scalemicrocenter

TeachingandTeacherEducation15(1999)849}859Recruitmentofruralteachersindevelopingcountries:aneconomicanalysisPatrickJ.
McEwanSchoolofEducation,StanfordUniversity,Stanford,CA94305-3096,USAReceived6April1998;receivedinrevisedform2November1998;accepted22January1999AbstractMonetaryandnon-monetaryincentivesforruralteacherrecruitmentareaprominentfeatureofdeveloping-countryeducationsystems.
Despitethewidespreaduseofincentives,thereislittletheoreticalorempiricalevidenceontheire!
ectiveness.
Thispaperinterpretsincentivepolicieswithintheframeworkoftheeconomictheoryofcompensatingdi!
erentials.
Thediscussionclari"estheimplicitassumptionsofincentivepoliciesandaidsinorganizingfurtherempiricalworkontheire!
ectiveness.
Existingevidenceoncompensatingdi!
erentials,mainlyintheUnitedStates,showsthatteacherstendtotradeo!
monetarywagesagainstnon-monetaryaspectsoftheirjobs,suchasgeographiclocationandclasssize.
1999ElsevierScienceLtd.
Allrightsreserved.
Keywords:Ruralareas;Geographicisolation;Incentives;Developingnations;Recruitment1.
IntroductionResearchontheteachingprofessionindevelop-ingcountriesobservesthatruralareaspresentmoreonerouslivingandworkingconditionsthanurbanareas(InternationalLabourO$ce,1991;Murnane,1993;Lockheed&Verspoor,1991).
Forinstance,ruralteachersareoftensubjecttosocialandprofessionalisolation,riskstopersonalsafety,andchallenginglivingconditions.
Ifruralschoolsaretorecruitpersonnelatleastasquali"edandableasthoseofurbanareas,researchersconcludethatmonetaryornon-monetaryincentivesarere-quired.
E-mailaddress:pmcewan@leland.
stanford.
edu(P.
J.
McEwan)Ruralincentivesarealreadyaprominentfeatureofmanyeducationsystems.
Developingcountrieshaveadoptedavastarrayofrecruitmentstrategies,includingwagepremiums,subsidizedruralhous-ing,specialin-servicetraining,andeventheforciblereassignmentofteacherstoruralareas.
Despitetheconsiderableresourcesdevotedtoincentives,thereisscanttheoreticalorempiricalworkwhichsystematicallyexplorestherationalefortheirimplementationorthecost-e!
ectivenessofcurrentrecruitmentpolicies.
Thispaperdelineatesaframeworkforanalyzingruralteacherrecruitment,groundedintheeco-nomictheoryofcompensatingwagedi!
erentials.
Thefundamentalideaisthatteachersvaluepecuni-aryandnon-pecuniaryaspectsoftheirjobs,andwillinglytradeo!
oneagainsttheotherinordertoachievehigherlevelsofsatisfaction.
Thepurposeofthediscussionistwofold.
First,itemphasizesthe0742-051X/99/$-seefrontmatter1999ElsevierScienceLtd.
Allrightsreserved.
PII:S0742-051X(99)00025-6assumptionsthatareonlyimplicitinmostaca-demicandpolicydiscussionsofruralteachingin-centives(indeed,theseassumptionsunderliemostofneoclassicaleconomics).
Foremostisthatteachersarerationalindividualswhomakedeci-sionsabouttheirworkandresidencesoastomaximizetheir`utilitiesa.
Second,thediscussionprovidesausefulmeansoforganizingfurtherem-piricalresearchintotheaspectsofruralworkwhichrequireincentivesandwhetherteachersexhibitagreaterwillingnesstoworkinareasthatprovidesuchincentives.
Thoughmyfocusisonruralteachers,theframeworkcaneasilybeappliedtoanyteachers,urbanorrural,whofacedi$cultlivingorworkingconditions.
Beforediscussingthetheory,thesecondpartofthispaperreviewsevidenceonthegapinschoolqualitybetweenruralandurbanareas.
Suchevid-enceisoftenusedtorationalizerecruitmentpol-icies.
Someexamplesofthesepoliciesarealsosummarized.
Part3describesthetheoryofcom-pensatingdi!
erentialsanditsapplicationtoteacherlabormarkets.
Part4operationalizesthevagueconceptof`ruralnessa,discussingspeci"caspectsoflifeandworkinruralareaswhichmayrequirecompensation.
Finally,part5exploresthelimitedempiricalevidencewhichexists*almostallofitintheUnitedStates*onteachers'implicitvaluationofsomeaspectsofruralteaching,aswellasestablishingsomecaveatstoitsinterpretation.
Theprincipalintentofthediscussionistomotivatesystematicempiricalworkonruralteacherrecruit-mentindevelopingcountries.
2.
RuraleducationandteacherrecruitmentBymanyyardsticks,ruralschoolsandstudentsareworseo!
thantheirurbancounterpartsthroughoutthedevelopingworld.
Enrollmentsofschool-agechildrenare,ingeneral,lowerinruralareas(Lockheed&Verspoor,1991).
InColombia,forexample,thenetprimaryenrollmentsare86%forurbanmalesand78%forruralmales(Molina,Alviar&Polania,1993).
Sometimeslowenroll-mentsareduetoasimplelackofschools;inothercases,existingschoolso!
eronlyafewgrades.
RuralschoolsconsistentlyshowhigherrepetitionandTable1Formaleducationofprimaryschoolteachers,1989CountryMeanyearsofschoolingUrbanRuralBrazil14.
213.
0CostaRica14.
413.
0Honduras12.
611.
5Panama14.
213.
0Source:Psacharopoulos,Valenzuela&Arends(1993).
drop-outratesthanurbanones,perhapsindicativeoflowstudentachievement.
