raising school quality in developing counties: investments ...€¦ · occurred on the relative...

THE WORLD BANK

DiscLssion Paper

EDUCATION AND TRAINING SERIES

Rrpopt No. EDT7

Raising School Quality inDeveloping Counties:

2 What Investments Boost Learning?

Bruce FullerSeptember 1985

Education and Training Department Operations Policy Staif

The views presented here are those of the .2uthor. and they should not be internreted as retlecting those of the World Bank.

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Discussion Paper

Education and Training Series

Report No. EDT7

RAISING SCHOOL QUALITY IN DEVELOPING COUNTRIES:WHAT INVESTMENTS BOOST LEARNING?

Bruce FullerEducation and Training Department

September 1985

The World Bank does not accept responsibility for the views expressedherein which are those of the author and should not be attributed to theWorld Bank or to its affiliated organizations. The findings, inter-pretations, and conclusions are the results of research or analysissupported by the Bank; they do not necessarily represent official policy ofthe Bank. The designations employed, the presentation of material, and anymaps used in this document are solely for the convenience of the reader anddo not imply the expression of any opinion whatsoever on the part of theWorld Bank or its affiliates concerning the legal status of any country,territory, city, area, or of its authorities, or concerning thedelimitation of its boundaries, or national affiliation.

Abstract

Low levels of student achievement and school quality persist withindeveloping countries. This paper reviews evidence on the importance ofschool quality - in raising literacy and influencing economic development.In addition, much has been learned about how to raise the quality andefficiency of schools. Major lessons are discussed within four areas.

School quality and development. First, evidence is now clear that low.* school quality largely accounts for low levels of literacy and achievement

among Third World children. After controlling for the effects of thechild's preschool background and economic context, school quality makes asubstantial difference in developing countries. Yet persistently lowlevels of school quality are constraining economic returns to educationalinvestments. In addition, initial evidence shows that little progress hasoccurred in improving school quality among the poorest developing nationssince 1970. Just as the potential force of school quality has becomeclearer, erosion of already limited levels of quality is accelerating.

Defining school quality. Second, a clearer definition of "schoolquality" would help refine investment strategies of governments anddevelopment agencies. School quality is often indicated by levels ofstudent achievement (output) or by school characteristics which areunrelated to pupil performance. Instead, investments should beconsistently targeted (a) on school characteristics which influenceachievement, and (b) on efforts which encourage more efficient managementand organization of material inputs by local school staff.

Isproving school quality. Third, much has been learned about whatelements of schools consistently boost literacy levels. The major lessonof 72 empirical studies are reviewed here. Those inputs most stronglylinked to the instructional process - textbooks, writing materials, andteacher quality - consistently influence achievement. Cost savings can berealized by reducing investment in other elements which are not related toachievement, including small class size, classroom laboratories, the papercredentials of teachers, and teacher salary levels. Very little researchhas occurred regarding the influence of teaching practices and classroomorganization on achievement within developing countries. These managementpractices might more effectively raise literacy and academic achievementthan via new investments in material inputs.

Boostiug school efficiency. Fourth, only a modicum of thinking hasoccurred on the relative cost-effectiveness (or internal efficiency) ofalternative school inputs and management practices in raising achievement.This paper does identify elements of schools which are not related tohigher student achievement, for which investment can be reduced withoutdetrimental achievement effects. But among those elements whichconsistently do influence performance, little is known about theirmagnitude of effect and their relative costs. Methods for studying theefficiency of various inputs are reviewed here.

Contents

Abstract

Summary Page 3

Chapter 1 The Issue: A Costly Neglect of School Quality? Page 7IntroductionIssue 1: Low Student Achievement,

Low School QualityIssue 2: Constraining the Economic Benefits

of School InvestmentIssue 3: Policy Tradeoff Between School Expansion

and QualityIssue 4: Eroding School QualitySummary

Chapter 2 Defining School Quality Page 22IntroductionDefining School QualityInternal Effectiveness of School QualitySummary

Chapter 3 What School Factors Boost Achievement? Page 29IntroductionLessons from Industrialized Countries?What School Characteristics Influence Achievement?

Overall School ExpendituresSpecific Material InputsTeacher QualityTeaching Practices and Classroom OrganizationSchool Management and Structure

A Narrow View of School Quality?Refining Production-FunctionsAbandoning Material Production-Functions

Summary

Chapter 4 What School Factors Boost School Efficiency? Page 66IntroductionSources of Internal EfficiencyMethod 1: General Comparison of Cost and BenefitMethod 2: Analysis of Discrete Marginal Cost and ProductMethod 3: Estimating Rates of ReturnSummary

Annex / School Quality Research in Industrialized Countries Page 75

Notes Page 85

References Page 87

Summary

"A year of schooling is not just a year of schooling.It can be anything from a delight to a torment dependingon the imagination and devotion, as well as the cognitiveskills, of the teacher. It can be anything from anexperience of growth and mastery to a stunting andconfining time [Note 1.1l]."

The efficacy of many Third World schools - even in providing basicliteracy skills - remains limited. Students in developing countries learnmuch less than pupils within industrialized nations, even when they haveattended school for the same number of years. Disappointing levels ofstudent performance, in large part, can be attributed to low schoolquality. And signs of dismal school quality are common in most Third Wiorldnations: a scarcity of reading materials, poorly trained teachers,classrooms with no tables on which to write, instructional programs whichfail to enliven children's curiousity.

This paper reviews major findings from the past 15 years of researchon school quality in developing countries. Two specific tasks areundertaken. The problem of low school quality is described in detail.Then, past empirical research is reviewed on which specific elements ofschools are most consistently related to student achievement. Lessons fromthis work could-help refine investment strategy aimed at improving theeffectiveness and efficiency of schools.

3nv serious is the school quality problem? Strong evidence now showsthat school quality contributes to higher student achievement in developingcountries, controlling for effects of the child's preschool background andeconomic setting. Earlier work demonstrating that material resources makelittle difference in shaping pupil performance within industrializednations does not hold within most Third World settings. Unfortunatelyschool quality remains low, even at the primary level, within mostdeveloping countries.

Early evidence suggests that low school quality is seriouslyconstraining the economic return to school investments - both forindividuals and for national economies. Initial analyses also suggest thata tradeoff faces governments and development agencies between furtherexpanding school enrollments versus improving school quality. Schoolquality (indicated by per pupil expenditures) has suffered most in thosedeveloping countries which have expanded enrollments rapidly in the pastdecade. No progress in improving school quality is evident in-the poorestdeveloping countries since 1970. The gap in school quality between low-income and middle-income Third World countries also has widened during thisperiod. Middle-income countries have made significant progress in raisingper pupil expenditures and in supporting more teachers relative toenrollments.

3

Now can school quality be more clearly defined? Vague definitions of"school quality" are contributing to imprecise strategies for improvingstudent achievement. Governments and development agencies often definequality in terms of the level of academic performance or retention inschool. But the quality of the school's impact cannot be inferredexclusively from the quality of student performance (i.e. output). Thisfails to account for factors outside the school which independentlyinfluence performance (including family income, early childhoodsocialization, and ongoing demand for the child's labor). Alternatively,quality is often defined in terms of the aggregate level of all inputs;this fails to discriminate between those school characteristics which leadto higher achievement from those that do not.

School quality is better defined in relation to the amount of learningimparted by the school after accounting for the effects factors whichoperate external to the school, such as the child's economic context andpreschool background. Two sets of school characteristics are involved inthe "value-added" to the child's academic skills or the increment oflearning directly by the school: (1) The concentration of material inputsper pupil allocated to the school (e.g., textbooks, desks, pencils andpaper), and (2) the social organization or management of these physicalfactors (e.g., the organization of instructional lessons, the headmaster'smanagement competence, or the character of interaction between teacher andstudent). The second set of factors largely determines the efficiency withwhich material inputs are organized to influence achievement levels. Thelevel of material resources available per student connotes a certain levelof quality. However, schools with scarce resources vary dramatically intheir level of efficiency and organizational proficiency in boostinglearning. Similarly, schools with abundant material resouces may utilizematerial resources inefficiently, failing to raise student achievementsignificantly.

What school quality investments raise achievement? Investments arecurrently made in a variety of school inputs and management practices,independent of each factor's actual relationship to student achievement.Yet a considerable body of empirical research has developed in the ThirdWorld over the past 15 years. One must first distinguish between thoseelements of school quality which are related to higher achievement fromthose which are not. This paper revievs findings of 72 empirical studiescompleted over the past decade and a half. In general, those schoolelements closely linked to the instructional process - books, libraries,writing materials, and teacher quality - most consistently influencestudent achievement. Investment strategies could also benefit fromknowledge of those school quality elements which hold no consistentinfluence on achievement (e.g., small class size, use of laboratories, orindividual teacher salary rates), identifying where substantial cost-savings ,can be generated.

Early research suggests that teaching practices also influence pupilperformance in the Third World. Despite extensive work in this area withinindustrialized nations, little research has occurred within developingcountries. Attempts to identify those material factors related toachievement have eclipsed efforts to understand how the local managementand social organization of material inputs increase pupil performance.

4

'What school quality investments increase the school's efficiency? Verylittle is known about the relative costs of alternative elements of schoolquality -- even within that set of school characteristics whichconsistently raise achievement. This area of ignorance limits our abilityto answer the question, What elements of school quality are the most cost-effective in raising student achievement? Here the concern is not onlywhether a specific ingredient of quality is effective in raisingachievement. But improving schools' efficiency involves two additionalquestions. First, what is the magaitude of the school factor's achievementeffect? Second, what is the cost of this strategy to improve qualityrelative to another factors? For instance, inservice training of teachersmay effectively boost achievement. But investing in additional textbooksor instructional materials may yield a larger achievement effect at anequal or lower cost, compared to the magnitude of the effect or cost ofteacher training. Methods for determining the relative efficiency (inraising achievement) of different elements of quality are reviewed here.

The search for more efficient investment strategies should perhapsstop at middle-range decision rules, focusing on those elements of qualitywhich are related to achievement and specifying the conditions under whichthese positive effects are likely to occur. Studies of cost-effectivenessor efficiency of different school elements should not assume an unrealisticlevel of precision in policy and investment implementation. Nor should thedesire to determine precise costs of material elements of school qualitydistract attention from a clearer understanding of teaching practices andschool management skills - social factors for which determination ofmarginal costs is difficult. Changing teaching behaviors is conceptuallysimply and potentially inexpensive; yet determining bow to improveclassroom practices is problematic. Governments and development agenciescould provide clearer information to local schools on how to improveteaching practices and how to create more motivating classrooms.

5

Acknovledgements

I want to thank Gary Theisen and Joe Farrell for their detailedcomments on an earlier draft of this paper. Stephen Heyneman and GeorgePsacharopoulos got me into this intellectual morass. Conversations withBarbara Searle, Peter Moock, Robin Horn, Birger Fredriksen, Nat Colletta,Emmanuel Jimenez, Bernardo Kugler, and Jee Peng Tan have been invaluable.

Bruce Fuller

6

Chapter 1The Issue: A Costly Neglect of School Quality?

Introduction

Education's contribution to national development has been widelyrecognized in the Third World since World War II. Government policymakersand development groups have emphasized the importance of boostingthe quantity of education. Expanding schools yields clear signals ofsocial and economic progress: Building more school structures, hiring moreteachers, and enrolling more children offer concrete signs of change. Yetthe quality of teaching and learning that occurs within this burgeoningnumber of schools is highly variable and very low in many developingcountries. Several recent events prompt rising concern over inadequateschool quality.

First, the issue of quality has been eclipsed by the Third World'spost-war desire to simply build national systems of education. Untilprimary enrollment rates reached reasonably high levels, improvement ofquality was not viewed as the immediate problem. Nations which haveexpanded primary schooling are now faced with a tradeoff between improvingquality of primary instruction or further expanding secondary and tertiaryschools.

Second, the recent decline in the level of investment capitalavailable to the Third World prompts an overdue question: How can schoolsmore efficiently increase literacy and academic achievement? When schoolsystems are rapidly expanding, policymakers and local school staff mayspend less time thinking about which elements of schools effectively booststudent achievement. But the slowing of growth in resources and enrollmentrates provides the opportunity to focus attention on improving theeffectiveness of schools. This is not to say that improving school qualityis less costly than continued school expansion. Recent research does,however, disentangle those elements of school organizations which areconsistently related to higher student achievement from those factors whichare not. Shifting resources from the latter to the former elements ofschooling could raise efficiency at no additional net cost to the educationsector.

Third, our knowledge about student achievement in the Third World andthe antecedent influence of school quality has grown substantially inrecent years. As attention of policymakers and researchers turns fromschool expansion to school improvement, the problem of inadequate qualitybecomes more stark. Four aspects of the school quality problem have beenilluminated by recent research:

o Literacy and academic achievement of children in developingcountries remains far below the performance of students withinindustrialized countries at the same grade level.

7

o Low levels of student achievement and poor school quality areconstraining the influence of school investments on subsequent economicdevelopment. Initial evidence demonstrates higher economic returns toschool investments when school quality and/or student achievement are atrelatively high levels in developing countries. These economic effects ofschool quality operate independently of school quantity (that is, eitherthe individual's length of schooling or the nation's enrollment rate).These potential economic benefits, however, are seriously constrained inmany settings due to low quality.

o The absence of a clear investment priority between improving thequality of primary schooling versus expansion of secondary and tertiaryschooling may be very costly. The evidence is clear that rates of returnto primary schooling are higher than returns to secondary and highereducation in most Third World countries. Improving the quality of primaryschools may spur economic development more strongly than expansion ofhigher levels of schooling.

o Initial evidence reveals that school quality has declined in thepoorest Third World nations since 1970. In addition, the gap in relativelevels of school quality between developing and industrialized countrieshas widened over the past 15 years.

This chapter speaks in turn to each of these aspects of the school qualityproblem.

Issue 1: Low Student Achievement, Lov School Quality

Only ten percent of Third World students, age 14, are as literate intheir native language compared to youth from industrialized nations. Thiswas a major finding of one cross-national study of school achievement.Mean exam scores in reading, math, and science were 50 percent lower withindeveloping countries compared to industralized nations. In reading, forinstance, the mean comprehension score (in native language) equalled 26 forall countries surveyed. The mean reading score, however, for Chileequalled only 14; for Iran, 8; and in India, 5. All youth given the examwere age 14 and all were enrolled in school at the time. This researchincluded only four developing countries (Chile, India, Thailand, and Iran).But the findings may accurately represent low levels of student achievementevident in most other Third World nations. For example, in Sierra Leoneonly 15 percent of all Grade 2 students are achieving in math at theexpected grade level. And by international standards, achievement may behigher in the four developing countries included in this cross-nationalstudy relative to lower-income Third World states.

Alternative measures of student performance yield equally discouragingsigns regarding student performance in the Third World. In Bangladesh,only one-fourth of all children complete the fourth grade. In Peru, 37percent of all children repeat the first grade. Even among the four lesspoor developing countries mentioned above - Chile, Thailand, Iran, andIndia -- the number of Grade 4 children was one-half the size of Grade 1enrollments [Note 1.2].

Variation in student performance is very wide within developing

8

countries. In Egypt, rural primary school students perform one grade levelbelow urban children. Reading scores among low achieving schools inBotswana average one-third the achievement level observed in highperforming schools. Similarly, in Kenya, rural school performance on thenational exam ranges as low as one-third below the mean performance ofurban schools (between-school mean scores). Urban Brazilian youth, age 12-15, on average have completed four years of schooling; yet rural youth havecompleted just one and one-half years of schooling. Each day, 15 percentof all urban Peruvian students and 30 percent of rural children reportedlyare absent from school [1.3].

The cost implications of frequent repetition of a grade-level and highdropout rates are enormous. An example of this inefficiency is clearlyapparent in Haiti. Due to a high repetition rate and low persistence amongrural Haitian children, over 20 student-years of school attendance areobserved. (Five pupils each attending school for two years equals 10student-years of attendance.] In Haiti's urban private schools, just sevenstudent-years of attendance occur per graduate (for a six year course ofstudy) [1.41.

Table 1 reports on the percent of pupils completing primary school fordeveloping and industrialized nations. These data do not capture variationin student persistence within nations. But inequities by levels ofnational wealth and region are important to note. For instance, among the23 low-income developing countries reporting data (GNP per capita less thanUS$ 405) only 60 percent of all children complete primary school. Thiscompletion rate rises to 75 percent for middle-income developing nations(GNP per capita of $840 for lower middle-income and $2,490 for uppermiddle-income nations). Virtually all children of industrialized nations(GNP per capita equalling $11,070) finish primary school. Studentperformance in sub-Sahara Africa remains acutely low on a variety ofbenchmarks. For example, adult literacy is below 20 percent in Chad andMali.

Table 1Student Performance: Primary School Completion Rates

Z Pupils Completing Number of CountriesPrimary School Reporting

Development StatusLow-income countries 60% 23Middle-income countries 75% 49Industrialized countries 93% 6

Developing CountriesBy RegionSub-Sahara Africa 65% 30East Asia 84% 11South Asia 66% 4Latin America 65% 19

Source: Unesco (1983a)

9

Documenting low levels of student achievement in developing countriesis straightforward. The task of explaining the antecedent causes of lowperformance is more difficult. Looking across nations, relative levels ofwealth are clearly related to students' average achievement levels [1.5].A nation's wealth boosts achievement throught at least two mechanisms.First, infants and young children in more affluent countries experience ahigher quality of life, especially in terms of nutrition, physical health,and cognitive stimulation. Therefore, children in wealthier nations enterthe primary school with stronger physical and intellectual competencies.Second, the demand for children's labor is much lower within moreindustrialized nations. These societies, in fact, often have enactedstrong institutional rules which constrain children's entry to jobs. Incontrast, within developing countries many children are required to work inagriculture or urban jobs for a good part of each day and especially duringplanting and harvest seasons.

Unequal levels of preschool development and labor demand for childrenalso exist within developing nations. Several studies from the Third Worldfind that the pupil's social class background does make a difference inshaping school achievement. However, the influence of the child's socialclass appears to be less overall and less consistent in developingcountries, compared to industrialized nations [1.61.

Quality of the child's school also makes a substantial difference indetermining achievement in the Third World. A decade ago the dominantspeculation was that, like industrialized countries, students' social classand surrounding community wealth eclipsed any potential influence of theschool itself. This is indeed the case for more industrialized nations, atleast to the extent that the correct school characteristics have beenstudied. When one looked across nations, the society's wealth was (and.remains) a powerful determinant of a nation's average level of achievement.But research over the past decade consistently shows that school qualitystrongly shapes student achievement within developing countries. Inaddition, the influence of school quality on student achievement isstrongest among poorer countries and among lower income students withindeveloping nations. And at times, school quality actually influencesschool achievement more strongly than does social class. Interestinglyschool quality is a stronger (and family background a weaker) determinantof achievement in curriculum areas which are not linked to indigenouslanguage or knowledge, such as achievement in mathematics or science (1.71.

Chapter 2 reviews in depth the evidence on whether aggregate levels ofschool quality influence achievement and which elements of quality mostefficaciously raise pupil performance. Part of the problem is thatresearch which demonstrates the importance of school quality has notpermeated policy circles within many countries and development groups.

Evidence which links school quality, to student achievement isencouraging. This research emphasizes that the school institution caninfluence literacy levels and human capital formation somewhat independentof material economic conditions - providing an important policy andinvestmient lever. Unfortunately school quality is very low in manydeveloping countries. And signs of low school quality are abundant. InMalawi, only one in eight children attending school has a seat. Just one

10

in 88 students is provided a desk on which to write. The ratio of pupilsto teachers is as high as 55/1 in Togo and 51/1 in Cameroon. Prior to alarge development program, schools in the Philippines had one textbook forevery 10 pupils. Shortages of books and basic reading materials arecommon. Sierra Leone spends US$ 0.68 per student on instructionalmaterials; Bolivia allocates $ 0.80 for textbooks and classroom supplies[1.8].

Large inequities in school quality are apparent across nations,particularly in terms of basic material resources available per student.Table 2 (Column A) reports expenditures per pupil for developing andindustrialized countries and by region for 1980. Among nations reportingdata, low-income nations spend just US$ 59 per student versus $195 formiddle-income developing countries. Industrialized nations spend $2,297per pupil. The picture is quite similar for per pupil resources allocatedfor (non-salary) instructional materials (Column B). Low-income countriesspend on average $1.69 per pupil versus $92.32 among industrializednations. These first two indicators of school quality are empiricallyrelated to student achievement (see Chapter 2). Expenditure data have notbeen adjusted for cross-national differences in purchasing power given thelack of indices available for individual developing countries. However,based on existing purchasing power information, the unequal quality levelsseen here would be compressed modestly if complete purchasing-poweradjustments could be made E1.9].

11

Table 2School ulity Idicatws for Primay School Level, 1980

(Ils r of Countrien Reporting)

(a) (b) (c)Total Recurreat Expunditurn Expenditurn cm Instructional Pupil/Teacher Ratio

Nation Per Pupil aterials Per Pupilgroups (1990 us S) (1980 us $S

Dvelopunt Statau

L-incm Countris S 59 $ 1.69 44(21) (34) (34)

Niddle-incu.e coutries S 195 $ 6.14 32(42) (32) (59)

Industrialized cuntrin $2,297 $92.32 l8(19) (16) IIBJ

Developing Countries Onlyby Region

Sub-Sabara Africa $ 92 $ 2.49 43(24) (16) (36)

liddl East &Nwth Africa S 221 3.28 30( 5) (4) (10)

East Asia & Pacific $ 210 $ 2.06 31( 6) ( 6) (12)

South Asia $ 17 $ 1.26 40( 6 (4) (q)

Latin Amrica & Caribbem S 209 8.9" 30(19) (15) (24)

Sources: Unesco (1"1, 1983a, 1983b)

Variation in school quality is also apparent across geographicalregions among developing countries. The six south Asian countriesreporting data for 1980 spend a total of just $17 per student and $1.26 perpupil for instructional materials. On the other hand, Latin Americannations allocate $209 per student and $8.99 on instructional materials.

The ratio of pupils per teacher represents a third measure of quality,representing the concentration of one material input per student. The

12

unequal distribution of material inputs is also observed here. Thepupil/teacher ratio is 44/1 in low-income nations, versus 32/1 in middle-income and 18/1 within industrialized states (Column C). Geographicalvariation is also seen here, with the highest pupil/teacher ratios reportedin sub-Sahara Africa (43/1) and South Asia (40/1).

As seen above in achievement patterns, inequities in school qualityalso are apparent within developing nations. In Honduras, for instance,about one-third of all rural teachers work in a one-room schoolhouse with60 children. The pupil/teacher ratio in Tanzania ranges from 60/1 to 31/iamong rural and urban regions, respectively. In northeastern Brazil, one-third of all teachers have four years of schooling or less, and 75 percentdo not meet minimum national qualifications. In El Salvador, 55 percent ofall rural students have no books available in their schools.

