Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized

PoNoy,Planning, and Reesarak

WORKING

PAPERS

Educationand Empoyment Populationand HumanResources Department

The WorldBank August1988 WP3 61

Student Performance and School Costs in the Philippines' High Schools EmmanuelJimenez, Vicente Paqueo, and Ma. Lourdesde Vera Private schools in the Philippines are substantially more effective than their public counterparts in teaching language skills, marginally less effective in teaching mathematics skills, and much less costly per pupil than public schools.

wokin pweu andto 3 Policy,Plng, and mRchCanplex R diatsbutsPFRWoiingPapentodatntethefindingpof ancange dithe exangeof ideu amongBankataffand al othemintead in devlopmnent ismes.Thus papea cam thenamesof ar the theanthua, rellea aly theirviews,and ad be used andcitedaccoringly.Te findings,intitatiams. and cmncuhions itmanagemen, ortanyofits memberem. to theWorldBank,it Boad of Dirmctis, autbW own.Theyshold notbe atitd

PoNcy,Pbnnin, a*d R_euech

Edotn and Employmentn

A key consideration in the policy debate on the appropriate role of private schools in preuominantly public school systems is cost effectiveness. The questions are: Do private school students leam more than their public school counterparts? And is it more or less expensive to educate students in private schools? Past studies in the Philippines and elsewhere have claimed that the educational achievement of students in the private schools is higher than that of students in public schools. These studies provide, however, only weak evidence regarding the relative cost effectiveness of public and private schools. A fundamental weakness is the potentially serious problem of selectivity due to unobserved differences between the student population of each type of school. Most of the studies do not compare costs in the two types of institutions. Taking selectivity into account, the paper finds that controlling for the effects of students' socioeconomic background, individual motiva-

tion, and innate ability, the private schools show a significant edge over public schools in both English and Pilipino (about 15 percent of the sample mean achievement scores). Public schools, on the other hand, had a slight (roughly 4 percent) advantage in mathematics. A comparison of cost per student reveals a substanlial ad, antage for private schools: public schools in the Philippines spend on average roughly twice as much as private schools. These findings strongly suggest that private schools are an efficient purveyor of secondary education in the Philippines, a conclusion that should be taken into account in the formulation of policy measures that could threaten the existence of such schools. This paper is a product of the Education and Employment Division, Population and Human Resources Department. Copies are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Teresa Hawkins, ir,,m S6-224. extension 33678.

The PPRWorkingPaperSeriesdisseminatesthe findingsof workunderway in the Bank-sPolicy,Plarning,and Research Complex.An objec'aveof the seriesis to get thesefmdingsout quickly,evenif presentationsare lessthan fullypolished. The findings,interpretations,and conclusionsin thesepapersdo notnecessarilyrepresentofficialpolicyof the Bank. Copyright© 1988by the InternationalBankfor Reconstructionand Developmentft'he WorldBank

Student Performance and Schools Costs In the Philippines' High Schools by Emmanuel imenez, VicentePaqueo, and Ma.Lourdesde Vera Tableof Contents Introduction....... ....................... .2 The BasicEmpiricalData . . .5 Data. . .9 ThwPhilippineContext . . . 9 S le ... 10 Acldevementand StudentBackground . ........................... 11 The Effectof Backgroundon Achievementin Publicand PrivateSchools 14 WhatDeterminesthe Choiceof SchoolType?.14 How DoesSocio-economic BackgroundAffectSchoolAchievement? 16 BackgroundConstant,Is Therea PrivateSchoolEffect?.19 Codusions ... 22 References................................................................................................................ 24 Tables

INTRODUCTION Few analystshave challengedthe notion that the public sector has a role to play in directing and stimulatingeducation. However, the nature of that role has been the focus of a recent lively 1 In particular,it has been argued exchange in the literature.

that, sincs private schools have built-in incentivesto provide educationefficiently,they should be allowed to competewith public schools on a (more or less) equal footing. Among developingcountries, there is an added policy dimension:tighteningfiscal constraintshave limited the ability of the public sector to expand its provision of free public education (World Bank 1986). A greater relianceon the private sector may become a financialnecessity if ambitious educational targets are going to be met in the near future. The key empirical questionsares Do private school stuc%ncs learn more than their public school counterparts? Is it more or less expensive to educate students in private schools? The debate is fueled by controversyover methodology,interpretationand date. The most importantmethodologicalissue is the difficultyin attributing differencesbetween the cognitive abilitiesof students in public versus private schools to school inputs alone, since a variety of non-school factors also affect achievement, such as socio-economic

background,

innate ability and individualmotivation. Moreover, these non-school factors also affect school choices made by families. For example,

I In the United States, the debate was sparked by the Coleman, Hoffer and Kilgore (1982) report which concludedthat private (Catholic) schools are more effectivethan public schools in helping students to acquire cognitive skills. Achievement in the Philippines .......