Thisisthecase,forexample,inGuatemala,Brazil,andColombia(Lockheed&Verspoor,1991;DepartamentoAdministrativoNacionaldeEstadmHstica,1993;Colbert,Chiappe&Arboleda,1993).
Urbanstu-dentshavescoredbetteronstandardizedachieve-menttestsinPeru,thePhilippines,Thailand,andColombia(Lockheed&Verspoor,1991;MinisteriodeEducacioHnNacional,1993).
Basicinstructionalinputsliketextbooks,libra-ries,andotherdidacticmaterialsareoftende"cientinruralschools.
Ruralschoolsmayhavesubstan-dardinfrastructure,lackingevenbasicservicepro-visionlikewaterorlatrines.
Akeydisadvantage,however,isthelowqualityofruralteachers,espe-ciallywhencomparedwiththeirurbancounter-parts(Dove,1982).
Table1showsthatruralteachersinseveralLatinAmericancountrieshavelessformalschoolingonaverage,ausefulthoughimperfectmeasureofteacherquality.
Inanationalsampleof"fth-grademathematicsteachersinColombianpublicschools,51and49%ofurbanteachershadsecondaryandpost-secondaryeduca-tion,respectively(InstitutoSERdeInvestigacioHn,1994).
Amongruralteachers,the"gureswere72and28%.
Urbanandruralteachershadanaverageof17.
5and11.
9yearsofteachingexperience,re-spectively.
InHonduras,ruralteachersscoredlowerontestsofbasicskills(AgenciadeCooperacioHnInternacionaldelJapoHn,1995).
Ingeneral,evidenceisconsistentthatruralschoolsattractteacherswithlowerlevelsofformaleduca-tion,experience,andsubjectknowledge,comparedtourbanschools.
850P.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859Countrieshaveadoptedavarietyofapproachestoproblemsofrecruitment(seeTable2forade-scription).
Theyhaveo!
eredsalarybonuses,sub-sidizedorfreehousing,travelallowancesfortripstourbanareas,subsidiesfortheeducationofteachers'children,andspecialtrainingwhichem-phasizespedagogyusefulintheruralsetting,suchasmultigradeteachingtechniques.
Subsequentsec-tionsexploreingreaterdetailthetheoreticalrationaleforthesepolicies,andtheempiricalmeansofexploringtheire!
ectiveness.
3.
Thetheoryofcompensatingdi4erentialsMostworkonruralteacherrecruitmentisnotexplicitlytheoretical,thoughitre#ectsaneconomicandrationalchoiceapproach.
Whiletheapproachisnecessarilyincomplete,itprovidesausefulstart-ingpointtoformalizeargumentsthathaveremainedatthelevelofintuition.
Thetheoryofcompensatingdi!
erentialshasalongtraditioninlaboreconomics,datingtothewritingsofAdamSmith.
ItissuccinctlydescribedbyChambers(1981;p.
332):Thebasicnotion)))isthatindividualscarebothaboutthequalityoftheirworkenviron-mentaswellasthepecuniaryrewardsassociatedwithparticularemploymentalternatives,andthattheywillseektoattainthegreatestpossiblepersonalsatisfactionbyselectingajobwiththeappropriatecombinationofpecuniaryandnon-pecuniaryrewards.
Similarly,employersarenotindi!
erentastothecharacteristicsoftheindi-vidualstowhomtheyo!
erparticularjobs.
Theresultofthesesimultaneouschoicesisthematchingofindividualemployeeswithem-ployers)))[which]revealsimplicitlythedif-ferentialratesofpayassociatedwiththeattributesofindividualemployeesandthework-ingconditionso!
eredbyemployers.
GeneraldiscussionsofruralincentivescanbefoundinMurnane(1993),Dove(1982),Zymelman&DeStefano(1993),Lockheed&Verspoor(1991),andInternationalLabourO$ce(1991).
Table2ApproachestotheproblemofruralteacherrecruitmentCountryPolicyArgentinaUpto80%bonusonbasesalaryBangladeshSpecialtrainingforruralserviceChilePer-pupilsubsidyforpublicandmostprivateschoolsincreasedinruralandpoormunicipal-itiesColombiaSpecialtrainingforruralserviceCostaRicaSalarybonusandextratimetowardsretire-mentEgyptSeniorityrequirementsreducedby2}4yearswhenapplyingtonewpositions;longervaca-tions;travelallowanceforteacherandfamily;recruitmentoflocalwomenGuyanaAcceleratedpromotionopportunities;travelallowancestopurchaseconsumergoodsHondurasThreeyearsofruralservicecountedas"veyearsincalculationofseniority;25%bonusonbasesalaryIraqFreehousingJamaicaRisetwostepsinpayscaleLibyaPaymentofsometravelandmovingexpensesMexicoReduced-renthousing;insomecaseshousesarebuiltforteachers;compensationfordam-agecausedbycrime;opportunitiesforad-vancedtrainingNicaraguaSpecialtrainingforruralserviceNepalUpto(andoccasionallyexceeding)100%bonusonbasesalaryPakistanHousingclustersforfemaleteachersPhilippinesUpto25%bonusonbasesalarySenegalHousingallowancesSierraLeoneHousingallowancesSyriaLoansforpurchaseofhouse;compulsoryruralserviceatbeginningofcareerVenezuelaTwelveyearsofruralserviceentitlesteacherto20%payraise;specialtrainingforruralserviceZimbabweLoansforpurchaseofhouseInternationalLabourO$ce(1991).
Carnoy&McEwan(1998).
McEwan(1998).
Carnoy&Torres(1994).
SecretarmHadeEducacioHnPuHblica(1996).
Dove(1986).
Lockheed&Verspoor,1991.
ImprovingtheE$ciencyofEducationalSystemsProjectSectorStudy(1988).
