Vous êtes sur la page 1sur 38

Accounting for Fertility Decline during the Transition to Growth Author(s): Matthias Doepke Reviewed work(s): Source: Journal

of Economic Growth, Vol. 9, No. 3 (Sep., 2004), pp. 347-383 Published by: Springer Stable URL: http://www.jstor.org/stable/40215875 . Accessed: 06/12/2012 12:16
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Journal of Economic Growth.

http://www.jstor.org

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

& W

ofEconomic Journal Growth, 9, 347-383,2004 Publishers. Academic Manufactured in The Netherlands. 2004 Kluwer

forFertility DeclineDuringthe Accounting Transition to Growth


MATTHIAS DOEPKE
doepke@econ.ucla.edu Los Angeles, CA90095 -J UCLA, 477, USA Department ofEconomics, Centre 90-98 Goswell ECIV 7RR,United Research, Road,London forEconomic Policy Kingdom In every theeconomic transition from to modemgrowth was developed country, pre-industrial stagnation transition from tolowfertility. Eventhough the is repeated, overall accompanied bya demographic high pattern arelarge inthe there variations andspeedofthe transition. What accounts for cross-country timing demographic thetransition to growth? To answer thisquestion, thispaperdevelops a unified fertility falling during growth model that delivers a transition from togrowth, Themodel is used stagnation bydeclining accompanied fertility. to determine whether that affect theopportunity costofeducation can account for crossgovernment policies infertility variations decline. the education subsidies tohaveonly arefound considered, country Among policies minor while for child labor is crucial. from thepolicies effects, accounting regulation Apart influencing fertility, also determine theevolution oftheincome distribution in thecourse ofdevelopment. child Growth, education, labor, Keywords: fertility, inequality JEL classification: 120, J13, O14,O40

1. Introduction decline is a universal feature of development. industrialized has Fertility Every country a demographic transition from to low a experienced high fertility, by large accompanied risein lifeexpectancy. Mostdeveloping arein themidst countries oftheir demographic transition timeago, economic and demographic today.Untila short changeduring in isolation: werestudied researchers economic tended to development studying growth abstract from while concentrated onexplanations for dynamics, population demographers the transition. More havestarted economists torecognize demographic recently, important between interactions and economic and demographic change, responded bydeveloping unified models which both the economic takeoff from togrowth and encompass stagnation thedemographic transition from to low fertility. high for behavior is an important for theories ofdevelopment Accounting fertility challenge because theeconomic affects ofa country ina number demographic change performance of ways.The most familiar concern is that dilutes thestock of high population growth andtherefore exerts a negative effect onincome Aneven more capital, physical percapita. channel works theaccumulation ofhuman rates important through capital. Highfertility tend to be associated with low education; countries with a highfertility ratetherefore accumulate lesshuman A third channel from the transition togrowth capital. demographic works inthe ofthe decline lowers through changes agestructure population. Rapidfertility thedependency ratiosinceinitially boththeold-ageand child-age cohorts are small

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

348
6.5 |-n^r, , 1 i

MATTHIASDOEPKE =r\

\ ^v. 60 - v \ 55 5.0* 45 4.0 3.53.02.5 2.0 Lj 1,000 \

Brazil

"""^-^^^ *^. \^ \ xv ^
s

s^

England Korea Nv ^s. -V \


\

" * ^ .^

\. \

%V

1 1,500

. 2,000

GDPper capita

. 2,500

1 3,000

1 3,500

4,000

1. Fertility decline from inBrazil, andEngland. Thegraph covers toGDP percapita Korea, Figure peakrelative theperiods1820-1914forEngland, 1945-1980forBrazil,and 1960-1985forKorea.Data approximately sources: Korean andBrazil:Fertility from from Chesnais Perm World Tables.England: (1992),GDP percapita from Chesnais from Maddison (1991). (1992)andLee andSchofield (1981),GDP percapita Fertility

relative to the working-age population.Hence, countriesthat undergoa fast fertility transition a if temporary, boost to theirlevel of outputper capita, sizable, experience because the size of the labor forceincreasesfaster thanthepopulationas a whole.1 The focus of this paper is to understand countriesundergodifferent why different is repeated, thespeed and transitions. theoverallpattern demographic Despitethefactthat thetiming to growth differs offertility declineduring thetransition from widely stagnation the paper is the across countries.An example thatwill be used repeatedly throughout restof paper). betweenBrazil and SouthKorea (referred to as Korea throughout contrast From the 1950s to the 1980s, the two countrieshad a verysimilargrowth experience. the late in incomepercapita started at thesame time,and throughout Substantial growth to be "miracle economies," withgrowth 1960s and 1970s bothcountries wereconsidered was similaras ratesof outputper capita exceeding5 percent fertility per year.Initially, rate of 6.0 in both countriesin 1960. After1960, however, well, witha total fertility decline was much slowerand more started to drop fastin Korea, while fertility fertility in out Brazil. spread As a raterelative to GDP percapitain thetwocountries. Figure1 plotsthetotalfertility from data fromEngland, the firstcountryto experience the transition benchmark,

ofEastAsia's for a large this cohort effect accounts 1 BloomandWilliamson that (1998) argue proportion miracle" between 1960and 1990. "economic

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITYDECLINE DURINGTHE TRANSITIONTO GROWTH ACCOUNTING

349

to growth, is also includedin thegraph.Relativeto incomeper capita,fertility stagnation fellmuchfaster in Korea thanit had in England,whereasthedeclinewas muchslowerin Brazil. Throughout most of the transition, for a given level of income per capita the differential between Brazil and Korea exceeds two children Similar fertility per family.2 in the speed of fertility variations decline can be observedacross manyothercountries. The transition was especially rapid in the East Asian *'miracle" economies such as include Malaysia, Taiwan, Hong Kong, and Singapore.Examples of a slow transition Mexico, and Costa Rica. Turkey, transition in Korea thanin Brazil? Why did the demographic proceed so much faster The aim of this paper is to explore the ability of one specific explanation,namely differences in educationaland child labor policies, to explain cross-country in variations forthisexplanation thespeed of fertility decline.3The motivation is twofold. most First, economicmodels of fertility choice are builton thenotionof a quantity-quality tradeoff of children and educationper child (see Becker and Barro, 1988; betweenthe number and educationare indeed joint decisions, government Becker et al., 1990). If fertility the opportunity cost of educationshould have a first-order effect on policies thataffect fertility. in educationaland child labor policies Second, we do in factobservelarge variations the across countries transition to Most countries introduce educationand during growth. child labor reforms at some pointduring their but the extent and timing of development, thesereforms varieswidely.Brazil and Korea are once again drastic in examples.Starting themid-1950s(after theKoreanwar), thetwo countries werepolar oppositesin terms of theireducationaland child labor policies. In Korea, child labor was almostcompletely eradicatedby 1960, and large amountsof resourceswere devoted to buildinga public education system.In termsof educational outcomes (enrollment rates, literacyrates, average schooling) Korea was far ahead of othercountriesat a comparable level of littleon basic education,and Brazil, on the otherhand, spentrelatively development. in termsof educationaloutcomes.Child labor lagged farbehindcomparablecountries laws werelax and looselyenforced, and as a consequencechildlaborwas widespread well intothe 1990s.4 In orderto examine the role of educationalpolicies duringthe fertility I transition, a phase of Malthusianstagnation model thatgenerates combinedwith develop a growth followedby a transition to a growth The model is highfertility, regimeand low fertility.

2 The figure wouldlookqualitatively similar if we considered thenumber of surviving children (say,to ofthenumber of live births thefraction of children (thefertility rate).In England, age 5) instead dying until in 1861,and then fellto 19 percent and Schoen, age 5 was 27 percent by 1911(Preston, Keyfitz, 12 percent of children woulddie by age 5 in 1960,and by 1980 1972).In Braziland Korea,18 versus these numbers haddeclined to 8 versus 3 percent Indicators of (World 2000). The ranking Development thethree transitions is closerto Korea in terms of surviving children staysintact, although England becauseofhigher child mortality. 3 In concentrating on theroleof political I abstract from other factors that couldalso affect the variables, transition. The aim hereis to derivethe implications of different educational fertility policieson the in isolation; transition theanalysis doesnotruleoutthat other factors couldalso matter. fertility 4 See Section a description ofeducational 6 for andoutcomes in BrazilandKorea. policies

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

350

MATTHIAS DOEPKE

and educationsubsidiescan used to explorethe degree to whichchild laborrestrictions and the speed of fertility decline.The accountforcross-country variations in thetiming An elements: theoretical framework consistsofthree function, agricultural production key model. The main role of an industrial and a function, fertility quantity-quality production the technology side of the model,whichderivesfromHansen and Prescott (2002), is to to growth. The preference side of the fromstagnation generatean economic transition introduced model is based on thedynastic framework by Beckerand Barro(1988). utility and on theeducationlevel oftheir Parents are altruistic and decide on thenumber children, is essential for generating tradeoff. This tradeoff thatis, they face a quantity-quality to growth. decline duringthetransition fertility the To assess whether the effects of government important, policies are quantitatively the calibratedmodel underdifferent model is calibratedto data. By simulating policy thefertility transition. I examinehow educationaland child laborpolicies affect regimes, on fertility The main finding is thatthese policies indeed have large effects duringthe the and childlaboris unrestricted, If parents have topay forschooling transition to growth. if educationis publicly transition starts laterand progresses slowly.In contrast, fertility declinesrapidly. on childlaborare strict, Quantitatively, fertility providedand restrictions educationsubsidieshave a relatively minorimpact,whereasaccountingforchild labor restrictions turns out to be crucial. thatwereinstituted at thestart I also examinewhathappensifthesame policyreforms ' of thetake-off in Korea are introduced witha delay.This *'English' regimeis designedto before 1800, whereasthe main started capturethe factthatin England industrialization after about 1870. In themodel,witha delayedpolicy policychangeswereonlyintroduced and declinesrapidly once the reform fertility stayshighin theearlyphase ofthetransition, in England of fertility withtheevolution is consistent reforms are introduced. This pattern on the also has largeeffects between1800 and 1914. The timing of theeducationreform of thetransition, incomedistribution. If thereform takesplace at thestart stays inequality a Kuznetscurveemerges:inequality thetransition. Witha delayedreform, low throughout increasesrapidly, and thendecreasesonce thepolicies are changed. first theopportunity cost of education Our resultsshow thatgovernment policies thataffect transition. in the differences are a promising for demographic explanation cross-country if thetrue we know as a test of also serve differences can theory: cross-country Addressing mechanismthatexplains fertility decline, we should also be able to explain why the in thispaperlend To thisend,theresults declinediffered so muchacrosscountries. fertility choice in general.Finally,the results model of fertility to the quality-quantity support debate. The same policies that on the growth-and-inequality providea new perspective influence choice in the model also have large effectson the evolutionof the fertility forcebehindthedistributional The maindriving thetransition. incomedistribution during between skilled and unskilledparents. differential effectsis the endogenous fertility effects on therelative educationalpolicies have large long-run Throughthisdifferential, on the skill premium.5 number of skilledand unskilledpeople, and therefore

role in Moav (2005), who pointsto cross-country 5 Fertility differentials fertility play an important andde la Croixand Doepke(2003), who forpersistent income differences as an explanation inequality,

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITYDECLINE DURINGTHE TRANSITIONTO GROWTH ACCOUNTING

35 1

This research adds to an emerging literature on long-run and population growth Galor andWeil(2000),KogelandPrskawetz and (2001),Jones (2001),Hansen dynamics. Prescott that models a transition from (2002),andTamura (2002)all develop generate pretomodern industrial transition. To the stagnation growth, accompanied bya demographic extent that is these models also have the feature that a in the rise fertilityendogenous, return tohuman is the force behind decline. Alternative theories of capital driving fertility declineare basedon changes in gender roles(Galorand Weil,1996;Lagerlof fertility motive (Morand, 2003), the old-agesecurity 1999), and changesin childmortality on childlabor is shared with HazanandBerdugo (Kalemli-Oczan, 2002).6Ouremphasis a rising between children and adultsas the (2002), who emphasize wage differential force for decline. While the literature has not driving fertility existing generally in fertility concentrated on cross-country differences the in these mechanisms decline, a number of alternative for such differences. For papers suggest explanations example, in preferences variation forchildquality, thecostof children, theskillcross-country ofproduction orthe function for human couldall intensity technologies, production capital havesomeimpact onfertility the transition. Someofthese areendogenized factors during in Galorand Moav (2002),whodevelop an evolutionary account of thedistribution of in a population, forchildquality and in Galorand Mountford (2003),who preferences concentrate on theinteraction of international thedemand forskilled and trade, labor, growth. population The restof thepaperis organized as follows. The following section introduces the model. Section 3 derives a number oftheoretical 4 ofthemodel, andSection properties discusses thebehavior ofthemodel intheMalthusian thegrowth andthe regime, regime, thetwo. InSection transition between thecalibration andSection 6 5,1discuss procedure, usesthe calibrated model toassessthe effect ofgovernment thetransition. policies during Section 7 concludes.

