DOCUMENT RESUME

ED 314 833 EA 021 521

AUTHOR Tsang, Mun C.TITLE Cost Analysis for Educational Policymaking: A Review

of Cost Studies in Education in Developing Countries.BRIDGES Research Report Series, No. 3.

INSTITUTION Harvard Univ., Cambridge, NA. Inst. for InternationalDevelopment.; Michigan State Univ., East Lansing.Coll. of Education.

SPONS AGENCY Agency for International Development (IDCA),Washington, DC. Bureau of Science and Technology.

PUB DATE Oct 88CONTRACT DDP-5824-A-5076NOTE 50p.; The Basic Research and Implementation in

Developing Education Systems Project (BRIDGES) isdirected by the Harvard Institute for InternationalDevelopment and the Harvard Graduate School ofEducation.

PUB TYPE Information Analyses (07u) -- Reports -Evaluative/Feasibility (142)

EDRS PRICE MF01/PCO2 Plus Postage.DESCRIPTORS *Cost Effectiveness; Cost Estimates; *Developing

Nations; Economic Development; Educational Economics;*Educational Finance; Elementary Education;Expenditures; Foreign Countries; Human Capital;Public Education

ABSTRACT

The cost of education to a country consists of totalpublic education expenditures, total direct private cost, and totalindirect private cost measured in terms of foregone earnings;improperly estimated educational costs focus only on governmenteducation expenditures. Considerable progress has been made inconceptual understanding of educational costs in the past 2 decades.Factors that determine the total amount of resources devoted toeducation and factors that determine expenditures within educationinfluence educational costs. With current tight budgetary constraintsand unmet demand for education in developing countries, the need tocontrol costs and to improve efficiency is very pressing. Anawareness of the disparities between educational costs in differentlocalities is necessary for decisionmakers to design proper policiesappli-.able to diverse settings Applications of cost-benefit aodcost-effective analyses address directly the concern regardinginefficiency in education; however, analyses of education indeveloping countries ere often plagued by unreliable and incompletedata. For educational policymaking and planning, data on educationalcosts, auantities, prices, and norms, as well as socioeconomic data,are required. The obvious cost of managing a database of educationalcosts is likely to be more than compensated for by the gain ofbetter-informed decisions. (132 references) (KM)

***********************************************************x*******A**** Reproductions supplied by EDRS are the best that can be made *

* from the original document. *

********************************r**********Y4***************************

trr 1.

yn

un A

pv 4_ -1.

ttit

BRIDGESBasic Research and Implementation in DevelopinG Education Systems

A project of the Harvard institute for International Development,the Harvard Graduate School of Education,

and the Office of Education. Bureau for Science and Technology.United Sto;es Agency for International Development

U.S DEPARTMENT OF EDUCATIONOffice ol Educational Research and ImprovementEDUCATIONAL RE SOURCES INFORMATION

CENTER (EPIC)

document has been reproduced asrimed from the person or organization

originating itr Minor changes have been made to improve

reproduction quality

Points of view or opinions e'ated in this document do not necesscriiv represem of hewOE RI position or policy

\

2

The Basic Research and Implementation in DevelopinG Education Systems Project (BRIDGES) is directed by the HarvardInstitute for International Development and the Harvard Graduate School of Education, under Cooperative Agreement No. DDP-5824 -A 50/6 with the Office of Education, 13umau for Science and Technology, United States Agency for International Development.Also participating h the Project are the Institute for International Research, Michigan State University, the Research Triangle Institute,and Texas Southern University,

The BRIDGES Group includes educators, researchers, planners and pelicymakers committed to improving opportunityand quality in Thirld World schools. The goal of their coilaborative effort is to identify policy options that will increas.:" children saccess to schooling. reduce the frequency of early school leaving and repetition, improve the amount and quality of what islearned, and optimize the use of fiscal and educational resources.

The FRIDGES Research Report Series is edited by the Harvard Institute for Internal.onal Development iho Series is acollection ot reviews of the state-Who-art in research, and original research reports on basic education in developing countries.Each review summarizes research about a particular policy issue and suggests policy options Original reports on BRIDGES-sponsored research pre ent new informadon about the impact and costs of specific alternatives that the reviews have identifiedas most promising.

The views expressed in this document are those of the author and do riot necessarily reflect those ot the United StatesAgency for Internationdl Development.

"Cost Analysis forEducational Policymaking:

A Review of CostStudies in Education inDeveloping Countries"

Mun C. Tsanu

Mun C. Tsang is Assistant Professor of Education, Michigan State university, The author acknowledgesthe helpful comments of Noel McGinn, Henry Levin, James Mosonlhal, and two anonymous reviewers,

as well as the assistance In literature search provided by Mauna Agyemang-Mensa.

Cost Analysis for Educational Policymaking:A Review of Cost Studies in Education inDeveloping Countries

Foreword

This paper is intended primarily for educators andofficials in ministries of education, rather than justfor economists and ministries of finance. That is notto say that the latter would not benefit from readingthe gaper. On the contrary, they will find much thatis useful in their work. The paper meets the highstandards that economists set in their profession,and decisions made in ministries of finance have far-reaching effects on the education sector.

The title and substance of the paper are likely,however, to put off some educators who are confusedby the special language of economics, and inclined tobelieve that because education deals with the trans-formation of human beings, discussion of the cost ofeducation is not appropriate. There is a sincere beliefthat to specify costs is to dehumanize the process ofeducaticn.

This belief is unfortunate because without consid-eration of costs, educators too often give up their rightand ability to contribute to important decisions aboutthe content and intens:*Ly of the education process.Decisions about educational policy are left to thoseoutside the practice of education who while expert intheir profession, have only a limited understandingof both meaning and method in education. Too oftenit is non-educators who make the final decisions onamounts and typ..s of teacher training, instructionalmaterials, school construction, instructional media,and other factors that affect the practice of educators.Too often the arena of decision making has beenabandoned by educators who wish only to discusswhat should be done, and not what can be done.

The avoidance of discussion of costs and econom-ics in education is often based on the belief that thesediscussions "technify" education, P " take away itsfundamentally human character. A articular stum-bling block for many is the concept of 4fficiency, usedby economists to describe the relationship betweenamounts of inputs and outcomes. There is a legiti-mate concern that the mechanical use of measures ofefficiency to choose options restricts human freedom.On the other hand, measures of efficiency requirethat goals and objectives be specified and designed interms of their relative importance. That is, the meas.

urement of efficiency requires a specification of val-ues, of the meanings that we attach to various kindsof outcomes. Similarly, the specification of costs ofinputs can require choices. As Tsang points out, "theproper definition of costs of an input to educationis...the value of the input in its best alternative use."Educators are highly qualified to identify those alter-native uses, and to specify which are best in terms ofoutcomes. Once those choices have been made theeconomist, using criteria specified by the choice-maker, can begin to calculate costs.

In other wordsand this theme is reoeatedthroughout the paperdiscussion of educational costsinvolves a political process in which social and educa-tional values are necessarily brought into play. The

Too often it Is non-educators who makethe final decisions on

teacher training, in-structional materials,

and other factors thataffect the practice

of educators.

paper is not about that political process, but itsattention to the specification of costs focuses andrefocuses our attention on issues of value and choice.By comparing costs of educational practices withsimilar outcomes, we are forced to consider what weare willing to give up in order to maintain a moreexpensive (less efficient) practice. By comparingdifferent outcomes of practices with similar costs, weare forced to consider which outcomes we prefer. Weare not required to choose what is more efficient, butwe are required to be more conscious about the choicethat we mako.

There is another risk in a paper that focuses onthe costs of education. The danger is that, by empha-sizing more efficient use of existing resources, atten-

tion will be drawn away from the problem of society'soverall level of support for education. Some arguethat the fundamental issue facing education in thepoor countries is not the choice of educational inputs,but the lack of educational inputs. The problem,these critics observe, is not one of efficiency, but oneof political commitment to the provision of educationfor everyone. Others argue that the first issue is whatr atcomes education should generate, not how to gen-erate them at the lowest possible cost.

In some countries, it is true, total spending oneducation falls below the norms specified by theUnited Nations, and those countries are maldnglittleif any progress toward objectives of universal pri-mary education. Often these are countries in whicha large proportion of public spending is for arma-ments, confirming once again the adage that if acountry does not spend on education, it will have tospend on guns. A reversal of priorities in favor ofeducation would make significant improvementspossible in both the quantity and quality of educa-tion. At that moment, it could be argued, issues ofefficiency would be important in an effort to attemptto meet all priorities.

But discussion of educational priorities goes oneven in countries that spend relatively little on edu-

iv

cation and, as argued above, there is an intimate linkbetween the concepts of priorities and efficiency. Forexample, when examination of the costs of educationindicates that one student year in a university costs10 to 20 times more than one student year in aprimary school, some shift must take place in therelative priority assigned to those two levels of thesystem. Further, when cost-benefit analysis sug-gests that returns to society are greater for invest-ments in primary education than higher education,there is a shift in the weight of arguments offered bypolitical groups favoring one or another option. Inother words, we cannot separate discussion of endsfrom discussion of means.

The objective of this paper is not, therefore, todisplace discussion of what the goals of an educationsystem should be, anymore than it is to offer formulasor algorithms that replace human choice among al-ternative education policies. Instead, this paperprovides a language that, properly used, can broadenthe dialogue among the inttny groups concerned foreducation, and expand our understanding of therange of alternative policy options among which wecan choose.

Noel McGinnOctober, 1988

Cost Analysis for Educational Policymaking:A Review of Cost Studies In Education inDeveloping Countries

Table of Contents

Executive Summary 2

Section I: Introduction 5Cost Studit s in Education in Developing Countries 5An Econonic Framework for Analyzing Educational Cost and 6

Efficiency

Section II: Educational Costing and Economic-Feasibility Testing 8Costing and Economic-Feasi? lity Testing of Educational 8

Interventions in Developing CountriesCost Estimation in Education 10

Concepts of Cost 10Classification of Educational Costs 11

Data Collection and Measurement 12Ways of Expressing Educational Costs 13

Section III: Behavioral Characteristics of Educational Costs 16Educational Expenditures and Unit Costs: Patterns and 16

DeterminantsPatterns and Determinants of Educational Expenditures 17Patterns and Determinants of Unit Costs 20

Utilization of Educational Inputs 23Rates of Utilization of School Resources 23Educational Cost Functions and Economies of Scale 24

Behavioral Characteristics of Educational Costs and 26Educational Policymaking

Section IV: Cost-Benefit and Cost-Effectiveness Studies in 20Education

Cost-Benefit Studies in Education in Developing Countries 28Cost-Effectiveness Studies in Education in Developing 30Countries

Section V: Policy Recommendations 35

References 37

Cost Analysis for Educational Policymaking:A Review of Cost Studies in Education InDeveloping Count/ les

Executive Surnmaty,

What are the costs of education? What are themajor determinants of educational costs? In whatwajs can cost analysis improve policymaking ineducation? And what are the informational needs forcost analysis in education?

These are the questions we set out to address inthis paper. We summarize what we have learnedfrom a review of educational cost studies in develop-ing countries in the past two decades.

An important contribution of the economic analy-sis of educational costs is the concept of opportunitycost. This enables us to estimate the real cost ofeducation which includes not only monetary expendi-tures, but also the value of other foregone opportuni-ties. In particular, the total cost of education to acountry consists of total public educational expendi-ture, total direct private cost, as well as total indirectprivate cost measured in terms of foregone earnings.Past cost studies have sometimes improperly esti-mated educational costs by focusing on governmente,-.9nditures on education only.

Previous experience indicates that there is nosingle response to the question of what the costs of

An important contribu-tion of the economicanalysis of educationalcosts Is the concept ofopportunity cost.

education are. In practice, what educational costsshould he measured depends on the decision contextin which the cost analysis is performed. In general,cost estimation is influenced by two issues: the costto whom, and the choice between average-cost analy-sis and marginal-cost analysis. For both accountingand analytical purposes, educational costs are usu-ally classified into distinct categories, such as recur-rent costs and capital costs, as well as personnel and

2

nonpersonnel costs. To reflect changes in the pricelevel, educational costs are usually expressed in bothcurrent dollars and constant dollars. Also, educa-tional costs are often compared on e per unit basis.The choice of educational unit depends on the pur-pose ofthe comparison. The most common basis is thecost per pupil enrolled.

In short, considerable progress has been made inour conceptual understanding of educational costs inthe past two decades.

A multitude of factors affects educational costs.Analytically, we can divide these factors into twogroups: factors that determine the total amount ofresources devoted to education, and factors that de-termine expenditures within education.

Total public educational expenditure represents alarge portion of the total amount of resources devotedto education. In general, it is influenced by factorsthat determine the supply of and demand for educa-tion. Supply factors include the rate of inflation, therange of competing demands of other public services,the rate of growth of the economy, tax revenues foreducation, the impact of foreign trade, as well as theavailability of foreign aid to education. Demandfactors, on the other hand, relate to the growth inpopulation and its age distribution, the importanceof education for social mobility, concern for equalityof educational opportunities, the choice and develop-ment of technology for economic production, andothers. These supply and demand factors are verycomplex and they are often beyond the control ofeducational decision makers. It does not imply,however, that all these factors are equally importantin all settings. Their relative importance is likely tovary over time and among countries.

The other group of factors is concerned with theinternal operation of education. Expenditures withineducation are strongly influenced by the technologyof educational production, compensation for teach-ors, the extent of utilization of educational inputs, aswell as rates of dropouts and repetitions. These arethe factors over which educational decision makershave more control, and thus are the targets of e(li.:ca-tional policies. With the current tight budgetary

...sound economic research and evciu-ation can Improve the efficiency in the

allocation of resources in education.

constraints and unmet demand for education indeveloping countries, the need to control costs andimprove efficiency is obviously very pressing in thesecountries today. Since the amount of discretionaryrecurrent expenditure in a given year is usuallysmall, educational policymakers will do well to beaware of the cost implications of past and present de-cisions, and to adopt a long-term perspective in theirplan to control costs and improve efficiency.

Our review has also documented the range of ap-plications of cost analysis which can contribute tobetter policymaking in education.

Cost estimation has been applied to a wide rangeof situations, from the costing of an educationalproject to the costing of a national education plan,and from the costing of a pedagogical intervention inthe traditional classroom to the costing of out-of-school distance education. For a given educationalintervention, cost estimation informs the policy-maker about: (1) the total cost required, (2) theeconomic feasibility of the intervention, (3) the short-run and long-run cost implications, and (4) the distri-bution of the cost burden. A sound cost analysis mayalso reveal serious policy errors that have to be dealtwith.

It also pays to understand the behavioral charac-teristics of educational costs by regularly construct-ing and examining educational cost indicators and bystudying resource utilization in schools. A close ex-amination of the patterns of educational costs panuncover opportunities for improving the efficiency ofeducational investment. An awareness of the dis-parities in educational costs in different settings isnecessary for decision makers to design proper poli-cies applicable to diverse settings. And the disclosureof areas of excessive wastage or underutilization ofeducational resources may lead to actions that reduceeducational costs without affecting school quality oractions that increase school output without incurringadditional cost.

Moreover, the applications of cost-benefi. .ialysisand cost-effectiveness analysis address directly theconcern about inefficiency in education. Cost-benefitanalysis can be used to assess the external efficiencyof education while cost-effectiveness analysis dealswith issues of internal efficiency. Prominent applica-tions to date include the estimation of the rates ofreturn to different levels and types of education, aswell as the evaluation of new educational media. Butcost-effectiveness studies of the traditional school arestill lacking. Efforts should be made to improve the

3

research basis of cost-effectiveness analysis and itsutilization in educational policymaking.

Previous experience indicates that the practice ofcost analysis in educational policymaking often fallsshort of its potential usefulness. There are "political"barriers that may not be easily penetrated. Yet thereare "t difficulties that are amenable toohm, ... For example, efforts can be made to incorpo-rat cost analysis in educational policymaking, totrain competent cost analysts, and to improve thedatabase for policy analysis.

In short, sound economic research and evaluationnan improve the efficiency in the allocation of re-sources in education.

Finally, analysis of education in developing coun-tries is often plagued by unreliable and incompletedata. In many developing countries, central-govern-ment budgetary data are what is available and acces-sible for cost analysis. Data are often not available onprivate costs, on costs at the school level or otherlevels of government, and on the relevant categoriesof costs. Efforts to collect these data should be encour-aged.

Given the wide range of applications of cost analy-sis, it is impossible to specify fully the data needs ofcost analysis. But for most applications for educa-tional planning and policymaking, wu can identifyfive kinds of educational data that are often required:

(1) Data on Educational CostsThe costs of education are supported by public,

private, and foreign sources. They can be classifiedinto institutional costs and household costs. Institu-tional costs are divided into recurrent costs andcapital costs. Recurrent costs are broken down into amatrix of input items and input functions. Householdcosts include direct and indirect cost items. Time-series data on institutional and household costs are tobe collected for various levels of education, types ofinstitutions, various levels of government, and incurrent and constant dollars.

(2) Data on Educational QuantitiesThese refer to the quantities of inputs to and out-

puts of education. They include data on student en-rollmen s, graduation, repetition and dropout rates,number of teachers and other school personnel (byage, experience, and qualification), and physicalinputs. They are used in costing educational inter-ventions. Also they can be combined with cost data toconstruct indicators of educational costs (e.g., unitcosts) for diagnostic purposes.

(3) Data on Educational PricesThese refer to the prices of school inputs. They

include information on the salary structure and othercompensations for teachers and other school person -uel, and prices for various school inputs. They areneeded in educational costing and in the constructionof educational price indices.

(4) Data on Educational NormsThese refer to the various norms or standards

used in school. They include information on classsize, the physical specifications of a. school, staff-student contact hours, the ratio of senior-to-juniorstaff, etc. They are needed in the estimation ofrecurrent and capital :Amts.

4

(5) Socio-economic DataThese include data on national output, cost of

living price indices, and public expenditures. Theyare used in constructing educational indicators andeducational indices.

Data for quantities, prices, and norms are to beprovided in sufficient details to match those of thecost data.

There is obviously a cost for managing a properdatabase of educational costs; but this cost is likely tobe more than compensated by the gains of betterinformed decisions. The need to strengthen theinformational basis of cost analysis is both obviousand urgent.

IL

IIMEMINNEMINVORMININCost Analysis for Educational Policymaking:A Review of Cost Studies In Education inDeveloping Countries

Section I:Introduction

Given the major challenge of improving educationunder tight budgetary constraints, educational poli-cymakers in developing countries today are con-cerned with issues on educational ccsts. Using aneconomic framework, this paper reviews the issuesand synthesizes the findings in a diverse literature oncosts of education in developing countries. Four keyissues on educational costs are considered: (1) Whatare the costs of education? (2) What are the majordeterminants of educational costs? (3) In what wayscan cost analysis improve policymaking in educa-tion? And (4) What are the informational needs forcost analysis in education? The paper concludes thatwhile cost analysis can contribute significantly toinformed decisions on education, greater efforts mustbe undertaken to strengthen the informational basisof cost analysis and to incorporate cost analysis ineducationa.1 policymaking,

Cost Studies in Education inDeveloping Countries

.1 major effort to promote the application of costanalysis in educational planning in developing coun-tries began in the late 1960s. The effort was organ-ized by UNESCO, with the participation of nineteencountries, twelve of which were Third World coun-tries. It included the preparation of monographs onthe methodology of cost analysis in educational plan-ning (Hallak, 1969; Vaizey and Chesswas, 1967;Woodhall, 1967) and the implementation of a large-scale research project consisting of twenty-seven casestudies (UNE`',CO, 1972). Cost analysis was appliedto education tc test the economic feasibility of educa-tion expansion plans, cost educational reforms andinnovations, and to guide efficient allocation of scarceresources to education. The focus was on formalschooling,

This effort occurred during a period of rapid in-creases in enrollment and public educational expen-diture, stimulated partly by the popular belief in theconsiderable economic value of education. The appli-cation of cost analysis to educational planning re-flected the acceptance by educational policymakersth, t educational spending was an investment activ-ity .menable to economic calculus.