P.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859851Ishalldiscusscompensatingdi!
erentialswithinthespeci"ccontextofruralteacherrecruitment,"rstfocusingonthedecisionofteacherstosupplytheirservicestoruralorurbanjobs,andthenonthedemandbyschoolsforteachers.
3.
1.
SupplyofteachersConsiderateacherdecidingbetweentwojobs,identicalexceptthatoneislocatedinaruralarea,theotherurban.
Forthemoment,assumethat`ruralnessaisasinglegood,R,forwhichteachersmustbecompensated(thisunrealisticassumptionisdroppedinsubsequentsectionswhenIdiscussparticularaspectsofruralteaching).
Rassumesvaluesofeither1or0,thelatterimplyinganurbanjob.
Theteachers'preferencesarerepresentedbythefollowingutilityfunction,whichsheseekstomaximize:;"u(=,R).
Utilityisincreasingin=,thewage,anddecreasinginR,theamountof`ruralnessa.
Theteacherwillbeindi!
erentbetweenthetwojobsifu(=0,1)"u(=3,0).
Inotherwords,thereissomewage=0whichissu$cientlyhightocompensateagiventeacherforthedi$cultyofruralservice.
Thepositivedi!
erencebetweentheruralandurbanwages(Z"=0!
=3)representstheteacher'spersonalvalu-ationof`ruralnessa.
Putanotherway,Zisthedi!
erentialrequiredbythatteacherinordertocompensatehimorherforundesirableaspectsofteachinginaruralarea.
NowassumethatRiscontinuous.
Theindi!
er-encecurveofFig.
1mapsoutvariouscombinationsof=andRthattheteacherisequallywillingtoaccept.
AtR"1,theteacherrequiresacompensat-ingdi!
erentialZ.
Assumetheteachercanchoosebetweentwojobs,oneo!
eringawageof=andruralnessR"1,theothero!
ering=withR"0.
Theverticaldistancebetween=and=,de"nedasM,istheruralwagebonuso!
eredtotheteacher.
ThediscussionofteachersupplyanddemandreliesuponRosen(1986)andChambers(1981).
Fig.
1.
Indi!
erencebetweenwagesand`ruralnessaNote:AdaptedfromRosen(1986:646)Fig.
2.
DistributionofZamongteachersNote:AdaptedfromRosen(1986:649)TheteacherwillchoosetheruraljobifM'Z,whichisthecaseinFig.
1.
Thatis,teachersprefertheruraljobiftheprevailingwagemorethancompensatesfortheundesirableaspectsofruraljobs.
Likewise,ifM(Zhewillchoosetheurbanjob.
Thoughallteachersfacethesameruralwagebonus(M),theywillalmostcertainlynotsharethesamepreferencesforlifeandworkinruralareas.
Thisimpliesthateachindividualpossessesauniqueindi!
erencecurvewhichmapsoutthevariouscombinationsofwagesandruralnesswhichareequallydesirabletothatindividual.
AssumethatvaluesofZaredistributedamongprospectiveapplicantsforteachingjobsaccordingtoabell-shapedcurve(seeFig.
2).
NotethatasmallgroupofteachersmayhavenegativevaluesofZ,indicatingthatsomewillrequirecompensationforjobsinurbanareas,contrarytoconventionalwisdom.
Thisconjecturehasdrivensomecountries'attemptstorecruitnativesofruralareas,whoarethoughttorequirelessorevennegativecompensationforru-ralservice.
ThoughthedistributionofFig.
2is852P.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859symmetrical,thereisnoguaranteethatthisisthecase.
Infact,therecouldbeathickertailextendingfarrightward,indicatingthatmanyteacherswouldrequireexorbitantcompensationinordertocon-sideraruralteachingjob.
Thisisonewayofinterpretingpreviousresearchers'assertionsthatincentivessimplywouldnotfunction.
Theprevailingruralwagebonus(M)islocatedalongthehorizontalaxis.
AllteacherswhosevaluesofZaregreaterthanMwillsupplytheirlabortourbanjobs,sincetheprevailingMisnotsu$cienttocompensatethemforruralservice.
Likewise,workerswithZ(Mwillchooseruraljobs.
AnincreaseinMshouldincreasethesupplyofteacherstoruraljobs.
3.
2.
DemandforteachersSchooldecision-makersarefacedwiththetaskofo!
eringacombinationofmonetaryandnon-mon-etaryincentivesthatappropriatelycompensatesteachersfortheassumingaruralteachingjob.
Assumethataruralschooldecision-makerfacesamarginalcostperteacherofreducingthe`rural-nessaofateachingjob,throughinvestinginnon-monetaryfactorssuchasimprovedteacherhousing,schoolinfrastructure,in-servicetraining,travelsubsidies,andsoforth.
De"nethisasB.
Liketeachersintheprevioussection,theschoolcom-paresBtoM,theprevailingwagebonusestab-lishedinthelabormarket.
IfB'M,theschoolsimplychoosestopayMasamonetaryinduce-ment,ando!
eracompletely`ruralajob.
Likewise,ifB(Mtheschoolinvestsinapackageofnon-monetaryincentivesfortheteacherandtries,totheextentpossible,toreduceorameliorateelementsof`ruralnessaattachedtothejob.
Di!
erentvaluesofBaredistributedacrossschools.
Someschoolswill"nditmorecostlytoinvestinnon-monetaryincentivesforteachersandwillthereforehavehighervaluesofB.
Forinstance,schoolmaintenanceorserviceconnectioncostsmaybehigherforespeciallyisolatedschools.
Fig.
3presentsabell-shapeddistributionofB;caveatsoftheprevioussectionregardingtheshapeofdis-tributionsapplyhereaswell.
SchoolstotherightofMwillo!
ercompletely`ruralajobsandpayFig.
3.
DistributionofBamongschoolsNote:AdaptedfromRosen(1986:652)amonetaryincentive.