2. The Model Theeconomy is populated ofpeople wholivefor twoperiods, byoverlapping generations childhood andadulthood. Children receive education orwork, do not and enjoy anyutility, do not to decide Adults can be either skilled or on their unskilled, get anything. depending In eachperiod education. there is a continuum ofadults ofeachtype; of Ns is themeasure skilledadults,and Nv is the measure of unskilled adults.Adultsdecide on their labor andon thenumber andtheeducation oftheir children. consumption, supply,
linkfertility differentials within a country to economic Interactions between differentials growth. fertility and inequality are also thefocusof Dahanand Tsiddon and Chen(2002), and Veloso (1998), Kremer (1999). 6 Otherrecentstudieswhichuse quantitative to evaluategrowth models withendogenous theory include FernandezVillaverde for (2001) on the role of capital-skill population complementarity Greenwood andSeshadri andNgai decline, transition, (2002) on theUS demographic explaining fertility totechnology (2000) on barriers adoption.

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

352

MATTHIASDOEPKE

2.1. Technology
The single consumptiongood in this economy can be produced with two different methods.There is an agricultural thatuses skilledlabor,unskilledlabor,and technology land as inputs,and an industrialtechnologythat only uses the two types of labor. Productionin each sector is carried out by competitive firms. The main task of the fromstagnation to growth:the revolution technologysetup is to deliver an industrial takeoff to be takesplace once the industrial becomes sufficiently technology productive introduced alongsideagriculture. The firm-level returns to scale, and is industrial function exhibitsconstant production given by:

y,= A>(is?-"(ivr,

(i)

whereytis output (/standsforIndustry), parameter, Aj is a productivity ls and lv are inputs of skilledand unskilledlabor,and theparameter a satisfies 0 < a < 1. Since there are no ~a = where and industrial is externalities, aggregate output YI Aj (LIS)1 (L/t/)a, LIS Lw are the aggregateamountsof skilled and unskilledlabor employedin the industrial sector. in The Since I wantto abstract there is no the function. from production bequests, capital setup is equivalent,however,to a model with capital underthe small-open-economy whichis the approachtakenby Galor and Weil (2000).7 assumption, The agricultural sectoruses thetwo typesof laborand land.The aggregate agricultural function is given by: production YF =AF {LFSf> (LFU)0" (Z)1-*-"". (2)

Here YF is agricultural (F standsforFarm), parameter, output AF is a productivity LFS and in the are the and unskilled labor amounts of skilled agricultural employed aggregate LFU sector,and Z is thetotalamountof land. I assume thattheparameters satisfy 9s,6u>0 and^-f 0/<l. To abstractfromland ownershipand bequests,I assume thatland is a public good. the agricultural From the perspectiveof a small individualfirm technology, operating is a limited amountof land,labor there are constant returns to labor.However,sincethere On the level of an on all otherfirms. inputby one firm imposes a negativeexternality individualfirm, forgivenlabor inputsls and lv of skilledand unskilledlaboroutput yF is given by:

yF-AF(ls)^(lu)^

(3)

on akf and thefixed return if theindustrial function is y = B [(ls)} t(/(/)a]1 7 Specifically, production that condition is equal to r. Solving k is given suchthat itsmarginal product by r,capital adjusts capital function thisback intotheproduction fork givesk = {aB/r)x^x~a\ls)l~a(lu)a, and plugging gives:
y = fl(1/i-)(fl/r)(/>--)/i-ij

The same argument whichis (1) by setting applies to the agricultural A, =B(1/1~a)(a/r)(a/1~fl). function. production

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

ACCOUNTING FOR FERTILITY DECLINE DURINGTHE TRANSITIONTO GROWTH

353

where AF=AF [(LFS)\LFV)9]-i^ (Z)1-^-"".

effect amount of labor has a negative on theproductivity ofan Thusthetotal employed external effect waschosen such that the individual individual firm. Thespecific form ofthe in thispaperis functions to (2) above.As faras theanalysis (3) aggregate production function is the main feature ofthe returns to concerned, agricultural production decreasing of decreasing returns is essential forgenerating theMalthusian labor.The assumption I assume that theindustrial sector is moreskill-intensive than theagricultural regime.8 sector: 1 The industrial thatis, theproduction sectoris moreskill-intensive, Assumption 1 - a > 9S. which implies function parameters satisfy (x<9Ut willbe important for decline thetransition to Thisassumption generating fertility during of bothtechnologies thegrowth The productivities growat constant, though regime. rates: different, possibly
A'F= yFAF, A't=yiAh (4)

of exogenous and constant where > 1. The simplifying productivity assumption yF,y7 that thespecific source ofproductivity does is madetoemphasize improvements growth features ofthemodel; itis only that notmatter for thequalitative necessary productivity to assumea changein therateof takesplace at all. Neither is it necessary growth to growth. Whileproductivity to the switch from growth explain stagnation productivity be of as an the economic mechanisms should variable, ultimately thought endogenous in this do notdepend on a particular oftechnological engine progress. explored paper Thestate x in this consists oftheproductivity levels vector economy AF andAj inthe andunskilled andindustrial andthemeasures sectors, agricultural Ns andNy of skilled restriction onthestate vector is that ithastoconsist adults: x = {AF,AhNSi Ny}. Theonly thestate this is given ofnonnegative numbers. Therefore spaceX for economy byX= R+. will in Inequilibrium, are a function of the state. It be shown 1 below wages Proposition willalways intheagricultural while theindustrial sector is that firms be operating sector, if thewagessatisfy thecondition onlyoperated WsC*)1*wu(x)a^At (1 -a)1"'^. in sector to the Theproblem of a firm ye {F, /}, is to maximize y,where profits subject Profit maximization that function, production taking wagesas given. implies wagesequal ineachsector. demand as a function labor ofthestate, for the marginal products Writing we the conditions: sector agricultural get following
of an external effect from on theother is notessential, and is usedonlyto 8 The assumption labor, hand, abstract from land ownership. The modelis equivalent to a setupin whichall land is ownedby ora separate class. foreigners land-owning

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

354

MATTHIASDOEPKE

^~*>Kki^z'--"-

(5)

in in total Notice that these labor wagesaredecreasing supply (ifLFUandLFSarevaried is operating, sinceQv+ 0S< 1. If theindustrial sector wageshaveto equal proportion) as well: marginal products

ws(x)=A,(l-z)(^Qy " Wu(x)=A,x(j^y

if Lls{x),Lw{x) > 0, if Lls(x),Llu(x)>0.

(7) (8)

since In this andunskilled labor on theratio ofskilled sector, supply wagesonly depend in thefixed factor landdoes notenter. Instead of writing out thefirms' problem the I willimpose conditions. definition ofan equilibrium below, (5)-(8) as equilibrium 2.2. Preferences and Policies and careabout I willnowturn Adults tothedecision oftheadults. consumption problem is anextension ofBecker Thepreference structure thenumber andutility oftheir children. of the The mainobjective of children. and Barro(1988) to thecase of different types the number ofchildren between a quantity-quality tradeoff is togenerate preference setup andchildlabor is essential for educational andeducation Thistradeoff policies perchild. to havean effect. inthe is decreasing factor andthediscount Adults discount theutility oftheir children, is the smaller adult In other children an the more number ofchildren. has, words, already units of adult who consumes c of an from another child. The theadditional utility utility children is given andnv unskilled theconsumption by: goodandhas ns skilled will children skilled andI assume 0 < ft< 1,0 < a < 1,and0 < s < 1. V'sis theutility foreseen both ofunskilled andVyis theutility children, bythe perfectly enjoyas adults, toconsumption, with ofutility theelasticity Theparameter a determines P respect parent. number tothe with of altruism and e is the is the level of altruism, respect elasticity general andaretherefore ofparents ofthecontrol ofchildren. Theutilities V'sandVv areoutside in thenext state vector on the The utility ofchildren taken as given. aggregate depends cannot be influenced is a continuum ofpeople, andsincethere byany aggregates period, ofpeople. finite number time between their andthey allocate oftime, areendowed with oneunit Adults working oftime. interms ofgoodsandinterms arecostly, both andchild-raising. Children Raising

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

ACCOUNTING FOR FERTILITY DECLINE DURING THE TRANSITION TO GROWTH

355

each childtakesp > 0 units oftheconsumption (j>> 0 ofthetotaltime good and a fraction available to an adult. Adults also have to decide on the education of theirchildren. need a skilledteacher to become skilled.It takesa fraction of a skilledadult's Children <j>s ifparents wantskilledchildren, timeto teachone child.Therefore, theyhave to sendtheir children to school and pay the skilled teacher.Childrenwho do not go to school stay childhood.Children can perform unskilled and workduring onlythe unskilledtask,and labor.The parameter is smallerthan one working child supplies(j)v unitsof unskilled <j>y birth and since theyare notas productive one since children do notworkfrom as onward, even after is stilla net adults.I also assume<j)u< (j>, so that forchildlaborthere accounting cost associatedwithhavingunskilledchildren. and an There are two government policies in the model: a child labor restriction educationsubsidy. Withthesepolicies,thegovernment can influence bothcomponents of theopportunity of child cost of educationfora child:thevalue of a child's timein terms amountsto loweringthe labor,and the directschoolingcost. A child labor restriction The government chooses a function how much parameter <^( ) whichdetermines <t>u. timechildren can only lower the legal work,dependingon the state.Since restrictions < <>v amount of childlabor,I requirethat0 < <t>u(x) forall x? The government also has theoptionof subsidizing a fixedamountof theschoolingcost forall children at school. This expenditure is financed witha flatincome tax, and budgetbalance is observedin The chooses a function b thatdetermines the fraction of the everyperiod. government where0 < S(x) < 1 forall x. Contingent on schoolingcost to be paid by thegovernment, thisfunction, the flattax t is chosen to observebudgetbalance. Withtaxes and thesubsidy, thebudgetconstraint of an adultof type/is givenby: c + p (ns + nv) + (1 - S(x)) <f)sws(x) ns

< (1- t(jc)) + *))*,(*) + 4>u to "Y/to [(1- <f>{ns */]

(9)

The right-hand side is the after-tax income of the adult plus the income fromsending unskilled children to work.Notice thatthetimecost <j> foreach childhas to be subtracted from thetimeendowment to computelaborsupply. On theleft-hand side are consumption, thegoods costforeach child,and thepartoftheeducation costfortheskilledchildren that is paid bytheparents. adultsare notrestricted to choose integer Forsimplicity, numbers of children. in thismodel. Whether Also noticethatthereis no uncertainty a childbecomes skilleddoes notdependon chance or unobserved butis underfullcontrol of the abilities, parent. In equilibrium, thewages and theutilities of skilledand unskilled people are functions of the statevector.The maximization problemof an adultof type/,whereie {5, U}, is describedby thefollowing Bellman equation:

= max {cff + fi(ns + nv)-*[nsVstf) + % W)]} Vt(x) c,nu,ns>}j


law of motionx! = g(x). (9) and theequilibrium subjectto thebudgetconstraint

In the applicationsbelow, I will considera one-timechange in child labor policy.

