Lucas是什么意思as在线翻译读音例句

MakingaMiracle

Author(s):,Jr.

Source:Econometrica,Vol.61,No.2(Mar.,1993),pp.251-272

Publishedby:TheEconometricSociety

StableURL:/stable/2951551

Accessed:07/12/201013:08

YouruseoftheJSTORarchiveindicatesyouracceptanceofJSTOR\'sTermsandConditionsofUse,availableat

/page/info/about/policies/\'sTermsandConditionsofUseprovides,inpart,thatunless

youhaveobtainedpriorpermission,youmaynotdownloadanentireissueofajournalormultiplecopiesofarticles,andyou

mayusecontentintheJSTORarchiveonlyforyourpersonal,non-commercialuse.

hercontactinformationmaybeobtainedat

/action/showPublisher?publisherCode=econosoc.

EachcopyofanypartofaJSTORtransmissionmustcontainthesamecopyrightnoticethatappearsonthescreenorprinted

pageofsuchtransmission.

JSTORisanot-for-profitservicethathelpsscholars,researchers,andstudentsdiscover,use,andbuilduponawiderangeof

nformationtechnologyandtoolstoincreaseproductivityandfacilitatenewforms

einformationaboutJSTOR,pleasecontactsupport@.

TheEconometricSocietyiscollaboratingwithJSTORtodigitize,preserveandextendaccesstoEconometrica.

Econometrica,Vol.61,No.2(March,

1993),251-272

MAKINGAMIRACLE1

,JR.

Thislecturesurveysrecentmodels

ofgrowthandtradeinsearchofdescriptions

of

technologiesthatareconsistent

withepisodesofveryrapid

isis

placedontheon-the-jobaccumulation

ofhumancapital:learningby

le

connectionsbetweenlearning

ratesandinternationaltradeare

discussed.

KEYWORDS:Growth,productivity,on-the-jobtraining,

learning.

UCTION

IN1960,THE

PHILIPPINES

ANDSOUTH

KOREAhadaboutthesamestandard

of

living,asmeasuredbytheir

percapitaGDPsofabout

$

countriesweresimilar

were28millionpeople

in

thePhilippinesand25million

inKorea,withslightly

overhalfofbothpopula-

sevenpercentof

Filippino\'slivedinManila,28

percent

countries,

allboysofprimary

school

agewereinschool,

andalmostallgirls,butonlyabouta

quarterofsecondary

schoolagechildrenwereinschool.

Only5percent

ofKoreansinthei沐猴而冠意思 rearly

twentieswereincollege,as

comparedto13percent

sixpercentofPhilippineGDP

wasgeneratedinagriculture,and28percent

in

a,thecomparable

numberswere37and20percent.

Ninetysix

percent

ofPhilippinemerchandiseexports

consistedofprimarycommodities

and4percentofmanufactured

a,primary

commoditiesmadeup

86percentofexports,andmanufactured

goods14(ofwhich8were

textiles).

From1960to1988,GDPper

capitainthePhilippinesgrew

atabout1.8

percentperyear,

abouttheaverageforpercapita

incomesintheworldas

a

a,overthe

sameperiod,percapitaincomegrew

at6.2percent

peryear,

arateconsistentwiththedoubling

oflivingstandardsevery11years.

Koreanincomesarenowsimilar

toMexican,Portuguese,

orYugoslavian,about

threetimesincomesinthe

Philippines,andabout

onethirdofincomesinthe

UnitedStates.2

Idonotthinkitisinanyway

anexaggerationtoreferto

thiscontinuing

transformationofKoreansocietyasa

miracle,ortoapply

thistermtothevery

similartransformations

thatareoccurringinTaiwan,HongKong,andSinga-

beforehavethelivesofsomanypeople

(63

millioninthesefour

areasin1980)undergone

sorapidanimprovement

oversolongaperiod,

nor

(with

thetragicexception

ofHongKong)isthereanysign

thatthisprogressis

1

Preparedforthe1991Fisher-Schultz

Lecture,giveninSeptember

attheEuropeanmeetings

of

theEconometricSociety.

IamgratefultoJoseScheinkman,

z,NancyStokey,

Alwyn

Young,andtherefereesfor

discussionandcriticism.

2Thefiguresinthefirstparagraph

aretakenfromthe1984World

incomeandpopulationfiguresin

thisparagraphandthenextarefrom

SummersandHeston(1991).

251

252ROBERT

,

JR.

ithappen?WhydidithappeninKoreaandTaiwan,and

notinthePhilippines?

efultobegin

simplybylistingsomeofthefeaturesofthesetransformationsinadditionto

heEastAsianmiracleeconomieshave

becomelargescaleexportersofmanufacturedgoodsofincreasingsophistica-

vebecomehighlyurbanized(noproblemforSingaporeandHong

Kong!)ve

pro-businessgovernments,followingdifferingmixesoflaissezfaireandmercan-

acts-or

atleastsomeofthem-must

figure

inanyexplanationofthegrowthmiracles,buttheyareadditionstothelistof

eventswewanttoexplain,notthemselvesexplanations.

Wewanttobeabletousetheseeventstohelpinassessingeconomicpolicies

plyadvisingasocietyto

\"followtheKoreanmodel\"isalittlelikeadvisinganaspiringbasketballplayer

to\"followtheMichaelJordanmodel.\"Tomakeuseofsomeoneelse\'ssuccess-

fulperformanceatanytask,oneneedstobeabletobreakthisperformance

downintoitscomponentpartssothatonecanseewhateachpartcontributesto

thewhole,whichaspectsofthisperformanceareimitableand,ofthese,which

ds,

in

short,atheory.

Therehasbeenagreatdealofinterestingnewtheoreticalresearchongrowth

anddevelopmentgenerallyinthelastfewyears,someofitexplicitlydirectedat

theAsianmiraclesandmuchmorethatseemstomeclearlyrelevant.

Iwilluse

thislecturetotryandseewhatrecentresearchoffers

towardanexplanationfor

ewwillbesharplyfocusedonneoclassicaltheoriesthat

ppenedoverthelast

30yearsthatenabledthetypicalKoreanoroverseasChineseworkertoproduce

6timesthegoodsandserviceshecouldproducein1960?Indeed,myviewpoint

willbeevennarrowerthantheneoclassicaltheoriesonwhichIdraw,since

I

intendtofocusonissuesoftechnology,withonlycursorytreatmentofconsumer

s

nodoubt

thattheissueofwhogetstherewardsfrominnovationisacentralone,anditis

notonethatcanberesolvedonthebasisoftechnologicalconsiderationsalone,

sothisnarrowfocuswillnecessarilyrestricttheconclusionsIwillbeableto

reisnopointintryingtothinkthroughhardquestions

of

industrialorganizationandgeneralequilibriumwithoutanadequatedescription

oftherelevanttechnology,sothisseemstometherightplacetostart.

IwillbegininSection2,\'withabriefsketchofsomerecenttheoretical

developments

andoftheimageoftheworldeconomythesedevelopmentsoffer.

Thisimagedoesnot,asIseeit,admitofanythingonecouldcallamiracle,but

itwillbeusefulinmotivatingmysubsequentemphasisontheaccumulation

of

humancapital,andinparticularonhumancapitalaccumulationonthejob:

learningbydoing.

InSection

3,

I

willreviewapieceofmicroeconomicevidence

onlearningandproductivity,justtoremindyouhowsolidtheevidenceisand

howpromising,quantitatively,ablishingthe

MAKINGAMIRACLE253

importanceoflearningbydoingforproductivitygrowthonaspecificproduction

processisverydifferentfromestablishingitsimportanceforanentireeconomy

asawhole,nnectionismuchmore

asbeenmade,inresearchbyNancy

StokeyandAlwynYoung,andIwillsketchthemaintechnologicalimplications

sgoodreasontobelieve,Iwillargue,that

somethinglikethistechnologyprovidedthemeansfortheproductivitymiracles

n5discussessomeoftheissuesinvolvedindevelopingmarket

equilibriumtheoriesinwhichdifferentiallearningratesaccountforobserved

growthratedifferences,andofferssomespeculationsabouttheimplicationsof

sionsare

inSection6.