..

page 2

if children from privilegedbackgroundsonly attendedprivate schools, it would be difficultto infer how they would do in public schools. Thus, unless non-schoolfactors are controlledappropriately,estimates of school effects will be contaminatedby what has become known as "selectivitybias." The.problemis that the researcher'smeasures of these factors,particularlythose that act &s proxies for ability and motivation,are far from perfect.2 Modern statisticaltechniqueshelp in controllingfor this bias, although recent researchhas revealedthat it is also importantto keep track of one's assumptionsin mod1lling (see Murnane, Newstead and Olsen, 1985, for a careful assessmentof the results of Coleman et al. and their critics). This paper contributesto the literaturein s4

important

dimensions. First, it extends the empirical evidencefo_ dsveloping countries by analyzingdata secondary level from the Household and School Matching Survey (HSMS) conductedby the EducationalDevelopment Projects ImplementingTask Force (EDPITAF)of the Philippinesduring the 1981-82 academicyear. The only other rigorouscomparisonsof publiciprivateschools in determiningachievementin developing countrieshave been conducted in Kenya (Armitageand Sabot, 1987), Colombia and Tanzania (Psacharopoulos,1987; Cox and Jimenez, 1987), and Thailand (Jimenez,Lockheed and Wattanawaha,1988). The Philippinesis a particularlyapt extensionbecause it has one of the highest rates of priv&ta school (about 402 of total secondary)enrollmentin the world. 2 Several studies have attemptedto use direct measures of ability through the use of tests specificallydesignedto measure innate ability (e.g , an I.Q. test) rather than cognitive achievement(Psacharopoulos and Loxley, 1985; Boissiere,Knight and Sabot, 1984 among others). Many analystshave questionedthe validity of these tests in distinguishing between ability and achievement. In any case, no one has ever suggested that such tests fully control for both ability and motivation.

Achievement in thePhilippines.;.page

3

Second, the paper also confronts the difficultmethodological questions that have arisen in other studies. An individual'sstatus as public or private school student is a choice made by student and parent. If this choice is systematicallycorrelatedwith personal characteristics,there may be sample selectionbias. We use some recent methodologicaladvancesto model and correct statisticallyfor this bias. Third, we use independentlygathered data to compare unit costs of public and private schools, before reachingconclusionsregarding the relativeefficiency (as opposed to the relativeproductivity)of public and private schools. In the next section of this paper, we outline the basic conceptualmodel and our approach in correctingfor selectionbias. Then, this is followedby sectionsont data, results on school effActs, results on relativecosts and conclusions.

Achievement in the Philippines .....................

page 4

THE BASIC EMPIRICALFRAMEWORK

Would a Filipinohigh school student, randomlyselected from the general studerntpopulatior,do better in a public or private school? In the absence of experimentaldata, a reliableanswer can be obtained from a cross-sectioncomparisonof public and private school students' performance in standardizedtests -- if we control for student background,motivation and innate ability. A standardmethod is to postulate the followingreduced form model: the "ith" private school student'sachievementscore (A) is a functionof a vector of observedbackgroundvariables (X) and unobserved variables (e)3

(la)

Aip

-

bp Xip + eipt

where each component of b measures the marginal effect of a characteristicon achievement. The "jth" public (or government)school student's score can be be similarlyexpressedby replacingthe subscript p* with Ug3s

(lb)

Ajg - b8 Xjg + ej*g

If the effects due to unobservedvariables,e, 4re randomly and normally distributed,ordinary least squares regressiontechniquescan be 3Alternatively, equations (la) and

(lb) can be estimated as one equation, with a dummy variable for private and public types of schools. However, statistical (F-) tests lead us to reject the hypothesisthat the coefficientsof all the other variablesare equivalentin both types of schools. Results are availablefrom the authors.

Achievement in the Philippines ........

page 5

used to estimate the parametersof equations (la) and (lb). PiLvate/publiccomparisonscan then be made using this information. The method would be to compare the predicted test score of a person with a given set of backgroundc

-

cteristicsin each of the public and

private school systems by u 'tg equations (la) and (1'). A critical problem arises if the observed public and private subsamplesare basically incomparabledue to selectionbias. This would be the case if studentswith a certain backgroundsystematicallychose one type of school over another. For example,if privilegedstudents chose only private schools,there would be no privilegedstudents enrolled in public schools. Thus, it might be misleading to uae equation (lb) to infer how privilegedstudentswould do in public schools. The error terms e are no longer normally distributedand OLS should not be used to estimate the above equations. To correct for sample selection,we use Heckman's (1979) two-step technique. The first step in this methodologyis to estimatewhat determinesthe choice of type of school. We assume that individualswill choose an educationalplan, includingthe type of school, that maximizes the child's economicwell-being,net of private costs. The solutionto this problem can be shown to result in the following choice equation for the "ith" child (Cox and Jimer.ez1987):

(3)

Ii* - k Yi + wi,

where Ii* is an unobservedvariablewhich characterizesthe propensity of a household to choose a certain type of school for the child. Since

Achievement in the Philippines ............................ * page 6

it

is unobserved,we use the indicatorvariable:

Ii - 1 Ii

O 0

iff II* > 0 and otherwise.

Y indicatesthe explanatoryvariables and w is a random error term. The second st p is to use the results of the first step to correct for the selectionbias in (la) and (lb). With selection,the expectedvalues of Ai are conditionalon the choice of public and private sector. This means that the error terms *i are correlatedwith wi.

The expectedvalue of ei will no longer be equal to zero and the

estimated parametersin (la-b)will be biased if OLS is applied. If we assume that (wi, *i) are jointly distributednormal with mean zero, then,

(4a)

E(eipVII>O)- ap Xip, and

(4b)

E(eiglli