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

356

MATTHIASDOEPKE

wehavetoplacea joint restriction Forthe ofa parent tobe always welldefined, problem factor is nothigher than one.In ontheparameters that ensures that theeffective discount of their children than on their other adults on the cannot words, utility weight placehigher of we have to on the number ownutility. Since thediscount factor children, depends adult which is reached consider thehighest number ofchildren, byan unskilled possible is 1/(0 - <j>u). number ofchildren whospends all income on children. Theresulting
and ytsatisfy: e, a, (f>, Assumption 2 The parametersft, <f>Uf

r(jh;)'"<lafter that is reached thebalanced Here(?/)*is thegrowth rate ofutility growth path along theswitch totheindustrial technology. leadsto a market decisions makeeducational The fact that notchildren, only parents, to borrow funds to finance children would be able With markets, imperfection. perfect between In equilibrium, would havetobe indifferent their owneducation. children going adults would be equalized. Since andunskilled toschool ornot, andnetincome ofskilled is the market inability income there areno differences orstochastic shocks, imperfection write that outthe inthis I alsorule tocreate model. possibility parents necessary inequality richer funds from couldborrow contracts thatbindtheir children. Otherwise, parents ofthelender. andhavetheir children adults, paybacktheloantothechildren corner solutions. hasonly in Section theadults' It willbe shown 3 belowthat problem there arenever both skilled ofthem; ornone Adults either sendall their children toschool, adults of a that It is possible, children within thesamefamily. andunskilled however, children to schoolornone.In all their between arejustindifferent sending specific type while others tohaveskilled decide that ofa given children, case,someparents might type all willbe that situation In equilibrium, thetypical choosetheunskilled variety. might with unskilled unskilled while there areboth skilled haveskilled children, parents parents is In other there to school. children children andunskilled whosendtheir words, parents mobility. upward intergenerational ofadults ofeach ofthefractions I havetokeeptrack In the definition ofanequilibrium of the fraction The function and children. who have skilled unskilled gives type ^,_>y() of x. Foreachtype ofthestate oftype adults oftype i whohavechildren j, as a function fractions haveto sumup to one: andfor all xeX these parent

+ ^i/to = lu^sW+*u^u(x)= I*s-*sto

0)

) givesthenumber whohaveyfor ofchildren The policy function /-type parents /iy-(i, the state. children as a function of type the are thelabor-market conditions The remaining constraints, clearing equilibrium and ofskilled for thenumber andthelawsofmotion ofthegovernment, constraint budget in aregiven ofan equilibrium definition anda formal unskilled adults. Theseconstraints A. Appendix

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

ACCOUNTING FOR FERTILITYDECLINE DURINGTHE TRANSITION TO GROWTH

357

3. Analytical Results willbe useful a number oftheoretical results that for the Thissection derives describing I behavior the model. the two and then I of First, sectors, analyze production equilibrium of an in the All are in turn to thedecision adult contained the economy. problem proofs Appendix. is that while theagricultural sector is always On thetechnology side,themainresult firms is sufficiently industrial onlyif industrial productivity operating, produce high relative towages.Thefollowing derive theconditions that arenecessary for propositions andagriculture. in industry production 1 Firmswillbe operating in theindustrial sector and onlyiftheskilled Proposition thecondition: unskilled wagesws(x)and wv(x)satisfy

wsto1 ""*(*)" <Ay(l- a)1-*".


In theagricultural willbe operating sector ; firms given anywages.

(11)

theindustrial willbe operated for a given It is easyto check whether sector of supply and unskilled labor.We can use conditions skilled (5) and (6) to compute wages in theassumption that there is agricultural Iftheresulting under production only. agriculture condition is used.Skilled andunskilled labor (11),theindustrial technology wagessatisfy is allocated suchthatthewage foreach skillis equalizedacrossthetwo sectors. If in is takes condition (11) violated, production place agriculture only. In equilibrium, is used.A feedback between initially onlytheagricultural technology and leads to a state where the size is quasi-steady population wages population growing there is positive in industry, and wagesare constant. Giventhat at productivity growth somepoint theindustrial becomes to be introduced sufficiently technology productive andan "industrial revolution" occurs. the takeoff arises from Thus, alongside agriculture, an interaction between theproperties of thetwoproduction sectors and thepopulation in themodel. Sincepopulation is determined I willnow decisions, dynamics byfertility an in turn to thedecision of adult the model problem economy. ofviewofan adult, theutility ofa potential From the skilled orunskilled child is point that cannot be influenced. There areno individual state and variables, given bya number theutility ofchildren is determined in theaggregate, decisions which adults byfertility take as given sincethere is a continuum ofpeople. Thisallowsus toanalyze thedecision of an adultwithout first to solve fora complete In this having problem equilibrium. we willanalyze thedecision ofan adult whoreceives section, (after-tax) problem wage w > 0 andwhoknows that skilled children willreceive utility period, Vs > 0 inthenext whereas unskilled children can expectVv > 0. I restrict attention to positive utilities becauseifchildren receive zeroutility, itis optimal nottohaveanychildren. In order to in terms of the keep notation simple,I will expressthe cost of children directly a skilled child is/?s, andthecostfor an unskilled child is consumption good.Thecostfor denoted as pv. Forexample, without we haveps = <j>w + <t>sws +p government policies

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

358

MATTHIAS DOEPKE

- ^>(jWv children aremore andPu = (j)w > pv\skilled + p. Wealways expensive have/?5 than unskilled children. as: ofan adult Wecan nowwrite themaximization problem

- PunuY ma* + nuvu\}+ P(ns + nu) ~B[nsvs J(w""Psns

02)

as choosing the theadults is to imagine An alternative this wayof formulating problem this costthat children andthe total education costE they onraising they fraction/of spend of children is thengivenby ns =fE/ps and The number spendon skilledchildren. oftheadult the maximization In this formulation, problem equivalent nv (1 -f)Elpu. is:

+ 0 ~f)Vu/Pu]}+ + o<Emaxf<i{(w-E) pEl~WPs (1 -f)/Pu)~'[fVs/Ps


(13) has only corner solutions: ofan adult The first is that thedecision result problem in(13) wehaveeither attains the maximum 2 Forany f= 0 Proposition pair {, /} that orf=l. in one mixbothtypes butnever Adults chooseeither skilled or unskilled children, is e = 0 (which with themodel result Wecangainintuition for this byconsidering family. and children from theutility in which case both outbyassumption), ruled having gained If we have of children. in thenumbers of thetwotypes are linear thecostof children = andany children and skilled unskilled between theadultis indifferent Vs/ps Vy/pu, the e > as if have we now convex combination ofthetwotypes. 0, assumed, However, will and the adult function a convex m becomes term + of/, (f/ps (1 -f)/Pu)~ 03) choosea corner solution. can be number of children theoptimal are onlycorner Giventhatthere solutions, are only there that choicesassuming theoptimal determined by separately computing which Wecanthen children. unskilled oronlyskilled utility. yields higher type compare i solve: oftype Parents whodecideto havechildren
0<nj<w/pj I

max hw-ptnM'+ ftntY-V,}) (14)

as: The first-order condition can be written

P(l - eXw-pnY-'V,= opfa)'.

a for condition andthesecond-order this There is a unique equation, positive ntsolving < The < s 1. < < 1 and 0 a 0 that we assume maximum is satisfied optimal given Thus in thechildren's number of children utility V{ andin thewagew>. ntis increasing ifthecostofchildren model. On theother children area normal hand, pfis goodin this with thewage.Thus decreases number ofchildren the tothewagewy optimal proportional theincome effect ifthecostof children is a puretime cost,thesubstitution outweighs effect.

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITY DECLINE DURINGTHE TRANSITIONTO GROWTH ACCOUNTING

359

Ifanadult skilled andunskilled thetotal is indifferent between on children, expenditure children doesnotdepend on thetype ofthechildren. 3 Anadult is indifferent between skilled andunskilled children Proposition ifandonly if children thecosts and utilities satisfy: of Vs

(PS)'"8" W-"

...

Vv

(15) ( '

is indifferent, thetotal on children doesnotdepend on thetype expenditure Ifan adult of children that is chosen. 1 and 3 have implications forintergenerational in themodel. mobility Propositions states that for utilities of skilled and unskilled children 3 the ratio ofthe given Proposition children determines ofskilled andunskilled whether sendtheir children to prices parents than school. As longas thewageforskilled laboris higher theunskilled wage,skilled children arerelatively for skilled sincews > wv implies: cheaper parents, +P + <t>sws +p + <t>swS <j)Ws ' ^ <t>Wy +p + </)WS(fryWup (frWy(fryWu ratio Theterm onthe left-hand sideis the ofthe for skilled andunskilled children for prices skilled andtheright-hand sideis theratio for unskilled adults. Thecostoftime is adults, than for skilled adults because theskilled theunskilled higher wageis higher wage.Since theopportunity costofchildrearing makesup a larger fraction ofthecostof unskilled more unskilled children are for skilled children, relatively expensive parents. Sincethe relative ofskilled andunskilled children differs for skilled andunskilled price itcannot be the casethat both ofadults areindifferent between the twotypes parents, types at thesame time.Since skilled children are relatively of children forskilled cheaper in equilibrium thereare alwaysskilledparents who have skilledchildren. parents, there children atall,which inequilibrium. would be noskilled cannot Otherwise, happen there arealways unskilled adults with unskilled children. Likewise, these facts three situations canarisein anygiven The exactly Taking together, period. first is that skilled skilled while unskilled children, possibility parents strictly prefer In thatcase, thereis no intergenerational unskilled children. parents strictly prefer The secondpossibility is that skilled are indifferent between thetwo mobility. parents while all unskilled haveunskilled ofchildren, children. The third is parents types option that all skilled have skilled whiletheunskilled adults are indifferent children, parents between thetwotypes. Thislastcase is thetypical onealongan equilibrium as will path, be explained in moredetaillater.In thissituation, there is upward intergenerational because someunskilled adults haveskilled butnodownward children, mobility, mobility. Thefollowing sums ofthese results for an equilibrium. proposition uptheimplications 4 In equilibrium, xeX suchthat must Proposition forany, following ws(x) > wu(x),the be true:

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

360

MATTHIAS DOEPKE

A positive and a positive adultshas skilled children, fraction fraction ofskilled of unskilled adults has unskilled children:

*s-s(*Mi/->/to>0.
one type thetwotypes Just between ofadultcan be indifferent ofchildren: *s-c/M > implies ku^s{x) = 0, > 0 implies A(/_5(jc) ks^u(x) = 0. Specifically, ks^v(x) > implies: ' + + (d>ws(x) <l,sws(x) p\ -= Vs(g(x)) and^u-^s(x) > 0 implies:

ffrtfrfr) 4+ pV "=V5fe(x)) <I>sws(x) * - 0/H>i/M + py v<K/M MsM)

4. Outline oftheBehavior of theModel inindustry that the starts ata time when is lowcompared Assuming economy productivity The Malthusian to agriculture, theeconomy evolvesthrough three different regimes: In the Malthusian the and the stagetheindustrial regime, transition, growth regime. like the model behaves inefficient to time. Therefore is too be used for some technology Malthusian one in whichthere is an agricultural sector only.The economy displays - wages stagnate, in features and populationgrowthoffsetsany improvements productivity. theMalthusian There arethree of themodelthat First, regime. keyfeatures generate increases once that children area normal This ensures growth good. property population the in lead to Second, technology agricultural wages. improvementstechnology higher that force. Thisfeature returns to thesize ofthelabor exhibits higher implies decreasing state. theeconomy backtothesteady wages,which pushes population growth depresses it Without this eachchild. is a goodscostp for Thethird is that there cost, keyassumption andwages aheadofproductivity would that be possible growth, population growth stays inwhich feedback create a Malthusian these features tozero.10 Taken converge together, in but raise resulting wages, growth depresses growth, population rising wages population a steady in which are constant. state wages a in theMalthusian can be determined Thevaluesfor all variables bysolving regime
as a subsistence eachchildleadsto similar 10 The assumption ofa goodscostfor consumption implications taken whichis theroute constraint on adultconsumption, by Galorand Weil(2000), among imposed others.