TICAL

BACKGROUND

Therehasbeenarebirthofconfidence-stimulated

inlargepartbyRomer\'s

(1986)contribution-thatexplicitneoclassicalgrowthmodelsinthestyleof

Solow(1956)canbeadaptedtofittheobservedbehaviorofrichandpoor

economiesalike,believe

wecanobtainatheoryofeconomicmiraclesinapurelyaggregativeset-upin

whicheverycountryproducesthesame,singlegood(andarichcountryisjust

one一二三四年级手抄报 thatproducesmoreofit)butsuchaframeworkwillbeusefulinstatingthe

problemandinnarrowingthetheoreticalpossibilities.

Consider,tobeginwith,asingleeconomythatusesphysicalcapital,k(t),

and

humancapital,h(t),toproduceasinglegood,y(t):

(2.1)y(t)=Ak(t)

[uh(t)]1.

HereImultiplythehumancapitalinputbyu,thefractionoftimepeoplespend

producinggoods.3Thegrowthofphysicalcapitaldependsonthesavingsrates:

dk(t)

(2.2)

dt=

sy(t),

whilethegrowthofhumancapitaldependsontheamountofquality-adjusted

3Oneoftherefereesforthispaperfoundmyuseoftheterm\"humancapital\"inthisaggregate

contextidiosyncratic,andIagreethataggrfgatetheoriststendtousetermslike\"technology\"or

\"knowledgecapital\"forwhatIamherecalling\"humancapital.\"Butthecostofhavingtwo

terminologiesfordiscussingthesamething,oneusedbymicroeconomistsandanotherbymacroe-

conomists,isthatitmakesittooeasyforonegrouptoforgetthattheothercanbeasourceof

relevantideasandevidence.

ItwastheexplicitthemeofSchultz

(1962)

thatthetheoryofhumancapital,theninitsinfancy,

wouldprovecentraltothetheoryofeconomicgrowth,andSchultzincludedthestockofhuman

capitalaccumulatedonthejobinhisTable1(p.S6).Hisfigureswerebasedonestimatesprovided

inMincer(1962),whoseestimationmethod\"treats\'learningfromexperience\'asaninvestmentin

thesamesenseasarethemoreobviousformsofon-the-jobtraining,suchas,say,apprenticeship

programs\"(p.S51).Myusageinthispaperis,Ithink,consistentwith30yearsofpracticeinlabor

economics.

,

JR.

timedevotedtoitsproduction:

dh(t)

(2.3)

dt

=

8(1

-

u)h(t).

Takingthedecisionvariablessand

uasgiven,whichIwilldofor

this

exposition,themodel(2.1)-(2.3)

isjustareinterpretation

ofSolow\'soriginal

modelofasingle,closedeconomy,withthe

rateoftechnologicalchange(the

averageSolowresidual)equalto,.t

=8(1-

a)(1

-

u)andtheinitialtechnology

levelequaltoAh(O)1\'system,thelongrungrowth

rateofbothcapital

andproductionperworkeris8(1

-

u),therateofhumancapital

growth,and

theratioofphysicaltohumancapitalconverges

ongrun,

thelevelofincomeisproportional

totheeconomy\'sinitialstock

ofhuman

capital.4

Toanalyzeaworldeconomymadeup

ofcountrieslikethisone,oneneeds

to

bespecificaboutthemobility

offactorsofproduction.

Abenchmarkcasethat

hasthevirtuesofsimplicityand,

Ithink,adecentdegreeofrealismisobtained

byassumingthatlaboriscompletely

immobile,whilephysicalcapital

isperfectly

,iftherearencountries

indexedbyi,assume

thattheworld

stockofphysicalcapital,

K

=

Ein_

ki,

isallocatedacrosscountriessoas

to

equatethemarginalproduct

ineachcountrytoacommonworldreturn,r.

Then

ifeachcountryhasthetechnology(2.1)

withacommoninterceptA,thisworld

returnisr

=

aA(K/H)a-1,whereH

=

Yiuihi

istheworldsupplyofeffective

domesticproductineachcountry

is

proportionaltoitseffective

workforce:

Ka

(2.4)

yi=A

H

uihi.

Ifeveryonehasthesameconstantsavingsrates,

thedynamicsofthisworld

economy

areessentiallythesameasthoseofSolow\'s

ldcapital

stockfollows(dK/dt)=sAKaHl

-a,

andthetimepathof

Hisobtainedby

summing(2.2)

overcountries,eachmultipliedbyits

owntimeallocationvariable

ui.

Thelongrungrowthrate

of

physicalcapital

andof

everycountry\'soutput

is

equal

untry\'s

incomelevelwillbe

proportionaltoits

initialhumancapital,notonlyinthelong

runbutallalong

oryisthusconsistent

withthepermanentmainte-

nanceofanydegreeofincomeinequality.

Itwouldbehardtothinkofanothertheory

assimpleasthisonethatdoes

a

betterjoboffittingthepostwarstatistics

inthebackoftheWorld

Development

terpreting

Solow\'stechnologyvariable

asa

country-specific

stock

40fcourse,essentiallythesameeconomicscanbeobtainedfrom

amodelinwhichconsumer

preferencesaretakenasgivenand

savingsandtimeallocationbehaviorare

derivedratherthan

wa(1965),Lucas(1988),

andCaballeandSantos(1991).Theparticular

model

sketchedinthetextissimplyonerather

arbitrarilyselectedexamplefromthe

largenumberof

similarlymotivatedmodelsthathave

,forexample,Jonesand

Manuelli

(1990),KingandRebelo(1990),and

Becker,Murphy,andTamura(1990).

MAKINGAMIRACLE255

ofhumancapital,amodel

thatpredictsrapidconvergence

tocommonincome

levelsisconverted

intoonethatisconsistentwithpermanent

incomeinequality.

Butthekeyassumption

onwhichthisprediction

isbased-that

humancapital

accumulationinanyone

economyisindependent

ofthelevelofhuman

capital

inothereconomies-conflicts

withtheevidentfactthat

ideasdevelopedinone

placespreadelsewhere,

thatthereisonefrontier

ofhumanknowledge,not

one

foreachseparateeconomy.

Moreover,asParenteand

Prescott(1991)observe,

ifthemodelaboveisrealistically

modifiedtopermiteacheconomy

tobesubject

toshocksthathavesome

independenceacrosscountries,

theassumptionthat

eacheconomyundergoes

sustainedgrowthduetoitsownhumancapital

growth

onlywouldimplyever-growing

inequalitywithinanysubset

-

tiveincomelevelswould

followrandom-walk-like

seehow

thispredictioncanbereconciled

withthepostwarexperience

of,say,theOECD

countriesorthe

ntriesoftheworldaretiedtogether,

economi-

callyandtechnologically,

inawaythatthemodel(2.1)-(2.3)does

notcapture.5

Onewaytointroducesome

convergenceintothemodel

Ihavesketched,

proposed

andstudiedbyParenteand

Prescott(1991),istomodify

thehuman

capitalaccumulation

technology(2.2)

soastopermitanyonecountry\'srate

of

humancapitalgrowth

tobeinfluencedbythelevel

ofhumancapitalelsewhere

mple,let

H(t)betheworld

effectivelaborvariabledefined

above,andletZ(t)=H(t)/Eiui

betheworldaveragehuman

capitallevel.

Replacethehumancapital

accumulationequation(2.2)

with:6

(2.5)

dt

=

5(1

-

u)h(t)1Z(t)6.

dt

Withthismodification,

thedynamicsof

theworldstocksofphysicaland

human

capital

areessentiallyunchanged,

butnowaneconomywithahumancapital

stocklowerthan

theworldaveragewillgrow

fasterthananaboveaverage

economy.

Forexample,ifthetime

allocationisequalacrosscount蟋组词2个字 ries,

so

that

H(t)

andZ(t)growatthe

rate8(1

-

u),

a

country\'s

relativehumancapital,

zi

=

hi/Z,

follows

d1

(2.6)

d-zj(t)=5(1-u)z(t)[z(t)--1].

Evidently,

zi(t)

converges

toone,andfrom(2.4),

thismeansthatrelative

incomesconverge

tooneatthesamerate.

