasters contribute significantly to high unit costs and are not likely to be sustainable as enrollments and employed teachers increase under UPPET. MoPS should consider altering the salary structure for teachers and headmasters who are newly recruited to the public service in order to reduce unit costs. While this would introduce a two-tier salary structure for publicly funded teachers, there is already a de facto two-tier structure with publicly funded and privately funded teachers receiving different pay for similar work.

Is Government Funding for Secondary Education Allocated Efficiently Across Districts?

Government funding for education should result in roughly equal spending per student across schools except for special needs which may require higher spending. These special needs may include low household income (poverty), remote rural location (more expensive to retain teachers, lack economies of scale), and high cost programs (vocational skills, science). A Unit Cost Study carried out in 2001 found large school teachers had salaries ranging from 70 to 100 thousand UShs per month. 65 This argument presumes equal experience and credentials and that private school teachers are not simply public school teachers working a second job. A more in-depth analysis of teachers’ wages would help to determine whether public secondary school teachers are paid an appropriate salary.

68 differences in average spending between types of schools.66 Contrary to what one would expect, the unit cost of rural secondary schools was less than 60 percent that of urban secondary schools.

While data on school budgets and expenditures are not available in the MoES EMIS, teacher salaries absorb a high percentage of school budgets, and a key determinant of a school’s wage bill is the number of teachers it employs. Thus, if expenditures are allocated equally across schools, one would also expect that the ratio of students to teachers would also be allocated approximately equally. Hence, in Figure F4 the distribution of the student/teacher ratio is presented for all the Government secondary schools in Uganda. As shown, the mean student teacher ratio is 17.8 with a standard deviation of 8.5. About 71.3 percent of all schools fall within one standard deviation of the mean.

FIGURE F4 . DISTRIBUTION OF TEACHERS ACROSS SECONDARY SCHOOLS

Government Secondary Schools student/teacher ratio/x -2 s.d. -1 s.d. mean +1 s.d. +2 s.d. 0 5 3 0 0 3 0 s l 5 o 2 o h c 0 s 0

f 2 o

r 0 e 5 b 1 m u 0 N 0 1 0 5 0 0 20 40 60 Student/Teacher Ratio

Source: EMIS 2005

Differences in student-teacher ratio may simply reflect differences in school and population characteristics across districts. Perhaps the distribution within districts is more closely concentrated around the average student-teacher ratio. However, in general, this is not the case. As shown in Figure F5 for one district, there can be large differences in student-teacher ratios even within relatively small geographic areas.

66 Reported by Bennell and Sayed (2002).

69 FIGURE F5. TEACHER DEPLOYMENT ACROSS SCHOOLS IN MBALE DISTRICT.

Government Secondary Schools 0 2 s l o 5 o 1 h c s

f o

r e 0 b 1 m u N 5 0 0 10 20 30 40 50 Student/Teacher Ratio

The fact that the distribution of student teacher ratios is so spread out may simply reflect the fact that the MoPS allocates teachers according to criteria other than simple student enrollment. One such criterion might be the poverty rate of the school or district, since children living in poverty are likely to be more costly to educate. To test this proposition, in Figure F6 the average student teacher ratio of each district is plotted against that district’s poverty level.67 The graph shows no relationship between poverty and student-teacher ratio, the proxy for Government funding.

67 Since information does not exist on the percent of a school’s children who are in poverty, this analysis cannot be carried out at the school level.

70 FIGURE F6. SECONDARY SCHOOL STUDENT TEACHER RATIO AND POVERTY RATES ACROSS DISTRICTS.

Government Secondary Schools 0 3 o i t a R

0 r e 2 h c a e T / t n e d 0 u 1 t S 0

0 10 20 30 40 50 60 70 80 90 100 Percentage of the Poor

Differences in student-teacher ratios across schools might also be justified if those schools having low student-teacher ratios are able to translate that resource advantage into better performance. One proxy for performance is the grade repetition rate. As shown in Figure F7, there is no apparent relationship between student teacher ratios and this proxy for school performance. Another proxy for performance is the ratio of S4 to S1 students, a measure of the academic survival rate. A plot of this proxy against student-teacher ratios also shows no apparent relationship.

71 FIGURE F7. STUDENT TEACHER RATIO AND REPETITION RATE ACROSS SCHOOLS s

e Government Secondary Schools s 0 s 2 a l C

g n i t 5 a 1 e p e R

t n 0 e 1 d u t S

f o

t n 5 e c r e P 0

0 10 20 30 40 50 Student/Teacher Ratio

In short, there is no obvious objective criteria for deploying Government-funded teachers, especially within districts. This can be seen in one final graph in Figure F8, showing the distribution of the student-teacher ratio across schools in Mbale district.

FIGURE F8. TEACHER DEPLOYMENT ACROSS SCHOOLS IN MBALE DISTRICT.

Government Secondary Schools 0 2 s l o 5 o 1 h c s

f o

r e 0 b 1 m u N 5 0 0 10 20 30 40 50 Student/Teacher Ratio

72 While the MoPS undoubtedly has rules for allocating teachers across schools, in practice the distribution of teachers across schools [see Figure 4] and even across districts [see Figure 6] is highly unequal. Furthermore, this distribution appears to be unrelated to the available measures of academic performance. Indeed, it is not obvious that having high student-teacher ratios is worse then low student-teacher ratios in terms of available quality measures. This relationship needs to be analyzed in greater depth with better control measures and better indicators of academic performance. However, if these results hold true, the efficiency of secondary education could be improved significantly by [a] increasing the average student-teacher ratio and [b] allowing schools and districts to deviate from that average only for reasons of special needs, small size due to remote locations, and special programmatic needs.

How can Government sustain the important private sector role in the finance and provision of secondary education?

According to the MoES’s Annual School Census, which understates the number of private schools and private school enrollments, there were 903 private secondary schools in 2005, representing 41 percent of all secondary schools and 37 percent of all secondary school enrollments. Private schools are free to set their own tuition fees, which vary greatly depending on the location and type of school—rural, urban day, and boarding. As shown in Table F4, the average annual fee is about 163,000 UShs. These fees cover the salaries of 19,000 private school teachers, in addition to other personnel and non-personnel recurrent costs.

TABLE F4. HOUSEHOLD EXPENDITURES ON SECONDARY EDUCATION, 2006.

All Public Schools Public Day Schools Private Schools Fees 125,296 72,187 163,120 Total Expenses 213,491 120,255 236,960 Source: Calculated from 2006 UNHS. Note: All public schools includes boarding schools. The average fees paid by all secondary school students is 161,432 UShs.

Public secondary schools are also allowed to charge fees, and while they are required to receive MoES approval for fee increases, such requests are seldom turned down. Revenues from fees are used for a number of recurrent expenditures, including the salaries of PTA contract teachers and salary supplements for government-funded teachers. The average annual fee in public secondary schools is 125,296 UShs. In addition, higher income households spend more on both school fees and total school- related expenditures than do lower income households. Overall, the richest income quintile spends about three times as much on fees and on total school-related expenditures as does the bottom income quintile68.

The role of private sector finance in secondary education in Uganda is thus very important. Just the fees that households pay to public secondary schools amount to about

68 See Annex Tables.

73 47 billion UShs annually. Since most fees are used for contracting and compensating teachers, this is equivalent to a 61 percent increase in the secondary education wage bill funded by Government. As shown in Table F1, households in aggregate spend three times as much on secondary education as does the Government. While Government spends 0.63 percent of GDP on secondary education, households spent an additional 1.88 percent of GDP. While this appears to show that Government is successfully leveraging private sector financing, it is more accurate to say that private sector finance comes about due to the lack of Government funding and provision.

The challenge facing the Government as it pursues Universal Post Primary Education and Training [UPPET] is how to expand publicly-subsidized secondary education [a] without reducing the demand for fee-paid private secondary education and [b] without reducing household contributions to public sector secondary schools. Government needs to think carefully about how to use its funding to create incentives for private supply and finance while proving new opportunities to households which cannot afford to pay the levels of fees charged currently. This will require creation of some form of transparent, explicit capitation grant to private institutions with the amount of the grant conditional on school location, student income, program cost, etc.69 Korea is often cited as a model where private finance and provision played a key role in educational development (See Box F1).

Is there strong accountability for academic and financial performance of public secondary schools?

In 2002 the Government created the Education Standards Agency [ESA] with the objective of strengthening inspection, especially of secondary schools. At the time experts agreed that an inspector:school ratio of 25:1 would be necessary to allow inspectors the time to observe teachers in their classes, to review budgets and expenditures, and to advise Boards of Governors [BOG]. With almost 2,000 secondary schools, this would require a staff of about 80 inspectors. At present the ESA has a total staff of about 50 to cover all secondary schools, tertiary institutions, and teacher training colleges. Not only is the inspection service seriously understaffed, but it also lacks the means of transport to regularly visit schools. In addition to ESA inspectors, the districts have their own inspectors, who don’t have adequate capacity to even inspect the primary schools, and the CCTs provide advisory and training services to the schools, but they don’t do inspection.

While Government inspection is deficient, secondary schools have two local governance institutions which give the community a role in monitoring and holding the school accountable. These are the BOG and the PTA. Unfortunately, the distinct roles of these two institutions are not always clear, and the headmaster of the secondary school is often viewed as untouchable. Still, the fact that parents contribute so significantly to the funding of both Government and private schools gives them a potentially strong voice in school governance.

69 See LaRocque (2007) for further elaboration of such proposals.

74 It seems unlikely that in the near future the Government will give ESA the financial and human resource it requires to be an effective inspection agency. In the absence of an effective inspection service for secondary schools, there is a need to strengthen the local governance institutions by training their members, clarifying their roles, giving them increased responsibility and authority, and providing them with the information they need to monitor the performance of the schools they govern. If there role in accountability is to be strengthened, the BOGs and PTAs must extend their current fund-raising function to a broader one of oversight.

BOX F2. PRIVATE SECTOR ROLE IN EXPANDING ACCESS IN KOREA

Source: World Bank (2005)

75 H. INTERNAL EFFICIENCY OF TERTIARY EDUCATION

A recent major study of tertiary education was jointly carried out by the MoES Department of Higher Education and the World Bank, resulting in several publications.70 The Uganda National Council for Higher Education has followed up this study with annual reports on higher education.71 Both the Bank and the NCHE reports analyze the efficiency of public expenditure on higher education and make a number of policy recommendations for improvements in higher education finance and efficiency of resource use. Given this recent, rich information and analysis, the current study mainly updates statistics and organizes the findings of previous studies to more explicitly respond to the policy questions being addressed in this report.

As of 2005, the tertiary (higher education) sub-sector in Uganda enrolled over 124,000 students in 157 institutions ranging from universities to specialized training institutions. This enrollment figure corresponds to 3.8 percent of the 19-25 year old cohort, about the average for Sub-Saharan Africa, and represents a rapid growth in enrollments since the year 2000. Table H.1 summarizes the distribution of enrollments across institutional types.

TABLE H.1. TERTIARY LEVEL ENROLLMENTS, 2005

Institutional Type MoES Budget Enrollments Share of Home Total Universities & Affiliated Higher 78,107 62.8% Colleges Education* National Teachers Colleges Teacher 12,096 9.7% [NTCs] Development Colleges of Commerce BTVET 14,479 11.6% Management Institutions BTVET 9,411 7.5% Other Technical BTVET 10,220 8.2% Source: NCHE (2006) * Public universities have their own budget lines.

Higher education enrollments in Uganda appear to be increasing rapidly. As UPE translates into higher numbers of primary school graduates and UPPET brings about larger numbers of secondary school graduates, the demand for higher education can be expected to increase faster still. It is important to plan for this rapid growth, rather than having to react in an ad hoc fashion it in future years. In addition, it is important to make specific plans and projections concerning the growth of publicly managed universities and non-university institutions [BTVET and NCTs] and the concomitant increase in public expenditures on higher education.

Allocation Across Tertiary Institutions

70 See World Bank (2004). 71 See NCHE (2005).

76 Tertiary institutions have their budgetary homes in three different MoES departments-- Higher Education, Teacher Development, and BTVET, and the major public universities have their own budget lines. While the Higher Education Department budget is for public universities, the budgets for Teacher Development and BTVET include expenditures at the primary and secondary levels as well as the tertiary level. This arrangement makes it difficult to accurately track total tertiary expenditures over time. For 2004 total public spending on all types of higher education represented about 15 percent of the MoES budget, and the distribution of that spending was 79% allocated to universities, 30% allocated to BTVET institutions, and 8% allocated to NTCs.72 The university recurrent expenditure for 2005/06 was UShs. 80 bn., or 15 percent of total MoES recurrent expenditure.73

Is the budgetary allocation across higher education programs appropriate?

The economic rationale for the public funding of higher education has several elements. The first one is equity, or to ensure that children from lower income households are not excluded from higher education for reasons of affordability. This rationale translates into government policies to provide financial aid or subsidies targeted to needy students. The second element is to ensure the nation has the type of highly skilled labor required for innovation and economic growth. This rationale often translates into subsidies for particular skill areas, such as science or engineering. The third element is to maximize the nation’s return on its higher education investment. When this return is measured in pecuniary terms, the public policy response is to ensure an adequate supply of that instruction which yields the highest rate of return, or the highest income relative to the costs of instruction. Finally, the fourth element is to ensure an adequate production of new knowledge, which is a pure public good that merits a high level of subsidy. The rationale translates into government funding of university research projects and funding to develop the capacity to carry out such research.

To answer the question about the allocation of the higher education budget across institutional types, one must at a minimum know how higher education fits in the country’s development plans and know the pecuniary returns to different types of instruction. Although there is evidence that the overall social and private rates of return to higher education are quite high, limited information exists to determine returns to specific types of instruction.74 Table H.2 shows the unit costs and subsidies for several degree programs, but there is no corresponding information on earnings by career program.

72 Calculated from World Bank (2004), Table 29. 73 Calculated from Table 2.1 of the MoES ESSAPR for 2005/06. (Makerere University + Mbarara University + Kyambogo University + Makarere University Business School + Gulu University + District Tertiary Institutions) 74 As shown by Appleton (2001), the social and private returns to higher education appear to be relatively high and to be increasing over time.

77 TABLE H.2. UNIT COSTS AND SUBSIDIES OF DEGREE AND DIPLOMA PROGRAMS (in US$)

Program Unit Cost Annual Fees Absolute Percent of an Subsidy Subsidy Enrolled Student Degree Programs Medicine (MUST) 4,588 1,535 3,053 66% Basic Science 1,971 773 1,198 61% Business/Commerce 1,278 1,085 193 15% Fine Art 1,434 800 634 44% Diploma Programs Education (science) 418 560 -142 -34% Technical 647 333 314 49% Health Professional 1,544 836 708 46% Hotel and Tourism 4,994 214 4780 96% Source: NCHE (2006)

Given the lack of information on returns, one can still ask whether the pattern of subsidies reported in Table H.2 appears to make sense. The highest income profession— medicine—has one of the highest subsidies, both in absolute and percentage terms. The most costly degree program—hotel and tourism—has the lowest fees and highest subsidy levels. And one program where it could be argued the Government may wish to encourage enrollments—science education—has a negative subsidy. On the other hand, business programs are only lightly subsidized, which could be argued is appropriate, and basic science is quite highly subsidized, which again could be argued is appropriate. On the face of it, there is plenty of room to improve the allocation of public subsidies across degree and diploma programs by raising fees and reducing subsidies for those programs which are especially costly and those which generate high incomes for graduates and to increase subsidies for those programs which arguably have high social benefits but relatively low incomes.

In terms of equity, one would hope that higher income students would receive lower subsidies than lower income students. In general, university students come from households with higher average incomes [Ush 3.1m] than do students in all tertiary programs [Ush 1.5m] and young people who never enroll in higher education [Ush 0.4m].75 An expenditure allocation that would improve equity is one that would give lower subsidies to degree than diploma students. Excluding hotel and tourism, that does not appear to be the case.

Do higher education financing policies promote internal efficiency?

75 Calculated from the 1999/2000 Integrated Household Survey as reported in World Bank (2004).

78 Another way of answering the question whether the allocation of public higher education expenditures is appropriate is to analyze whether the mechanisms used to fund higher education programs promote either efficiency or equity. Do funding mechanisms provide incentives for universities and other institutions to operate efficiently, or do they pose obstacles to the wise use of resources?