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITY DECLINE DURINGTHE TRANSITION ACCOUNTING TO GROWTH

361

system of steady-state equations under the assumption that only the agricultural is used. For a wide range of parameter choices, thereis a solutionto the technology in which both and the ratio of skilledto unskilled adultsare steady-state equations wages In the steadystate,average fertility constant.11 is higherforunskilledadults.While all children are some unskilled skilledadultssendtheir to school,there withunskilled parents children and others withskilledchildren. and the Wages dependon preference parameters rateof agricultural butare independent of the level of productivity, growth productivity, is exactlyoffset since productivity by population growth growth. Even though function is used in theMalthusian onlytheagricultural production regime, the productivity of the industrialtechnologyincreases over time. At some point, in industry reachesa level at whichindustrial is profitable at the productivity production in that the Malthusian From that time the industrial on, wages prevail regime. technology will be used alongside the agricultural Since populationgrowthdoes not technology. to growwithproductivity in wages and incomepercapita start depresswages in industry, the industrial sector.While the model assumes thatthe growth rate of productivity in is used, all thatis necessaryis that is constanteven before the technology industry in industry is boundedaway from zero. This is a natural productivity growth assumption in industry and agriculture is complementary, in thesense that sincetechnology inventions whichare useful for agriculture also have industrial uses. For example, JamesWatt's contributions to thedesignof steamengineswould nothave been possible without prior discoveries in mechanics and metallurgy that were originallyaimed at improving (and perhapsmilitary) agricultural technology. The evolution of fertility and theincomedistribution once thetransition starts depends on the specificproperties of the industrial function. Given Assumption1 production is more skill-intensive than productionin agriculture), the (productionin industry of the industrial increases the introduction for skilled labor. technology wage premium This increasesthe returns to education,and more unskilledadults will choose to have in highersocial mobility. skilled children, The overall effecton fertility is resulting The increased demand for expensive skilled childrenwould tend to lower uncertain. butthensince wages start to grow,theutility of children relativeto their fertility, parents The transition can be influenced increases,which tendsto increase fertility. by public lower the relativecost of policy.Both an educationsubsidyand child labor restrictions and as a resultboth policies have a positive effecton the numberof skilledchildren, children on fertility, are different. Since a subsidy however, goingto school. The effects an educationsubsidyincreasesfertility lowersthe cost of children, among parentswho children wouldhave senttheir to school even without thesubsidy. At thesame time, more send theirchildren to school, whichwould tendto lower fertility as long as the parents cost of educationis not zero. The net effect is generallyambiguous;even opportunity in all computedexperiments rises in responseto an educationsubsidy. though, fertility

11 If theschooling it is possible costis very that no solution exists and thefraction of skilled adults high, to zero.Also,ifproductivity andthecostof children are very that converges growth highit is possible is nothigh to offset Neither case arisesin thecalibrated population growth enough productivity growth. model.

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

362

MATTHIAS DOEPKE

Child labor restrictions, on the otherhand,unambiguously increasethecost of children, of publicpolicies during and therefore lead to lowerfertility. The quantitative importance will be assessed below. the transition If productivity of output in industry is sufficiently high,thefraction producedin growth industrywill increase over time, until the agriculturalsector ultimatelybecomes negligible.The economywill thenreacha balanced growth path:thegrowth regime.Here the model will behave like one in whichthereis the industrial only,and all technology variablescan be computedby solvinga systemof balanced-growth equations.Whether is higher in thegrowth regimeis determined by the fertility regimethanin theMalthusian relative importanceof skill in the two technologies.If the industrialtechnologyis will go to school. Since in the growth skill-intensive, sufficiently regimemost children On the other and populationgrowth. schoolingis costly,thiswill tendto lower fertility This becomes negligible. hand,as wages grow,thephysicalcost p of children ultimately whichwill tendto increase effect makeschildren regime, relatively cheaperin thegrowth Unless the schoolingcost is veryhigh,the ratio of skilled to unskilledadults fertility. are reaches a fixed numberin the growthregime.12Populationgrowthand fertility technical and and at the rate of constant, progress.Average wages consumption grow is lower for skilled than for unskilledadults. This would be trueeven if the fertility schoolingcost were zero and if therewere no child labor,since thentheonlyremaining is a puretime cost of children wouldbe a timecost. It was shownin Section3 thatifthere With related. cost forhavingchildren, are negatively positiveschooling wages and fertility more and sincerelatively cost and childlabortherelative costof skilledchildren increases, differential increase the fertility skilled adults have skilled children,this will further adultsin the forunskilled is higher betweenthe two typesof adults.Given thatfertility children. adults have skilled some unskilled balancedgrowth it has to be the case that path, unskilled Otherwise the fraction of unskilledadultswould increaseover time.Therefore adultsare just indifferent betweenthetwo typesof children. In thelimit-economy without bytheratioof skilledto wages are determined agriculture, theratioof skilled unskilledlabor supply.The onlyrequiredstatevariablesare therefore be The setup can further and unskilledadults and the productivity level in industry. and is of the that the function form, constant-elasticity by noting simplified periodutility thatare thatwages are linear in the productivity level. This resultsin value functions in the industrial level: homogeneous productivity

withtheratioof to a one-dimensional This reducesthegrowth system, regimeessentially skilledto unskilledadultsas the statevariable. the I will now turn to thequestionof whether during publicpolicycan have largeeffects and simulatethe In the followingsections,I calibratethe model parameters, transition. on policies. model underdifferent assumptions
to zeroovertime. adults 12 Iftheschooling thenumber ofskilled costis toohigh, converges

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

TO GROWTH ACCOUNTING FOR FERTILITYDECLINE DURINGTHE TRANSITION

363

5. Calibration I describe In this section theprocedure for themodel SinceI use calibrating parameters. differences the toexamine whether across countries can account for different model policy it transition would be to choose to experiences, counterproductive parameters closely I choose match datafrom onespecific suchthat inthegrowth Rather, country. parameters matches certain features ofmodern industrialized inthe the model while countries, regime the resembles a pre-modern Malthusian model Forthegrowth I regime economy. regime, while for thepre-modern I rely usecurrent US data, on England, where economy mostly theavailable datais ofrelatively Sincethecalibration high quality. procedure pinsdown two ends of the time we can use the transition the for the line, only period testing model. I start the choices that are determined ofthe bydescribing parameter byfeatures growth The the rate of in the industrial sector, technological progress regime. parameter yh inthe the rate ofper-capita Inthe determines United States, growth output growth regime. realGDP percapita increased on average intheperiod from 1960 by1.9percent peryear to 1992.1therefore chosetheyearly rate ofproductivity intheindustrial sector to growth be 2 percent. Sincea model is 25 years, this of 1.64. period givesa valuefor yy there arestrict lawsandchild inthe Given that labor restrictions compulsory schooling United States theparameters for thegrowth are calibrated under the regime assumption that child labor is ruled out.The schooling costparameter determines thenumber of </)s andthestudent-teacher ratio in thegrowth teachers to theDigestof regime. According inthe Education Statistics ofEducation, United States teachers on (US Department 1998), all levelsof education makeup 1.5 percent of theAmerican andthere are population, in themodel, in therealworld about16 students eachteacher Since,unlike perteacher. teaches more than one generation of students, we cannot match both valuesat thesame Ifthe time. ratio ofteachers topopulation is matched, classsizeswould be toobig,andif classsizeis matched, there would be toomany teachers. As a compromise, I chose</>s tobe results ina classsize of21, and 1.7percent ofthepopulation areteachers.13 0.04,which Thetime cost<t> wasthen chosen tomatch thetotal onchildren intheUnited expenditure estimated for1992byHaveman andWolfe to their States, estimates, (1995).According were orabout 38 percent ofpercapita GDP.I chose $9,200 parental per-child expenditures tobe 0.155,which leadstothesameratio ofper-child toGDP percapita. <j> expenditures Knowles calibrates the same to other estimates of the costofchildren (1999) parameter andarrives at a similar valueof0.15. datafrom Jones that inBritain intotal thedifference 1975, (1982) finds Using fertility between women with education orlessandwomen with orhigher elementary secondary education is about valuefor theUnited is about States 0.5. 1 0.4,andthecorresponding chosethepreference and/? tobe consistent with a fertility differential of ex, parameters 0.5 between skilled andunskilled whohaveunskilled anda total children, parents parents rateof 2.0, which matches current in theUnited States. The choices fertility fertility c = 0.5,e = 0.5,andj8= 0.132areconsistent with these observations. theDigestofEducation in 1994total on education Statistics, According expenditures
13 Variations ofthemodelwhich match either statistic lead to essentially thesameoutcomes. exactly

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

364

MATTHIAS DOEPKE

orroughly twothirds were7.3 percent ofGDP, whilepublicexpenditures were4.8 percent, and like of thetotal.These numbers exclude expenditures students, textbooks by parents than for less two thirdsof all and transportation. The government therefore pays S of education costpaid bythe In themodel,I chose thefraction educationalexpenditures. to be 0.5. government The technology a, the share of unskilledlabor in the industrial production parameter It determines theratioof skilledto unskilled function, regime. people in thegrowth mainly is hardto matchthisratioto data, since thereare more thantwo skill levels in the real world.If we defineskill to mean completedhigh school, skilledpeople would make up most of the population,since alreadytoday almost 90 percentof recentschool cohorts the thiscriterion. On theother hand,ifskilledmeans completed college education, satisfy Since college educationis rarein number of skilledpeople would dropbelow 30 percent. with some of thecountries and timeperiodsthatI am interested in,I chose a compromise in 75 a was chosen to be 0.22, whichresults higher weighton highschool. The parameter of thepopulationin thegrowth regimeto be skilled. percent are identicalto the growth I now turnto the Malthusianregime. Most parameters the we need to calibrate agriculturaltechnologyand the child labor regime only I and can be chosenarbitrarily. children is a scale parameter The fixed cost for p parameter. in the the rate of chose p = 0.001. The parameter agricultural progress technological yF, ratewas in theMalthusian determines sector, regime.In Britainthetotalfertility fertility about 4.0 in 1700, and values in other European countrieswere similar.The value ratein theMalthusian fertility regime.Notice thattargeting yF = 1.32 yieldsthisfertility in themodel, from childmortality leads to a counterfactual while abstracting prediction: in is veryhighin theMalthusian growth phase, whileactualpopulation population growth Unless child revolution. was rather low beforethe onsetof the industrial mostcountries and population in themodel,itis notpossibleto matchbothfertility is integrated mortality themain be would While such an extension in the possible, phase. growth pre-industrial matches thecalibration nottheMalthusianphase. Therefore, focushere is thetransition, had alreadyincreased at theonsetof fertility conditions decline,whenpopulationgrowth is high and income per capita is are thatfertility Our key requirements substantially. Also noticethatwhile thecalibrated transition. the onset of the constant before roughly to takeplace: is notnecessaryforthetransition thisfeature value foryFis smallerthanyl9 is all that is needed. 0 > y7 evidenceon is onlylimited The childlaborparameter ^v is hardto calibratesince there revolution. the extentof child labor beforethe industrial However,4>vhas to be chosen does not growth regime.If <^ is small,population sufficiently largeto allow a Malthusian catch up with technologicalprogress,and per capita outputgrows alreadybeforethe is too large,thenumber On theother of theindustrial introduction hand,if <$>v technology. is theindustrial to zero after of skilledagentsconverges My choice technology introduced. bothwithMalthusianstagnation of <j)u= 0.07 is in the middlegroundand is consistent forlaborsupply, of skilledagents.Adjusting and a growth regimewitha positivefraction The behaviorof the are equivalentto one unskilledworker. thisimpliesthatninechildren of to thechoice of 0^; by making<j>u modelis sensitive large,we can increasetheeffects to ascertainthat4>uis notunrealistically childlabor.It is therefore important eliminating of the parent,the key question is how much of household large. From the perspective