Intheworldasawhole

inthe

postwarperiod,

income

dispersion

acrossall

countriesappears

,ofcourse,

therearemanyreasons

to

believethattheassumption

offreeworldtradethatleads

to(2.6)isavery

bad

5Aninformativerecentdebate

onincomeconvergence

hasbeenstimulatedbythe

exchange

betweenBaumol(1986),

DeLong(1988),and

BaumolandWolff(1988).My

statementinthe

text

simplyechostheshared

conclusionoftheseauthors.

6Thisexternaleffectmight

betterbecapturedthrough

thehumancapital

levelofthemost

advancedcountries,rather

thantheworldaverageZ(t).

Buttheuseofthelatter

variable

keeps

the

algebrasimple,andIdon\'t

thinkthedistinction

iscriticalforanyconclusions

Iwishtodrawhere.

,

JR.

approximationformuchof

theworld,andtherearecertainlydifferences

across

countriesintheincentivespeople

havetoaccumulatebothkindsofcapital,

implyingdifferences

oversubsets

ofcountries,orregionsofcountries,

wherefactorandfinalgoodsmobility

is

high(liketheEEC

)convergencecanbeobserved.7

BarroandSala-i-Martin(1992)obtain

aregressionestimateofanaverage

convergencerate

ofrelativeincomes关于春天的名言 ,conditionedonvariables

thatmaybe

interpretedascontrollingforacountry\'s

adherencetotheaboveassumptions,

ofslightlylessthan

.02

(Table3,p.242).

Astheyobserve,ifoneinterpretsthis

coefficientasreflectingdifferential

ratesofphysicalcapitalaccumulation

ina

worldinwhichincomedifferencesreflect

mainlydifferencesincapitalper

worker,thisrate

ofconvergenceismuchtoolowtobeconsistentwithobserved

capital

atively,interpretingthisfigureasan

estimate

of

(l/z)(dz/dt)in(2.6),theirestimateimplies068(1

-

u)

=.8(1-u)is

theaveragerateofhuman

capitalgrowth,also

about.02inreality,this

interpretationyields

anestimated0ofunity,whichfrom(2.5)

wouldmeanthat

humancapitalaccumulation

inanycountrydependsonlocalefforttogether

withworldwideknowledge,independent

thisviewpoint,theBarro-Sala-i-Martin

estimateseemshigh.

Allofthisisbywayofaprelude

tothinkingaboutgrowthmiracles-about

deviationsfromaveragebehavior.

Ihavedescribedamodelofaworldeconomy

reasonablyrealisticinitsdescription

ofaveragebehaviorofcountriesat

differentincomelevels-in

whicheveryonehasthesamesavingsrate

and

etheprospects

forusingthesame

theorytosee

howvariationsacrosseconomiesintheparameters

sanducan

inducevariation

inbehaviorofthemagnitudeweseektoexplain?Herethe

exercisebegins

to

gethard.

TheEastAsianeconomiesdoindeed

havehighinvestmentrates.

Thecurrent

ratioofgrossdomesticinvestment

toGDPinKoreaisabout.29,ascompared

toaveragebehaviorof

anandHongKong,the

investment

ratiosare.21and.24respectively.

InSingapore,itisaremarkable.47.

Inthe

Philippines,

for

comparison

itis.18.8Inaworldwiththeperfectcapitalmobility

usedinmyillustrationabove,these

differencesininvestmentrates

wouldhave

noconnection

with

savi荒凉的近义词是什么 ngs

rates:

anycountry\'shigher

than

averagesavings

withnointernationalcapitalmobility,

to

translateagivendifference

in

savings

ratesintoadifferencesinoutputgrowth

ratesonemustmultiplybythe

returnoncapital(since

a1dy

d1dydk

ay

ds

ydt

dsydkdt

dk

7See,

for

example,

Ben-David

(1991).

8Allthe

figures

io

forTaiwanisfromthe1987TaiwanNational

ersarefromthe1986WorldDevelopment

Report.

MAKINGAMIRACLE257

from(2.2)).Ifthereturnoncapital

weretenpercent,then,theKorea-Philip-

pinesinvestmentratedifferenceof

.11canaccountforadifferenceof.011

in

outputgrowthrates,oraboutone

iseffectis

only

transient,sinceinthelongrundifferences

insavingsratesareleveleffects

only.

Nowapplyingthesameroughcalculation

totheSingapore-Philippinesinvest-

mentratedifferenceof.29,onecanaccount

foradifferenceinoutputgrowth

ratesofnearlythreepercentagepoints

(andmore,ifahigherandstilldefensi-

blereturnoncapitalisused)which

isclosetothedifferentialsIamcalling

\"miraculous.?\'

Indeed,Young(1992)demonstrates

thatoutputgrowthinSinga-

poresincethe1960\'scanbe

accountedforentirelybygrowthinconventionally

measuredcapitaland

laborinputs,withnothingleftovertobeattributedto

ng\'s

point,underscoredby

hisparalleltreatment

ofSingaporeandHongKong,istheexceptional

characterofgrowthinSinga-

pore,andnotthattheAsian

miraclesingeneralcanbeattributed

tocapital

accumulation.

Growthaccountingmethods,appliedcountry-by-country

asinYoung\'sstudy,

canquantifytherole

ofinvestmentdifferentialsinaccounting

forgrowthrate

ral,thesedifferentialsleave

mostmeasuredoutputgrowth

nclusion,

whichseemstomesoclear,

ations

betweeninvestmentratiosandgrowthrates,

whichtendtobepositive,arefrequently

citedbutdonotsettleanything.

If

growth

isdrivenbyrapidaccumulationof

humancapital,oneneedsrapid

growthinphysicalcapitaljust

tokeepup:lookatequation(2.4)!

Itmaybethat

byexcludingphysicalcapital

fromthehumancapitalaccumulationequation

(2.3)or(2.5)

I

haveruledoutsomeinterestingpossibilities:

Onecannot

accumulateskillas

s

physicalcapital

willassumeamoreimportantrolewhenthetechnology

for

accumulatinghumancapital

isbetterunderstood,but

ifso,itwillbeatbesta

supportingpart.

Letuslookelsewhere.

IntheframeworkIamusing,theotherpossible

sourceofgrowthrate

differentialsisdifferentialratesofhumancapital

accumulation,stemming

from

differencesinsocieties\'time-allocation

ancapitaltakesmany

formsanditsaccumulationoccurs

in

manyways,so

therearedecisions

in

emphasis

choice,

I

think,iswhether

tostress

humancapitalaccumulationatschool,

oronthejob.

Ifoneinterprets(2.3)or(2.5)asdescribingknowledge

accumulationthrough

schooling,

theseequationsimplythatdoubling

thefractioninschoolwould

doublethehumancapitalgrowthrate,addingonly

another.02totheaverage

,ofcourse,thelinearity

of(2.3)probablyleadstoanoverstate-

mentoftheeffect

ofso

large

a

change.

AsIremarkedinmyintroduction,

the

fastgrowingAsianeconomies

arenot,ingeneral,betterschooledthansome

of

is

onformal

schooling,then,

seemsto

involvetheapplicationof

amodestmultipliertoveryslightdifferences

in

behavior,leading

tothesamediscouragingconclusionforhumancapital

thatI

arrivedatinthecaseofphysicalcapital.

,

JR.

Thisconclusionmayseemaninappropriateinferencefroman

oversimplified

model,butIthinkitisinfactreinforcedbythinkingmoreseriously

aboutthe

schoolingdecisionstake

placeinalife-cyclecontext,

withschoolprecedingworkandeachindividual

decidingonthelengthofthese

twocareerphases.(Thisisasimplification,too,butabetteronethan

thinking

ofarepresentativeagentdividinghis

timeinperpetuity.)Nowinasteadystate

orbalancedpathofaneconomyinwhicheveryone

spendsafraction1

-

uof

hisworkinglifeinschool,workerswithschoolinglevel1-uare

retiringfrom

thelaborforceatexactlythesamerateasnewworkerswiththesame

education

erwhatthevalueofuisinsuchasteady

state,allof

thisinvestmentisreplacementinvestmentandthereisno

increaseinthe

(2.3)isanhypothesisaboutnet

investment,onecannotthenidentifythevariable1

-

uwithtimespentin

dentifyincreasesin

average

choolinglevelsare

increasinginvirtuallyallsocietiestoday,thisisapossibilityworth

developing,

butitcannotbepursuedwithinasteadystateframework.