The public funding mechanism varies by institutional type. For the recurrent budget, public universities receive a subvention per government-funded student, and the Government sets both the subvention amount and the number of students to be funded. This is equivalent to a block grant to the university and provides no specific efficiency incentives. On the other hand, the equity implications of this grant are almost certainly negative as it is the students from advantaged homes and advantaged secondary schools who are most likely to obtain the scores that “win” a scholarship. Furthermore, since Government-funded students’ attendance is assured, the universities have incentives to play such students in academic disciplines or career tracks which face low demand from fee-paying students. While this may appear efficient from the university’s perspective, it almost certainly is not from society’s perspective.

Public post-secondary institutions managed by the BTVET directorate do not receive a block grant. Rather, they receive funding based on the inputs authorized by the MoES and MOF. Input-based funding always carries with it rigidities in terms of how funds can be spent, and in practice in Uganda is unrelated to enrollments or outputs.76 This funding mechanism provides no incentives to institutions to manage their resources efficiently and often results in non-personnel inputs being underfunded.

Government funding of public higher education in Uganda should be reformed to [a] directly tie funding to the total number of students enrolled in an institution, [b] adjust per student capitations to account for program cost differences, [c] provide incentives to institutions to generate their own revenues, [d] reward institutions which graduate students on time, and [e] provide additional per pupil funding for program areas of high national priority.

Do higher education financing policies leverage public monies?

Two of the basic higher education policy decisions faced by governments are [a] what percent of costs in public institutions should be covered by fees charged to students and [b] what incentives should be provided to increase the supply of places offered by private institutions? The answers to these questions determine the extent to which Government funding leverages private sector funding. In general, public institutions should be given the autonomy to set their own fees, but Government can influence the fees charged different types of students or different types of instruction through selective subsidies. The trickier question is what emphasis should be given to expanding enrollment vs. ensuring quality of instruction and research.

76 Kasozi (2003) finds no correlation between public subventions received by institutions and total student enrollment.

79 By simultaneously restricting public subsidies [e.g., the number and per student subsidy of sponsored university students], allowing institutions to set student fee levels, and facilitating the entry of privately managed institutions, higher education finance in Uganda encourages private finance, at least at the university level.77 As shown earlier in Table A 2, private finance is almost as important as public finance at the tertiary level. One result of this policy is that the public subsidy per university student in Uganda [e.g., US$ 459 at Makerere University and US$ 591 at Kyambogo University] is lower as a ratio to GDP per capita [less than 2.0] than is true for Sub-Saharan Africa as a whole, where the ratio is over 4.0.78 The fees paid by students in public universities is approximately equal to the recurrent unit cost of educating those students, but the same is not true of the non-university or “other tertiary institutions”. Of course, encouraging private finance means more simply charging fees, as illustrated by the case in Box H1.

BOX H1. DAR ES SALAAM MANAGEMENT REFORM.

Source: Mkude (2001) and World Bank (2002)

Despite having an enlightened policy with respect to university fees, there are no explicit incentives to stimulate private finance and provision. While the number of private universities has grown in recent years, especially among universities and colleges

77 This is true despite occasional political interference, such as Parliament’s rejection of Makerere’s fee increase in 2005. Institutional autonomy is in principle ensured by the Universities and Other Tertiary Institutions Act of 2001. 78 World Bank (2004), Table 35. Unit costs are also low relative to estimates of what they should be to provide higher education of adequate quality. See Kasozi et.al. (2002) and a recent mimeo update of Kasozi’s work for evidence.

80 of commerce and management, they enroll a disproportionately small proportion of total enrollment. For example, private universities represent 81 percent of all universities but only 29 percent of total enrollments. Selective subsidies to private institutions, especially for development expenditures and for children from poor households or for skill areas of high national priority, could be a cost-effective way of rapidly increasing enrollment rates. While there has been rapid growth of private provision of university education and business and commerce education in Uganda, the same appears to not be true for some areas in the non-university [“other tertiary institutions”] sector. Thus, perhaps it is with the more trade-oriented schools where the Government should experiment with selective subsidies to encourage private provision and finance. At the university level, Government currently sponsors and fully pays for a fixed number of students selected on the basis of merit. Changing this policy to provide partial subsidies rather than full subsidies to meritorious students attending public institutions would further leverage private finance, especially if the magnitude of the subsidies were made contingent on household income79.

Do higher education financing policies ensure adequate quality?

By allowing public universities to charge fees to non-government sponsored students, Uganda has ensured that prestigious universities like Makerere have been able to maintain a reasonable level of quality. The alternative policy—prohibiting tuition fees while expanding enrollments with fixed government budgets—would have led to rapid declines in quality, a policy and consequence that is commonly found in some Latin American countries. Despite its policy regarding the setting of fees, the Government’s policy does not necessarily ensure an adequate level of quality of instruction and research.

Relative to per capita incomes, universities in developing countries have expensive cost structures. The market for university professors is increasingly global in nature, meaning that universities in poor countries must pay competitive salaries, which are high relative to per capita incomes, in order to retain their best and brightest faculty members. For this reason, the unit cost of higher education for countries in Sub-Saharan Africa is a much higher percentage of GDP per capita [422%] than it is for richer countries in Europe and Central Asia [36%] or even South Asia [74%].80 At the same time, low per capita incomes serve to constrain what even high income households can pay in terms of university fees. And as access to secondary education in Uganda opens up to lower income households, the average household income of secondary school graduates will decrease, further constraining what universities can charge in tuition fees. In short, universities are constrained in terms of their capacity to generate private financing, meaning that if Uganda wishes to maintain and increase the quality of universities like Makerere and Kyambogo, Government will in the end need to provide adequate levels of funding. While it is not fiscally possible to fund high quality for all tertiary level institutions, it may be feasible to do so for a very limited number. 79 See Johnstone (2004) for examples in Africa of cost-sharing in higher education. Also, the World Bank [2004] notes that a disproportionate number of sponsored students come from the highest income strata of families in Uganda and have the capacity to pay full fees. 80 World Bank (2004), Table 35.

81 Actions that Would Improve Efficiency in Government Tertiary Education Expenditures.

The 2005 NCHE report The State of Higher Education and Training in Uganda provides detailed recommendations for improving efficiency in the use of resources in tertiary education81. Adoption of the following actions are consistent with those recommendations:

 Develop an explicit strategy to accommodate growth in tertiary enrollments over the next decade, including options for funding both capital and recurrent expenditures and specifying the role of the private sector in funding and provision.

 Reform government funding of universities to eliminate full-funding of government-sponsored students and replace it with capitation grants and student loans targeted on financially needy students and on programs of high national priority.

 Substitute input-based funding of non-university [BTVET and NTCs], other tertiary education with capitation grants and institutional and managerial autonomy [and accountability] in the use of funds.

 Improve information systems on tertiary education to regularly report unit costs by program and institution, and track program graduates to monitor the impact of tertiary programs on earnings and employment.

 Stimulate the private provision of tertiary education, and especially the BTVET institutions, by facilitating access by both publicly and privately managed institutions to lines of credit for development expenditures aimed at increasing capacity or quality82.

81 Further, detailed recommendations are reported in the World Bank (2004) report on tertiary education. 82 One example might be the schemes being developed with the assistance of the International Finance Corporation {IFC] in Ghana and Kenya to facilitate capital expenditures at the secondary level.

82 NEXT STEPS

This efficiency analysis has covered a large number of issues across several sub- sectors. Still, it has not been possible to treat all sub-sectors in equal depth, and there is a need to complete the analysis by undertaking additional survey and analytic work on, especially, BTVET and teacher education. These are sub-sectors where a significant amount of data collection and survey work will need to be undertaken prior to doing the analysis itself.

In addition, this study has proffered several recommendations for policies and programs to address some of the efficiency problems and issues that have been identified. However, it has not been possible to systematically assess the costs, benefits, and administrative and political feasibility of adopting and implementing these recommendations. This kind of careful evaluation is required before setting priorities for action and before converting the recommendations into programs that can be implemented.

HIGH PRIORITY ISSUES

For the reasons given above, it is not possible to rank order the various recommendations made in this report. However, in the absence of detailed financial information, it is still possible to identify recommendations where the magnitude of benefits relative to costs should be highly favorable. These are listed below by sub- sector.

Primary Education.

Reduce Headmaster Absenteeism. Headmasters, who are well paid relative to teachers, have double the absenteeism rate of teachers. The frequent absence of headmasters both sets a negative role model for teachers and constrains schools in their attempts to implement quality-enhancing reforms. Thus, the potential benefits of reducing headmaster absenteeism are large. There are costs associated with monitoring absenteeism, but simply improving transparency on absenteeism rates [possibly by involving PTAs and SMCs] may be effective.

Increase Teacher Classroom Time. Teachers are also excessively absent from school, and even when at school often absent from the classroom, leading to insufficient student-teacher contact time and limited learning. The potential benefits of reduced teacher absenteeism are to some extent conditional on changes in teaching practices, but attempts to improve learning are not likely to bear fruit so long as teachers fail to teach. Headmasters are directly responsible for monitoring teacher attendance—both at school and in the classroom—and they should be held accountable for performing this function.

Improve Accountability Arrangements. The weakness of the district school inspection system limits the Ministry’s and districts’ ability to hold schools accountable

83 for performance. Given the cost and difficulty of improving inspection, local governance should be strengthened by [a] giving PTAs and SMCs additional financial and monitoring responsibilities and [b] providing them with annual school report cards. This will require strengthening the capacity of the CC’s to provide assistance and training to PTAs and SMCs and will require reorienting the Ministry’s EMIS to the delivery of district and school level report cards on financial and academic performance.

Rationalize Teacher Deployment. Large disparities in class size and student teacher ratios across schools are neither equitable nor efficient. Staffing guidelines and teacher deployment practices need to be revised to reduce disparities, perhaps by giving schools the authority to approve or disapprove teacher transfers. This action may be politically difficult, but its administrative cost is minimal.

Prioritize Grades 1-3. Students who fail to enter school on time and who fail to achieve mastery over reading by the end of grade 3 are candidates for repetition and dropout. For reasons of equity and efficiency, greater priority should be given to grades 1-3. Average class sizes should be smaller in grades 1-3 than grades 4-7. At least as much emphasis should be put on what percent of P 3 students exhibit mastery of reading as on what percent of P7 students pass the PLE. Incentives to strengthen learning in grades 1-3 could include linking the headmaster’s annual evaluation to improvements in P3 indicators.

Secondary Education.

Reduce Teaching Costs. Teaching costs at the secondary level should be reduced by increasing the student-teacher ratio by simplifying the curriculum and rationalizing teacher deployment. Double-shifting should also be seriously considered as an option as the demand for secondary increases under UPPET.

Ensure Quality. As the demand for secondary school places continues to grow in the face of tightly constrained government spending, there is the risk of lower standards. MoES will need to be able to quickly identify quality related problems that may result from double-shifting or other cost-reducing measures. In the face of this risk, the ESA’s already stretched capacity to monitor quality is critical, and it will be necessary to significantly increase its capacity with the aim of giving feedback to the schools and to the MoES for the purpose of quickly identifying and correcting quality- related problems.

Facilitate Privately Funded Expansion83. Even with more effective and efficient use of teachers, it is not feasible for the Government to fully fund a massive expansion of secondary education. Government will need to continue to rely on the private sector to both provide and to finance secondary education, and it is important that it develop a strategy to leverage public funds by facilitating and stimulating private finance and provision. Among other things, it should consider working with the private

83 See LaRocque (2006) for an in-depth discussion of how Uganda could foster public-private partnerships in education.

84 capital market to ensure that secondary schools can borrow at reasonable cost for the purpose of expanding their operations.

Tertiary Education.

Reform Government Finance. Current Government funding of “sponsored students” is neither equitable nor efficient and should be replaced by financial aid targeted to qualified students who could not otherwise attend a higher education institution and by capitation grant funding that reflects national interests. At the same time, the line-item, input-oriented budgeting of some tertiary level institutions should be replaced by formula-funding capitation grants along with giving managers greater resource allocation autonomy along with responsibility for results.

Stimulate Privately-Financed Supply. Government should work with the private capital market to develop mechanisms that give fee-charging tertiary institutions —both public and private--access to private funding to develop infrastructure and expand supply. It should also stimulate the supply of privately managed institutions by providing an appropriate regulatory framework and selective subsidies.

MoES.

Strengthen Information and Analysis Role. The current EMIS does a good job of reporting student enrollment and teacher employment data. Its scope needs to be increased—either on a sample or census basis--to collect and report financial and expenditure information to provide continuous monitoring of financial indicators for decision making. Furthermore, this information should be linked to school performance results such as the PLE, SLE, and UNEB tests. The resulting data should be reported back to the districts and schools in the form of report cards. In addition, the MoES should strengthen its capacity to analyze data, possibly locating this function in an expanded EMIS unit.

Carry out Selected Studies. The basic cost and output information required for analyzing efficiency is missing for some sub-sectors, such as BTVET and Primary Teacher Colleges and school construction. Important policy questions—such as whether BTVET should be expanded along with general secondary education, or whether some of the PTCs should be consolidated—cannot be answered in the absence of information and analysis. The MoES should commission special surveys and studies covering these areas.

85 J. REFERENCES

Akyeampong K, Furlong D & Lewin K M, (2000) The Costs and Financing of Teacher Education in Ghana. MUSTER Discussion Paper No 18, Centre for International Education, University of Sussex.

ARD, Inc. (2006). UPE Capitation Grant Tracking Study, FY 2005-2006. USAID/Uganda, Draft Report.

Bennell, Paul and Yusuf Sayed (2002) Improving the Management and Internal Efficiency of Post-Primary Education and Training in Uganda. Draft.

Bregman, Jaap, Andrew Clegg and Wout Ottevanger (2007). Uganda Secondary Education Curriculum, Assessment and Examination (CURASSE): Roadmap for Reform. Draft paper dated August 2007. The World Bank, Kampala.

Development Cooperation Ireland (2004). Critical Characteristics of Effective Primary Education in the Rwenzori Region of Uganda. Kampala.

Eilor, Joseph (2005). Impact of Primary Education Reform Program on the Quality of Basic Education in Uganda. Association for the Development of Education in Africa.

Habyarimana, James (2004). Measuring and Understanding Teacher Absence in Uganda. Draft. Harvard University.

Habyarimana, James (2007). Characterizing Teacher Absence in Uganda: Evidence from 2006 Unit Cost Study. Draft report.

Hanushek, Eric A. and Ludger Wossmann (2007). The Role of Education Quality in Economic Growth. World Bank Policy Research Working Paper 4122. The World Bank.

Hicks, R.B. (2005). Baseline Study: Learning and Teaching Strategies and Outcomes in Four Districts. Draft Report.

Higgins, Cathal, et.al. (2006). Multigrade Teaching in Uganda: A Case Study of the Kalangala Multigrade Pilot.

International Development Consultants (2006). A Tracking Study on the Allocation and Utilization of Resources in Conflict Districts in Northern Uganda in Primary Education Sub-Sector. Ministry of Education and Sports, Republic of Uganda, Kampala.

Johnstone, D. Bruce (2004). Higher Education Finance and Accessibility: Tuition Fees and Student loans in Sub-Saharan Africa. Journal of Higher Education in Africa.

86 Kasente, Deborah [Leader, Project Research Team] (2006). Increasing Retention Through Improved Literacy and Learner-Friendly Primary Schools in Uganda. Makerere Institute of Social Research, Makerere University, Kampala.

Kasozi, A.B.K. (2003) Draft Issues Paper for Writing a Strategic Plan for Higher Education. NCHE. Kasozi, A.B.K. and N.B. Musisi (2002). Decentralization and Tertiary Institutions of Learning in Uganda. Makerere Institute of Social Research, Kampala.

Kunje, Demis and Keith Lewin (2000). The Costs and Financing of Teacher Education in Malawi. MUSTER Discussion Paper No. 2. Centre for International Education. University of Sussex.

LaRocque, Norman (2006) Options for Increasing the Scope of Public-Private Partnerships at the Universal Post-Primary Education and Training Level in Uganda. MoES.

