You are on page 1of 32

Fertility, Education, and Development: Evidence from India Author(s): Jean Dreze and Mamta Murthi Reviewed work(s):

Source: Population and Development Review, Vol. 27, No. 1 (Mar., 2001), pp. 33-63 Published by: Population Council Stable URL: http://www.jstor.org/stable/2695154 . Accessed: 05/11/2011 06:56
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.

Population Council is collaborating with JSTOR to digitize, preserve and extend access to Population and Development Review.

http://www.jstor.org

Fertility, Education, and Development: Evidence fromIndia


JEAN DRtZE MAMTA MURTHI

INDIAIS IN the midstof a demographic transition exhibits that striking spa-

Withfewexceptions, and tial differences. have been defertility mortality but levelsand at different clining, from stateto greatly varying speedsfrom the state.Aside from intrinsic of thesepatterns importance understanding of of decline,the diversity the Indian experience providesan opportunity variousinterpretations thefertility of to reexamine transition. India,it may be recalled,was one of the first countries the worldto introduce nain a tional family in planningprogram, the 1950s. In the earlydays,"populationcontrol" itwas thencalled)appearedto assumesome urgency, (as with critics such as Paul Ehrlich(1968) warning the impending authoritative of of bomb" and ofthe specter famine overIndia itself. "population hovering that Then came a moregentleapproach, is stressing "development thebest this contraceptive." Initially was takentomeanthateconomic would growth reducepoverty and slow the growth population.The noof automatically itself evolvedas awarenessgrewthateconomicgrowth tionofdevelopment in per se did not mean a rapidimprovement the qualityoflife.Overtime, fromeconomicgrowthto social development, the focusshifted with the to lattercallingforeconomicgrowth be supplemented direct by action in fields such as publichealth,elementary and social security. The education, emphasison social development gained acceptanceas a growing body of research substantiated view thatpublicactionin these fields the empirical both to betterlivingconditions had much to contribute and to reducing populationgrowth. In recent about the abilyears,however,doubtshave been expressed an ityofthe social development approachto foster adequatelyrapidslowAttention been drawnto the successesof has down ofpopulationgrowth. and TamilNadu in achievingrapidfertility Bangladesh declines,allegedly by placingmore emphasison vigorousfamily planningprograms than on
POPULATION AND DEVELOPMENT REVIEW 27(l):33-63 (MARCH 2001)

33

34

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

Doubts about the effectiveness the social developof social development. some to arguethat"contraceptives the mentapproachhave prompted are after Renewedconcernabout the so-calledpopulaall. bestcontraceptive" tion explosion,notablyfrom environmental tiltedthe lobbies,has further balance in favorof energetic if family planningprograms, Chinese-style needed.' The alarmist backlash even takenconcrete has forms. Recentfamily refer "population to control" and the singleplanning slogans,forinstance, laws barring childnorm,and severalIndianstateshave introduced parents ofmorethantwo children from local elections. contesting Many otherproposals in the same veinhave been floated.2 We arguethattheIndianexperience does notwarrant disenchantthis mentwiththe social development approach.India is not a model of social development any means,butmanyIndianstatesare makingreasonable by decline throughnonauthoritarian methods. This progresswith fertility owes a greatdeal to the improvement femaleliteracy of progress and the declineof childmortality, much morecan be achievedin those direcand withauthoritarian tions.Experiments have had intervention, contrast, by results.3 is notto dispute This thatmorecan and shouldbe done in disastrous of Indiainthefield family convenient informed and planning. Indeed, providing accessto contraception nonterminal is com(including methods) an essential ponentofthesocialdevelopment muchneglected Indiaso far. in approach, The relation betweenfemale educationand fertility a crucialbearhas ing on thisdebate.Indeed,femaleeducationplaysa key role in the social development approach.A largebodyofIndian and international evidence femaleeducationin lowering pointsto the roleofrising In fertility.4 recent years,however,questionshave been raisedabout thenatureand interpretationoftheIndianevidence(Jeffery Basu 1996b). Severalstudies and have evidenceofa positive linkbetweenwomen'seducationand "fefoundlittle doubton one ofthemajorpathways maleautonomy," casting which through the former was supposedto reducefertility and (see, e.g., Jeffery Jeffery some studies-mainlyat 1996; Vlassoff 1996; and Visaria1996). Further, the villagelevel-reportno significant correlation betweenfemaleeducaIn tion and fertility. these studies,observeddifferences fertility in across to groupsor overtimeare attributed otherfactors, reductions including in infant and the rising mortality (Kolenda 1998), family planningprograms cost of children(Vlassoff risk 1996), and different environments (Jeffery and Jeffery the 1997). Without necessarily asdisputing generalstatistical betweenfemaleeducationand low fertility, bodyofliterature sociation this assertsthatthe associationis neitheruniversalnor well established, and thattheprocess whichfemale education influences through fertility-if such a causal linkexistsat all-remains farfrom clear (Jeffery Basu 1996b). and to Thisarticle attempts addresstheseissues.It examinesthe determiin nants of fertility India in a multivariate framework, usingdistrict-level