In the 1970s, these applications of cost analysiscontinued to receive attention in developing coun-tries (McMeekin, 1975; Zymelman, 1976b). How-ever, during this period, a major focus was placed onthe costing and cost-effectiveness evaluation of neweducational media (Jamison et al., 1978; UNESCO,1977; Eicher and °rivet, 1980). This was promptedby the realization of the high cost of a linear expan-sion (simply increasing the size) of the traditionaleducation system. The capability of television andradio to reach large audiences and remote geographi-cal regions, as well as the potential savings in the costof education per student, generated much enthusi-asm in new educational media. At the same time, theinterest in non-formal education for rural develop-ment also led to studies on the economics of non-formal education (Ahmed, 1975; Hunter, 1974).

Since the late 1970s and especially in the pest fewyears, stagnant economic growth and severe fiscalconstraints have shifted attention to the control and/or reduction of educational costs, as well as on em-ploying alternative mechanisms for financing educa-tion (Eicher, 1984; Schiefelbein, 1986; Wolff, 1985;Psacharopoulos, Tan, and Jimenez, 1986; WorldBank, 1985).

The literature on educational cost analysis in de-veloping countries is quite diverse. It includes stud-ies that vary in their scope of analysis, such as the

...stagnant economicgrowth and severe fis

cal constraints haveshifted attention to thecontrol and/or reduc-

tion of educationalcosts, employing alter-native mechanims forfinancing education...

modes ofeducational delivery (formal and non-formaleduntion), levels of schooling, types of education(public and private), geographical locations (urban

1

(

A sound cost arolysis may alsoreveal serious policy errors that haveto be dealt with

and rural), trends over time, as well as the scale ofeducational interventions (an educational project, aneducational innovation, and an educational plan).The diversity of the studies is further multiplied bythe different types of economic analysis within thesestudies, such as costing, economic-feasibility testing,cost reduction, cost-effectiveness comparison, cost-benefit comparison and others.

To make the review manageable, this paper haslimited itself to cost studies of formal education andhas concentrated on the findings for public educa-tion'; but it has placed no restriction on the type ofeconomic analysis to be considered. The sources forthe review include published and unpublished worksavailable in the public domain, as well as a limitednumber of governmental reports. These sources arealmost exclusively in English2. A substantial numberof these works were conducted with the sponsorshipof the World Bank or UNESCO.

An Economic Framework for AnalyzingEducational Costs and Efficiency

The conventional economic approach to the studyof education regards education as similar to economicproduction (Lau, 1979; Hanushek, 1979). In eco-nomic production, given production objectives, pricesand technology, inputs are transformed into desiredoutputs. The internal process that transforms inputsto outputs is represented by a production functionthat is a relationship indicating the maximumamount of outputs that can be produced for giveninputs. Let us consider the application of this frame-work to education.3

The objectives of education refer to the tasks to beaccomplished by education that are assigned by thestate. They may include general ones such as "pro-ducing good citizens" and learning cultural heri-tage" to specific ones such as computational andreading skills.

Outputs of education consist of educational effectssuch as cognitive and noncognitive skills that aretaught to students. Presumably, these added skillsare consistent with the objectives placed upon educa-tion. Besides these outputs, education may also gen-erate benefits such as higher productivity and earn-ings.

Inputs to education are the various ingredientsused in producing outputs, including students, teach-ers, instructional materials, equipment, physicalfacilities, and others. The resources devoted to these

inputs constitute the costs of education,The technology of education denotes the internal

process encompassing the curriculum, pedagogicalmethods, school organization, management, andmonitoring procedure. Alternative technologies ofeducation exist, such as the traditional school andout-of-school distance teaching. The relationshipbetween inputs and outputs is represented by an edu-cational production function (EPF).

Using this educational production framework, wecan distinguish several concepts of efficiency in edu-cation to which cost analysis can be applied. Theseconcepts include internal efficiency, external effi-ciency, technical efficiency, and economic efficiency.

The internal efficiency of education col Lkpares thecosts of education to the outputs or effects withineducation, such as the acquisition of cognitive andnoncognitive skills. Education production is said tobe more internally efficient when it can produce moredesired outputs given the same resources.

The external efficiency of education compares thecosts of education to the benefits of education that areexternal to educational production, such as higherproductivity and earnings in post-schooling work. Itprovides a measure of the profitability of investmentin education.

While external efficiency and internal efficiency'are defined with respect to the boundary of educa-tional production, technical efficiency and economicefficiency concern the very nature of educationalproduction. Consider a given amount of financialresources. This amount of financial resources can beused to nurchase a certain combination of inputs atprevailing prices. Given this combination of inputsand technology, educational production is technicallyefficient when the maximum amount of school out-come (outputs or benefits) is produced; that is, aschool is operating on the production-function"curve." Educational production can be technicallyinefficient when some of the given inputs are under-utilized. When a school is technically inefficient,school outcome can be raised without incurring addi-tional cost, just by utilizing existing inputs morefully.

Given prices of inputs, the same amount of finan-cial resources can be used to purchase different com-binations of inputs, for example, more or fewer teach-ers as opposed to textbooks or physical facilities.Educational production is economically efficientwhen, given prices, technology, and financial re-sources, the maximum amount of school outcome is

it pays to understand the behavioralcharacteristics of educational costs by con-structing educational cost Indicators and by

studying resource utilization In schools.

produced by selecting the right combination of in-puts. When a school is ecorn.. tically inefficient,school outcome can be raised without incurring addi-tional cost, just by altering the combination of inputs.

Thus the efficiency of education, internally or ex-ternally, can be promoted by "technical" and/or "eco-nomic" means.

The above discussion of the efficiency of educationhas been confined to a given technology of schooling.This, however, does not have to be the case. Givenavailable resources, it is possible to raise schooloutcome by developing and using alternative tech-nologies in terms of changes in organization, man-agement, curriculum, or pedagogy. The costs andoutcome of the alternative f' m of schooling andthose of the traditional school can be compared.

So far, we have considered the efficiency of educa-tion from the perspective of a production system,using an educational production function to relateochool outcome to inputs. An alternative but equiva-lent way to study efficiency (Shepard, 1970) is to usean educational cost function (ECF). An ECF is a re-lationship which, under the prevailing technologyand input prices, indicates the minimum cost neededto produce a given level of outcome.

It is not difficult to see that the several concepts ofefficiency mentioned above can be analyzed using anECF. External or internal efficiency is increasedwhenever cost can be reduced in achieving a givenlevel of external benefit or internal effect respec-tively. Educational production is technically efficientwhen, given input prices, input combination andtechnology, minimum cost is incurred to produce agiven level of school outcome; that is, when the schoolis operating on the ECF "curve." When school is nottechnically efficient, educational cost can be reducedby cutting excess inputs without affecting the level ofschool outcome. Finally, economic efficiency is at-tained when, at prevailing input prices and technol-ogy, ;he right mix of inputs is chosen to produce ar±: on level of school outcome so that the cost incurredis minimal. When school is not economically efficient,cost can be reduced by changing the mix of inputswithout affecting the level of school outcome.

Viewing education (or part of it) as a productionsystem, we can conveniently place educational coststudies into three categories; (1) educational costingand cost-feasibility studies, (2) studies analyzing thebehavioral characteristics of educational costs, and(3) cost-benefit and cost-effectiveness studies.

Studies in the first category are concerned with

inputs to education only. The major teaks are toidentify, classify, and measure the costs of variousinputs to education. These studies are conducted forpurposes such as costing and testing the economicfeasibility of an educational plan, measuring thestart-up costs and operating costs of a major educa-tional intervention, and estimating the short-termand long-term cost impacts of a project.

Studies in the second category are concerned withrelationships among inputs and how inputs are util-ized in the educational production process. Themajor tasks are to determine the distribution of costsamong educational levels and educational inputs,identify the various factors affecting total costs andunit costs, estimate the impacts on cost of the level ofutilization of educational inputs, and analyze therelationship between educational costs and the sizeof an educational establishment. These analyses pro-vide a diagnosis of the behavioral characteristics ofeducational costs. They assess the level of resourceutilization and thus the opportunities for improvingtechnical efficiency. They can also uncover problem-atic cost patterns to identify strategies for improvingeconomic efficiency.

And finally, studies in the third category relateinputs to educational outcome. By comparing boththe costs and benefits or effects of alternative educa-tional interventions, cost-benefit or cost-effective-ness studies can inform educational decision makersabout efficient allocation of educational resources.The studies reviewed here include those that con-sider improvement in economic efficiency throughthe use of different mixes of inputs under a giventechnology of educational production, for example,more textbooks or smaller c' ass size in the traditionalschool. They also include those considering efficiencyimprovement through i he use of alternative educa-tional technology, such as educational media.

In reviewing educational cost studies in each ofthese three categories, tho paper attempts to clarifythe issues involved, synthesize the findings, and;ndicate knowledge gaps for further research. Thepurposes of the review are twofold: to document howeducational cost analysis can contribute to improvedpolicymaking in education, and to identify the infor-mational needs for cost analysis.

The rest of the paper is divided into four sections.The following three sections review the three catego-ries of cost studies. The last section presents recom-mendations on cost studies in education.

Cost Analysis for Educational Policymaking:A Review of Cost Studies in Education inDeveloping Countries

Section 11:Educational Costing andEconomic-Feasibility Testing

Costing and testing the economic feasibility ofeducational interventions are common applicationsof cost analysis in educatiun. Analytically, they areconcerned with identifying the various inputs to edu-cational production and measuring the costs of theseinputs. In this section, we first discuss the applica-tions of educational costing and then consider thevarious conceptual and practical issues relevant tothe estimation of educational costs.

Costing and Economic-Feasibility Testingof Educational Interventions In DevelopingCountries

Before reviewing studies in the costing of educa-tional interventions, it is necessary to provide a briefdiscussion of the methodology of costing.

The cost of an intervention can be estimated usinga simple and logical approach called the ingredientsor resource approach (Levin, 1983). According to thisapproach, the ingredients used in the interventionare identified and costed. In cost analysis, the cost ofan ingredient is its opportunity cost, that is, the costincurred as a result of the ingredient being used in thegiven intervention and thus not available for use inalternative activities. It is measured as the worth ofthe ingredient in its best use. The sum of all the in-gredient costs is the total cost of the intervention.

In the ingredients approach, it is important todifferentiate between the total cost of an interventionand the costs incurred by those who pay for it. Eachingredient is paid for by someone (e.g. the centralgovernment, a local government, an individual, orforeign aid). The total cost of the intervention is thus

A frequent error in edu-cational costing is toestimate the total costof an intervention byconsidering costs in-curred by the govern-ment only.

8

often distributed among several sources of support.Information about the sources of support for an inter-vention is needed to assess the economic feasibility ofthe intervention. A frequent error in educationalcosting is to estimate the total cost of an interventionby considering costs incurred by the governmentonly.

Consider educational costing studies in develop-ing countries. An eF.rly example is provided by the1962 reform of primary education in Madagascar (TaNgoc et al., 1972). In trying to a thieve the goal ofuniversal primary education, the leaders of Mad-agascar realized that a strategy of linear expansion ofthe existing primary education system was economitally infeasible and would not meet the needs of thecountry, To expand educational opportunity for ruralchildren, a reform of primary education was intro-duced. It had four major features: (1) reorienting theprimary school curriculum to better match the needsof rural life in Madagascar, (2) reducing the cycle ofprimary school from six years to four years for ruralareas, (3) reducing teacher costs by creating a newcategory of teachers with lower qualifications, and (4)adopting a new pattern of financing which placedmore financial responsibilities on provincial govern-ments and rural communities.

A detailed costing of the reform found that the planplaced an excessive burden on provincial govern-ments and poor rural communities. It also showedthat the potential savings would be smaller andslower because of costs of training teachers for thenew curriculum and the fact that low-salaried teach-ers could not replace high-salaried teachers quickly.Thus, parts of the initial 1962 reform had to beredesigned. The study indicated that cost estimationwas considerably constrained by the lack of relevantinformation on the breakdown of school enrollmentsby grade levels and school cycles, on the variation ofteacher/pupil ratios by different types of school andgrade levels, and on the age and salary distribution ofthe teaching force. The ability of each participatingunit to finance the reform was also a major con sideration. Nevertheless, the example shows that a soundeconomic analysis can be useful in alerting decision

14%

makers to the potentially serious errors in educa-tional policies and in indicating the type of dataneeded for policy formulation.

Another early example is the costing of an educa-tional innovation in elementary schools in Barbadosin the 1960s (Durstine and Hudscn, 1972). In 1965,Barbados undertook a five-year experiment in team-teaching to promote student learning. The experi-ment was carried out in five schools, with a budget of$297,500. Each school had an average of about 650students, with five teams of four teachers each. TheBarbados teachers involved were trained to workeffectively in a team. iBefore the end of the experi-ment, a decision had to be made whether or not toextena team teaching to all elementary schools. Acost analysis was conducted to inform decision mak-ers about the cost implications of expansion. Since a"marginal" decision (whether or not to expand teamteaching beyond what already existed) was at issue,the "marginal cost" (additional costs) of the expan-sion was estimated. The calculation found that theadditional cost per student in the expansion of teamteaching would decline considerably to about one-quarter of tnat of the original experiment. This esti-mate, when combined with subsequent findings ofpedagogical evaluation, would constitute the infor-mation basis for decision making. It should be noted,however, that the estimate was an improper measureof the cost per student because was based on thebudget alone; resources from other sources were notconsidered. This is a frequent error in educationalcosting.

The costing of new educational media is anotherimportant application of cost analysis in education.Traditional primary schools are characterized by avery labor-intensive technology of educational pro-duction; in other words, schools rely heavily on teach-ers and other personnel to educate children. Thecapability of mass media (radio and television) toeducate a large number of students at presumablylow cost and to reach children in thinly populatedareas has made it an attractive innovation in devel-oping countries (for example, Nicaragua, El Salva-dor, Brazil, Mexico, Korea, Ivo," Coast, Senegal,Malawi, Kenya, and Thailand) A significant amountof cost-analysis work has been done on new educa-tional media projects (Jamison et al., 1978; Carnoyand Levin, 1975; UNESCO, 1977 & 1980; Eicher etal., .1.982; Perraton, 1982; Wagner, 1982). The find-ings indicate that the unit cost (cost per br)adcasthour) varies significantly among countries (Lichei et

al., 1982: 56) and that the cost per student usuallydecreases with studciit enrollment (Eicher andOrivel, 1980). Also, these projects usually involvehigh start-up costs. A positive result of the costingeffort is the standardization of the measurement andclassification of the costs of new educational media.But quite a few of these projects were implementedwithout a urior evaluation of the cost-effectiveness ofnew educational media. The cost-effectiveness ofnew educational media will be considered in the sub-section entitled "Cost-Effectiveness Studies in Edu-cation in Developing Countries."

As a last illustration, let us consider the costingand feasibility testing of the education plan of Thai-land in the 1960s (Reiff, 1972a & 1972b). Thailand'sNational Economic and Social Development Plan for1967 - 1971 called for a set of development objectivesand quantitative targets for the Thai education sys-tem (1972b: 268-272). However these objectives andtargets were developed without examining their costimplications. A cost analysis was conducted onlyafter the targets were set. Crude cost data for theanalysis came from budgetary data of the centralgovernment as well as a sample survey of over ahundred schools.

Some of the findings of the study were supportiveof the educational plan. For example, the plannedcai ital expenditure for new educational facilities andthe planned public funds for the recurrent costs ofpublic schools at the pre-tertiary levels were ade-quate for meeting the financial requirements. Butfindings also emerged that questioned the feasibilityand desirability of some of the targets of the plan. Forexample, the per student recurrent cost of highereducation was ten times that of secondary education.Since the plan called for a 30% increase in universityenrollment over the five-year period, a significantportion of the educational fund would have to bedevoted to higher education. One might thus ques-tion whether the benefits of higher education inThailand were large enough to justify the proposedallocation between the education levels. Also, be-cause they ignored repetitions and dropouts inschools, the enrollment targets would fail to meet thetargets in graduates. Revisions of the plan had to bemade subsequently. The process of decision makingcould have been improved by incorporating costanalysis in the planning stage.

We can go on and on.' But the above examplesillustrate the broad scope of s.pplications of costanalysis in education, from an intervention in the

classroom and an innovation in education, to reformof primary education and the planning of an entireeducation system. In looking back over the previousexperiences in developing countries, we can quicklypoint out the importance of cost analysis in informingdecision makers about the cost implications of aneducational intervention and whether or not theintervention is financially feasible. Incorporatingcost analysis into the educational planningprocess ishighly desiralae.

However, previous experiences have also demon-strated the existence of a number of barriers thatcould limit the usefulness of cost analysis for in-formed decision making in education. These barriersare socio-political in nature. For example, many edu-cational plans were drawn up by decision makers forsymbolic purposes, to legitimize Vie actions andpower of a political regime, and to comply with therequirements for external financial assistance. Theeconomic feasibility anri implementation issues wereof secondary importance (Weiler, 1978). A well-executed cost analysis might present findings incon-sistent with the hidden intentions of decision mak-ers. Besides, educational production also takes placewithin a social and political context. A detailed costanalysis of education might also reveal significantinequit'es in the distribution of educational re-sources by social class, gender, ethnicity, and region(Tilak, 1985). This could potentially lead to socialtensions that a regime would like to avoid. The use ofthe findings of a cost analysis might thus be re-stricted.

But we could also easily point nut how previousevaluations of genuine educational interventionswere undermined by a failure to consider educationalcosts. This failure often resulted from a lack ofawareness of the importance of cost analysis on thepart of decision makers, a shortage of competent costanalysts, or from the lack of good data for cost analy-sis (Eicher, 1984: Part I; McMeekin, 197C: Chapter3). These barriers are "technical" in nature. They aremore likely to be overcome than the political ones.

Cost Estimation In EducationEven though there is no dearth of studies on

educational costing, determining the costs of educa-tion is not a simple matter. Previous studies havemade it clear that there is no single response to thequestion, What is the cost of education? (Psacharo-poulos and Woodhall, 1985: Chapter 7). Althoughconsiderable progress has been made in conceptual

10

understanding of the costs of education, significantpractical and theoretical issues remain that makeeducational costing less than a hard science. Besidesbeing familiar with basic cost concepts and analyticalskills, a competent cost analyst needs to be ingeniousabout using existing cost data (which are often crudefor developing countries), be able to make simplifyingbut not off-the-mark assumptions under conditions ofincomplete information, and be sensitive to the relia-bility of available data (Coombs and Hallak,1987). Inthe following pages, the conceptual and practicalissues in cost estimation in education are reviewedunder these headings: concepts of cost, classificationof costs, data collection and measurement, and waysof expressing costs.

Concepts of CostFrom the viewpoint of economic analysis, the

proper definition of cost (real cost or economic cost) ofan input to education is its opportunity cost, which ismeasured by the value of the input in its best alterna-tive use. Applying this concept to the cost of aneducation system, the real cost of education includesnot only public expenditure on education, but alsoprivate costs (Bowman, 1966). Privatecosts of educa-tion include both direct monetary expenses for tui-tion, textbooks, ana other maintenance items, andthe indirect cost of students' time measured by theforegone earnings in employment. Public educa-tional expenditure can significantly underestimatethe real cost of education. In a recent study of India,Tilak (1985: 22) estimated that the indirect privatecost in terms of foregone earnings accounted forabout 40% of the real cost to education, based on1977-78 data. For nine eastern African countries, thetotal direct private cost of secondary education perstudent averaged 80% of government educational ex-penditure per student (Wolff, 1985: 51-55).