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITYDECLINE DURINGTHE TRANSITION TO GROWTH ACCOUNTING

365

ina sample income is contributed Horrell andHumphries that of (1995) find bychildren. around contributed between 27 and32 families inEngland 1,800children working-class in themodelincome ofhousehold whereas from childlabormakesup income, percent in families with 25 of income children. only percent working I chose6S = 0.1 andQy= 0.5. With oftheagricultural Fortheparameters technology, these the share ofoutput toland rents is 40 percent, andskilled adults make choices, going intheMalthusian A lower 5 percent ofthe bound ontheland population regime. upabout can be derived from theEnglish National Accounts share for1688in Deane and Cole time rents madeup 27 percent ofnational income. Sincesomeofthe34 (1969).Atthat income tonon-rent arealso derived ofnational income andprofits going percent property is from 30 a reasonable lower bound for the share of land. On the other land, percent hand, contracts allocated 50 of to the land owner. Since share-cropping typically percent output not all ofnational income wasderived from 50 percent is then anupper bound agriculture, thelandshare. between these for bounds. Mychoiceliesin themiddle 6. Computational Experiments calibrated Thissection usesthe model todetermine whether canhavelarge policy changes the I simulate the under effects transition. model three different on during assumptions to thecases of Brazil,KoreaandEngland roughly government policies, corresponding I start inmore mentioned earlier. these detail. bydescribing policies 6.1. ActualPoliciesand ModelPolicies KoreaandBrazilarepolarcases in terms ofeducational andchildlabor Korea policies. instituted a system offree, education from the after compulsory agesof6-12 in 1949, right After theKorean warthegovernment instituted a Compulsory Education independence. enrollment ratereached 96 percent. Plan,and by 1959 theprimary Accomplishment enforced education served as aneffective constraint onchild in labor Strictly compulsory additionto directchild labor restrictions. to the International Labor According ofthechildren from zeroto 15 years were (ILO), in 1960only1.1 percent Organization and 0.3 of the 1985 children between 10 and 14 active, by economically only percent ages in thelabor market. participated In Brazil, ofthelawfree andcompulsory education wasintroduced bytheletter already in 1930.In practice, was notavailable in many rural however, areas, primary schooling andtothe schools areoften ofpoor In 1965theprimary present daytheavailable quality. enrollment rate wasstill below50 percent inrural areas. Theneglect ofprimary education in Brazilcan also be gaugedfrom education finances. In 1960,less than10 percent of on education was directed to the while the sector, public spending primary corresponding was about 70 percent for Korea.Childlabor in Brazil is lessrestrictive figure regulation in Korea.WhileKoreasigned than an ILO convention that rulesoutchildlaborunder theage of 14,Brazildidnot. Theminimum inBrazil is now12,and age for employment eventhis limit is notalways In 1985 18.7percent enforced. ofthechildren between ages

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

366

MATTHIAS DOEPKE

10 and 14 participated in thelabormarket. Child laboris even moreprevalent amongmale 14 with 25.3 listed as children, percent economicallyactive in 1987. In England,educationwas notwidespread at thebeginning oftheIndustrial Revolution. In 1780, only about 50 percentof brides and grooms were able to sign theirname. Education expanded only slowly throughout most of the nineteenth Public century. but therewere largeregional educationalso started to expand in thenineteenth century, differences and no universal access to affordableeducation. The situationchanged The ForsterEducation Act of 1870 drasticallywith the Victorianeducationreforms. placed primaryeducation under public control.In 1880 compulsoryschooling was and starting educationwas free.While schoolingqualityis in 1891 primary introduced, were successful.While in 1880 hardto measure,literacy data suggeststhatthe reforms about 15 percent of groomsand 20 percent of brideswereunableto signtheir name,these decreasedto less than2 percent numbers by 1910. Actsof 1833. Onlya The first childlaborrestrictions wereputin place withtheFactory small set of industries was affected, however,and Nardinelli(1980) concludesthatthe Acts wereamendedin 1844 and 1874,when impacton childlaborwas small.The Factory became theminimum for child laborers was raised to 10. At thattimetherestrictions age also insteadof being limitedto certainindustries. laws universal, Compulsory schooling had an effect late in on childlabor.As a result, theincidenceofchildlaborwas decreasing 10 to 14 reacheda peak 1861, thenineteenth ratesfor children century. Activity aged from when29 percent in thatage groupwereeconomicallyactive.In 1871, the of all children but thenit fellto about 20 percentin 1881 and 1891, 17 number was stillat 26 percent, in 1901, and 14 percent in 1911. percent at thestart of a strong In summary, we see that Korea instituted publiceducationsystem outlawedearlyon as well. In contrast, theeconomictakeoff. Child labor was effectively bothin terms Brazil laggedbehindmostother at a similarstageof development countries of the of educationaland child laborpolicies. Englandis a mixedcase. At thebeginning in theareas of public intervention industrial therewas hardly revolution, any government theend of was a drasticpolicychangetowards educationand childlabor.However,there theEnglishpolicies resembletheKorean case. thenineteenth and afterwards century, The different government policies observedin the real world are given a simplified in the In the stylizedBrazilian policy,thereis no policy reform. model. representation the In thestylized Koreanpolicy, and child laboris unrestricted. Parents for education, pay to leave the starts the Once out with the same starts economy economy policies. Malthusianregime,however,thereis an immediateswitchto the policies thatwere the thegrowth basis forcalibrating pays for50 regime.Underthesepolicies,thegovernment

In both countries outcomes. in educational 14 The different policiesin Braziland Koreaare also reflected in Korea.In 1960,the is muchfaster overtime, buttheincrease increase average yearsof schooling and a year(see Barro between Koreaand Brazilwas less than in average of schooling difference years Koreans on average so that than four to more increased Lee 1993,2001). By 1985,thedifference years, successis given of educational measure Another as Brazilians. than twice as much havemore schooling of to 29 percent wereilliterate, Brazilians In 1960,39 percent ofadult rates. compared illiteracy byadult are almost whileKoreans is stillat 16.7percent, In 1995,in Brazililliteracy adultKoreans. completely ofonly2 percent. with an illiteracy rate literate

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITY DECLINE DURINGTHE TRANSITIONTO GROWTH ACCOUNTING

367

inthe skilled andchild labor is ruled out.While ofthe education costfor children, percent once the industrial is takesplace immediately Koreanpolicythereform technology in the same reforms are carried out with a of the introduced, stylized English delay policy after thebeginning ofthetransition. The policies usedin the three (or 75 years) periods arenot intended tobe exact real-world but tobe stylized simulations copiesofthe policies, between thepolicy for versions that thedifference Brazil, regimes. emphasize Especially in themodelis not "fair" formorerecent for thepolicydescription years:support in thelastdecades(andfertility hasbeenfalling education has increased faster). The initialvalues foragricultural All economies start from identical conditions. of skilledand unskilled adultsare chosento start the and the number productivity in intheMalthusian state. The is chosen such that economies industry steady productivity twoperiods after thestart ofthesimulation. from thedifferent thetransition starts Apart theBrazilian ofpolicy reforms under immediate reform under (no reform timing policy, all simulations theKorean reform under the areidentical. policy, delayed English policy), and Growth 6.2. Effects on Fertility can be summarized as follows. of policy The outcomes of thesimulations The timing the reforms hasa major on the and of transition. Without a impact timing speed fertility in fertility thedecrease thetransition to growth is small.Whena reform, during policy it causes an immediate reform is introduced, and the transition to dropin fertility, within twogenerations after thereform. In addition, is completed replacement fertility increases in GDP percapitasincethe theopportunity costofeducation growth lowering inthe workers increases. Out of the fraction ofskilled twopolicies considered, population while childlabor restrictions account for thebulkoftheresults, theeffects ofeducation The samepolicy reforms subsidies arecomparatively small. also havea major on impact distribution. As longas nopolicies areintroduced, theevolution oftheincome inequality rises the since theskill reform lowers transition, during premium goesup.Oncethe policy theopportunity costofeducation, thefraction ofskilled workers increases over however, which theskillpremium andoverall lowers time, inequality. 2 shows thesimulated for GDP percapitaandthetotal rate under Figure path fertility In the for the Korean and the Brazilian thedifferent English policies. graphs policies, casetofacilitate outcomes areincluded dashed as the benchmark (with lines) comparisons. Thetime axisis labeled such that eachmodel to25 years, andthestart period corresponds of thetransition to thegrowth is placedin theyear1950fortheBrazilian and regime andin 1800for Korean theEnglish GDP percapitais initially constant, policies, policy. intheMalthusian where sincethesimulation starts Oncethe regime wagesareconstant. industrial is introduced, incomes start to grow. technology thestart The totalfertility rateincreases before of thetransition, and then slightly in all simulations. The speedand timing declines of fertility varies decline, however, thesimulations. With decreases across no reform, to a total fertility onlyslowly fertility rate ofabout 3. With an immediate thetotal rate below3 already at reform, fertility drops andreplacement inlessthan thestart ofthe is reached twogenerations transition, fertility after thetake-off. The maximum difference in fertility acrosstheBrazilian and Korean

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

368

MATTHIAS DOEPKE

8f

" - "Brazi Noreform 1 ' '

- "Brazil" Noreform ' '

1900

0'

""r

1950

2000

'

2050

1900

1950

2000

2050

- "Korea" Immediate reform

Immediate reform"Korea"
'^

I4
1900

o'

- *

^<-'
2000

1950

2050

^3 nJ ^!
\ ~~ ^ "^

1900

1950

2000

2050

- "England" reform Delayed

- "England" reform Delayed

1750

o'

""^

1800

1850

1900

1750

1800

'

1850

1900

2. Outcome as a benchmark). with different linesareBrazilian outcomes (dotted Figure policies

simulationin any year is about 1.5, or about 60 percentof the maximumdifference betweenKoreanand Braziliandata in Figure1. Thus,whilethemodeldoes notaccountfor Withthedelayed all of the observeddifferential, theeffect of thepolicies is substantial. after thebeginning of declinestarts of theEnglishsimulation, policyreform fertility right are carried out (1875 in the the transition as well, but speeds up afterthe reforms themodelwitha delayededucation Relativeto theactualEnglishexperience, simulation). decline aroundtheend of the reform does a good job in accounting fortherapidfertility 19thcentury. The main deviationbetweenmodel and data is thatthe model does not theinitialincreasein fertility between1700 and 1820 in England,as well as the reproduce level of about 6.0 in Korea and Brazil. Across all policy options,the peak fertility is 4.8. maximumfertility level reachedin the simulations affect thespeedof thesame policies that Since fertility and educationdecisionsinteract, rateis in themodel. The highest decline also influence economicgrowth growth fertility Underthe Korean educationreform. achieved underthe Korean policy of an immediate rateof GDP per capita between1950 and 2000 is 4.2 percent, policy,theaveragegrowth

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITYDECLINE DURINGTHE TRANSITIONTO GROWTH ACCOUNTING

369

which toa growth rate ofonly 3.3percent under the Brazilian Inthe data, Policy. compares a fast we also observe that as wellas other countries with Korea, transition, demographic in GDP percapitathan faster slow-transition countries likeBrazil. experienced growth in to the Penn World the rate GDP between Tables, average growth According percapita in Korea,and only2.8 percent in Brazil.Even if we 1960 and 1990 was 6.7 percent excludes consider theshorter from 1960to 1980(which theLatinAmerican debt period inthe1980s), was6.2 percent inKoreaversus crisis 4.4 percent inBrazil. average growth in thedatais larger rate differential than thedifferential inthemodel, which Thegrowth andeducation decisions arenottheonly that socialpolicies factor the suggests explaining relative ofBrazilandKorea. Given that the twocountries aredifferent performance along a number itwould infact be very ifthepolicies ofdimensions, considered here surprising fordifferences in economic werethemainexplanation the Nevertheless, performance. results that the and their effects on and education decisions are an fertility suggest policies oftheoverall explanation. important part inthe Thereason for the differential model is that with education subsidies growth-rate theskilled fraction of thepopulation andchildlaborrestrictions, increases which fast, theindustrial raises since is skill-intensive.15 Under theKorean technology output policy, inthepopulation increases in 1950 toalmost thefraction ofskilled adults from 5 percent in2000.Under the Brazilian the increase is from 70 percent 5 toonly about policy, percent In thedata,we observe in differences skill between Korea and 33 percent. large average ofthepopulation Brazilas well.In Korea, ofschooling average years aged25 andover increased from 3.2 in 1960to7.8 in 1985.In Brazil, theincrease was from 2.6 in 1960to of thethree channels which thespeedof fertility only3.5 in 1985.Noticethat through affects channel is present in themodel. decline GDP percapitaonlythehuman-capital Thischannel that becauseofthequantity-quality increase the tradeoff, captures parents whenfertility falls.The modelabstracts education of their children from thecapitalthecapital dilution channel increases stock andthe (lower population growth perworker) cohort-size channel decline lowers thedependency If we extended ratio). (rapid fertility the model tocapture these effects as well, the differential would be even predicted growth larger. 6.3. Effects on theIncomeDistribution and output, the income Apartfrom influencing fertility policychangesalso affect distribution. Thegraphs sideofFigure 3 shows on theleft-hand income Ginicoefficients In themodel, for thesimulations. is determined in which inequality bytheskill premium, turn on the relative of skilled and unskilled labor. The Korean depends supply policy increases therelative number of skilled which offsets theincreased demand for adults, inthe skilled labor industrial sector. inthe simulation with animmediate Therefore, policy reform is generally lowanddecreases thetransition. Theinitial slightly inequality during
is exogenous, 15 Because productivity the effect on growth ratesis transitory, but thereis a growth leveleffect as longas policies differ. permanent

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

370

MATTHIAS DOEPKE

- "Brazi 1" Noreform


0.6 ^^

' 4" 3

- "Brazil" Noreform
* "*-

5
ol 1900

/
, 1950 . 2000 1 2050

2
o> 1900 1950 2000

j
- - Unskilled

1 2050

- "Korea" Immediate reform 0.6 [ 4 *50.4


\s,.