Schoolstotheleftwillinvestinnon-monetaryincentivesinorderto`urbanizeatheirjobsasfaraspossible.
Ofcourse,theremaybealimittotheextentthatruralschoolscanattractgoodteachersbysimplyimprovingnon-monetaryincentives(Chambers,1981).
Thisisparticularlysoiftherearenon-linearitiesinthejobcharacteristicsofinterest.
Forexample,teachersmaygreatlyvalueasmallerclasssizeinsomeinitialrangeofvalues*say,40to20*andbewillingtotradeo!
agoodportionofmonetarywages.
Afterthat,however,theymayvaluesmallerclasssizelessandless.
Thereisequilibriumintheteacherlabormarketwhenthereisequalityofdemandandsupplyforteachersoneachtypeofjob.
Inacompetitivelabormarket,thevalueofMadjustsinordertoequalizethepartitionsofFig.
2andFig.
3(Rosen,1986).
Ofcourse,teacherlabormarketsindevelopingcoun-triesarerarelyperfectlycompetitive,withwagesadjustinginresponsetoshiftingdemandandsup-ply.
TeacherwagesandruralincentivesareoftenrigidlygovernedbyMinistriesofEducation.
Ifwagesandotherincentivesareinsu$cient,fewteacherssupplytheirservicestoruralteachingjobs,ortheteachersthatdoapplyforruraljobswillbelessquali"ed.
Thoughlocalschoolmanagersmayattempttocompensatebyinvestinginnon-monet-aryincentivesorwagebonusesabovethestandardpayscale,theyareinvariablyconstrainedbycen-tralizedsystemsofschool"nancewhichprovidefewdiscretionaryresourcestolocaldecision-makers.
4.
Themeaningof`ruralnessaThissectiondelineateswhatruralteachersarebeingcompensatedfor*what,inotherwords,mayP.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859853beundesirableaboutlivingandworkinginaruralarea.
Itreinforcesthepointthatnotwocountries'incentivepackageswillbethesame,ifonlybecausetheirruralareasandschoolswilldi!
erwidelyinthecharacteristicsoutlinedbelow,aswellasteacherpreferencesforsuchcharacteristics.
Characteristicsofruralschoolsandareasaredividedintotwogroups:alterableandinalterablefactors.
Alterablefactorsarethoseelementscommonlyassociatedwithruralschoolswhichmaybeunderthecontrolofschoolpolicy-makers,suchastheavailabilityoftrainingopportunitiesorclasssizes.
Thekeyim-plicationisthatpolicy-makerscouldchoosetoalterthesecharacteristicsinanattempttomakeruralteachingamoreattractiveoption,ratherthansimplyapplyingmonetaryincentives.
Inalterablefactors,ontheotherhand,areelementsofruralareaswhichareimmutableandbeyondthereachofschoolo$cials,suchasclimate,crime,orphysicalisolation.
4.
1.
AlterablefactorsRuralschoolstypicallyisolateteachersfromop-portunitiesforinteractionwiththeirpeers,profes-sionaldevelopment,andpromotionthroughthesystem.
Somecountrieshaveputinnovativepro-gramsinplacetocombatsuchisolation.
Forin-stance,Malihascreatedaprofessionalbulletinforruralteachers(Thomas&Shaw,1992).
Colombiahasinstituted`microcentersaforsometeachers,wheretheycanmeettoexchangeideasandexperi-ences(Colbertetal.
,1993;Schiefelbein,1992).
Anumberofcountries(seeTable2)haveallowedruralservicetoincreaseteachers'seniority,therebyquickeningthepaceofpromotion.
Housingofappropriatequality,particularlyforsinglewomen,maybedi$cultto"ndinisolatedareas.
Manycountrieshaveresortedtoprovidingsubsidizedorfreehousingtoteachers,particularlyforwomen(seeTable2).
Insomecases,housingislocatedontheschoolpremises.
Chambers&Fowler(1995)makeasimilardistinction.
BothDove(1982)andMurnane(1993)includeageneraldiscussionofsuchfactors.
Schoolandclassroomcharacteristicscoulda!
ectthequalityoftheteacher'sworkenvironment.
Thestudent}teacherratio,thoughrarelylinkedtostu-dentachievementindevelopingcountries(Fuller&Clarke,1994),signi"cantlya!
ectshowdi$cultaclassistomanage.
Thoughstudent}teacherratiosinsomeruralschoolsareexceptionallylargerthanurbanschools,sometimesthesituationisquitere-versed,whenlowpopulationdensitiesimplylowratios(InternationalLabourO$ce,1991).
Eitherwouldtendtoa!
ecttherelativeattractivenessofruralteachingopportunities.
Theabsenceofbasicschoolinputsandinfrastructuresuchastextbooks,alibrary,blackboards,electricity,sanitaryservice,andwatermakestheteacher'staskextremelychal-lenging.
Suchde"cienciesareoftenexacerbatedinruralschools.
Finally,ruralteachersmustoftenteachinamultigradeenvironment*thatis,class-roomswherestudentsareheterogeneousinageandability.
Researchersgenerallyconcurthatmultigradeteachingrequiresgreatertrainingandpreparationoutsidetheclassroomthanstandardpedagogy(Bray,1987;Benveniste&McEwan,1996),andthereforemayrequireincentives.
Afewcountrieshavemadeintensivee!
ortstoendowmultigradeteacherswithspecialtrainingandma-terials,suchasColombiaanditsEscuelaNuevaprogram(Colbertetal.
,1993;McEwan,1998).
4.
2.
InalterablefactorsIfruralteachersarenativesofurbanareas,theymaybeisolatedfromfriendsandfamilies.
Inothercases,teachersmaynotsharethelanguageorcul-tureofthecommunity,andmaybeconsideredoutsiders.
Singleteachersmighthaveamoredi$-culttime"ndingaspouse.