Thisisanimportant

andneglectedrespectinwhichneitheradvancednor

mostbackwardeconomies

canbeviewedasmoving

alongbalancedgrowthpaths.

Alternatively,wecanthinkofabalancedpathonwhichtimespentinschool

isconstantbutthequalityofschoolingisimprovingduetoincreasesin

general

ssibilityisanalyzedinStokey(1991a),fromwhich

the

paper,therateofexpansionof

knowledgeistakentobeanexternaleffectofthetimespentin

school,the

this

hypothesisdoesnotsalvagethemultiplierargumentsIapplied

above,unless

oneiswillingtoassumethatincreasesin

generalknowledgeaccrueequallyfrom

timespentin

primaryschools

quantify

amodellike

Stokey\'s,onewouldneedamuchsharperempiricalidentific电视剧全集免费播放 ationofthesetof

activitiesthatleadtonewknowledge-tonetinvestmentina

society\'shuman

capital-than

is

providedbyanyaggregateindex

oftotal

schooling

wouldbeamostinterestingavenueto

explorebut

Iamnot

preparedto

doso

here,soIwillendthisdigressionandmoveon.

Humancapitalaccumulationalsooccursat

work,asweknowfromthefact

theexperiencedworkersand

managersearnmorethaninexperiencedones.

Thisaspectofhumancapitalaccumulation-on

thejobtraining-could

alsobe

(andhasbeen)modeledasatime-allocationdecision.

Alternatively,

ina

multiplegoodworld,onecouldthinkofonthejob

accumulation-learningby

doing-as

associatedwiththe

type

of

processoneisengagedin.

That

is,

one

might

thinkofsomeactivitiesas

carrying

withthema

high

rateofskill

acquisition

and

others,routineortraditionalones,asassociatedwithalowrate.

If

so,themixofgoodsasocietyproduceswillaffectitsoverallrateofhuman

erstandingdiversity,Ithinkthisroute

has

promise:Thevariationacrosssocieties,oratleastthose

engaged

in

internationaltrade,inthemixof

section,

I

MAKINGAMIRACLE259

havetriedtomotivatea

focusonthissourceof

diversitybyaprocessof

elimination:Neither

physicalcapitalaccumulationnorhuman

capitalaccumula-

tionthroughschooling

seemstohavemuchpotential,

atleastwithinthe

thisnextsection,Iturntomuch

moredirect,

microeconomicevidenceonthesamepoint.

ERTYSHIPMIRACLE

InLucas

(1988)

I

usedamulti-goodmodel,

adaptedfromKrugman(1987),in

whichdifferentgoodswere

associatedwithdifferent

learningratestocapture

theideathatthechoiceofwhich

goodstoproducecanbeviewed

asanimplicit

choiceofahuman

ldof

openeconomies,

comparative

advantage-previously

accumulated,good-specifichumancapital

holdings-will

determinewhoproduceswhat,and

themixofgoodsthatthis

processassignstoaparticular

economywilldetermineitsratesofhuman

capital

ndofformulation

hasbeentakenin

interestingdirectionsby

BoldrinandScheinkman

(1988)andMatsuyama(1992).

Itisattractive,for

presentpurposes,because

therearesuchwidedifferencesin

productmixacross

countriesandbecausethefast

growingAsianeconomieshave

undergonesuch

dramaticchangesinthegoods

theyproduce.

Butthehypothesisthat

differentgoodsareassociatedwith

permanently

differentlearningpotentials

conflictssharplywithavailableevidencein

two

,examinationofgrowthintotal

factorproductivity(Solowresidu-

als)acrossbothindustriesand

time(asconducted,for

example,byHarberger

(1990),showsnodecade-to-decade

stabilityinthehigh

productivitygrowth

andwood

productscanrank14thinthe

1950\'s,firstinthe

1960\'s,anddisappearfromthelistof

leadersaltogetherinthe1970\'s.9

Second,

evidencewehaveon

learningonnarrowlydefinedproductlines

invariably

shows

high

initial

learningrates,decliningovertimeas

productioncumulates.

Thesetwokindsofevidence

reinforceeachother,andseem

decisiveagainstthe

formulationKrugman

bservationshaveled

Stokey(1988)and

Young(1991a)toaverydifferent

formulation,onethatismuchmore

tightly

groundedinmicroeconomic

eviewthisformulationin

Section4,

butbeforedoingsoIwantto

reinforcethemotivationwithareminderof

just

howimpressivethe

evidenceontheproductivityeffectsof

learningbydoingcan

be.

ThebestevidenceIknowof

thatbearsonon-the-job

productivitychangeina

single,largescaleproduction

pro\'cess,

wasutilizedinstudies

(1945)g

(1965).Bothstudiesuseddataonthe

produc-

tionofa

singletypeofcargovessel-the

rds

December,1941,through

December,1944,these

yardsproducedatotalof2458

LibertyShips,alltothe

samestandardized

eralindividual

yards,Searleplotted

man-hourspervesselagainst

9

Harberger(1990),Table3.

260ROBERT

,

JR.

BASICDATAFOR2

YARDSBUILDINGLIBERTYSHIPS

MAN-HIOURSPER

VESSEL

MAN-HOURSPERVESSEL

IHOUSANOS)

(

THOUSANDS)

2000----2000

1000

10

800~~~~

700

700

600

600

500YARD

2

500

400

400

300

-~__------300

1

23

4

56

78

910

20

30

40

SUCCESSIVE

GROUPS

OF

5VESSELS

FIGURE1.-Reductionsinman-hourspervesselwithincreasingproduction.

Merchantshipyards.

ults

eresultsovertenyards

aregiveninFigure2,erty

Ships,\"thereductions

inmanhours

pership

witheach

doubling

ofcumulative

outputrangedfrom12to24percent.\"\'10

StimulatedinpartbyKennethArrow\'s(1961)theoreticalsuggestionthat

learning-by-doingmightserveasthekeyfactor

in

growthfor

aneconomyasa

whole,RappingincorporatedSearle\'sandotherevidence

withinaneoclassical

edthedataforallyardsandestimateda

Cobb-Douglasproductionfunction,controlling

for

changes

in

capitalperyard,

withcumulatedyard(notindustry)

obtainedestimatesofthelearningeffect,comparabletoSearle\'s,rangingfrom

showedthattheinclusionofcalendartimeadded

nothing(thetrendcameoutslightlynegative!)totheseresults.

Idonotthinkthereisanythinguniquetoshipbuildinginthefindingsthat

tonConsultingGroup(1972)hasob-

tainedfairlycleanlearningcurves,withslopessimilartothoseestimatedby

SearleandRapping,foravarietyofindustries,andotherresearchershavedone

l?Searle(1945),p.1144.

MAKINGAMIRACLE261

MAN-HOURS

PER

VESSEL

MA\"-HOURS

PER

VESSEL

MANHORSPE

VSASEL

(THOUSANDS)

0THOU0ANDS

1600

1400

i_

|

t

1

1400

1200

_200

900

]

>t

-

j

-

<

X

9oo~~~~~~~~~~~~~~~~~~~~~~~

1200

ANKERS_

/

8

0

700

|_

LIBERTY

SHIPS

x

l

|

1

700~~~~~VICTRY

HIP

1000

-

-000

900~~~~~~~~~~~~~~~~~~~~~~~~~~~Sooo

70070

0C600So

500

_

_____

.

500

400

_

_1_________

1

1

,,,,

._

.

_

1400

1941

1942

1943

1944

UNITED

STATEi

DfPAITMENT

OF

LABOR

BUREAU

OF

LABOR

STATISTICS

FIGURE

2.-Unit

man-hour

requirements

for

selected

shipbuilding

programs.

Vessels

delivered

December

1941-December

1944.

262ROBERT

,JR.

uniqueabouttheLibertyshipdataisthattheshipswere

builtaccordingtoexactlythesameblueprintsoveraperiodofseveralyearsand

2,whichgivesSearle\'slearning

curvefortheindustryasawhole,isnotnearlyassharpasthecurvesinFigure1

forindividualyards,presumablybecauseindustryexpansionisamixofin-

creasedproductionbyexistingyardsandtheentryofnew,inexperiencedyards.

Productiondataevenfromnarrowlydefinedindustriesmaskcontinualmodel

andotherproductmixchangesovertime,whichmakesitdifficulttousethemto

exceptionalabouttheLibertyship

evidence,Ithink,isthecleannessoftheexperiment,notthebehaviorit

documentssobeautifully.