Private costs of education are important to con-sider not only because they constitute a significantpart of the real cost of education, but also becausethey can affect the demand for schooling. For chil-dren in rural communities in developing countries,going to school means not helping parents in farmwork and other household chores, with a correspond-ing reduction in family welfare. Thisprivate sacrificeexplains the high dropout rates in rural primaryschools, frustrating the government's attempt inproviding universal primary education (Haddad,979). Also, under the current tight budgetary con-

straints, governments may have to rely on additional

private contributions to education. Information onthe existing burden on parents is necessary in designin; alternative finance strategies (Tan, 1985b). Un-fortunately, data on private costs are lacking in mostdeveloping countries.

What costs to measure depends heavily on thedecision context in which cost analysis is conducted.A key issue to consider is the cost to whom. If one isconcerned with the social efficiency of educational in-vestment, then both public and private costs have tobe included. If, however, one is concerned with thefiscal implications of an educational intervention,then it is the costs to the government that have to beestimated, even though fiscal costs and real costs maydiffer considerably. Another key issue to consider iswhich cost measure is appropriate, average cost ormarginal cost. If the decision is to determine whetheror not to expand an existing program (as in theBarbados example), the amoi mt of the additional costis what is at issue; marginal. cost analysis is appropri-ate. Average cost analysis ;s relevant if the decisioninvolves a choice between different programs, for ex-ample, for providing in-service training for unquali-fied teachers (Taylor, 1983).

Classification of Educational CostsThe classification of educational costs remains an

unsettled area of cost analysis, Except for the costs ofnew educational media, there exists no internation-ally standardized classification of educational costs.Presumably, educational costs can be classified ac-cording to criteria that are economic (the real re-sources used), institutional (the sources of support),financial (the timing of expenditure), and technical(the function of inputs) (Psacharopoulos andWoodhall, 1985:169-170). But the complexity ofthese issues and the variations among countriesmake a standardized classification difficult. Eachdeveloping country tends to have its own system ofcost accounting. This creates problems for a cross-national compariPon of educational costs.

Nevertheless, these are some generally accepteddistinctions among educational costs (Vaizey andChesswas, 1967; Coombs and Hallak, 1972; Eicher,et al., 1982: Chapter 3). These include, for example,the distinctions between economic cost and expendi-ture, direct costs and indirect costs, recurrent expen-ditures and capital expenditures, variable costs andfixed costs, unit costs and marginal costs, public costsand private costs, personnel costs and nonpersonnelcosts, as well as instructional costs and nonin-

,c

struction al costs.Tilak (1985) suggests a taxonomy for organizing

different kinds of educational costs. The taxonomybegins with an institutional distinction (costs bysources) between public costs (referred to as institu-tional costs by Tilak) and private costs. Institutionalcosts consist of direct (visible) institutional costs andindirect (invisible) institutional costs (referred to asopportunity costs by Tilak). Visible institutionalcosts are divided into two categories: recurring costsand nonrecurring costs. Recurring costs consist ofteachers' salaries, salaries of other staff, scholar-ships and stipends, depreciation, and other expendi-tures. Nonrecurring costs include costs for buildings,furniture, equipment, etc. The invisible institutionalcosts are not specified.

On the other hand, private cost are costs borne byindividuals. They consist of direct (visible) privatecosts and indirect (invisible) private cost. Visibleprivate costs consist of tuition cost (tuition fees plusother fees) and nontuition cost (maintenance costrelated to individual spending on books and station-ery, hostel, transport, unifor-as, etc.). Invisible pri-vate cost is the earnings foregone by individuals.

There are two distinguishing features of this tax-onomy. First, it is suitbble for calculating the real orsocial costs of education. Second, it focuses on )oththe sources of educational costs and dn. cost ofvarious input items to education. But there ar e alsosoma deficiencies. It ignores other sources of supportfor education, such as contributiorm from private or-ganizations and industry, as well as lxternal aid. Forsome developing countries, external lid may accountfor a significant portion of public expenditure oneducation (World Bank, 1980: Chapter 8; Coombsand Hallak, 1972: 106-107). Also, there is no consid-eration of how institutional costs are utilized techni-cally, and no information about the functions of theinputs in educational production. A similar, but moredetailed classification of educational costs whichaddresses these deficiencies is proposed by Tsang(1987a) and is presented in Figure 1 (see page 12).Figure 1 highlights the financial sources of educa-tional costs, since information on financial sources iscrucial in assessing economic feasibility of educa-tional interventions. The figure indicates that recur-rent costs are to be broken down into a matrix of inputitems and their functional uses.

New educational media are an area where thereis now a generally accepted classification of costs.Beginning in 1977, UNESCO undertook a compre-

Figure 1: A Classification ofCosts of Education

[ Costs of Education

J

Sources: Public, Private, Foreign

Institutional Costs

_J--Household

CostsRecurrent Costs Capital Costs

Buildings & Funiii,ureEquipmentLandOthers

Direct Costs Indirect CostsTuition OpportunityOther School Fees Costs (farm work/Uniforms foregone earnings)Transportation

Function (or Programs)Books & othersupplies Factors Instruction Admin. Food/Dorms Health Care Ge '1 Maintenance

Personnel:OthersTeachersA iministrative

staffOther staff

Non-Personnel:TextbooksOther teaching

aidsSuppliesUtilitiesScholarshipsStudent WelfareMaintenance and

repairs

hensive re Tiew and synthesis of the works done onnew educational media (UNESCO, 1977, 1980;Eicher et al., 1982). One of the outcomes of the effortis a standard procedure for costing new educationalmedia. This procedure reflects a technical approachto cost accounting. Costs are divided into four catego-ries: costs of general administration, costs of produc-tion, costs of distribution or transmission, and costsfor reception (Eicher et al., 1982: 41-64).

Data Collection and MeasurementTo date, the information basis for educational cost-

ing remains in a primitive state in developing coun-tries and wide gaps exist between what data are

12

needed and what data are available (Eicher, 1984:Part I).

In many developing countries, budgetary datafrom the central government are usually all that isavailable for educational costing. The shortcomingsof ;,here data for educational costing are by now wellknown. First, the data are often given for plannededucational expenditures, not actual educationalexpenditures. Experience has shown that there couldbe significant differences between them. And whenactual expenditures are available, they are moredifficult to obtain and may come too late to inform adecision at hand (Eicher, 1984: 7).

Second, expenditures from other levels of govern-

...tho information basis for educational cost'ngremains in 6 primitive state in developing countries

and wide gaps exist between what data areneeded and what data are available.

ment are often not available. The underestimation ofpublic expenditure on education can be significant ina country like India with a decentralized educationalfinance system (Tilak, 1985).

Third, attentiveness must be exercised in identify-ing public costs of education from the central govern-ment. Expenditures on education are made not onlyby the department or ministry of education, but alsoby other departments or ministries. For example,building costs may come from the public works de-partment, awl teacher pensions may be paid by thefinance department. On the wither hand, tome of theexpenditures by the department of education are notrelated to education, for example, allocations formuseums and the arts (Vaizey and Chesswas, 1967:14-16). Proper .Ausion of expenditure by noneduca-tion departments and exclusion of expenditure by thedepartment of education are warranted.

Fourth, central government data can be too aggre-gated to be useful. For example, governments usu-ally provide information on teacher salaries, butother inputs are lumped into one broad category"other current expenditures," thus not showing sepa-rate expenditures on the other inputs. Also thecentral level data for the expenditures on , -18inputs are not relevant for determining the costs ofindividual items such as textbooks or equipment atthe school level. "Micro" data on costs are needed toconduct a cost-effectiveness comparison of alterna-tive educational interventions at the school level(Eicher, 1984: 9). Costs can vary by schools in differ-ent regions.

Fifth, central government budgetary data are notusually classified into categories that are relevant forsome applications, for example, cost analysis of edu-cational programs or subjects. Cost data are avail-able for some input items but they do not show howthe costs are spent for different programs or subjectareas.

Sixth, the data are expenditures, not economiccosts. They do not include the costs associated withthe direct and indirect private costs of schooling.Such data are thus insufficient for estimating thesocial costs of education.

And finally, central government budgetary datacan be unreliable. There are often inconsistencies inhow cost categories are defined over time, and differ-ent governmental sources may yield contrad; dory re-sults (Eicher, 1984: Chapter 2).

Several guidelines for collecting cost data .'eresuggested by McMeekin (1975: Chapter 3). First,data should be collected and processed on a timely

basis and be easily accessible. Second, it is importantto have reliable data. Since there are incentives forrespondents to misrepresent information, such asinflating attendance rates and deflating dropoutrates, data must be carefully checked against pos-sible biases in reporting.

Third, while the most accurate information shouldbe gathered, data collections should also be guided bya cost-effectiveness mentality. One should weigh theadditional gain in the precision of a data item againstthe additional resources devoted to its collection. Andfourth, data collection should be made economical, ifpossible. For example, surveys can replace census-type data collection on schools and households todetermine school costs and private costs respectively.

Finally, previous experience has shown the exist-ence of a number of problematic areas in the meas-urement of educational cost. An important one is themeasurement of capital costs. In many educationalprojects, capital costs are not properly estimated(Psacharopoulos and Woodhall, 1985: 178-184). Oneproblem is the failure to treat capital investment asa flow by annualizing it (using a social discount rateand making an assumption on the expected life ofservice). Another common problem is the underesti-mation of capital costs by ignoring future expendi-tures that are required as a result of the capitalinvestment (e.g. maintenance, repair, and staff). Theeconomic evaluations of new educational media inthe early 1970s often underestimated capital costsand thus led to misleading findings about the attrac-tiveness of new educational media (Carnoy andLevin,1975). Another common error in cost measure-ment to be avoided is double counting, usually involv-ing transfer payments. A transfer payment is a shiftof purchasing power from one agent (an individual,an institution, or a unit of government) to anotheragent, without generating additional income or na-tional product. Double counting occurs when theexpenditures from a higher level of government anda lower level of government are both counted withoutsubtracting the transfer payment between the twolevels of government. It can also occur in the transferpayment between a university and a student, that is,student scholarships, fellowships, and other subsi-dies (Coombs and Hallak, 1987: 50-51).

Ways of Expressing Educational CostsThe cost of an educational innovation or plan is

often expressed in terms of its Total cost to indicatethe total value of real resources devoted to ic. But inmany situations, unit costs are more meaningful,

especially for diagnostic, comparative, and evalu-ative purposes. Here we have to define what aneducation unit is.

From the economic framework of educational pro-duction, a unit for educational cost refers to a unit ofoutput of educational production. Yet education hasmultiple outputs. These outputs have been measuredin terms of student achievement, number of gradu-ates, student literacy, numeracy, and others. Giventhe diffuse nature of outputs of education, cost ana-lysts have often looked at inputs as units of analysis,for example, students enrolled and number of teach-ers. Previous works have identified a number of unitcosts of education (Hallak, 1969:41-44; Coombs andHallak, 1972: Chapters 7 & 8; Jamison, 1977: Sec-tion III; Tilak, 1985:6-8): (1) cost per pupil enrolled,(2) cost per pupil actually attending school, (3) costper graduate, (4) cost by level of education attained,(5) cost per pupil of the relevant age-group popula-tion, (6) cost of education per capita, (7) cost per class,(8) cost per hour, (9) average recurrent cost perteacher, and (10) capital cost per place.

In the unit costs listed above, cost per pupil enrolled is sometimes known as the "normal" cost ofeducation. It is the most commonly used -snit cost inplanning education at all levels. But since there canbe a large gap between reported enrollment andactual attendance in lower levels of schools in devel-oping countries, actual attendance of pupils is alsoconsidered. Cost per pupil attending school indicatesthe unit cost associated with attendance. Cost pergraduate is referred to as the "effective" cost of edu-cation. The gap between cost per pupil enrolled andcost per graduate is a measure of the extent ofrepetitions in school. The ratio between them issometimes taken as an indicator of the internalefficiency of education (Psacharopoulos andWoodhall, 1985; Chapter 8). Cost per graduate isparticularly relevant for manpower-planning pur-poses as it relates to school completers. While costper graduate measures the unit cost at the end of aneducation cycle, cost per education level measuresthe unit cost of a given level of education. The use ofcost per education level reflects the view that a pupilwho quits school at a certain level of schooling hasnevertheless acquired some useful skills. Cost perpupil of the relevant age group and cost of educationper capita are measures of the coverage of an educa-tion system or a subsystem. Sometimes cost analystsfind it car ienient to use cost per class in cost projec-tions. But its usage should be discouraged whenthere is significant variation in class size (in different

14

regions, by education levels, and across countries).Cost per hour is often used to express the cost perbroadcast hour in new educational media projects; itis seldom used elsewhere. Average recurrentcost perteacher is also used in costing because of the labor-in-tensive nature of educational production and thedominance of teacher salaries in public educationalexpenditure. Finally, capital cost per place is used inprojecting the costs of construction and equipmentper pupil place provided.

Although each unit-cost measure has its own ap-plication, and no one measure is suitable for allpurposes of costing, cost per graduate and cost perpupil enrolled should probably be considered in mosteducational costing work. From the viewpoint of edu-cational production, cost per graduate is the appro-priate unit cost of education. Although its computa-tion is sometimes hampered by the lack of informa-tion about the number of graduates (by level ofeducation, type of program, and type of school (pub-lic vs. private schools, rural vs. urban schools)), itsusage can be promoted by a better effort in data col-lection. Despite the fact that it is an input measure,cost per pupil enrolled is of great practical value tcpolicymakers in that it estimates the cost of puttinistudents into the classroom. Dataon student enroll-ments are usually available. In the rest of the paper,unit cost refers to cost per pupil enrolled, unlessstated otherwise.

Another major issue in expressing education costsis the distinction between cost in current dollars andcost in constant dollars. Cost in constant dollarsdiffers from cost in current dollars in that the formerreflects the real value ofresources and is obtained byadjusting the latter with price indices for changes inthe price level.

In many developing countries, high rates of infla-tion often nullify a large portion of the considerableincrease in public expenditure on education in cur-rent dollars. In India, for example, educational ex-penditure increased by about 30 times in currentdollars between 1950 and 1975, but it increased byonly 7.5 times in constant dollars for the same period(Tilak and Jandhyala, 1983:5). 7thiopia, the 380%increase in current expenditure _etween 1970 and1982 was reduced to a 46% increase in constant ex-penditure. And for Mexico, the increase in expendi-ture for the period 19'10 -82 was 3,866% in currentdollars and only 941% in constant dollars (UNESCOdata). :'here figures demonstrate the importance ofestimating educational costs in constant dollars inorder to assess the amount of real resources to educa-

tion. Teachers are often victimized by inflation astheir salary increases lag behind the rates of infla-tion.

The need to adjust costs for price changes withprice indices was recognized early (Vaizey, 1958). Indeveloped countries, significant advances have beenmade in the theoretical development of price indice3and applications to education (Wasserman, 1963;Chambers, 1979). Suggestions for constructing priceindices for developing countries have often beenmade (Hallak, 1969: 52-53; Johnstone, 1981 :104-

106; Tile ;, 1985:14), but educational price indicesare rare in these countries.

Conceptually, educational price indices are basedon the actual mix of physical and human resourcesconsumed in educational production and their pricelevels. The mix and prices of these resources arelikely to be different from other buyers or the typicalconsumer. In developing countries, price adjustmentfor educational cots is usually done by using eitherthe cost-of-living price index or the wholesale priceindex. However, this practice can lead to misleadingresults because these two indices do not necessarilyreflect changes in price levels in education. The prob-lem is often compounded by basing these indices onurban prices; rural prices and urban prices can differsignificantly.

For educational decision makers, ed .nationalprice indices are important not only for estimatingthe real resources to education, but also for under-standing the patterns of the real value of variouseducational inputs, especially teacher costs, to in-form the wage bargaining process. Despite the use-fulness of price indices, they have not been con-structed in developing countries.

In reviewing the cost studies on educational cost-ing in developing countries in the past two decades, I

have mixed feelings about the progress made in costestimation.

We may take some comfort in the fact that signifi-cant advances have been made in the conceptualunderstanding of educational costs. These includethe explication of an economic concept of cost, anunderstanding of the impact of the decision contexton the costs to be estimated, the progress in theclassification of education al costs in general and costsof new educational 'media in particular, an apprecia-tion of the relative advantages of alternative ways forexpressing educational costs, and clarifications in theprocedure for measuring costs, especially capitalcosts. Further work is warranted in the developmentof cost classification and price indices applicable todeveloping countries.

But we definitely cannot take comfort in the primi-tive state of the information basis for cost analysis inmany developing countries today. One cannot avoidbeing struck by the tremendous gap between whatdata are required and what data are available for costanalysis. In the face of severe fiscal constraints andthe call for efficient utilization of scarce educationalresources, the need to strengthen the informationalbasis of cost analysis for educational decision makingis both self-evident and urgent. A cost-of-educationdatabase is not adequate if it does not contain reli-able information on the relevant categories of publiccosts (broken down by input items and functions) andprivaie costs (both direct and indirect costs) for vari-ous levels of education, over time, and at variousadministrative levels (school, local, regional, andcentral levels of government). There is obviously acost tbr developing or improving such a database; butthe expense of uninformed educational decisionmaking is considerably higher, in the long run.

InewmwmaiwCost Analysis for Educational Policymaking:A Review of Cost Studies In Education inDeveloping Countries

Section 111:Behavioral Characteristicsof Educational Costs

This section reviews cost studies in education indeveloping countries for their analysis of the behav-ioral characteristics of educational costs. Althougheducational costs in developing countries can varysignificantly among levels, programs, and regions,they also exhibit common patterns or "behavior"because schools in these countries are operated insimilar ways. From the economic framework of edu-cational production, these studies are concerned withthe relationships among inputs and the utilization ofinputs in the production process. They cover a widerange of analyses, from "macro" investigations ofnational education systems to "micro" investigationsof individual schools.

...cross-nationalcomparison is useful forIndicating the relativemagnitude of educa-tional costs and for un-covering the patternsand determinants ofthese costs.

The review is divided into two parts. The first partconsiders studies on the patterns and determinantsof educational expenditures and unit costs of educa-tion. This part pertains mainly to the application ofcost, analysis for diagnostic purposes and for improv-ing economic efficiency. The second part considersstudies on the analysis of resource utilization inschools. It is concerned primarily with the technicalefficiency of schools.

Educational Expenditures and Unit Costs:Patterns and Determinants

Comparative studies based on national time-se-ries data represent a major approach to the analysisof educational expenditures and unit costs (Gandhi,1971; Zymelman, 1976 & 1982; Eicher and Orivel,1980b; Eicher, 1984; Wolff, 1985; Schiefelbein, 1986;

16

Mingat and Psacharopoulos, 1985). They emphasizeunderstanding the trend in educational expenditure,uncovering the patterns in unit costs, and identifyingthe factors that determine educational expendituresand unit costs.

Eicher (1984: Chapters 2 & 3) points wit a num-ber of methodological issues to be considered inmaking a cross-national comparison of educationalcosts. The first issue has to do with specifying themeasures of educational expenditures and unit coststo be used in the comparison. Total public educa-tional expenditure is not a meaningful variable forunderstanding the trend in cost because changes inpublic educational expenditure depend on the initialpercentage of government budget devoted to educa-tion and on changes in the government budget overtime. What is needed is some indicator of the "effort"devoted to education under given resources. Twomeasures are commonly employed in the literature:public educational expenditure as a percentage ofnational income (national-effort indicator) and pub-lic educational expenditure as a percentaf e of totalpublic expenditure (fiscal-effort indicatoz). Our dis-cussion in the subsection entitled "Cost Estimation inEducation" has also identified a number of measuresfor unit cost. The most, common measure used inthese studies is the cost per pupil enrolled.