Lewin, Keith (2007) Financing Universal Post Primary Education and Training (UPPET) in Uganda. MoES.

Lewin, Keith (2006). Mapping the Missing Link: Planning and Financing Secondary Education Development in Sub Saharan Africa. Second Regional Conference on Secondary Education in Africa.

McEwan, Patrick and Lucrecia Santibañez. 2004. “Teacher and Principal Incentives in Mexico. Manuscript, Washington: World Bank.

Mkude, D.J. (2001). Reforming Higher Education: Change and Innovation in Finance and Administration: A Case Study of the University of Dar es Salaam. Washington, DC: The World Bank.

MoES (2006). Sector Annual Performance Report (ESSAPR). Financial Year 2005/06.

MoES (2007). Needs Assesment Draft Report for the P.7 Enrolling Institutions in Uganda. Business Technical Vocational & Education Training Section.

MoES (1996). Primary Teachers’ Colleges Cost Effectiveness and Efficiency Study. Final Report. Education Planning Department.

MoES (1999). A Three Year Primary Teacher Development and Management Plan, 2000/2001 – 2002/2003. The Teacher Education Department.

MoES (2006). Uganda Education Statistics Abstract, 2004.

MoES (2004). Education Sector Strategic Plan 2004-2015. Education Planning Department.

87 MoES (2003). The Quality of Education: Some Policy Suggestions Based on a Survey of Primary (SACMEQ) Schools in 2000. Kampala.

Morley, S., and D. Coady (2003). From Social Assistance to Social Development: A Review of Targetededucatoin Subsidies in Developing Countries. Center for Global Development and International Food Policy Research Institute, Washington, DC.

Musisi, Nakanyike B., D. Kasente, and A.M. Balihuta (2006). Attendance Patterns and Causes of Dropout in Primary Schools in Uganda. Makerere Institute of Social Research, Makerere University, Kampala.

Nannyonjo, Harriet (2006). An Analysis of Factors Influencing Learning Achievement in Primary Six. The World Bank.

National Council for Higher Education (NCHE) (2006). The State of Higher Education and Training in Uganda, 2005. www.unche.or.ug

National Council for Higher Education (NCHE) (2005). The State of Higher Education and Training in Uganda, 2004.

PFK Consulting Ltd. (2003). SFG Programme: Value for Money Audit. Draft Report.

Pillay, Pundy and Ibrahim Kasirye (2006). Efficiency in the Education Sector in Uganda. Draft.

PricewaterhouseCoopers (2003). Final Report on the Study of Cost Efficiency and Effectiveness in Human Resources Deployment in Social Sectors. Ministry of Public Service.

Reinikka, Ritva, and Jakob Svensson (2002). “Fighting Corruption to Improve Schooling: Evidence from a Newspaper Campaign in Uganda,”

Shinyekwa, Isaac (2006) Report on Universal Post Primary Education and Training: An Analysis of Schools, Budgets, and Expenditure. MoES.

Smith, Ian (2006). A Study on Deployment, Utilisation & Management of Secondary Education Teachers Under UPPET. MoES.

Sugrue, Ciaran, Nansozi Muwanga, Zerabubel Ojoo (2003). Primary Teacher Development and Management Plan. Funded by Irish Aid and prepared for MoES.

USAID (2006). UPE Capitation Grant Tracking Study, FY 2005-06. Kampala.

Vegas, Emiliana (Ed.). 2005. Incentives to Improve Teaching: Lessons from Latin America. Washington, DC: The World Bank

88 Villegas-Reimers Eleonora. 1998. “The Preparation of Teachers in Latin America: Challenges and Trends”. Latin America and the Caribbean Region Human Development Department, Paper 15. Washington DC: The World Bank

Ward, Michael, Alan Penny, and Tony Read (2006). Education Reform in Uganda – 1997 to 2004. Department for International Development.

World Bank (2002). Constructing Knowledge Societies: New Challenges for Tertiary Education. Washington DC: The World Bank.

World Bank (2004a). Uganda Tertiary Education Sector Report. Africa Region Human Development Working Paper Series No. 50.

World Bank (2004b). Central America: Education Strategy Paper. Report No. 29946. Latin America and Caribbean Region, World Bank.

World Bank (2004c).

World Bank (2005). Expanding Opportunities and Building Competencies for Young People: A New Agenda for Secondary Education. Washington DC: The World Bank.

World Bank (2007a). Teachers in Uganda. Draft report dated June 2007.

World Bank (2007b). Uganda: Fiscal Policy for Growth. Public Expenditure Review 2007. Part II: Main Report. Report No. 40161-UG. Africa Region, World Bank.

World Bank and Ministry of Education, Cambodia (2007c). Teaching in Cambodia. East Asia and Pacific Region, World Bank.

89 GOVERNMENT BUDGET FOR EDUCATION IN 2005/06 {MTEF CLASSIFICATION (in Ush bn.) Recurrent expenditure Development expenditure Total Total Domestically- Donor- expendit recurrent Wage Non-Wage financed financed ure Uganda Management Institute 0.4 0.0 0.4 0.0 0.0 0.4 Education and Sports (incl Prim Educ) 46.9 7.3 39.6 19.3 39.5 105.7 Makerere University 33.4 0.0 33.4 0.1 17.6 51.1 Mbarara University 6.4 4.0 2.4 0.4 0.0 6.8 Kyambogo University 11.0 6.1 4.9 0.3 0.0 11.3 Education Service Commission 2.2 0.5 1.8 0.1 0.0 2.3 Makerere University Business School 4.2 0.0 4.2 0.0 0.0 4.2 Gulu University 3.2 0.0 3.2 1.2 0.0 4.4 District Primary Educ incl SFG 287.5 254.0 33.5 51.0 0.0 338.5 District Secondary Education 82.8 76.3 6.5 0.0 0.0 82.8 District Tertiary Institutions 24.0 15.7 8.3 0.0 0.0 24.0 District Health Training Schools 4.2 2.5 1.8 0.0 0.0 4.2 506.3 366.4 139.9 72.3 72.3 650.9 Source: MOFPED, MTEF tables GOVERNMENT EXPENDITURE ON EDUCATION

Education (Ush, billion) 98/99 99/00 00/01 01/02 02/03 03/04 04/05 05/06 Total education expenditure (MTEF number) 326 356 415 512 561 587 638 656 Recurrent 241 258 286 360 396 436 484 538 Wages 158 153 180 238 274 310 358 396 Non-wages 83 106 106 122 123 127 126 142 Development 85 98 129 152 164 151 154 118 Domestic 36 67 87 96 95 81 88 69 Donor 49 31 42 56 69 70 65 49

Primary (WB estimate based on MTEF) 205 237 285 349 366 383 407 415 Recurrent 152 172 196 245 259 285 309 334 Wages 99 101 124 162 179 202 248 276 Non-wages 52 70 73 83 80 83 61 58 Development 53 65 88 104 107 98 98 81 Domestic 22 44 59 66 62 53 56 52 Donor 31 21 29 38 45 46 42 29

Secondary (WB estimate based on MTEF) 72 71 77 98 110 114 126 109 Recurrent 53 51 53 69 78 84 96 100 Wages 35 30 34 46 54 60 71 88 Non-wages 18 21 20 23 24 25 25 12 Development 19 20 24 29 32 29 30 9 Domestic 8 13 16 18 19 16 17 4 Donor 11 6 8 11 14 14 13 5

Tertiary (WB estimate based on MTEF) 49 48 53 64 85 91 104 131 Recurrent 36 35 37 45 60 67 79 104 Wages 24 21 23 30 41 48 39 32 Non-wages 12 14 14 15 19 20 40 72 Development 13 13 17 19 25 23 25 27 Domestic 5 9 11 12 14 12 14 13 Donor 7 4 5 7 10 11 11 15 Source: World Bank attempt to reclassify total government spending on education into private, secondary and “others”

90 NON-WAGE EXPENDITURES PER STUDENT IN PUBLIC AND PRIVATE SCHOOLS Mean spending per student (Ush) Non- teacher pay Total non- and wage School teacher expenditure materials Repairs Utility bills lunch No. schools Government 4,728 3,120 696 111 801 130 Private 30,965 20,756 3,399 2,348 4,461 19 Government aided (private and NGO) 6,149 4,127 1,084 217 722 7 Total 7,987 5,314 1,042 388 1,243 156

91

GOVERNMENT SCHOOLS: MEAN AND MEDIAN PER STUDENT HOUSEHOLD EXPENSES ON EDUCATION, BY URBAN/RURAL AND EXPENSE ITEM PUBLIC SCHOOLS PRIVATE SCHOOLS

UNIFORMS TOTAL UNIFORMS TOTAL AND SPORTS BOOKS AND BOARDING SCHOOL SCHOOL AND SPORTS BOOKS AND BOARDING SCHOOL SCHOOL FEES CLOTHES SUPPLIES FEES OTHER EXPENSES FEES CLOTHES SUPPLIES FEES OTHER EXPENSES MEANS PRIMARY urban 32,182 7,784 9,603 1,701 14,751 66,107 161,775 13,331 18,046 19,501 27,117 233,601 rural 6,701 4,102 5,218 993 3,831 20,326 71,223 6,777 8,824 10,445 16,322 112,451 Total 9,006 4,435 5,616 1,055 4,823 24,568 98,066 8,745 11,592 13,075 19,602 149,643 SECONDARY urban 354,126 20,348 44,812 45,846 82,794 487,519 331,663 22,270 48,281 82,116 57,949 545,547 rural 247,221 18,043 32,334 42,349 55,818 363,633 224,759 16,961 31,545 42,145 55,840 344,247 Total 270,123 18,522 35,053 43,053 61,821 391,512 259,336 18,705 37,040 55,179 56,530 413,703 MEDIANS PRIMARY urban 7,000 5,500 5,200 - 1,000 26,000 114,000 10,000 11,000 - 5,000 155,800 rural - 4,000 3,600 - 600 11,500 33,500 5,000 5,000 - 2,100 54,000 Total - 4,000 3,600 - 700 12,200 49,200 6,000 6,000 - 3,000 79,000 SECONDARY urban 270,000 15,000 26,000 - 22,000 405,000 230,000 20,000 28,000 - 20,000 362,000 rural 165,000 15,000 20,000 - 16,000 273,000 150,000 15,000 21,500 - 15,000 227,500 Total 183,000 15,000 20,500 - 20,000 291,000 170,000 15,000 24,000 - 18,000 283,800

92 TREND IN HOUSEHOLD PRIMARY SCHOOL EXPENDITURE: MEAN AND MEDIAN PER STUDENT ANNUAL HOUSEHOLD SCHOOL EXPENSE BY ITEM*. Uganda Household Surveys 2002 and 2006. In Ugandan Shillings. 2002 2006, constant 2002 prices 2006, current prices SCHOOL SCHOOL SCHOOL FEES AND BOOKS TOTAL FEES AND BOOKS TOTAL FEES AND BOOKS SCHOOL BOARDING/ AND SCHOOL SCHOOL BOARDING/L AND SCHOOL SCHOOL BOARDING/ AND FEES LODGING UNIFORM SUPPLIES EXPENSES FEES ODGING UNIFORM SUPPLIES EXPENSES FEES LODGING UNIFORM SUPPLIES MEANS ALL SCHOOLS ALL STUDENTS 8,273 8,720 2,909 2,547 15,658 11,893 12,511 3,447 4,182 23,776 14,254 14,994 4,132 5,012 URBAN 39,509 41,977 4,632 4,480 54,341 52,254 54,553 6,538 8,011 80,211 62,625 65,380 7,836 9,601 RURAL 5,134 5,378 2,735 2,353 11,770 7,227 7,650 3,090 3,739 17,251 8,661 9,168 3,703 4,481 PUBLIC SCHOOLS - - - - ALL STUDENTS 3,712 3,969 2,749 2,299 10,452 4,892 5,068 3,125 3,696 14,604 5,863 6,074 3,745 4,430 URBAN 21,357 22,492 4,267 3,597 33,942 25,712 26,980 5,063 6,481 47,466 30,816 32,334 6,068 7,767 RURAL 2,521 2,719 2,647 2,211 8,866 3,236 3,326 2,971 3,475 11,991 3,879 3,986 3,560 4,164 NONPUBLIC SCHOOLS - - - - ALL STUDENTS 34,979 36,466 3,758 3,970 45,576 50,997 53,729 5,339 5,945 73,617 61,118 64,393 6,398 7,125 URBAN 65,690 68,989 4,876 5,575 81,499 94,064 98,607 8,971 9,728 132,740 112,733 118,177 10,751 11,659 RURAL 19,653 20,235 3,199 3,169 27,648 32,498 34,453 3,778 4,320 48,221 38,948 41,290 4,528 5,178 MEDIANS ALL SCHOOLS ALL STUDENTS - - 2,500 1,500 6,000 - 278 2,920 2,670 10,013 - 333 3,500 3,200 URBAN 14,000 15,000 3,750 2,800 26,250 12,516 12,516 4,589 4,172 36,922 15,000 15,000 5,500 5,000 RURAL - - 2,500 1,500 5,800 - - 2,698 2,503 9,262 - - 3,233 3,000 PUBLIC SCHOOLS ------ALL STUDENTS - - 2,500 1,500 5,533 - - 2,781 2,503 8,511 - - 3,333 3,000 URBAN - - 3,333 2,100 12,750 4,172 4,673 4,172 3,504 18,294 5,000 5,600 5,000 4,200 RURAL - - 2,500 1,500 5,400 - - 2,642 2,503 8,205 - - 3,167 3,000 NONPUBLIC SCHOOLS ------ALL STUDENTS 17,500 18,000 3,000 1,800 23,000 25,032 25,032 4,172 3,588 38,215 30,000 30,000 5,000 4,300 URBAN 55,400 55,400 4,000 3,100 67,000 75,096 75,096 6,675 5,757 99,794 90,000 90,000 8,000 6,900 RURAL 9,000 9,000 2,500 1,333 15,200 12,516 13,350 2,920 3,087 25,338 15,000 16,000 3,500 3,700

93 TREND IN HOUSEHOLD SECONDARY SCHOOL EXPENDITURE: MEAN AND MEDIAN PER STUDENT ANNUAL HOUSEHOLD SCHOOL EXPENSE BY ITEM*. Uganda Household Surveys 2002 and 2006. In Ugandan Shillings. 2002 2006, constant 2002 prices 2006, current prices

SCHOOL SCHOOL SCHOOL FEES AND BOOKS TOTAL FEES AND BOOKS TOTAL FEES AND BOOKS SCHOOL BOARDING/ AND SCHOOL SCHOOL BOARDING/ AND SCHOOL SCHOOL BOARDING/ AND FEES LODGING UNIFORM SUPPLIES EXPENSES FEES LODGING UNIFORM SUPPLIES EXPENSES FEES LODGING UNIFORM SUPPLIES MEANS ALL SCHOOLS ALL STUDENTS 123,770 128,524 8,430 8,946 150,794 134,699 142,201 12,333 19,869 198,539 161,432 170,423 14,781 23,812 URBAN 145,034 154,728 7,417 9,609 176,407 166,775 167,421 16,184 23,976 232,387 199,875 200,649 19,396 28,734 RURAL 113,741 116,165 8,909 8,634 138,714 115,625 127,204 10,043 17,426 178,412 138,573 152,450 12,037 20,885 PUBLIC SCHOOLS ALL STUDENTS 121,185 123,458 7,541 7,991 144,776 104,547 118,709 11,204 14,189 169,404 125,296 142,269 13,427 17,006 URBAN 129,624 130,111 8,122 9,914 153,916 55,055 55,055 13,270 8,070 110,940 65,981 65,981 15,904 9,672 RURAL 118,872 121,634 7,382 7,464 142,269 121,946 141,086 10,477 16,341 189,956 146,148 169,087 12,556 19,584 NONPUBLIC SCHOOLS ALL STUDENTS 120,243 121,041 8,998 8,756 142,870 136,107 138,128 13,841 22,067 197,718 163,120 165,542 16,588 26,447 URBAN 160,352 160,622 7,112 7,397 178,627 187,360 187,360 19,738 29,123 259,665 224,545 224,545 23,656 34,903 RURAL 94,667 95,802 10,201 9,622 120,069 101,640 105,019 9,875 17,322 156,060 121,813 125,863 11,835 20,759 MEDIANS ALL SCHOOLS ALL STUDENTS 108,000 115,000 6,500 8,000 143,800 111,809 112,644 10,013 14,519 173,137 134,000 135,000 12,000 17,400 URBAN 106,000 144,000 2,500 7,200 189,000 138,927 138,927 16,688 23,085 222,367 166,500 166,500 20,000 27,667 RURAL 108,000 114,000 8,000 8,000 131,500 100,128 101,129 8,344 12,516 147,688 120,000 121,200 10,000 15,000 PUBLIC SCHOOLS ALL STUDENTS 114,000 115,000 5,000 6,500 129,000 100,128 100,128 10,013 9,011 134,505 120,000 120,000 12,000 10,800 URBAN 120,000 130,000 8,000 6,000 163,500 - - 11,682 2,086 54,236 - - 14,000 2,500 RURAL 108,000 115,000 5,000 7,000 129,000 112,644 112,644 8,344 10,013 147,688 135,000 135,000 10,000 12,000 NONPUBLIC SCHOOLS ALL STUDENTS 90,000 90,000 6,000 8,000 114,000 120,988 125,160 10,847 16,688 174,556 145,000 150,000 13,000 20,000 URBAN 144,000 144,000 - 6,750 189,000 175,223 175,223 17,940 25,032 249,485 210,000 210,000 21,500 30,000 RURAL 90,000 90,000 10,000 8,000 101,000 87,612 87,612 6,953 12,516 125,160 105,000 105,000 8,333 15,000