JEAN

DREZE

/ MAMTA

MURTHI

35

the censuses,1981 and 1991.5A district panel data linking twomostrecent and is the basic unitofadministration is the lowestlevel at whichspatially is on information fertility available (thereare over 400 disdisaggregated for tricts India). The panel aspectofthedata allows us to control districtin otherwise effects might that correlation between producea spurious specific variables.Even aftercontrolling disfor and fertility various explanatory women'seducationand childmortality emergeas themostimtrict effects, differences acrossthe country and over factors fertility portant explaining to also By time.Low levelsofson preference contribute lowerfertility. conof such as urbanand trast, generalindicators modernization development bear no significant and association ization, poverty reduction, male literacy decline. withfertility

Issues and hypotheses


Female educationand fertility on Althoughmuch has been written the subjectof femaleeducationand as whichthe relathereis a lack ofclarity to thepathways through fertility, Guio Followingour earlierworkwithAnne-Catherine tionshipoperates. (see Murthi,Guio, and Dreze 1995), we findit usefulto decompose the of into effect femaleeducationon fertility threedistinct effects, pertaining to of educationon (1) desiredfamily size, respectively the influence female between desiredfamily size and planned numberof (2) the relationship to and births, (3) women's ability achievetheplannednumberofbirths. Female educationcan be expectedto reducedesiredfamily size fora from in of number reasons, fertility goals greater autonomy defining ranging to to enhanced receptiveness modernsocial norms,reduceddependence and on sons forsocial statusand old-agesecurity, the higheropportunity betweenfemale cost of time foreducated women.6The negativerelation size in India is borne out in a wide rangeof educationand desiredfamily studies.7In additionto reducingdesiredfamily size, femaleeducation is size and planned between desiredfamily the likelyto affect relationship One reasonforthisis thatfemaleeducationreducesinnumberofbirths. Educatedmothers thusneed to plan fewer births fantand childmortality.8 femaleeducationmayasa desiredfamily size. Finally, in orderto achieve sist in achievingthe planned numberof births, especiallyby facilitating and women's barknowledgeofand access to contraception by enhancing For gainingpowerwithinthe family. example,the NationalFamilyHealth women in India have Survey(1992-93) foundthat6 percentof illiterate no knowledgeof any contraceptive method,comparedwithless than 0.5 of highschool.Amongthosewithsome percent womenwho have finished 16 women did not know knowledgeof contraception, percentof illiterate

36

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

for figure women withhigh whereto obtaina method;the corresponding Communication betweenspouses school educationwas around 1 percent. witheducation.These indialso sharply contraception increased regarding goals, thateducatedwomen not onlyhave different fertility catorssuggest Further theiraspirations into reality. but are also betterable to translate of linksbetweenfemaleeducationand fertility, indications these different are HealthSurvey, presented Table 1. in also based on theNationalFamily issue in thiscontext) Notethat"female autonomy"(a much-discussed mediatethelinkbetweenfemaleeduthatpotentially is one ofthevariables for women greater control overtheir cationand fertility, instance giving by betweenfemaleeducationand autonomyis itThe fertility.9 relationship thatthetwo are,in fact, Some studies poorlycorselfcontroversial. suggest female These Much also dependson how one defines autonomy."I related.10 bearingon the overallrelation issues,however,have a limited outstanding since betweenfemaleeducationand fertility, femaleautonomy(regardless
TABLE 1 Relationship between female education and fertility:Relevant indicators from the National Family Health Survey (1992-93) Woman,'s education Indicatorsrelated to Desired family size Mean ideal number of children Percentof women unsure about the acceptability of family planningmessages on radio or televisiona Child mortality Percentof childrenaged 12-23 monthswithout any vaccination rateb Under-five morality Contraceptive practice Percentof women who have never discussedfamily planningwithhusbandc married Percentof currently women not aware of any source of modern contraception Illiterate Literate,but below "middle school complete" Middle school complete High school and above