The second issue concerns data sources. Most ofthe analysis was based on the time-series data col-lected by UNESCO and the World Bank, AlthoughUNESCO and the World Bank have made significantcontributions to the standardization of the procedurefor data collection, caution must be exercised in usingthese time-series data for international comparisonbecause of problems associated with the data col-lected: (1) inconsistencies in a time series can occurdue to changes in definitions and/or administrativeorganization for data collection; (2) UNESCO andWorld Bank data are sometimes not comparablebecause of some differences in their data collectionprocedure; (3) for a given country, contradictions incosts exist as a result of using different data sources(provisional account vo. final account) or differencesin the choice of a base year for enrollments (calendaryear vs. school year); and (4) variation exists in the

quality or reliability of the data collected.The third issue relates to comparison of findings

for different country groups. Countries are oftengrouped together and the group averages compared.Experience has shown that the values of group aver-ages are sensitive to the ways countries are grouped(by region, GNP per capita, number of countries in agroup, etc.), the choice of the base year for compari-scn, and especially the choice of the averagingmethod (arithmetic and geometric averages;weighted and unweighted averages). Different waysof grouping and averaging often lead to very differentfindings and different policy recommendations. Forexample, by using arithmetic averages weighted byGNP in each country, a UNESCO study found thatthe ratio of educational expenditure to GI P has in-creased from 3.3% in 1970 to 4.0% in 1979 in devel-oping countries (UNESCO, 1982). Zymelman (1982)found the ratio to fall from 3.29% in the early 1970sto 3.16% in the late 1970s in developing countries; hiscomputation was based on geometric averages. Simi-larly, a World Bank study found the unit cost ofhigher education in China in 1979 ($880 in 1975prices) to be high relative to two comparable countrygroups ($534 and $675 respectively in 1975 prices)(World Bank, 1983). But if a different averaging pro-cedure were used, China's unit cost would be rela-tively low.

Finally, it should be pointed out that educationalexpenditures are not the same as educational costs;they do not include the resources contributed byprivate sources to education.

These issues suggest care in interpreting cross-nationl findings on educational costs. Nevertheless,cross-national comparison is still useful for indicat-ing the relative magnitude of educational costs andfor uncovering the patterns and determinants ofthese costs.

Patterns and Determinants of EducationalExpenditures

Most cross-national studies have focused on pub-lic, not private educational expenditures, because ofa lack of data for the latter. They are almost exclu-sively based on the data collected by UNESCO andthe World Bank, and contain analyses of both devel-oped and developing countries. A common approachto cross-national comparison is based on countrygroupings by geographical regions. Prominent ex-amples include Zymelman's study of 68 developingcountries and 13 developed countries for 1970-72

211

(Zymelman, 1976a) and the late 1970s (Zymelman,1982), Eicher and Orivel's study (1980) of about 140developed and developing countries for the period of1960-1976, and compilations by UNESCO (1982) forthe period of 1970-78, and the World Bank (1980) forselected years and intervals between 1960 and 1974.While detailed results on the patterns of the expendi-ture indicators for individual countries and countrygroups can be found in these works, the major find-ings regarding the pattern of public educational ex-penditure over time, the distribution of public educa-tional expenditure, and private educational expendi-tures are briefly summarized here. Consider first thepattern of public educational expenditure over time.

Three significant findings emerge from Eicher'sreview of the time pattern of public educationalexpenditure (1984: 45-47).

First, despite differences between countries andregions, the overall effort in favor of education in thepast two decades has been nothing less than phe-nomenal. Total public educational expenditure in-creased by 250% in real terms. Public educationalexpenditure as a ratio of national income increasedby 75% in developing countries and by about 50% indeveloped countries. Public educational expenditureas a ratio of total public expenditure increased bymore than 30% in both developing and developedcountries.

Second, there are some discernible trends in thenational-effort and fiscal-effort indicators. Considerfirst the national-effort indicator. Between 1960 and1970, the average indicator (arithmetic average)went up from 2.9% to 4.2%. In both developed anddeveloping countries, public educational expenditureincreased faster than GNP. The upward trend con-tinued in th 1970s. Between 1970 and 1974, the rateof increase of the indicator slowed down and theindicator reached an average of 4.3% in 1976. But inthe 1970s, especially after 1974, the patterns be-tween countries began to differ. The indicators fordeveloped countries and Asian countries started tostabilize, but the average indicator for sub-SaharanAfrica continued its upward movement while theaverage indicator for Latin American countries wasmoving downward.

Similar developments have happened to fiscal ef-fort. Between 1960 and 1974, the average fiscal-effort indicator increased from 11% to 15%, the pat-tern being similar for both developed and developingcountries. But in the 1970s, almost every developedcountry had begun to reduce their level of fiscal effort

...the need for cost-reducing measures and,more generally, for policies turned toward cost-effectiveness Is everywhere present and is be-coming urgent in many countries."

for education. In contrast, developing countriescontinued the upward trend until the mid-1970s, andstarted to stabilize their fiscal effort afterward.

Third, the effort in favor of education is now lev-eling off. Most countries have stabilized their na-tional and fiscal efforts. Only the sub-Saharan Afri-can region maintains its upward trend. But the slow-down in educational effort is widespread.

Two sets of factors appear to be relevant in ex-plaining the level of total public educational expendi-ture and its changes over time (Coombs and Hallak.1972). The first set relates to the supply of educationand the amount of funds available to the governmentfor financing education. The determinants of educa-tional supply and budgets are quite complex. Theyinclude factors such as the rate of inflation, the rangeof competing demands of other public services, therate of growth of the national economy, the diversityof the tax bases for education and the sensitivity ofeducational revenue to changes in national income,the impact of foreign trade, and the availability offoreign aid to education. The second set of factorsrelates to the demand for education. They includethe growth in population and its age distribution, theimportance of education for social mobility and na-tional development, the choice and development oftechnology for economic production, the level of cov-erage of education to diverse seg-ments of the population, the directand indirect impacts of nationaland international developmentpolicies, and others. These two setsof factors are "external" to educa-tion in that they lie outside thelocus of control of school adminis-trators and policymakers.

Eicher and Orivel (1980b), andZymelman (1982) have providedstatistical analyses of the determi-nants of the effort for education.Using data for 122 countries for1960-76 and three models, Eicherand Orivel found that the most im-portant explanatory variable forthe nationil-effort indicator wasthe enrollment r. tio in the 6-11 agegroup; and the rate of change in theenrollment ratio was the most im-portant explanatory variable forthe increase in effort between 1960and 1976. The variable GNP per

capita was found to be less important and its coeffi-cient was small. Finally, "culture" appeared to be asignificant factor; Francophone Africa exhibitedhigher effort than Anglophone Africa (1980b: 42-55).

Zymelman found that educational expendituredid not depend on the level of economic development(measured by GNP per capita), but was influenced byunit cost as a ratio of GNP per capita, and theenrollment ratio. The relative importance of thesetwo factors varied by countries and by regions. Hisstudy was bared on data for 81 countries in sevenregions for the period of the 1970s. From the abovestudies, Eicher suggested that the slowdown in totalpublic educational expenditure in developing coun-tries was due to a combination of three factors: slowerrates of economic growth, a relative decrease in thedemand for education, and a change in governmentattitude toward education (1984: 55). Having consid-ered the current fiscal constraints and demographictrends, he concluded that "the need for cost-reducingmeasures and, more generally, for policies turnedtoward cost-effectiveness is everywhere present andis becoming urgent in many countries" (Eicher, 1984:59).

The distribution of public educational expendi-ture is also of interest to decision makers. Table 1presents the distribution of public educational expen-

Table 1: Public Educational Expenditure as a Percentage of GNPby level of education in the early 1970s and the late 1970sRegion Total Primary Level Secondary Level Higher Level

of Countries) Aa Bb A B A B A B

Eastern (14) 3.76 4.16 2.04 2.23 1.13 1.25 .59 .67Africa

Western (12) 3.93 3.61 1.64 1.75 1.20 1.19 .49 .66Africa

East Asia & 3.42 3.45 1.87 1.57 .96 1.06 .59 ,81Pacific (7)

South Asia 2.12 1.60 .72 .87 .89 .39 .51 .34(6)EMENA (12) 4.09 3.40 1.71 1.58 1.54 1.18 .84 .64

Latin (17) 3.02 2.78 1.63 1.39 .84 .80 .55 .59America

"Averaged(68)4.18 4.93 1.68 2.18 1.79 1.80 .71 .95Region"

Notest e: A.early 19701b: 11.1ate 19701

EMENA Includes 12 countries around the Mediterranean region including Northern Africa and the MiddleEast

d: based on geometric average

18

Sourest Zymelezan (198249)

...public educational expenditure isdominated by personnel costs,

especially teacher salaries...

diture by levels of education for seven regions in theworld at two time intervals. For developing coun-tries, public expenditure on primary education ac-counts for half of total public educational expendi-ture, while secondary education claims about onethird and higher education absorbs about one fifth ofthe total. The distribution of public educationalexpenditure among levels of education is stable overthe two ..me intervals. In OECD countries, highereducation accounts for slightly less than one fifth ofthe total, while the remainder is split about equallybetween primary education end secondary educa-tion. Since enrollments differ considerably amonglevels of education, these distributional figures masksignificant disparities in unit costs for various levelsof education (to be discussed later). In developingcountries, public expenditure on higher educationhas increased at a faster rate than that for primaryeducation.

The distribution of public educational expendi-ture between personnel and nonpersonnel categoriesis also striking. In all educational systems, publiceducational expenditure is dominated by personnelcosts, especially teacher salaries, although the ratioof personnel costs to nonpersonnel costs decreaseswith the level of education. This obviously reflectsthe labor-intensive technology of educational proc ac-tion. For example, in Eastern Africa, around 1D80,teacher emoluments alone accounted for 37%, 64%,and 40%, respectively, of public expenditure on pri-mary education, secondary education, and highereducation (Wolff, 1985:45 & 59). In the 1970s, LatinAmerican countries spent an average of 88% and 85%of their funds on teacher salaries for primary educa-tion and secondary education, respectively (derivedfrom Schiefelbein, 1986: 9). The dominance ofteacher costs is also found in Asian countries (Tilak,1985; Bennett, 1975; Alles et al., 1972).

Aho, given the salary increases associated withautomatic promotion in the pay structure for teach-ers, there is a built-in tendency for teacher costs torise over time in these countries. Moreover, com-pared to developed countries, developing countriesspend a higher percentage of their budgets on teachercosts. During periods of tight budgets and pressuresfrom teachers for higher salaries, developing coun-tries have often shifted resources away from non-salary items such as textbooks and other teachingaids to salary items. Thus nonsalary expenditure asa portion of total public expenditure tends to decreaseover time in ma. developing countries, probably

affecting the optimal mix of inputs in educationalproduction.

Finally, the lack of information on private costs ofeducation makes any attempt to compare privateeducational expenditure difficult. Central govern-ment accounts sometimes provide data on privateeducational expenditure, but such data are problem..atic in that they often include other private consump-tion 0.7penditure or are unreliable. In a recent study,Tan ,A.985a) used such data to examine the macro-trend in private expenditure for 1960-1978. Hersample included 42 developed and developing coun-tries. The study found that there was significantvariation in private educational expenditure in de-veloping countries and that it ranged from less than1% to about 3% to 4% of total private consumptionexpenditure.

Schiefelbein reported some data on private educa-tional expenditure in Latin American countries. Theratio of private educational expenditure to the totalexpenditure of the ministry of education was found tobe 1.11, .51, .32, .13, and .04 for Brazil, Colombia,Chile, Venezuela, and Argentina respectively (1986:29). In a detailed study of rural primary schools inColombia, private contributions accounted for anaverage of 30% of the total education cost per pupil(Paulsen, 1981).

Wolff provided measures of the direct private costsfor students in secondary schools in nine EasternAfrican countries (1985: 51-55). The ratio of totaldirect private cost to total cost per student variesaccording tr, the type of secondary schools and thecountry. It ranged from 0% for day schools inSomalia for 1981-82 to 81% for assisted Harambeeschools in Kenya for 1981-82. The ratio for boardingschools was consistently higher than that for dayschools. On the average, direct private costs ac-counted for one third of the total cost per pupil. Highdirect private costs were also reported in Tan's studyof secondary schools in Tanzania (1985b). She foundthat even though state school students paid no fees,their school-related expenditure added up to US $13dper student in 1981. The direct private costs forstudents in private schools were even higher (US$439).

Tilak's (1985) study of the costs of education inIndia indicates the large amounts of both direct andindirect private costs. For 1979-80, public educa-tional expenditure was found to be about 3.9% ofGNP, indirect private cost (foregone earnings) wasestimated to be 4.2% of GNP, and household expendi-

ture on education (mostly direct private cost) was ashigh as 1.9% of GNP. Thus the real cost of educationin India in 1,)79 -80 was about 10% of GNP; over halfof it was private cost. The study also computed theshare of public educational expenditure for both thecentral government and state governments. For1976-79, the central go "ernment contributed to lessthan 10% of public educational expenditure. Thisundersccres the large contributions by noncentralgovernments, especially for countries with a decen-tralized system of educational finance.

In summary, although information on privatecosts of education is still lacking in most aevelopingcountries, the available evidence indicates that directprivate costs are significant in secondary schools ir.some countries. If we also take into account theindirect private cost of foregone 4 arnings, total pri-vate cost will be considerable.

etc.), for different curricula (general vs. vocational),and for different programs or subject areas. Whilethere are numerous ways of expressing unit costs, themost common one is based oil per-pupil enrolled. Twomajor findings have emerged from these studies:there are significant disparities in unit costs of edu-cation ir. each of the above situations, and there areclearly identifiable patterns in unit costs.

Typical unit costs (per pupil basis unless stateotherwise) for various levels of education at the coun-try level are given in Tables 2 and 3. Since countriesvary greatly in their level of per capita income, unitcosts are usually compared by using the ratio of perpupil public educational expenditure to per capitaGNP. Included also in the two tables are the cost-atios among the three levels of education. Severalobservations can be made: (1) there are considerable

Table 2: Costs per pupil of public education at various levels as aproportion of GNP per capita, and cost ratios between levels(around 1980)

Region Costs per pupil as a Pro-portion of GNP per capita

Primary Secondary Higher

Cost RatiosSecondary/ nightly'

primary primary

Eastern Africa 0.16 0.85 10.40 5.3 65Francophone 0.29 1.43 8.04 4.9 28

AfricaSouth East Asia

and Pacific0.11 0.20 1.18 1.8 11

South Asia 0.08 0.18 1.19 2.3 15Latin America 0.09 0.26 0.88 2.9 9.8All Developing 0.14 0.41 3.70 2.9 26Countries

Developed 0.22 0.24 0.49 1.1 2.2Countries

Source: Based on Wolff (1985), Table 1, p. 2. Cost ratios were computed by thisauthor.

Patterns and Determinants of Unit CostsPrevious studies have estimated unit costs of

education or a wide range of situations: for educationat various leN is of aggregation (centre? state, re-gional, local, and rIchool), for different levels of educa-tion (primary, secondary, and territory), for differenttypes of schools (urban vs. rural, public vs. l.. ivate,

20

differences in both the effortratios and the relative costratios between developed anddeveloping countries, betweenregions in the world, and be-tween countries in a region; (2)compared to developed coun-tries, developing countrieshave shouldered a heavierburden in their provision ofeducation, since they havelower levels of per capita GNPand higher unit-cost ratios;and (3) the unit cost of highereducation is very high relativeto that of primary education indeveloping countries.

Significant variations inunit costs can also exist be-tween states or provinces in acountry. For a given country,unii. costs are usually com-pared on a dollar-per-pupilbasis. For example, using datafor 14 states in India in 1976-1977, Tilak found that the ratio

of the unit cost of the highest spending state to thatof the lowest spending state was respectively 5.2 forprimary education, 6.3 for middle school, 1.9 for highschool, and 2.4 for higher education (1985: 38). Also,the relative cost ratio between higher education andprimary education in 22 states ranged between 3.07in Uttar Pradesh to 40.77 in Sikkim in 1976-77 (p.45). But this relative cost ratio was not significantly

correlated with the level of economic development ineach state (as measured by per capita state income).

Tilak (1985:43) provided further evidence of dis-parities in unit costs between: drat areas and urbanareas for several levels of education in the state ofAndhra Pradesh in India. He found that for bothprimary and middle schools, rural areas had higherpublic cost per pupil than urban areas. But sincestudents in urban areas had higher foregone earn-ings and direct private costs, the resultant social costper pupil was higher in urban areas than in ruralareas for these two levels of education. But for higherlevels of education, both the public cost per pupil andthe total private cost per pupil were higher in urbanareas, again resulting in higher social cost per pupil

in urban areas.Higher public cost per pupil in rural schools was

also reported fu;. Thailand. Based on a survey of tenrural secondary schools and ten urban secondaryschools, Reiff found that the public cost per pupil inrural secondary schools was about 40% higher thanthat of urban secondary schools. This was due to thefact that the former schools had a lower pupiUteacherratio and the latter schools usually had a largerenrollment (1972a: 216).

Cost studies of rural schools in developing coun-tries are few in number. A well-designed study of thecosts of rural primary schools was conducted in Co-lombia (Paulsen, 1981; OFISEL, 1982). The entirestudy consisted of six individual case studies, one for

each primary school in six dif-ferent rural contexts. The re-sults warned against treatingall rural primary schools thesame because the costs perpupil varied significantlyamong the schools. They alsoindicated that private contribu-tions were a significant part ofthe total cost for rural primaryschools in Colombia, rangingfrom 15% to 48% of the total costin the six case studies.

Ge-sierally, local support ofpublic primary schools is animportant source of funding,and private contributions inkind are common in rural areas,especially the poorest ruralareas (Schiefelbein, 1986: 32).

Besides geographical vari-ations, cost disparities also ex-ist for different types of schools.For secondary education,boarding schools are morecostly than day schools. InEastern Africa in the early198Cs, the relative cost ratiobetween boarding school andday school was found to be 1.1,1.3, 4.0, 1.4, and 1.03 respec-tively for Botswana, Lesotho,Somalia, Uganda, and Zambia(Wolff, 1985: 51-55). For four-teen countries in Latin Amer-ica, Asia, and Africa, secondary

Table 3: Costs per pupil as a proportion of GNP per capital, and costratios, Eastern Africa (around 1980)

Countries GNP percapita (US$)

Costs per pupil as a propor Cost ratiostion of GNP per capital secondary/ Higher/

Primary Secondary Higher Primary Primary

Botswana 780 .20 1.03 6.5 5.2 33Burundi 230 .23 1.52 :19.7 6.5 55Comoros 260 .18 .,',2 - 2.9 -

Djibouti 480 .63 .80 - 1.3 -

Ethiopia 140 .19 .66 11.1 8.5 58Kenya 390 .14 .52 9.9 3.7 7.1

Lesotho 540 .07 .43 11.4 6.1 163Madagascar - - - - -

Malawi 200 .06 1.08 15.9 18.0 265Mauritius 1080 .11 .19 2.9 1.7 26

Rwanda 220 .13 1.73 14.0 13.3 108Somalia 280 .10 .30 3.2 3.0 32Sudan 360 .12 .29 4.3 2.4 36Swaziland 760 .09 .36 3.6 4.0 40Tanzania 280 .12 2.94 30.9 24.5 258Uganda 210 .03 .20 10 5 6.7 350Zaire - -

Zambia 580 .12 .66 6.3 5.5 53Zimbabwe 870 .16 1.24 12.7 7.8 79

Average 450 .16 .85 10.4 5.3 65

Source: Based on Wolff (1985), Annex I, Tables 5 & 18.by this author. GNP per capita io in current dollars.