94 TREND IN HOUSEHOLD PRIMARY AND SECONDARY SCHOOL EXPENDITURE: MEAN PER STUDENT TOTAL ANNUAL HOUSEHOLD SCHOOL EXPENSE BY GROUPS OF HOUSEHOLDS. Uganda Household Surveys 2002 and 2006. In Ugandan Shillings. 2002 2006, constant 2002 prices 2006, current prices households with households households households households at least one households households with with at least households with children with at least with children member in with children children only in one member in with children only in one member in households with only in school of any only in primary secondary school of any only in primary secondary school of any children only in secondary level school school level school school level primary school school MEANS ALL SCHOOLS ALL STUDENTS 38,686 15,658 150,794 61,443 23,776 198,539 73,637.5 28,494.5 237,943.5 URBAN SCHOOLS 94,505 54,341 176,407 148,816 80,211 232,387 178,351.4 96,130.8 278,509.3 RURAL SCHOOLS 29,519 11,770 138,714 45,913 17,251 178,412 55,025.6 20,674.6 213,821.4 PUBLIC SCHOOLS ALL STUDENTS 20,251 10,452 144,776 27,759 14,604 169,404 33,267.9 17,503.0 203,025.2 URBAN SCHOOLS 55,576 33,942 153,916 72,771 47,466 110,940 87,213.8 56,886.0 132,958.2 RURAL SCHOOLS 17,449 8,866 142,269 23,685 11,991 189,956 28,386.2 14,370.9 227,656.8 NONPUBLIC SCHOOLS ALL STUDENTS 66,779 45,576 142,870 114,172 73,617 197,718 136,831.3 88,227.2 236,959.5 URBAN SCHOOLS 101,050 81,499 178,627 178,050 132,740 259,665 213,387.8 159,084.5 311,200.5 RURAL SCHOOLS 46,869 27,648 120,069 80,634 48,221 156,060 96,637.0 57,791.2 187,033.0 MEDIANS ALL SCHOOLS ALL STUDENTS 9,750 6,000 143,800 16,354 10,013 173,137 19,600 12,000 207,500 URBAN SCHOOLS 63,000 26,250 189,000 99,877 36,922 222,367 119,700 44,250 266,500 RURAL SCHOOLS 8,250 5,800 131,500 13,100 9,262 147,688 15,700 11,100 177,000 PUBLIC SCHOOLS ALL STUDENTS 6,333 5,533 129,000 9,700 8,511 134,505 11,625 10,200 161,200 URBAN SCHOOLS 19,100 12,750 163,500 23,889 18,294 54,236 28,630 21,925 65,000 RURAL SCHOOLS 6,000 5,400 129,000 9,262 8,205 147,688 11,100 9,833 177,000 NONPUBLIC SCHOOLS ALL STUDENTS 36,480 23,000 114,000 62,163 38,215 174,556 74,500 45,800 209,200 URBAN SCHOOLS 78,500 67,000 189,000 127,329 99,794 249,485 152,600 119,600 299,000 RURAL SCHOOLS 21,667 15,200 101,000 36,296 25,338 125,160 43,500 30,367 150,000

95 96

ANNEX 2: SCHOOL GRANTS.

What are School Grants? School grants are transfers of financial resources and authority from governments or non-governmental organizations directly to schools or small networks of schools. School grants are managed by the school director, a school council, or parent-teacher association (PTA) with the legal authority to receive and spend funds. School grants are often supported by education development projects financed by bilateral and multilateral organizations.

School grants can be either unconditional or conditional. Unconditional school grants are those that the receiving organization may spend as it wishes. An example is Nicaragua’s Autonomous School model, where the Ministry of Education transfers a monthly lump sum payment to secondary schools who then independently decide how to spend funds. Conditional school grants are financial resources transferred to the school level for the purpose of purchasing specific school inputs such as textbooks or teacher training or to fund school improvement projects.

What Are the Objectives of School Grants? The objectives of school grants vary widely. It is precisely this capacity to address multiple and different objectives that makes them an attractive policy tool. The improvement of the quality and relevance of school inputs—better teacher performance, increased provision and relevance of textbooks and school materials, improved school infrastructure—motivate many school grant projects. The Small Grants for School Improvement Program in Guinea enables teachers to take responsibility for their own professional development. Teams of up to 10 teachers work together to determine their own professional development needs, and to compete for small grants.

Some school grant programs also have the objective of improving school access and/or equity and use a targeting mechanism to meet the needs of populations underserved by the education system. With the aim of serving the poorest and most isolated communities, El Salvador’s Education with Community Participation Program channels education funds through parents’ organizations at the community level to hire teachers and manage educational services first-hand.

Improvement of management and efficient utilization of resources represents a fourth stated objective of school grant schemes. In most countries, teachers’ salary expenditures eclipse essential non-salary expenditures. School grants that are earmarked for non-personnel inputs are one means of ensuring minimum provision of such inputs.

What are Some Design Features of School Grants? School grant funds are often formula- based, with poverty rates and student population determining the funding amount. Some school grant plans incorporate a targeting mechanism to reach underserved populations. School grants can be competitive or simply based on fulfillment of particular criteria. School grant schemes can also offer incentives based on performance. Ethiopia’s Community-Government Partnership Program bases the opportunity for continued program participation on approved financial and subproject management of previous grants, and schools progress through three phases of funding. Each phase is worth increasingly more funding, and the application criteria become increasingly more rigorous. Alternatively, Chile’s National Teacher Performance Evaluation System (SNED) awards its incentive grants based on student achievement.

To increase accountability for funds, a variety of programs include safeguards. Indonesia’s School Improvement Grants Program requires that two members of the school committee, the head teacher, and the community representative sign to open the school’s bank account and to approve each withdrawal and use of funds. At each phase in Ethiopia’s Community-Government

Partnership Program, the school sponsors an open-house to inform the larger community about school improvement efforts. After completion of the project, the school holds another open-house to convey its accomplishments.

Examples of International Experience

1. Chile--National Teacher Performance Evaluation System (SNED)

Established in 1994, Chile’s National System of Performance Assessment (SNED) awards teacher incentive grants to schools based on an index of school excellence measures.

Objectives of School Grant. The SNED creates competition among schools to encourage teachers to improve their performance.

Design Features. Chile’s National System of Performance Assessment (SNED) program mandates that schools spend grants in the form of teacher incentive awards and teacher bonuses. The teacher incentive grants are conditional in that awarded school directors must use 90 percent of the grant for teacher bonuses based on hours worked. The school director is to allocate the residual 10 percent to “outstanding” teachers at his/her discretion to avoid the “free-rider” problem. Another design feature of the SNED program is that the teacher incentive grants are distributed through a competitive process. Schools are stratified within regions by socioeconomic status and other external factors that affect school performance. This ensures that the process is competitive among comparable establishments. Every two years, schools are ranked according to an index of school performance measures using the national System for Measuring Educational Quality (SIMCE) test as the basic criterion. Schools can win the teacher incentive grants repeatedly.

2. Guinea— Programme de Petites Subventions aux Écoles (PPSE)

Since 1994, Guinea has been implementing a unique and promising World Bank funded program that integrates school improvement with professional development for teachers known as the Small Grants Staff Development and School Improvement Program (PPSE). PPSE is a conditional school grant program that engages primary school teachers to participate in the process of education quality improvement through competitive small grants of approximately $1000 that are awarded to school-based teams of teachers.

Objectives of School Grant. The overall objective of Guinea PPSE is to improve the quality and relevance of specific school inputs, which in this case are teachers. As a means of improving the quality of primary education, PPSE provides organizational support and the incentives necessary for teachers to assume primary responsibility for their own professional development and to determine what is most appropriate in their local context for improving teaching practices. Furthermore, this program seeks to give teachers greater professional autonomy to analyze teaching and learning problems at the classroom level, define the problems or issues to be addressed in a 1-year project, propose and implement solutions, then evaluate and report results.

Design Features. Diverging from the traditional top-down approach of in-service teacher training where central education authorities mandate workshop contents for large groups of teachers, PPSE allows teams of teachers to design professional development programs unique to their local context and compete for grant funds to implement their own programs. Teachers learn about PPSE grant competition through a series of workshops led typically by a pedagogical advisor or a regular school teacher, who presents the program’s operational manual and proposal-

writing guidelines. Interested teams of teacher then go through a two-cycle, highly structured competition. First, teacher teams determine the contents of their projects, prepare their own budget, and then submit preliminary proposals for their own professional development program to a prefectural jury, which is presided over by the prefectural director of education (DPE) and composed of retired teachers and local education leaders. Second, once promising proposals are selected, pre-selected teacher teams are invited to revise their proposals with help from the facilitators based on critical comments received from a prefectural jury, and then submit their final proposals to a regional jury who makes final decisions of which team will receive grants. The regional jury is presided over by the Regional Inspector of Education (IRE) and composed of local educational leaders. Selected teacher teams are granted full funding, provided with project implementation support from the project facilitator, and visited by an evaluator, who is typically a prefectural or regional jury member, three times throughout the 1-year project cycle. In addition, since PPSE also has performance incentives as one of its design features, teacher teams are given the option of renewing their grant if they show that their projects attained good results. School grant schemes can also offer incentive based on performance.

3. Indonesia’s School Improvement Grant Program

Indonesia initiated the School Improvement Grant Program (SIGP) for primary and junior secondary schools as part of the large school safety net program to mitigate the impact of the economic crisis of 1997.

Objectives of School Grant. The SIGP, funded by the Royal Netherlands Government through a World Bank Trust Fund, targets large one-off grants to a small number of schools based on the following three categories:

1. Schools coping with a large increase of displaced students due to civil unrest and social conflict 2. Schools with sufficient damage from natural disaster 3. Schools among the poorest 10 percent in the poorest 10 percent of districts

Design Features. Though the program depends upon district-level education officials for school selection, funds go directly from the Government of Indonesia to schools. School committees, composed of teachers, local government authorities, and community members allocate funds depending on the category in which the school falls. For category one, school grants might be used towards purchasing furniture or more textbooks to cope with the influx of internally displaced students. For categories two and three, grant monies might be used towards the repair of water supplies and improvement of toilet facilities. Before SIGP funds, lack of adequate toilet facilities forced students to use nearby streams or fields, discouraging parents from sending girls to school and disrupting classes through frequent requests for toilet breaks. Many steps incorporated into the implementation of the SIGP seek to enhance grant effectiveness. For example, with construction as a central component of many subprojects, on each district committee, the SIGP demands the participation of a member of the district’s department of public works. At the school-level, the head teacher and one community representative must sign for each withdrawal of grant funds and inform the community through the school notice board what the school committee will use the SIGP funds for. Finally, category three grant seeks to fine-tune the allocation of grant funds so that resources reach those with greatest needs.

4. Tanzania—Community Education Fund (CEF)

Started in 1995, the Community Education Fund (CEF) program is a non-competitive matching grant program that has been designed to increase the allocation of public funding for non-salary expenditures at the school level and to empower communities to improve their primary schools.

Objectives of School Grant. The objectives of the CEF program are as follows:  Increase community involvement in the management of school resources and create a sense of ownership for school inputs.  Overcome lack of public financing and improve quality of school inputs.  Increase access to schooling and improve quality of education performance (test scores).  Ensure funding reach school in a timely manner and resources are available for the quality control of schools.

Design Features. There are several criteria that schools have to meet in order to be eligible to participate in the CEF program. These criteria include registration as a primary school to operate in Tanzania, constitution of a school committee that is comprised of members elected by the parents, and development of a Three-Year School Plan with a contractual agreement between the school and the school community. The community decides by majority vote whether they would like to participate in the CEF program. If they decide to participate in the program, the community may raise funds by collecting cash contributions from parents and local businesses, and through various fund-raising activities. The government will then match the money raised by the community on a 1-to-1 basis, but the amount is not to exceed Tsh. 6,000 per pupil. One of the salient features of the CEF program is that funds are being allocated directly to primary schools. The schools, however, may not exclude students from attending if the parents are unable to contribute to the CEF. Furthermore, students may not, under any circumstances, be used to generate funds for the schools by providing labor outside the proximity of the school property. The CEF is a matching fund as well as a targeted fund program. This is because disadvantaged schools are eligible to receive additional targeted subsidy of 0.5-to-1.

One way of ensuring accountability, the CEF grant program is designed in such a way that not all schools may continue to participate in the CEF program even if they satisfy the initial eligibility criteria. There are several requirements that participating schools have to meet in order to maintain their status. In addition to complying with their own Three-Year Plan, schools must show consistent improvement in school enrollment without jeopardizing student attendance and performance.

5. Armenia-School Improvement Program (SIP)

Armenia’s School Improvement Program (SIP) aims at promoting community and parental participation in school financing and management, and at building capacity for reform management. The Ministry of Education and Science implemented SIP through support from the World Bank in 1998, and the program is headed by a SIP Board chaired by the Ministry of Education and Science. The SIP is a conditional and competitive grant program that supports implementation of policies for school autonomy and innovative school improvement projects (micro projects) aimed at increasing education quality by channeling resources directly to schools.

Objectives of School Grant. The objective of the SIP is to improve education quality by decentralizing school management, increase school autonomy by formalizing community and parental participation in school management, and support capacity building at the school level.

Design Features. The SIP makes grant funds available to individual schools based on the budget determined by their elected parent-teacher board. Schools have to meet a list of specified criteria to qualify for submitting grant proposals on an annual basis and to receive funding of up to a maximum of US$ 15,000 for micro-projects. Some criteria for eligibility include (i) autonomous school, which can be demonstrated through having an active school board composed of elected teachers and parents following national guidelines; and (ii) an original micro project business plan for improving school quality, according to locally defined objectives, prepared by elected parent-teacher board. The SIP has several design features:

Matching—it is a requirement for community to contribute at least 10% of total project costs. Among schools with similar conditions, the SIP is biased towards schools making a larger contribution to the total project costs. School contributions may include both monetary and in- kind, such as provision of school materials, equipments and labor.

Formula based--the central government uses a simple formula based on pupil population to allocate funds across clusters of schools. Each individual grant provides the school with out-of- budget resources for further school development as determined by the school council itself. Grant scope of activities include training of teachers and administrative staff, new organization of school management with community and teacher participation, and integration of children with special needs into the school process. Competitive—presented project must not duplicate the design of school improvement projects initiated by others or ongoing projects.

Conditional—the micro-project must be implemented by the elected parent-teacher board and the quantity of work carried out must conform to the timetable submitted with the proposal.

ANNEX 3: TEACHER ABSENTEEISM

Characterizing Teacher Absence in Uganda: Evidence from 2006 Unit Cost Study by James Habyarimana, March 2007.