Quantitatively,theseresultsareinterestingtoaneconomistlookingfor

threeyearperiodcoveredbyRapping\'s

study,industryoutputpermanhourincreasedata40percentannualrate!

e

individualworkerwhoisdoingthelearning?Themanagers?Theorganization

asawhole?Aretheskillsbeinglearnedspecifictotheproductionprocesson

whichthelearningtakesplace,ormoregeneral?Doeslearningaccruesolelyto

theindividualworker,manager,ororganizationthatdoestheproducing,oris

someofitreadilyappropriablebyoutsideobservers?Thesearequestionsthat

thetheoryofgrowthneedstoaddress,butIwillpassoverthemhere.

Amoreurgentquestion,Ithink,iswhetherthekindofbehaviorRappingand

Searledocumented,foroneproductlineforonebriefperiod,canbelinkedto

productivitygrowthforanentireeconomyoverperiodsofthirtyorfortyyears.

Thisisthetopicofthenextsection.

NGMODELS:TECHNOLOGY

Inordertoexaminethepossibleconnectionbetweenevidenceoflearningon

individualproductlinesandproductivitygrowthinaneconomyas

a

whole,

considerthelabor-onlytechnology:

(4.1)x(t)

=

kn(t)z(t)

,

wherex(t)istherateofproductionofa

good,

kisa

productivityparameter

thatdependsontheunitsinwhichlaborinputandoutputaremeasured,n(t)is

employment,andz(t)representscumulativeexperience

inthe

production

of

tiveexperienceisinturndefinedbythedifferentialequation:

(4.2)

dz(t)=

n(t)z(t)

a,

dt

andtheinitialvalue

z(to),

assumedtobegreaterthanorequaltoone,of

the

experiencevariableonthedate

to

when

production

was

begun.

The

general

MAKINGAMIRACLE263

solutionto(4.2)is

]1/1-

(4.3)

z(t)=(Z(tO))l\'a

+

(1

-

a)ftn(u)

du

Theimplicationsofthismodelforthedynamicsofproductionofasingle

e,totakethesimplestcase,thatemployment

(4.1)and(4.3)implythatproductionfollows

x(t)

=

kn[z(to)1-a+

(1

-

a)n-(t-

to)|a1-

Productiongrowswithoutbound,andtherateofproductivitygrowthdeclines

monotonicallyfromanh(z(to))initialproductivitylevel

z(to)

>1andanyemploymentlevel(orpath)productivityatdatetisan

increasingfunctionofthelearningratea.

Noticethatthetechnology(4.2)impliesascaleeffect:alinkbetweenthelevel

rriestheunwel-

comeimplicationthatacountrylikeIndiashouldhave

anenormousgrowth

afeatureofanylearning

bydoingtheory,butIagreewithMatsuyama(1992)thatifoneisthinkingabout

anentireeconomyorsizeablesectorofaneconomy,itisanuisanceimplication

thatwewanttodisposeof.\"Matsuyamaproposesthinkingofapopulationas

containingafixedfractionofentrepeneurs,andofatechnologythatrequires

ublingthepopulation

meansdoublingthenumberofenterprisesthataresubjecttothelearning

technology,keepingthesizeofeachfixed,ras

learningeffectsarepartlyexternaltothefirm,asIthinkthey左忠毅公逸事朗读 are,thisdevice

doesn\'tquitework,andoneneedstothinkofsomeotherlimitationon

scale-city

size,say.

Iwill

simplyignorethesescaleeconomiesinwhatfollows,

assumingthatsomeexplanationalongthelinesofMatsuyama\'swillbediscov-

eredtorationalizethisneglect.

Withthetechnology(4.1)-(4.3),onecanobviouslyobtainmiraculousratesof

productivitygrowthbyshifting

alargeamountoflaborontoasingle,new

edthatn(t

-

to)

islargerelativetoinitialexperience(which

isthewaymostpeopleinterpretstatisticallearningcurves),therateofproduc-

tivitygrowth

tyearsafterproductionisinitiatedisapproximatelya/((1

-a)t).

Usingthevalue

a

=

0.2estimatedbyRappingandSearle,productivitygrowth

oneyearafteraproductisintroducedisa/(1

-

a)

=

woyears,the

growthrateisreducedbyhalfto0.125,hmiraclesustained

foraperiodofdecadesclearlymustthusinvolvethecontinualintroductionof

newgoods,

ifnew

goodsareintroduced,

ashiftofworkersfromold

goods

withlowlearningrates

tonewgoodswithhighratesinvolvesaninitialdrop

in

productivity:people

are

11Backus,Kehoe,andKehoe(1991)isanempiricalexaminationofscaleeffectsongrowth

rates,

ndsomeevidenceofsucheffectsinmanufacturing,andnone

foreconomiesasawhole.

,JR.

betteratfamiliaractivitiesthantheyare

tevenclearhow

thesefactorsbalanceout.

Topursuethisquestion,IfollowStokey(1988)

andconsideraneconomyin

whichavarietyofgoods,indexedbys,isproduced,

whereahigherindexs

ey(1988)

and,indifferentwaysinYoung(1991a)

andGrossmanandHelpman(1991b),specific

assumptionsonconsumerprefer-

encesorthetechnologygive

aprecisemeaningtothesenseinwhichonegoodis

mmediateobjectives,

itwillbeadequateto

considerasmall,openeconomy

andtouseanassumedschedulep(s,t)

=

elsof

worldpricestosummarizethequality

ofgoods:abettergoodmeansagood

thattheeconomyprogressesby

introducingbetterquality(highers)

goodsintoproductionovertime,

andlet

S(t)betheindexofthegoodthatisfirstproduced

atdatet.(Iwillalsouse

r(s),whereiistheinversefunctionoftheincreasing

functionS,todenotethe

dateonwhichgoodsisfirstproduced.)

Thenifx(s,t)isproductionofgoods

atdatet,thevalueoftheeconomy\'stotalproduction

is

(4.4)

y(t)=S(t)ePsx(st)ds.

Letn(s,t)beemploymentongoodsatt,

andz(s,t)becumulatedexperience.

Theniflearningproceedsindependently,

goodbygood,(4.1)

and(4.3)imply

a/1

-a

(4.5)

x(s,

t)

=

kn(s,

t)

(z(s,Tr(s)))1a+(

a)

t

n(s,u)duI

Equations(4.4)and(4.5)togetherdescribe

theimplicationsfortotalproduction

ofagivenwayofallocatinglaboracrossproduct

linesthroughtime.

Considerthefollowingspecificlaborallocation.

Lettherateofnewproduct

introductionbeaconstantA,sothatS(t)=Atandr(s)=s/ea

densityfunction

withcdf

iP,

andsupposethatforalls

E

(0,At],n(s,t)=

p(t-s/A)(that(p(t

-s/A)workersareassigned

toproducethegoodsofage

t

-s/A)andthattheremaining

1

-

?P(t)workersproduceagood

0

onwhichno

thatinitialproductivityisthesameforallgoods,atthe

levelz(s/A,s)=

6

>heseassumptions,(4.4)

and(4.5)implythatthe

valueoftotalproductionis

(4.6)y(t)=1-?P(t)+kAetAtfeAu(u)[61-a+(1a)P(u)]a/1-a

du

Theasymptoticgrowthrateforthiseconomyisevidently

/A.

Thisratedoesnot

dependeitheronthelearningparameter

a

oronthedistributionSDofthe

s

ineitherofthesefactors

insustainedgrowth

atallinthisframework,itis

necessarytoassumethatbettergoodsbecome

producibleatsomeexogenously

givenrateA,which

thenalongwiththequalitygradient

/i

dictatesthelongrun

growthrateof

thesystem,independentoflearningbehavior.