Cost ratios were computed

21

boarding schools built in the 1960s were two to uhreetimes more expensive than day schools built in thesame period (Hutton and Rostron,1971 ). Similar costratios were found for secondary schools in Morocco(Radi, 1982). The difference in cost is mainly due tothe much higher building costs of boarding schools.

The unit costs of academic education and voca-tional education at the secondary level have been thesubject of many economic analyses, often fueled bythe long-standing debate over the relative merits ofthe two types of curriculum to meet the employmentand manpower needs of a developing economy(Grubb, 1985). In general, unit costs of vocationaleducation are found to be higher than those of aca-demic or general education. But since there aredifferent types of vocational training and differentinstitutional arrangements for vocational education(such as vocational schools, company-affiliated voca-tional schools, schools with a diversified curriculum,etc.) in different countries, the difference in costbetween the two types of curriculum varies widely.For example, in Latin America, the ratio of expendi-ture per pupil of industrial/agricultural education tothat of academic education was 4.2 for Paraguay in1983, 1.5 for Colombia in 1981, and 1.4 for Chile in1969 (Schiefelbein, 1986: 8). In a recent study ofdiversified secondary schools (called INEM schools)and traditional secondary schools in Colombia, theunit costs for academic and co -nercial tracks inINEM schools were found to be higher than those inthe corresponding traditional schools, but the situ-ation was reversed for the industrial, agricultural,and social service tracks (Psacharopoulos andLoxley, 1985).

In Thailand, data from the late 1960s showed thatat the secondary level, vocational education wassignificantly more costly than either academic orcomprehensive education, and that technical educa-tion was even more costly. Unit cost averaged 4869bahts for technical schools, 2971 bahts for vocationalschools, and only 1162 bahts for the other academicand comprehensive schools. Compared to academicand comprehensive schools, the average teacher costper pupil was four times as expensive in technicalschools, and twice as expensive in vocational schAnd for nonteacher cost per pupil, the correspondingcost ratios were 8.8 and 6.4 respectively (Bennett,1975: 50-56). In a study of an automobile company inChina, the cost of vocational curriculum was alsofound to be significantly higher than that of academiccurriculum. According to school administrators, the

22

per pupil cost of a secondary vocational school affili-ated with the auto company is 1000 YUAN per year,while that of a secondary academic school is 250YUAN per year (Min and Tsang, 1988).

But in Tanzania, the differences in unit cost be-tween vocational and academic streams in publicsecondary schools are relatively modest compared tothose for other countries. In 1981, agricultural, tech-nical, and commercial streams were respectivelyonly 20%, 13%, and 9% more expensive than theacademic stream (Psacharopoulos and Loxley,1985).

Obviously, cost comparison alone is incomplete forinforming the debate over the choice of curriculum insecondary education; the benefits and effects of dif-ferent curricula have to be taken into account, too.Cost-benefit and cost-effectiveness studies on thisissue will be taken up in the next major section of thispaper.

Cost disparities by curriculum also exist in highereducation. Data for a sample of developing countriesshow that university subjects like agriculture, sci-ences and engineering are on the average more thantwice as coaly as general subjects. And in terms ofthe magnitude of recurrent cost per pupil, the de-scending order for the subjects was: agriculture,sciences, engineering, arts, humanities, and socialsciences (Psacharopoulos, 1982). Subjects near thetop of the listhave higher recurrent expenditures dueto lower student-teacher ratios and higher capital ex-pendituroc.

The t. we discussion documents the significantdisparities in unit costs of education between coun-tries, states/provinces, regions, levels of education,types of school, and curricula. But it also displaysconsistent patterns or behavioral characteristics inunit costs that reflect similar operations in educationacross countries.

For education systems in developing countries,expenditure per pupil exhibits the following pat-terns: (1) it rises with the level of education, (2) it isdominated by personnel costs, although the propor-tion for personnel costs decreases with the level ofeducation, (3) it is higher for boarding schools thanday schools at the secondary level, (4) it is generallyhigher for vocational education than academic educa-tion at the secondary level, (5) it is higher for engi-neering and science subjects than arts and humani-ties at the tertiary level, and (6) it has a built-intendency to rise over time.

The factors affecting unit costs in developingcoun-tries have been considered by a number of studies

40 C

...policymakers should be warnedagainst designing an "average" pol-icy to be applied to diverse settings.

(Coombs and Hallak, 1972; Eicher, 1984; Wolff, 1985;Tibi, 1986). The findings of these studies are synthe-sized below.

A major factor affecting educational costs is thetechnology of educational production. In countriesacross the world, education takes place predomi-nantly in the traditional school and university, withsimilar organization, curriculum, pedagogical meth-ods, management and monitoring procedure. Alter-native technologies of education, such as out-of-school distance teaching, operate only at the marginof the education system. The adoption of a commontechnology is the major reason for the similar pat-terns of unit costs found around the globe. Changesin educational production in terms of the staffingpattern, class size, school organization, and otheractions will affect unit costs.

The labor-intensive technology of traditionaleducation explains the dominance of teacher costs.Teacher compensation is an important cost. determi-nant. Expenditure on teacher salaries depends on anumber of factors. These factors include the salarystructure, the current pattern of qualifications ofteachers, the age composition of the teaching force,the average salary of teachers, and the pupil-teacherratio. In most developing countries, salary levels aregraduated on scales based on qualifications and yearsof service. The average salary of teachers is deter-mined by the supply of and demand for teachers,alternative employment opportunities, and the bar-gaining power of teacher unions. There are signifi-cant regional dirm.ences in the average level ofteacher salaries. For example, primary school teach-ers' salary as a proportion of per capita GNP is low inLatin America and Asia, but quite high in West Africaand Francophone Africa. In many developing coun-tries, teacher salaries in current dollars have risenover time, but the real level of teacher salaries hasactually declined as a result of high rates of inflation.

Two other factors that affect unit costs are therates of utilization of educational resources, andrates for dropouts and repeate's. If educationalresources are more fully utilized, then more pupilscan be schooled at a given level of educational expen-diture, and thus expenditure per pupil will drop.Studies on utilization rates are reviewed in flit) fol-lowing subsection. Similarly, rates of dropouts andrepeaters affect the effective cost of educating a pupil.Lower rates of dropouts and repeaters imply lowerunit costs ui education.E

Thus, for developing countries, per pupil expendi-

23

ture on primary education de.,iends mostly on teachersalaries and the pupil-teacher ratio. The effectivecost of expenditure per graduate is affected also byrates of dropouts and repeaters. At the secondarylevel, nonteacher costs such as boarding facilities forboarding schools and equipment costs for vocationalschools are significant cost determinants as we'l asteacher salaries and the pupil-teacher ratio. At thehigher education level, teacher salaries, the pupil-teacher ratio, and capital costs are all importantfactors. The higher unit costs at higher levels ofeducation can be attributed to higher teacher sala-ries, lower pupil-teacher ratios, and higher per pupilcapital costs.

Utilization of Educational InputsThe rates of utilization of educational inputs in

schools also have direct implications for unit costsand educational efficiency. To the extent that thereis considerable underuilization of school resources,educational expenditures can be reduced withoutaffecting the number of students served. Alterna-tively, the number of students served can be in-creased without incurring additional cost. In eithercase, unit cost is lowered and the technical efficiencyof the school is raised.

Studies on the utilization of educational inputs indeveloping countries can be classified into twogroups. The first group consists of school-level stud-ies that survey the rates of utilization of schoolresources like teachers, and school facilities. Thesecond group estimates cost relationships to explorethe existence of economies of scale. These two groupsof studies are reviewed separately in the followingsection.

Rates of Utilization of School ResourcesAn early study of resource utilization in secondary

schools was conducted by the Asian Regional Insti-tute for School Buildinp Research for the governmentof Ceylon, now Sri Lanka (ARI, 1969). The studyconsisted of a survey of eleven schools in four prov-inces, encompassing both rural and urban schools, aswell as schools at different altitudes. It found signifi-cant wastage in the utilization of space due to anumber of factors: the use of large rooms for smallclasses, the use of some teaching spaces for only partof the school day, an the failure to use availablefacilities in neighboring schools. '30, in someschools, an excess of teachers led heads of schools to

deliberately provide more subjects and smallerclasses so as to create enough class periods for teach-ers to qualify for salary. The report recommendedthat an "area check" be used by the government toevaluate requests by local school administrators foradditional facilities. The area check was based on astandard space of14 square feet per place for primaryschools and 32 square feet per place for secondaryschools.

Significant underutilization of school rP Jurces atthe secondary level appears to be a widespread phen-omenon in many developing countries. In a compre-hensive Falvey of100 schools in 14 countries in LatinAmerica, Asia, and Africa, Hutton and Rostrcn(1971) found low rates of utilization: teacher contacttime ranged from 8 to 19 hours per week, laboratoryutilization had a median time of 18 hours per week,and the median time for the use of the whole schoolwas under 3 hours per day. The low rates of utiliza-tion were probably due to: (1) a standardized class-room size for all grades even though the number ofpupils decreased with the grade level, (2) each classhad its own general classroom, and (3) special class-rooms and laboratories were not used for generalteaching. The authors estimated that by utilizingschool resources more fully, most of the countries intheir study could easily increase secondary enroll-ment by 30%.

Both of these early studies call for better manage-ment of and planning capability for school resources.But the problems of utilization have not disappearedover time in developing countries. In a recent studyof12 countries in Latin America, Asia, and Africa, lowrates of utilization were still reported in someschools. The study found that the variation in teacherutilization was strongly influenced by the quality ofeducational administration and its ability to affectthe distribution of teachers across educational insti-tutions and services. And in general, the utilizationof resources in schools is depe)Ident on the relation-ship between available resources to education andthe provision of school places to pupils, and on thedistribution of available resources (Tibi, 1986).

Actually, a wide variety of cost-reduction strate-gies for primary and secondary education has beenput forward (Eicher, 1984; Wolff, 1985; Mingat andPsacharopoulos, 1985: AID, 1974). They includestrategies that focus on teacher salaries (loweringteacher salaries when they are considered high, andrevising pay scale), low-cost teachers (volunteerteachers, teachers with lower qualifications, and

24

setting a lower requirement for teacher licensing),increasing teaching load (more hours per day, andmore days per year), increasing class size (if it issmall), school reorganization (double shifts, schoolconsolidation, and cluster school/regional coopera-tion), and alternative educational technology (use ofnew educational media). A common drawback ofthese strategies is that they are not based on empiri-cal evidence re rating the impact of cost-reductionactions schi)o1 outputs like student achievement.

Estimates of the rates of utilization of resources inhigher education are available for a number of devel-oping countries (Psachal opoulos,1982: 114-115). Forexample, in Sierra Leone in 1975, actual utilizationin four institutions ranged from 40% to 71% of capac-ity. In Zambia in 1978, faculty wastage rates rangedfrom a low of 17% in agriculture to a high of 48% innatural sciences. But conditions vary among coun-tries. In 1978-79, the student per faculty ratio was aslow as 5 for universities in Botswana, Burundi, andTanzania, but it was 20 for Cameroon, 30 for Ma-dagascar, and 72 at a university in Sudan (Hinchliffe,1985: 81-82). Student contact hours are much lowerat the University of Malawi than they are at Ken-ya:la University in Kenya (p. 53-54).

Several observations can be made from these utili-zation studies. By revealing costly wastage practicesin schools and universities, a careful examination ofresource utilization can lead to significant cost sav-ings. Universities, in particular, should receive closescrutiny because of their high unit costs. Given thelarge number of secondary schools, the costs of find-ing out the utilization rates at the secondary level canbe greatly reduced by surveying a representativesample of schools. And lastly, in analyzing strategiesfor cost reduction, careful attention should also begiven to the potential impact of these strategies onschool quality. Cost reduction at the expense ofschool quality does not raise the technical efficiencyof the school.

Educational Cost Functions and Economiesof Scale

An education cost function (ECF) relates the mini-mum cost of education to tl, e level of educationaloutput, given input prices and the technology ofeducational production. An ECF can be estimatedstatistically to determine the relationk hip betweenaverage costs and marginal costs, and thus the econo-mies or diqeconomies of scale. If educational outputcan be increased at relatively low additional costs,

marginal costs will bo lower than average costs.Then each additional unit of output bermes cheaperto produce; there are economies of scale. Economiesof scale can result from improved technology or or-ganization of production as output is increased. If,however, educational output can only be increased atrelatively high additional costs, marginal costs willbe higher than average costs. Then each additionalunit of output becomes more expensive to produce;there are diseconomies of scale. Diseconomies ofscale often result from inefficiency of technology ororganization of production as output is increased.Finally, ifmarginal costs are equal to average costs asoutput increases, there are constant returns to scale.

Previous studies on ECFs have been subject to anumber of empirical and theoretical difficulties (Fox,1981). To begin with, the unit of analysis often doesnot correspond to the research question at issue.Data limitations have forced researchers to use dataat a higher level of aggregation (e.g. state or schooldistrict level) than they would like (e.g. school level).Because of the multi-dimensional nature of educa-tional output and measurement problems, research-ers have resorted to using alternative "output" meas-ures, such as student enrollment or average dailyattendance. Also, inputs related to capital costs areoften emitted and this can lead to significant estima-tion errors. Moreover, the use of expenditure data asa cost proxy raises a serious conceptual difficulty inthat expenditure levels are often determined in apolitical context and are thug unlikely to be costminimizing. The estimated equation would not be anECF and economic interpretation of the results isdifficult. Furthermore, restrictive functional formsare employed in estimating an ECF. These func-tional forms impose prior assumptions on the tech-nology of educational production that can be unwar-ranted.6 Finally, in estimating an ECF, most studiesdo not control for the quality of schooling or schooleffectiveness. This significant flaw undermines thevalidity of the findings.

Since the late 1960s, numerous studies have beenconducted to estimate cost functions and economiesof scale for the traditional school and university indeveloped countries (see, for example, Riew, 1966;Cohn, 1968; Maynard, 1971; Hind, 1977; Kenny,1982). But such attempts have been relatively recentand few in number in developing countries.

A recent study has looked at traditional primaryand secondary schools in two Latin American coun-tries, Bolivia and Paraguay (Jimenez, 1986). The

sample consisted of 43 primary and secondary Boliv-ian schools and 41 Paraguayan schools. By estimat-ing an ECF based on a flexible functional form(translog function) and using enrollment data ad-justed for school quality (based on student testscores), the study found that the average primaryschools in these two countries did exhibit economiesof scale; the same finding was also true ofthe averageBolivian secondary school. The study found that sub-stitution between personnel and nonpersons in-puts was possible for the Bolivian schools, and thattho size of the physical plant for Paragw.yan schoolswas excessive.

The existence of economies of scale, however, wasnot reported in an earlier study of secondary schoolsin Uganda (Chesswas and Hallak, 1972). Using datafrom eleven secondary schools, the researchers foundno apparent pattern between per pupil expenditureand school size. No ECF was estimated by this study.

A review of 34 studies on American elementaryand secondary schools concludes that "per pupilschool costs appear to be characterized by a U-shapedaverage cost curve. Scale economies do exist over alimited range of student populations" (Fox, 1981:285-286). But so far there is not enough empiricalevidence to draw any definite conclusions for primaryand secondary schools in developing countries.

It appears that there are economies of scale inhigher education in developing countries. Using datafrom 5L leveloping countries, 18 developed coun-tries, and seven oil-producing countries, Psacharo-poulos (1982) observed that the cost per pupil inthese countries declined with the level of tertiary en-rollment. The estim ted cost function indicated thatunit cost declinea rapidly with university enrollmentup to the point of a 3% enrollment ratio and decreasedslowly afterwards.

A more recent study also documents the existenceof economies of scale in higher education. Using datafor 123 developing countries and 20 developed coun-tries in 1979, Lee (1984) found that unit cost for auniversity would decline dramatically up to an en-rollment of 500 students. The rate of decrease re-mained significant between 500 and 10,000 students,but unit cost would level off after 10,000 students.But contrary to the Psacharopoulos study, Lee foundthat enrollment ratio had no independent effect onunit cost. Enrollment ratio was correlated with unitcost only because it was highly correlated with totalenrollment, which influenced unit cost.

Evidence of significant economies of scale was

A careful examination of costpatterns can uncover opportunitiesfor improving the efficiency ofeducational Investment.

found in a study of 136 institutions of higher educa-tion in China (World Bank, 1986). According to thisstudy, recurrent cost per student declined as enroll-ment increased. There would be substantial savingsin per student cost if the level of enrollment wasraised to about 8000 to 10,000 students.

Although there are few studies on the cost func-ticns oftraditional education in developing countries,extensive analyses of cost functions of new educa-tional media projects exist (Jamison and McAnany,19';d; Jamison, Klees, and Wells, 1978; Eicher andOrivel, 1980; Perraton, 1982). A number of conclu-sions can be drawn from these analyses. First, unitcost is sensitive to the choice of the social discountrate. Cost comparison should be informed by compu-tation baseu on a range of racial discount rates (forexample, 0%, 7.5%, and 15%). Second, small mediasuch as radio are much less costly than big mediasuch as television, the ratio of unit costs being about1/5. Third, there are clearly economies of Fcale fornew educational media. For example, a study onprimary instruction using television in the IvoryCoast found that unit cost decreased sharply with thenumber of students enrolled until enrollmentreached 300,000 students (Eicher and Orivel, 1980a).And fourth, such projects usually involve large start-up costs. To reach levels of unit cost comparable tothat of the traditional school, these projects need tohave large enrollments and a long period of operation(10-20 years).

Obviously we cannot compare new educationalmedia with the traditional school on the basis of unitcosts alone; measures of effectiveness are also rele-vant. The cost-effectiveness of new educatit. nalmedia is reviewed in the section titled "Cost-Benefitand Cost-Effectiveness Studies in Education."

In conclusion, the above discussion indicatesclearly that there are gaps in our understanding ofcost relationships in traditional education in develop-ing countries. Further research on cost functions andeconomies of scale in traditional education should beencouraged. But such an effort should not be confinedsolely to a statistical estimation of cost functions andtheir properties; it should also be complemented bydetailed in-school studies of resource utilization.Statistical analyses can reveal patterns of coy rela-tionship, but not the mechanisms that underlie then.In a substantial way, the limitations and methorlologic al issues related to ECFs parallel those of edu-cational production-function studies of student

26

achievement (Hanushek, 1.979). What is needed is anunderstanding of the "process," not correlations only.

Behavioral Characteristics of EducationalCosts and Educational Policymaking

The two preceding sections have reviewed coststudies on the behavioral characteristics of educa-tional costs in developingcountries. The implicationsof the major findings of these studies are consideredhere.

The review documents the existence of cost pat-terns in education. A careful examination of thesecost patterns can uncover opportunities for improv-ing the efficiency of educational investment. Con-sider a number of examples.

Unit costs of higher education are considerablyhigher than those of primary education. A close lookat public expenditures in higher education revealslarge subsidies to university studies. We may ques-tion whether or not the significant cost gaps arecompensated by higher social benefits of highereducation. We need to reexamine the mechanisms forfinancing education. Reallocation of resources be-tween educational levels and alternative financingstrategies may be called for.

Many educational systems have vocationalized ordiversified their schools because of the presumedproductivity and employment advantages of the vo-cational curriculum. In light of the significantlyhigher unit costs of vocational training, we need toexamine the evidence for the presumed benefits andeffects of vocational training.