Introduction and Motivation

The relationship between schooling inputs and educational outcomes continues to receive wide attention in discussions about how to improve educational outcomes. A predominant share of educational inputs in developing countries are publicly provided. Primary among these inputs is a key input in cognitive achievement: teacher instruction. The importance of teacher instruction is underlined by the fact that it is difficult for households to find alternative substitutes, particularly in developing countries where the markets for private instruction is very thin and parental levels of education are too low (to support significant self-provision).

In addition to being a vital input in the production function for learning, teacher salaries account for a large fraction of recurrent expenditure in all levels of schooling, and even more so at the primary level. Bruns et. al. (2003) estimate that primary school teacher remuneration in developing countries account for between 50-80% of recurrent expenditure.

Despite the importance of instructional time in the production function and its demands on meager public resources, recent evidence from a cross-section of studies around the world has revealed very high levels of teacher absenteeism. The best estimates come from a multi-country study in which teacher absence is measured using direct observation (Chaudhry et. al. 2006). The multi-country surveys were conducted at the end of 2002/early 2003 and found absence rates ranging from 11% in Peru to 27% in Uganda.

This chapter presents the results of a follow-up study to measure and characterize teacher absence in Ugandan schools. The timing of the study is crucial to understand and possibly evaluate a number of important policies that have been implemented in the intervening period. Chief among these is the increased control (and experience) that district authorities are exercising over the management of primary education. Many proponents of decentralization argue that assigning control to managers with better information and possibly incentives leads to higher output conditional on inputs.

The main findings reported in this chapter are that: 1. The level of teacher absenteeism is high. Nearly 20% of all teachers could not be found in the school at the time of enumerator visits. 2. A large fraction of teachers that are present were not in class at the time of the enumerator verification. In fact nearly one third of teachers were outside of the classroom when they were found.

3. Teacher absence is very heterogenous. Variation between districts accounts for less than 2% of variation, while districts and schools account for only 18% of total variation. 4. We find a strong negative association between parental involvement (parental contributions of resources and a higher frequency of parent meetings) and school level absence. 5. We document a strong negative association between the number of functioning teacher housing units available and the level of teacher absence 6. The presence of neighboring non-government owned schools is associated with lower levels of teacher absence. This is potential evidence of the effects of competition. 7. Individual characteristics such as teaching in the district-of-birth and age are significantly associated with teacher absence. 8. The results of the study conducted at the end of 2006 tentatively suggest that absenteeism has fallen relative to the levels measured in 2002/2003. We discuss the possibility that this result is affected by the composition of the sample.

Understanding the sources of such high levels of teacher absence is crucial for the design of more effective systems to produce the cognitive skills required to be productive. A number of studies have attempted to evaluate the extent to which the measured absence is a result of weak incentives (salaries) and/or weak enforcement (monitoring by managers of education or parents). While earlier studies primarily used non-experimental data (King and Ozler (1999)), recent studies present evidence from randomized interventions targeted at different aspects of instructional time. (Duflo and Hanna (2006); Glewwe et. al. (2006); Miguel et. al. (2007) and Karthik (2007)). The results of these studies provide a measure of confidence on the reliability of different types of interventions that are likely to generate improvements in teacher attendance.

The rest of this chapter is organized as follows: section 2 discusses data and sampling issues, section 3 explains the methodology and presents baseline estimates of teacher absence. Section 4 discusses the reasons for teacher absence. Section 5 compares the results to estimates from the 2002 survey. Section 6 presents the individual-, school- and district-level correlates of absence. Section 7 discusses the results and concludes.

Data

The data used for this exercise comes from 160 schools in six districts from three regions in Uganda. There are no districts from the Northern region in this sample owing to an on- going parallel and region-specific study. All schools were visited in November 2006. The districts were selected from the eligible universe (excluding Northern districts) with a sampling probability proportional to population share. Population estimates were drawn from estimates provided by the Uganda Bureau of Statistics (www.ubos.or.ug). Care was

taken to ensure that the drawn sample reflected the broad attributes of the population with respect to socio-economic status (see appendix for details).

Having selected the districts, the list of schools was stratified by ownership into two categories: government-owned schools and non-government owned schools. The list of the schools and their attributes comes from the 2005 Education Management Information System (EMIS) school returns data. The stratification was done in order to generate a sample of non-government owned schools that was large enough to produce reliably precise measures of central tendency for non-government owned schools. Within the each strata, schools were selected with sampling probability proportional to pupil enrollment. 23 government schools and 7 non-government schools were selected for each district. By design, about ¾ of the schools sampled are government owned and the rest non- government owned. Non-government owned schools include community schools, private-owned and schools owned/managed by non-governmental organizations (chiefly religious organizations). The appendix shows the characteristics of schools in selected districts and sampled schools relative to non-sampled schools in the country and selected districts respectively.

Owing to time constraints, only 160 out of the sampled 180 schools could be visited before the end of the third term. Of the 160 schools that were visited 127 were government schools and 33 were non-government schools. Table 1 below presents means of selected characteristics. The average school in the sample had an average of 11.6 teachers and a pupil-teacher ratio of 48.4. Only 5% of the schools in the sample were multi-grade schools, where more than one grade is taught in the same classroom. 20% of the schools were located in peri-urban or urban areas, with nearly two thirds in peri-urban areas. One third of schools is located within 5 kilometers of a main road and nearly two- thirds are located within 5 km of a taxi stage.

Table 1: School characteristics, selected means by ownership Non- Government Total government

Number of teachers 10.64 11.90 11.63 (0.98) (0.42) (0.39) Pupil teacher ratio 38.42 51.09 48.44 (3.29) (1.23) (1.25) Proportion of mother’s literate 0.72 0.60 0.63 (0.07) (0.04) (0.03) Multi-grade school 0.06 0.05 0.05 (0.04) (0.02) (0.02) Proportion of schools within 5km of main road 0.39 0.32 0.34 (0.09) (0.04) (0.04) Proportion of schools within 5km of taxi stand 0.79 0.62 0.65 (0.07) (0.04) (0.04) Proportion Urban 0.36 0.15 0.20 (0.09) (0.03) (0.03) Number of observations 127 33 160

Since the primary focus of this chapter is on teachers, it is worth pointing out a number of important teacher characteristics. A total of 1837 teachers were surveyed as part of this exercise. This includes all teachers in schools with less than 20 teachers and a maximum of 20 for schools that had more than 20 teachers. A little more than 10% of schools had 20 or more teachers. As figure 1 shows, 60% of all teachers are class teachers/permanent teachers; 14% are heads of department and 20% are head-teachers or deputy head teachers. The rest are volunteer, private or other part-time teachers. Figure 1: Teacher composition

Teacher Rank Head Teacher 1%1%0%4% 9% Deputy/Asst. Head 10% Director of Studies/Head 20% Class Teacher Permanent/Regular 14% Private Temporary/Probationary Volunteer 41% Other

The demographic structure of our sample of teachers is presented in table 2 below. 58% of teachers are male. The average age in the sample was 35.6 and they had been teaching in the school for an average of 4.4 years. Only about a quarter of all teachers had completed A-levels. More than ¾ of teachers are married and fluent in the language spoken in the area around the school. More than half the teachers are teaching in their district of birth. This is in part an outcome of autonomy in hiring by the districts and the tendency for districts boards to hire ‘locals’. 63% of all teachers live in the same parish as the school.

In the sections that follow, we restrict our analysis to full time teachers. This involves dropping about 1.7% of observations that correspond to part-time, private, volunteer teachers or teachers that have been transferred but are still on the school roster.

Table 2: Teacher characteristics, selected means Mean (Standard Error)

Gender (1=Male) 0.58 (0.01) Age, years 35.60 (0.32) Tenure of current posting 4.40 (0.12) Proportion that have completed A-levels 0.24 (0.01) Proportion teaching in district of birth 0.54 (0.01) Proportion that live in the same parish as school 0.63 (0.01) Proportion fluent in language spoken around school 0.77 (0.01) Proportion married 0.76 (0.01) Number of Observations 1843

Measuring Absence Teacher absence is measured in three ways. Firstly we ask the head teacher or the primary respondent (typically a deputy head teacher or senior teacher) about the attendance of up to 20 teachers per school.84 We call this the head teacher spot absence report. Secondly we ask the teacher about the duration of absence for all teachers over the last 30 days. We call this the head teacher absence duration report. Finally, we use direct observation to assign absence status; the enumerator marks a teacher as present, if the enumerator can find that teacher during his/her visit during working hours. A teacher is absent if the enumerator cannot find this teacher within the school boundaries. All measures rely on the fact that each of the school visits is unannounced.85

The head teacher reports of absence are unreliable for a number of reasons: 1. The head teacher/primary respondent has incentives to misrepresent the absence status of teachers since this reflects on his/her capacity to effectively manage the school. 2. The head teacher is unlikely to be aware of the absence status of all teachers at all times. 3. Recall bias is likely to be significant given a 30-day recall period and the fact that the head teacher must report a duration for every teacher on his roster.

84 In schools with 20 or few teachers, all teachers are sampled. In larger schools, the 20 teachers are drawn randomly using a random number table. This random number table differs from school to school. 85 Care was taken to ensure that schools were not informed of when or if enumerators would visit the school. District education offices were informed of the survey but had no idea which schools had been selected to participate in the survey.

Both of the head teacher based measures could be improved if there was an attendance register in which teacher attendance was accurately recorded. In general, it is not clear how accurately and/or if teacher attendance registers are used. Evidence from this survey indicates that head teachers in 2 schools reported that no such register existed. Furthermore, 55 teachers reported as present by the head teacher did not sign the register, while 50 teachers reported as absent signed the register. It is only coincidental that these two quantities offset each other. Evidence from the previous survey (Habyarimana, 2004), suggests that the bias associated with a reliance on attendance registers is on the order of 5 percentage points.

If absence durations over at 30-day recall period were reported with small/no biases, one could use the estimate as a measure of the probability of a teacher being absent. An unbiased measure of absence duration also allows us to determine the concentration of absence by teacher. In the absence of any recall bias, the probability that a teacher was absent is equivalent to the ratio of monthly absence duration to the total number of days in the month. As the result below demonstrates, absence durations from head teacher recall predict a much lower level of spot absence. According to this data and assuming a teaching month of 22 days, the spot absence rate suggested by head teacher recall is on the order of 10%, about half the size of the corresponding spot absence measure.

Table 3: Head Teacher Response: Spot Absence and Absence Duration over last 30 days Variable Mean Implied spot absence

Head teacher report, duration 2.08 0.1 (0.09) Head teacher report, spot 0.21 (0.01)

However, the dispersion in the days absent is potentially informative if recall bias is independent (or proportional to) of the duration of absence. An examination of the data suggests that head teachers report that about 50% of all teachers are present throughout the previous month; 10% are absent for 1 day; 12% for 2 days; 9.5% for 3 days; 5.5% for 4 days. 10% of teachers have absence durations between 5-10 days. 3% of teachers are absent 10 or more days. To the extent that head teachers have incentives to under-report absence, this distribution suggests that absenteeism is not restricted to a small fraction of “ghost” teachers. As has been shown in other surveys (Chaudhry et. al (2006) and Glewwe et. al. (2003)), between 50-80% of all teachers have taken an absence episode at least once over the previous 30 days.

Given the shortcomings of both the head teacher reports of absence, we rely instead on the measure of absence that comes from direct observation of the teacher’s status. This measure has a number of advantages over other measures that have been used in the literature. Firstly, it does not suffer from biases resulting either from recall or from mis- reporting. Secondly, it provides a measure that is of direct policy interest. Since the enumerator’s arrival in the school is random, the estimate of directly-observed absence

tells us the fraction of teachers one can expect to find in school during working hours. In addition, this strategy allows us to characterize the allocation of teacher effort for teachers that are present.

This measure of absence does have an important drawback. With only one school visit, we are unable to measure teacher-level absence rates with any precision. One visit, cannot distinguish between teachers who are absent a lot of the time, and teachers who are only absent a few times. This has implications for the teacher-level analysis that we will perform – measurement error of teacher absence rates is likely to produce noisy estimates of associations at the teacher level. As a result, we use the school level estimate of absence which is measured with greater precision.

18.3 % of teachers could not be found by the enumerator over the duration of the school visit. This is a non-trivial fraction of the workforce. We investigate the existence of simple associations in the data in the table below. In particular we look at the extent to which absence rates vary by ownership, location and district. Surprisingly, there is very little variation in average absence across ownership and location categories. In fact, government run schools have a lower rate of absence than non-government owned schools (albeit not statistically significant). Similarly, absence rates do not differ across bucolic status. Both peri-urban and rural areas have absence rates of 18% compared to 20% for urban areas (difference not significant). There is however, variation in absence rates across districts. Average absence rates range from 10% in Mukono to 23% in Mayuge. Table 4a: Mean Absence Rates: Ownership, Location and District Category Mean Absence (std. Error) Ownership Government 0.18 (0.01) Private 0.23 (0.03) Community 0.21 (0.04) Other 0.04 (0.04) Location Urban 0.20 (0.04) Peri-urban 0.18 (0.02) Rural 0.18 (0.01) District Kibaale 0.19 (0.02) Luweero 0.17 (0.02) Mayuge 0.23 (0.02) Mukono 0.10 (0.02) Ntungamo 0.17

(0.03) Tororo 0.21 (0.02) 0.18 Total (0.01)

As the table below shows, the absence rate varies significantly across teacher rank. Head teacher absenteeism stands at 27% compared to the average of 18%. The absence rate of other teachers is on par with average absence rates.

Table 4b: Teacher Absence rates by teacher rank Teacher Rank Absence Rates

Head Teacher 0.27 (0.04) Senior Teachers 0.16 (0.02) Regular Teachers 0.18 (0.01) Others 0.13 (0.04)

Overall 0.18 (0.01)

We turn now to an examination of the reasons for teacher absence. Direct observation does not permit the possibility of characterizing the absence episode. In order to establish reasons for absence, we asked the head teacher/primary respondent to report why a particular teacher was not in school on the day of the survey visit. The figure below shows the distribution of primary respondent explanations for teachers that they report as not being in school. Figure 2a: Reasons for Teacher Absence

Absence for official reasons accounts for 20% of absences; the bulk of which are teaching related. Each of the following categories account for 1/6 of all absences  Illness  Authorized leave  Different shift/expected to arrive later  Un-authorized leave Finally observations for which the head teacher could not determine the reason or where teachers had left early, account for another 1/6 of all absences. While these reasons are subject to recall and other biases, they provide a guide to understanding and characterizing teacher absence. Firstly, it is instructive to look at the breakdown in absence reasons by teacher rank. We split the sample of teachers into the head teacher and other teachers (including the deputy head teachers, heads of department and other teachers). Figure X below provides a breakdown of teacher absence by teacher rank:

Figure 3:

Reasons for Teacher Absence Head Teacher Regular Teacher

5.4% 10.4% 0.3% 13.9% 2.1% 2.1% 16.1% 2.3% 2.0%

45.8% 20.8% 6.2% 15.9% 5.7%

6.3% 14.2% 18.1% 12.5%

Official teaching related duty Official non-teaching duty Assigned elsewhere/transferred Sick Authorized leave Expected to arrive later Left early Don't know Unauthorized absence Suspended Other - please specify

Source: Uganda Unit Cost Study, 2006

Nearly half of all head teachers’ absences are due to official teaching related duties compared to only about 14% of regular teacher absences. 16% of regular teacher absences are unauthorized compared to 2.1% of head teacher absences.

Given the absence rates and teaching days per month, this is equivalent to approximately 0.7 teaching days per month devoted to official teaching related duties such as seminars, exam invigilation and training. As shown in the table AA above, absence rates for head teachers are 50% higher than other teachers. In addition, nearly half of these absences are due to official teaching related reasons. Consequently, head teachers are away from school about 3 days a month for official teaching related reasons. In the multi-country studies cited above, head teacher absence rates were about 3 percentage points higher than that of regular teachers.