- "Korea" reform Immediate Xx ^ . L... :...^^


1950
"

2
1950
.

vL

** - ..

1900

Ol

2000

2050

1900

0'

2000

- - Unskilled 1

2050

- "England" reform Delayed

- "England" reform Delayed

0.6 [ i
S 0.4

I /
1800
.

4f" 2
*

~v> -
1800
.

""""""X

\
1850
.

1850
- - Unskilled J

ol

1750

1900

1750

ol

1900

and fertility differentials. Figure 3. Gini coefficients

for education inthe sincethe demand increase ofthe reform occurs Giniatthetime policy areskilled. teachers increases. Thistemporarily raises theskill because suddenly premium of the In theBrazilian adultscontinue to makeup themajority simulation unskilled translates intoan increased The increased demand forskillin theindustry population. a delayed With thetransition. so that increases policy during wagepremium, inequality transition. the first then decreases reform, during inequality increases, thepronounced Thusonlywith theEnglish inverse-U-shaped policydo we observe Inthe realworld, Kuznets curve. between and known as the income relationship inequality that inthe countries wasobserved this the ofdevelopment ofinequality over course pattern andtheUnited industrialized first States), 1955,lookedat England, (Kuznets, Germany, is mixed. theevidence butforcountries that started to growmore Specifically, recently was generally In Koreainequality curve. neither KoreanorBrazilexhibit a Kuznets low, Between 1955and1985, andthere intheincome distribution overtime. was little change with wasalways the varies 0.3 and0.4.InBrazil, Ginicoefficient between high, inequality

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITYDECLINE DURINGTHE TRANSITION TO GROWTH ACCXDUNTING

37 1

a Gini between0.5 and 0.6 (see Deiningerand Squire, 1996). The incomesharegoingto oftheKoreanpopulation is aboutthesame as thesharegoingto the thepoorest 20 percent in Brazil. Even that from some features 40 considering themodelabstracts poorest percent the resultsdo show that which are relevantforexplainingthe income distribution,16 interactions ofpublicpolicieswithfertility decisionscan have largeeffects on thegrowthThis mayprovidean explanation whyKuznetscurvesshow up in relationship. inequality butnotin others. some countries, A moredirecttestof thedistributional of themodel would look at wage implications inequality, especially the premiumfor skilled labor, since the skill premiumis the in themodel. The skillpremium reactsto the forcebehindchangesin inequality driving effect education and the of decisions on the fertility policies through general-equilibrium effect is relativesupplyof skilledand unskilledlabor. Clearly,thisgeneralequilibrium falls fastin the Korean simulation centralto the results:fertility preciselybecause the whichmeans thateducationis affordable even for policies delivera low skill premium, unskilledparents.Park et al. (1996) compare wage inequalityin Brazil and Korea in withthepredictions of themodel,theyfind muchhigher skill 1976 and 1986. Consistent premia in Brazil than in Korea. Moreover, there is a substantialdecline in wage in Korea between1976 and 1986. Specifically, Parket al. estimatea standard inequality on educationand work that the of is is, earnings equation, logarithm earnings regressed In school in Brazil earned a 60 percentpremium 1976, high graduates experience. was only35 percent relative to workers who finished school,whilethepremium primary in Korea. For university the premium(relativeto primary school) was 201 graduates, in Brazil versus 121 percentin Korea. In 1986, the differences were even more percent In Brazil,thepremiaforhighschool and university graduateswere still55 pronounced. whilein Korea thepremiahad fallentojust 10 percent and and 188 percent, respectively, 76 percent. in fertility In themodel,changesin overallfertility are drivenby differences between The graphs on theright-hand side ofFigure3 breakdownthe skilledand unskilled parents. In theBraziliansimulation ratein each simulation totalfertility by thetypeof theparent. Unskilledchildren thereare no educationsubsidiesor child labor restrictions. are cheap and therefore continue relative to skilledchildren, unskilled to have manychildren parents thetransition. increasesforbothgroupsof parents, Initially, throughout average fertility and theaggregate totalfertility ratefalls onlybecause skilledadultsbegin to make up a of thepopulation. The fertility differential betweenskilled and unskilled largerfraction differential betweenthe two parentsstayslarge. Underthe Korean policy,the fertility after declinesrapidly thepolicychange.The abolitionof child laborand typesof parents educationsubsidies implythatskilled childrenare only slightly more expensive than unskilledchildren,which resultsin a small fertility differential between skilled and

16 Forthecomparison of BrazilandKorea,theroleof landownership is essential. After a landreform at in Korea,butalwayshighly theendofJapanese landownership was dispersed concentrated occupation, in Brazil.Indeed, ifconcentrated landownership is introduced, themodelgenerates with high inequality almost no time under theBrazilian trend policy.

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

372

MATTHIASDOEPKE

- Labor Child restriction only

- Labor restriction Child only

1900

0'

1950

~r^

2000

2050

1900

1950

2000

2050

Education subsidy only

Education subsidy only

ei I' /^ M ^y i \
1900 1950 2000 2050 1900
4. Outcome with isolated Figure policies.

1950

2000

2050

after thenew policies unskilled The same effect occursin theEnglishsimulation parents. are introduced. In the data, in line withthe predictions, the available evidence shows small fertility in Brazil. According to United Nations in Korea and large differentials differentials rateof 6.7, educationhad a totalfertility formal (1995), in 1986 Brazilianwomenwithout while for women with seven or more years of educationthe numberis 2.4. Thus the Alam and Casterline children. differential amounts to morethanfour (1984) report fertility educationwas 5.7, formal thatin 1974 the totalfertility rateforKorean women without whileforwomenwithsevenormoreyearsofeducationtherateis 3.4. This givesa fertility in Brazil.17England,as well as other of 2.3, roughly half of the differential differential withinthe populationtoday in has small differentials industrialized countries, fertility countries. to comparison developing decline on fertility In summary, we see thatthepolicies consideredhave large effects for into further translate to growth. These differences implications duringthe transition and the income GDP per capita, average education levels, fertility differentials, distribution which line up well withthe evidence forour examples Brazil and Korea. of two policies, an to a combination So farthe analysis of public policy was restricted is whichofthetwo to ask A natural educationsubsidyand childlaborrestrictions. question It turns out thatthe described above. the results is for more important generating policies child labor restriction has a much biggerimpact than the educationsubsidy.Figure 4

fallin theoverall thesameyear. there is no datafor BrazilandKoreafor 17 Unfortunately, However, given in 1986than in Koreawas evenlower that thefertility differential it seemslikely by education fertility, in 1974.

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

TO GROWTH ACCOUNTING FOR FERTILITYDECLINE DURINGTHE TRANSITION

373

ifeither education subsidiesorchildlabor showsoutcomesfor GDP percapitaand fertility is introduced, the restrictions are introduced in isolation.If onlya child labor restriction is similarto the Korean simulation, which evolutionof income per capita and fertility there is an unambiguous increasein introduces bothpolicies.Witha childlaborrestriction even faster whichreduces fertility the cost of unskilledchildren, comparedto the case when thereis only an educationsubsidy, whenbothpolicies are combined.In contrast, and thesubsequent declineis muchsmaller thanin the actuallyincreasesinitially, fertility The reasonis thattheeducationsubsidyworksin two different Korean simulation. ways. Since skilled children become cheaper,more unskilledparentsdecide to educate their which lowers fertility. At the same time,however,fertility increasesfor the children, children even without thesubsidy. skilledparents who wouldhave decidedto educatetheir are partially and the overall is small. These results The two effects offsetting, impact of including the cost of the children'stimein the opportunity underline the importance In fact, ifwe did notaccountforthecostofthechildren's cost ofeducation. it would time, thatin Englandfertility fell fastest be rather just at the timewhenfreepublic surprising educationwas introduced. A welfare by thefactthatskilledand unskilled analysisof thepolicies is made difficult have interests. Therefore the policies cannotbe rankedusingthePareto people opposing criterion. The educationsubsidy, however,is beneficialforadults of both skills in the Futuregenerations of skilledpeople have lower period when the subsidyis introduced. A child labor restriction withthe policy because of a lower skill premium. alone utility hurtsboth skills in the period when it is introduced, since the unskilledparentslose of skilledchildren falls. However,future of unskilled income,and theutility generations A combination from child laborrestrictions. of thetwo policies yieldsthe people benefit skilledgenerations are stillbetter off highest growth pathforGDP per capita,but future without thepolicies.

7.

Conclusions

This paperdevelopsa theory that is consistent witha phase of stagnation whichthe during exhibits Malthusian followed a transition a to features, economy by growth regime.The economictransition to growth is accompaniedby a demographic transition from highto The model is used to assess whether low fertility. child labor restrictions and education subsidiescan have largeimpactson thefertility declinethataccompaniesthetransition to growth. The possibility ofchildlaborimpliesthat thevalue of a child's timebecomespartofthe cost ofeducation.18 Without forthevalue of thechildren's it time, opportunity accounting would be verypuzzlingthatin manycountries fell fastest fertility just at thetimewhen freepublic educationwas introduced. When thereis no alternative use fora child's time,
18 Theroleofchildlabor in determining has beenanalyzed andEvenson (1977). In fertility byRosenzweig linewith theresults find that theeconomic contributions ofchildren child laborarean here, they through determinant for rates. important fertility