The dominance of teacher costs in recurrent ex-penditure implies the importance of keeping teachercosts under control. In many developing countries,teacher costs account for over 90% of the recurrentexpenditure on primary education, leaving very lim-ited funds for nonteacher items. The negative conse-quence of an imbalance in the mix of educationalinputs for student achievement has to be confronted.

The significant underutilization of edt, rationalinputs in some schools in developing countries shouldalso be of concern to educational policymakers. Byutilizing teachers and school facilities more fully,more school outputs may be produced without incur-ring additinnal costs.

Last, the costly nature of a linear expansion oftraditional education has prompted policymakers toexplore alternative technologies of education. Given

...disparities in educational costsalso reflect inequities in the distribu-tion of educational resources across

regions and social classes.

a large enrollment and a long operating life, neweducational media can achieve low unit costs. Newmedia may also be an effective strategy to reach stu-dents in thinly populated areas.

Thus, an understanding of the behavioral charac-teristics of educational costs can cont _cute to theidentification of three distinct classes of strategies forimproving efficiency in education: strategies for util-izing existing resources more ly (technical effi-ciency), strategies for realloc;.. .g resources in edu-cation (economic efficiency), and strategies involvingalternative technologies of education.

Although opportunities for increased efficiency ineducation exist, the findings of studies reviewed herealso imply limitations and cautions for educationalpolicymakers. Among the many factors affectingexpenditures on education, there are important"external" determinants that lie outside the locus ofcontrol of educational policymakers. Also, given thepractice of automatic promotion and salary in-creases for teachers, there is a built-in tendency forrising teacher costs. In fact, in a given yeas, theproportion of discretionary expenditure in the totalrecurrent fund is quite small.

Poticymakers are well advised to recognize theimpacts of past and pr:sent decisions on cost require-ments in the future. decision, for example, to raiseteacher qualifications now implies a commitment tomeet the substantial amount of additional teachercosts in many years to come.

Another major finding of the studies is the consid-erable disparity in educational expenditures and unitcosts across diverse settings. This fin dingh as several

policy implications. First, policymakers should bewarned against designing an "average" policy to beapplied to diverse settings. The policy is likely to havevery different cost implications for schools in differ-ent settings. Schools may have different seeds thatrequire different policy treatments. Second, the di-versity of educational settings implies the impor-tance of basing policies on accurate information fromthese settings. Third, disparities in educatio..al costsalso reflect inequities in the distribution of educa-tional resources across regions and social classes.Reducing such inequities can be a desirable policyobjective. And fourth, differences among countriesand regions warn against the danger of uncriticallyadopting practices found useful elsewhere.

Finally, our review demonstrates the usefulness ofindicators of educational costs for understanding thebehavioral characteristics of educational costs. Indi-cators of educational costs include, for example,measures of national effort and fiscal effort, unit costsby level of government, level of education, type ofeducation, curriculum, region, and time, as well asindicators related to the distribution of costs bysources and by input categories. These cost-of-educa-tion indicators provide an anatomical examination ofthe resources used in education. From the perspec-tive of policymakers, indicators are useful for a num-ber of diagnostic purposes: they indicate the state ofaffairs in education, uncover areas of abnormalities,and provide the bases for gaug.ng progress in educa-tional interventions (Johnstone, 1981; Oates, 1986)Thus there are strong arguments for regularly con-structing and examining these indicatees.

Cost Analysis for Educational PolicymakingA Review of Cost Studies in Education inDeveloping Countries

SectionCost-Benefit and Cost-EffectivenessStudies in Education

In contrast to the studies reviewed in the previ-ous two sections that focus on input costs and charac-teristics of educational costs respectively, the studiesreviewed hen consider both input costs and out-comes of education. Two classes of such studies canbe identified: cost-benefit studies that compare edu-cational benefits such as increased earnings to edu-cational costs, and cost-effectiveness studies thatcompare educational effects such as student achieve-ment to educational costs. In this section, cost-benefit studies and cost-effectiveness studies arediscussed separately.

The rates of returnindicate underin-vestment In edu-cation, particularlyat the primaryeducation level.

Cost-Benefit Studies in Education iiiDeveloping Countries

Cost-benefit comparison in education is Jften con-ducted to assess the external efficiency of education.The theoretical and me,liodological bases of cost-benefit analysis have received extensive treatment inthe literature (Woodhall, 1967; Psacharyoulos andWoodhall, 1985); and cost-benefit studie. of educa-tion in developing countries are numeroos. The fol-lowing presents a brief summary of the major find-ings of these studies and their policy impli_7tions aswell as ',he debate concerning the theoretical andmethodological bases of cost-benefit studies in edu-cation.

Cost-benefit studies in education are mostly basedon the rates of return approach to evaluating educa-tional investment and on the human capital theoryregarding the economic benefits of education. Ac-cording to human capital theory, education enhances

28

the skills (human capital) of an individual and thusraises his/her productivity. There is a direct andpositive relationship between education and produc-tivity. In a competitive labor market, a more produc-tive individual is paid a higher wage. Thus, on theaverage, a more educated individual has higher earn-ings. The profitability of education can be measuredby comparing the benefit of education in terms ofadditional lifetime earnings to the cost of education(Schultz, 1961; Becker, 1964; Mincer, 1974). One candistinguish between private rates of return and socialrates of return. Private rates of return to educationcompare the benefits of education to an individut.i tothe costs of education to the individual; they informprivate decisions regarding educational investment.Social rates of return to education compare the bene-fits of education to society and the costs of educationto society; they guide public policies regarding educa-tional investment.

A convenient way to review the major findings ofcost-benefit studies in education in developing coun-tries is to draw upon the work of Psacharopoulos(1973, 1981, 1985) who has collected and analyzednumerous studies from countries in different regionsover the years. As an illustration, Table 4 givesaverage estimates of the private and social rates of

Table 4: Average returns to education (percent)

'legion;Countrytype

SocialPrimary Secondary Higher

PrivatePrimary Secondary Higher

Africa 26 17 13 45 26 32Asia 27 15 13 31 15 18Latin 26 18 16 32 23 23

AmericaInter-

mediate13 10 8 17 13 13

Advanced NA 11 9 NA 12 12

Source: Psacharopoulos (1985), p.586.Note: NA=not available because of no control group ofilliterates.

5 4.1k

return by level of education for countries in differentregions. In his most recent compilation, Psacharo-poulos (1985) analyzed the results for over 60 coun-tries. A nimber of consistent findings can be identi-fied: (1) investment in education is very profitable;rates of return to investment in education are wellabove the benchmark (10%) rate of return to invest-ment in capital; (2) the social and private rates ofreturn to primary education are highest among alleducation levels; (3) private rates of return are higherthan social rates of return at all levels of education,particularly at the university level; (4) in developingcountries, the average return to a given level ofeducation is higher than that in developed counts ies;(5) the rates of return to investment in women'seducation are higher than those for men in develop-ing countries; and (6) at eke secondary level, theaverage return to the traditional academic curricu-lum (16%) is higher than that of the vocational-technical curriculum (12%). At the higher educationlevel, rates of return for programs in humanities andsocial sciences are higher than those for technicalsubjects.

Psacharopoulos discussed the implications ofthese findings regarding policies on educational in-vestrizta. The rates of return indicate underinvest-ment in education at all levels in developir.g coun-tries, particularly at the primary education level.Investment in primary education should receive toppriority. The significant gap between social rates andprivate rates at the university level reflects consider-able public subsidies to university students. Effi-ciency and equity in education can both be improvedby reducing public subsidie to students (mostly fromwell-to-do families), reallocating the savings to pri-mary education, and providing scholarships to students of low-income families (World Bank, 1985).The findings also support expanding women's accessto education. They also cast doubts on the presumedeconomic advantages of vocational training overacademic training, though vocational training is stillprofitable (Metcalf, 1985).

The social rates of return approach to public edu-cational investment is attractive to cost analysts inthat it is analytically simple, its results have expliciteconomic interpretations, and it is grounded in con-ventional economic theory. But even within thetheoretical framework of human capital theory,most rates of return studies are subject to a numberof methodological problems. First, the results arebased on past conditions; they may not be reliablepredictors of future rates of return in dynamic set-

tings. Second, most studies use cross-sectional datainsteat: of longitudinal data in assessing the earningslevel of an individual over time. Third, most studiesuse the quantity of schooling as a measure of humancapital and ignore issues of educational quality andrelevance, thus creating difficulties in interpretingthe findings. Fourth, most studies ignore significantnoneconomic benefits of education and factors otherthan education that influence an individual's em-ployment and earning opportunities, resulting inbiased estimates of the rates of return to education.While some studies have tried successfully to resolvethese methodological problems, most of the rates ofreturn studies, especially those in developing coun-tries, have not. And fifth, in using earnings as a proxyfor productivity, rates "return studies assume thatthe labor market is perfectly competitive. This is notlikely to be true in developing countries where gov-ernments are big employers. However, it seemsunlikely that these methoth,ogical problems willinvalidate the conclusions of the rates of return lit-erature discussed above.

But alternative analyses of the economic benefitsof education have emerged since the early 1970swhich challenge human capital theory and questionthe relevance of social rates of return to education forguiding public educational policies. Psacharopouloshas rightly obse-vA that the "rate of return subjectis still highly controversial in the literature" (1981:329). Soine analysts have put forward differentexplanation.; of the relationship between education,productivity, a id earnings'; they question the theo-retical basis of the social rates of return approach.

Thurow (1972) presents a job-competition modelthat suggests that earnings and productivity aredetermined by demand factors (job structure), notsupply factors (e.g. education). Productivity is acharacteristic of a job, not an attribute of an individ-ual. Education does not raise the productivity of anindividual. He argues that, if everything else is thesame, a more educated individual requires a lowercost of on-the-job training; he or she is put into a moreproductive job with higher earnings. Thus the posi-tive relationship betwe,- -cation and earnings isnet due to the enhance 'human capital, but isa result of the cerrelat ,ween education andtraining costs.

Spence (1973) ext positive relationshipbetween education and earnings by pointing out thescreening function of education in a labor marketcharacterized by imperfect infornition. Educationdoes not raise an individual's prod,,-tivity. Rather it

reflects the ability of an individual. If everything elseis the same, it costs less for more able individuals toacquire education and thus hay*: higher educationalattainment. Employers use education as a signal toidentify more able individuals and pay them higherwages because of their better performance. Sinceeducation reflects ability only, expenditures on edu-cation will not raise the productivity of individuals.Don 1976) points out that job screening using edu-cational credentials often puts intense pressure onthe education system to expand. The expansionprocess is a vicious cycle; as the growing crowd ofyoung people secure diplomas, ever higher educa-tional credentials become necessary to obtai jobs.At the same time, educational expansion let...lb to asevere financial burden on the gcvernment.

The existence of dual or segmented labor marketscan mediate or even fundamentally change the rela-tionship between education and income (Carnoy,1980). This perspective argues that the labor marketis not homogeneous. Instead, it is divided into seg-ments. Workers in some segments (e.g., primarylabor market) have stable and well-paid jobs whileworkers in other segments (e.g., secondary labormarket) have temporary employment and a meagerit ^me. Education is an important factor in decidingin hich segment a worker is located. Thus, theeconomic benefits of education depend on the divi-sions in the labor market. These divisions are theresults of a historical process shaped by economic andpolitical forces. According to this perspective, theassumption of a single, competitive labor market onwhich the social rates-of-return approach is based isuntenable.

Recent analyses indicate that the relationshipbetween education and productivity is more complexthan the direct and positive relationship suggestedby the human capital theory (Tsang, 1987b; Tsangand Levin ,1985). Empirical evidence has shown thatunderutilization of education in production can leadto lower work effort and lower productivity. Thus theeconomic benefits of education depend not only on thequantity and quality of education, but also on theutilization of education in the workplace. Underutil-ized education can be counterproductive.

Finally, some analysts have pointed out that thefocus on t ie productivity and earnings benefits ofeducation is too narrow. It ignores the analysis of thecentral function of education in reproducing the so-cial relations of production in a capitalist economy(Bowles and Gintis, 1976), and the analysis of thedialectical process of and tensions in educational

30

developments in the context of a capitalist, demo-cratic state (Carnoy and Levin, 1985). In reality,decisions on public educational investment are influ-enced by broader economic and political considera-tions, and are not based on social rates of return.

Given the complexities of the issues involved,controversies regarding the relationship betweeneducation and the economy will likely remain. Edu-cational decision makers should be informed aboutthe advantages and limitations of cost-benefit analy-ses in education that are based on the rates of returnapproach.

Cost-Effectiveness Studies in Educationin Developing Countries

By comparing the costs and effects of alternativeeducational interventions, cost-effectiveness analy-sis can inform decisions to improve the internalefficiency of education. Compared to cost-benefitstudies, cost-effectiveness studies are relativelyuncommon in education in developing countries.Most studies for school improvement have concen-trated on school effectiveness, without also combin-ing effectiveness with information on costs. Also,cost - effectiveness studies tend to rely only on thecognitive measures of school effectiveness such asstudent achievement. In the following, we considercost-effectiveness studies of both new educationalmedia and the traditional school.

A prominent application of cost-effectivenessanalysis in education is in the area of new educationalmedia. As the previous two sections of this paperhave already discussed the costing and cost functionsof new educational media projects, issues of cost-effectiveness are considered here.

A comprehensive overview and synthesis of thestudies on the costs and effectiveness of new educa-tional media was conducted by Eicher et al. (1982).The synthesis was based on about 30 case studies andprevious partial syntheses by Jamison and McAnany(1978), and Jamison et al. (1978), These case studiescover a dozen developing countries and several devel-oped countries. They analyze a range of media,including radio, television, computer, and multime-dia systems. The findings of the synthesis by Eicheret al. can be grouped under three topics: methodol-ogy, conclusions on cost-effectiveness comparison,and ovaluation of new educational media.

Consider first the issues on methodology. Thecase studies have followed similar guidelines in theestimation of costs. However, their measurement of

3 u

effectiveness is questionable. In general, new educa-tional media have a variety of effects, encompassingboth internal effects and external effects. Internaleffects refer to effects that are produced during thelearning process, such as cognitive skills, and affec-tive or attitudinal development. External effects (orbenefits) concern the impacts on graduates and soci-ety at large, such as the impacts on a graduate'searnings and occupational mobility, and effects oneconomic growth. The case studies tend to concen-trate on the cognitive-skill measure, thus ignoringaffective and other external effects of new educa-tional media. This preoccupation with cognitiveskills partly reflects a narrow assessment of the goalsfor some of these projects. This narrow focus makesthe comparison of new educational media with thetraditional school problematic in that the latter alsoprovides other socialization and certification servicesto its students.

Actually, few projects of new educational mediaescape the criticism of non comparability with thetraditional school. On the one hand, some of themwere carried uut to serve students not covered by thetraditional school, so these projects were not meant tobe comparable to the traditional school. On the otherhand, few projects were designed in such a way as tocontrol for other variables so that the effects couldunambiguously be attributed to the project treat-ment. Internal validity is difficult to establish forthese case studies. But the use of an experimenialdesign for these projects is costly and will createunrealistic conditions for assessing the applicabilityofthe findings of these projects to concrete situations.

Despite these limitations in methodology, Eicheret al. were able to draw several tentative conclusionsregarding the cost-effectiveness of new educationalmedia:

(1) Students do learn from new educational media.But the use of such media in traditional schools is notcost-effective, at least at the primary level. Withoutaltering the pupil-teacher ratio, new educationalmedia generally add costs without raising effective-ness significantly.

(2) "Little media" such as the radio are more cost-effective than "big media" such as television. Butadvances in microcomputers may change this situ-ation in the future,

(3) Distance teaching may be the only way to edu-cate children in thinly populated areas. Distanceteaching still requires teachers to ^rovide face-to-face contacts with students, and teachers have to betrained in the use of new educational media. The

total cost of primary education cannot be reduced byrelying exclusively on these media.

(4) The effective use of new educational mediarequires a technical staff for operation and mainte-nance. But these technical personnel are in shortsupply in some developing countries.

(5) Given a situation, it is difficult to determinewhich medium is the most suitable. There is no supermedium and several media are often combined in aproject.

Several observations can be made about the evalu-ation of new educational media. Previous studieshave indicated that it is costly to conduct such evalu-ations. Also, lack of neutrality in evaluation exists,especially for the earlier studies. The close relation-ship between the evaluator and a project often leadsto biases, such as the underestimation of costs andgenerous assumptions about project effectiveness(Carnoy and Levin, 1975). Such biases often resultin a favorable evaluation for new educational media.The scope of the evaluation is also limited to issues ofinterest to the sponsoring agency. Moreover, experi-ence has shown that the success of a project is not dueto the application of media technology per se, butdepends on a series of circumstances concerninghowthe technology was applied; that is, on the implem-entation process. This implies that an importantaspect of project evaluation has to do with the ini-plementation process and the contexts for successfulapplication.

Eicher et al. (1982: 130-133) concluded by suggest-ing further research on both the effectiveness andcosts of new educational media. Effectiveness re-search is needed in four areas: effects of distance-learning systems, effects of nonformal education,external effects, and problems in design and meas-urement. Further cost studies should evaluate thecosts of adult education and higher education pro-grams in developing countries.

Jamison and Orivel (1982) reviewed the cost-effectiveness of fourteen distance-teaching projectsfor school equivalency. These projects lead to stan-dard educational certification at various levels ofeducation. Eight of the fourteen projects are in-school projects, and the remainder are out-of-schoolprojects. The distinction between the two types ofprojects is an important one in that it reflects signifi-cant differences in the technology of educationalproduction. In-scl ,00l equivalency projects involvefrequent group meetings between pupils and teach-ers; media do rr substitute for teachers but for theskill of teachers. Out-of-school equivalency projects

have infrequent meeti.igs between pupils and teach-ers; there is more substitution of teachers by media.

The costs of the projects were estimated by em-ploying well-developed costing procedures for neweducational media. Again, effectiveness proved to bemore difficult to measure. Evidence on effectivenessis fragmentary. Since these projects are for schoolequivalency purposes, it is assumed that they providetheir students with a service similar to that of thetraditional school so that the cost-effectiveness ofthese projects and their traditional counterparts canbe compared by their unit costs only. But such anassumption is problematic, as the distance- teachingprojects and the traditional school have differenteffects on students that cannot be captured by theirschool-equivalency purpose alone.

Several findings emerge from this review. First,these projects have large fixed costs, often exceeding50% of the total cost. This cost structure is thus verydifferent from that of the traditional school which isdominated by teacher costs. Also projects using radiohave lower unit costs than those using television.

Second, given the high fixed costs, cost per pupil issensitive to enrollment. For these projects to beviable in terms of per pupil cost, they must have largeenrollment and long operating life. For example, atthe secondary level, projects with less than 10,000pupils per year are risky. But one cannot generalizethis figure to all situations. In general, the thresholdfigure is determined by a number of factors: the levelof school equivalency, the choice of media, the cost ofthe closest alternative program, the ratio of fixedcosts to variable costs, and other factors that arecountry-specific.

Third, distance-teaching equivalency projects ap-pear to be effective and cost-effective. For manyemployed adults, these projects provide the only wayfor them to study. The projects also expand theopportunities for education to groups not previouslycovered by the education system. Most of theseprojects have lower costs per pupil than their tradi-tional counterparts. The significant reduction incosts for out-of-school projects is msinly due to thelarge increase in the pupil-teacher ratio.

In short, new educational media appear to be at-tractive when they are used outside the traditionalschool setting. But given the lack of a systematicevaluation of the cost-effectiveness of distance teach-ing, firm conclusions are unwarranted.