Theinfrastructurewhichmightreducesuchisolation,suchastelephones,publictransportation,orroads,isofteninpoorrepairifavailable.
Manycountrieshavesoughttorecruitteachersnativetothecommunitiesinwhichtheywillteach*therationalebeingthatsuchteacherswill"ndconditionsmorefamiliarandlessonerous,thoughevidenceonthispointisscarce.
Ruralareaslackseveralamenitiesmorecommontourbanareas,whichcoulda!
ectthedesirabilityorcostofrurallife.
Theseincludeeducational854P.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859opportunitiesforthechildrenofteachers;employ-mentopportunitiesfortheteacher'sspouse;andculturalattractionslikerestaurantsandtheaters.
Sometimesruralteachersaresubjecttopersonaldangers,becausetheyaregovernmentrepresenta-tivesorsimplythemostprivilegedlocalcitizens.
NotethatMexicoo!
ersruralteachercompensa-tionfordamagescausedbybandits(Table2).
InColombia,teachersoftenworkinzonesofarmedcon#ictwithnegligiblegovernmentpresence,andtheirstudentsmayberelatedtoarmedparticipantsononesideofthecon#ictoranother(McGinn,1996;Loera&McGinn,1992).
Finally,teachersmayacquiresatisfactionordis-comfortfromthecharacteristicsofthestudentstheyteach,suchassocioeconomicstatus,oftencorrelatedwithrurallocation.
Foravarietyofreasons,stu-dentsoflowersocioeconomicstatusmaybeabsentmoreoften,requiremoree!
orttoteachwhentheyarepresent,orbesubjecttoadditionalbehavioralproblems(Chambers&Fowler,1995).
5.
Empiricalevidenceoncompensatingdi4erentialsTheprecedingsectionsdescribedaframeworkforanalyzingincentives,andspeci"cruralcharac-teristicswhichmayrequireincentives.
Stilllackingisameansofempiricallycorroboratingwhetherteachersimplicitlyvaluethecharacteristicsde-scribedabove,andhowmuchcompensationisre-quired.
Thissectiondescribesthelimitedempiricalevidenceoncompensatingdi!
erentialsinteacherlabormarkets,andgivessomeguidelinesforitsinterpretation.
5.
1.
PriorstudiesEmpiricalresearchineconomicshastestedthetheoryofcompensatingdi!
erentialsinseverallabormarkets,usuallyfocusingonindustrialworkersandjobsafety.
Researchhasexploredifandtowhatextentworkersappeartotradeo!
wagesagainstincreasedrisksofinjuryordeathonthejob.
Inessence,thesestudiesestimatethepre-Polachek&Siebert(1993)reviewthisliterature.
vailingbonusinthelabormarketforassumingjobswithrelativelyhigherrisks.
Afewresearchershaveconductedsimilarstudiesofteacherlabormarkets,invariablyintheUnitedStates.
Theyestimatethedegreetowhichteachersimplicitlyvalueaspectsoftheirjobsandlocations,usingvariablesthatareintendedtoproxymanyofthecharacteristicsde-scribedintheprevioussection.
Researchersemployasimilarempiricalstrategy,usingmultivariatere-gressiontoestimatewageequationsofthefollow-ingform:Wage"f(teachercharacteristics,jobcharacter-istics,locationalcharacteristics).
Usingordinaryleastsquares,thewagesofteachersareregressedonavarietyofcharacteristicsthoughttoa!
ectteacherproductivity,suchasedu-cation,experience,andacademicability.
Aseriesofadditionalindependentvariablesisaddedwhichproxythejoborlocationalcharacteristicsofinter-est.
Signsandmagnitudesofthelattervariables'regressioncoe$cientsareexaminedinordertodeterminewhetherworkersattachimplicitvaluestothosecharacteristicsinthelabormarket.
Forexample,apositivecoe$cientimpliesthatteachersarepaidhigherwageswheretheyconfrontgreateramountsofaparticularjobcharacteristic,holdingconstantteacherandotherjobcharacter-istics.
Antos&Rosen(1975)foundthatwhiteteachersintheUnitedStatesrequireadditionalcompensa-tioninordertoassumeteachingpositionsinschooldistrictscomposedofmainlyblackstudents.
Levin-son(1988)foundthesamee!
ect,thoughitdisap-pearedwhenadditionalindependentvariablessuchasstudentpovertyandtestscores*correlatedwithrace*wereaddedtotheregression.
Thisseemedtoindicatethatteachersmaysimplypreferteachingwealthier,better-achievingstudents,butnotnecessarilywhitestudents.
AntosandRosendidincludelocationalvariables,thoughtheirre-sultsweremixed.
Theyfoundwhitefemaleteachersrequiredasmallpremiumtoteachinruralareas,relativetotheinnercity,whileblackfemalesap-pearedtorequireapremiumnottoteachinruralareas.
Kenny&Denslow(1980)estimatedwageequa-tionsforUSteachersthatincludedindependentP.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859855variablesmeasuringclasssize,crimerates,thede-sirabilityoftheclimate,andthecostofliving.
Theyfoundseveralexpectede!
ects:thatwagesadjustdownwardinthepresenceofsmallerclasses,lesscrime,amoredesirableclimate,andalowercostofliving.
Harbison&Hanushek(1992),whilenotintendingtostudycompensatingwagedi!
erentials,doestimatewageregressionsforasampleofruralBrazilianteachers.
They"ndthatthestudent}teacherratioispositivelycorrelatedwithwages,allelseequal,perhapsindicatingthatteacherspreferworkinginsmallerclasses.
Thismustbeinterpretedcarefully,sincesmallclasssizesareprobablymoreprevalentinmoreisolatedareasduetolowerpopulationdensitiesandtheydonotcontrolforotheraspectsof`ruralnessaintheirequation.