MAKINGAMIRACLE265

Thoughtheproductionofnewgoodsiscontinuouslyinitiatedinthisexample,

ey(1988)thisrateis

madeendogenousthroughtheassumptionthattheexperienceaccumulatedin

producinggoodsreducesthecostofproducinggoods\'>s.(Itmayreducethe

costofproducings\'

thedirectionofimprovingproductivityonthemoreadvancedgood.)Asa

specificinstanceofStokey\'shypothesis,verycloseto

thatproposedbyYoung

(1991a),letusmodifythelastexamplebypostulatingthattheinitialvalue

z(s,r(s))inthelearningcurve(4.3)dependsontheexperiencethathasbeen

ethataneconomyatsomefixed

datethasexperiencesummarizedbyz(s,t)

fors<

S(t),

buthasyettoproduce

anygoodwithindexaboveS(t).Assumethatifproductionofagoods>S(t)is

initiatedatt(ifr(s)

=

t)thenitsinitialz-valueisproportionaltoanaverageof

theeconomy\'sexperienceonpreviouslyproducedgoods:

s

(4.7)z(s,7(s))

=

06|e-\'(S-U)z(u,7(s))

du.

Equation(4.7)expressestheinitialproductivityongoodsasanaverageof

lently,wecanexpressthe

initial

productivityonthegoodintroducedatt,goodS(t),asanaverageofexperience

ongoodsintroducedearlier:

(4.8)z(S(t),t)

=

0O8te-5[s(t)-s(t-v)]z(S(t

-

v),t)S\'(t

-

v)dv,

0

integratingoveragesvinsteadofgoodss.

Assume,next,thatproductiononanewgoodisinitiatedwheneverthe

expressions(4.7)

and

(4.8)reach

a

triggervalue

7>1,takenas

a

givenconstant.

Underthisassumption,theleftsideof(4.8)isreplacedwiththisconstant(,

implyingthatthefunctionS(t)whosederivativeisthe

rateatwhichnewgoods

areintroducedmustsatisfy

(4.9)

6

05

te-5[S(t)-S(t-V)]z(S(t

-

v),t)S\'(t-v)

dv.

Asinthepreviousexample,wecontinuetoassumethattheallocationof

employment

at

any

dateisdescribed

by

a

densitySD

andcdf

\',

where?P(u)is

thefractionofpeopleemployedproducinggoodsthatwereintroducedlessthan

resentcase,eachgoodhastheinitialproductivitylevel

(,soinserting

thesolution

(4.3).

for

z(S(t-v),t)

withthisinitialvalueinto

(4.9)

yieldsasingleequationinthefunctionS(t).Forlargevaluesoft,thesolution

S(t)to

this

equation

willbehavelike

S(t)=At,wherethe

constantAsatisfies

(4.10)

=8

05Afe

-Av[-a

+(1-)cP(v)]1/l-dv.

Therightsideof(4.10)isjustanaverageofthepositive,increasingfunction

0[f1-a

+(1

-a)P(v)]\"1(1-a),takenwith

respect

toan

exponential

distribution

tisapositive,decreasingfunctionof8A,tending

,JR.

towardthevalueO0as8A-*oo

andtowardthevalue

0[f1a

+1

-

a1/(l-a)as

8A->0.(IfthelatterexpressionislessthanfatA

=

0,thentheeconomydoes

notaccumulaterelevantexperiencefastenoughtointroducenewgoodsinthe

steadystate.)Forfixed8A,therightsideof(4.10)isanincreasingfunctionof0,

a,andk,anditalsoincreasesasthedistributionoflaborsp(v)becomesmore

concentratedonlowervaluesofv(onnewergoods).Henceifapositive

solutionAexists,itisinverselyproportionaltothedecayrateofspillover

experience,anincreasingfunctionofthespilloverparameter0andthelearning

ratea,andincreasesasemploymentismoreheavilyconcentratedongoodsthat

areclosertotheeconomy\'sproductionfrontier.

Theformula(4.6)forthevalueoftotalproductioncontinuestoholdinthis

secondexample,andtheeconomy\'slongrungrowthrateisA,u,

underthissecond,spillover,technology,economiesthatdistributeworkers

ac

course,thisconclusionisnotbasedpurelyontechnologicalconsiderations:The

value(ofinitialproductivitythatisassumedtotriggertheinitiationof

productionofanewgoodisofcentralimportance,andneedsaneconomic

rationale.

OnemightviewthespillovertechnologiesofStokeyandYoungasreconciling

theKrugmanhypothesisofamanufacturingsectorwithaconstantrateof

productivitygrowth,basedonlearning,withthefactthatlearningrateson

mple,one

couldinterpreteitheroftheexamplesinthissectionasdescribing

asectorofan

view,

thecontributionofStokeyandYoungistobreakdownanassumedsectoral

learningrateintoitscomponents,a,0,and8(inmynotation),andtorelatethis

ratetothewayworkersaredistributedovergoodsofdifferentvintages.

Thisinterpretationseemsfinetomeaslongasoneisdiscussingtheconse-

quencesofagivenworkforcedistribution,butifonehas

inmind

applying

the

theoryofcomparativeadvantagetodeterminingthewayworkersineach

countryareallocatedtotheproductionofdifferentgoodsitceasestomake

man\'stheory(asinLucas(1988))itisasectorasawholethat

toralinterpreta-

tionofStokeyandYoung\'stheories,eachsectorconsistsofmanygoodsand

trycanbe

expectedtohave

a

comparativeadvantage

in

manufacturing

in

general,oreven

incrudeaggregateslikeChemicalsandAlliedProductsor

Printing

andPublish-

ativeadvantagewiilbeassociated

with

categories,likeacetyleneor

paperbackeditionsofEnglishpoetry,thatareinvisibleeveninthefinest

allseeinthenextsection,thisfeature-besides

being

a

steptowardsgreaterrealism-leads

toanentirelydifferentviewof

tradeandgrowththanisimpliedbytheKrugman

technology,

the

superficial

similarityof

thetwo

notwithstanding.

Themainattractionofalearningspillover

technology

suchasthatdescribed

inthesecondexampleofthissectionisthatitoffersthe

potential

of

accounting

MAKINGAMIRACLE267

forthegreatdifferenceinproductivitygrowthratesthatareobservedamong

se,littleisknownaboutthecrucial

spilloverparametersaand0-on

whichthelearningcurveevidencedescribed

inSection3providesnoinformation-but

surelyanessentialfirststepistofind

aformulationthatiscapable,undersomeparametervalues,ofgeneratingthe

behaviorwearetryingtoexplain.

NGANDMARKETEQUILIBRIUM

Theobjectiveofthelastsectionwastosetdownonpaperatechnologythatis

consistentwithagrowthmiracle,whichistosay,consistentwithwidediffer-

sbeen

done,followingStokeyandYoung,inawaythatIthinkisconsistentwiththe

mainfeaturesoftheEastAsianmiracles,allofwhichhaveinvolvedsustained

growingeconomyorsectorunderthistechnologyisonethatsucceedsin

concentratingitsworkforceongoodsthatarenearitsownqualityfrontier,and

thusinaccumulatinghumancapitalrapidlythroughthehighlearningrates

associatedwithnewactivitiesandthroughthespilloverofthisexperiencetothe

ypothesesareconsistentwithcommonly

knownfacts,,however,I

havesaidnothingabouttheeconomicsthatdeterminethemixofproduction

activitiesinwhichaneconomyorsectorofaneconomyinfactengages.

ThepapersofStokey(1988),(1991b)andYoung(1991a)developmodelsof

marketequilibriumwithlearningtechnologiesundertheassumptiontheeffects

case,

laborissimplyallocatedtotheusewiththehighestcurrentreturn,independent

econstantreturnstechnologytheseauthorsassume,

th

isthesimplestcase,soIwillbeginwithittoo.

Insuchasetting,Stokey(1991b),studiesnorth-southtrade,where\"north\"

pecificassumptions

aboutconsumerpreferencesforgoodsofdifferentqualities,sheobtainsa

unique

world

equilibrium

inwhichthesouth

produces

anintervaloflow

quality

goods,thenorthproduces

anintervalof

highqualitygoods,

andthereisan

eetrade

(asopposedtoautarky)learning-by-doingisdepressed

inthe

poorcountry,

whichnowimportshigh-qualitygoodsfromtherichcountryratherthanat-

tempting

to

producethem

seethatwith

dynamics

asassumed

inStokey(1988),bothcountrieswillenjoygrowthbutthepoorcountrywill

remainforeverpoorer.