Applications of cost-effectiveness analysis to tra-ditional school inputs and conventional educationalinterventions in developing countries are relatively

32

few. It is only in the past few years that some wellexecuted cost-effectiveness studies in developingcountries have begun to surface. This state of affairscan probably be attributed to a combination of rea-sons: the neglect of educational costs and an earlypessimism about the effectiveness of school factors.

Cost-effectiveness analysis is a methodology forevaluating alternative strategies that takes intoaccount both the costs and effects of the strategies.But in the past, most of the efforts at improving schooltended to concentrate on the study of the determi-nants of student achievement. General evaluationstudies often do not incorporate a cost component intheir analyses (Levin, 1987).

The evolution of our understanding of the deter-minants of school effectiveness is also pertinent. Inthe past twenty years, a large amount of research wasconducted on the determinants of student achieve-ment, mostly based on an input-outputor production-function approach. Based on the works of Coleman(1966), Jencks (1972), and others in the U.S., theresearch reviews in the early 1970s concluded thatthe most important determinants of student achieve-ment were family background and socioeconomicfactors and that school variables had little impact onachievement (Averch et al., 1974). The negativeassessment of the impact of school variables stu-dent achievement was reiterated in a review of stud-ies of input-output relationships in nine developingcountries (Alexander and Simmons, 1975). Thisimplies that interventions involving school variablesare not effective strategies for raising studentachievement.

This negative assessment, however, has beenchallenged by further studies in recent years. In theU.S., studies on school effectiveness have indicatedthat although school expenditure and teacher quali-fication are not significant factors, school climate andschool organization are relevant factors (Brookoveret al., 1979; Purkey and Smith, 1933). It was alsofound that school variables had stronger effects onachievement in poor countries than in rich countries(Heyneman, 1980; Heyneman and Loxley, 1983).Moreover, studies in developing countries have foundthat school inputs like textbooks (Heyneman, et al.,1981), teachers (Husen, et al., 1978) and school-management variables (Arriagada, 1983) did have animpact on student achievement. Recent reviews havealso concluded that school variables can be effectivein raising student achievement (Fuller, 1985;Schwir.e, et al., 1986). These recent findings have ledsome researcher« to shift their attention from the

consideration of whether or not school variables areeffective to analyses of the cost-effectiveness ofschool interventions (Psacharopoulos and Woodhall,1985: Chapter 8).

A conspicuous example is the provision of text-books to promote student achievement. Actually, thisschool intervention is not a new one. In Mexico, forexample, the provision of free textbooks to primary-school students was a key part of a national educa-tional policy (Neumann and Cunningham, 1982).This policy, however, was based on a belief ratherthan cost-effectiveness analysis. Supportive evi-dence came from a recent study of a textbook projectin the Philippines (Heyneman, et al., 1984). Thestudy found that the provision ofbetter textbooks andthe increase of the ratio of textbook to pupil from 1 :10to 1:2 bad significant and consistent improvement instudent achievement in science, mathematics, andlanguage. The additional cost incurred was only amodest 1% increase in per pupil educational expen-diture per year. Moreover, students of low-incomebackground appeared to benefit most in terms oflearning. The study also found that raising thetextbook-pupil ratio to 1:1 was not cost-effective.This and other studies (Jamison, et al., 1981) suggestthat the provision of textbooks is a cost-effectivestrategy to improve student achievement, especiallyat the primary level. And with regard to studentachievement, the research indicates that it is a mis-allocation of resources not to spend (or to spend verylittle) on textbooks.

Another area for cost-effectiveness analysis isteacher training. A recent study of teacher trainingin Kenya using correspondence and radio courses(Hawkridge, et al., 1982) found that this method ofteacher training was effective in terms of plannedobjectives; but it was also very costly. The study,however, did not compare the cost-effectiveness ofthis method with other teacher training methods.

Taylor (1983) studied two distance-teaching proj-ects for upgrading teacher qualification in southernAfrica, the Bophuthatawana Teacher UpgradingProject (BTUP) and the Lesotho In-Service Educa-tion for Teachers scheme (LIET). The major taskswere to estimate the costs of these projects Lind tocompare their cost-effectiveness with traditionalteacher-training programs. In contrast to the earlierstudies whip h reported significantly lower unit costsfor the distance-teaching method, Taylor found thatth' recurrent cost per pupil for BTUP was not lowcompared to that of traditional secondary schools in

southern Africa. The unit cost of .1.41ET was morethan three times as high as that of BTUP, mainlydue to a larger staff and higher salary costs. Butthese comparisons were based on direct costs. Taylorspeculated that if indirect costs (foregone earnings)had been taken into account, then the distance-teaching projects would have become relatively inex-pensive. BTUP and LIET were also comparable totheir traditional counterparts in terms of passingrates in certifying examinations, a crude measure ofeffectiveness. These two projects would become morecost-effective if their enrollments were substantiallyincreased.

In general, good estimates of both the costs andeffects of alternative teacher-training methods arestill deficient. Further research in this area is neededin order to draw more definite conclusions.

The cost-effectiveness of alternative curricula atthe secondary level is the subject of some recentstudies. The interest in this subject is also derivedfrom the long-standing debate over the merits anddemerits of the vocational and general curricula. Thecosts and rates of return of the two curricula havebeen treated in Section III and Section IV respec-tively.

A detailed study of the impacts and costs of cur-riculum diversification was conducted in two devel-oping countries, Colombia and Tanzania (Psacharo-poulos and Loxley, 1985). In Colombia, diversifiedsecondary schools (INEM schools) were comparedwith traditional secondary schools on the basis ofunit costs and student achievement in academic andvocational subjects: industrial, agricultural, aca-demic, commercial, and social service. Using data onthe 1981 cohort, the study found that cost-effective-ness varied with the tracks. Industrial, agriculturaland social service tracks in INEM schools had lowerunit costs than those of the traditional school andtheir students had higher test scores in academic andvocational subjects. Academic and commercialtracks in INEM schools had higher unit costs thanthose of the traditional school and their students hadhigi er test scores. In short, for the 1981 cohort,INEM schools did ',aise student achievement andthey were more cost-effective than the traditionalschool in some curricul_um tracks.

In Tanzania, all lower-secondary schools providea diversified curriculum with academic and voca-tional streams. The study thus compared the costsand student achievements between the streams,Using data on the 1981 cohort, it found that, corn-

...evidence of cost-effectiveness for tradi-tional school inputs is either lacking,incomplete, or not yet generalizable.

pared to the academic stream, the technical streamwas more cost-effective. The agricultural streamenjoyed a modest increase in achievement; but it was19% more expensive than the academic stream. Thecommerce stream was 9% more expensive than theacademic stream and it had mixed results on studentachievement.

Cost-effectiveness information on other tradition-al school inputs is lacking.

What have we learned from the studies of the cost-effectiveness of alternative school interventions forimproving the internal efficiency of education? Not alot. The provision of textbooks has been found to beeffective and cnst-effective. This finding warnsagainst any decision for cutting non-teacher costs(including expenditure on textbooks) to an extentthat upsets the balance in the mix of school inputs.But evidence of cost-effectiveness for other tradi-tional school inputs is either lacking, incomplete, ornot yet generalizable. Given the need to utilize scarceeducational resources efficiently, particularly duringthe current period of tight fiscal constraints andunmet demands, this situation is indeed unsatisfac-tory. The situation can probably be improved bycom-bining efforts to strengthen the research basis of cost-effectiveness analysis and the utilization of researchfindings in educational policymaking.

The research basis for cost-effectiveness analysisin education can be strengthened by simultaneousactions in three areas. First, evaluation of educa-tional interventions using the cost-effectiveness ap-proach should be encouraged. As pointed out previ-ously, many past evaluations considered school ef-fects only, without taking costs into account and werethus incapable of informing decisions about efficiencyin education. Evaluation studies can be improved byadding a cost-estimation component. Also, althoughthe analysis of the costs and effects of an educationalintervention can be analytically separated into aprior conventional evaluation of effectiveness and asubsequent cost-estimation component, it is metho-dologically more superior and probably less costly tomake the analysis of costs an integral part of theevaluation process (Levin, 1987).

Second, the usefulness of cost-effectiveness com-parison is strongly affected by the availability andreliability of data on costs. Obviously, cost compari-son cannotbe made without data on costs, and resultsof cost-effectiveness comparisons can be very mis-leading if inaccurate cost data are used. For manycost-effectiveness applications, information on the

34

costs of individual items at the school level is needed,such as the price of a textbook or an equipment.Macro-level budgetary data are of no use for suchapplications. Only surveys of costs at the school levelcan yield the required information (Eicher, 1984).Variation in price and other disparities in costs forschools in different regions also warn against usingaverage or aggregate estimates uncritically.

Third, school produces many effects on individu-als. But studies of effectiveness, cost-effectiveness,and internal efficiency tend to focus on cognitiveeffects and ignore other important non-cognitive ef-fects of schooling. To the extent that non-cognitiveeffects can influence a person's performance in theworkplace or the quality of life after school, theyshould also be considered in effectiveness research.We should be aware that schools may be internallyefficient in raising test scores but externally ineffi-cient in socializing students for adult life. In addi-tion, cost-effectiveness studies depend on informedresearch on school effectiveness. It is only by ade-quately understanding the factors that contribute toschool effectiven ss that we can appropriately iden-tify the alternatives for improving schools and com-pare their cost-effectiveness.

The utilization of findings of cost-effectiveness7.esearch should receive no less attention than theresearch effort itself. Consideration of post- effective-ness should be part of the decision making process.Cost-effectiveness analysis is useful in informingdecisions in that it nrovides a framework tor identify-ing, collecting, and analyzing relevant informationon both costs end effects. Previous experience hasshown that some cost-effectiveness studies wereconducted only "after the fact." Such a practicereduces the usefulness of cost-effectiveness analysisin informing decisions.

But cost-';Nect4.veness comparison should not beused as the sole basis for a decision; other considera-tions are relevant. For example, alternative educa-tional interventions may have different distributionsof effects and costs on different social groups notreflected by such comparison. Thus different inter-ventions may have different equity implications.Also, some interventions may be more easily imple-mented than others (Schwille, et al., 1986: 91-97).These considerations can affect the choice and thelikelihood of success of an educational policy. Theseother considerations should be combined with cost-effectiveness comparison in informing policy deci-sions.

Cost Analysis for Educational Policymaking:A Review of Cost Studies In Education inDeveloping Couniries

Section V:Policy Recommendations

Based on the review of educational cost studiesin developing countries, we can make the followingrecommendations regarding cost analysis and poli-cymaking in education.

(1) Incorporating cost analysis into educationalpolicymaking. Educational decision makers shouldbe made aware of the range of applications of costanalysis to policymaking in education. These appli-cations include estimation of costs, evaluation of fi-nancial feasibility, cost-benefit and cost-effectivenssanalyses, as well as the regular anatomical examina-tion of resource utilization in education. Cost analy-sis should be integrated into the process ofpolicy-making in education.

(2) Strengthening the informational basis of edu-cational costs analysis. The need to strengthen theinformational basis of cost analysis for educationalpolicymaking is both self-evident and urgent. Adatabase on educational costs should be constructedto provide reliaole and timely information on therelevant categories of public costs (broken down byinput items and functions) and private costs (bothdirect and indirect costs), for various levels and typesof education, over time, and at various administra-tive levels (school, local, regional, and central levelsof government). Besides data on educational costs,four other types of data are also needed for most ap-plications to edtcational planning and policy analy-sis (Tsang, 1987a).

The four types of data are: educational quanti-ties, educational prices, educational norms, and so-cio-economic data. Educational quantities refer tothe quantities of inputs to and outputs from educa-tion. They include data on student enrollments,graduates, repetition and dropout rates, number ofteachers and other school personnel (by age, experi-ence, and qualification), and physical inputs. Educa-tional prices refer to the prices of educational inputssuch as the salary structure and other compensationsfor school personnel, and prices for other school in-puts. Educational norms refer to the various normsor standards used in school. They include informa-tion on class size, the physical specifications of aschool, staff -student contact hours, the ratio of seniorstaff to junior staff; and so forth. Socio-economic

data refer to Lata on national output, cost-of-livingprice indices, and public expenditures. Data forquantities, prices, and norms are to be provided insufficient details to match those of the cost data.Figure 2 summarizes the inputs to and applicationsof a database for cost analysis. There is obviously acost for managing a proper database of educationalcosts, but this cost is likely to he more than compen-sated by the gains from better informed decisions.

(3) Research on educational costs. This review hasalso identified at least three areas for further re-search on educational costs. First, studies should beconducted to examine how educational resources areutilized at the institutional level. Such studies will

The need to strengthenthe informational basis

of cost analysis for edu-cational policymaking

is both self-evident andurg9rIt.

assess the extent of underutilization of educationalinputs or imbalance in the mix of educational inputs.Second, the gaps in our understanding of cost rela-tionships in traditional education can be addressedby further research on cost functions and economiesof scale in traditional education, to be complementedby detailed in-school studies of resource utilization.Third, there is a lack of cost-effectiveness studies oftraditional school inputs (other than textbooks).There is not much information, for example, on thecost-effectiveness of alternative teacher-trainingstrategies, or the cost-effectiveness of differentschool inputs for raising school output in developingcountries.

(4) Controlling costs and improving efficiency.Given the current tight budgetary constrain' andunmet demand for education in developing countries,the need to control costs and improve effieciency is

35 4

Figure 2: Inputs to and Applications of a Datt,base for Cost Analysis

ObjectivesEducational Planning and

Policy Analysis

Input DataEducational costs

Data ProcessingConstruction of

cost indicato's andcost indices

Educational quantitiesEducational pricesEducational normsSocio-economic data

7Applications

CostingFinancing/feasibility testing

Cost-reduction evaluationCost-benefit evaluation

Cost-effectiveness evaluationDiagnosis of education

very pressing in these countries today. While educa-tional decision makers can seldom influence thesupply and demand factors which d?termine the totalamount of resources devoted to education, they cannevertheless attempt to deal with cost determinantswithin education. These determinants include thetechnology of educational production, compensationfor teachers, the extent of utilization of educationalinputs, as well as drop out and repetition rates.Since the amount of discretionary recurrent expendi-ture in a given year is usually small, educationaldecision makers will do well to be aware of the cost im-plications of past and present decisions, and to adopta long-term perspective in their plan to control costsand improve efficiency.

NOTES

1. Public formal schooling is the most dominantmode of education. Private provision of formalschooling is also significant in some countries, es-pecially at the secondary level (World Bank,1980: 125-126). For some recent cost analyses ofprivate education, see Schiefelbein (1986: 34-42),and Tan (1985a). Non-formal education is not thefocus of this paper. See Ahmed (1975) andHunter (1974) for an economic treatment of non-formal education.

36

2. A major source in Spanish on Latin Americancountries is the RAE collection (Indice de Resu-mes Analiticos Sobre Educacion en AmericaLatina y el Caribe).

3. Some analysts have questioned the appropriate-ness of using the concepts of efficiency and pro-ductivity, and the conventional framework ofeconomic production to understand education(Vaizey et al., 1972). Such an approach can be tootechnocratic and narrow; it also ignores the socialand political dimensions of education (Camayand Levin, 1985).

4. See the other case studies in UNESCO (1972),such as Bennett (1972c), Proust (1972), Ta Ngoc(1972), Arrigazzi and Simone (1972), Auerhanand Solomon (1972), Fachin (1972), andWoodhall (1972).

5. High rates of dropouts and repetitions are wide-spread in developing countries (World Bank,1980: 116-119). There can be a big differencebetween the actual number of years to completean education cycle and the normal length of theeducation cycle. The ratio between the two isreferred to as the inefficiency index. See Haddad(1979) for a revic-v of studies on school wastage.

6. To estimate an ECF, a functional form is used torelate input and output variables. A functionform is said to be flexible if it imposes no a priorirestrictions on the educational production proc-ess, particularly regarding the degree of substi-tution among inputs (elasticity offactor substitu-tion), and how output changes when all inputsincrease by the same multiple (homotheticity);otherwise it is restrictive. The conventionalforms, such as the Cobb-Douglas function and theCES function, are restrictive in that they presup-pose homothetic technology or constant elasticityof factor substitution. A functional form that isflexible is the translog function (see Hendersonand Quandt, 1980).

7. For a review of human-capital studies, see Blaug(1976). Blaug (1985) also provides a brief over-view of alternative perspectives to human capitaltheory. For more rt,..ent developments, see Car-noy and Levin (1985', and Tsang (1987b).

Cost . Analysis for Educational Policymaking:A Review of Cost Studies in Education inDeveloping Countries

References

AID. Manual for Reducing Educational Unit Costsin Latin American Countries. Washington, DC:Agency for International Development, Departmentof State, 1974.

Ahmed, M. The Economics of Non formal Education:Resources, Costs, and Benefits. New York: Praeger,1975.

Aklilu, H., & Heyneman, S. Education for NationalDevelopment: World Bank Activities. Prospects,13(4), 1983.

Alexander, L. & Simmons, J. The Determinants ofSchool Achievement in Developing Countries: TheEducational Production Function. Washington, DC:World Bank, Staff Working Paper 201, 1975.

Alles, J.; Fernando, W.; Wijegunasekera, D.;Gunaratne, D.; de Silva, M.; & Hettiarachchi, C."Ceylon: Costing First- and Second-level GeneralEducation." In Rducational Cost Analysis in Action.Case Studies for Planners. Vol 1. Paris: IIEP,UNESCO, 1972.

ARI. A Study of Utilization, Design and Cost ofSecondary Schools. Colombo, Ceylon: Asian Re-gional Institute for Building Research. (ERIC docu-ment: E4037985), 1969.

Arriagada, A. Determinants of Sixth-grade StudentAchievement in Peru. Washington, D C: EducationDepartment, World Bank, 1983.

Arrigazzi, L., & de Simone, J. "Chile: ImprovingEfficiency in the Utilization of Teachers in TechnicalEducation." In Educational Cost Analysis in Action:Case Studies for Planners. Vol III. Paris: IIEP,UNESCO, 1972.

Auerhan, .T., & Solomon, E. S. "Asia: Cost Analysisin an Asian Model of Educational Development." InEducational Cost Analysis in Action: Case Studies forPlanners. Vol 1. Paris: IIEP, UNESCO, 1972.

Averch, H.; Carroll, S.; Donaldson, T.; Kiesing, H; &Pi' 7.us, J. How Effect:.vo is Schooling? A CriticalReview of Research. Rand Educational Policy Study.Englewood Cliffs, N.J.: Educational TechnologyPublications, 1974.

Becker, G. Human Capital. New York: NationalBureau of Economic Research, 1964.

Bennett, N. "Uganda: The Use of Cost Evaluation inPlanning of Makerere University College." In Edu-cational Cost Analysis in Action: Case Studies forPlanners. Vol III. Paris: IIEP, UNESCO, 1972.

Bennett, N. Problems of Financing the Thai Educa-tional System in the 1960s and 1970s. Paris: TheUnesco Press, 1975.

Blaug, M. "The Empirical S4-atus of Human CapitalTheory: A Slightly Jaundiced Survey." Journal ofEconomic Literature, 14, 1976, pp. 827-55.

Blaug, M. "Where are We Now in the Economics ofEducation?" Economics of Education Review, 4(1),1985, pp. 1;-28.

Bowles, S. & Gintis, H. Schooling in CapitalistAmerica. New York: Basic Books, 1976.

Bowman, M. J. "The Costing of Human ResourceDevelopment." In E. A. Robinson & J. Vaizey (Eds.).Economics of Education. London: Macmillan, 1966,pp. 421-50.

Brookover, W. et al. School Social Systems andStudent Achievement. New York: Praeger, 1979.

Carnoy, M. " Segmented Labor Markets." In Edu-cation, Work, and Employment, 2, Chapters 1-3.Paris: UNESCO, 1980.