3.1

2.6

2.4

2.1

33

10

40 141

17 84

8 65

4 43

58

42

35

29

15.9

4.8

3.0

1.1

aMostother consider acceptable. family planning messages women of bProbability dyingbeforeage fiveyears. marriedwomen who are not sterilized and know at least one contraceptive method. 'Base: currently Institute Population Sciences 1995: Tables 6.2 (p. 131), 6.28 (p. 173), 6.29 (p. 174), 7.8 (p. 193), 8.6 for SOURCE: International are women aged 13-49 years. (p. 215), 9.11 (p. 250). These findings based on an all-India surveyof about 90,000 ever-married

JEAN DREZE

/ MAMTA

MURTHI

37

is variables. of how defined) onlyone of the possibleintervening Thus,the abouttheempowerment doubts thathave been raised value offemale educaand extended therelation to tion(Jeffery Basu 1996b)shouldnotbe casually the and That as between latter fertility. relation, willbe seenbelow,is robust. of Otherdeterminants fertility Many of the foregoing arguments applyto men as well as women. Thus, in improvements male educationmayalso lowerfertility. the Nevertheless, of is influence male educationon fertility likelyto be smallerthan thatof are decisions dominated men),because femaleeducation(unlessfertility by for women bear theprimary responsibility childrearing. of is The effect incomeon fertility harderto predict thanthatof education.'2Atleasttwobasicissuesare involved. incomeeffects likely are First, are as to depend on whetherchildren perceived an economicburdenor a asset.In the literature family on economicsin developedcounproductive has as the tries, tendency been to see children a consumption good,leading and inter to a focuson thecostofchildren thequantity-quality alia tradeoff. incomes makechildren moreaffordable, negaIn thisframework, but higher on are tiveincomeeffects fertility also possible-forexample,ifparents subas for stitute qualityl quantity incomerisesor ifhigherincomesare associated witha higher cost opportunity oftime.In developing countries, the on otherhand,children as maybe regarded economicassetsbysomeparentsforexample,because theyare a sourceoflaborpowerand old-agesecurity. to as Thisis likely reinforce incomesreduce negativeincomeeffects, higher the economicdependenceofparents theirchildren. on Indeed,the notion thatchildrenare economicassets effectively turnsthe affordability argufor ment(themainargument positive incomeeffects fertility) itshead. on on are of Second,incomeeffects not independent the sourceofadditionalincome. For instance,if higherincomesreflect highadult wages or a high of participation women in thelaborforce(bothofwhichraisethe opportubetweenincomeand fertility likely is nitycostoftime),the relationship to be negative.On the otherhand, if higherincomesreflect higherendowassets(such as land) thatalso raisethemarginal mentsofproductive product of childlabor,theymaybe associatedwitha higherdemand forchildren.'3Unfortunately, availabledatado notenableus to distinguish between sourcesofincome. different Accessto publichealthservices fertilmayalso playa rolein reducing of and Asidefrom direct ity, effects independently education income. through access to contraception, improved publichealthservices mayreducefertilityby enhancingchildsurvival. These effects maybe smallwhere services are of poor quality, is truein much of northIndia. In our earlierwork as (Murthi, Guio,and Dreze 1995), we foundthataccessto publichealthserv-