AsimilarequilibriumischaracterizedinYoung(1991a),usingaparameteri-

zationofpreferencesandthelearningtechnologythatpermitstheexplicit

calculationofthenorth-southequilibrium,including

afull

description

ofthe

remanypossibleequilibriumevolutionsofhis

,JR.

north-southsystem,dependingonthepopulationsofthetwore人生若只如初见何事秋风悲画扇 gionsandon

Stokey\'s(1991b)analysis,theadvancedcountryproduceshighqualitygoodsand

adeslowslearningand

g\'sframework,

thereareequilibriainwhichthepoorcatchuptotherich,butonlywhentheir

oesnot

emphasizethispossibilityand,asIhavesaidearlier,Idonotwishtoeither.

TheequilibriaofStokeyandYoung,then,involvesustainedgrowthofboth

richandpoor,atpossiblydifferentrates,andthecontinuousshiftingof

l

comparativeadvantageisnotpermanent,asinKrugman\'sformulation,sincea

richcountry\'sexperienceinproducinganygivengoodwilleventuallybeoffset

bythefactthatthegoodcanbeproducedmorecheaplyinalessexperienced

rearenogrowthmiraclesinthese

theseequilibriacouldreadilybemodifiedtoincludecross-

countryexternaleffects,andhencecatchingup(forreasonsunrelatedto

economiesofscale),asIhavedonewiththeSolowmodel,therewouldbe

nothingonewouldwishtocallmiraculousaboutthisprocess.

InthemodelsofStokeyandYoung,allhumancapitalbenefitsareassumedto

rningandgrowththatoccursisalways,inasense,acciden-

odelscontainaspectsofprivatelyheldknowledge,sothatindividual

agentsfacethecapital-theoreticproblemofbalancingcurrentreturnsagainst

ama(1991)studiesa

two-sectorsysteminwhichworkerscomparethepresentvalueofearningsina

traditionalsectortothevalueofearningsinamanufacturingsectorinwhich

(1991b)augments

an

andHelpman(1991a)postulatetwoRandDactivities-innovation,doneonly

inadvancedeconomies,andimitation,donebypooreconomiestoo-with

lags

thatletthediscovererorsuccessfullow-costimitatorenjoyaperiodofsuper-

ronecallsthedecision

problemsthatariseintheseanalysesoccupationalchoice,orresearchand

development,

or

learning,

allinvolveadecisionontheallocationoftime-at-work

thatinvolvesbalancingcurrentreturnsagainstthebenefitsofincreasedfuture

earnings,

andallhaveasimilar

capital-theoreticstructure.

Droppingtheassumptionthatlearninghasexternaleffectsonlyiscertainlya

steptowardrealism,one

thatraises

manyinterestingtheoreticalpossibilitiesyet

usonlyconjecture,butIwouldguessthatthemain

featuresoftheequilibriathathavebeenworkedoutbyStokeyandYoungwill

turnouttostandupverywellunderdifferentassumptionsabouttheownership,

ifIcanusethatterm,ingspillovertechnologygives

thosewhooperatenearthecurrentgoodsfrontieradefiniteadvantagein

movingbeyond

advantageisdecisivewhendecisions

aretaken

myopi-

cally;

Idonotsee

why

itshould

disappear

whensomeofthereturnsfrom

doing

MAKINGAMIRACLE269

soareinternalizedandworkersandfirmslooktothefutureintheirindividual

decisionproblems.

Inshort,availablegeneralequilibriummodelsofnorth-southtradedonot

predictmiraculouseconomicgrowthforthepoorcountriestakenasagroup,

nordoIseeanyreasontoexpectthattheequilibriaofmoreelaboratetheories

adisappointment,perhaps,butitdoesnotseemto

retheoriesdesignedtocapture

themaininteractionsbetweentheadvancedeconomiestakenasagroupand

thebackwardeconomiesasawhole,withinatwo-countryworldequilibrium

tisafactthatthepoorareeithernotgainingontherichor

aregainingonlyveryslowly,onewantsatheorythatdoesnotpredictotherwise.

Asuccessfultheoryofeconomicmiraclesshould,Ithink,offerthepossibility

ofrapidgrowthepisodes,butshouldnotimplytheiroccurrenceasasimple

ldbeasconsistentwiththe

purposeofexploringthese

possibilities,theconventionsofsmall,openeconomytradetheory

aremore

suitable(aswellassimplertoapply)thanthoseofthetheoryof

aclosed,

echnologyavailabletoindividualagentsfacingworld

priceshasconstantreturns,locationswillyield

highexternalbenefitsandgrowthinproductionandwages;otherswillnot.

Therewillbealargenumberofpossibilities,withindividualagents

in

equilib-

riumindifferentbetweencoursesofactionthathaveverydifferentaggregative

tically,onecanshutoffsomeofthesepossibilitiesby

introducingdiminishingreturnsintherightplaces,butIamnotsure

thatthese

multiplicitiesshouldbeviewedastheoreticaldefects,tobepatchedup.

Ifour

objectiveistounderstandaworldinwhichsimilarlysituatedeconomiesfollow

verydifferentpaths,ant

returns(atthelevel涿鹿之战 ofindividualproducingunits)learningspillovertechnology

askistounderstand

diversity,thisisanessentialfeature,notadeficiency.

Asecondattractionofthelearningspillovertechnologyisthatitisconsistent

withthestrongconnectionweobservebetweenrapidproductivitygrowthand

ertwosmalleconomiesfacingthesameworldprices

andsimilarlyendowed,ethat

Koreasomehowshiftsitsworkforceontotheproductionofgoodsnotformerly

producedthere,andcontinuestodoso,whilethePhilippinescontinues

to

cordingtothelearningspillovertheory,

1960,KoreanandPhilippine

incomeswereaboutthesame,sothemixofgoodstheirconsumersdemanded

sscenariotobepossible,Koreaneededtoopenup

alargedifferencebetweenthemixofgoodsproducedandthemixconsumed,

a

argevolumeof

tradeisessential

toalearning-basedgrowthepisode.

Onecanusethesamereasoningtoseewhyimport-substitutionpoliciesfail,

despitewhatcaninitiallyappeartobesuccess

in

stimulatinggrowth.

Consider

,JR.

aneconomythatexports,say,agriculturalproductsandimportsmostmanufac-

economyshiftstowardautarkythroughtariffandother

barriers,itsworkforcewillshifttoformerlyimportedgoodsandrapidlearning

sisaone-timestimulustoproductivity,andthereafterthemix

ofgoodsproducedinthisclosedsystemcanchangeonlyslowly,astheconsump-

atthisargumenthastodoonlywiththepaceof

changeinaneconomy\'tinvolvescale,thoughit

canobviouslybereinforcedbyscaleeconomies.

Idonotintendtheseconjecturesabouttheimplicationsofalearning

spillovertechnologyforsmallcountriesfacinggivenworldpricestobea

is,onewould

needtotakearealisticpositionontheseissuestouchedoninmydiscussionof

Rapping\'sandSearle\'thenatureofthehumancapital

accumulationdecisionproblemsfacedbyworkers,capitalists,andmanagers?

Whataretheexternalconsequencesofthedecisionstheytake?Thepapers

citedhereconsideravarietyofpossibleassumptionsontheseeconomicissues,

butitmustbesaidthatlittleisknown,andwithoutsuchknowledgethereis

littlewecansayaboutthewaypoliciesthataffectincentivescanbeexpectedto

influenceeconomicgrowth.

SIONS

Ibeganbyaskingwhatcurrenteconomictheoryhastosayaboutthegrowth

entliteratureonwhichIhavedrawntoanswer

thisquestionisfragmentary,,

theimageofthegrowthprocessandtheroleoftheseremarkableeconomies

withinthisprocessthatemergesis,Ithink,surprisinglysharp,certainlycom-

onclude

bysummarizingit.

Themainengineofgrowthistheaccumulationofhumancapital-of

knowl-

edge-and

themainsourceofdifferencesinlivingstandardsamongnationsis

alcapitalaccumulationplaysanessential

apitalaccumulationtakesplacein

schools,inresearchorganizations,andinthecourseofproducinggoodsand

engaging

in

isknownabouttherelativeimportanceofthese

differentmodesofaccumulation,butforunderstandingperiodsofveryrapid

growthinasingleeconomy,learningonthejobseemstobebyfarthemost

hlearningtooccuronasustained

basis,itisnecessarythat

workersandmanagerscontinuetotakeontasksthatarenewto

them,to

continuetomoveupwhatGrossmanandHelpmancallthe\"qualityladder.\"

Forthistobedoneonalarge

scale,theeconomymustbealargescaleexporter.