Ideally, better cross-national measures of school quality would beavailable. Expenditures per pupil are general indicators which fail tocapture variation in how resources are spents, particularly as distributedamong administrative costs, teacher salaries, and instructional materials.This indicator, however, is empirically related to student achievement in amajority of studies completed (Chapter 3). Yet the pupil/teacher ratioreported by Unesco does not take into account multiple shifts of classes.It simply reports the national ratio of enrollments to teachers. Thebenchmark may reflect average class size; but class size is notalways related to achievement levels. More work is needed in developingcross-national indicators of school quality.

Issue 2: Constraining the Economic Returnto School Investments

The economic cost of low school quality is sizable. Research on thematerial effects of school quality is quite young. Yet initial evidencesuggests that variability in quality is related to economic gains both interms of individual income and national economic development. Importantly,the economic effects of investing in school quality occur independently ofschool quantity. Not surprisingly, how much the child learns appears tosignificantly influence future earnings, beyond how many years the childattends school. Therefore, unless a priority is placed on improvingquality, the economic benefits of increasing school investments will beconstrained.

Research on the economic benefits of school quality has followed twostreams. First, research has looked at the influence of school quality onthe individual's subsequent success in the labor force. That is, childrenattend schools of varying quality and learn differing amounts of knowledgeor skills. This variation among individuals may influence the level of jobattained in the labor force and subsequent personal earnings. For example,one longitudinal study followed 1,205 Chilean youth through school and intotheir first jobs. All youth'surveyed had completed primary school in 1970.Their family background and the quality of school attended each helped toexplain the level of their first occupation. School quality was measuredthrough several indicators, including teachers' educational level, textbookavailability, and the quality of school facilities [1.10].

13

Several studies within industrialized nations find that school quality(measured in terms of expendituTes per pupil) significantly affectsearnings of school graduates. Early work on estimating economic returns toeducation used global measures of school quantity, such as the length ofschool attendance. Similar to the findings for academic achievement,family background has been found to dwarf any effect of the youth's amountof schooling on subsequent earnings. Yet this more recent work reveallarger income effects for schooling when discriminating between varyinglevels of quality within industrialized nations.

School quality may be more efficacious in shaping individual earningswithin developing countries. For instance, a recent study of 6,171Brazilian males, age 15-35, found that the quality of school attended (interms of mean educational level of teachers) influenced subsequent earningsmore strongly than the length of schooling completed. Quality along thisbenchmark was low and highly variable. Teachers on average had attainedless than nine years of schooling. Variation in their level of schoolingwas quite wide; two-thirds of the teachers had received between five andtwelve years of formal education. The Brazil study suggests that theeconomic return to school quantity may be commonly over-stated unless oneaccounts for differences in school quality across regions and individualschools. For example, the social rate of return to greater investment inschool quality exceeded the yield associated with increasing the number ofyears of school attendance (even after accounting for the additional costof educating teachers for a longer period of time). In addition, the rateof return on improving quality also exceeded the ten percent rate-of-returnstandard commonly used in judging investment strategies [1.11].

Whether governments can actually reap such levels of return is anotherquestion. School quality differences are often bounded by economic andinstitutional structures. For example, many developing nations have notaddressed inequities in school quality levels found between urban and ruralareas. One recent study found that children of white-collar workers gainnearly six times as much benefit from school expenditures (includingsupport of school quantity and quality) as do children of farmers. In somedeveloping countries, such as within francophone Africa, the distributionof educational benefits is ten to one. Additional benefits in schoolquality may further benefit only urban students unless institutionallyentrenched inequities are reduced. Also the strong earnings benefitobserved in Brazil may only occur when taking into account variation inincome across urban and rural areas. But these encouraging rate-of-returnfindings assume that no geographical, informational, or institutionalbarriers hamper movement among sectors of the labor force, acrossgeographical regions, and among job sectors within urban or rural regionst1.12].

These weaknesses are avoided in a recent study of income effects ofactual school achievement among 205 workers in Kenya and 179 workers inTanzania. All individuals were employed within the modern economic sectorby formal firms, avoiding the criticism expressed above that institutionalbarriers may mitigate against income effects of school quality.Importantly, this study disentangled the length of schooling attained byworkers from how much they learned while in school. Achievement wasmeasured on standard literacy and numeracy exams. This is not a directmeasure of school quality. Some degree of literacy may also be acquired

14

outside of formal schooling. But actual achievement levels reflect, inlarge part, on the school's quality and capacity to raise achievement,independent of the child's length of school attendance. The research groupfound small income effects from the length of schooling completed and fromwhether a school degree was actually obtained (regardless of actualachievement level). But returns to the level of literacy and numeracyachievement were substantial in both countries, and the benefits wereobserved for both manual and white-collar workers. A rough test of"ability" was also used to control for the effects of preschool cognitiveskills (which explained little variation in subsequent earnings) [1.13].

A related study examined the influence of educational attainment andschool quality on farmer productivity among 683 households in Nepal. Thisstudy has a strong advantage over prior research in that it used actualmeasures of productivity rather than assuming that wage levels act as areliable proxy for the worker's productivity level. Agriculturalproductivity was not sensitive to variation in school quantity amongfarmers completing one to six years of schooling. However, farmers withseven or more years of schooling were significantly more productive thanthose with lower levels of school attainment, after controlling for theinfluence of physical production inputs (land and capital). This suggeststhe presence of a threshold level of achievement or school retention whichis related to subsequent productivity gains.

Similar to the Kenya/Tanzania study, the Nepal research also obtainedinformation on farmers' numeracy levels, linking the study to schoolquality and actual levels of achievement. Variation in farmers' numeracywas not related to overall productivity but did influence output of wheat,a recently introduced crop in Nepal. Sizable rates of return were foundfor higher levels of numeracy (based on the elasticity of the numeracyachievement effect and schooling costs). Importantly, this researchfocuses on the economic benefits of actual school achievement, not only thelength of school attendance. This study design also suggests that schoolquality - already shown to correspond with student achievement - yieldscompetitive economic benefits. In addition, these benefits are independentof and at least comparable to returns from school expansion [1.14].

A second stream of research is looking at whether nation-levelinvestments in school quality influence total economic output over time.This research design also avoids the problem of using income proxies forproductivity, since actual levels of output are assessed. At theindividual level, a close correlation may be observed between the worker'slevel of schooling and his or her wage rate - even when a national economyis experiencing no growth in productivity. By looking at aggregateeconomic output and production inputs (including levels of investment inschool expansion and school quality) the weaknesses of using individual-level data are avoided. In addition, returns to education may also vary aslabor demand patterns change over time. Looking at the relationshipbetween school investments and economic growth over historical periods canavoid the danger of inferring long term relationships from data collectedat one point in time.

Yet here too, most nation-level historical research has examined theinfluence of school quantity (national expenditures, average schoolattainment, or length of schooling for different occupational groups).

15

However, one recent study of economic development in Mexico found positiveeconomic output effects from investments in school quality (in terms ofexpenditures per pupil and primary student persistence rates) and fromrising literacy. This study estimates the influence of school quality andliteracy effects on economic output among Mexico's state for a 60 yearperiod, controlling for variation in physical production inputs (land,labor, and capital). The influence of literacy, a general measure ofschool achievement, was strongest within the urban-based manufacturingsector [1.15].

Issue 3: Policy Tradeoff Between School Expansionand School Quality

This evidence suggests that the economic payoff to educationalinvestment stems not only from school expansion; economic returns toimproving school quality also are significant. And potential benefits ofcontinued school expansion may be seriously constrained unless low schoolquality is also addressed. Many governments and development groups havenot yet faced this tradeoff between further school expansion versusimprovement in school quality.

Two studies have examined explicitly this tradeoff between quality andquantity. The Brazil study found that the social rate of return wasgreater for marginal improvements in quality (again, as measured byteachers' educational levels) compared to increasing the length (years) ofschool attendance. In addition, an individual income effect was discoveredfor the interaction between length of schooling (quantity) and schoolquality. These findings suggest that higher aggregate productivity canresult from improving school quality for a constant number of students,rather than expanding enrollment or increasing the amount of schoolingavailable for a constant number of youth. An illustration is offered bythe researchers: Assume that (1) school places were provided to one-halfthe number of pupils actually enrolled, and (2) the resources saved wereredirected to improvements in quality. The researchers then accounted forthe costs of improving quality with, and income benefits stemming from,this alternative use of educational resources. They found that aggregateincome would be 18 percent higher with this allocation pattern than if allpupils below the mean length of school attendance (three years) werebrought up to this level at the current level of school quality [1.16].

A second study looked at the possible causes of wide variation inschool quality found among developing countries. One question within theinvestigation: Do nations with higher enrollment rates and higher rates ofenrollment growth have lower levels of school quality? Two indicators ofschool quality were used for this cross-national analysis, per pupilexpenditures and the ratio of pupils to teachers. Looking across allnations, countries with higher primary school enrollment rates also havehigher quality schools. However, after controlling for differences innational wealth, developing countries with higher enrollment rates actuallyhave lover school quality in terms of expenditures per pupil. And over the.1970-1980 period, expenditures per pupil declined in those countries withthe highest growth in enrollment rates. The concentration of resources perstudent simply failed to keep pace with the rising number of students.These findings did not hold when using the pupil/teacher ratio as a measure

16

of quality. This evidence further points to a discrete tradeoff betweenschool expansion and improvement of school quality [1.17].

The tradeoff between school quantity and quality becomes moresignificant in light of higher rates of return to primary schooling. Table3 indicates that in some regions the social rate of return from primaryschooling is twice the yield on investment in higher education, due to thelatter's much higher costs and limited income benefits for most societies.Rates of return to larger investments in secondary school also fallconsistently below social income gains provided by primary schooling.Further, social rates of return are inversely related to national wealth.That is, the highest rates of return to schooling are observed in lower-income developing countries. This is largely due to more acute shortagesof literate and skilled workers in poorer nations. Given these diminishingrates of return (a) from the quautity of secondary and tertiary schoolingand (b) from school expansion in general among middle-income developingcountries, a stronger emphasis on improving quality may be advisable.Given the recent research, improvement in quality may offer rates of returnwhich are more comparable to the yield on school investments observedduring early periods of school expansion.

Table 3Social Rates of Return to Education (Length of Attendance)

Region or Level of SchoolingDevelopment Status Primary Secondary Tertiary

Low-income countriesAfrica 28% 17% 13%Asia 27 15 13Latin America 26 18 16

Middle-income Countries 13 10 8Industrialized Countries - 11 9

Source: Psacharopoulos (1985)

In sum, more research is necessary to achieve consistent estimates ofthe economic benefits of school quality and to specify the prerequisiteconditions under which these returns will be felt. Yet this early researchconsistently suggests that the quality of schooling and how much isactually learned by students significantly contribute to productivity andincome growth. Education's contribution to economic development has beenimplicitly thought about in terms of school quantity. This growing liue ofresearch suggests that school quality exerts an independent and (at least)a comparable influence on economic development.

Until significant improvement in school quality occurs, the economicimpact of additional school investments will be constrained. An emphasison school expansion may yield diminishing economic returns by ignoring theindependent influence of quality and by failing to take advantage of the

17

potential interactive influence of accompanying improvements in quality.Initial empirical work also substantiates the intuitive feeling held bysome policymakers that both allocation and benefit tradeoffs exist betweenschool expansion and quality [1.181.

Issue 4: Eroding School Quality

Since World War II Third World governments and development agencieshave emphasized school expansion, not improvement of school quality. Thispolicy priority has shown dramatic results: In 1950, just 37 percent ofall children in the relevant age cohort were enrolled in primary schoolamong developing countries. This rate grew to 58 percent by 1970 and isestimated at 72 percent for 1985. Secondary enrollment rates similarlyrose from 5 to 43 percent between 1950-1985. This level of schoolexpansion is even more impressive when noting that the base number ofchildren was increasing at about three percent annually over this period(1.191.

Yet little knowledge is available on the long term effects on schoolquality resulting from such high rates of expansion. Recent evidence doessuggest that resources available to the education sector have been levelingwhile school expansion has continued. This may indicate a diminishinglevel of expenditures per student and a thinning quality of education. Forinstance, government expenditures as a percent of GNP fell from 21 to 18percent during the 1970's among developing countries. And the averageshare of government budgets allocated to education slipped from 16 to 11percent during this time 11.20]. It is possible that nations experiencingsignificant economic growth may have been able to maintain prior levels ofper pupil expenditures even while enrollments were expanding.

Recent data on two school quality indicators - per pupil expendituresand pupil/teacher ratios - were examined to better understand recenttrends. Table 4 reports changes in school quality between 1970-1980 forall countries reporting these data. Among low-income nations, expendituresper pupil dropped from (constant 1980 US) $109 to $75 over the 1970's.These figures convert expenditure data into constant US dollars forcomparative purposes and adjust for country-specific inflation rates. Perpupil expenditures in middle-income developing nations increased from $127in 1970 to $195 in 1980. Industrialized countries almost doubled theirsupport per student in constant terms over the 1970's.

18

Table 4School kality Treds for the Primary School Level

by Deelopm"t Status, 1970-1980(Muber of Countrie Reporting)

School Quality Lo-incoe Middl-income IndustrializedIndicator Cmutrin Cntries Countries

Primary School ExpenditurePer Pupil(Canstant 1990 USS

1970 $109 127 1,2051990 S 75 S195 82,343ln) 11) (33) (17)

Pupil/Teacher Ratio

1970 44 36 231980 45 31 19(n) (33) (57) (18)

Sourcet Unesco (1983b1)

Similar trends are evident when looking at the ratio of pupils perteacher. Low-income countries showed no real change over the 1970-1980period. Yet middle-income countries reduced this ratio from 36/1 in 1970to 31/1 in 1980. A significant decline is also apparent amongindustrialized nations, falling from 23/1 to 18/1.

Recent trends in school quality among Third World nations are mixed.Middle-income developing countries have made progress in improving theconcevtration of resources applied to each student, even during a period ofsignificant school expansion. But school quality is eroding in the poorestnations, those unable to expand student spaces and maintain prior resourcelevels for each pupil. And the already large gap in school quality betweenall developing countries and industrialized nations is widening.

One might argue that unit costs should go down as enrollments expand,given apparent economies of scale. However, note that middle-incomedeveloping countries have been able to raise simultaneously enrollmentrates and expenditures per pupil. In addition, expenditures per pupil arepositively related to higher student achievement levels. Therefore,alleged economies of scale may be false when falling units costs depressschool quality, pupil achievement, and subsequent economic returns toeducational investments.

19

Suzmary

Knowledge about school quality has grown substantially in recentyears. This experience and research has illuminated four salientcomponents of the school quality problem:

- Low levels of academic achievement and correspondingly lowquality among Third World students and schools, respectively;

- Constrained economic returns to investments in education due toan emphasis on school expansion and relatively little improvement in schoolquality;

- Inadequate policy attention to the tradeoff between furtherschool expansion versus improvement of school quality; and

- A recent decline in school quality among the world's poorestnations and an increasing gap in quality between developing andindustrialized countries.

These issues suggest the desirability of shifting policy attention andresources from further school expansion to improvements in quality. Thisshift is occurring within some Third World governments and developmentgroups. For instance, prior to 1967 the World Bank invested educationresources almost exclusively in school construction. By 1982, just 37percent of the Bankls education capital went to construction. Improvementin school quality has become a major goal for the Bank. Between 1967-1982the number of projects which included support of textbooks rose from threeto 37; the number sponsoring curriculum improvements increased from threeto 56. Yet in many countries and within numerous education developmentinitiatives the emphasis remains on further school expansion [1.211.

Beyond the question of whether more resources should be allocated forschool quality, lies the second-generation issue of how quality should beraised. Some qualities of schools may not be related to higher studentachievement. Other elements of school quality are clearly linked to pupilperformance. The remaining chapters focus on the task of clearlydistinguishing between these two sets of school ingredients. First, Ipropose a more specific definition of "school quality" (Chapter 2).Second, research is reviewed on which specific elements of school qualitymost consistently influence student achievement (Chapter 3). The objectivehere is to identify how schools can be improved to more efficiently impartliteracy and academic skills.

Before proceeding, I should note that the issue of whether educationsector resources should be shifted from further expansion to improvingquality will be set aside for the moment. The push to raise school qualityand student achievement is founded, in part, on the assumption that schoolimprovements can lead to the more efficient use of scarce educationalresources. Higher levels of achievement presumably are possible evenwithin static levels of material school inputs, particularly to the extentthat quality and efficiency are linked to the management and socialorganization of inputs. Improvement in quality can also generate costsavings by spending less on those elements of schools which are unrelated

20

to student achievement. But this argument does not assume that seriousimprovement of school quality is a less costly project than the financingof further school expansion.

Three allocation decisions are involved in addressing low schoolquality: (1) The level of resources allocated to the entire educationsector of a nation or development group, (2) the distribution of educationsector resources between school expansion and improvement of quality, and(3) targeting investments on those elements of school quality that are mosteffective and efficient (taking into account variable costs) in boostingstudent achievement. More careful allocation of limited resources to thoseelements of school quality which most efficiently raise achievement (forinstance, books and instructional materials) could raise the efficiency ofschools. In addition, moving resources away from ingredients of schoolswhich are unrelated to achievement (for example, small classes) willconserve resources. These issues fall within just the third allocationdecision. The magnitude of benefits from carefully attacking low schoolquality are influenced at the first two points. Nevertheless, we assumethat the pursuit of higher school quality can be accomplished independentof the first two policy decisions.

21

Chapter 2Defining School Quality

Introduction

The term "school quality" often is evoked within developing countries.Low student achievement levels - manifest in low test scores, high dropoutand grade repetition rates - often provoke greater concern with schoolquality. Here. "quality" is defined in terms of output or achievement. Theacademic performance of children, however, stems from factors other thanthe character of their school. In addition, a diagnosis of the schoolquality problem that focuses on outputs (pupil performance) fails toinform us about the antecedent causes (inputs and practices). Investmentstrategy founded only upon general output indicators is less likely toaddress specific school factors which are actually related to schoolachievement. A clearer definition of "school quality" could sharpen how weconceptualize the problem and formulate investment strategies.

Second, this chapter summarizes research related to the influence ofschool quality on achievement. Debate continues over the relative strengthof the school versus contextual factors in determining literacy andacademic achievement. As outlined earlier, school quality now does appearto significantly influence achievement relative to the effects of thechild's family background and the community context (in contrast to thelatter two factors' overwhelming influence within industrializedcountries). A decade of research has provided considerable evidence on theaggregate effect of school quality on student achievement. However,research from industrialized nations, and continuing skepticism over theschool's impact, continues to influence development policy. Prior todelineating which school characteristics shape achievement, a review isprovided of the school's aggregate force in boosting achievement.

A Value-Added Definition of School Quality

One useful way to grasp the distinct meaning of school quality is toback-up and ask, What factors inside and outside the school influencestudent achievement? We know that levels of literacy and schoolachievement are low within many Third World countries. But this lowperformance should not be entirely attributed to the school. Table 5illustrates three forces which determine achievement levels. This simplemodel identifies school-related and contextual factors which operate as thechild develops from birth.

At Time 1, prior to entering school, social contexts faced by youngchildren vary enormously within developing countries. During thispreschool period (Box 1) material conditions shape the child's health andnutritional status, which in turn shapes cognitive capacities. Parentaland community beliefs regarding literacy and education, exercised early inthe child"s life, also have been found to influence later school

22

Table 5Factors Influencing Student Achievesent

Chili FI- -- Ti e I liee 2 Time 3 liee -Pre-school to-school Leaving school Post-school

…--- -- --------- ------- ------ --------------------------… ---- … ---------------- -----…----…------------

2. SCHOOL WUALITV

- Naterial inputs- Tacher Duality- leaching practices and

classroom organization- School *anageaent and

struchure

1. PRESCHOOL DEVELWHENH 4. SCNOOL WUICONES 5. POST SCHOOL OUTCONES

- Health and nutritlion - Literacy level and -p - Job attainmenttd - Cognitivn grouth academic achievement and earnings

- Parents literacy and - Social skills - Social qualityendorsement of schooling of life

3. SOCIAL I ECUOHNIC CONTEIT A

- Demand for child's labor- Opportunity costs ofattending school

………_… - -… - -_________________--_

achievement. Parenting practices also vary in terms of parents' nurturanceof cognitive skills through various behaviors and interactions. Certainforms of interaction between parent and child are related to higherintellectual development and school achievement later in the child's life.This has been observed particularly among parents who more frequently asktheir child questions, allow the child to solve problems with some degreeof independence, and encourage early reading [2.1].

In Time 2, once the child enters school, characteristics of theinstitution may influence his or her level of literacy and academic skills.This potential degree of influence exercised by the school stems from thequality of school attended. Here the emphasis is on the quality of theinstructional process experienced by each student - the school's efficacyin imparting higher levels of literacy and cognitive skills.

Table 5 also illustrates factors external to the school whichinfluence the child's eventual level of academic achievement (Box 3).Within many developing countries the strongest intervening factor is demandfor the child's labor and his or her potential earnings, particularlywithin rural and poor urban families. On the other hand, more affluentfamilies do not.require that children contribute to the household's income.This frees considerable time for school attendance. Opportunity costs ofentering and persisting through school also vary. This factor operatessomewhat independently of the family's wealth. For instance, evidence fromMexico and Brazil suggests that children of more productive rural familiesattend school less, given higher opportunity costs [2.2]. Finally,immediate material and health conditions, including drought and disease,represent contextual factors which influence eventual school achievement.

School quality in defined here as (a) the level of material inputsallocated to schools per pupil (resource concentration), and (b) the levelof efficiency with which a fixed amDunt of material inputs are organizedand managed to raise pupil achievement.

This definition reduces the imprecision evident in looser uses of theterm "school quality." First, the concentration of material resources perstudent disguishes quality from the capacity or size of a school system.This is the fundamental conceptual shift for policymakers and localeducators: Moving attention from how many children are being served, towhat is the quality of instruction for each child?

Second, the definition emphasizes the school's value-addedcontribution to the child's achievement, independent of preschoolbackground, community context, and ongoing demand for his or her labor.School quality cannot be validly measured in terms of the quality ofoutputs, unless factors outside the school organization are taken intoaccount. A low quality school may graduate relatively high-achievingstudents - if it serves children from more affluent families. Similarly,a school that effectively raises the literacy of, for instance, poor ruralchildren would be of high quality - even when graduates are achievingbelow urban children. The focus here is on the difference made by theschool organization itself in boosting on children's literacy and academicskills.

Third, all characteristics and inputs of schools are not related to

23

achievement levels. The definition specifies that only inputs andmanagement practices empirically linked to pupil performance are validindicators of school quality. Other elements of schools (for instance,classroom laboratories) may come to be symbols of a "higher qualityschool." But labs would not be useful measures of quality, according tothe definition, since they are not related to higher pupil performance.Similarly, marginal decreases in average class size yield no consistentgain in pupil performance (Chapter 3). But schools with somewhat smallerclasses may be perceived as being of higher quality. Investing in a widearray of school quality may yield important symbolic benefits for bothgovernments and development agencies. But unless the elements of schoolquality are related to achievement, the investment will be ineffective andlower the efficiency of schools in providing higher pupil performance.