Carnoy, M., & Levin, H. "Evaluation of EducationalMedia: Some Issues." Instructional Science, 4(314),1975, pp. 385-406.

37 4

Carnoy, M., & Levin, H. Schooling and Work in theDemocratic State. Stanford, CA: Stanford UniversityPress, 1985.

Chambers, J. Educational Cost Differentials andAllocation of State Aid for Elementary SecondaryEducation. Journal of Human Resources, 13(4),1979, pp. 459-481.

Chesswas, J., & Hallak, J. "Uganda: Behaviour ofNon-Teacher Recurring Expenditures." In Educa-tional Cost Analysis in Act )n: Case Studies for Plan-ners. Vol III. Paris: IMP, UNESCO, 1972.

Chesswas, J. "Tanzania: Factors Influencing Changein Teachers' Basic Salaries." In Educational CostAnalysis in Action: Case Studies for Planners. Vol 1.Paris: IIEP, UNESCO, 1972.

Cohn, E. "Economies of Scale in Iowa High SchoolOperations." Journal of Human Resources, 3(4),1968, pp. 422-34.

Coleman, J.; Campbell, E.; Hobson, C.; MacPar-land, J.; Mood, A.; Weinfall, F.; & York, R. Equalityof Educational Opportunity. Washington, DC: U. S.Government Printing Office, 1966.

Coombs, P., & Hallak, J. Managing EducationalCosts. New York: Oxford University Press, 1972.

Coombs, P., & Hallak, J. Cost Analysis in Education.Baltimore, MD: Johns Hopkins University Press,1987.

Durstine, R., & Hudson, B. "Barbados: MarginalCost for Marginal Decisions-The Case of TeamTeaching." In Educational Cost Analysis in Action:Case Studies for Planners, Vol. II. Paris: IIEP,UNESCO, 1972.

Eicher, J. C. Educational Costing and Financing inDeveloping Countries: Focus on Sub-Saharan Africa.Washington, DC: World Bank, Staff Working PaperNo. 633, 1984.

Eicher, J. C., & ()rival, F. "Cost Analysi of PrimaryEducation by Television in the Ivory Coast." In TheEconomics of New Educational Media. Vol. 2. Paris:UNESCO, 1980a.

38

Eicher, J. C., & Orivel, F. The Allocation ofResourcesto Education Throughout the World. Paris:UNESCO, 1980b.

Eicher, J.C.; D. Hawkridge; E. McAnany; F. Mariet;& F. Orivel, (Eds.). The Economics of New Educa-:ional Media. Vol. 3. Cost and Effectiveness: Over-view and Synthesis. Paris: UNESCO, 1972.

Fachin, R. "Brazil: Costing an Expansion Pro-gramme for Secondary Education in Rio Grande doSul." In Educational Cost Analysis in Action: CaseStudies for Planners. Vol II. Paris: IIEP, UNESCO,1972.

Farrell, J. " Educational Expansion and the Drive forSocial Equality." In P. Altbach; R. Arnove; and G.Kelly, (Eds.) Comparative Education. New York:Macmillan Publishing Co., 1982, pp. 39-53.

Fox, W. F. "Reviewing Economies of Size in Educa-tion." Journal of Educational Finance, 6, 1980, pp.273-296.

Fuller, W. Raising School Quality in DevelopingCountries: What Investments Boost Learning?Washington, DC: World Bank, Education and Train-ing Department, Report No. EDT7, 1985.

Gandhi, V. Some Aspects of Public Education Ex-penditure in Africa. Washington, DC: World Bank,Staff Workitig Paper No. 100, 1971.

Grubb, W. N. "The Convergence of Educational Sys-tems and the Role of Vocationalism." ComparativeEducation Review, 29(4), 1985, pp. 526-548.

Haddad, W. Educational and Economic Effects ofPromotion and Repetition Practices. Washington,DC: World Bank, Staff Working Paper No. 319,1979.

Hallak, J. The Analysis of Educational Costs andExpenditue. Paris: International Institute for Edu-cational Planning, UNESCO, 1969.

Hanushek, E. Conceptual and Empirical Issues inthe Estimation of Educational ProductionFunctions." Journal of Human Resources, 14(3),1979, pp. 351-388.

:A.

Hawkridge, D.; Kinyanjui, P.; Nkinyangi, J.; &Orivel, F. " In-service Teacher Education in Kenya."In H. Perraton (Ed.), Alternative Routes to FormalEducation. Baltimore, MD: Johns Hopkins Univer-sity Press, 1982.

Henderson, J., & Quandt, it Microeconomic Theory.New York; McGraw Hill, 1980.

Heyneman, S. "Differences between Developed andDeveloping Countries: Comments on Simmons andAlexander's 'Determinants of School Achievement.'"Economic Development and Cultural Change, 28(2),1980, pp. 403-06.

Heyneman, S. (1983). "Education Duringa Period ofAusterity: Uganda, 1971-1981." Comparative Edu-cation Review, 27(3), 1983, pp. 403-413.

Heyneman, S.; Farell, J.; & Sepulveda-Stuardo, M."Textbooks and Achievement in Developing Coun-tries: What We Know." Journal of Curriculum Stud-ies, 13(3), 1981, pp. 227-246.

Heyneman, S., & Loxley, W. "The Effect of SchoolQuality on Academic Achievement across 29 High-and Low-income Countries. American Journal of So-ciology, 88(6), 1983, pp. 1162-94.

Heyneman, S.; Jamison, D.; & Montenegro, X."Textbooks in the Philippines: Evaluation of thePedagogical Impact of a Nationwide Investment.Educational Evaluation and Policy Analysis, 6(2),1984, pp. 139-150.

Hind, I . W. "Estimates of Cost Functions for Primary8chools in Rural Areas." Australian Journal of Agri-cultural Economics, 21, 1977, pp. 13-25.

Hinchliffe, K. Issues Related to Higher Education inSub-Saharan Africa. Washington, DC: World Bank,Staff Working Paper No. 780, 1985.

Hunter, J. M. et al. Economics and Non-formalEducation. Program of &tidies in Non-formal Edu-cation. Washington, DC: AID, 1974.

Husan, T.; Saha, L.; & Noonan, R. Teacher Trainingand Student Achievement in Less Developed Coun-tries. Washington, DC: World Bank, Staff WorkingPaper No. 310, 1978.

39

Hutton, J., & Rostron, M. Comparative Study ofSecondary School Building Costs. Paris: UNESCO.(ERIC document: ED 063616), 1971.

Jamison, D. Cost Factors in Planning EducationalTechnology Systems. Paris: IIEP, UNESCO, 1977.

Jamison, D.; Klees, S.; & Wells, S. The Costs ofEducational Media: Guidelines for Planning andEvaluation. Beverly Hills, CA.: Sage, 1978.

Jamison, D., & McAnany, E. Radio for Educationand Development. Beverly }Tills, CA.: Sage, 1978.

Jamison, D.; Searle, B.; Golda, K.; & Heyneman, S."Improving Elementary Mather dies Education inNicaragua: An Expe, 'mental Study of the Impact ofTextbooks and Radio on Achievement." Journal ofEducational Psychology, 73(August), 1981, pp. 556-67.

Jamison, D., & Orivel, F. "Cost-Effectiveness of Dis-tance Teaching for School Equivalency." In H. Perra-ton (Ed.) Alternative Routes to Formal Education.Baltimore, MD: Johns Hopkins University, 1982.

Jenck, C. Inequality: A Reassessment of the Effectof Family and Schooling in America. New York:Basic Books, 1972.

Jimenez, E. "The Structure of Educational Costs:Multiproduct Cost Functions for Primary and Secon-dary Schools in Latin America." Economics of Educa-tion Review. 5(]),1986, pp. 25-39.

Johnstone, J. Indicators of Education Systems.London: Kogan Page/Paris: UNESCO, 1981.

Konny, I. W. "Economies of Scale in School." Eco-nomics of Education Review, 2, 1982, pp. 1-24.

Lau, L. "Educational Production Functions." InNational Academy of Education, Economic Dimen-sions of Education. Washington, DC, 1979, Chapter2.

Lee, K. Further Evi terxe on Economics of Scale inHigher Education. Washington, DC: Education De-partment, World Bank, 1984.

Levin, H. Cost-Effectiveness: A Primer. BeverlyHills, CA: Sage Publications, 1983.

Levin H. "Cost-Effectiveness Analysis." In D.Cordrai, H. Bloom, & R. Light (Eds.), EvaluationPractice in Review. San Francisco; Jossey-Bass,1987, pp. 83-99.

Maynard, J. Some Microeconomics of Higher Educa-ticn: Economies of Scale. Lincoln: University ofNebraska Press, 1971.

McMeekin, R. Educational Planning and Expendi-ture Decisions in Developing Countriee, with a Malay-sian Case Study. New York: Praeger, 1975.

Metcalf, D. Economics of Vocational Training: PastEvidence and Future Evaluations. Washington, DC:World Bank, Staff Working Paper No. 732, 1985.

Min, W. & Tsang, M. "Vocational Education andProductivity: A Case Study of the Beijing GeneralAuto Industry Company." Mimeo, 1988.

Mincer, J. Schooling, Experience, and Earnings.New York: Columbia University Press, 1974.

Mingat, A., & Psacharopoulos, G. Education Codsand Financing in Africa: Some Facts and PossibleLines of Action. Washington, DC: Education andTraining Department, World Bank, Report No.EDT13, 1985.

Neumann, P., & Cunningham, M. Mexico's FreeTextbooks: Nationalism and the Urgency to Educate.Staff Working Paper No. 541. Washington, DC:World Bar 1982.

Oates, J. Educational Indicators: A Guide for Policy-makers, New Brunswick, NJ: Center for PolicyResearch in Education, Rutgers University, 1986.

OFISEL, Ltd. "Cost of Rural Primary Education:Case Studies Methodology." Bogota, Colombia:Mimeo, 1982.

Paulsen, A. The Unit Costs of Education. Bogata:OFISEL. (RAE 2.700), 1981.

Perraton, H. (Ed.) Alternative Routes to FormalEducation. Baltimore, MD: Johns Hopkins Univer-sity Press, 1982.

40

Proust, J. "Morocco: Costs of Public Secondary Edu-cation Analysis of the Results of a GovernmentalSurvey." In Educational Cost Analysis in Action:Case Studies for Planners. Vol III. Paris: IIEP,UNESCO, 1972.

Psacharopoulos, G. Returns to Education: An Inter-national Comparison. San Francisco: Jossey-Bass,1973.

Psacharopoulos, G. "Returns to Education: An Up-dated International Comparison." Comparative Edu-cation, 17(3), 1981, pp. 321-341.

Psacharopoulos, G. "The Economics of Higher Edu-cation in Developing Countries." Comparative Edu-cation Review, 26(2), 1982, pp. 139-59.

Psacharopoulos, G. "Returns to EducatiG..: A Fur-ther International Update and Implications." Jour-nal of Human Resources, 20(4), 1985, pp. 583-604.

Psacharopoulos, G., & Loxley, W. Diversified Secon-dary Educaticn and Development: Evidence fromColombia and Tanzania. Baltimore, MD: JohnsHopkins University Press, 1985.

Psacharopoulos, G., & Woodhall, M. Education forDevelopment: An Analysis of Inves&ment Choices,New York: Oxford University Press, 1985.

Psacharopoulos, G.; Tan, J.; & Jimenez, E. TheFinancing of Education in Latin America: Issues andLines of Action. Washington, DC: Education andTraining Department, World Rank, Report No.EDT32, 1986.

Purkey, S., & Smith, M. "Effective Schools: A Re-view." Elementary School Journal, 83,1983, pp. 427-4 52.

Radi, M. Study of Operating Costs of SecondarySchools in Morocco. Paris: IIEP, UNESCO, 1982.

Reiff, H. W. "Thailand: Educational Cost Analysis."In Educational Cost Analysis in Action: Case Studiesfor Planners. Vol 1. Paris: UNESCO, 1972a.

Reiff, H. W. "Thailand: 2he Use of Cost Analysis inEstimating the Total Cost of an Educational Plan andTesting its Feasibility." In Educational Cost Analy-

sis in Action: Case Studies for Planners. Vol . Paris:UNESCO, 1972b.

Riew, 0. "Economies of Scale in High SchoolOperation." Review of Economics and Statistics,48(3), 1966, pp. 280-87.

Schiefelbein, E. Education Costs and FinancingPolicies in Latin America: A Review of AvailableResearch. Washington, DC: Education and TrainingDepartment, World Bank, 1986.

Schultz, T. "Investment in Human Capital." Ameri-can Economic Review, 51: 1961, pp. 1-17.

Schwille, J.; Beeftu, A.; Navarro, R.; Prouty, R.;Raudenbush, S.; Schmidt, W.; Tsang, M.; & Wheeler,C. Recognizing, Fostering and Modeling the Effec-tiveness of Schools as Organizations in Third WorldCountries. Cambridge, MA: Project BRIDGES,Harvard University & USAID, 1986.

Shephard, R. Theory of Cost and Production Func-tions. Princeton, NJ: Princeton University Press,1970.

Spence, M. "Job Market Signaling." Quarterly Jour-nal of Economics, 87, 1973, pp. 355-374.

Ta Ngoc, C. "Ivory Coast: The Cost of Introducing aReform in Primary Education." In Educational CostAnalysis in Action: Case Studies for Planners. Vol II.Paris: IIEP, UNESCO, 1972.

Ta Ngoc, C.; Hallak, J.; & Coombs, P. "!,:adagascar:The Role of Cost Analysis in the Introduction andImplementation of the 1962 Reform of PrimaryEducation." In Educational Cost Analysts in Action:Case Studies for Planners. Vol II. Paris: IIEP,UNESCO, 1972.

Tan, J. "Private Enrollments and Expenditure onEducation: Some Macrotrends." International Re-view of Education, 31(May), 1985a, pp.103-117.

Tan, J. "The Private Direct Cost of SecondarySchools in Tanzania." International Journal of Edu-cational Development, 5(1): 1985b, pp. 1-10.

Taylor, D.C. "The Cost-Effectiveness of Tear:herUpgrading by Distance Teaching in Southern

41

Africa." International Journal of Educational Devel-opment, 3(1), 1983, pp. 19-31.

Thurow, L. " Education and Economic Equality." ThePublic Interest, 28, 1972, pp. 66-81.

Tibi, C. "The Determinants of Educational Coat."IIEP Newsletter, 4(2), 1986, pp. 1-2.

Tibi, C., & Wheeler, A.C.R., "Studies of chool Costs."Report of a Workshop Held by IIEP (13-16 October1980), IIEP, November 1980, Paris, (RAE 3.241).

Tilak, J. Analysis of Costs of Education in India.Occasional Paper No. 10. (ERIC document: ED261946), 1985.

Tilak, J., & Varghese, N. V. Resources for Educationin India. (ERIC document: ED 259963), 1983.

Tsang, M. A Cost-of-Education Data System forNational Educational Planning and Policy Analysis.Report prepared for Harvard University for ProjectBRIDGES, 1987a.

Tsang, M. "The Impe'...t of Underutilization of Edu-cation on Productivity: A Case Study of the U.S. BellCompanies." Economics of Education Review, 6(2),1987b.

Tsang, M., & Levin, H. "The Economics ofOvereducation." Economics of Education Review,4(2), 1985, pp. 93-104.

UNESCO. Educational Cost Analysis in Action: CaseStudies for Planners, Volumes 1-3. Paris, 1972.

UNESCO. The Economics of New Educational Me-dia. Vol 1. Present Status of Research and Trends.Paris, 1977.

UNESCO. The Economics of New Educational Me-dia. Vol. 2. Cost and Effectiveness. Paris, 1980.

UNESCO. Public Expenditure on Education in theWorld, Regional and Country Trends, 1970-1979.Paris: Office of Statistics, 1982.

UNESCO. Policy, Planning and Management inTechnical and Vocational Education, Paris: Author,1984.

Vaizey, J. The Costs of Education. London: GeorgeAllen & Unwin, Ltd., 1958.

Vaizey, J., & Chesswas, J. The Costing of Educa-tional Plans. Paris: International Institute for Edu-cational Planning, UNESCO, 1967.

Vaizey, J.; Norris, K.; & Sheehand, J. The PoliticalEconomy of Education. London: Duckworth, 1972.

Wagner, L. The Economics of Educational Media.London: Macmillan, 1982.

Wasserman, W. Education Price and QuantityIndexes. Syracuse, NY: Syracuse University Press,1963.

Weiler, H. "Towards a Political Economy of Educa-tional Planning." Prospects, 8(3), 1978, pp. 247-267.

Windham, 1). Internal Efficiency and the AfricanSchool. Washington, DC: Education and TrainingDepartment, Worij Bank, Report No. EDT47, 1986.

Wolff, L. Controlling the Costs of Education in East-ern Africa: A Review of Data, Issues, and Policies.Washington, DC: World Bank, Staff Working PaperNo. 702, 1985.

Woodhall, M. Cost-Benefit Analysis in EducationalPlanning. Paris: International Institute for Educa-tional Planning, UNESCO, 1967.

42

Woodhall, M. "India: The Use of Cost-Benefit Analy-sis as a Guide to Resource Allocation in Education."In Educational Cost Analysis in Action: Case Studiesfor Planners. Vol II. Paris: IIEP, UNESCO, 1972.

World Bank. Education. Washington, DC: WorldBank, 1980.

World Bank. China: Socialist Economic Develop-ment. Vol. 3. The Social Sectors. Washington, DC:1983.

World Bank. Financing Education in DevelopingCountries: An Exploration of Policy Options. Wash-ington, DC: 1985.

World Bank. China: Management and '"finance ofHigher Education. Washing ton, DC: 19E.,

Zymelman, M. Patterns of Educational Expendi-tures. Washington, DC: World Bank, Staff WorkingPaper no. 246, 1976a.

Zymelman, M. Economic Evaluation of VocationalTraining Programs. Baltimore: Johns HopkinsUniversity Press, 1976b.

Zymelman, M. Educational Expenditures in the1970s. Washington, DC: Education Department,World Bank, 1982.

BRIDGES Publications Staff:

Billie Jo Joy, Publications CoordinatorPamela Summa, Editorial AssistantKin Bing Wu, Editorial AssistantBrenda Salyer, Graphic Designer

For more information about this or other BRIDGES documents, please contact:

BRIDGES Publications442 Gutman LibraryHarvard UniversityCambridge, MA 02138USA

(617) 495-9720

or

4

USAIPOffice of EducationBureau for Science and TechnologyDepartment of StateWashington, D.C. 20523 USA(703) 235-8980

. t . '- ,,-;,,,ic.:?..t.-,..40 VA

tu.c.,* %'..20, ",...---

Vfl:C1 i 1 , ''.:'_"' - .' -...T ,,. , 6, A..e.P...., 'rti-- \,,.. ,I,",9,c ',., 1 , C'..-- ,-

v ;:t7;t5v., P4,19,4%.,.2.t' ,,,i,"- , '`,,, ,::...1-.

). , , , .

e4 1. >''.' .

A,,i,,:,:

,,, 1; "V. ,',4,-N.

? .

470:#;:frA.

A program of:USAID. Bureau for Science & Technology

xy

Directed by:The Harvard University Institute for International Developmentand the Harvard Graduate School of Education

In Cooperation with:Academy for Educational Planning and Management. PakistanCentre de Perfectlonnement et de Formation en Cours d'Emploi, BurundiInstitute for international ResearchMichigan State UniversityThe Open University, IndonesiaResearch Triangle InstituteTexas Southern UniversityNational Education Commission of ThailandNational Institute of Education, Sri LankaUSAID Missions and Ministries of Education in Burundi, Egypt, Indonesia, North Yemen, Pakistan, Sri Lanka, and Thailand