ThemostcomprehensivestudyisthatofChambers&Fowler(1995),notableforitsexhaust-iveattemptstoidentifytherelevantnon-wagechar-acteristicsofteachers'jobsandconstructcarefulempiricalproxies.
Theyanalyzedanationwidesampleof40,484USpublicschoolteachers.
They"ndthatforteacherswhoaremorethan160milesawayfromtheclosestcity,salariestendtobearound6%higherthanthoseoftheurbanrefer-encegroup,holdingallelseequal.
Theyalso"ndcompensatingdi!
erentialsforclimate(gaugedbymeantemperaturesandsnowfalls)andthenumberofviolentcrimespercapita,aswellasamodesttrade-o!
betweenhigherwagesandlargerclasssizes.
5.
2.
EmpiricalchallengesThereareseveralchallengestotheestimationandinterpretationofthemodelsdescribedabove.
First,themeasurementof`ruralnessaandruraljobcharacteristicsisnotstraightforwardandmustbeadjustedtoparticularcontexts.
Theimmediatein-stinctmightbetomeasureruralcharacteristicswithasimpledichotomousvariable(ruralvs.
ur-ban).
Ihavealreadyarguedthatdoingsowouldobscureimportantaspectsofruralnessthatmaybeimportanttoteachers.
Avarietyofempiricalpro-xiescouldbedevisedindi!
erentcountrieswhichre#ecttherelevantcharacteristicsdescribedinpre-vioussections.
Second,noneofpreviousstudiesofteacherlabormarketshaveaccountedforthepossiblebiasesinducedbyimperfectlymeasuringtheproductivityofindividualteachers.
Teachersmaybeproductiveinwaysthatareunobservablebytheresearcher.
Nonetheless,suchvariablesbelonginthewageregressionequation.
Ifomittedproductivitymeasuresarecorrelatedwithruralcharacteristicsandwithwages,thenthecoe$cientsofruralvari-ableswillbebiased.
Forexample,teachersmayselectruralpositionswhohaveapreferenceforshirking.
Ruralschoolsare,ingeneral,subjecttolesssupervisionbycentraleducationalauthorities(Keith,1989),andtheymaylackon-siteprincipalscapableofobservingteacherbehavior.
Moreover,parentsmayhavelittleauthoritytomonitorandsanctionteacherbehavior,especiallyincentralizedsystems.
Iftheunmeasuredpropensitytoshirkispositivelycorrelatedwiththeruralnessofschoolsandnegativelycorrelatedwithwages,thenruralcoe$cientswilllikelybebiaseddownward.
ThepossiblebiasfromomittingteacherproductivitymeasuresisillustratedinFig.
4.
Severalresearchersineconomicshaveattemptedtoresolvetheseprob-lemswithadditionaleconometricmethods,includ-inginstrumentalvariables(Garen,1988)andtheuseoflongitudinaldataonindividuals(Brown,1980;Duncan&Holmlund,1983).
Perhapsthebestsolutionisobtainingbetterdataonthebackgroundandabilitiesofteachers.
Third,compensatingdi!
erentialswhichareob-servedamongcurrentteachingsta!
maynotbeanadequateguidetotheincentivesrequiredtoattractadditionalorbetter-quali"edteacherstoruralareas.
Assumethattheempiricalanalysisproducesunbiasedestimatesofthecompensatingdi!
eren-tialsrequiredbycurrentteacherstoliveandworkinruralareas.
Butthe"rstteacherstochooseruraljobsarethosewiththesmallestvaluesofZ*inotherwords,thosewhorequirerelativelysmallercompensationtoassumeruraljobs(Rosen,1986;Foracompletediscussionofcompensatingdi!
erentialsandunobservedproductivity,seeHwang,Reed&Hubbard(1992).
TheproblemofteacherabsenceisonemotivationbehindprogramssuchasEDUCOinElSalvador,whichpromotescommunity-managementofschools(Jimenez&Sawada,1998).
856P.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859Fig.
4.
E!
ectsofomittedproductivityvariablesonestimationofcompensatingdi!
erentialsNote:AdaptedfromPolachekandSiebert(1993:189)Polachek&Siebert,1993).
Thereisatemptationtointerprettheregressioncoe$cientsasthepremiumthatnon-ruralteacherswouldrequiretomove.
However,suchteachers(thosecurrentlyinurbanjobs)probablyhavehighervaluesofZ.
Thus,theestimatedcompensatingdi!
erentialsareonlylower-boundstopremiumsthatmightactuallyberequiredifadditionalteachersaretoberecruitedinruralschools.
Fourth,theinterpretationofempiricalresultsmusttakeplacewithinthecontextofacountry'sinstitutions.
Assumeonceagainthatunbiasedesti-matesareobtained,butthatnoevidenceofcom-pensatingdi!
erentialsisfound.
IsthisnecessarilyevidencethatteachersdonotrequirecompensationforruralteachingpostsInfact,itmaysimplyindicatethatanincentiveprogramwasgraftedontodysfunctionalinstitutions,whichwerein-capableofselectingthebestteachersfromavailableapplicants.
Incentivesmayprovee!
ectiveatat-tractingalargebodyofquali"edapplicants.
ButinHonduras,forexample,politicalinterferenceandcorruptionoftenresultinless-quali"edteachersbeingdrawnfromtheavailablepoolofapplicants.
Hanson(1986)suggeststhatlocalpoliticiansinColombiaholdconsiderablein#uencein"llingru-ralteachingvacancieswithfavoredapplicants.
Incasessuchasthese,higherpayresultingfromincen-tiveswillnotnecessarilyleadtoamorequali"edteachingsta!
.
6.
SummaryandimplicationsIncentivesforruralteachersareapervasiveas-pectofeducationalsystemsindevelopingcoun-tries.