Fourth, the definition emphasizes that a school's level of qualitydoes not equal the aggregated sum of various material inputs (per pupil).In short, more is not necessarily better. Problems arise when one tries tomeasure total school quality in this way. The level of material inputs perpupil undoubtably influence student achievement levels. This includes theavailability of textbooks, desks, and writing materialst for each pupil.The length of the instructional program (per day and over the school year)also are determined by the level of material resources. And empirically, aschool's global level of expenditures per pupil is a consistent predictorof student achievement, controlling for students' background.

However, looking only the level of material inputs ignores thequestion of whether these resources are efficiently managed. Thedefinition proposed here emphasizes that the social organization ormanagement of the school and the classroom also denotes the level of schoolquality. Research has identified three subsets of factors:

(a) Teacher quality, especially in terms of their own schooling,social class background, verbal proficiency, and motivation;

(b) Teacher behaviors within their classrooms, includingefficient use of instructional time, the level of performance standards andexpectations set for students, the extent to which teachers evaluatestudents' performance, and teachers' ability to place students in activelearning roles, not simply requiring passive behavior by students; and

(c) Organization of the school, including the headmaster'smanagement capability, a feeling of camaraderie among teachers, norms ofachievement set by the headmaster, and the school's legitimacy within thesurrounding community.

Simply increasing the level of material inputs does not guarantee thatthese management and social elements of quality will improve. Holdingconstant material resources per student, schools vary enormously in thequality of their management. Inadequate attention to these socialvariables may severely constrain the impact of additional material inputs.

Fifth, the definition distinguishes between the concepts of schoolquality versus internal efficiency. Schools which expend more resourcesper students are of higher quality in that they are more likely empirically

24

to graduate higher achieving students. Or developing countries that decideto shift scarce resources from expanding the number of schools to spendingmore per pupil, will improve school quality from the standpoint ofconcentrating funds on fewer children. In each case, schools will bebetter able to invest in those material inputs (e.g. textbooks) which willboost pupil achievement. These schools will become more internallyeffective. However, this does not necessarily mean that they will be moreinternally efficient. Among those school characteristics which are relatedto higher student achievement, a subset comprise the most efficient devicesfor raising student performance. For instance, textbooks may raiseachievement less than would an equal investment in writing materials. Orthe increment of achievement realized from textbooks may be at a cost threetimes the level required relative to an alternative school input. Thus theexercise of improving school quality or effectiveness is only a first steptoward raising efficiency. The first requires investing in those inputsand organizational practices which relate to achievement. The second stepnecessitates identifying the most effective and least costly schoolcharacteristics on which investments can be focused [2.31.

I do not address one essential element of school quality - thecontent of cur4iculum - given its variation across cultural settings andschools. In contrast, material inputs and social organization areimportant elements of school quality, largely independent of the school'ssocialization or curriculum goals. However, latter two components ofschools are neither culture-free nor neutral in their effects. Forinstance, schools and classrooms in China and Japan often are structured toencourage cooperation and interdependence among students. In contrast,schools in Western cultures generally emphasize individual competition anddiscourage interdependent forms of learning and working. The level ofmaterial inputs applied to schools variably serve to legitimate such socialrules. And the social organization of schools may or may not fit ways inwhich a culture organizes families or workplaces. Yet the assumption madehere is that under any type of formal socialization - within families,apprenticeships, within work settings, or in separate schools -improvements in educational quality require attention to material inputsand social organization. I focus on these two sets of schoolcharacteristics given their importance across diverse cultures andeducational arrangements.

The Internal Effectiveness of Schools

The model of student achievement sketched in Table 5 prompts afundamental policy question: What is the strength-of school quality -relative to the child's preschool development and the community context -in shaping achievement levels? As introduced in Chapter 1, work withinindustrialized nations consistently finds that schools have little effectin determining academic performance or eventual economic success, aftercontrolling for the student's preschool development and communitybackground. These findings initially came from the Coleman Report in theU.S. and the Plowden Report in England over 15 years ago.

This early line of research continues to fuel skepticism over thewisdom of increasing school investments in developing countries. Forinstance, in 1975 a World Bank paper (Alexander & Simmons) addressed the

25

issue of whether schools influence academic achievement, reviewing researchfrom Western Europe, the U.S., and initial studies from the Third World.Based on the research at that time, the authors concluded that schools madelittle difference in raising literacy and academic skills after accountingfor the family background of children and their community context.However, the subsequent decade of research within the Third World yieldedconsiderable evidence that school quality makes a substantial difference.And the impact of school quality relative to external factors appears to begreatest among the poorest developing countries.

The social class background of children - operating through preschoolexperiences and ongoing community pressures - at times does influenceacademic achievement in the Third World. One study in 37 Sri Lankansecondary schools found a high correlation between pupils' social classbackground and school performance. Evidence from India, Peru, and Malaysiasuggests that this correspondence is stronger for achievement in readingthan for student performance in math or science. One review of 33 studiesof school achievement determinants focused, in part, on the frequency withwhich effects from students' background were discovered. The authorsreported significant social class effects in 73 percent of the cases invhich the influence of family background was empirically investigated.This evidence clearly suggests that preschool development and the communitycontext makes a significant difference. But this work does not speak tothe magnitude of social class effects relative to the influence of theschool institution itself.* -

This issue is directly addressed in one recent analysis of thecontribution of school quality to student achievement within 16 developingand 13 industrialized nations (Heyneman & Loxley). An index of schoolquality was constructed for each country, based on those materialcharacteristics of schools that were significantly related to pupilachievement (in science). Many material elements of school quality wereincluded in the analysis, such as expenditures per pupil, textbookavailability, length of the instructional program, and teacher credentials.The share of variation in student achievement explained by this schoolquality index relative to the variance explained by students' backgroundwas reported for each nation. Methodological difficulties arise in tryingto distinguish between the relative strength of school quality versusstudent background when these two factors are highly correlated. That is,when the highest quality schools are located in the most affluentcommunities, it is difficult to disentangle the independent force of eachfactor. Covariation was found within some countries, including Finland,Japan, and Chile. Yet no correspondence between school quality andstudents' background was observed in other nations, including India,Thailand, and Italy [2.5].

Table 6 reports the extent to which school quality and students'family background contributes to student achievement in science. Theirrelative strength is expressed in terms of the share of variance in pupilachievement each factor explains. That is, achievement scores vary widelyamong youth within each country. The policy related question is whether astudent's family background or the quality of school attended more stronglyaccounts for this variation in achievement. The percent of varianceaccounted for by each is reported for the poorest seven countries and forthe seven most affluent nations included in the study.

26

In general, the aggregate index of school quality is a strongerdeterminant of achievement within the developing country group. Among all29 nations included in the study, a strong negative correlation wasapparent between the amount of variance in achievement explained by schoolquality and the nation's wealth. This offers initial evidence that schoolquality substantially shapes achievement, after controlling for the effects,of the student's background. The strength of this pattern across diversecountries minimizes concern over some distortions which may result from theoccasional correspondence between school quality and student backgroundwithin a particular nation.

Table 6Relative Influence of School Quality and Student Background

GNP Per Capita Variance in Achievement Explained by:Country US$ 1971 Student Background School Quality

India $110 3% 27%Uganda 130 6 5Botswana 160 6 14Bolivia 190 11 24Thailand 210 - 6 25Egypt 220 6 14Paraguay 280 23 16

Netherlands 2,620 22 11Australia 2,870 17 7French Belgium 2,960 14 16Flemish Belgium 2,960 12 16Germany 3,210 17 14Sweden 4,240 18 7United States 5,160 21 13

Source: Heyneman & Loxley (1983)

Summary

The allocative efficiency of educational investment stems from theclarity of how a problem is defined. General output definitions of "schoolquality" have resulted in rather vague ameliorative investments. A cleardefinition of quality should focus on the concentration of materialresources per student and the efficiency with which these inputs aremanaged and structured within schools and classrooms. Quality should notbe denoted by school charateristics which are unrelated empirically tohigher student achievement. Nor should we ignore the influence of factorsexternal to the school on student performance. The child's preschooldevelopment, for instance, plays a large role in later achievement inschool.

27

On the other hand, the school's potential influence in raisingliteracy and academic skills appears to be considerable. This is trueespecially within the lowest income countries. WThether the child actuallybenefits from this potential effect depends largely on the quality of theirschool. Efforts to raise school quality are efficient given theinstitution's potential impact. These efforts, however, should move fromcareful definitions of the problem.

28

Chapter 3

What School CharacteristicsBoost Achievement?

Introduction

The influence of aggregate levels of school quality on studentachievement is now firmly established for most Third World settings.However, much less is known about which elements of school quality areeffective in shaping achievement. The first generation of research onschool quality was concerned with determining whether the schoolorganization made a difference. Therefore, the aggregate influence ofvarious elements of school quality was investigated. Delineating whichspecific school factors boost pupil performance is a different task. Inaddition, shifting investments from school expansion to improving schoolquality and effectiveness is conceptually simple. But fine tuninginvestment policy to target resources on the most efficacious ingredientsof schools is a more complex endeavor.

First, those elements of school quality which are effective in raisingstudent achievement must be separated from those that are unrelated topupil performance. Second, the magnitude of influence exerted byefficacious elements of school quality must be estimated. Third, therelative cost-effectiveness of significant school factors must be examined,accounting for differing magnitudes of influence and varying costs amongdifferent school quality elements. Whether policymakers and local schoolmanagers feasbily can move through each analytic step in detail is notclear. However, this simple stream of questions can certainly be pursuedin many decision-making settings.

This chapter focuses on the first step: Identifying those elements ofschool quality which consistently boost pupil achievement and isolating theother set of school characteristics which make no difference. Research onthis question has blossomed in the Third World over the past 15 years.Here the findings of 72 studies are reviewed. Little knowledge exists onthe relative magnitude of effects from different school factors. And evenless evidence is available on the relative costs of alternative schoolfactors. More efficient use of resources would occur if scbool resourceswere shifted from ineffective to effective elements of quality. But evengreater efficiency would result from centering investments on the mostcost-effective ingredients of schools. Chapter 4 reports on what littleresearch has been conducted in this area.

Lessons from Industrialized Countries?

School quality research in developing countries remains largely

29

reactive to the policy question initially asked two decades ago: Doschools make a difference in shaping the academic achievement andoccupational success of young people? Only recently has research shiftedfrom looking at whether school quality makes a difference to identifyingwhich elements of quality consistently influence student achievement.

Yet research methods employed in the school quality field (within theThird World) continue to fit the more global question. That is, the focusis on the aggregate influence of school quality on achievement, afterempirically controlling on the child's family background and communitycontext. Typically these investigations study the general correspondencebetween student achievement levels and various school factors (forinstance, class size, expenditures per pupil, and teacher qualifications)based on large scale school surveys. It is not unusual for suchproduction-function studies to look at the independent influence of 15-30elements of school quality on student achievement. The review of thisempirical work which follows does report on specific school factors whichdo or do not consistently shape achievement. But this is really asecondary purpose of this line of research. Only the most recent workcarefully assesses the relative impact of specific elements of schoolquality.

The school quality field is bounded by a second lasting effect ofearly research on school effects. The production-function metaphorcontinues to relied upon to represent the schooling process. That is, theschool organization is seen as a firm which receives material inputs andtransforms these resources into the production of educated children.Following my review of Third World research, I examine the utility ofapplying a material production model to the largely social process ofteaching and learning.

Overview / What Elements of School Quality Influence Achievement?

This summary of research is based on a review of empirical studieswhich examine the influence of school quality on pupil performance indeveloping countries. The studies vary widely on what particular schoolquality elements are examined. All investigations, however, control on thestudent's social class and economic background in some manner. That is,this literature examines the effects of school characteristics independentof family background and community influences. Each study also employedsome means of testing for the statistical significance of any observedeffect. Only statistically significant effects of school quality elementsare reported in this review (p<.05 for beta coefficients, production-function elasticities, or f-values within experimental analysis of variancedesigns) [This review build on previous research summaries, 3.11.

Table 7 summarizes the findings from all studies reviewed. Column 1indicates the specific elements of school quality which have beeninvestigated. Column 2 reports the expected direction of correlationbetween the specific quality factor and pupil performance. Column 3reports the number of statistical observations made between the particularschool factor and achievement. For instance, among all studies reviewed,11 analyses were conducted on whether expenditures per pupil was related toachievement. Each study examined a subset of school quality elements. On

30

Table 7Influence of School Quality Elesents an Student Achievement

(1) (2) (3) 14) (5) (b)School Quality Expected Total Nusber Number of Analyses Number of Analyses ConfirmatiIndicator Direction of of Analyses Confirming Effect Reporting No or Rate

Relationship Negative Effect

School Espenditures n

1. Expenditures per pupil + 11 6 5 54%2. Total school ependitures + 5 2 3 40%

Specific Material Inputs

3. Class siz - 21 5 16 24%4. School size + 9 4 5 44%5. Instructional materials

Texts and reading uterials + 22 14 B 64%Desks + 3 3 0 IOOI

6. Instructional media (radio) + 3 3 0 100%7. School building qulity + 2 2 0 100%. Library size and activity + 18 15 3 83%

9. Science laboratories + II 4 7 36%10. Nutrition and feeding progrm + 5 5 0 100%

Teacher Quality

I1. Teacher's length of schoolingTotal years of teacher's schooling + 25 11 14 441Years of tertiary & teacher training + 30 21 9 70%

12. Inservice teacher training + 5 4 1 80%13. Teacher's length of eperience + 23 10 13 43%14. Teacher's vwerbal proficiency 4 2 2 0 100%15. Teacher's salary level + 13 4 9 31%16. Teacher's social class background + 10 7 3 70%17. School's percent of full-time teachers + 2 1 1 50%19. Teacher's punctuality & (low) absenteeis + 2 0 2 0%

Teaching Practices / Classroom Organization

19. Length of instructioal program + 13 11 2 89520. Homork frequency + 7 5 0 71%21. Active learning by students + 2 0 2 Ox,22. Teacher's expectations pupil performance + 3 3 0 100l23. Teacher's tine spert co class preparation + 5 4 1 80%

School Managemnt

24. Quality of principal + 7 4 3 57%25. Multiple shifts of classes each day - 3 1 2 33%26. Student boarding + 4 3 1 75%27. Student repetition of grade 4 5 1 4 20%

31

the other hand, an individual study may include analyses with the samequality factor for different pupil subgroups. Continuing with the example,Column 4 then reports that in seven of the 11 analyses a (statistically)significant relationship was found (again, always controlling the effectsof the student's background).

Table 7 is useful in illustrating what elements of school quality havebeen most often studied. Both material inputs and teacher quality(proxies) have received the most attention from researchers. Despitegrowing research interest in the influence of teaching practices and schoolmanagement within industrialized countries, these areas have received verylittle attention in the Third World literature.

In addition, Table 7 differentiates those elements of quality whichare not consistently related to student achievement from those factorswhich more frequently appear to exert an influence on performance.Elements of school quality which are significantly related to achievementin at least one-half of the analyses are assumed to hold a "consistent"influence. This definition is somewhat arbitrary. Yet this rule of thumbis based on the fact that most of the studies reviewed included more than10 elements of quality in multivariate models. Therefore, even when anelement is related to achievement it will not appear to be statisticallyrelated if it covaries with another indicator of school quality.

Given this body of evidence, school factors which influence pupilachievement can be distinguished from those which do not.

Set 1 Quality elements uot consistently related to achievement:

1. Class size (no effect in 16 of 21 analyses)2. Laboratories (no effect in 7 of 11 analyses)3. Individual teacher (no effect in 9 of 13 analyses)

salary levels

Set 2 Quality elements consistently related to achievement:

4. Expenditures per pupil (effect in 6 of 11 analyses)5. Instructional materials (effect in 17 of 25 analyses)6. School library activity (effect in 15 of 18 analyses)7. Teacher training (effect in 21 of 30 analyses)

(tertiary level)8. Length of instructional (effect in 11 of 13 analyses)

program9. Teacher's social class (effect in 7 of 10 analyses)

Patterns of influence are mixed for several other elements of schoolquality. For instance, the teacher's total length of schooling (includingyears in primary and secondary school) affected achievement in 10 of 24studies. Whether this is a "consistent pattern" is difficult to infer.Other quality factors show promise, particularly frequency of homework, useof instructional radio, and inservice teacher training. But aninsufficient number of studies have been completed from which togeneralize.

32

Interpretation of these general patterns must be couched within thefeatures of specific studies. The following discussion details researchwithin each area of school quality. The review will also highlightfindings on the magnitude of reported achievement effects from specificelements 'of school quality. However, this area is problematic. Any onefactor's level of influence on achievement depends upon many aspects of theresearch design. Empirical models and data sets vary enourmously in (a)the number of factors included in each multivariate model, (b) the level ofobserved variance for specific elements of quality, (c) the amount ofmeasurement error for indicators of quality, (d) the degree of colinearityamong quality variables, (e) the level of instruction studied, and (f) thediversity of settings within which schools are studied. It is difficultfor researchers to employ even roughly consistent designs and impossible tohold constant scope conditions within which findings would apply. Andagain, central concern with the aggregate question - does school qualitymakes a difference - diverts attention from the critical issue of what isthe magnitude of observed effects among different school factors.

I turn next to reviewing findings within each of the primary areas ofschool quality. A table summarizing the literature accompanies eachdiscussion below. Citations have been kept to a minimum in the text, sincethe tables contain formal reference information.

Overall School Expenditures

Table 8 summarizes existing empirical studies that have examined theinfluence of expenditures per pupil and total school expenditures onstudent performance. Each expenditure measure is a global indicator ofquality. Per pupil expenditures more directly indicate the concentrationof school resources on each student enrolled in school (Indicator 1). Thebulk of resources go to teacher and school staff salaries. Therefore, thisglobal measure does not indicate teacher quality nor specify the level ofresources allocated to books, instructional materials, or other inputs moredirectly linked to the instructional process. Different analyses look atvarious areas of achievement (Column 4). Any reported effect (Column 5) isstatistically significant (p<.05). Some findings are limited to certainconditions (specified in Column 6). And the studies reviewed utilizevarious methods (see Column 7 and notes at the foot of Table 8).

Summarized above, a majority of studies have found a positiverelationship between school expenditures per pupil and achievement. Theinfluence of this school quality element is strong in some analyses. Forexample, in the Colombian study of academic achievement (4,233 secondaryschool graduates), the influence of per pupil expenditures was exceededonly by the youth's verbal ability (comparing t-statistics among 24 studentbackground and school factors tested). The influence of this factor wasonly slightly less strong for achievement within the commercial curriculum.This same study found more moderate, but significant, achievement effectsin Tanzania (2,803 secondary graduates).

The magnitude of the achievement effect from per pupil expenditureswas thoroughly examined -within an early study of primary and secondaryschool students in Kenya. No significant effect was found at the primaryschool level. At the secondary level, per pupil expenditures did influence

33

Efficacy of School Quality Elesents:School Expendituresu

(1 (21 (3) 14) 16) (71School Quality Hypothesis Country kbievesent or Economic Direction of Effect Setting StudyIndicator ISchool Level) Outcoee Neasured 4 O - (Method)

1. School expenditures Higher expenditures mill Argentina Science achievement 4 Heyneman I Louley (MR)per pupil provide higher quality IPrisary)

teachers and learning Kenya National exam 0 Thias & Carnoy (PFIresources, leading to IPrimary)higher achievement (Secondary) 4

Malaysia Comprehensive exam 0 Beebout (IRI(Secondary)Nexico Agricultural 6 * Fuller et al. (PF)IPrisary) manufacturing output

Science 0 Neyneman & LaIleyColombia kademic I vocational + Psacharopoulos I(Secondary) Loxley (MRITanzania Vocational knowledge 4 For 2 of 3 tracks

Lagvuage 0Bolivia Science * Urban schools Norales F PinellsileswPrimary) 0 Rural schools IMRI

2. Total school Overall expenditures Chile Science 4 Heyneman & Louleyexpenditures reflect higher quality, bruil 0

leading to higher student Paraguay 0achievement Colombia 0

Mexico 4

* Column headingsi Colmn 3 indicates both the country within which the research was conducted and the school level of students studied (primary or secondarygrades). The school level last specified applies to the current country, reading down the column, unless a new school level is indicated.

Coluen S reports a positive or negative effect only when the study found a statistically significant influence of the particular school quality input (p ( .05).Ihen a '0' is indicated in Coluon 5, this means that no significant effect was found for this school quality input. The direction of the effect reported relatesto the direction indicated in the stated hypothesis (Column 21.

Column 6 specifies a specific condition underuhich the findings hold. If know scope condition is specified, the effect vas found for the entire sample ofstudents being investigated.

Column I mentions the type of empirical analysis utilized by the researcher(s). The following sysbols vere usedi KR m sultiple regressions with reporting ofbeta coefficients; PF * a log-linear production function with a report of elasticities; CS an experimental or cosparison (controll group study was conducted; CO* correlation coefficients were reported; bivariate correlations reported only Nhen no relationship was found.

achievement. The researchers then estimated the magnitude of this effect.Second, they identified the strength of relationship between highernational test scores (indicator of achievement) and future earnings. Theauthors found that raising national exam scores among sampled secondarystudents by five percent would require a 50 percent boost in expendituresper pupil. Yet the rate of return to this investment in terms of increasedincome would be significant (six percent) -- but only among low achievingstudents. For relatively high achieving students, income benefits gainedfrom higher test achievement were entirely offset by the cost of raisingachievement.

A growing amount of research within industrialized countries suggeststhat per pupil expenditures - as a concrete indicator of school quality -is related to individual income of graduates. This claim has not beenextensively tested in developing countries.

Four analyses have found no achievement effects from higher per pupilexpenditures, including well designed studies in Malaysia and Tanzania.In addition, one might argue that schools must reach a threshold sizebefore achievement can be significantly influenced (Indicator 2). Yet interms of total expenditures, just one in four analyses discovered astatistically significant effect.


The second set of school quality elements includes discrete materialinputs which operate more closely to the instructional processs (Table 9).First, Third World findings on possible achievement effects from smallerclass size are quite similar to research within industrialized nations(Indicator 1). Within normal ranges, the presence of fewer students perclassroom has no consistent affect on achievement. No effect from smallerclass size has been found in 11 of 21 analyses. In five additionalstudies, students working within larger classes actually performed athigher levels. We should be quick to point out that "normal ranges" areoften exceeded in the poorest developing countries, given that the meannational ratio of pupils to each teacher equals 44 within the poorestnations as reported in Chapter 1. Note that due to double shifts, this isnot an exact measure of class size. Further, achievement effects forspecific ranges of class size have not been investigated in the ThirdWorld.

Yet given this existing evidence, only substantial reductions in classsize would likely raise achievement. And modest increases in class sizewould free up a significant level resources while not diminishing overallstudent achievement. Just one study has found that smaller classes at theprimary level are significantly related to higher achievement. This reportcomes from an analysis of science achievement among 837 urban students inColombia. The bulk of studies simply reports no effect for class size.