38

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

but effect fertility.'4 on Beices reduceschildmortality has no significant betweenthe 1981 and 1991 censuses,furcause of changesin definition stand. of as therinvestigation thisissue is difficult things has role of urbanization also been emphasizedin the literature The is to (e.g.,Schultz1981, 1994). Urbanization believed reducefertility because to to are and children less likely contribute householdproduction are more Insofar fertility as to in declineis in part difficult supervise an urbansetting. it is also likelyto proceedat a faster "diffusion a process," pace in urban areas,wherepeople have greater exposureto mass media as well as wider of to opportunities observeand discussthelifestyles othersocialgroups. factors affect also and cultural Diverseregional in fertility patterns the and Moore 1983; Basu 1992; Indian population(see Sopher 1980; Dyson For Maharatna1998a; and Bhat 1998,amongothers). instance, rates fertility tendto be higher amongMuslimsthanin othercommunities, althoughthe holds after for extentto whichthisrelationship controlling various socioIndianMuslims(e.g.,lowerincomes economicdisadvantages experienced by of have distinct and literacy rates)is a matter debate.'5Tribalpopulations and genderrelations, kinship patterns including higherratesof femalelathat bor forceparticipation, may encouragelower fertility. Similarly, the in higherstatusof women and weaker hold of patriarchy south India are to rates. believedto contribute lowerfertility of In additionto thesefamiliar determinants fertility, examinethe we in This possiblerole of son preference enhancingfertility. concernarises observation thatdesirefora specified from common-sense the numberof withthe transition towardsmall-family sons ofteninterferes norms,parif of in ticularly northIndia. To illustrate, the probability a newbornchild adulthoodis, say,0.75 (a plausiblevalue forstatessuch as Uttar reaching wantsthe riskof endingup withoutan Pradesh),a mother who, in effect, to adult son to be lowerthan0.05 has to givebirth threesons; thiswould on if are requiresix births average.16 By contrast, sons and daughters conto sideredequallyvaluable (so thatthepredicament avoid is thatofending of are up withno adultoffspring regardless sex), threebirths enough.Ifthe to of rises 0.75 are probabilitysurvival adulthood from to0.8,twobirths enough. of the As thissimple and exampleillustrates, combination sonpreference high causeofhighfertility. childmortality be a particularly may important Endogenousfactors we be So far, have focusedon variablesthatmight reasonably expectedto but influence fertility not be influenced it. Examples of variablesthat by standin a relationof mutualinterdependence withfertility femalelaare and For bor force femalelaborforce participation childmortality. instance, lower fertility.'7 participation may both lead to and resultfrom Similarly,

JEAN DREZE

/ MAMTA

MURTHI

39

betweenchildmortality fertility. one shouldexpectmutualinfluences and is for On the one hand, high fertility likelyto raise child mortality both biologicaland behavioralreasons.'8On the otherhand,highlevelsofchild mortality tendto raisefertility, inducing by parentsto have morechildren desirein anticipation losingsome (so-calledhoardof than theyultimately ingbehavior),and also to replacelostchildren. Bearingin mindthesemubetweenmortality fertility,is important view fertiland it tual influences to context the"demographic of transition" from itydeclinein thebroader high and to low ratesoffertility mortality.'9 are variables influenced the dependent When explanatory variable by or (in thiscase fertility), whenbothare influenced thesame unobserved by estimators biased and inconsistent. are Consistent estivariables,standard if is mation(e.g.,usingtwo-stage techniques) possible adequate instruments but credibleinstruments are can be found forthe endogenousvariables, In for this hardto findin thiscontext. the case ofchildmortality, instance, an thataffects child mortality would requirespecifying exogenous factor withfertility. avoid thesedifficulties, To correlated this but is not otherwise in articlefocusesmainlyon reducedforms, whichwe considerthe effects be to ofvariables thatcan reasonably expected be exogenous.The estimated variablemeasuresitstotalimpacton fertilof coefficient each independent effects of indirect ity, including involving adjustment the endogenousvariof the the ables. To illustrate, coefficient female literacy captures totaleffect of femaleeducation,including any impactit might have via reducedchild While thisprocedure mortality. helpsto removethe simultaneity bias, we as remainuncertain to thespecific contribution say,declining childmorof, to reduction. thepenultimate we In to tality fertility section, attempt idenbased on two-stage this estimation. tify effect As to son preference, thisarticle treatthisindicator an exogin we as of of enous variable.Our interpretation the coefficient thisvariableis conditionalon the exogeneity which appears to be plausibleas assumption, are The to stand.20 otherconclusions quiterobust withrespect differthings of indicator. enttreatments theson-preference

Statistical analysis
Data variableanalyzedin thisarticle the district-level feris The dependent total rate (TFR),availablefrom Government India (1997) forboth 1981 of tility and 1991. The TFRis thesum ofprevailing rates age-specific fertility (births census questions per woman per yearin a givenage group),derivedfrom the on births of during previous yearand number children everborn.Thus, TFR measuresthenumberofchildren thatwould be bornto a woman dur-