How much do these reasons tell us about the sources of high teacher absence? Firstly, we need to point out the need for caution in interpreting head teacher reported reasons for absence. However, conditional on the reliability of these reports, it is clear that there is no predominant source for absenteeism. Illness, which has been identified as a major source of absenteeism in a number of studies (Das et. al. 2004; Bennel (2005) and Bell et. al. (2003)), only accounts for little more than 3% of all observations. Official reasons account for only 3.6% of observations. The picture painted by this evidence is one of weak incentives and an education system that requires teachers, and particularly, head teachers to be away from school.

Understanding why teachers are absent and what fraction is absent is crucial. But given that the input of interest is instructional time, it is important to examine the allocation of teacher effort even for those teachers that are present. The figure below shows the distribution of activity across a number of categories.

Figure 4: What are teachers doing when the enumeration team finds them?

2006 0% In class, teaching 18% 19% In class, not teaching 2% Out of class, break 8% Out of class, in school Administrative work 18% Can't find teacher

With surveyor 35%

18% of teachers are in class teaching; only 2.4% are in class but not teaching. 17.6% are out of class on a scheduled break. More than 1/3 of teachers are out of class and in school and 8% of teachers are doing administrative work. While the foregoing has emphasized absence rates that are very high, figure 4 above suggests that improving teacher attendance alone is not enough to increase instructional time. One concern with the result that 1/3 of teachers are out of class is that it reflects the odd timing of the survey (at the end of term during the examination/marking period). In the figure below we present evidence from the 2002/3 absence survey. 2002 In class teaching 3% In class but not teaching 25% 28% Out of class on a scheduled break Out of class, but in school Doing administrative work 6% 11% Cant find teacher/absent 18% 9% Accompanying surveyor

The figure above shows that 18% of teachers are outside the classroom, when they should be teaching. The timing of the 2002 survey provides a potential lower bound of the proportion of time that teachers spend in the classroom.

Robustness of the Absence Estimates One of the concerns with our measure of teacher absence is that it relies on the effort of the enumerators to find the teachers. In this case, a teacher will be classified as absent when he/she is present in school but beyond the reach of an enumerator. The other concern is that physical verification requires that the enumerator go down his/her list until all teachers are accounted for. There is a danger that the time it takes to complete all teachers might be too long enough to invalidate any verification that takes place outside school hours. Alternatively, teachers that live close by might be alerted that an absence survey is being conducted and show up.

To address some of these concerns, we provide two pieces of evidence that increase our confidence in the reliability of the absence estimates. We employ data from the classroom observation as the first piece of evidence. As part of the survey, classrooms from grades 1-4 were visited to collect information on the classroom environment, including the number of textbooks, desks and subject being taught at the time of the observation.86 The grade was chosen depending on whether the first and last digit of the school identification number is odd/even. The classroom observation collected data on nearly 330 classrooms in the 160 schools. Data from the classroom observation shows that nearly 20% of all classrooms observed (during school hours) had no teacher in them. The figure below shows the distribution of enumerator observations of the classrooms visited.

86 Note that for this piece of evidence to be valid, we require that absence rates of teachers that teach in these lower grades not be systematically different from the population.

Figure 5: Classroom observation Class room observation

15.8% 24.5% 2.6% 3.1%

14.8%

3.6% 35.7%

Quiet, teacher around Orderly, Teacher Disorderly, Teacher Disorderly, No Teacher Other Orderly, No Teacher No class

Source: Uganda Unit Cost Study, 2006

In about ¼ of the classrooms visited, the teacher was quietly conducting his/her class. 36% of classrooms had a teacher in them with the pupils doing an assignment. In about 17.4% of the classrooms there was no teacher (and the class was predominantly disorderly). This number is close to the level of absence that we obtain (18%). In one sixth of the classes, the pupils were either not in class or had already gone home. This fraction is larger for the lowest grades since they typically only study half day (see figure below)

Figure 5b: Classroom observation by grade Class room observation P1 P2

17.4% 26.3% 24.2% 21.1% 3.7% 4.6% 2.1% 9.5% 15.6% 2.1% 32.1% 35.8% 5.5%

P3 P4

8.8% 10.2% 1.3%1.3% 22.5% 4.6% 3.7% 29.6% 18.8% 15.7% 6.3% 0.9% 41.3% 35.2%

Quiet, teacher around Orderly, Teacher Disorderly, Teacher Disorderly, No Teacher Other Orderly, No Teacher No class

Source: Uganda Unit Cost Study, 2006

The figure above shows that there is very little variation in the proportion of classrooms with no teachers. The lower classes (P1 and P2) are more likely to have ended by the time the enumerators conducted the classroom observations.

The second piece of evidence involves dropping any observations for which verification of attendance occurred outside normal school hours. We are able to do this because we record the exact time that a verification was done. There are two reasons why we this needs to be done. Firstly, a teacher is not technically absent if he/she cannot be found outside of school hours. Secondly, it is possible that absent teachers are alerted to the enumerators presence and return to school. While this can happen at any time, it is more likely to happen later in the day. Both of these would generate unreliable measures of teacher absence. We define normal school hours to be between 8AM and 5PM. This is the modal school time range in our sample. Using this criteria changes absence levels only by a small fraction (see table 6 below).

Table 5: Absence measures: robustness test Variable All observations Excludes observations verified after 5pm.

Proportion absent (excludes head 0.174 0.168 teachers) (0.010) (0.010)

Proportion absent (includes head 0.183 0.177 teachers) (0.009) (0.009)

The absence rate does not change much when we exclude observations identified after 5pm.

Comparison of 2002 and 2006 Surveys Both the 2002 and 2006 surveys use the same methodology to measure teacher absence and visited schools at the same time of the year. However, differences in the estimates produced by both surveys could be due either to differences in behavior – absence rates have changed or due to use of different samples.87 We can examine the extent to which the differences are due to one of these two reasons by restricting our analysis to two districts that were visited in both surveys. We begin by comparing the samples of teachers and schools that were visited across the two surveys. Selected means are presented in the table below to determine if there are systematic differences across the two samples.

Table 6: Sample comparisons: 2002 vs 2006 Selected Characteristic 2002 October Visit 2006 November Visit

13.44 11.63 Number of Teachers (0.56) (0.39) 50.52 48.44 Pupil to Teacher Ratio (1.69) (1.25) 0.57 0.63 Mother's Literacy (0.03) (0.03) 0.89 0.79 Proportion government owned (0.03) (0.03) 0.11 0.05 Multi-grade teaching (0.03) (0.02) 0.37 0.34 Distance to road (< 5km) (0.05) (0.04) 0.73 0.65 Distance to taxi stand (< 5km) (0.05) (0.04) 0.16 0.20 Urban location (0.04) (0.03) 94 160

Un-weighted absence rate (%) 25.2a 18.4

Weighted absence rate (%) 23.7a 18.3 a - excludes northern districts from calculation.

An examination of average characteristics of schools visited in both surveys suggests that the two samples are comparable across most dimensions. Only the proportion of schools that are government owned is the share significantly different with the former sample

87 The 2006 survey uses a stratified sample (by ownership) while the 2002 survey had no stratification (so non-government owned schools are over-represented in the 2006 sample).

containing 90% of government owned schools compared to 79% of the current sample. On the basis of the similarity of these two samples, one can make a quick comparison of absence across the two samples. A comparison of the simple mean is not particularly informative. Using information about the design of the sample allows us to produce national estimates for both samples – note that the 2006 sample is not drawn from a national sample of districts. The table above suggests that the weighted mean of absence has fallen from nearly 24% in 2002 to 18.3 in 2006. While the weighting does attempt to control for differences in the district-composition of the sample, it is possible that differences in the districts remain that the weighting scheme cannot deal with. To control for this we restrict our analysis to two districts that were visited in both surveys. We look at changes in mean absence rates in Luweero and Tororo across the two years as check on what is likely driving the differences in national absence rates. The chart below shows the results of this analysis.

Figure 6: Changes in Absence rates: 2002-2006

Changes in Abs ence Rates , 2002-2006

40

35

30

25

20

15

10

5

0 Tororo Luweero

2002 2006

The results show what looks a like a dramatic decline in absence rates in Tororo district but a much smaller and statistically insignificant decline in the absence rate in Luweero districts. Absence rates drop by 40% in Tororo between the two surveys. Is this difference capturing real changes in the provision of instruction time or does it reflect subtle differences in the timing of the visits (October/early November vs late

November). It is difficult to say. Census data or a repeated panel would be the ideal tool to determine whether absenteeism has fallen since 2002.88

88 The high absence rates obtained in 2002 are driven by 3 districts where more than 1/3 of teachers are determined to be absent. This includes Tororo district. In the visit in 2006, the district with the highest absence rate, Mayuge, had 23% of its teachers absent.

Determinants of Absence In this section we examine the correlates of absence at the individual, school and district levels. The objective of this exercise is to identify dimensions of association that could potentially be affected by policy. It is important to point out that the results that are reported below are not causal, but rather associations that are potentially causal to absence.

We begin by presenting a theoretical framework that is a rough model of the teacher’s attendance decision. In words, a teacher’s chooses not to attend when the benefits of being away from school exceed the costs of being absent from class.

The benefits and costs of this decision can be driven by three broad categories of factors.  Individual determinants such as the experience and seniority, marital status, education levels (that determine outside opportunities), gender, residential location e.t.c  School based factors such as the availability of housing units, the extent to which parents are involved in monitoring and enhancing school quality, school location and quality of the schooling environment.  District based factors include poverty rates, the quality of management, and in particular monitoring capacity of the district and crucially the likelihood that absenteeism is punished by the controlling authority.

We can model the teacher’s choice in a regression framework that includes each of these factors. In particular assume that we estimate a model:

yijk = + Xijk + +  Sjk + Qk + i + ijk

Where yijk takes on the value of 1 if the teacher is absent and 0 otherwise. Xijk represents individual factors thought to affect the absence of teacher i in school j and districk k, Sjk are school level characteristics in school j in district k, and Qk are district level factors and i is unobserved fixed teacher characteristics. We estimate the model above using a linear probability model and as a limited dependent variable probit estimation. Note that both of these models attempt to model the probability that a teacher is absent on the left hand side. As noted earlier, with one school visit our measure of teacher absence probabilities is very noisy. As such we would expect that this will tend to produce imprecise point estimates of the strength of association. We present below a dependent variable that is measured with less error. The individual level variables include teacher age and age squared, length of posting at current school, teacher gender, rank, marital status, proximity of residence to school, schooling, and ties to the local community. The school level variables include the number of functioning teacher houses, teacher enrollment, pupil teacher ratios, a dummy that takes on the value of 1 if the school is government owned and 0 otherwise, an infrastructure index, school remoteness and location, multi-grade teaching, the degree of competition, measured by the number of non-government owned schools less than 1km

and 1-5km away.89 We include district level variables such as the proportion of schools in the district that have been visited in the term of the survey, the district poverty rates (from the poverty mapping exercise) and the proportion of teachers in the district who are involved as local council members. The results are presented in tables A1 and A2 in the appendix. While the results of the probit analysis are more reliable, the results from the linear probability model are easier to interpret and are similar to the probit results. I will report the LPM results in this chapter. Eight specifications are estimated that cumulatively add individual, school and district controls. Specification (8) includes district fixed effects. As you can see from the regressions, there are three robust associations at the individual level: 1. There is a ceteris paribus moderate U-shaped relationship of absence with age. Absence rates fall with age, until a minimum that is reached around age 25 and then increase therafter 2. Teachers working in their district of origin are about 3.5 percentage points more likely to be absent. (Decentralization?) ceteris paribus. 3. Teachers that live in the same parish as the school are 9 percentage points less likely to be absent, ceteris paribus. At the school level there are three robust associations: 1. Increasing the degree of parental involvement (contributions, and frequency of parents and PTA meeting, reduces absence rates ceteris paribus. 2. Increasing the number of non-governmental schools within 1km of a school reduces absence rates by about 1 percentage point, ceteris paribus. 3. Teachers in schools that practice multi-grade teaching are nearly 10 percentage points more likely to be absent, ceteris paribus. The other school level controls such as teacher enrollment, pupil-teacher ratios, school location and measures of school infrastructure are not significantly associated with teacher absence. At the district level, there are no robust associations. District poverty rates are positively but insignificantly related to absence rates. Similarly, measures of district monitoring capacity do not seem to exert any real influence on absence rates.

One reason to be skeptical of these results is that a single visit to the school does not provide a consistent estimate of teacher-level absence rates. Measurement error (assumed to be uncorrelated with the variables of interest) reduces the precision of the estimates, and possibly conceals relationships of interest. We investigate this by running a regression at the school level, where the dependent variable

Ajk =  +  Sjk + Qk + j + ik

Where Ajk is the fraction of teachers in school j in district k that are absent. Sjk are school level characteristics and Qk are district level factors and j is unobserved fixed school characteristics. We estimate this model using ordinary least squares.90 The results are 89 The infrastructure index is the constructed as the leading eigen value of principle component analysis of the availability in the school of the following factors: electricity, water, library, staff-room, playground and computers. 90 The most important concern with this formulation is that unobserved factors are likely correlated with variables of interest and lead to biased inferences.

shown in table A3 in the appendix. 9 specifications are estimated that start by sequentially adding school level factors that affect teacher behavior directly. There are three robust associations at this level: 1. Increasing the number of functioning teacher houses is associated with a reduction in average absence rates, holding other factors, including the number of teachers constant. In particular, an increase in housing units from the current average of 1 house for 7 teachers to 1 house for 2 teachers would reduce absence by 5 percentage points.91 2. Increasing the degree of parental involvement in the school is associated with a reduction in absence rates. A one unit increase in this index (equivalent to a move from the 10th to the 90th percentile) is associated with a reduction in absence rates of 4.6 percentage points. 3. Finally an increase in the number of neighboring non-government owned schools is associated with a reduction in average school absence rates. An increase of 1 non-government owned school within 1km of a school is associated with a 2 percentage point reduction in absence rates.

The result of this school level analysis does not throw up other correlates of interest as we had suspected. Instead, it increases our confidence in a number of policy-amenable factors.

1. School infrastructure. The results of this analysis specifically suggest that the construction of more teacher housing would potentially reduce teacher absence. The channels are myriad, and subject to the caveat pointed out above, include greater proximity to the school that increases the monitoring levels of the community and peers and greater satisfaction (through high remuneration). 2. Parental/community involvement. As with the results pointed out in the introduction, increased involvement by parents and communities is associated with lower absence rates. It is difficult to isolate the sources of this effect. On the one hand, it is possible that community involvement means greater monitoring by the community. On the other, it is possible that parents that care will support a learning environment that increases the utility of teaching and therefore reduces absence. 3. Competition effects. Being surrounded by well-run schools exerts imperatives for schools to perform better. It is also possible that this reflects some of the effects outlined in 2 above (better parents and improved working environment).

Discussion and Conclusion This chapter has presented the results of a survey of teacher attendance in six districts in Uganda at the end of 2006. The results discussed in this study highlight very high levels of teacher absence. 1 in 5 teachers cannot be found in school at any given time. Given the importance of instruction time and the share of teacher remuneration in the budget, high

91 It is possible that this result, like the proximity result in the individual level regressions reflects the ease with which teachers can respond to an attendance check. Teachers close by will be able to appear as though they are present simply because there is an enumerator looking for them.

teacher absence represents a gaping source of inefficiency. In addition, an examination of teachers that are present in the school suggests about a third of teachers are outside of the classroom. While this is plausibly due to end of year examinations, it suggests that interventions that simply keep teachers in school are not likely to raise instruction time to the optimal levels required.92 Even at the lower bounds suggested by the March/April visits suggests that pupils are deprived of a large fraction of the interaction between pupils and teachers that comprises learning. Teacher absence is very heterogenous and reflects a wide array of factors from weak incentives, poor working conditions and a debilitating disease environment. This chapter explores the possibility that teacher absence rates have fallen between 2002 and 2006. We find some evidence in support of this hypothesis, but it is very weak at best. In order to determine trends in teacher absence, better data will need to be collected. The correlates of absence that we identify at the individual level are age, residential proximity and working in the same district of birth/origin. The latter raises an important question about decentralization. One of the major trends in teacher hiring, and indeed in other sectors, is that an increasing fraction of new hires are indigenous to the district. Our results suggest that one possible explanation of this association is that individuals with strong social ties are more likely to be pulled away from work either for personal benefit or to help relatives and friends. School level correlates include greater competition generated by neighboring private schools, increased participation and involvement of parents and teachers in schools that practice multi-grade teaching are more likely to be absent. We do not find any strong evidence that district level factors affect teacher attendance.