Two studies have examined the magnitude of the class size effect (orlack thereof). First, the Malaysia study of 89 secondary schools found asignificant effect of smaller class size and pupil achievement in languagelearning (in Malay). The researcher then estimated the marginal(achievement) product associated with spending one additional dollar to

35

Table 9Efficacy of School Quality Elements.


I1) (2) 13) 14) 15) (A) (7)School quality Hypothesis Country Achievesent or Econonic Direction of Effect Setting StudyIndicator ISchool Level) OutEcno Neasured * 0 - (ethod)

…--------------------- --------------------------------------------…------_-____-___ -_________-___-_--_-__-__------------…---------------------…-- -

I. Class size Fewer students per lotswana Skience Heyneaun & Losleyte acher mill improve the (Prinary) Loxley INR)quality of interaction Thailand Reading L scienceand raise achievesent India 0

Chile 0 Comber & Keeves l1R)

ISecondary) 0 Schiefelbein & Farrell- Poorer schools (1973)

Iran 0 Thorndike (MA)Reading & oath I Ryan (MIA

Egypt Literacy I numeracy 0 Hartley & Swanson (MR)(Primary)Kenya National enas 0 Thias I Carnoy (MR)Malaysia Comprehensive ex u Beebout (MR)(Secondary)(Prisary) 0 Karon (KR)Puerto Rico Reading Carnoy I(R)(Primary/sec.)Tanzania Language * Psacharopoulos L

LoxleyVocational knowledge 0

Indonesia Letters I arts exas 0 Sembiring L(Secondary) Livingstone IMR)Bolivia Science - Urban schools Morales I Pinellsiles(Primary) Reading 0 Urban sthoolsrgentina 0 Echart et a. (MR)Colombia Science * Urban schools Arriagada (1901, MR)

2. School size Whei a threshold school Chile Reading I science t Comber L Keevessize is reached, higher IPrimary)C equality resources mill Thailand 4

be available, raising Iran tachievesent Kenya Cambridge exam Thias & Carnoy

IPrimary)ISecondary) IMalaysia Comprehensive exam Beebout(Primary) 0 HaronCongo Language & math Youdi (KR)Bolivia Reading & science O Norales & PinellsilI(Secondary)

3. Instructional Greater availability of Uganda Comprehensive exam (Actual counts) Heyneman & Jamison Imaterials of tests and reading (Primary)materials vill raise (Secondary) t Heyneman (NRIthe quality of Egypt SCience * (Student/teacher Heyneman & Loxleylearning activities, El Salvador t reports of adequacy)increasing achievement Paraguay

Boliviai MexicoPeru 0Brazil 0

Comprehensive exam n Student owns text Wolff (RR)Colombia Reading & math 0Argentina 0India Reading & science Science texts Comber I KeevesChile Language & math * Schiefelbein & Farrel

(197:Nicaragua Math t Jamison et al. IC6)Philippines Science * Heyneman it al. ICS)Indonesia Arts & science O Sembiring &(Secondary)

LivingstoneThailand Comprehensive exam 0 Fuller & Chantanvani,(Primary)Malaysia Rural schools Haron

0 Urban schoolsShana Reading comprehension I Smart (CO)ISecondary)Malaysia Language stills 0 6eebout

Provision of desks will Egypt cience * Heyneuan I loxileyintrease opportunity to IPrisary)read and writel raising El Salvador aachievesent Peru Reading + Arriagada (1983; IINI

4. Instructional media Radis and other Nicaragua Kath t Jamison at al. (IR)instructiona) sedii (Primary) Student promotion a Jaison (IMR)efficiently raises Philippines Language, oath & science 4 Heyneuan et a1.student achievement

5. Physical Better facilities Uganda National eals + Composite of Heynesan h Jadisonfacilities provide sore motivating (Primary) facility quality

conditions for learning Peru hieding & science a Electric lights krrialada 11983)(Primary)

6. Library The presence and active Argentina Science a Heynesan I Loaleyactivity use of a sckool library (Primary)

mill boost reading hruil lachiovement 0 Costa t1R)

Hexico 4 Heyneoan L LoIleyBolivia 4ParaguayEl Salvador Botsiana 4

Peru k Arriagada (19831Chile 4 tbarodikeWPrimarylsec.1Iran 4

Thailand +India 4

Nalaysia Comprehensive exam BeeboutfSecondary)(Prisary) a Volumes in libraries Haron

0 In-class librariesIndonesia Arts * science 0 Sembiring &(Secondary) LivingstoneSotsiana Reading & math * Loxley(Secondary)

1. Laboratories Tbe presence and India Science 4imstractional tise spent IPrisary)in laboratories mill raise Thailand science achieveaent Iran *

Brail +Chill Peru 0Paraguay 0Notieo 0Coloabia Argentina Bolivia 0

3. Shool feeding aluletritilm Vill loawr Guatn ala Verbal still enrollment + Balderston rt al. 11programs student achievement tPriuary I sec.)

Egypt Reading & math + Ocurece of Hartley i SwansonIPrisary) illnessChile 4 Schiefelbein & Farre

l197Thailand + Fuller & ChantavanicTUganda Comprehensiv exa + Hynean & Jamison

LA) - - - -

help lower class size. Raising student achievement by just one percent (onstandardized exams) would cost an additional $50 per student if allocatedto help lower class size. In contrast, this same increment in higherstudent achievement could be accomplished at one-third the cost ifresources were allocated to teacher training. Cost data used for theseestimates are admittedly rough. But this method for comparing the cost-effectiveness of alternative elements of school quality is instructive.

A second study examined the relative magnitude of achievement effectsresulting from the introduction of textbooks, from establishment of radioinstruction (in Nicaragua), and from lowering class size (in the U.S.).Various experimental-studies provided an estimate of how strongly eachfactor influenced achievement (in terms of standard deviation gains betweencontrol and treatment groups). The researcher then estimated the cost ofachieving the same gain in achievement from each element of school quality.To obtain the achievment benefit gained from raising the availability oftextbooks at a constant increment of cost, schools must lower average classsize from 40 to 10 pupils per teacher! This method of equating the costsfor alternative inputs against a standard gain in achievement yields aclear picture of the relative efficiency of alternative interventions. Theefficiency df introducing radio instruction was even higher than boostingthe availability of books in the Nicaraguan case. The main point remains:In most situations lowering class size with the intent of raisingachievement is not an efficient strategy [3.23.

The positive impact of instructional materials - especially thosedirectly related to reading and writing - is consistent across severalstudies. The availability and use of textbooks (measured, for instance, interms of the number of textbooks in a classroom per student) have beenlooked at in 22 analyses (Indicator 3). Significant effects were observedin 14. Early research in the 1970's relied on IEA survey instruments whichasked students and teachers about simple availability of textbooks inclassrooms [3.3]. This factor also was significant in many Latin Americancountries which employed these measures. Actual counts of textbooks inUganda also revealed a significant influence'on pupil achievement.

This survey research generally indicates a moderate influence oftextbooks and instructional materials on achievement. In Uganda, forinstance, textbook availability strongly influenced achievement in English,dwarfing the effects of the child's social class (based on 1,907 studentsin 61 primary schools). However, averaging across curriculum areas to lookat total achievement levels, the influence of textbooks was smaller thansocial class, preschool competence, pupil health, aud the teacher's verbal(English) proficiency (comparing t-statistics). In Malaysia and Chile,textbook availability was related to higher achievement. But thecorrelation between these two variables was less than .20, prior tocontrolling for the effects of student background or other school qualityelements. Textbooks did not explain more than four percent of thevariation in achievement among all students sampled within each country.

The influence of textbooks appears to be stronger within rural schoolsand among students from lower income families. In rural Brazil, forinstance, students with parents who had received no schooling were almostthree times as likely to pass primary school if they had used two or morebooks (67 percent graduating), compared to students in this same group who

39

had no textbooks in school (only 24 percent graduating). Among studentswith parents who had completed primary school, 73 of all pupils with atleast two books passed primary school, versus 61 percent of those with nobooks (total sample equalled 1,006 primary school students). Similarly,the study of 6,056 Malaysian youth found that the availability of books inschool was more strongly related to achievement among lower income childrenfrom Chinese and Indian ethnic groups.

Clearer evidence on the magnitude of textbooks' effect comes from morerecent studies which have employed experimental research designs - therebyholding constant student background and other school factors. Forinstance, a'controlled evaluation in the Philippines provided textbooks to2,295 first and second grade pupils within 52 schools. A control group ofsimilar schools was also selected. Books were then introduced at ratios of2 pupils per book and 1 pupil per book in alternate classrooms.Achievement gains resulting from the intervention were substantial. Infirst grade science, performance was .51 of a standard deviation higherwithin the experimental classrooms, .30 higher in mathematics, and .32higher in Pilipino. The .51 change (in units of the standard deviation)indicates that the mean score achieved by 50 percent of all students wasobtained by 69 percent of those students in the treatment group. Thisimprovement is twice the impact of what would be gained by lowering classsize from 40 to 10 students (Philippines textbook data, U.S. class sizedata).

The influence of the textbook program on achievement was greater forchildren with parents who had received less schooling. The correlationbetween the child's social class and science achievement was modest for allstudents. Yet this association was not at all evident for pupils receivingtextbooks. Nor did the child's social class influence gains in achievementscores. Interestingly from an efficiency viewpoint, the concentration oftextbooks (2:1 versus 1:1 pupil to book ratio) made no difference on levelsof pupil achievement. The magnitude of effect on Pilipino and mathematicswas more modest. This may be the result of greater difficulty in usingthese latter texts relative to the science volume, as reported by theteachers.

Less robust, yet significant effects of textbooks were found in anexperimental program in Nicaragua. Eighty-eight (88) first-gradeclassrooms within rural and urban schools participated in the program,including 1,098 children. These classrooms were split into three groups:those receiving textbooks, those receiving radio instruction, and thoseserving as control classrooms (the findings for radio instruction arereviewed below). The interventions were applied at the beginning of theschool year. Post-tests were given at the end of the same academic year.Pupils who received textbooks scored four percent higher on the mathematicspost-test (one-third of a standard deviation). The researchers note thatNicaraguan teachers were less schooled than teachers in the Philippines.The latter also had more experience with textbooks either as studentsthemselves or during their teacher training.

A variety of measures have been used in survey studies to determinethe presence and use of textbooks. The initial IE research askedindividual students to simply report whether a textbook was "available" tochildren to help in their studies (used in India, Iran, and Chile). Other'

4o

studies have asked students whether and how many textbooks are owned (inChile and Brazil). This indicator is more relevant in Latin America wheretexts are often sold by private booksellers to parents and students. Inthe Uganda study, the researcher actually counted how many textbooks wereavailable and used within each classroom setting. No work has been foundon how often and within what context students read textbooks. We do notknow whether texts are read in class, at home, alone, or with otherstudents. The impact of textbooks may be understated if measurement erroris high. Distilling out disturbance caused by factors related to how andwith what frequency books are read may sharpen our understanding of howstrongly texts influence achievement. Future research might examine thisfiner-grain question. In addition, some uniformity in measures used byresearchers would aid future comparisons of findings across studies.

Desks in classrooms represent an additional input within theinstructional material category. All three analyses examining the effectof this concrete element of school quality have found significantachievement effects. For instance, a recent study of 324 sixth-gradestudents in Peru, discovered that the percent of children with desks atschool was more strongly related to reading achievement than was theinfluence of social class background (comparing t-statistics).Surprisingly little research has examined the effects of simple inputsrelated to opportunities to read and write inside the classroom (as well asoutside the school).

As with textbooks, the availability of a desk is easily measured andobserved achievement effects obviously are important. But we know littleabout how children's classroom time is structured, particularly how thematerial desk fits into opportunities to read and write. In some instancesthe desk may hold more utility as a symbol of constructing a "modernschool." Whether teachers structure lessons to encourage the actual use ofdesks is a separate issue. Encouragement of students to write may be moreimportant than reading from the standpoint of motivation. Writing is aproductive form of literacy. Rather than passively reading material,writing involves active creation of ideas and organization of information.But we have little understanding of how instructional materials can betterencourage writing activity.

The radio is another instructional device which has receivedconsiderable attention from researchers (Indicator 4). Initial productionof curriculum and classroom lessons is difficult and costly. Yet followingthe initial investment, delivery of radio instruction via simple receiverscan be efficient in terms of per pupil costs. The Nicaragua textbookprogram outlined above also included a radio instruction project. Theradio component was more effective than the textbook element in raisingachievement. Participating first-grade students received mathematicsinstruction over a radio for 20-30 minutes within their classrooms. Post-broadcast lessons were often given by the teacher, and worksheets wereprovided students to practice the material. At the end of the first grade,participating students scored 17 percent higher on the post-test comparedto control-group pupils. This gain exceeded one standard deviation on thepost-test [3.41.

A school library is another instructional resource which maysignificantly influence pupil achievement (Indicator 6). Significant

'1

student performance effects have been found in 15 of 18 analyses. The mostconsistent findings come from Latin America where multiple measures ofschool library utilization were used. Building on the early instrumentsfrom the Coleman and IEA projects, a survey of school quality was conductedin seven Latin American countries in 1975. The number of books on loanfrom a school library was significantly related to student achievementlevels within Argentina, Mexico, and Brazil. At the individual level ofanalysis, students reporting that they used the library more frequentlyperformed at higher levels (as always, controlling for pupils' familybackground). Other research indicates that the simple presence of a school-library is related to the school's average achievement level in ElSalvador, Botswana, and Uganda.

These findings come from one major analysis of the influence of schoolquality on science achievement (by Beyneman & Loxley). Within this work, aminimum criterion of statistical significance was utilized (p<.05) todetermine whether the school library measure was significant aftercontrolling for family background factors. The efficacy of the libraryvariables is confirmed by earlier findings from the IEA study withinThailand, Iran, Chile, and India. Yet neither project analyzed themagnitude of-this consistent effect of from the presence of a schoollibrary and utilization rates. The study of Malaysian primary schools.(6,056 students) did find significant effects of school library size(number of volumes) and academic achievement. However, the bivariatecorrelation between the two was small. And within multivariate analyses,the achievement effect of library size was significant yet much less strongthan social class control variables and availability of textbooks. InBotswana, a study of 869 standard seven students found statisticallysignificant achievement effects for how often the library was used(individual student report) and the number of new volumes purchased by thelibrary (school report). The strength of this relationship was comparableto the significant influence of social class background, appearing for bothreading and mathematics achievement.

The early IEA survey included questions which asked schools about theuse of classroom laboratories in the teaching of science (Indicator 7).This particular material input remains controversial. Governments anddevelopment agencies have invested sizable resources in building schoollaboratories and in financing the purchase of laboratory equipment.Whether this investment is paying off in higher levels of achievementremains an unanswered question. The early IEA survey asked schools aboutthe presence and utilization of classroom laboratories. These measureswere rather consistently related to higher achievement in three of the fourdeveloping countries included in the study: India, Thailand, and Iran. Twoindicators of laboratory use - number of students in laboratory classesand time spent in laboratory work - were related to achievement. Howeverthe subsequent Latin American survey used these same items and found nosignificant relationship with pupil performance [3.5]. No analyses werefound which examine the effectiveness of laboratories relative to otherelements of school quality. Nor do the earlier lEA findings mean thatlaboratories are cost-effective. The high cost of building, equipping, andmaintaining school laboratories may outweigh subsequent incremental gainsin achievement.

Finally, school feeding programs have received only slight attention

42

from researchers (Indicator 8). More is known about the effects of poorhealth, malnutrition, and hunger on children's school achievement than theameliorative effects of school feeding efforts. The effects of poor healthcan be dramatic. In a recent study of 3,699 primary school students anddropouts in Egypt, health status was one of the strongest predictors ofacademic achievement. For instance, children who had suffered from aserious illness in the preceding year scored 20 percent lower on theliteracy test and about 10 percent lower on the numeracy exam. Childrenwho had a major physical disability or suffered from malnutrition alsoachieved at significantly lower levels.

An earlier study in 61 Ugandan primary schools provided similarresults. Multiple measures were used, asking the children whether they hadsuffered from malnutrition, how often they had chills or a fever, how oftenthey had stayed in a clinic overnight, and whether they had seen blood intheir stools. Among the 1,907 surveyed, 75 percent reported havingsuffered from Malaria; 37 percent answered that they had stayed overnightin a clinic or hospital; 12 percent had seen blood in their stools; 5percent claimed they had been treated for malnutrition. This latter 5percent alone performed a quarter of a standard deviation below all otherstudents on a standard achievement test. Controlling for family backgroundand other school quality elements, pupil health (overall index) contributedsignificantly to academic achievement. Similar findings also are reportedfrom studies in Chile, Thailand, and a longitudinal investigation inGuatemala. The Guatemala work included study of 512 children from fourrural villages. Longitudinal achievement effects for diet were found aftercontrolling for earlier levels of health and nutritional intake. Yet verylittle empirical evidence substantiates the effects of child health andnutrition interventions on eventual academic performance.

Teacher Quality

At this point only material inputs to the schooling process have beendiscussed. But how are material resources managed by the teacher instructuring learning activities? One way to approach this question is tolook at the quality of teachers. Since most school quality research in theThird World has relied on large scale surveys, easily measured proxies ofteacher quality have been used most often. For instance, many studies haveasked about teachers' qualifications in terms of their total length ofschooling or their amount of postsecondary teacher training. -Whether andhow these proxies are related to the teacher's proficiency in organizinginstructional activities and in motivating children remains very cloudy.

Yet even rough measures of teacher quality are related to higherlevels of student achievement (Table 10). Findings are mixed on the effectof the length of primary and secondary achooling completed by teachers(Indicator 1). This factor has been studied in 25 analyses across manycountries. A significant relationship was found in 11 of these studies.The strength of this relationship is moderate in a few of these analyses.The IEA survey of literature achievement in Chile included 103 schools and1,311 students. A moderate correlation between teachers' school attainmentand pupil performance was observed (r-.34); the signficance of this factorremained when the factor was analyzed after controlling for studentbackground and other school quality elements. An early study of school

43

Tible 10Efficacy of School Quality Elemestsu

Teacher Quality

(l} (21 (3) (4) lSI (O 71School Quality ypothesis Cutry Achieveemt or Economic Direction of Effect Setting Studylndictor ISchool Level) Outcome Neasured t - I(ethod)

…-- ------------------- - - - -- --- ------------------------------------- …

1. Pre-srvice teKcher Ibe teacKhws years of India Scince Heyneman & LoSleytraiuig prisary and secondary IPrimary)

schoolieg Mill raise their Uganda +verbal .kills, boosting Bolivia tstudent achievenut Brail 4

lSecondary) 0 Norales I PinellsilesChile * Husen (MR)Paraguay 0 Heynnean I Louley(Primary)Neuieo 0Peru 0Colombia 0Argentina 0Thailand Language I math 0 Fuller & Chantavanich IIR)htsuama Reading 0 Husen (Cal

Reading & math * LoxleyNexico Language & math 0 lzquierdo I Guman ICO)42 LC's Drop-out rates 0 Levy lHR)

[Cross-national design)

The teacher s years of btsiaa Science t Heyneman L Loxleypost-secondary instruction (Primary)and teacher training India Reding & science t Reading teachers Cosber I reeveswill boast teaching skills, Chile * Reading teachersleading to higher student Thailand B Science teachersachievement Iran * Science teachers

Egypt Reading & math +Lituacy t Hartley I SwansonNumeracy 0

Negico Science 4 Heyneuan & Loxley(Prisary)(Secondary) 4 Klees (MR)

Argentina 4 Heynesan & LoxleyIPrioary) Readinq & oath 0 Echart et ii. IlR)lSecondary)Peru Science u Heyneuan & LoxleyIPrisary)Bolivia +Paraguay 4IPrisary) Reading o math + Rivarola & Corvalan (KRl(Secondaryl 0Brazil 4 Costa (KR)

Sciente 0 Heyneman & LouslyColombia 0

t Arriagada (19811 NV)Thailand Language & eath * Fuller & ChantavanichMalaysia Readin Ellwson lCS)

Comprehensive ne xa Rural schools NuaronO Urban schools

Uganda National exam 0 Heyneman l1i761 NR)O Somerset (CO)O Silvey (CO)

Sierra Le ne 0 Nindham ICO)6hana 4 Bibby & Peil (KRI(Secondary)Indonesia Arts * letters 4 Senbiring &

Scien 0 Livingstone

The teacher's total years Puerto Rico Reading * Carnoyof formal schooling *ill (Priaary)raise verbal skills and Iran Language I math 0 Ryanteaching effectiveness, Halaysia Comprehensive exa + Beeboutraising student (Secondary)achievement Congo Language & math + Youdi

Kenya National exam 0 Thias I CarnoyChile Language I math 0 Schiefelbein & Farrell

(1973)Tanzania Akadeic & vocational 0 For 3 of 4 Psacharopoulos &

tracks toxleyBrazil Individual income * Behrman & Birdsall IPF)(Primary/sec.1Uganda Comprehensive exam 0 Silvey (KR)(Secondary) 0 Somerset (KR)Sierra Leoe 0 Mindhas (CO)IPrimary)

Mlore highly schooled Brazil Students' mean yearS 4 Dehraan I Birdsallteachers vill boost lPrimary) of school completedparents' demand 4ormore schooling, andraise academic achievement

2. InserWVice teacher Upgrading the skills Egypt Litercy & nueracy H Iartley Swunsontraining of current teacking staff (Primary)

will raise the quality of Indonesia Arts & science 4 Seebiring I Livingstoneinstruction, leading to Languge * Nasoetion at *I. (NRIhigher student achievement - Botswana Reading & math 0 Loxley

* Husen

3. Teacher's length Teacers with longer El Salvadr Science Heyneman L Losleyof experience tenure develop stronger lPrisary)

instructional skitls, Iran raislig student achievment India Reading teachers

Chile t Reading teachersUganda National exis 0 Heyneman (1976)(PriurylChile Language I math 0 Schlefelbein & Farrell(Secondary) (1973)Egypt Literacy I nuseracy 0 Hartley I SNanson(Prisary)Puerto Rico Reading Carnoy(Primary)Puerto Rico 0(Secondary)Kenya National exam t Thias & Carnoyfalaysia Comprehensive exam + SeeboutlPrimary) 0 HaronTunisia Grade point average Carnoy & ThiasISecondary)Congo Language I math 0 YoudiColonbia Reading & science A Arriagada (19B1)Nexico Lasguage I math -MeesIndonesia Arts t science 0 Seebiring I

LivingstoneBolivia Reading 0 forales Pinellsiles(Prisary)Paraguay 0 Riivarola & CorvalinBrazil t Costargentina 0 Echart et al.(Prisary i sec.1

Botswana Reading L math * Veirs teaching LoxleyIPrimary) 0 Age

Teacher's verbal Teachers uith greater Uganda National uoal Heyneman & Jasison IMB)proficiency verbal skills increase IPrisary)

the quality of student Iran + Rural schools Ryanand teacher interactions,increasing achievement

i. Teacher's salary Higher paid teachers Peru Science 4 Heynesan Louxleyposseus higher language (Primary)and instructional skills, Bolivia sraising student achievesent Brazil 0

Paraguay 0Neuico 0Argentina 0Kenya National vxsa Thias & CarnoyColombia Heyneman & LoxleyColombia Acdemic & vocational Psacharopoulos I(Secondary) knowledge LoxleyTanzania Language 0

Vocational knowledge - For 2 of 3 tracksBolivia Reading L science + Urban schools horales & Pinellsiles

0 Rural schools

Teacher's Teachers vith more Argentina tience t Heyneman & Loxleysocial class highly educated parents IPrisary)

will possess greater Peru +language proficiency Bolivia +and increase achievement Paraguay +

Brazil tMexico 0Colombia 0Chile Verbal & math * Farrell & SchiefelbeinIran + RyanUganda National exam 0 Heyneman (19761

Teacher's Teachers displaying Thailand Language & math 0 Fuller & Chantavanichabsenteeism and greater puntuality (Primary)punctuality are more highly sotivated Mexico 0 lzquierdo & Suzuan

and effective

1. Full-time versus Full-time teachers will Indonesia Arts & letters 4 Seahiring &part-time teachers possess moe w *otivation (Secondary) Science 0 Livingstone

and teaching effectiveness

quality in Puerto Rico found that teachers' schooling level was moststrongly related to achievement of primary students and among pupils fromlower-income families. No significant relationship was found for secondaryschool students [3.6].