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

374

MATTHIASDOEPKE

In shouldlead to higher freeeducationlowersthecost of children and therefore fertility. education we the cost of once account for full contrast, includingpotential opportunity childlaborincome,educating a childis costlyeven if schools are free.A steepdeclinein is therefore after just whatwe shouldexpectto see. fertility right majoreducationreforms on thefertility thateducationand childlaborpolicies have largeeffects Indeed,we find transition. The policies also determine the evolution of income per capita, average to growth. When thetransition and theincomedistribution of a country education, during of educationsubsidiesare relatively thetwo policies, theeffects small,while comparing forchild laborregulations is crucial. accounting the forunderstanding forpolicychangesis important The results show thataccounting This from to the transition of different countries stagnation growth. during experience role as well. For does not rule out thatotherfactorscould play an important finding rates,even thoughdeclining example, the model abstractsfromchanges in mortality as declining is an equally important mortality aspect of the demographictransition can be As a simple extension to the model, changes in child mortality fertility. In the calibrated until adulthood. a of survival incorporated by introducing probability the numberof birthsincreases,but the model, for small increases in child mortality numberof surviving childrendeclines. Feeding realisticsurvivalrates into the model of policy numbers moreclosely,buttheeffects makes it possible to matchactual fertility In order childmortality.19 of changing theeffects changesare stillthesame and dominate to allow formore theframework wouldhave to be extended to incorporate adultmortality, thantwo periodsof life. in the Another policies in themodel are implemented open questionis how thestylized which are subjectto an obviousenforcement real world.Child laborregulations problem, is wheredirect muchchildlabortakesplace in agriculture, is severegiventhat supervision do littleto child labor restrictions difficult. A number of studieshave foundthatofficial constrain child labor. In practice,compulsory schoolinglaws oftenappear to be a more Froman administrative effective constraint on childlaborthandirect pointof regulation.20 show up in a givenclassroomat a certain time,as view,it is easier to check thatchildren thattheydo notengagein illegal workat any timeand anyplace. opposed to ascertaining in the model, theyare best even Therefore, thoughthe policies can be used separately in a of as thought joint policy practice. are taken A questionbeyondthescope ofthispaperis how thegovernment policies that to understand Two recent determined. as exogenoushereare ultimately paperswhichtry are Acemogluand Robinson(2001), wherea rich thepoliticaleconomyof such reforms and Galor and Moav (2003), of revolution, reforms to reduce a threat elite introduces in order to reap human capital where the elite creates social institutions voluntarily reflects In the theory of Galor and Moav, the timingof educationreforms externalities.

on effects rateshave only minor 19 Within the dynastic framework, changesin childmortality utility see Doepke choiceare takenintoaccount, and sequential survival even if stochastic fertility fertility (2004). are especially childlaborrestrictions that 20 Forexample, (1996) find usingUS data,Margoand Finegan laws. ifcombined with effective schooling compulsory

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

ACCOUNTING FOR FERTILITYDECLINE DURINGTHE TRANSITIONTO GROWTH

375

inthedemand for skill. From this theexpansion ofeducation inthe changes perspective, latenineteenth in Britain rolefor skilled century mayhavebeencausedbyan increased inproduction labor inthesameperiod. that (2004) argue Similarly, DoepkeandZilibotti after anincreased skill demand for ledgrowing number ofparents tosend their children to will those the introduction of child labor since no school, laws, parents support they longer on childlaborincome. of on theevolution Thus,whilethispaperconcentrates depend for for future there work tobe somescopefor fertility given government policies, appears unified theories thatexplainchanging and government policies alongsidefertility growth.21 In thediscussion ofthetransition from to growth I concentrated on crossstagnation differences in A decline. different is to the country fertility challenge explain why take-off occurs at different timesin different countries. the Clearly, government policies considered herecannot tookofffirst, economic explain whyEurope giventhat growth started theintroduction ofbroad education andchildlabor When longbefore legislation. across the havethepotential tocausea delayin countries, comparing developing policies thetake-off of about10 to 20 years, whichis stilltoo little to explaintheobserved In order differences. tomake this itwillbe necessary tomove line, progress along beyond theassumption ofexogenous andintroduce a theory oftechnological productivity growth The results in thispaperdo notdepend in anyway on theassumption that progress. is As a first it is productivity growth exogenous. beyond step exogenous growth, possible tolink therate oftechnological tothenumber ofskilled progress peopleintheeconomy. Sucha model canexplain the rate ofeconomic increased the industrial why growth during instead of jumping to thegrowth-regime level right of revolution, away.The effects on growth would be amplified. government policies

A. Definition ofan Equilibrium Appendix This section introduces the remaining with the conditions, equilibrium starting determination of laborsupply. Skilledadultsdistribute their timebetween working, and teaching their own children, and teaching children of unskilled raising parents. Therefore thetotal ofskilled labor supply by: Ls is given + 05) is^sW s(S,x) - <j> **(*) =[!-(* Xs^u(x) nvfrx)]Ns -tslu^sWnsiU^Nu. (16) Unskilled labor adults andbychildren whodo not byunskilled Lv is supplied go toschool:
21 Concerning thecontrast of Koreaand Brazil,an additional factor thatis likely to be important is the initial distribution of wealth. Thisshould be especially relevant for suchas theonesconsidered policies, that havea largeimpact on theevolution of inequality in a country. Brazilalways had here, Historically, a high concentration ofwealth in general andoflandownership in particular, whereas Koreacarried out a landreform after that leveledthedistribution of wealth. See Galor, Moav,and Vollrath independence somework on theroleoflandinequality in development. (2003) for

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

376

MATTHIASDOEPKE

+ <t>u to fo-c/to to Mtf>*) MM *s+ *(/-(/ ty/].

(17)

Inequilibrium, I assume labor hastoequallabor demand for eachtype oflabor. that supply skilled adults canperform both theskilled andtheunskilled while unskilled adults work, cando unskilled fallbelowthe work Under this theskilled wagecannot only. assumption, unskilled wage: "sM > Hfc(*). are: The market-clearing conditions for thelabor market = if ws(x)> w^x), LFS(x)+ LIS(x) < Ls(x), LFU(x)+Llu(x)=Lu(x) + [Ls(x)-LFS{x)-Lls(x)]. (19) (20) (18)

The flat taxt which has to be chosen to observe finances theeducation budget subsidy thetotalexpenditure on schooling balance.The correct tax rateis givenby dividing subsidies bytotal wageincome:

TW

+ <l>sN'sWs(xr Ls(x)ws(x)+Lu(x)wu(x)

[ >

that for thecomputation oftotal where number ofskilled children. Notice N'sis thetotal tothewageoftheusualworkers. labor income we havetoaddtheincome oftheteachers in themodel adults Teachers receive and are taxedlikeall other work wagesfortheir economy. SinceI abstract for The final condition is thelaw of motion population. equilibrium is givenby the from childmortality, thenumber of adultsof a giventypetomorrow number ofchildren ofthat today: type **s= ^sto ns(S,x)Ns + A^jc) ns{U,x)Nv, = K ^s- i/toi/(M Ns + kv^v(x) nv(U9x)Nv. (22) (23)

an equilibrium. areneeded todefine We nowhaveall theingredients athandthat 1 (Recursive Competitive Definition policy Equilibrium) Given a government tax value a a recursive consists t, function functions of competitive equilibrium {<t>uid}> labordemand laborsupply functions functions LIS,and LFS,LFU, Vs and VUt Ls andLUf mobility functions ky^s* and ^u^>u>a^ As_>s, A.s_>u, Liu, wagefunctions ws and wUf X into functions mapping nsandnvmapping {5, U} x X into R+, anda lawof R+,policy motion X into such that: g mapping itself, i. Thevalue the fori e {5, U}: equation functional functions satisfy following = Vk(x) ma* {c" + P(ns+ ) -[s^(^) + t/^(x')]}. (24)

= g(x). xf tothebudget constraint (9) and thelawofmotion subject

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

ACCOUNTING FOR FERTILITY DECLINE DURINGTHE TRANSITIONTO GROWTH

377

in (24). ii. For i,j e {5, U}, ifAj_>j(x) themaximum > 0, tij(i,x)attains ill. Thetax r satisfies thegovernment constraint (21) function budget iv. ThewageswsandWy and labordemand (5)-(8) and LFS, LFU, LIS,andLw satisfy (18). v. Laborsupply (16) and(17). Ls andLv satisfies vi. Laborsupply (19) and LFS,LFU, LIS,andLw satisfy Ls andLv and labordemand (20). and ^u^u satisfy vii. Themobility (10). functions A.s_>s, A.s_>Ut ly _>s, variable x is given viii. Thelaw ofmotion by(4), (22), and (23). gforthestate a market-clearing constraint for thegoods The equilibrium conditions do notinclude In Walras' Law. condition because it holds market, (ii) above,it is automatically by one children. In understood that choose of other that words, only parents type saying = = in the maximum means that and the attains (24) ns(S,x) {ns ns(S,x)1nu 0} constraint maximize Maximization is c that results from thebudget utility. consumption if a positive in question. number of parents choosethetypeof children onlyrequired constraint. that Condition Condition (iv) requires (iii) is thegovernment budget wages andthat theskilled products wagedoesnotfallbelowtheunskilled wage, equalmarginal to population andeducation condition condition time, (vi) is the (v) linkslaborsupply thelabor condition that for eachtype condition for market, (vii)requires market-clearing and unskilled children sum to and condition the fractions skilled ofadult one, (viii) having defines thelaw ofmotion.

B. Proofs forall Propositions Appendix intheindustrial ofa firm sector 1. Theprofit-maximization Proof ofProposition problem is given by:

"a - ws(x) maxU (ls)1 (/)" ls - wv(x) /}.


The first-order condition for a maximum with to lv gives: respect

(25)

- {ls} = - IT*
thisexpression backinto(25) yields a formulation of theprofit maximization Plugging as a function ofunskilled labor problem only:

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

378

MATTHIAS DOEPKE

max h_ -Wu(x) j^M ^^y1'1"'*/,, /f,J.

(26)

Since thisexpressionis linearin lUy in the industrial sectorwill be profitable production only if we have:

whichcan be rearranged to get:

whichis (11). The first-order conditions fora maximum oftheprofit-maximization necessary problem of a firm in agriculture are given by the wage conditions (5) and (6):

^-Awa-.z'-'-'--

(27)

are also sufficient fora Since theobjectivefunction is concave, thefirst-order conditions maximum.It is therefore sufficient to show thatforany ws(x),wu(x) > 0 we can find values for skilled and unskilledlabor supplyLFS and LFU such that(27) and (28) are satisfied. The requiredvalues are givenby:

and:

Lf5=(atv UfeJ
A

\O/l-0s-0</)

\(l-*s/l-*s-*tf)

Uk)

/ Q

\(*i//l-*s-W

/ A

/ q \(l~6u/l~9s~0u) / q \(os/\-es-eu) \(l/l-es-0u)

whichare positiveforany positivewages ws(x) and wu(x). maximization Proof of Proposition2. We considerthefollowing problem:

- Ef + 0 "(///>s + (1 -f)Vv/Pu]}. + (1 -f)/PurVVs/Ps >max<o{(w


the parameters )S, a and e are all strictly biggerthanzero and strictly By assumption, smallerthanone. It is also assumed w.l.o.g. that0 < pv < ps and Vs, Vv > 0 hold. We wantto show thatthereare no interior solutionsin/, thatis, if thereis a solutionto the zero or one. maximization /is either problemabove, the optimal the assumethat we have alreadydetermined To showthat there are no interior solutions, optimal E. Given this E and the fact that the functionto be maximized is twice

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITYDECLINE DURINGTHE TRANSITION TO GROWTH ACCOUNTING

379

werean interior theoptimal differentiable solution, /would have in/,ifthere continuously fora maximum. I solve fortheunique/which to satisfy firstand second-order conditions and show thatthis/ does not satisfythe second-order solves the first-order condition, condition. This provesthatthereare onlycomer solutions. of U withrespectto/ I will namethemaximand and secondderivatives U( ). The first are:

8/

.-.L.(I_V/+!V)~Y/!k+iWEi) Pu J \Ps \Ps Pu J [ \Ps PuJ


\Ps Pu J \Ps PuJl

9/2

\Ps PuJ \Ps Pu J

[(1+^i_v/+iz/yY^+(w)^)_2^_M'. Pu J \Pu PsJ_ \Ps PuJ\Ps Pu J \Ps [


is positive.For In thefirst for term within theouterbrackets 0 < / < 1, thefirst derivative, an interior solutionto be possible, it has to be the case thatVs/ps< Vy/py,because condition cannotbe satisfied. thesecondterm is also positiveand thefirst-order otherwise if Vs/ps > Vy/py,we are done. For thecase thatVs/ps< Vy/pv,setting the Therefore first derivative equal to zero and solvingfor/yields:
r_ \PS Pu) \PS PuJ

v1 cwuyPsPuJ\Ps Pv)
I will now plug this value for/ into the second derivativeto verifythatthe second The second derivative derivative so that thecritical is is positive, pointis nota maximum. if the holds: positive following inequality

\Ps PuJ \Ps

Pu J

\ Ps

Pu

\Ps PuJ

some algebra: Pluggingin our value for /yields after

Pu \Ps Pu\Ps Vu<Vs.