What can be done? This section briefly reviews 4 studies that have attempted to address teacher performance either directly or indirectly. Two studies are located in India and the other two are located in Western Kenya.

Teacher Incentives in Western Kenya This study conducted by Glewwe, Illias and Kremer (2003) attempted to measure the impact of a randomized intervention in which high performing teachers would be rewarded with prizes at the end of the school year. Prizes were substantial and included bicycles, mattresses and other household durables. Performance was determined by the rank of the school in the district and the degree of improvement relative to the previous year’s exam results. The authors measured a variety of inputs before and after the intervention. The authors find improvements in test scores in treatment schools. However, this improvement is driven primarily by teachers using different teaching strategies – particularly teaching to the test. There is no significant different across treatment and comparison schools in teacher absenteeism (teacher absence rates are around 20% in these schools).

92 The corresponding numbers from the previous survey range between 18 and 22.5% (the latter corresponds to visits in March and April).

Girls Scholarships in Western Kenya Miguel, Kremer and Thornton (2005) evaluate a randomized intervention in which a series of pupil scholarship possibilities are announced at the beginning of the school year. About 200 girls are eligible for scholarships in 60 schools in two districts (about 15% of the eligible enrollment). The scholarships pay for tuition for the next two years and parents receive an unconditional cash transfer of $12 per recipient. The results of this intervention were quite dramatic. Performance of all pupils, including non-eligible scholarship recipients, perform much better than control schools. In addition, teacher absenteeism falls by 6.5 percentage points in treatment schools. The authors attribute this to an improvement in working conditions engendered by increases in pupil effort.

Monitor-less monitors in India Duflo and Hanna (2006) evaluate a randomized intervention in community schools in which an NGO provides cameras to teachers and institutes attendance-dependent remuneration. Teachers are expected to take pictures every morning, and teachers will be paid depending on the number of “full” days attended. The results of this intervention were surprisingly large. Teacher absence fell by about half from a high of 36% in comparison schools to 18% in program schools. The authors discuss the political- economy of this intervention and conclude that it is not a realistic option for national scale up. Group vs Individual Incentives in India A suite of interventions is evaluated in a randomized-control design in India. The incentives include individual and group-based incentives, additional teachers and block grants (equivalent expenditure). Karthik and Sundararaman (2006) find evidence in support of teacher incentives. Learning outcomes are higher in treatment schools. However, like the teacher incentives study in Kenya, teacher attendance is not affected. Instead, teachers choose to increase “cheap effort” – assigning more homework and practice tests rather than show up.

A number of studies have been carried out in Latin America and show large effects of community involvement on school performance. When control was transferred to the community, so that parents could hire and fire teachers, teacher attendance and test scores went up (see Lewis (2005) for a recent review). While the associations here are very strong, it is difficult to interpret these results as causal.

Bibliography 1. Abhijit Banerjee and Esther Duflo 2005. “Addressing Absence” Journal of Economic Perspectives 2. Chaudhury, N., Hammer, J., Kremer, M., Muralidharan K, and Rogers H. 2005. “Missing in Action: Teacher and Health Worker Absence in Developing Countries” Journal of Economic Perspectives. 3. Das, J., Dercon, S., Habyarimana, J. and Krishnan P. 2006. “Teacher Shocks and Student Learning: Evidence from Zambia” forthcoming Journal of Human Resources Fall 2007 4. Duflo, E. and Henna, R. “Monitoring Works: Getting Teachers to Come to School” mimeo MIT. 5. Ehrenberg, R. G, Ehrenberg R. A, Rees, D and Ehrenberg E, (1991) “School District Leave Policies, Teacher Absenteeism, and Student Achievement” Journal of Human Resources, Vol. 26:1, 72-105 3. Glewwe, P, Nauman I, and Kremer M. (2003) “Teacher Incentives,” mimeo, Harvard University. 4. Jacobson, Stephen, (1989) “The Effects of Pay Incentives on Teacher Absenteeism” Journal of Human Resources 243:2 280-6 5. Jimenez, E. and Y. Sawada. 1998. "Do Community-Managed Schools Work? An Evaluation of El Salvador’s EDUCO Program." Working Paper Series on Impact Evaluation of Education Reforms Paper No. 8. Washington, DC: Development Research Group, World Bank. 6. King, E.M., and B. Ozler. 2001. “What’s Decentralization Got To Do With Learning? Endogenous School Quality and Student Performance in Nicaragua.” Washington, D.C.: World Bank. Mimeo. 7. Kremer, M. and Glewwe, P., 2005. "Schools, Teachers, and Education Outcomes in Developing Countries," forthcoming in Handbook on the Economics of Education 8. Lewis, Maureen. 2005. “Decentralizing Education: Do Communities and Parents Matter?” Center for Global Development. Washington DC. 9. Miguel, T., Kremer, M., and Thornton, R. 2004. “Incentives to Learn” NBER Working Paper #10971, December 2004) 10. Park, A and Hannum E, (2002) “Do Teachers Affect Learning in Developing Countries?: Evidence from Student-Teacher data from China”, mimeo 11. PROBE (Public Report on Basic Education in India) (1999), Oxford University Press.

Table A1: Correlates of Absence: Individual level linear probability model Dependent Variable: Indicator -- Cannot find teacher/teacher absent (1) (2) (3) (4) (5) (6) (7) (8)

Gender -0.004 -0.056 -0.004 -0.006 0.007 -0.019 -0.005 -0.019 (1=Male) (0.018) (0.036) (0.018) (0.020) (0.023) (0.020) (0.022) (0.019) Teacher age, -0.008 -0.009 -0.008 -0.011 -0.013 -0.009 -0.012 -0.010 years (0.003)** (0.003)** (0.003)** (0.003)** (0.004)** (0.004)** (0.004)** (0.003)** * * * * * * Age squared 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 (0.000)** (0.000)** (0.000)** (0.000)** (0.000)** (0.000)** (0.000)** (0.000)** * * * * * * * * Tenure current 0.003 0.003 0.003 0.002 0.005 0.002 0.006 0.003 posting (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) Tenure current -0.000 -0.000 -0.000 -0.000 -0.000 -0.000 -0.000 -0.000 posting squared (0.000)* (0.000)* (0.000)* (0.000)* (0.000)** (0.000) (0.000)** (0.000) * * Completed A- 0.023 0.021 0.023 0.031 0.053 0.025 0.033 0.032 levels (0.021) (0.021) (0.021) (0.024) (0.033) (0.027) (0.035) (0.025) Head teacher 0.033 0.029 0.033 0.042 0.052 0.042 0.060 0.044 (0.034) (0.034) (0.034) (0.041) (0.054) (0.041) (0.054) (0.035) Deputy teacher -0.033 -0.035 -0.033 -0.023 -0.005 -0.017 0.005 -0.016 (0.031) (0.031) (0.031) (0.030) (0.038) (0.030) (0.038) (0.031) Head of dept -0.025 -0.028 -0.025 -0.036 -0.072 -0.001 -0.009 -0.007 (0.027) (0.027) (0.027) (0.028) (0.032)** (0.026) (0.033) (0.030) Teaches in 0.023 0.021 0.023 0.024 0.033 0.036 0.048 0.035 district of origin (0.020) (0.020) (0.020) (0.021) (0.027) (0.022)* (0.028)* (0.021)* Lives in same -0.096 -0.096 -0.096 -0.093 -0.054 -0.102 -0.066 -0.099 parish as school (0.019)** (0.019)** (0.019)** (0.024)** (0.028)* (0.024)** (0.027)** (0.019)** * * * * * * Teacher fluent 0.033 0.032 0.033 0.029 -0.005 0.017 -0.004 0.018 in local language (0.023) (0.023) (0.023) (0.024) (0.029) (0.022) (0.027) (0.025) Teacher is -0.016 -0.050 -0.016 -0.017 -0.033 -0.013 -0.030 -0.011 married (0.023) (0.031) (0.023) (0.027) (0.034) (0.027) (0.033) (0.023) Male * married 0.069 (0.042) Government 0.013 0.039 0.035 0.059 0.041 owned (0.031) (0.033) (0.034) (0.032)* (0.027) School -0.020 -0.020 -0.005 -0.005 0.007 infrastructure index (0.022) (0.025) (0.019) (0.020) (0.019) Distance to 0.021 0.019 0.031 0.015 0.031 transport point index (0.025) (0.030) (0.026) (0.026) (0.021) Parental -0.036 -0.038 -0.040 -0.037 -0.044 involvement index (0.021)* (0.030) (0.023)* (0.031) (0.016)** *

District -0.026 -0.048 -0.003 -0.020 0.008 monitoring index (0.045) (0.061) (0.043) (0.059) (0.030) Urban 0.011 0.041 0.049 0.036 0.008 (0.027) (0.041) (0.039) (0.042) (0.029) Number of 0.003 -0.001 -0.000 -0.004 -0.002 teachers (0.002) (0.003) (0.003) (0.003) (0.002) Pupil teacher -0.000 -0.000 0.000 0.000 -0.000 ratio (0.001) (0.001) (0.001) (0.001) (0.001) School practices 0.090 0.155 0.133 0.210 0.136 multi-grade teaching (0.053)* (0.051)** (0.059)** (0.048)** (0.060)** * * Total number of -0.012 -0.010 -0.007 -0.003 -0.004 NGO schools within 1km radius (0.007)* (0.009) (0.007) (0.008) (0.007) Total NGO -0.002 -0.003 -0.003 -0.004 -0.003 number of schools within 1-5 km radius (0.002) (0.002)* (0.002)** (0.002)** (0.002) District 0.026 0.051 0.034 recognition award (0.035) (0.047) (0.030) District poverty 0.002 -0.000 rates (0.002) (0.002) District -0.233 -0.036 Inspections (0.400) (0.435) District local -1.041 -1.998 council (0.700) (0.851)** Average literacy 0.036 0.024 of grade 4 pupil mums in the school (0.038) (0.035) Constant 0.329 0.365 0.329 0.335 0.395 0.310 0.489 0.361 (0.077)** (0.080)** (0.077)** (0.092)** (0.115)** (0.114)** (0.147)** (0.089)** * * * * * * * * Observations 1643 1643 1643 1574 998 1574 998 1574 R-squared 0.08 0.08 0.08 0.10 0.11 0.11 0.13 0.11 Notes: Standard errors in parentheses significant at 10%; ** significant at 5%; *** significant at 1%. All specifications include controls for day of visit. Specification 8 includes district fixed effects.

Table A2: Correlates of Absence: Individual level probit Cannot find teacher/teacher absent (1) (2) (3) (4) (5) (6) (7) (8)

Gender (1=Male) -0.004 -0.056 -0.004 -0.004 0.008 -0.016 -0.002 -0.017 (0.020) (0.039) (0.020) (0.020) (0.022) (0.020) (0.021) (0.020) Teacher age, years -0.005 -0.005 -0.005 -0.007 -0.009 -0.005 -0.008 -0.006 (0.004) (0.004) (0.004) (0.003)* (0.004)** (0.004) (0.004)* (0.003)* Age squared 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 (0.000)** (0.000)** (0.000)** (0.000)** (0.000)** (0.000)** (0.000)** (0.000)** * * * * * * * * Tenure current posting 0.004 0.004 0.004 0.003 0.013 0.003 0.012 0.004 (0.005) (0.005) (0.005) (0.005) (0.007)* (0.005) (0.007)* (0.005) Tenure current posting -0.000 -0.000 -0.000 -0.000 -0.001 -0.000 -0.001 -0.000 squared (0.000)* (0.000) (0.000)* (0.000) (0.000)** (0.000) (0.001)* (0.000) Completed A-levels 0.026 0.024 0.026 0.035 0.054 0.029 0.035 0.036 (0.024) (0.024) (0.024) (0.025) (0.033)* (0.027) (0.035) (0.027) Head teacher 0.027 0.023 0.027 0.039 0.051 0.038 0.055 0.041 (0.040) (0.040) (0.040) (0.040) (0.051) (0.039) (0.050) (0.037) Deputy teacher -0.034 -0.035 -0.034 -0.022 -0.004 -0.012 0.014 -0.011 (0.029) (0.029) (0.029) (0.029) (0.038) (0.030) (0.039) (0.030) Head of dept -0.021 -0.024 -0.021 -0.031 -0.068 0.000 -0.020 -0.006 (0.028) (0.027) (0.028) (0.026) (0.027)** (0.027) (0.032) (0.029) Teaches in district of 0.023 0.022 0.023 0.025 0.031 0.038 0.043 0.035 origin (0.019) (0.019) (0.019) (0.020) (0.025) (0.020)* (0.024)* (0.020)* Lives in same parish as -0.097 -0.097 -0.097 -0.090 -0.055 -0.102 -0.067 -0.098 school (0.025)** (0.025)** (0.025)** (0.024)** (0.027)** (0.024)** (0.026)** (0.021)** * * * * * * * Teacher fluent in local 0.037 0.036 0.037 0.034 0.005 0.025 0.008 0.025 language (0.022)* (0.022)* (0.022)* (0.024) (0.029) (0.023) (0.027) (0.025) Teacher is married -0.022 -0.058 -0.022 -0.026 -0.045 -0.019 -0.037 -0.017 (0.028) (0.039) (0.028) (0.028) (0.034) (0.028) (0.032) (0.024) Male * married 0.070 (0.048) Government owned 0.010 0.038 0.029 0.055 0.034 (0.033) (0.033) (0.033) (0.027)** (0.025) School infrastructure -0.019 -0.014 -0.005 0.002 0.006 index (0.025) (0.026) (0.021) (0.020) (0.019) Distance to transport 0.023 0.023 0.033 0.020 0.031 point index (0.025) (0.029) (0.026) (0.025) (0.021) Parental involvement -0.030 -0.032 -0.034 -0.029 -0.037 index (0.017)* (0.024) (0.019)* (0.025) (0.015)** District monitoring -0.024 -0.043 -0.003 -0.013 0.006 index (0.040) (0.049) (0.037) (0.045) (0.028) Urban 0.015 0.049 0.058 0.038 0.015 (0.032) (0.049) (0.049) (0.050) (0.032) Number of teachers 0.003 -0.001 -0.000 -0.005 -0.002 (0.003) (0.003) (0.003) (0.003)* (0.002) Pupil teacher ratio -0.000 -0.000 0.000 0.000 -0.000

(0.001) (0.001) (0.001) (0.001) (0.001) School practices multi- 0.085 0.153 0.149 0.278 0.160 grade teaching (0.059) (0.060)** (0.077)* (0.081)** (0.084)* * Total number of NGO -0.015 -0.012 -0.008 -0.001 -0.004 schools within 1km radius (0.009)* (0.011) (0.009) (0.010) (0.008) Total NGO number of -0.003 -0.004 -0.004 -0.005 -0.004 schools within 1-5 km radius (0.003) (0.003) (0.003) (0.003)* (0.003) District recognition 0.035 0.060 0.043 award (0.039) (0.049) (0.033) District poverty rates 0.002 -0.001 (0.002) (0.002) District Inspections -0.289 -0.317 (0.412) (0.418) District local council -0.924 -1.847 (0.674) (0.761)** Average literacy of 0.053 0.040 grade 4 pupil mums in the school (0.042) (0.037) Observations 1643 1643 1643 1574 998 1574 998 1574 Notes: Standard errors in parentheses significant at 10%; ** significant at 5%; *** significant at 1%. All specifications include controls for day of visit. Specification 8 includes district fixed effects.