Teacher schooling effects have been negligible in several otherstudies. For example, a study of over 27,000 Thai primary school studentsfound a statistically significant, yet small association (r-.ll). This lowlevel of magnitude is troubling from a policy viewpoint. Allocatingadditional resources to increase teacher candidates' total length ofschooling may be an inefficient strategy for raising pupil achievement.One analysis based on IEA data for Chile and India found that moving theaverage length of teachers' schooling to the 75th percentile (of the rangeof teacher schooling previously found in these countries) would boostpupils' reading scores by 10 percent. But the cost of such an improvementwould be extraordinary [3.71.

In contrast, achievement effects are more consistent for teachers'length of tertiary schooling and number of teacher training courses. Theearly IEA survey included items on both of these areas; identical measureswere used in the subsequent Latin America survey. As seen in Table 10,either the original IEA research group or the later analysis by Heynemanand Loxley found significant effects from at least one of these factors in11 countries. Independent work in 10 other countries has revealedsignificant effects. In total, 21 of 30 studies have found a significantassociation between teachers' level of tertiary or teacher training andlater achievement of their students.

The magnitude of this relationship at times is moderate. One study of89 secondary schools in west Malaysia (7,674 pupils) found a notablecorrelation between length of teacher training and student achievement(r-.27). This factor retained its significance when entered into amultivariate model which controlled for the effects of pupils' familybackground and other elements of school quality. This research alsodiscovered an achievement effect for the interaction of teacher trainingand length of teaching experience. Similarly, the IEA study found a modestbivariate correlation for science achievement among 3,153 students from 124schools in India (r-.23). And the level of teachers' credentials was thestrongest predictor of student achievement within a sample of 463 primaryschool graduates in Uganda when tested against several other elements ofschool quality.

Very little evidence exists on the effectiveness of inservice teachertraining programs (Indicator 2). This scarcity of knowledge is in starkcontrast to the increasing level of resources invested in upgrading theskills of incumbent teachers. For instance, in the last decade two-thirdsof the World Bank's education projects have included inservice teachertraining components. Only four studies have examined the influence of suchefforts. The strongest effect was found in a survey of 124 Indonesiansecondary schools. Half the teaching staff in the arts stream had attendedinservice training courses. For science teachers, participation was evenhigher. This factor was the second strongest school quality predictor ofpupil achievement (bivariate correlations equalling .46 for arts and .35for science). Within multivariate models, controlling for students' familybackground and many other quality elements, the frequency of inservice

48

training remained as a moderate predictor of student achievement. Thisfactor was far stronger than the influence of pupils' family background[3.8].

Significant findings are reported in an earlier study in Indonesiawithin 40 fifth grade classrooms. The experimental design comparedclassroom groups which differed on textbook introduction and the level ofinservice teacher training. The training "treatment" occurred independentof the introduction of textbooks. The former factor was significantlyrelated to pupil achievement although the magnitude of the effect was weak,indicated in both the bivariate relationship (r-.15) and within amultivariate model. Inservice teacher training also was found to hold asignificant association with achievement in Botswana and Egypt. Much moreresearch is needed in this area, particularly as governments anddevelopment agencies invest a burgeoning amount of resources in trying toupgrade teachers' skills. The research to date finds that such initiativesdo influence subsequent pupil achievement. But even less is known aboutwhat types of inservice programs are more effective and more efficient.

Only a few studies have examined how a teacher's skills relate topupil achievement. Proxies such as the teacher's level of schooling orinservice training assume that additional competencies are obtained whichsubsequently increase pupils' achievement levels. More direct assessmentof actual skills which are clearly related to the teacher's effectivenessis very rare. The school quality literature from industrialized countries,for instance, emphasizes the strength of teachers' verbal competence inboosting pupil performance (summarized in the annex). Work in the ThirdWorld is beginning to examine these types of teacher attributes. Forinstance, the previously mentioned study of 1,907 primary school studentsin Uganda found consistent and strong achievement effects from teachers'English proficiency on achievement in both language and mathematicsachievement (Indicator 4). This factor influenced pupil performance morestrongly than students' social class background, preschool ability, and twoother school quality elements included in the model (comparing t-statistics). Teachers' achievement level (on a secondary school-leavingexam) was correlated with their pupils' performance in a study of 797second grade students in Iran. The bivariate association was weak for thestudent sample overall (ri.14). However, within a multivariate analysis ofstudents from rural and/or poor backgrounds, teachers' achievement levelmore stongly influenced their pupils' subsequent achievement.

The teacher's social class has been studied as a proxy for verbal andcognitive attributes which may influence achievement (Indicator 6). Theargument is simply that teachers coming from more highly schooled familieswill be more effective teachers. For example, teachers' own verbalproficiency may enrich the quality of verbal interaction and cognitivestimulation experienced by their students. Teachers' social class issignificantly associated with pupil achievement in 7 of 10 analyses. Sevenof these analyses come from one study. The Latin American school qualitysurvey included a single measure of social class background: level ofschooling attained by the teacher's father. This factor was consistentlyrelated to the performance of teachers' own pupils in Peru, Bolivia,Paraguay, Argentina, and Brazil. In addition, the Chilean study of 3,530grade 8 students found a significant, yet small, effect between teachers'social class and pupil achievement. The study of Iranian second grade

49

students mentioned above found a modest correlation between an indexmeasuring the quality of teachers' secondary schools and their pupilssubsequent performance (r-.l9). This factor remained significantlyinfluential when included in a multivariate model.

Two alleged indictors of teacher quality hold little consistentinfluence in boosting achievement. Teacher experience did appear to make asignificant difference in the early IEA survey, at least for readingteachers in Chile, India, and Iran (Indicator 3) [3.9]. However, worksince then has led to more skeptical findings. A recent study of 869students from 37 primary schools in Botswana found that the influence ofteachers' experience rivalled the influence of father's occupation (used asa background control variable). Interestingly, a second study fromMalaysia found that the length teachers' experience was associated withpupil achievement, but only among teachers receiving more preservicetraining courses. lievertheless, 13 of 23 studies looking at teacherexperience have found no significant achievement effect.

Similarly, teacher salary levels in general are not related to pupilperformance (Indicator 5). Two recent analyses from Tanzania (2,803secondary students) and Colombia (4,233 secondary students) found thathigher achieving vocational students actually were taught by lower paidteachers. The negative influence of teacher salaries in Colombia wasstrong, exceeding the achievement effect of students' social classbackground and other elements of school quality. In contrast, an earlyschool quality study of 115 secondary schools in Kenya found a significanteffect for higher teacher salaries (with just three other quality elementsin the model). As mentioned above, the researchers estimated the magnitudeof this relationship, then determined the link between higher pupilachievement and future earnings. Both relationships were stronger forgraduates from lower-income families. They estimated that additionalinvestment in raising teacher salaries would yield a seven percent annualrate of return to graduates from poor families [3.10]. However, the rateof return was less than one percent for graduates of average social classbackground, despite the statistical significance of the association betweenteacher salary and pupil achievement.

Overall only 4 of 13 analyses have found an association betweenteacher salary levels and pupil achievement. The cost implications ofincreasing teacher salaries in general and of paying higher wages toteachers with longer tenure are enormous in developing countries. Giventhe limited achievement effect of salary level and experience, budgetsavings could be generated in this area and redirected to more promisingelements of school quality. No detrimental effect on student achievementwould likely occur.

Teaching Practices and Classroom Organization

Despite the burgeoning literature witin industrialized countries onhow teachers manage instructional resources and organize their classrooms,very little work has occurred within developing countries. Table 11summarizing the findings of studies which have been conducted. The lengthof instruction stands out as a consist predictor of student achievement(Indicator 1). This element of school quality suffers from inconsistent

50

Tdble 11Efficacy of School Quality Elements:

Teaching Practices / Classroom Organization

(1I (21 13) (4) (5) 1h0 (7)School auality Hypotbesis Country Achievement or Economic Direction of Effect Setting StudyIndicator (School Level) Outcome Neasured * 0 - (Method)

1. Length of Nore hours or days of India cience f Heyneean & Laxtleyinstructional instruction will IPrisary)progras increase achievesent Iran +

Thailand fChile * Hours reading

science text* Schiefelbein L Clavel (INR

Peru Reading t Arriagada (19831Coloubia 4 Arriagada (19B11Colombia Vocational knowledge * Hours spent on Psacharopoulos I Loxley(Secondary) vocational

instructionTanzania Language & math t Clams time spent

(Acadeeic students only] on science &social studies

0 Clams time spenton language

(Yocational students only] * Number course periodsin vocational area

Brazil Cosprehensive exam u Rural schools Wolff(Prisary) 0 Urban schools

2. Hoiumork Assignoent and close lotsmana Heading r oath + Heyneaan & Laxleyevaluation of homework Chile Reading L science 4 Biology homework Tharndike (1R)will boost learning (Prisary)

Chile Language o math + Schiefelbein & Farrell(Secondary) (19731Tunisia Language 4 Urban students Simuons (KR)WPrimarylsec.)Iran Reading & science + Comber & KeevesIndi aThailand 0

3. Activn teaching Time spent explaining a Peru Reading & science 0 Arriagada (l?831and learning lesson by the teacher vill (Primary) hath 0roles in classrooss limit the student's oun

engagement in the uaterial,lowering achievesent

Classroom discussioas mill botsana Reading I oath 0 Loileyencourage active student lPrimary)learning nd raise ackievenent

4. High teacher leachers who expect Hong Kong Language I math Rome ICE)expectations fo high achievement receive lPrimary)student performance stronger comitmunt Uganda Cosprehensive uexa 4 Durojaiye (BR)

and performance from IPrimary h sec.1students Thailand Fuller & Chantavanich

IPrimary)

5. Teacher's tim Nore hours spent India Science 4 Heyneman & Loxleyspent on class preparing for class (Primary)

n preparation vill raise the quality Iranof instructios and boast Chile tachievement Thailand 0

Botswana Reading I math fHours marking papers Loxley

4. Teacher evaluations Nore frequent feedback to Colombia Science + Urban schools Arriagada 181)of pupil performance pupils and parents on

achievement levels Villencourage higher performance

definition and the use of varied measures, ranging from the number of daysin the school year to how many hours science is studied during the schoolweek. Yet in general, the length of instruction was significantly relatedto achievement in 11 of 13 analyses.

The length of instruction offered by schools is bounded, in part, byavailable material resources. Yet in many settings, the length of theschool day, time spent on particular curriculum areas, and the efficientuse of instructional time within classrooms is more strongly determined bymanagement practices than by material parameters. Classrooms varyenormously in the amount of time actually spent on instructional tasksrather than keeping order, checking each student's homework, or arranginglessons. More efficient use of classroom time is strongly related to pupilperformance within industrialized nations. Considerable progress on thispotential source of school efficiency could be made - by sharpeningclassroom management and teaching skills - within existing levels ofmaterial inputs.

In Brazil, the study of primary school achievement introduced above(1,006 pupils) included a question for teachers on the length of theirschool day. This simple measure was significantly related to achievementin rural, but not urban schools. The influence of this indicator wasmodest, though similar to the magnitude of pup.ils' family background.Research in Colombian primary schools (826 sixth grade pupils) found thatthe hours of class per year moderately predicted reading acbievement. Thisfactor's strength was comparable to the significant influence of twomeasures of students' social class background. This same measure ofinstructional time also significantly affected reading achievement in asubsequent study of 324 grade 6 pupils in Peru. And the recent studies ofsecondary school achievement in Tanzania (922 academic students) andColombia (4,233 pupils) found that'the number of class periods spent inacademic or vocational courses help predict performance on correspondingachievement exams. The strength of these instructional time measuresusually exceeded the influence of pupils' social class background and werecomparable in magnitude to other school factors.

Finally, Heyneman and Loxley's reanalysis of the IEA data on scienceperformance revealed consistent achievement effects for different measuresof instructional time. For instance, hours of instruction (per year) ingeneral science were significantly associated with achievement in India,Thailand, and Iran. Hours of instruction in reading also helped predictachievement in Chile and India. This study did not estimate the magnitudeof these instructional time effects; but the consistency of positivefindings across different indicators and countries is notable.

The assignment of homework - a second aspect of the organization ofinstruction - also shows promise in raising student achievement (Indicator2). The early IEA survey accounts for two analyses, where positivefindings were reported in Chile and Iran (but not for India and Thailand).The magnitude of effect was small, although multivariate modelssimultaneously tested for possible effects of many school factors. Onemodest study of 83 urban Tunisian students found that reported conditionsat home for studying were related to language achievement. This measure isnot an indicator of teaching practices. But the factor was stronger thanother school quality elements and equalled the magnitude of effects from

53

pupils' social class.

Additional mechanisms for encouraging higher student performance havereceived only slight attention from researchers working in developingcountries. Two studies have examined the extent to which active learningroles are created for students in classrooms (Indicator 3). Self-reportsby teachers of the amount of time they spent explaining academic materialto students had no relationship to reading or math achievement withinPeruvian primary schools. One recent study in Botswana (37 schools) foundno relationship between the frequency of classroom discussions reported byteachers and achievement on national reading and math exams.

In contrast, the level of teacher expectations for higher pupilperformance is related to their actual achievement in three independentstudies (Indicator 4). For instance, the large survey of 23,555 thirdgrade students in Thailand found a modest relationship between pupils'actual achievement and teachers' assessment of their "learning ability"(r-.22). This factor remained significant, though not strong, when enteredinto a multivariate model. Significant effects were also found byresearchers in Hong Kong (stratified sample of 100 low and high achievingstudents) and Uganda (540 pupils from standards I, II, and III). However,the magnitude of these effects was comparable to findings from the Thaistudy.

Finally, alternative indicators of how much time teachers spend inclass preparation have been used in two studies (Indicator 5). Forexample, analysis of the IEA data showed that the amount of time scienceteachers spent in preparing lessons (self-report) was significantly relatedto pupil achievement in Chile, Iran, and India, but not in Thailand. Threedifferent measures were used: hours spent preparing lessons in and out ofschool, and hours spent marking papers. The latter measure was also usedin the more recent Botswana study of 37 schools and standard 7 pupils.This factor was significantly related to both math and reading achievement.The magnitude of the effect was just statistically significant (p<.05).However, the full model included 10 student background variables and 17school factors 13.11].

School Nanagement and Structure

Beyond organization of the classroom, management of the entire schoolalso represents an important component of school quality. The school'sorganizational structure drives the efficiency with which inputs aremanaged efficiently. The first four sets of school quality factors --

expenditures per pupil, specific material inputs, teacher quality, andteaching practices - apply to various types of schools. Once a culturedecides to formalize the socialization and training of their children,these four sets of quality elements are important in improving varioustypes of schools. But this fifth school quality area - management andorganizational structure - raises issues which at times are pertinent toparticular kinds of schools. For instance, the structure of vocationalschools may differ substantially from basic education in the rural primaryschool. In addition, the social rules which comprise management practicesoften are tied to the local culture or grow from social norms within thegovernment sector. A hierarchical style of school management would be

54

viewed as desirable in some national contexts; a more participatory andprofessional school structure would be normative in other culturalsettings.

Organizational structure refers to two features of schools. First,schools vary in terms of their goals. For instance, different schoolsemphasize vocational, academic, or artistic instruction. Also schools varyin their social goals. Readmasters may emphasize tightly disciplinedclassrooms which stress obedience, or instead stress more active roles forstudents and individual initiative by both teachers and pupils. Second,the management practices of headmasters can vary enormously, at timesindependent of the school's official goals. Headmasters employ a varietyof means in supervising staff, in managing the school budget, in motivatingteachers to improve their practices, in working with parents, or indisciplining errant pupils.

The "quality" of school management can be conceptualized within aunilinear framevork. A prindipal aquires a variable level of skills andcredentials necessary in becoming a strong leader and supervisor. Or"aquality" of management can be view in a multidimensional way. Forexample, headmasters differ qualitatively on several aspects of schoolmanagement: (a) whether they enforce a hierarchical power structure orencourage participation of teachers in addressing problems; (b) thefrequency with which and how they evaluate teachers' performance inclassrooms; (c) the extent to which they prescribe curricula to teachers orencourage different approaches and professional judgement; and (d) theircompetence in budgeting and accounting for material inputs.

Research on management practices of headmasters is blossoming withinindustrialized nations [3.12]. Unfortunately, very little is known abouthow headmasters in the Third World act to improve a school's instructionalprogram. Table 12 summarizes the work that has occurred in developingcountries. This research has largely employed proxies which presumablyindicate the quality of a school's headmaster. For example, one recentstudy in Egypt of 60 primary schools found that students performed betterin schools with principals that had attended more training courses and hadlonger teaching experience prior to becoming a principal (Indicator 1).This finding also appeared for science achievement among primary andsecondary school students in Paraguay, coming from the Latin Americanschool quality project. In each case, the principal's level of trainingwas significantly related to achievement; the magnitude of these effectswas not reported.

Stronger findings come from the Indonesian study of 124 secondaryschools. Two characteristics of headmasters were among the strongest threeschool quality elements associated with pupil achievement levels. Thisinquiry found moderately high associations between pupil performance withboth the headmaster's salary level (r-.50) and the headmaster's length ofteaching experience (r-.33). Both factors remained strongly associatedwith achievement within a large model which estimated achievement from 13student background factors and 17 other elements of school quality. Themagnitude of these headmaster effects exceeded the significant influence ofmost student background factors (comparing t-statistics). The analysisfailed, however, to explore whether these headmaster characteristics wereacting as proxies for other aspects of school quality. The headmaster's

55

Table 12Efficacy of School luality Elements:

khool Nanagement and Structure

ll) 12) 13) 10) IS) l&) (7)School Quality Hypothesis Coumtry Achieveent or Economic Direction of Effet Settian StudyIndicator ISchool LeNvI) Outcome Neasured * 0 - (Nethod)

1. Duality of The primcipal's length Egypt Science Heyneman I Loxleybeadeaster or of formal training vill IPriury)principal boost mangesent and Paraguay Reading I science +

instructlon-related skills, Bolivia Urban schools Norales I PinellsilesIndirectly increasing (Prisary) 0aclievleenat ISertafy) O Ml schools

Botswana Reading I math 0 leadmasters LoxilyIPrimary) experience

Higher salaries will attract Indonesia Reading I mth I Sembiring Istronger principals, improve iSecmndary) Livingstom 1IR)the ismtructional proirm,and raise achivevent

2. Number of Norm tha one shift of Egypt Literacy I muneracy 0 Hartley I Swansonclass shifts Classes eacb day will strais (Primary)

the effectiveness of falaysia Comprehensive exua + eeboutresources and lower (Secondary)achilevemt Chile Language I math 0 Schiefelbein I Farrell

(Primary & seK.) (19131

3. Student boarding Living at the school Kenya Cambridge esua * hias I Carnoywill raise student (Secondirylmotivation and achievement unisia 6rade point average Carnoy & Thias

Congo Language & math t Youdiflalaysim Comprehensive exam 0 Beebout

4. Student Holding low achieving Tunisia 6rade point average Carnoy & Thiasrepetition students at a grade ISecondary)

level will boist Egypt Reading & math exam 0 Hartley I Swansonacademic performance IPrimary) (Lowering) dropout rate

Thailand Academic exm * Fuller & ChantavanichBrazil 0 Costa

1. School sonitoring Nore frequent sonitoring Botsuama Nith * Loxleyby the education sinistry IPrimary) Reading 0will improve instructionand raise achievesent

1. Student achieVelent Natiomal exus will Kenya Comprehesive exa In affluent schools Somerset (CB)information isprove school management, (Primary/c.) 0 In poorer schools

raising achievement

1. Akademic Tracking students into Chile Level of first Job 4 Schiefelbein I Farrellversus vocational vocational curricula, ISecondary) (1994)curriculua where academic achievement Colombia kcdemic exam 0 For 3 of 4 Psacharopoulos & Loxley

is not emphasized, vill vocational tracks (NIRIlower student achievement Vocational exam - For 4 of 4

vocational tracksTanzania Language I math For 3 of 3

vocational tracks

Vocational curriculum Colombia Earnings (short-run) 0 Psacharopoulos Ixi II prepare youth ore lSecondary) labalza (CS)effectively for the labor Colombia Rate of return (income) t Compared to Psacharopoulos lCOJC6)force and mill increase academic trackearnings Philippines

TurkeyThailandIndonesia - ClarkCross-national National iuoee 0 Over 1950-1970 Benavot (Il)R

S. School versus In-plant training is South Korea Individual income t Lee (CS)in-plant vocational more cost-effective (Secondary)training

------ ----- ------ --- - - ------ ----- -----_----- - ---___---- - - -- - - --- --_----- ----- ------ ----- ------ ----- ------ ---- - ----__----- --____-----_----

salary level holds no logical relationship to pupil achievement. Yetsalary could be acting as a proxy for length of training or experience.This work does encourage deeper investigations into an important issue:What specific actions do headmasters engage in which indirectly affectstudent achievement?

In Bolivia, a study of 53 primary and secondary schools found arelationship between student achievement and the headmaster's length oftertiary schooling. The magnitude of this factor was comparable to theinfluence of expenditures per pupil in a simple model which included justseven other school quality factors. Each of these two significantpredictors of achievement was stronger than the student's social classbackground (when comparing t-statistics). However, these results heldonly for urban, not rural schools. And the recent study in Botswanamentioned above found no effect from the the headmaster's years ofexperience among 37 schools enrolling grade 7 students. (Note, however,that values from only 37 cases were assigned to 869 pupils, constrainingthe variance in this headmaster-experience measure.)