PuJ Pu I \Ps PuJ'

PuJ

\Ps PuJl

Pu \Ps PuJ

PuJ Pu\Ps

Thus if Vs > Vy thesecond-order condition fora maximum is notsatisfied, and therefore

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

380

MATTHIASDOEPKE

there is no interior there maximum. ifVs weresmallerthanV\j, wouldbe only Conversely, unskilled children forsure,since theyare cheaperto educate.Thus once again we obtaina cornersolution.Thus in everycase thereare onlycornersolutions.Families have either unskilledor skilledchildren, but theydo notmix. of theproblemin which 3. It is helpfulto considertheformulation Proofof Proposition for adultschoose the totaleducationcost E, so thatthe numberof children equals //?, adults who choose to have children of type i. The maximizationproblem in this formulation is:

O<E<W/P; K

' max l(w - Ef + p(E/Pi)1 V,l


)

(29)

This can also be written as:

+ xj(w-r ft)"^J.
_^_ = _^-

(30)

an adultis indifferent Since thecosts and utilities of children enteronly in the last term, betweenskilledand unskilledchildren if and only if:

(ft)'"

(Pa?-"

(31) ( ]

Noticethat thiscondition does notdependon thewage of theadult.Also, ifcondition (31) of whether is satisfied, adultsface thesame maximization theydecide regardless problem forunskilled This impliesthattheoptimaltotaleducationcost E does or skilledchildren. will be cost of havingskilledchildren notdependon thetypeof thechildren. The higher of children. exactlymade up by a lower number exceeds the 4. We assume ws > wU9 thatis theskilledwage strictly ProofofProposition is moreexpensive.We thenhave: unskilledwage, thetimeof skilledparents

+P +P + <i>sws + <t>sWs <j>ws < 4>Wu ' +p + <j)Ws (j)[JWup (frWu(fruWu


the ratioof the total cost of skilled child to thatan unskilledchild is lower forskilled are relatively thanforunskilledparents. In otherwords,skilledchildren cheaper parents for unskilledparents. for skilled parents,and unskilledchildrenare relatively cheaper of childrenof each type,at least Since in equilibrium theremustbe a positivenumber must and at least some unskilled some skilledparents musthave skilledchildren, parents bothtypesof parents 3 that it followsfrom have unskilled children. Moreover, Proposition betweenbothtypesofchildren can be indifferent onlyifthepriceratiobetweenskilledand the unskilledchildrenis the same. This cannotbe thecase since the wage and therefore can be indifferent. timecost differs, thusjust one typeof parent

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

FOR FERTILITY DECLINE DURINGTHE TRANSITION TO GROWTH ACCOUNTING

381

C. Noteson Computation Appendix andthegrowth themodelcan be computed Within theMalthusian via regime regime, standard valuefunction iteration ona discretized state The initial for the value space. guess is computed thenumber ofchildren tozero, so that stems functions from bysetting utility a given state the for finds a state x! in iterations, x,the only. algorithm consumption During decisions ofadults areconsistent the next such that the with state period resulting fertility law ofmotion can be usedto compute y. After theiterations theequilibrium converge, canbe computed values. thesteady state a system of bysolving steady-state Alternatively, that defines thesteady state. equations In principle it is possible to compute theentire theMalthusian model, encompassing with the and the the same method described above. regime, transition, growth regime, sincethestate vector is four-dimensional, would either take However, computations very I use a shooting be imprecise. Therefore that algorithm directly longorwould computes from valuex0 for theequilibrium thestate vector. The pathoverT periods anystarting theindustrial are started a number ofperiods before becomes computations technology ofperiods T tobe computed hastobe chosen Thenumber competitive. sufficiently large is closetothebalanced suchthat theeconomy at T.Whether this is thecase growth path canbe checked the as described abovebefore the bycomputing growth regime computing The algorithm transition. is notguaranteed to converge, butitworks wellin practice. Acknowledgments Thispaper is basedon myPh.D.dissertation at theUniversity ofChicago(2000). I am toGary Edward Robert andespecially Robert Lucas Becker, Prescott, Townsend, grateful for their andencouragement. atChicago, theSED annual guidance Workshop participants the Max PlanckInstitute forDemography, the FederalReserveBank of meetings, Boston Western Wharton, Ontario, UCLA, Minneapolis, Pennsylvania, University, Stanford, Rochester, UCL, Cambridge, IIES, Illinois, Minnesota, USC, UC Virginia, I also benefited comments. from Riverside, Duke,and Brown provided many helpful Dirk Lee Ohanian, OdedGalor, Dana Zakharova, suggestions bySylvain Dessy, Krueger, Rui Zhao, and an anonymous referee. Financialsupport by the NationalScience Foundation SES-0217051)is gratefully (grant acknowledged. References
of Political American Economic Review D., andJ.A. Robinson. Transitions," (2001). "A Theory Acemoglu, 91(4),938-963. inRecent World Alam, I., andJ.B. Casterline. (1984).Socio-Economic Differentials Fertility. Fertility Survey Studies No. 33, Voorburg, Netherlands: International Statistical Institute. Comparative R.J., andJ. W.Lee. (1993)."International ofEducational Journal Barro, Attainment," Comparisons ofMonetary Economics 32, 363-394. R. J.,andJ.W.Lee. (2001). "Schooling in a CrossSection ofCountries," Economica Barro, 68(272), Quality 465-488.

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

382

MATTHIASDOEPKE

G. S.,andR. J.Bamo. ofthe ofFertility," Economic Journal Becker, (1988).A "Reformulation Theory Quarterly 53(1), 1-25. ofEconomics G. S., K. M. Murphy, andR. Tamura. andEconomic Journal 1990."Human Becker, Growth," Fertility, Capital, 98, 12-37. ofPolitical Economy in Emerging Transitions andEconomic Miracles Bloom,D. E., andJ.G. Williamson. (1998). "Demographic BankEconomic Review Asia," World 12(3),419-455. de la Croix, andGrowth: Differential American Matters," D., andM. Doepke. (2003). "Inequality Why Fertility Economic Review 113. 93(4), 1091-1 J.C. (1992). TheDemographic Press. Transition. Oxford: Oxford Chesnais, University Income andEconomic Growth," Dahan,M., andD. Tsiddon. Transition, Distribution, (1998). "Demographic Journal Growth 3, 29-52. ofEconomic 1688-1959.Cambridge: Economic Growth Deane,P., andW. A. Cole. (1969). British Cambridge University Press. Income World BankEconomic K., andL. Squire. (1996). "A New Data SetMeasuring Deininger, Inequality," Review 10(3),565-591. Model Fit theFacts?" and Fertility Decline:Does theBarro-Becker Doepke,M. (2004). "Child Mortality Journal Economics ofPopulation (forthcoming). ofChild LaborRegulation. M.,andF. Zilibotti. (2004).TheMacroeconomics Unpublished Manuscript, Doepke, UCLA andIDES. andPopulation FernandezEconomic Growth J.(2001).WasMalthus Villaverde, Dynamics. Unpublished Right? ofPennsylvania. Manuscript, University Journal Selection andtheOrigin ofEconomic Growth," Galor, O., andO. Moav.(2002). "Natural of Quarterly 117(4),1133-1 Economics 192. ofthe Demise ofthe ClassStructure. Galor, O., andO. Moav.(2003).Das Human Unpublished Kapital: ATheory Brown andHebrew Manuscript, University University. andOvertaking inthe andthe ofDivergence Galor, O., O. Moav,andD. Voilrath. (2003).LandInequality Origin andHebrew Brown Growth Process: andEvidence. University. Manuscript, University Theory Unpublished Brown forProductivity. Galor,O., and A. Mountford. (2003). Trading Manuscript, Unpublished Population andHebrew University University. American Economic Review andGrowth," Galor, O., andD. N. Weil. 86(3), (1996). "The Gender Gap,Fertility, 375-387. From Malthusian to the andGrowth: Galor, O., andD. N. Weil.(2000). "Population, Stagnation Technology, Review Transition andBeyond," American Economic 90(4), 806-828. Demographic Review American Economic andA. Seshadri. Transition," Greenwood, J., (2002)."The US Demographic Papers andProceedings 92(2), 153-159. Economic Review toSolow,"American G. D., andE. C. Prescott. Hansen, 92(4), 1205-1217. (2002). "Malthus and A Review ofMethods Attainments: ofChildren's Haveman, R., andB. Wolfe. (1995). "The Determinants Journal Literature 33, 1829-1878. Findings," ofEconomic TheEconomic Journal andEconomic Growth," Hazan,M., andB. Berdugo. (2002). "Child Labor,Fertility, 810-828. 112(482), andthe Labor ofLittle Children: Child Horrell, S.,andJ.Humphries. (1995)."The Exploitation Family Economy in theIndustrial in Economic 32,485-516. Revolution," History Explorations editions. Various Geneva:ILO. International LaborOrganization (ILO). YearBookofLabourStatistics, overtheVery Economic Growth Inevitable? Revolution C. I. (2001). "Was an Industrial Jones, LongRun," 1. Advances inMacroeconomics 1(2),Article World in Achieved E. F. (1982). Socio-Economic Jones, Comparative Survey Fertility. Fertility Differentials Institute. Statistical Netherlands: International Studies No. 21, Voorburg, Journal Economic Growth?" DeclinePromote S. (2002). "Does theMortality Kalemli-Ozcan, ofEconomic Growth 7(4), 411-439. theMalthusian Growth andtheEscape from (2001). "Agricultural Productivity Kogel,T, andA. Prskawetz. Growth 6, 337-357. ofEconomic Trap,"Journal of DecisionsExplainUS IncomeInequality? J. (1999). Can Parental Knowles, Working paper,University Pennsylvania.

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

TO GROWTH ACCOUNTING FOR FERTILITY DECLINE DURINGTHE TRANSITION

383

with Journal Kremer, M., and D. Chen.(2002). "IncomeDistribution Dynamics Endogenous Fertility," of Economic Growth 7(3), 227-258. American Economic Growth andIncome Review S. (1955). "Economic 45, 1-28. Kuznets, Inequality," andLong-Run Journal Growth N.-P. Growth," (2003). "Gender 8(4),403-426. ofEconomic Lagerlof, Equality in theEighteenth in Floud, Lee, R. D., andR. S. Schofield. R., andD. (1981). "British Population Century," Press. since1700.Cambridge: (eds.),TheEconomic History ofBritain Cambridge University McCloskey A Long-Run A. (1991). Dynamic View.Oxford: Forcesin Capitalist Maddison, Development: Comparative Oxford Press. University andSchoolAttendance inTurn-ofR. A.,andT. A. Finegan. (1996). "Compulsory Schooling Legislation Margo, America: A 'Natural Economics Letters 53, 103-110. the-Century Experiment' Approach," andthePersistence ofPoverty," TheEconomic Journal Moav,O. (2005). "Cheap Children (forthcoming). Income andGrowth," Journal Growth O. F. (1999). "Endogenous Distribution, Morand, Fertility, ofEconomic 4(3), 331-349. C. (1980). "ChildLaborandtheFactory Acts,"Journal Nardinelli, 40(4),739-755. ofEconomic History of and theTransition to Modern Growth. Unpublished Manuscript, University Ngai,L. R. (2000). Barriers Pennsylvania. andtheInequality ofPayinBrazil and andR. H. Sabot.(1996)."Educational Park, Y.-B.,D. R. Ross, Expansion Education in Brazil.Inter-American Korea,"in Birdsall, N., andR. H. Sabot(eds.),Opportunity Forgone: Bank:Washington, D.C. Development New andR. Schoen. S. H., N. Keyfitz, Preston, (1972).CausesofDeath:LifeTables forNational Populations. York: Press. Seminar andthe Economic ofChildren in M. R.,andR. Evenson. Contribution (1977)."Fertility, Schooling, Rosenzweig,

This content downloaded by the authorized user from 192.168.72.221 on Thu, 6 Dec 2012 12:16:13 PM All use subject to JSTOR Terms and Conditions

Vous aimerez peut-être aussi