Table A3: Correlates of Absence: School level regressions

(1) (2) (3) (4) (5) (6) (7) (8) (9)

No of -0.013 -0.012 -0.013 -0.012 -0.011 -0.011 -0.006 -0.011 -0.011 functioning teacher housing units (0.005)* (0.005)* (0.005)* (0.005)* (0.005)* (0.005)* (0.007) (0.005)* (0.005)* ** ** ** * * * * * Number of 0.002 0.002 0.002 0.002 0.002 0.002 -0.002 0.001 -0.001 teachers (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) (0.003) (0.004) Pupil teacher -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 ratio (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Multi-grade 0.075 0.077 0.075 0.084 0.077 0.042 0.094 0.100 School (0.060) (0.060) (0.061) (0.062) (0.062) (0.080) (0.065) (0.065) Other teacher 0.012 0.014 0.013 0.005 -0.010 0.007 0.019 conditions index (0.024) (0.026) (0.025) (0.025) (0.029) (0.027) (0.027) Government 0.004 -0.001 -0.003 0.022 0.006 0.014 school (0.035) (0.036) (0.036) (0.041) (0.037) (0.037) Distance to 0.008 0.004 0.020 0.003 0.024 0.024 transport points (0.027) (0.027) (0.028) (0.034) (0.028) (0.029) Urban -0.021 -0.016 0.028 0.055 0.053 0.030 location (0.040) (0.039) (0.043) (0.053) (0.052) (0.043) PTA/parents -0.046 -0.045 -0.042 -0.048 -0.053 involvement index (0.022)* (0.022)* (0.026) (0.023)* (0.023)* * * * * District/minist -0.000 -0.002 -0.035 0.008 0.015 ry supervision index (0.040) (0.039) (0.054) (0.040) (0.040) Total number -0.019 -0.016 -0.018 -0.014 of NGO schools within 1km radius (0.010)* (0.012) (0.011) (0.011) Total NGO -0.002 -0.003 -0.003 -0.002 number of schools within 1-5 km radius (0.003) (0.003) (0.003) (0.003) District 0.001 poverty rates (0.002) District 0.134 Inspections average (0.424)

District local -0.921 council involvement average (0.751) District 0.019 recognition award (0.041) Average -0.007 mother literacy (0.051) Constant 0.181 0.209 0.209 0.212 0.208 0.238 0.272 0.257 0.268 (0.033)* (0.052)* (0.052)* (0.056)* (0.056)* (0.057)* (0.080)* (0.094)* (0.064)* ** ** ** ** ** ** ** ** ** Observations 158 155 155 153 153 153 96 153 153 R-squared 0.05 0.06 0.06 0.06 0.09 0.13 0.14 0.15 0.17 Notes: Standard errors in parentheses significant at 10%; ** significant at 5%; ***

ANNEX 4: FUNDING FORMULAS FOR BASIC EDUCATION.

Box 1: New Zealand Funding Formula

Background

New Zealand schools’ Board of Trustees manages educational services and receives resources directly from the Ministry of Education via a funding formula. Schools’ Boards of Trustees consists mainly of elected parents and school principal, and they are responsible for the day-to-day administration of the school and overseeing its resources without other intermediary or regional agencies being involved in the decision- making process. In order to ensure sound accountability, all Board of Trustees are required to produce an annual financial report which are to be made available to the Ministry of Education as well as to the school community.

Schools receive funding to cover their operational activities and most of their required inputs, but the direct payment of teachers’ salaries in the majority of the schools remains to be the responsibility of the central government. Since 1996, schools may choose to be funded directly for the costs of teaching personnel (Directly Resourced Schools Program) or to have the component of teachers’ salaries be paid directly by the Ministry of Education (Centrally Resourced Schools Program). Although the Directly Resourced Schools (DRS) program devolve a higher level of resources directly to the school level and thereby give the Boards of Trustees greater flexibility in making decisions in resource allocations, most schools choose the Centrally Resourced School (CRS) approach for allocating teachers’ salaries. School budget allocation in New Zealand is divided into three distinct parts: operational funding, major capital projects, and staffing of teaching personnel and senior managers. The formula of resource allocation from the central government to schools consists of both operational and teaching personnel for DRS schools, and only operational funding for all other schools, but funding for major capital projects is retained at the central level.

The operational funding in New Zealand includes all four components of a funding formula: basic student allocation, curriculum enhancement, student supplementary educational needs, and school site needs. The funding framework takes into consideration actual costs of educational resources and develops cost structures in terms of needs of different groups from various schools.

Component 1: Basic student allocation Teaching and non-teaching personnel, relief teaching funding (e.g. costs of employing substitute teachers), and operational funding, are allocated predominantly by the proportion of student enrollment. In the case teaching personnel, the staffing entitlement is allocated based on a teacher to student ratio of each grade level, and the allocation of non-teaching personnel, e.g. management staffing, is calculated separately based on a weighted enrollment factor which varies by different level of education. Costs of employing substitute teachers may not be a recurrent one, but substitute teacher funding are provided on a teacher-entitlement basis and rates are differentiated by the size of the core staffing and management component allocated to each school.

Component 2: Curriculum Enhancement Additional staffing and resources are provided for schools with advanced curriculum or specialized programs such as computer technology or Maori language programs. In addition, per-student funding rates increase with as students get older because funding for school-to-work is built into the per-student funding component at a rate of grade level because funding for school-to-work program is built into the per-student funding component at a rate of $17.19 for Y9-10 and $19.06 for Y10-15.

Component 3: Student supplementary educational needs Targeted Funding for Educational Achievement (TFEA) provides additional per-student funding to schools that cater to students from socio-economically disadvantaged communities. Further, additional funding is made available to upper secondary school students for resources on career guidance. This funding is shown to benefit disadvantaged students most because they have a tendency to leave schools early and are therefore at a greater risk of unemployment. Disadvantaged students are evaluated on the basis of

household income, the concentration of workface in manual and unskilled occupations, household crowding, parents’ academic qualifications, family welfare benefit dependency, and the proportion of school enrollment made up of Maori and Pacific Island students. These indicators were selected because investigation shows that they have a high correlation with average school achievement scores. Since the first five indicators are drawn from the Population Census data instead of school-level survey, information used to construct the composite indicator cannot be manipulated unfairly by schools.

Ongoing Resource Scheme (ORS) ensures a school that caters to students with disability has sufficient resources and additional teaching staff to support students so that they can participate both in special and mainstream schools. Students are assessed by independent verifiers on the basis of their learning support needs rather than type of learning disability. Since students are assessed by an independent entity, this preserves the integrity of the indicators. Once this group of students has been identified, they are further categorized into level of needs. Funding allocation is therefore differentiated based on the level of students’ needs.

Component 4: School site needs

The Ministry of Education is responsible for major property upgrades and the establishment of new classrooms. However, funding is made available to the schools’ Boards of Trustees for school site maintenance, utility costs (e.g. heat, light, and water), and minor improvement projects. Specific site characteristics are taken into consideration for funding allocation, such as the size and location of school sites. Small schools (with less than 160 student enrollment), for example, benefit from additional cores staffing allocation as well as a supplementary ‘base funding’, which is allocated on the basis of an enrollment range designed to offset ‘diseconomies of scale’ for very small schools. This type of assistance guarantees delivery of education to very isolated areas, however, one of the unintended outcomes is that it could prevent small schools from amalgamating with other smaller schools situated in the same area. In the case school site location, schools are compensated for higher costs incurred in the delivery of educational services as a result of being located in isolated rural communities. Schools are categorized as being ‘isolated’ if it is at least 30 kilometers away from a trade and service center, with a population of 2,000 or more, and if there is an absence of schools within close proximity offering educational services to students of the same education level or age group.

Box 2: South Africa Formula-Based Distribution of Resources

National-to-Provinces Funding Allocation Provinces in South Africa receive unconditional block grants from the national government determined by the equitable shares formula. The level of funding transferred to each province is determined in a two-step process. First, the national government decides how to divide the total national revenue among different levels of government (e.g. national, province, and municipal). Second, the pool of funds available to provincial government is distributed based on a weighted average demographically driven funding formula, where the weights reflect the proportion of national spending allocated to each major social service sector, such as education, health, and welfare. In the case of education, equitable shares for each province are computed based on a 40 percent weight, and determined by actual student enrollment and school-age population, with the latter given twice the weight the former. This formula represents a compromise to overcome the problem of over student enrollment and under student enrollment. For example, a formula that is based on school-age population alone will not give provinces the incentive to reduce the proportion of school-age population not enrolled in school. On the other hand, a formula that is based on actual student enrollment alone may encourage inefficiency and over student enrollment.

Non-Personnel and Non-Capital Recurrent Province-to-Schools Funding Once each province receives its fiscal transfer, it determines how the cost allocation of various categories will fit within the overall provincial education budget. School budget allocation in South Africa is divided into three major categories: non-personnel and non-capital recurrent, major capital, and personnel expenditures. Financing of schools by provincial governments is determined by the National Norms and Standards for School Funding, which only imposes distributional requirements to provinces but does not

stipulate a minimum basic allocation per student. This national program was designed to give poorer schools more non-personnel and non-capital recurrent funding (e.g. utilities, maintenance of building, teaching materials, and non-emergency building repairs) as a means or promoting greater equal educational opportunity.93 However, non-personnel funding are limited since it only consist of what is left after each province has meet all its personnel costs commitments, which usually amounts to an average of about 90 percent of the total provincial education budget.94 In order to distribute the funds in a progressive manner, it requires each Provincial Department of Education (PED) to first rank all its schools according to a poverty index, which is calculated by applying a 50 percent weight to the relative poverty of the school community (as measured by the characteristics of the parents of students attending the school or characteristics of the community where the school is located in) and a 50 percent weight to the physical conditions of the school itself (as measured by classroom to student ratio and access to basic services such as water and electricity), and then dividing the list of schools in five quintiles, from poorest to least poor. Allocation of resources for each component in the formula will be made on a per-student basis favoring the poorer segment of the population. Based on this formula, 60 percent of available recurrent non-personnel and non-capital resources will go to 40 percent of the poorest schools in each province.

Capital Non-Recurrent Funding Non-recurrent capital resources for construction of new schools and classrooms are retained at the provincial level and are targeted at the neediest population. The objective of this formula is to eliminate backlogs of physical facilities and provide sufficient school places in all provinces. Similar to the formula in which recurrent non-personnel resources are allocated, PED rank geographical areas from neediest to least needy based on proportion of children who are not enrolled in school or who are in existing overcrowded schools, then prioritize funding allocation to the neediest and most crowded areas.

(Note: If you think this section is too long already you may choose not to include the paragraph below) There are four weaknesses to this formula-based distribution of resources. (1) The ability to effectively implement this program varies across provinces because not all provinces have the technical and analytical capacity to rank schools according to the specified criteria. (2) Non-personnel funding are limited since it only consist of what is left after each province has meet all its personnel commitments. (3) Redistribution of funds had to occur within provinces, but the degree of distribution possible within each province depended on its mix of schools. For example, a poor province with a few wealthy schools could do much less than a rich province with a more even distribution of wealthy and poor schools. As a result, provinces vary tremendously in the funds available for learners in the poorest schools. (4) The program is designed to redistribute funds for recurrent spending and not on capital spending

Personnel (Teaching and Non-Teaching Personnel) The PED is responsible for providing teaching and non-teaching personnel (e.g. administrative and support staff) to the schools. The Ministry of Education is responsible for determining the norms and standards for the provision of teaching personnel at the school level but not non-teaching personnel. Teachers are provided to schools on a formula basis driven by type of curriculum, number of students, and schools’ circumstances. The National Norms and Standards recommend achieving personnel to non-personnel cost ratio of 85:15 by the year 2005 to improve adequacy and equitable finance of non-personnel education services.

School Fees Independent schools also receive funding from the PED on a pro-poor basis, and anchored on the per- student spending level in public schools. The level of subsidy level, however, is reduced for schools charging high fees since it is indicative of the socio-economic well-being of a school’s community. School fee level is determined by the School Governing Body (SGB) fee revenue is used to supplement state’s public funds to improve the quality of education (e.g. recruiting additional staff). Parents qualify for full exemption or partial exemption of school fees depending on the level of the combined annual gross income of both parents.

93 Fiske and Ladd (2004, pp. 116). 94 Fiske and Ladd (2004, pp. 116).

Box 3. Chile Per-Capita Funding Formula

Chile decentralized public education to municipalities in 1981, but the central government remained responsible for financing education. The Chilean education system is mixed, that is, schools may be public (municipal), private subsidized, or fully private. The central government finances education by transferring a formula-based per-capita subsidy (Unidad de Subvención Educacional or USE) to both private subsidized and public schools. However, in the case of public schools, resources are transferred to the Municipal Department of Education or the Municipal Department who then manages and provides for schools.

Non-Personnel Recurrent Funding

The Chilean per-capita subsidy consists of three main components: basic student allocation, supplementary educational needs, and school site needs. The per-capita funding is a base subsidy weighted according to the grade level of students, type of school attended by students (e.g. vocational, adult, and boarding), special needs of students, and location of schools. The funding weights increase from primary to secondary school, but the weight is higher for regular schools compared with adult schools (see Table below). Of all the different categories of students, children with special needs receive the highest weight on per-capita subsidy. In terms of school site needs, a ‘zone assignment’ factor is included in the per-capita subsidy since 1974 to take into account high cost or disadvantaged areas of the country. Another factor that accounts for school site needs is the ‘rural factor’, which was included in the per-capita subsidy since 1987 to provide supplemental funding to small schools (less than 85 students). However, it is included only if the school is located more than five kilometers away from the nearest urban center and other schools of the same type. The reason for restricting its application to the rural area only is to avoid funding schools that have diminishing enrollments because they are underperforming. Further, extending its application to the urban area would eliminate the incentives to increase student enrollment and insulate schools from competition.95

The monthly per-capita subsidy that schools receive is based on the average attendance for the previous three months, and there is a penalty for misrepresenting student attendance report. A school’s monthly funding amount is calculated as follows: (Average Months Attendance) x (USE) x (student weight) x (zone weight) x (rural weight). Private and public schools both receive per-capita subsidy on an equal basis, but if private subsidized schools charge fees, the per-capita subsidy they receive will be reduced accordingly. In order to prevent private subsidized schools from excluding students who cannot afford to fees, a compulsory scholarship scheme was introduced since 1997. Private subsidized schools and the government both contribute to the scholarship fund and the scholarships can be awarded either in full or partial exemptions.96

The Chilean per-capita funding formula is intended to have both equity and market regulation function. It has an equity function because it takes into account cost differentials due to level of schooling, supplementary educational needs, and regional and school-site cost variations. However, there are two weaknesses to the per-capita funding formula. First, while the weights are intended to reflect variation in the cost of providing education, they are not based on a detailed analysis of schooling costs (Parry, 1997). On the other hand, the weight assignment for rural schools has been improved recently through the development of cost models attempting to repoduce cost structures of schools located in sparsely populated areas.97 Second, although the per-capita subsidy is supposed to have a market regulation function, the funding formula does not directly transfer incentive to schools simply because school managers do not have much financial and managerial discretion, and the municipal administration, in a way, acts as a middleman between the Ministry of Education and the schools.

Capital Non-Recurrent Funding

95 Gonzalez, (2005). 96 Gonzalez, (2005). 97 For more details on subsidy correction for rural areas in Chile, see Gonzalez (2005).

Since the per-capita subsidy system only covers operating costs of schools but not capital expenditures, the FNDR (National Fund for Regional Development) was designed for municipal schools to finance capital facilities. These funds are targeted to regions on a needs-based basis and are for municipal schools only. Private schools on the other hand finance their own capital facilities from small fees, donations from parents, business or church congregations, and bank loans (Parry, 1997). Municipalities compete with other municipalities within their region for funding by submitting a proposal for new schools, health centers or other related investments on capital facilities to the regional representative (SEREMIS) of the Ministry of Education.

Table 1. Weights on per-capita subsidy Categories Weights 1980 Weights 1989 Weights 1992 Pre-primary 0.880 0.909 0.909 Primary Education, Grades 1-6 1.000 1.000 1.000 Primary Education, Grades 7-8 1.077 1.107 1.107 Primary Education, Adults 0.308 0.316 0.474 Secondary Education 1.210 1.245 1.245 Secondary Education, Adult 0.365 0.375 0.563 Special Students (Mentally or Physically Challenged) 2.250 2.312 3.000 Vocational-Agriculture 1.210 1.245 1.970 Vocational-Industrial 1.210 1.245 1.480 Vocational-Commercial 1.210 1.245 1.300 Boarding School (Meals) -- -- 0.125 Source: Parry, Tarry. (1997). “Achieving Balance in Decentralization: A Case Study of Education Decentralization in Chile.” World Development, Vol. 25, No. 2, pp. 211-225.