Research has looked sporadically at other areas of school structure,including the number of class shifts each day and student boardingpractices. An insufficient number of studies have occurred from which togeneralize. However, two of three studies found no detrimental effect onachievement from the number of class shifts which operated each day(Indicator 2). In Malaysia (89 secondary schools), a negative achievementeffect was observed where schools bad two sessions of classes each day.However, the marginal achievement return from investing resources ineliminating double sessions was moderate. The researcher estimated that a$100 investment toward reducing double sessions would yield a seven percentincrease in language achievement. The analysis does not clarify whetherthis level of investment was affordable or even sufficient to eliminatedouble-sessions. This level of marginal effectiveness was higher than thebenefits gained from investing in lower class sizes, but would yield alower return than directing the same resources to improve teachertraining.

Another area of school structure - repetition of grade levels by lowachieving students - also holds significant cost implications (Indicator4). Four analyses have examined whether repeting a grade improvesperformance in the long run. This is a difficult issue to model and studyempirically. Yet positive effects been observed in just one study. Therecent well designed study of 8,570 primary students in Egypt found noeffect from grade repetition on pupils' academic achievement. Nor didrepetition lower the probability of eventually dropping out of school,holding constant pupils' skill level.

The relative effectiveness of vocational versus academic curriculum isone piece of school structure which has generated considerable research.Table 12 includes a summary of this work (Indicators 7 and 8). Thisingredient of school structure also has enormous cost implications. ThirdWorld governments and development agencies continue to invest in technicaltraining schools and in vocational components of comprehensive (ordiversified) secondary schools. However, earlier optimism that vocationaltraining would raise student motivation, school achievement, and subsequentearnings has not be substantiated by the research. This element of school

58

structure relates to specific types of institutions. Since the presentreview focuses on elements of school quality which are present acrossvarious kinds of schools, the vocational issue will not be discussed indetail. For readers curious about this area, two recent publicationsreview work which has been done this question [3.131.

A Narrow View of School Quality?

As a body of work, this collection of school quality studies isimpressive. Over the past 15 years much has been learned about theinfluence of school quality in the Third World The research hassubstantiated the importance of school quality in shaping achievement, netthe influence of pupils' family background and community context. Inaddition, this literature has begun to delineate which specific schoolcharacteristics efficaciously boost student achievement.

But we should step back from this body of research and question itsbasic premises and weaknesses. A critical analysis is useful beforeapplying research findings to investment strategies or local effortsdesigned to improve school quality. Questioning the research conducted todate may also prompt more refined investigations.

Here we summarize two sets of limitations. The first set comprisetechnical problems with how the basic production-function has been appliedto output of learning. The second set raises more fundamental questions asto whether this model, borrowed from the process of materialtransformation, is even appropriate in trying to understand the socialprocesses of teaching and learning.

Refining Production Functions. Early in this chapter the schoolquality literature was introduced by placing it in an historical context.Large scale school surveys were undertaken in developing countries toanswer the basic question which still troubles industrialized societies: Doschools make a difference in shaping children's school achievement andeventual occupational attainment? The positive finding for developingcountries importantly contradicts more limited effects withinindustrialized nations. But the methods employed to respond to this initialpolicy question are not always well suited to answer the second generationissue: Which specific elements of school quality effectively boost studentachievement? This history constrains how we represent the process ofteaching and learning within formal models and empirical studies. [Inaddition to the following discussion, the limitations of productionfunction studies in education have been explored elsewhere, see Note 3.14.]

First, large scale surveys of easily countable school characteristicshelp in answering the original aggregate question, Do schools make adifference? But by including a grand number of school variables, thebehavior and influence of a few characteristics are more difficult tostudy. The true influence of a particular school attribute may not beobserved, given the simultaneous effects of other school factors. Forinstance, the study of Indonesia secondary schools found an overwhelmingeffect of the principal's level of training. But the functions typicallyincluded 45 different school characteristics. Thus the influence of otherschool characteristics were eclipsed by this strong proxy of principal

59

quality.

This suggests that production-function studies may be more useful indisconfirming that a certain school characteristic influences achievement,rather than identifying which elements are efficacious. Where productionfunction studies include a large number of variables, the least ambiguousproxies of quality may show the most robust statistical findings. Thisoccurs not because the school factor exerts the strongest influence. Butthis input may be more easy to measure. For instance, the teacher's levelof training may simply be a proxy for verbal skill in the classroom. Yetquickly measuring the first is possible on a survey form. The secondconstruct is more problematic to assess. On the other hand, the fact thatclass size (a discrete variable) rarely is related to student achievementis useful knowledge. In this way, the process of disconfirmation is easierwithin large scale production-function studies.

Second, large scale surveys also encourage the measurement of materialaspects of schools. If their is pressure to gather data on a great numberof variables from a large number of schools, material features can be morequickly counted. The assessment of underlying social and managementprocesses is more problematic, including how teachers use textbooks,laboratories, or writing materials.

The recent experimental school quality studies in Nicaragua and thePhilippines go very far in addressing weaknesses of production-functionstudies mentioned thus far. By focusing on a small number of schoolcharacteristics the relative magnitude of achievement effects can be moreclearly determined. By looking just at the influence of additionaltextbooks and the introduction of radio instruction within specificschools, the research design controlled for the intervening influence ofother school quality elements (assuming control-group schools were wellmatched). Concentrating only on these two quality inputs, the researchgroup could then estimate the relative magnitude of discovered gains inachievement. Experimental designs also minimize the danger of inferringthe wrong direction of causality which is present in correlational studies.It may be that more competent students attend higher quality schools,rather than higher quality schools actually increasing pupils' achievement.Cross-sectional data (from one point in time) can not fully rule out thefirst direction of causality). But with an experimental design, assessmentof the school quality effect follows introduction of a discreteintervention within specific schools. Longitudinal surveys within theproduction-function framework would at least strengthen the case for thepostulated direction of causality.

A third limitation of production-functions employed to date is theimplicit assumption that each school characteristic behaves independentlyin shaping pupil achievement. But the interaction of two or more schoolcharacteristics is clear to any student of education. For instance, theinfluence of textbooks is bounded by the skills of teachers in using thesematerials. Interaction terms are commonly included within the production-function frame in other fields. This would be one step toward morerealistic representations of the instructional process. Identifying themost efficient mix among several essential school characteristics may bemore important than determining the independent influence of any onefactor.

60

Fourth, typical models assume that school characteristics influenceachievement in a linear fashion and that each can be substituted foranother. Researchers often infer that a positive effect of, for instancelengthening the school day, is generalizable to all conditions. However,achievement gains would be constrained if teachers were not competent ineffectively using additional instructional time. Investing more resourcesto lengthen the school day may have no influence on achievement untilteachers' skills are upgraded. Here the influence of one schoolcharacteristic depends upon a threshold level of a second factor.

Production-function analysis could be improved to better capture therelative advantages of input mixes. In modeling agricultural production,for example, the relative efficiency of various input mixes (land, labor,and capital) often are compared in determining productivity advantages.Such work examines not only the magnitude of an input's effect onagricultural product but also its cost. In addition, farmers vary on theircapacity to manage constant mixes of inputs. This resembles the managementand organizational issues addressed below with regard to schools. The mainpoint here: Education production-functions could look more carefully at theadvantages of different mixes of inputs [3.15].

Fifth, researchers should be more careful in specifying the conditionsunder which findings apply. Large scale surveys invite inferences thatcertain school characteristics influence achievement across all settings.Little work has clarified under what specific conditions school factors doand do not shape pupil performance. An important exception to this generalveakness: School quality elements (relative to family background) appear toexert a stronger influence in rural areas and among lower achievingstudents. Within relatively affluent urban areas, school quality effectsdiminish somewhat, though by no means entirely. This specification of thesituation within which the finding applies is important. Yet beyond thisconsistently discovered conditionality, few studies systematically examinehow quality effects may differ by grade level, academic area of study, orfor different ethnic groups.

Finally, the traditional production-function literature rarelyidentifies the costs of various elements of school quality. This isparticularly troublesome given policymakers need to determined whichelements of quality effectively boost achievement and which are cost-effective. Again, the historical priority placed on establishing thatschools do or do not have an effect has eclipsed concern over cost-effectiveness. Making schools more effective is somewhat independent ofmaking schools more efficient. Both goals require establishing whichschool factors increase student achievment. But once efficacious elementsof school quality have been identified, analysis of their relative costs isnecessary. Chapter 4 will address the question of how school qualityresearch can better look at this issue of cost-effectiveness or efficiency.

Abandoning material production-functions. The literature'shistorical context has led to a second, more serious, set of constraints.The production-function metaphor has encouraged study of material inputswhich flow into the school rather than focusing on variability in thesocial organization of these inputs and in human behaviors which enhance orhinder learning. Within the production-function tradition the teaching-

61

learning process is defined simply as a system which combines materialinputs to produce achievement. From this viewpoint, the production-function approach has ignored non-material behavioral aspects of schoolquality which can not be measured through large-scale surveys -- especiallyhow teachers and beadmasters manage and organize material characteristics.This point of criticism suggests that the production-function framework beabandoned, replaced by new methods and a fresh set of school qualityelements which may be more strongly linked to student achievement.

Rather than viewing the school as a firm which transforms materialinputs, the school can also be viewed as an organization which managesmaterial inputs with variable levels of efficiency. That is, the ways inwhich the headmaster manages school staff and how teachers arrangeinstruction significantly determine levels of student achievement.Material inputs clearly contribute to the quality of instruction andmanagement of the school. But the emphasis here is on pin-pointingelements of the school's social structure which strongly influenceachievement. For instance, research within industrialized countriesconsistently demonstrates that children learn more when more class time isspent on actual engagement in instructional tasks. In contrast, studentsoften spend time talking to friends, idly waiting for the teacher to getorganized, and sitting passively while the teacher checks the work of otherchildren. Similarly, headmasters may be craftful in evaluating teachers'performance, in stimulating improvements, and in creating a sense ofcamaraderie among staff and students. Or the headmaster may remain in hisor her office and attend to normal administrative tasks. But theseelements of school quality are social in nature, not material. Materialresources provide the instruments and simple technology necessary foreffective instruction. But the use and management of these material inputsoccurs through social practices.

This second line of research has been more fruitful in explainingstudent achievement within industrialized nations than has the study ofmaterial determinants. Several reviews of school quality research withinindustrialized countries have been conducted and are summarized in theannex. In general these reviews have found only a slight relationshipbetween global measures of material inputs of school quality (for example,expenditures per pupil) and achievement, after controlling for theinfluence of community wealth and pupil background. Exceptions do arise.For instance, several studies in the U.S. found that the length of theschool year does affect achievement. But in general, the level of materialinputs fails to significantly boost pupil performance within industrializedsettings.

School quality effects within industrialized countries are more robustwhen looking at social factors, including teacher quality, classroomstructure, teaching practices, and school management. For example,empirical research frequently finds achievement effects stemming from theteacher's verbal ability, teacher experience, and level of teachertraining. In developing countries, we have seen that material resourcemake a greater difference in raising achievement. But even in Third Worldcountries, it appears that material inputs must be directly related to theinstructional process to wield influence on pupil performance.

Regretably, little research on the management and social organization

62

of instruction is being conducted in the Third Wiorld. Intellectual andresource constraints make surveys of material school features less costly(in the short run). Research on management and social factors - arelatively new field within industrialized countries - is slow toinfluence the older line of thinking. Third World research continues tofollow the production-function form of investigation, even though thisapproach has been largely abandoned among investigators in industrializedcountries.

Historical emphasis on material indicators of school quality mayinadvertantly discourage deeper examination of classroom practices. Theschool effects literature within industrialized countries currently focuseson two promising areas: (a) behaviors between teacher and student which arerelated to achievement gains, and (b) effective ways of organizingclassrooms which encourage more active learning roles for students. Forinstance, the process of motivating students involves formulation by theteacher of concrete expectations for pupil performance, opportunities forthe pupil to attack specific questions, and adequate evaluation of thestudent's performance by the teacher. Clearly material inputs (such as abook or writing pad) play a sizable role in Third World schools. But thestructure of social action between teacher and student - and mobilizationof these material inputs - exerts an independent force. And themanagement of material inputs can not be studied through large scalesurveys.

A recent review summarizes several classroom-level studies whichattempt to relate teaching practices and student achievement in developingcountries. This work is still in an embryonic stage, limited bymethodological problems and unconvincing evidence. Yet these early studiesdo enumerate several characteristics of teachers and ways of organizingclassrooms which may help improve learning (Avalos & Haddad, Note 3.16).

Teacher's attributes and skills1. Academic and intellectual proficiency2. Creativity and inventiveness3. Internal motivation for teaching4. Participation in inservice teacher training5. Teacher holds high expectations for student performance6. Knowledge of subject matter7. Teachers' beliefs about the purpose and utility of schooling

Classroom social structure and management8. Teacher lectures at the class versus encourages questions and

discussion with students9. Learning occurs within cooperative groups versus individual

study and competitive achievement10. Lessons emphasize problem-solving skills versus subject matter11. Teacher praises student achievement12. School climate or organizational norms which encourage effort,

trust, and mutual respect among students and teachers

Unfortunately, research on these non-material, social areas of schoolquality is just beginning. But attention to a subset of these factors andcareful empirical research would be very helpful. Deliberate observations

63

of teachers and classroom organizations are necessary. Considerableprogress within industrialized countries has occurred in in defining andmeasuring these social factors with sufficient validity and reliability.Hard evidence is abundant on several of these social elements of schoolquality within industrialized countries. But the traditional production-function method must be adapted in studying these factors. A narrower anddeeper focus on, for example, 40 varied classrooms would provide a wealthof data on the relative effectiveness of different teaching practices.This approach would move us far beyond looking only at proxies of quality;instead evidence would accumulate on specific behaviors of teachers andconcrete ways of organizing classrooms which are empirically linked tohigher student achievement [3.17].

Finally, materialist assumptions underlying the production-functionframework ignore the symbolic dynamics of school quality. The body ofresearch reviewed in this chapter stands upon the rationalist assumptionthat policymakers want to identify those elements of quality which increasestudent achievement. But attention to (or neglect of) improving schoolquality occurs within institutional contexts, be they governments ordevelopment agencies. And these institutions must necessarily be concernedwith both the visibility and the perceived legitimacy of their actions.Early in this paper, the point is made that school expansion ofternreceives more attention than improving quality. This may stem from theconcrete and salient signals provided by the former. The sight of a newschool building in a village or urban neighborhood is a powerful signal of"modernization" and change. Simply sending new textbooks into an oldschool is not as effective in symbolic and political terms.

Little thinking has occurred on how the project of improving schoolquality can incorporate more vivid signals of progress and come to be seenas a more legitimate initiative. The lack of political will, in part,stems from the lack of clear, effective signals that quality is importantand possible within the context of national development. Reform ofnational examinations may be one device for making the issue of schoolquality more visible and for providing benchmarks for progress. China,Malawi, and Kenya recently have been considering improvements in theirexamination systems. Exams could become useful tools for mapping changesin student achievement, and even discovering what school factors appear toboost pupil performance. More concrete and public recognition of excellentteachers may be another way of building popular support for the improvementschool quality. Importantly, the task here centers on understanding howalternative elements of schools can send crisp signals, not which factorsmaterially influence student achievement.

SuDary

Two lessons from this review of empirical research should beemphasized. First, these 72 empirical studies have advanced our knowledgeenormously on the question of which material school inputs do and do notraise student achievement. Investments on ineffective inputs by ThirdWorld governments and development agencies continue to be sizable. Thecorresponding loss in efficiency and waste of scarce educational resourcesis equally large. On the other hand, we know much about what works inschools. In particular, those material inputs that are closely linked to

64

the instructional process consistently influence achievement.

Second, policymakers and researchers are becoming more clear on whatwe do not know. Among those school inputs which do shape achievement, weunderstand little about their relative costs and the magnitude ofachievement benefits which they yield (that is, their relative efficiency).Also several studies show that the influence of school factors is strongestfor children in the poorest countries and/or from lower-income backgrounds.But beyond this important condition, we know very little about the settingswithin which elements of school quality are more or less potent. Mostimportantly, we are largely ignorant of how headmasters and teachers managematerial inputs. Differing forms of management and variation in classroomorganization are apparent to most observers of Third World schools. Butlittle empirical research has tried to relate these differing ways oforganizing material inputs with levels of student achievement. Here workoccurring within industrialized countries may contribute much to oureventual understanding of the organization of schools within developingcountries.

65

Chapter 4

What School CharacteristicsBoost Efficiency?

Introduction

Research reviewed in Chapter 3 identifies the relative effectivenessof various elements of school quality. That review sorted out those schoolfactors which help raise student achievement from those which do not. Butamong this former subset of efficacious school factors, very little isknown about the efficiency with which these inputs raise studentachievement. Wise investment decisions rest not only on understandingwhich school characteristics effectively boost achievement. In addition,understanding both the magnitude of achievement effects and the cost ofalternative school characteristics could help boost the efficiency withwhich scarce school resources are allocated. This chapter reviews the fewefficiency studies which have been conducted.

More efficient schools provide higher levels of achievement atconstant, or ideally at lower, material cost. One way of achieving greaterefficiency is to invest in those material elements of school quality whichare most cost-effective or efficient. This requires choosing among variouspossible school inputs and practices. For instance, we now know with somecertainty that textbooks and school libraries influence studentachievement. Determination of which is the more efficient strategy forboosting achievement requires two additional pieces of information. First,we need to know how strongly textbooks versus school libraries influencepupil achievement. That is, the uagnitude of each intervention's effectshould be determined. Second, the cost of each strategy must bedetermined. Together these two variables - magnitude and cost - indicatethe relative cost-effectiveness of investing in textbooks versus schoollibraries.

Building a school library, for instance, might boost studentachievement 10 percent higher than would a doubling of available textbooks.Yet the former strategy may cost 50 percent more than the latter.Directing resources to textbooks, therefore, would be more cost-effectivein raising achievement (per constant unit of investment).

Investing in those school factors which rarely correlate with higherachievement is very inefficient. In fact, resources allocated toineffective material inputs or tanagement practices could be reduced. Thiswould generate cost-savings which could be redirected to more effectiveelements of school quality. Mentioned above, the research literature ismost useful in identifying those school factors which hold no consistentinfluence on achievement, not necessarily in identifying what works. Forinstance, the consistent *finding that (within reasonable limits) increasingclass size does not diminish achievement is very important from an

66

efficiency standpoint. The cost-savings of even incremental increases inclass size can be substantial.

This chapter speaks primarily to the issue of internal efficiency.The policy issue here: Given a limited level of material and humanresources, can the school attain higher levels of student achievement? Asecond aspect of school efficiency deals with the relationship between thecost of schooling overall and students' subsequent streams of income,Schooling costs and eventual income benefits floving to graduates varygreatly for different types of schools (for example, primary or secondary,vocational or academic). Therefore, schools differ on their externalefficiency as well as their internal efficiency. A few studies reviewedbelow look at the relationship between school quality and eventual incomeeffects experienced by graduates. This work crosses over the two spheresof school efficiency.

Sources of Internal Efficiency

The initial efficiency studies summarized here are just beginning toaddress the weaknesses of the production-function metaphor and model. Thiswork necessarily (a) identifies the magnitude of the influence of a schoolquality factor on achievement, and (b) analyzes the cost of this elementrelative to the magnitude of its influence on achievement. However,weaknesses remain. Technical efficiency depends not only on selecting theright inputs and management practices; care must also be taken in creatingefficient mixes of ingredients, as previously discussed.

Most troubling, the area of internal efficiency - by emphasizingidentification of costs - ignores management practices and teachingbehaviors for which prices are difficult to determine. For instance, whena headmaster successfully moves teachers to improve their method ofinstruction, student achievement gains may be significant. Butconventional surveys neither capture this organizational behavior, nor cancosts be readily assigned. A major challenge is to integrate conventionalefficiency analysis with the study of management practices and socialorganization in schools. Otherwise, the study of internal efficiency willcontinue to focus only on material inputs.

Only a few studies have been completed which examine the internalefficiency of alternative school inputs or management practices. Thisresearch falls into three methods of analysis.

Method 1: General Comparison of Cost and Benefit

General comparisons between the relative costs and achievementbenefits of school quality inputs can be quite revealing. For instance,one cost study in Bolivia found that achievement levels were no higher inschools with lower class size. This finding stems from a comparison of arandom sample of 63 urban and rural schools (29 private and 34 publicinstitutions). This does not rule out the possibility that the independentforce-of lower class size acts to improve achievement. School quality.factors associated with higher class size could be exerting an offsettinginfluence to increase student performance (for instance, a critical

67

threshold level of textbook availability). But this lack of relationshipbetween class size and achievement suggests that the allocation ofadditional resources to lower class size would not raise efficiency [4.1).

The Bolivian study also found that unit costs were 10 percent lover inprivate schools compared to pubic schools in both the fourth and twelfthgrades. Reading scores among private school students were 40 percenthigher in grade 4 and 30 percent higher in grade 12, compared to publicschool pupils. The difference in unit costs was more strongly linked tohigher class size in private schools, not lower teacher salaries. Inaddition, the investigators found that the size of school enrollment heldno effect on unit costs. Units cost were lower in schools with betterqualified principals. More sophisticated models which look at theindependent influence of various factors would be necessary to estimateunit costs. But these straightforward comparisons strongly suggest wherecost-savings can be obtained and how some schools could become moreefficient.

Simple evidence on the cost of school quality elements can also pointto possible efficiencies. For instance, a study of (preservice) teachertraining costs was conducted in Pakistan. The analysis included cost datafrom six of Karachi's colleges involved in teacher training and a survey of400 recent graduates. Two findings are especially relevant. First, theunit cost of training one new teacher varied by a factor of six among thecolleges. Cost accounting in education is not a precise science, nor aredata reported on the graduates' average achievement level across theseinstitutions (necessary in determining cost-benefit). Yet this widevariation in costs does suggest that some colleges are operating moreefficiently than others.

Second, only 40 percent of the graduates were actually employed asteachers 13 months following their graduation from the training colleges.Over 55 percent were not working in any job. Despite a documented teachershortage in this region of Pakistan, particularly in rural areas, manygraduates were not teaching. Only 31 percent reported that they would bewilling to teach in a rural area. Employment as a teacher was unrelated totheir area of specialization in their training programs (for instance,mathematics, science, or language instruction).

These findings suggest a low overall level of efficiency in teachertraining programs. The variation in unit costs also suggests that greaterefficiencies are feasible within particular programs. Again, more carefulanalysis is necessary to see how teacher quality may covary with highertraining costs. But this simple analysis of costs points to unambiguousinefficiencies [4.2].