Thispicturehasthevirtueofbeingconsistentwiththerecentexperienceof

dbeequallyconsistentwithpost-1960

icturethatis

consistentwithanyindividualsmalleconomyfollowingtheEastAsianexample,

MAKINGAMIRACLE271

not

appeartobeconsistentwiththethirdworldasawholebeginningtogrow

at

EastAsianrates:Thereisazero-sumaspect,withinevitablemercantilist

overtones,toproductivitygrowthfueledbylearningbydoing.

Canthesetwoparagraphsbeviewedasasummaryofthingsthatareknown

abouteconomicgrowth?Afterall,theyaresimplyasketchofsomeof

the

propertiesofmathematicalmodels,

purelyfictionalworlds,thatcertain

soneacquireknowledge

aboutrealityby

workinginone\'sofficewithpen

andpaper?Thereismoretoit,ofcourse:Some

ofthenumbersIhavecitedareproductsofdecades-longresearchprojects,and

allofthemodelsIhavereviewedhavesharpimplicationsthatcouldbe,and

havenotbeen,,

Ithinkthisinventive,

model-buildingprocessweareengaged

inisanessentialone,andIcannot

imaginehowwecouldpossiblyorganize

andmakeuseofthemassofdata

derstandtheprocess

ofeconomicgrowth-or

ofanythingelse-weoughttobecapableofdemonstrating

thisknowledgeby

creatingitinthesepen

andpaper(andcomputer-equipped)laboratoriesof

owwhataneconomicmiracleis,

we

ought

tobeabletomakeone.

omics,TheUniversity

ofChicago,1126E.59thSt.,Chicago,IL

60637,U.S.A.

Manuscript

received

April,1992;

final

revisionreceived

August,1992.

REFERENCES

ARROW,

KENNETHJ.(1961):\"TheEconomicImplicationsofLearningbyDoing,\"Review

of

EconomicStudies,29,155-173.

BACKUS,

DAVIDK.,,

(1991):\"InSearchofScaleEffects

inTradeandGrowth,\"FederalReserveBankofMinneapolisWorkingPaper.

BARRO,

ROBERTJ.,

AND

XAVIER

SALA-I-MARTIN(1992):

\"Convergence,\"JournalofPolitical

Economy,100,223-251.

BAUMOL,

WILLIAMJ.(1986):

\"ProductivityGrowth,

Convergence,

andWelfare,\"AmericanEco-

nomicReview,76,1072-1085.

BAUMOL,

WILLIAMJ.,(1988):

\"ProductivityGrowth,

Convergence,

and

Welfare:

Reply,\"

AmericanEconomicReview,78,1155-1159.

BECKER,

GARYS.,KEVIN

M.

MURPHY,

ANDROBERTTAMURA(1990):\"Capital,

Fertility,

and

EconomicGrowth,\"JournalofPoliticalEconomy,98,S12-S37.

BEN-DAVID,

DAN(1991):

\"Equalizing

Exchange:

AStudyoftheEffects

ofTradeLiberalization,\"

National

BureauofEconomic

Research

Working

PaperNo.3706.

BOLDRIN,

MICHELE,KMAN(1988):\"Learning-by-Doing,

International

Tradeand

Growth:ANote,\"

inSantaFeInstituteStudies

intheSciencesofComplexity,TheEconomyas

anEvolvingComplexSystem,285-300.

BOSTONCONSULTING

GROUP(1968):Perspectiues

:

Boston

Consulting

Group.

CABALLE,

JORDI,

AND

MANUEL

(1991):\"OnEndogenous

GrowthwithPhysical

and

Human

Capital,\"WorkingPaper.

DELONG,RD(1988):\"ProductivityGrowth,

Convergence,

andWelfare:Comment,\"

AmericanEconomicReuiew,78,1138-1154.

DIRECTORATE-GENERAL

OF

BUDGET,

ACCOUNTING

AND

STATISTICS,

EXECUTIVEYUAN(1987):Na-

tionalIncomeinTaiwan

:

VeteransPrintingWorks.

272ROBERT

,

JR.

GROSSMAN,GENEM.,ANDELHANANHELPMAN(1991a):\"Quality

LaddersandProductCycles,\"

Quarterly

JournalofEconomics,106,557-586.

(1991b):InnovationandGrowthinthe

dge:MITPress.

HARBERGER,ARNOLDC.(1990):\"ReflectionsontheGrowth

Process,\"WorkingPaper,U.C.L.A.

JONES,LARRYE.,LI

(1990):\"AConvexModelofEquilibriumGrowth:

TheoryandPolicyImplications,\"Journal

ofPoliticalEconomy,98,1008-1038.

KING,ROBERTG.,ANDSERGIOREBELO(1990):\"PublicPolicyandEconomicGrowth:Developing

NeoclassicalImplications,\"JournalofPoliticalEconomy,

98,S126-S150.

KRUGMAN,PAULR.(1987):\"TheNarrowMovingBand,the

DutchDisease,andtheConsequences

er:NotesonTradeinthePresence

ofScaleEconomies,\"JournalofDevelopment

Economics,27,41-55.

LUCAS,ROBERT

E.,JR.(1988):\"Onthe

MechanicsofEconomicDevelopment,\"JournalofMonetary

Economics,22,3-42.

MATSUYAMA,

KIMINORI(1991):\"IncreasingReturns,

Industrialization,andIndeterminacy

ofEqui-

librium,\"Quarterly

JournalofEconomics,106,617-650.

(1992):\"AgriculturalProductivity,

ComparativeAdvantageandEconomicGrowth,\"

Journal

ofEconomicTheory,58,317-334.

MINCER,JACOB(1962):\"On-the-JobTraining:Costs,

Returns,andSomeImplications,\'\"

Journalof

PoliticalEconomy,70,S50-S79.

PARENTE,STEPHENL.,AND

Tr(1991):\"TechnologyAdoption

andGrowth,\"

NBERWorkingPaper.

RAPPING,LEONARDA.(1965):\"LearningandWorldWar

IIProductionFunctions,\"Reviewof

EconomicsandStatistics,47,81-86.

ROMER,PAUL(1986):\"Increasing

ReturnsandLong-RunGrowth,\"JournalofPoliticalEconomy,

94,1002-1037.

SCHULTZ,

THEODOREW.(1962):\"ReflectionsonInvestment

in

Man,\"Journalof

PoliticalEconomy,

70,S1-S8.

SEARLE,

ALLAND.(1945):\"ProductivityChangesinSelectedWartimeShipbuildingPrograms,\"

MonthlyLaborReview,61,

1132-1147.

SOLOw,ROBERT

M.(1956):\"AContributiontotheTheoryofEconomicGrowth,\"Quarterly

Journal

ofEconomics,70,

65-94.

STOKEY,NANCY

L.(1988):\"LearningbyDoingand

theIntroductionofNewGoods,\"Journalof

PoliticalEconomy,96,701-717.

(1991a):

\"HumanCapital,ProductQuality,andGrowth,\"Quarterly

JournalofEconomics,

106,587-616.

(1991b):\"TheVolume

andCompositionofTradeBetweenRichandPoor

Countries,\"

ReviewofEconomicStudies,58,

63-80.

SUMMERS,

ROBERT,

ANDALANHESTON(1991):

\"ThePennWorldTable(Mark5):AnExpandedSet

ofInternationalComparisons,1950-1988,\"Quarterly

JournalofEconomics,106,327-368.

THEWORLDBANK(1984,1986):World

:OxfordUniversity

Press.

UZAWA,

HIROFUMI(1965):\"Optimum

Technical

Change

inanAggregativeModelofEconomic

Growth,\"

InternationalEconomicReview,6,

18-31.

YOUNG,ALWYN,(1991a):\"LearningbyDoing

andthe

Dynamic

EffectsofInternationalTrade,\"

Quarterly

JournalofEconomics,106,369-406.

(1991b):

\"InventionandBoundedLearningbyDoing,\"

MITWorkingPaper.

(1992):

\"ATaleofTwoCities:FactorAccumulation

andTechnical

Change

inHongKong

andSingapore,\"NBERMacroeconomics

Annual1992,forthcoming.