Nevertheless,thereislittleevidencewhichcouldguidetheimplementationandevaluationofsuchpolicies.
A"rststeptosuchananalysisisbeingmoreexplicitaboutthetheoreticalassump-tionswhichareembeddedinpolicies,principallythatteachersarerationalutilitymaximizers.
Isug-gestedthattheeconomictheoryofcompensatingdi!
erentialsisausefultoolforthinkingaboutin-centivesandforguidingempiricalstudy.
Thefewempiricalstudieswhichexist,mainlyintheUnitedStates,givelimitedsupporttotheideathatteacherstradeo!
monetarywagesagainstsomenon-monet-aryaspectsofjobsandtheirlocation.
Forinstance,teachersappeartovalueworkinginschoolswherethestudentsareofarelativelyhighersocio-economicstatusandwhicharerelativelyclosertourbanareas.
Muchremainstobedone,however.
Whereteacherincentivesareconcerned(andteacherlabormarketsingeneral),thereisanenormousgapbe-tweenpolicyandresearchindevelopingcountries.
Wagebonusesandnon-monetaryincentivesareimplementedwithlittleregardtotheire!
ectivenessinattractingteachers.
Moreover,itisunclearwhethernon-monetaryincentives,suchason-sitehousing,mightattainthesameendsinamorecost-e!
ectivemanner.
Assume,forexample,thata$100wagebonusandareducedclasssizebothcontributeequallytowardsattractingagiventeacher(thatis,theteacheriswillingtotradeo!
$100ofwagesagainsttheprospectofteachingasmallerclass).
Butsupposethatthemonetarycostofreducingtheclasssizeisonly$85,whichsuggeststhatitisthemorecost-e!
ectivemeansofrecruitingteachers.
Thoughonlyillustrative,ithighlightsthenecessityofdesigningapackageofincentives*monetaryandnon-monetary*whichmaximizesrecruitmentbene"tsataminimalcosttotheeduca-tionalsystem.
ThiscanonlybeaccomplishedP.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859857throughcountry-speci"cempiricalworkofthesortdescribedintheprevioussection,becauseschool-ingcontextsandteacherpreferenceswilldi!
erwidelyacrosscountries.
Increasingly,thereisthepotentialtocollectthenecessarydataforsuchempiricalwork,ortotakeadvantageofexistingdatasetscollectedforotherpurposes.
Chile,forexample,conductedateachercensusin1996whichgatheredawiderangeofdataonteacherquali"cationsandbackground.
Ifitweretoincorporateadditionalinformationonsalaries,thedatacouldeasilybeusedtostudycompensat-ingdi!
erentials.
Brazilconductedarecentnation-widesurveyofteacherswhichdoesincludeswagedata.
OthercountriessuchasColombiahaveongo-ingprojectstocollectdataonstudentachievementandcharacteristicsoftheirteachersandschools.
Themainintentistoconductschoole!
ectivenessorproductionfunctionstudies,buttheaccompany-ingteacherquestionnairesarearichsourceofdataonteachercharacteristicsandcontexts.
Datasuchastheseholdgreatpotentialtoshedlightonthee$cacyofcurrentrecruitmentpoliciestowardruralteachersindevelopingcountriesandprovideadditionalguidepoststopolicy-makers.
Itishopedthatthispaperwillspuradditionaltheor-eticalandempiricalresearchinthisrelativelyne-glectedarea.
AcknowledgementsThanksareduetoMartinCarnoy,CyrusDriver,JulieScha!
nerandtwoanonymousrefereesfortheircomments.
ReferencesAgenciadeCooperacioHnInternacionaldelJapoHn(1995).
Inves-tigacio&nsobreelniveldeconocimientodelosdocentesdelosnivelesprimarioymediodeldistritocentralenlasasignaturasdematema&tica,cienciasnaturalesyesudiossociales.
Tegucigalpa:AgenciadeCooperacioHnInternacionaldelJapoHn.
Fuller&Clarke(1994)summarizemanyofthesestudies.
Antos,J.
R.
,&Rosen,S.
(1975).
Discriminationinthemarketforpublicschoolteachers.
JournalofEconometrics,3(2),123}150.
Benveniste,L.
,&McEwan,P.
(1996).
Constraintstoimplemen-tingmultigradeschools:theroleoflocalcapacityandwill.
Unpublishedmanuscript.
Bray,M.
(1987).
AresmallschoolstheanswerCost-ewectivestrategiesforruralschoolprovision.
London:TheCommon-wealthSecretariat.
Brown,C.
(1980).
Equalizingdi!
erencesinthelabormarket.
QuarterlyJournalofEconomics,94(1),113}134.
Carnoy,M.
,&McEwan,P.
J.
(1998).
Isprivateeducationmoree!
ectiveandcost-e!
ectivethanpublicThecaseofChile.
Unpublishedmanuscript,StanfordUniversity.
Carnoy,M.
,&Torres,C.
A.
(1994).
Educationalchangeandstructuraladjustment:acasestudyofCostaRica.
InJ.
Samo!
(Ed.
),Copingwithcrisis:austerity,adjustmentandhumanresources.
London:Cassell,UNESCO.
Chambers,J.
G.
(1981).
Thehedonicwagetechniqueasatoolforestimatingthecostsofschoolpersonnel:Atheoreticalexposi-tionwithimplicationsforempiricalanalysis.
JournalofEdu-cationFinance,6(3),330}354.
Chambers,J.
,&Fowler,W.
J.
(1995).
PublicschoolteachercostdiwerentialsacrosstheUnitedStates.
Washington:U.
S.
Gov-ernmentPrintingO$ce.
Colbert,V.
,Chiappe,C.
,Arboleda,J.
(1993).
TheNewSchoolprogram:moreandbetterprimaryeducationforchildreninruralareasinColombia.
InH.
Levin,&M.
Lockheed(Eds.