The low efficiency of some teacher training programs was also revealedwithin the early Chilean study of school quality (grade 8 students within439 classrooms) [4.3]. Whether teachers were trained in the university ornormal schools was unrelated to student achievement. Yet university-basedtraining was much more costly, in terms of direct cost of instruction andeventual salary levels. Based on actual cost data, the researchersestimated if only normal school graduates had been hired as teachers in theprevious five years, a cost-savings equal to two percent of the nationaleducation budget would have been generated. In addition, the researchers

68

estimated that a 15 percent increase in average class size would have nodiminishing effect on pupil achievement - and would free resources equalto five percent of the education budget. Moving to double shifts inschools, also unrelated to achievement levels, would create a cost-savingsof an addition six percent of the sector's budget.*

Simple comparisons of costs and achievement benefits at times buildfrom disconfirmation that a specific school factor has an effect onachievement. For example, the Bolivian analysis observed no associationbetween higher unit costs and reading performance between public andprivate schools. The researchers then determined that variation in unitcosts was better explained by the pupil/teacher ratio than by variation insalaries. The school quality literature already has found that class sizemakes little difference in shaping achievement levels. The advantage ofboth the Bolivian and Chilean studies is that they focus on the costaspects of this ineffective element of school quality.

Method 2: Analysis of Discrete Marginal Cost and Product

More precise methods have been employed to estimate the cost-benefitof specific school quality elements. This approach has two advantages overgeneral cost and benefit comparisons. First, the independent effect of aquality element is isolated either through an experimental design where anintervention is applied to a treatment group; the magnitude of the effectis then compared to the control group of classrooms or schools.Second, the cost of alternative quality improvements can be calibratedagainst a marginal increment (say a US$ 100 investment in quality) inalternative school inputs (for instance, textbooks or lengthening theschool day). The size of marginal products (or achievement benefits) canthen be compared with the alternative investments, holding the marginalcost increment constant.

Experimental studies. The Nicaragua textbook and radio experimentdiscussed in Chapter 3 also included a cost-benefit analysis. Here thebenefit or outcome variable utilized was the first grade passage rate amongpupils receiving radio lessons in school versus students in the controlclassrooms. Among control-group students, an estimated 45 percent failedor dropped out before completing the first grade, versus 33 percent ofstudents in the radio mathematics program (Table 14). The cost-savings(benefit) resulting from this lower failure rate was calculated. Theoverall cost per primary school student in Nicaragua at that time was$54.00. The cost per first grade graduate, including resources expended onpupils who failed or dropped out, equalled $99.50. By increasing the gradepassage rate, the cost per graduate in the radio mathematics programapproximated $12.16 less than in the control classrooms. The per studentcost of the intervention was $3.00 per student.

69

Table 14Nicaragua Radio Efficiency Study

Nicaraguan ClassroomsWith Radio Without RadioInstruction Instruction (controls)

1. Mean math test 52.5 37.6score

2. Percent children 33% 45%failing grade 1

3. Cost per graduate $87.34 $99.50from grade 1

Source: Jamison (1978); Jamison, Searle, Galda & Heyneman (1981)

Relative costs can also be calculated backwards, holding constant theincrement of benefit (or achievement gain) which is observed. For example,students who received textbooks within the Nicaragua mathematics projectscored .22 standard deviations higher than pupils in the control groups ontests given following the intervention. The cost per student of thetextbook component of the project equalled $2.00. This level ofachievement gain was then calibrated against the magnitude of achievementgain which is observed when class sizes are lowered (based on U.S.literature). To accomplish this same magnitude of achievement gain,average class size must be lowered from about 40 to 10 students. Clearlythe cost of this change would be enormous.

Mulivariate survey studies. Experimental studies are rare in schoolquality research that is conducted in developing countries. The largescale survey, as discussed above, is more commonly used. The marginalproduct of incremental investments in elements of quality can be estimatedfrom empirical models which attempt to report the independent influence ofseveral factors. Based on a conventional production-function analysis(using a Cobb-Douglas model), the Malaysian study of 89 schools estimatedthe marginal gain in achievement associated with a constant level of newinvestment 14.4]. The researcher determined the cost of each element ofschool quality which significantly influenced achievement: level of teachertraining, double sessions of classes, and class size.

The marginal gain in language test scores was then estimated for each$100 increase in investment. Table 15 illustrates the various levels ofeffect found for this constant cost increment.

The production-function transformation provides informatioi on themagnitude between a change in the school quality factor and the achievementeffect (a one percent change in the independent factor leading to an X%change in achievement, depending upon the observed elasticity). The onepercent change in the school quality element can be costed out. Whencalibrated against the $100 increment in investment, the size of eachachievement effect can then be estimated.

70

Table 15Malaysia Efficiency Study

School Quality Percent Gain in LanguageElement Test Score Per $100 Investment

1. Teacher training 8%

2. Reduce use of double 8%session

3. Reduce class size 3%

Sources: Beebout (1972), Alexander & Simmons (1975)

This form of analysis first provides information on how strongly thequality improvement will boost achievement. In the Malaysian case, a $100increase in cost is very sizable given the moderate achievement effectobserved. Second, the analysis indicates the relative efficiency withwhich alternative elements of quality will yield higher achievement,holding the cost increment constant.

One cost-benefit study pulled from both experimental and multivariatestudies, comparing the relative efficiency of increasing instructionaltime, reducing class size, or using a pupil tutoring program to boostachievement. All data used in the analysis come from the U.S. The latterintervention involved tutoring of young children (grades 1-4) by olderchildren (from grades 5 and 6). Several experimental and multivariatestudies were reviewed to estimate the relative effect of the three schoolquality interventions. First, the effect sizes of each program on readingachievement gains are calculated in terms of standard deviation units(Table 16, Column A). Second, the cost per student per year is determinedfor each intervention (Column B). Third, the effect size is calibrated per$100 increment of cost (for the base program, not for an additionalincrement of investment as with the Malaysia study).

This form of analysis provides information on the relative efficiencyof different strategies for improving school quality. In addition, theunit cost data demonstrates which intervention is most feasible withinfixed resources. For example, the peer tutoring program is very efficient.But within developing countries, this approach would be more feasible ifits unit cost could be lowered.

71

Table 16U.S. Efficiency Study (Reading)

(a) (b) (c)School Quality Effect Size Cost Per Effect Size PerElement Student $100 in Cost

1. Lengthen instructional .07 $ 61 .12time

2. Reduce class size .22 $201 .06(fm 35 to 20 pupils)

3. Peer tutoring .48 $212 .22

Source: Levin, Glass & Meister (1984)

Method 3: Estimating Rates of Return

Two studies which link school quality and achievement to graduates'subsequent earnings were discussed in Chapter 1. The recent study inBrazil found that school quality (measured by the length of teachers'schooling) more strongly influenced pupils eventual stream of income thandid (pupils) length of school attendance. Similarly, the research inKenya and Tanzania found that workers' earlier level of school achievement(measured with standardized literacy exams) also significantly influencedachievement. And a considerable body of research within industrializedcountries demonstrates that the quality of school attended significantlyaffects graduates' eventual earnings (4.5].

Empirical data could be collected which (a) estimates the efficiencywith which alternative elements of school quality influence achievement orliteracy, and then (b) estimates how strongly levels of school achievementor literacy shape eventual earnings after leaving school. The early studyof secondary school graduates in Kenya did undertake this type of two-steprate of return study. In part, the analysis reports on the relativeefficiency of alternative school quality elements. Here the benefits arein terms of future earnings. The achievement gain associated with oneschool quality factor is seen as an intermediate factor which influencessubsequent earnings. This work examined a limited number of schoolcharacteristics. But the method employed is important to note.

The analysis calibrated costs against a one-point increase in thenational exam. For example, the cost of increasing the average test scoreby one point -- via an increase in per pupil expenditures - was estimatedat 31.25 (Kenyan) pounds per year (Table 17, Column A). The study also hadfound a significant achievement effect for higher teacher salaries.Assuming that this input measure was not acting as a proxy for anotherfactor, the cost of boosting the average test score by raising salaries wasestimated at 26.3 pounds per year. As reported in Chapter 3, lowerachieving students (as a subgroup analyzed separately) realized a higherearnings benefit from higher school achievement. For instance, a one-point

72

gain in achievement was associated with a 15 pound per year difference inearnings, controlling on individual background factors (Column B). But forhigh achieving students (the second subgroup), a one-point gain inachievement was associated with only a 1 pound gain in earnings. Theresearchers then calculated rates of return by relating the cost of eachelement of school quality (per pupil expenditures and teacher salaries)with the realized earnings benefit. The intermediate process here is theschool factor's influence on national exam scores. The calculated rates ofreturn are reported in Column C.

Table 17Kenya Rate-of-Return Study

(a) (b) (c)School Quality Cost Per Year Earnings Benefit Rate-of-ReturnElement Per 1-point Gain Per Year

(pounds) (pounds)

Lower AchievingStudents

1. Expenditures 62.50 15.00 5.0%per pupil

2. Teacher salary 52.60 15.00 7.2%

Higher AchievingStudents

1. Expenditures 62.50 1.00 0.4%per pupil

2. Teacher salary 52.60 1.00 0.5%

Source: Thias & Carnoy (1972)

Note that earning gains associated with each quality element areequal, given that the marginal cost of each quality element was calculatedon the basis of a one-point gain in earnings. A one-point gain in thenational exam is related to an equal gain in earnings regardless of whichelement of school quality determines that one-point increase.

This type of analysis is helpful in estimating the relative efficiencyof different elements of quality, in terms of eventual gains in earnings.For instance, the influence of per pupil expenditures is somewhat greaterthan teacher salaries. In addition, this analysis can determine whichstudents will benefit most from improvements in school quality. Here wesee that quality investments at the margin will not significantly benefitpupils who are already achieving at relatively highly levels.

73

Summary

Empirical research is now sufficient to determine which school factorsare most consistently related to student achievement. However, among theseefficacious elements of school quality, which factors most efficientlyraise achievement? Very little empirical work has been conducted whichfocuses on this issue. We remain largely ignorant of both the magnitude ofbenefit (that is, pupil performance levels) rendered by alternative schoolinputs and management practices. Nor is much known about the relativecosts of different inputs.

Yet the identification of internal inefficiencies is not alwaysdifficult. Simply knowing that a moderately higher class size or morelimited use of classroom laboratories will not lower student achievementidentifies where substantial cost-savings can be realized.

More precise methods for determining the efficiency of alternativeschool factors can also be employed. Care should, however, be taken to notonly examine the efficiency of easily countable material inputs. Costs andachievement benefits are also associated with various ways in which schoolsare managed and classrooms are organized. The study of internal efficiency- by policymakers, local school staff, and researchers -- should focusboth on material inputs and organizational practices.

74

Annex R Review of School Quality Research from Industrial Countries

A brief review of school quality research from industrializedcountries was undertaken. This exercise brought together several articlesand books which have summarized the influence of school characteristics onstudent achievement. Since the Coleman and Plowden reports in the late1960's, the school quality field has grown dramatically and broadened intoa variety of specialized fields. The extent of Third World research,however, is considerably more modest. For developing countries it wasuseful to pull together the individual empirical studies which have beencompleted.

Table A-1 summarizes the several research reviews which apply toschool settings in Western Europe or the U.S. Individual review papers,cited at the end of Table A-i, can be consulted for information on specificempirical studies.

Material Inputs

The influence of material inputs on achievement is inconsistent withinindustrialized countries. Even when a material variable influences pupilperformance, the effect often is quite small. This is the case withsmaller class sizes. One exception: schools with longer academic yearstend to raise student achievement more efficaciously (at least within theU.S.).

Teacher Quality

The quality of teachers more consistently influences achievement.The most marked effects are for teachers' level of experience and verbalcompetence. Teacher salary levels appear to have little consistentinfluence on achievement. Mean salary levels within a school district,however, may determine its ability to attract staff with higher verbalskills.

Classroom Organization

Consistently positive achievement effects are found for severalfactors related to the organization and management of schools.Particularly potent classroom characteristics include time spent oninstructional tasks, assignment and close evaluation of homework, placingstudents in active learning roles (not passively listening to lectures),tight evaluation of students' performance, and the teacher's clarity inpresenting material.

75

Table A-1Influence of SchoDl Reality Elueoets o Student Achievementi

Research froe Industrial Countries

School Duality Elenent Hypotiesis Evidence I Nethod Study--- -------- -- ---- - - --…---------…

A. NATERIAL INPUTS

1. Expenditures per pupil A higher level of rnsources mill NIIED effets found in three U.S. studies; kidge, Judd I NooKkrain studlent Achieveent lwVels POSITIVE effcts found i to other studies, 11979)

vith expeuditwre opeating indirectly viapupiliteacher ratio(NethodaCS,INY SiglTB Leovlfl,S,RI

NO or NEGATIVE effects found in 50 of 55 U.S. Nanushek 11981)studies(llethod:CS,NV SigalT Lovelil,S,R)

POSITIVE effects found for total slasuean I DiniailnovinstruKtioual expenditures Six models (19819reported in four U.S. studies(NethodIVjE,CS Sig:BT Levelail

NO EFFECT found is four U.S. studiem; Jamison, Suppes I wellsPOSITIVE effect found is one U.S. study 11974)(N.thodmNV,CS SiglT Levelail, 9)

2. Class size Faevr studnts per teacher mill POSITIVE effect of smaller class size found Heyueen & Luuley (1981)improve the quality of instructioe - . for science classes is Belgium, 6ersany, and

the U.S.1 NO EFFECT in six other industrialcountrieslflethod:CS,INV SigIT Levelal)

POSITIVE effects found in 10 models estiuated 61asman I Oiniaminovvithin five U.S. studies; 119U1)NEGATIVE effect is two aodels report in samesame studies(lethod3NV,CS SigaDT Levelal)

NO EFFECTSfound is 100 of 109 U.S. studies Hanusbek (19811revieved Also reviewed in(llethodsCS,IV SigtlT Levelil,S,R) Averch et ai. (1974)

NO EFFECT found for clams size difference Glass I Seith (1970)within the 20-40 students/teacher range witkin Ratter (19031the U.S. and England; POSITIVE EFFECT for classsizes of less this 20(NethodCS,tNV SigtBT Levelil,S)

NO EFFECT found in comparing class size Stevenson (1983)of Japanese and U.S. classrooms(IethodsCS,CO SigilT Lve*ll,S)

3. School SiZe Sualler schools mill increase the NO EFFECT found is sis U.S. studies Bridge, Judd & Noockamount and quality of teacher-student (hethodaCS,IY SigiDT levelaI,SIRI (1979)interaction

IIIED and inconsistent effects found in six Rutter (l9531studies reviewed from England and the U.S.(NethadsNS)

4. Instructional materials Higher availability of testbooks and POSITIVE effect found for the school's Neyneman I Losley (1983)other instructional materials 1ill espenditures for books in 6eruany I Hungarylboost the quality of learning activities, NO EFFECT found for nine other countriesespecially in reading instruction (NethodiCS,NV Sigifl Level:l)

5. Length of school daylyear lore total hours spent in school, PDSITIYE and consistent effetts found Bridge, Judd I Noockon instructional activities will boost in four U.S. studies (1979)achievement (Nethod:CS,NY SigjlT Levelil,S)

4. Pbysiral facilities letter facilities provides better, NO EFFECT found in 64 of 71 studies reviewed: Hanushek (19S11sore motivating conditions for leaning (NethodsCS,HV Sig5sr LeveltI,S,R)

7. Library Libraries provide higher levels and NO or NIXED EFFECT found in all eight U.S. Bridge, Judd I Hoockvariety of reading eaterials studies reviewed 11979)

(Nethod:CS,NH SigiBT Levell:,S,R)

3. TEACHER DUALITY

1. Pr-service training Nre years of college instrKtion will POSITIVE effect found for reading teachers in HeNy.man I Losley (1983)boost teaching effectiveaess Hungary, Nk lealand, nd England; NO EFFECT

found is ten other industrial nations(IethodsCS,NV SigulT Levelsll

Kore years of graduate teacher-traiunig NO or NEGATIVE effects founi in seven ridge, Judd I NooKvill boost instructional skills U.S. studies; POSITIVE affect found in 119791

two studies(IHethodsCS,HI MiuT Leveltl,S,R)

2. Teacber experinece Teachers with loner tenure will develo POSITIVE effects found in five U.S. studies; ridge, Judd & Naockstronger instructional skills NIXED or NO effects found in five other (19791

studies depending on specifiC studentachievesent variable examined(HNthod:CS,HV SigaBT Levels l,5.R

POSITIVE, NODERATE effects found in revici Jaeison, Suppes I Vellssix U.S. studies; NO EFFECT in one additional (1974)study

(Nethod:CS,NY SiguBT Lnvel)il

POSITIVE effect found is 30 U.S. studies Hanushek 11911)NO EFFECT reported in 74 other examinations Also reviewed inof this factor Avercb at *L. (1974)(HethodsCS,Iv SigiT Levelfl,S,R)

POSITIYE effects found in all 23 sodels slassun & Biniauiw Yreported in eight U.S. studies (19811(hethodsNV,CS SigilT Levelill

3. Teacher's vweral ibility Teachers ith greater verbal skills will POSITIVE effects found in four U.S. studies; Bridge, Judd I Hoockiscrease the quality of student-teacher NIIED or ND EFFECT found in three other (1979)interaction researches

(NethodsCS,NV SilsBT LevelsS)

POSITIVE, NODERATE effects found for verbal Jamison, Suppes, Vellsskill of teacher, standardized test (1974)(NothOd:CS,NV SigulT Levelul)

POSITIVE effects found for teachers verbal rlasulam & inlulmiovachicvesuot from all IS sodels reported i (1981four U.6. studies(IIthodtV,CB 6ipifr lvrelili

4. Ic u hr salary Nigbr salurins ill attract bhtter N or HIIED effecs for d i nla s1s 1.. studiesi Iridge, Judd I Knockqualified people and sre strongly IKNthodaCS,KV SigloT Lvevltl,B,klsotivat teachers

NO EFFECT found i 51 of A0 studies reviu ed nanushik ll9SIIUlethIodaiS,V Sigtl Lwdalatl,5,3)

POSITIVE effects found in tuo U.B.studies Jal uon, Suppes I mellsN0 or NEGATIVE ufficts, found in five otker (1971Ostudies Also reviewed in 61asmanlNethodiC8,NI liSha Ievelill & iuiamiov 119811

C. CIASSROON STRUCRN I TEACHIhS PRACTICES

1. Classroom ti, spent on Nore timn spent n concrete learI ing POSITIVE effset found for time reading Neymaan u I Loley (1903)instrsctimoal tasks activities, versus tie as disciplising science text in seven industrial nationsl

studeatrs managil rtcords, or hO EFFECT found Is five other countriesarranglig lessons, vill increase (NethodiCSIIVN Sig1uT Lvelil)-Echievent

POSITIVE effsct found for hours of Hicynuuan I Lozley 119B3)instructiom spent per seek on reading isbermany I NungarylllthodaCS,IV SigilT Lvevltll

POSITIVE effeti found for auvber of Nalberg a Shanahan 11993)ecidesic tourses completed Iversusvocational or elective courses) innational sample of U.S. studentslhethadlVY SigilT levelil)

Assigament and close evaluation of POSITIVE effects found in two British Hetter 119031bonusuork ill boost student learning studies

ElbthodCiC,ZN SiglTy LlovelaS6

2. Copurative inslractIual Arrangesut of ad rewards vith POSITIVE effect found in 29 avalustioms of Sviav 119831tasks cooperative imstructional tasks pill cooperative instructional tasks in the U.6.1 Also reviewed In

Ilcroase Istractiom among students NO EFFECT found In 21 other studies Vebb 11982)aid subsequent lcarsiog lIthodaEl,CO Sigall levelil)

3. Vigh teacher xpectattims Teachers sho expect bhl stuaddrds of MITIVE ofdcts foumd when teacher nspecta- Brophy 119791o, studet perforume perfrsaece receiV stronalr couitent tines are satchd sith slt fctivet classroom Valkier (11941

aid ackiave met from students sagetent and utive learing exercises,U.S. studies

IIlehodIlYCU,K lgaPLevehsl,B)

4. ctive teacubil aid T Ie level ifinteractiom NMtiutchr POSIT IVE mi cooi.t.t fildiMgs. thb amount rookovur at &1. 119791learning riles aid studet, aid with leasuing materials,i oltiu teachers spend isteracting uith the flutir 119831

m111 increase studest achieveast class, oat only with isdividual students,Is related to studest achlevemeet, efficacyof Instructitoal materials in facilitatinlattive lairuing remains unlearl hritlsh andILS. studieshNethodsC,iIV SiglT Ieveil,Si

0 5. Tigit ovaluation of stdeet Close cntionoeny betun student effort MPITIVE comrelatio between students Fer- k9okover at &I. R19791achievement aid tachr rards mill boost ception that their classroom efforts sere

studet as UtivtiXoi and ainveset noticed and rewarded and tbeir achievementin 12S V.I. schoolslHthodaCS,INV SilglT LOvelill

POSITIVE effcit found for consistat Valtbrg, 1914recognition of high student performanceIflethoda not BpKcifled is reviewl

A siodest's percption of efficacy mad POSITIVE offetts foun in four U.S. studies Bridge, Judd N Koockability to insflssce the techcer or IlethodsCS,K Sigall levelIll (19191Kchool structure say lead to hiherotivation and ACHievM1nt

A. Clariky of teachwrs Clear slplations of eaterial ad POSITIVE flidings found is 50 studies Rosenshine a Furst 119711presntation will raioe student comprehensios I reviewed fro the U.S. Also reviewed in

clarity of evaluation criteri (NIethodire S1il1 Levelil) Averch et al. (19711

7. Individualized instruction Curriculue which allows each student NO EFFECT found is 39 U.S. studies; Bangert, Kulik I Kulikto progress at his/her own pace will POSITIVE effects in 10 reports (1993)increase mastery of material and (NethodiEl Level: lstudent motivation

0. SCHOOL STRUCTURE

1. Academic vs. vocational Tracking students iato vocational NEGATIVE effect of vocational track found in Heyneman I LEoley (1983)curriculus curriculum, mhere academic achievement Scotland, Neu Zealand, Netherlands I Austrial

is not emphasized, will lower student NO EFFECT found i miste other industrialperformance nations (NethodzCS,IV SigiRS)

2. Student tracking hy Placing students in different tracks will NIXED effects found is nine studies from the futter (19931competence level hinder performance of students who are U.S. and England

expected to perform at lower levels (Nethod:NS)

3. School selectiVity Admitting awe able students will POSITIVE effects found: a school's student Butter 119BI)encourage higher achievement standards cosposition moderately effects the individual

OD student's achievesent, after controlling onthe litter's family background; findings fromeight British and U.S. studies(NethodtCS,NV Sig:lT Levelil,S)

4. Participatory management Greater imvolvement of teachers i. school POSITIVE effects found in British and U.S. Rutter (19031decision-making Will improve teacher studiesmotivation and commitment (NethodtET)

5. Size of administrative staff Ample staff to deal with administration NO or lIXED effects found in six U.S. studies; -Sridge, Judd I Noock111 improve quality of teaching and POSITIVE effects found in just one study (1979)

efficient use of resources lNethod:CS,NY Sil.BT Levolul,S,R)

Administrators with sure training and NO EFFECT found in 50 of 54 studies Hanushet 119811froa higher quality universities vill IlNthodECS,NW SigST LevelIl,5,RIsore elficiently utilize resources

6. Integration of school fore tie spent i. out-of-school work NlIED results from U.S. studiest moderate D*Amico (1904)and work will lower school achievement levels of work outside school may positively

influence school commitment and achievement(Nethod:CS,WV Sig:DT Leveloll

a The folloving codes are *sed is Colum I to siglify the method of analysis used in specific studies.

NETHOD's of analysis.CS s cross-sectiomal data from one point in timeL6 * longitudinal tlee-suries dataCO s simple correlational relationshipNY multivariate method used to control on family hackgroumd, roeunity wealth,

or other ntecendest deteriuasts of school mrhieesentEl a experimental design with classrooe or chool control groupsET - ethnographic case studies, often linkiag observd school characteristics

with empirical dataNS - method not specified is review article

Statistical importance or SlSnificancesRS a the factor explains at least five percent of the variance in

school achieveuot measureBT - bets coefficient is sigpificant at .10 level or better in regressions

or at .05 level or better with bivariate correlations

LEVEL or unit of obhervationsI * Individual studentS * schoolR a region, community, or stateI a aggregate national level

School guality Reviews from Indutrialized Countries

Averch, H., Carroll, S., Donaldson, T., Kiesling, H. & Pincus, J. (1974)How Effective Is Schoolinq? Englewood Clifis, N.J.: Educational TechnologyPublications.

Bangert, R., Kulik, J. & Kulik, C. (1983) Individualized Systems ofInstruction in Secondary Schools. Review of Educational Research. 53 2,143-158.

Bridge, R., Judd, C. & Moock, P. (1979) The Det erinants of EducationalOutcomes: The lagct of FLai ,L Peers Tieachers and Schools. Cambridge,Mass.: Ballinqer.

Brookover, W., Beady, C., Flood, P., Schweitzer, J. & Wisenbaker, J. (1979)scgool Social Svsteas and Student Achievement: Schools Can Make aDifference. New York: Praeger.

Brophy, J. (1979) Teacher Behavior and Its Effects. Jo rnal of EducationalPsXrhology,& 7I, 733-750.

D'Amico, R. (1984) Does Employeent During High School Impair AcademicProgress? Socialogy of Educatio_n, 57, 152-164.

Slasman, N. & Biniasino,, I. (1981) Input-Output Analyses of Schools.Review of Educatianal Research. 51 . 4, 509-539.

Glass, S. & Saith, M. (1?78) Meta-analysis of Research on the Relatimshgjof Class Size and Achievement. San Francisco: Far West Laboratory forEducational Research and Development.

Hanushek, E. (1981) Throwing Money at Schools. Journal of Policy Analy isand ManalmeLntL h 1, 19-41.

Heynesan, S. & Loxley, W. (1983) The Effect of Primary School Quality onAcademic Achievesent across Twenty-nine High and Low-income Countries.American Journal of Sociology , 1162-1194.

Jamison, D., Suppes, P. & Wells, S. (1974) The Effectiveness of AlternativeInstructional Media: A Survey. Review of Research in Educations 44.U 1, 1-67.

Rosenshine, B. & Furst, J. (1971) Current and Future Research on TeacherPerformance Criteria. In B.L. Smith (ed) Research n Teache Educatio n: ASywsiys. Englewood Cliffs, N.J.: Prentice-Hall.

Rutter, M. (1983) School Effects on Pupil Progress: Research Findings andPolicy Implications. Child Devel ogent. §, 1-29.

Slavin, R. (1993) Coogerative Learning. New York: Longean.

Stevenson, H. (1963) Comparisons of Japanese, Taiwanese, and AmericanMathematics Achievement. Stanford, Ca.: Center of Advance Study in theBehavioral Sciences.

83

Walberg, H. (1964) Improving the Productivity of America's Schools.Educational LeadershiX i1 8, 19-30.

Walberg, H. E Shanahan, T. (1983) High School Effects on IndividualStudents. Educational Researcher, 12. 7, 4-9.

Webb, N. (1982) Student Interaction and Learning in Small Groups. Review ofEducational Research. 521 3, 421-445.

84

Notes

Full citations to the studies listed below appear in the referencelist (page ).

Summary and Chapter 1

1.1 Dore (1976)1.2 Unesco (1983a)1.3 Egypt data, Hartley & Swanson (1984); Kenya, Somerset (1983);

Brazil, Behrman & Birdsall (1983); Peru, Cespedes (1984).1.4 de Regt (1982)1.5 Heyneman & Loxley (1983); this issue is discussed in detail in

Chapter 2.1.6 For reviews see, Heyneman (1980) and Niles (1981).1.7 Heyneman & Loxley (1983)1.8 Heyneman, Jamison & Montenegro (1983)1.9 Kravis (1984)1.10 Schiefelbein & Farrell (1982, 1984)1.11 Brazil study, Behrman & Birdsall (1983)1.12 Distribution of educational benefits, Mingat & Tan (in press).1.13 Boissiere, Knight & Sabot (in press)1.14 Jamison & Moock (1984). Studies from industrialized countries

on the economic effects of school quality are reviewed inSolmon (1985).

1.15 Fuller, Gorman & Edwards (in press)1.16 Behrman & Birdsall (1983)1.17 Fuller (1985)1.18 Current work is looking at (a) the tradeoff between expanding

enrollments versus increasing quality, in terms of expenditures perpupil, and (b) tradeoffs in investing among different schoolcharacteristics. For a draft paper, see Mingat & Tan (1984).

1.19 Unesco (1983a)1.20 World Bank (1984)

Chapter 2

Many studies referred to within the text are clearly cited in Tables 8-12.Only papers not clearly identified in the narrative and tables are notedbelow.

2.1 For examples of this research, see Bing (1963), Hess & Holloway, (1984)2.2 Mexico, Bowman (1984); Brazil, Armitage, Ferreira Gomez, Holsinger

& Leite (1984).2.3 This paper focuses only on the internal effectiveness and

efficiency of schools. External efficiency is also an importantissue related to school quality. That is, do elements of schoolquality influence the student's occupational success and social

85

quality of life after leaving school? This question has spawned alarge amount of research - which will not be reviewed here.

2.4 For review, see Schiefelbein & Simmons (1981).2.5 Heyneman & Loxley (1983)

Chapter 3

3.1 My review builds from earlier summaries of the school qualityliterature. Especially helpful reviews include Simmons & Alexander(1978); Husen, Saha & Noonan (1978); Heyneman, Farrell & Sepulveda-Stuardo (1978); Schiefelbein & Simmons (1981).

3.2 Jamison, Searle, Galda & Heyneman (1981)3.3 For examples, Thorndike (1973) and Rivarola & Corvalan (1976).3.4 The cost-benefit analysis for Nicaragua appears in Jamison (1978).3.5 Summarized in Heyneman & Loxley (1983).3.6 The Chilean data are reported in Husen (1977). For Puerto Rico

findings, see Carnoy (1971).3.7 For review, Husen, Saha & Noonan (1978).3.8 For a report on the World Bank's lending in education, see

Romain (1974).3.9 See Busen et al. (1978) and Heyneman & Loxley (1983).3.10 The efficiency analysis appears in Thais & Carnoy (1973).3.11 The tEA analysis is reviewed in Husen et al. (1978).3.12 For reviews, see Edmonds & Frederiksen (1979), Rosenholtz (1985),

Fuller, Izu & Berman (1985).3.13 Metcalf (1985) and Psacharopoulos & Loxley (1985)3.14 For examples, see Levin (1976) and Klees (1984).3.15 Noted earlier, initial results of ongoing work is reported in

Mingat & Tan (1985).3.16 Avalos & Haddad (1981)3.17 For reviews of reliable measures of observable teaching and

classroom management practices, see Flanders (1985), Karweit(1985), and Peterson, Micceri & Smith (1985).

Chapter 4

4.1 For the Bolivian analysis, see Morales & Pinellailes (1977). Foran efficiency study of class size reduction, see Jamison (1982).

4.2 Klitgaard, Siddiqui, Arshad, Niaz & Khan (1985)4.3 This efficiency analysis for Chile is detailed in Farrell &

Schiefelbein (1974).

86

References

Complete citations for papers indicated in Table A-1 directly follow thetable and do not appear below.

Armitage, J., Ferreira Gomes, J., Holsinger, D. & Relio Leite, R. (1984)School Quality and Achievement in Rural Brazil (draft mimeo). Paperpresented at the World Bank, November.

Arriagada, A. (1981) Determinants of Sixth Grade Student Achievement inColombia (Mimeo). Washington, D.C.: The World Bank, Education Department.

Arriagada, A. (1983) Determinants of Sixth Grade Student Achievement inPeru (Mimeo). Washington, D.C.: The World Bank, Education Department.

Avalos, B. & Haddad, W. (1981) A Review of Teacher Effectiveness Research.Ottawa: International Development Research Centre.

Balderston, J., Wilson, A., Freire, M. & Simonen, M. (1981) MalnourishedChildren of the Rural Poor. Boston: Auburn House.

Beebout, H. (1972) The Production Surface for Academic Achievement: AnEconomic Study of the Malaysian Secondary Education. PhD dissertation,University of Wisconsin.

Behrman, J. & Birdsall, N. (1983) The Quality of Schooling: Quantity Aloneis Misleading. American Economic Review- 73, 5, 928-946.

Benavot, A. (1982) "Secondary Schooling and the Impact of VocationalEducation on Economic Development." Paper read at Midwest SociologicalAssociation, Des Moines, Iowa.

Bibby, J. & Peil, M. (1974) Secondary Education in Ghana: PrivateEnterprise and Social Selection. Sociology of Education 47, 399-418.

Bing, E. (1963) Effects of Child Rearing Practices on Development ofDifferentiated Cognitive Abilities. Child Development 34, 631-648.

Birdsall, N. (in press) Public Inputs and Child Schooling in Brazil.Journal of Development Economics.

Boissiere, M., Knight, J. & Sabot, R. (in press) Earnings, Schooling,Ability, and Cognitive Skills. American Economic Review.

Bowman, M. (1984) An Integrated Framework for Analysis of the Spread ofSchooling in Less Developed Countries. Comparative Education Review-28, 4, 563-583.

87

Carnoy, M. (1971) Family Background. School Inputs, and StudentsPerformance in School: The Case of Puerto Rico (Mimeo). Palo Alto:Stanford University.

Carnoy, M. & Thias, E. (1974) Second Tunisia Education Research Project(dract mimeo). Washington, D.C.: The World Bank.

Cespedes, A. (1984) Peru Primary Education Proiect: Staff ApDraisal Report.Washington, DC: World Bank.

Clark, D. (1983) How Secondary School Graduates Perform in the LaborMarket: A Study of Indonesia (SWP No. 615). Washington, D.C.: The WorldBank.

Comber, L. & Keeves, J. (1973) Science Education in Nineteen Countries. NewYork: Halstead Press.

Costa, M. (1977) School Outputs and the Determinants of ScholasticAchievement: An Econometric Study of Urban Schools in Sao Paulo (Mimeo).Stanford, Calif.: Stanford University.

Currie, J. (1977) Family Background, Academic Achievement and OccupationalStatus in Uganda. Comparative Education Review. 21, 14-27.

de Regt, J. (1982) Haiti: Staff ADpraisal Report (Basic Education Project).Washington, DC: World Bank.

Dore, R. (1976) Human Capital Theory, the Diversity of Societies and theProblem of Quality in Education. Hizher Education, 5, 79-102.

Drysdale, R. (19 ) Factores Determinantes de la Desercion Escolar enColombia. Revista del Centro de Estudios Educativos.

Durojaiye, M. (1974) The Role of Non-cognitive Factors in School Learningof Uganda Secondary School Pupils. West African Journal of Educational andVocational Measurement, 2, 1, 35-39.

Echart, E., Meir, J., Manuelli, R. & Binimelis, M. (1976) Los Determinantesde la Educacion en Argentina. Buenos Aires: ECIEL.

Ellson, D. (1973) "Programmed Teaching: Effective Teaching by 'UnqualifiedTeachers'." In T. Husen et al. (1978).

Farrell, J. & Schiefelbein, E. (1974) Expanding the Scope of EducationalPlanning: The Experience of Chile. Interchange. 5, 18-30.

Flanders, N. (1985) Human Interaction Models of Teaching. In T. Husen & T.Postlethwaite, International Encyclopedia of Education (Volume 4, F-H).Oxford: Pergamon Press.

Fuller, B. (1985) School Quality Trends in the Third World (mimeo).Washington, DC: World Bank, Education & Training Department.

88

Fuller, B., Izu, J. & Berman, P. (1985) Explaining School Cohesion: WhatShapes the Organizational Beliefs of Teachers? (mimeo). University ofMaryland.

Fuller, B., Gorman, K. & Edwards, J. (in press) The Influence of SchoolInvestment Quality on Economic Growth: An Historical Look at Mexico,1880-1940. In S. Heyneman & D. Siev White (ed) The Quality of Educationin Develolinx Countries. Washington, D.C.: The World Bank.

Fuller, W. & Chantavanich, A. (1976) A Study of Primary Schoolinz inThailand: Factors Affecting Scholastic Achievement of the Primary SchoolPupils. Bankok: Office of the National Education Commission.

Haddad, W. (1978) Educational Effects of Class Size (SWP No. 280).Washington, D.C.: The World Bank.

Haron, I. (1977) Social Class and Educational Achievement in a PluralSociety: Peninsular Malaysia (Doctoral Dissertation). Chicago:University of Chicago.

Hartley, M. & Swanson, E. (1984) "Achievement and Wastage: An Analysis ofthe Retention of Basic Skills in Primary Education" (Draft). Washington,D.C.: The World Bank, Development Research Department.

Hess, R. & Holloway, S. (1984) Family and School as EducationalInstitutions. In R. Parke (ed) Review of Research in Child Development.

Heyneman, S. (1976) Influences on Academic Achievement: A Comparison ofResults from Uganda and More Industrialized Societies, Sociolorr ofEducation. July.

Heyneman, S. (1980) Differences Between Developed and Developing Countries:Comment on Simmons and Alexander's "Determinants of School Achievement."Economic Development and Cultural Change, 28, 2, 403-406.

Heyneman, S. & Jamison, D. (1980) Student Learning in Uganda: TextbookAvailability and Other Factors. Comparative Education Review 24, 206-220.

Heyneman, S. & Loxley, W. (1983) The Effect of Primary School Quality onAcademic Achievement across Twenty-nine High and Low-Income Countries.American Journal of Sociolofv. 88, 6, 1162-1194.

Heyneman, S., Jamison, D. & Montenegro, X. (1983) Textbooks in thePhilippines: Evaluation of the Pedagogical Impact of a NationwideInvestment. Educational Evaluation and Policy Analysis 6, 2, 139-150.

Husen, T. (1977) "Pupils, Teachers, and Schools in Botswana: A NationalEvaluation Survey of the Primary and Secondary Education." Gaborone:Government Printing Office.

Husen, T., Saha, L. & Noonan, R. (1978) Teacher Training and StudentAchievement ia Less Developed Countries (SWP 310). Washington D.C.: TheWorld Bank.

89

Izquierdo, C. & Guzman, J. (1971) Una Exploracion de los FactoresDeterminantes del Edimento Escolar en la Educacion Primaria. Revista delCentro de Estudios Educativos. 1, 2, 7-27.

Jamison, D. (1978) Radio Education and Student Repetition in Nicaragua.Washington, D.C.: World Bank (Reprint Series No. 91).

Jamison, D. (1982) Reduced Class Size and Other Alternatives for ImprovingSchools: An Economist's View. In G. Glass, L. Cahen, M. Smith & N. Filby,School Class Size: Research and Policy. Beverly Hills, Calif.: Sage.

Jamison, D. & Moock, P. (1984) Farmer Education & Farm Efficiency in Nepal.World Development. 12, 1, 67-86.

Jamison, D., Searle, B., Galda, K. & Heyneman, S. (1981) ImprovingElementary Mathematics Education in Nicaragua: An Experimental Study ofthe Impact of Textbooks and Radio on Achievement. Journal of EducationalPsycho loty. 73. 4 556-567.

Karweit, N. (1985) Should We Lengthen the School Term? EducationalResearcher, June/July.

Klees, S. (1975) Television and Other Determinants of ScholasticAchievement in Mexican Secondary Education (Mimeo). Ithaca, N.Y.:Cornell University, Economics Department.

Klees, S. (1984) The Need for a Political Economy of Educational Finance.Comparative Education Review, 28, 3, 424-443.

Klitgaard, R., Siddiqui, K., Arshad, M., Niaz, N. & Khan, M. (1985) TheEconomics of Teacher Education in Pakistan. Comparative Education Review29, 1, 97-110.

Lee, C. (1985) "Financing Technical Education in LDC's: EconomicImplications from a Survey of Training Modes in the Republic of Korea."Washington, D.C.: Education Department, The World Bank.

Levin, H. (1976) Concepts of Economic Efficiency and EducationalProduction. In J. Froomkin, D. Jamison & R. Radner, Education as anIndustry. Washington, DC: National Bureau of Economic Research.

Levin, H., Glass, G. & Meister, G. (1984) Cost-Effectiveness of FourEducational Interventions (mimeo). Stanford, Calif.: Institute forResearch on Educational Finance and Governance.

Levy, M. (1971) Determinants of Primary School Dropouts in DevelopingCountries. ComDarative Education ReviewX 15, 1, 44-58.

Loxley, W. (1984) Quality of Schooling in the Kalahari (Mimeo). Paper read,Comparative and International Education Society, Houston.

Metcalf, D. (1985) The Economics of Vocational Traininu (SWP 713).Washington, DC: World Bank.

90

Mingat, A. & Tan, J. (1984) On the Quantity-Quality Tradeoff in Education(draft mimeo). Washington, DC: World Bank, Education & Training Department.

Mingat, A. & Tan, J. (in press) Who Profits from the Public Funding ofEducation? Comparative Education Review.

Morales, J. & Pinellsiles, A. (1977) The Determinant Factors and the Costsof Schooling in Bolivia (Working Paper No. 4-77). La Paz: UniversidadCatolica Boliviana.

Nasoetion, N., Djalil, A., Musa, I. & Soelistyo, S. (1976) The Developmentof Education Evaluation Models in Indonesia. Paris: InternationalInstitute for Educational Planning, Unesco.

Niles, F. (1981) Social Class and Academic Achievement: A Third WorldReinterpretation. Comparative Education Review, 25, 419-430.

Peterson, D., Micceri, T. & Smith, 0. (1985) Measurement of TeacherPerformance: A Study in Instrument Development. Teachinx and TeacherEducation, 1, 1, 63-77.

Psacharopoulos, G. (1973) Returns to Education: An InternationalComparison. San Francisco: Jossey-Bass.

Psacharopoulos, G. & Loxley, W. (in press) Diversified Secondary Educationand Development. Washington, D.C.: The World Bank, Education Department.

Psacharopoulos, G. & Zabalza, A. (1984) The Destination and Early CareerPerformance of Secondary School Graduates in Colombia. Washington, D.C.: The World Ban;

Purves, A. (1973) Literature Education in Ten Countries. Stockholm:Almqvist & Wiksell.

Rivarola, D. & Corvalan (1976) Determinante del Rendimiento Educativo en elParaguay (Mimeo). Asuncion: ECIEL.

Romain, R. (1984) Lending in Primary Education: Bank Performance Review FY1963-83 (draft mimeo). Washington, DC: World Bank.

Rosenholtz, S. (1985) Effective Schools: Interpreting the Evidence.American Journal of Education, May, 352-388.

Rowe, E., Lau, G., Lee, G., Li, A. & Rodd, R. (1966) Failure in School:Aspects of the Problem in Honz Kon&. Hong Kong: Hong Kong UniversityPress.

Ryan, J. (1973) Educational Resources and Scholastic Outcomes: A Study ofRural Primary Schoolint in Iran. PhD Dissertation, Stanford University.

Schiefelbein, E. & Clavel, C. (1977) Stability Over Time of EducationalInput-Output Relationships (Mimeo). Santiago: University of Chile,Economics Department.

91

Schiefelbein, E. & Farrell, J. (1973) Factors Influencing AcademicPerformance among Chilean Primary Students (Mimeo). Santiago: Centro deInvestigaciones y Desarrollo de la Educacion.

Schiefelbein, E. & Farrell, J. (1982) Ei2ht Years of Their Lives: ThroughSchoolinz to the Labour Market in Chile. Ottawa: International DevelopmentResearch Centre.

Schiefelbein, E. & Farrell, J. (1984) Education and Occupational Attainmentin Chile: The Effects of Educational Quality, Attainment, andAchievement. American Journal of Education. 125-162.

Schiefelbein, E. & Simmons, J. (1981) Determinants of School Achievement: AReview of Research for Developing Countries (mimeo). Ottawa:International Development Research Centre.

Sembiring, R. & Livingstone, I. (1981) National Assessment of the Qualityof Indonesian Education. Jakarta: Ministry of Education and Culture.

Silvey, J. (1972) Long Range Predicition of Educability and Its Determinants inEast Africa. In L. Cronbach & P. Drenth (eds.) Mental Tests and CulturalDeDrivation. Paris: Mouton Publishers.

Simmons, J. (1970) Towards an Evalution of Adult Education in a DeveloPin2Country (Volumes 1 and 2). Paris: Unesco.

Simmons, J. & Alexander, L. (1978) The Determinants of School Achievementin Developing Countries: A Review of the Research. Economic DeveloDmentand Cultural Change. 26, 2.

Smart, N. (1978) The Densu Times: Self Made Literacy. DeveloDmentCommunication Re-port. 21.

Solmon, L. (1985) Quality of Education and Economic Growth. Economics ofEducation.

Somerset, H. (1968) Predictinu Success in School Certificate:A Uzanda Case Study. Nairobi: East Africa Publishing House.

Somerset, R. (1983) Examinations Reform: The Kenya Experience (mimeo).Sussex: University of Sussex, Institute of Development Studies.

Thias, H. & Carnoy, M. (1973) Cost Benefit Analysis in Education: A CaseStudy of Kenua. Baltimore: Johns Hopkins Press.

Thorudike, R. (1973) Readinr ComDrehension in Fifteen Countries. New York:Halsted Press.

Windham, G. (1970) Occupational Aspirations of Secondary School Students inSierra Leone. Rural Sociolo2v. 35, 1, 40-53.

Unesco (1983a) Trends and Proiections of Enrolment bvy Level of Educationand bv Age 1960-2000. Paris: Office of Statistics.

92

Unesco (1983b) Statistical Yearbook. 1983. Paris: Office of Statistics.

Wolff, L. (1970) Why Children Fail in First Grade in Rio Grande do Sul:Implications for Policy and Research. Washington, D.C.: U.S. Agency forInternational Development.

World Bank (1984) World Develovment ReDort. 1984. Washington, DC & NewYork: Oxford University Press.

Youdi, R,. (1971) An Exnloratorv Study of Achievement and Attitudes gj NighSchool Students in the Con-o: An Aspect of Socialization for NationalDevelopment. PhD Dissertation, Stanford University.

93

raising school quality in developing counties: investments ...€¦ · occurred on the relative...

Documents