),Ewectiveschoolsindevelopingcountries.
London:TheFalmerPress.
DepartamentoAdministrativoNacionaldeEstadmHstica(DANE).
(1993).
LasestadmHsticassocialesenColombia.
BogotaH:DANE.
Dove,L.
A.
(1982).
Thedeploymentandtrainingofteachersforremoteruralschoolsinless-developedcountries.
Interna-tionalReviewofEducation,27(3),3}27.
Dove,L.
A.
(1986).
Teachersandteachereducationindevelopingcountries.
London:CroomHelm.
Duncan,G.
J.
,&Holmlund,B.
(1983).
WasAdamSmithrightafterallAnothertestofthetheoryofcompensatingwagedi!
erentials.
JournalofLaborEconomics,1(4),366}379.
Fuller,B.
,&Clarke,P.
(1994).
Raisingschoole!
ectswhileignoringcultureLocalconditionsandthein#uenceofclass-roomtools,rules,andpedagogy.
ReviewofEducationalRe-search,64(1),119}157.
Garen,J.
(1988).
Compensatingdi!
erentialsandtheendogeneityofjobriskiness.
ReviewofEconomicsandStatistics,70(1),9}16.
Hanson,E.
M.
(1986).
Educationalreformandadministrativede-velopment:thecasesofColombiaandVenezuela.
Stanford:HooverInstitutionPress.
Harbison,R.
W.
,&Hanushek,E.
A.
(1992).
Educationalperfor-manceofthepoor:lessonsfromnortheastBrazil.
Oxford:OxfordUniversityPress.
Hwang,H.
,Reed,W.
R.
,&Hubbard,C.
(1992).
Compensatingdi!
erentialsandunobservedproductivity.
JournalofPoliticalEconomy,100(4),835}858.
858P.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859InstitutoSERdeInvestigacioHn(1994).
Sistemanacionaldeevalu-acio&ndelacalidaddelaeducacio&n:calendarioA,1993.
BogotaH:InstitutoSERdeInvestigacioHn.
InternationalLabourO$ce(ILO)(1991).
Teachersindevelopingcountries:asurveyofemploymentconditions.
Geneva:ILO.
Jimenez,E.
,&Sawada,Y.
(1998).
Docommunity-managedschoolsworkAnevaluationofElSalvador'sEDUCOpro-gram.
WorkingPaperSeriesonImpactEvaluationofEduca-tionReforms,No.
8.
Washington:WorldBank.
Keith,S.
(1989).
Improvingsupportforruralschools:amanage-mentperspective.
Policy,Planning,andResearchWorkingPaper302.
Washington:TheWorldBank.
Kenny,L.
W.
,&Denslow,Jr.
D.
A.
(1980).
Compensatingdi!
er-entialsinteachers'salaries.
JournalofUrbanEconomics,7,198-207.
Levinson,A.
M.
(1988).
Reexaminingteacherpreferencesandcompensatingwages.
EconomicsofEducationReview,7(3),357}364.
Lockheed,M.
E.
,&Verspoor,A.
M.
(1991).
Improvingprimaryedu-cationindevelopingcountries.
Oxford:OxfordUniversityPress.
Loera,A.
,&McGinn,N.
(1992).
Larepitenciadegradoenlaescuelaprimariacolombiana,EducationDevelopmentDis-cussionPapers.
Cambridge,MA:HarvardInstituteforInter-nationalDevelopment.
McEwan,P.
J.
(1998).
Thee!
ectivenessofmultigradeschoolsinColombia.
InternationalJournalofEducationalDevelopment,18(6),435}452.
McGinn,N.
(1996).
Requirementsforcapacitybuilding:EscuelaNuevainColombia.
Mimeo.
MinisteriodeEducacioHnNacional(MEN)(1993).
Primerosre-sultados:matematicasylenguajeenlabasicaprimaria,ColecioHnDocumentosdelSaberNo.
1.
BogotaH:MEN.
Murnane,R.
J.
(1993).
Economicincentivestoimproveteaching.
InJ.
P.
Farrell,J.
B.
Oliveira(Eds.
),Teachersindevelopingcountries:improvingewectivenessandmanagingcosts.
Washington:WorldBank.
Polachek,S.
W.
,&Siebert,W.
S.
(1993).
Theeconomicsofearn-ings.
Cambridge:CambridgeUniversityPress.
Psacharopoulos,G.
,Valenzuela,J.
,Arends,M.
(1993).
Teachers'salariesinLatinAmerica:Acomparativeanalysis.
WPS1086.
Washington:TheWorldBank.
Rosen,S.
(1986).
Thetheoryofequalizingdi!
erences.
InO.
AshenfelterandR.
Layard(Eds.
),Handbookoflaborecon-omics,volume1.
Amsterdam:ElsevierSciencePublishersBV.
Schiefelbein,E.
(1992).
RedexningbasiceducationforLatinAmerica:lessonstobelearnedfromtheColombianEscuelaNueva.
Paris:UNESCO,InternationalInstituteforEduca-tionalPlanning.
SecretarmHadeEducacioHnPuHblica(1996).
Estudionacionalsobrelaformacio&ninicialdedocentesdelaeducacio&nprimaria.
Tegucigalpa:SecretarmHadeEducacioHnPuHblica.
Thomas,C.
,&Shaw,C.
(1992).
Issuesinthedevelopmentofmultigradeschools.
WorldBankTechnicalPaperNumber172.
Washington:TheWorldBank.
Zymelman,M.
,&DeStefano,J.
(1993).
Factorsa!
ectingteachers'salaries.
InJ.
P.
Farrell,J.
B.
Oliveira(Eds.
),Teachersindevelopingcountries:improvingewectivenessandmanagingcosts.
Washington:TheWorldBank.
P.
J.
McEwan/TeachingandTeacherEducation15(1999)849}859859