40

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

if in ingherlifetime at each age she were to bear children accordancewith rate. the prevailing of age-specific estimates fertility fertility Whileindirect of based on census data involvesubstantial it margins error, is reassuring withthe estimates the Sample Registraof thattheyare broadly consistent and tion System(SRS), as faras interregional intertemporal are variations the concerned. instance, correlation For coefficient between SRS-based cenand of sus-based estimates state-level rates fertility in 1991is0.968.21Thetwosources of of estimates theextent fertility declinebetween1981 also producesimilar and 1991 (0.8 and 0.7 births woman,respectively). per is The district a usefulunitof analysis, bearingin mindthe social didecisionstend to be highlyinterdemensionof fertility change. Fertility of pendent,owinginteralia to the influence social normsand diffusion efin fects.Capturingsocial effects household-levelanalyses raises serious difficulties.22 alternative An conceptualand empirical approachis to treat rate as thanan individual the fertility itself a social rather variable,and to as an integral effects thinkof the interpersonal partof the observedrelaand tionsbetweenfertility othervariables(e.g.,educationor urbanization). considerthe relation betweenfemaleeducationand fertility. To illustrate, It is reasonableto expectan increasein educationlevelsto foster spread the of small-family normsnot only among educated women themselves but A also among uneducatedwomen.23 district-level analysishelpsto capture of thissocialdimension theeffect education fertility.standard of on A houseon hold-levelanalysis, the otherhand,would missthe effect a woman's of
education on the fertility decisions of other women.24

On the otherside, iffertility decisionsare, in fact,drivenmainlyby individualand household characteristics (with social effects playinglittle role),thenhousehold-level in analysesare moreappropriate, bearing mind the potentialaggregation involvedin treating district the the problems as unit of analysis.However,giventhe well-knownimportance social efof fectsin this field,a district-level analysisseems at least as plausibleas a household-level of analysis.The two approaches, course,are not mutually exclusiveand are indeedbestseen as complementary. to variables(listedin Table 2), our indicator Turning the explanatory offemaleeducationis literacy the 15+ age group("adultfemaleliteracy" in and similarly forshort), withmale education.25 is Poverty measuredby the ruralheadcountindex: the proportion the ruralpopulationbelow the of line.26 Urbanization the share of the populationresiding uris poverty in ban areas.The sharesofscheduledcastes,scheduledtribes, Muslimsin and the populationare used as indicators the social composition different of of districts. use five We variables identify to dummy regional patterns: "South" fordistricts the statesofAndhraPradesh,Karnataka, in Kerala,and Tamil for in and Nadu; "North" districts MadhyaPradesh, Rajasthan, Uttar Pradesh; "Bihar"fordistricts Bihar;"East"fordistricts Orissaand WestBengal; in in

JEAN DREZE

/ MAMTA

MURTHI

41

TABLE 2
Variable

Variable definitions and sample means, 1981 and 1991


Definition 1981 1991

TFR Female literacy Male literacy Poverty Urbanization Scheduled castes Scheduled tribes Muslim Son preference South dummy Northdummy Bihar dummy East dummy West dummy Child mortality Access to drinking water

Total fertility rate Percentofwomen aged 15 and older who are literate Percentof men aged 15 and older who are literate Percentof populationbelow the poverty line Percentofpopulationresident in urban areas Percentof scheduled castes in totalpopulation Percentofscheduled tribes in totalpopulation PercentofMuslimsin total
population

5.1 22.2 (15.2) 52.3 (14.3) 40.3 (13.6) 20.5 (15.1) 15.9 (7.1) 8.9 (15.2) 9.8
(9.1) (0.9)

4.4 29.9 (17.7) 59.4 (14.1) 32.3 (13.8) 22.4 (16.1) 16.7 (7.2) 8.9 (15.5) 10.7
(10.1) (1.0)

Ratio of female-to-male mortality


(5qo) in 1981

1.07
(0.13)

in Dummy= 1 fordistricts AndhraPradesh,Karnataka, Kerala, and Tamil Nadu in Dummy= 1 fordistricts Madhya Pradesh,Rajasthan, and UttarPradesh in Dummy = 1 fordistricts
Bihar

0.21 (0.41) 0.39 (0.49)


(0.29)

0.21 (0.41) 0.37 (0.48) 0.12

0.10

(0.32)

in Dummy = 1 fordistricts Orissaand West Bengal in Dummy = 1 fordistricts Gujaratand Maharashtra thata childwill Probability die beforereachingfifth birthday 1,000) (x Percentof households with access to safe drinking water

0.09 (0.28) 0.14 (0.34) 156.7 (42.9) 34.7 (22.8)

0.08 (0.28) 0.13 (0.34) 106.3 (36.7) 59.8 (22.4)

NOTES: Means are unweighted. Standarddeviationsin parentheses. of SOURCES: Government India (1997), forTFR and childmortality.For poverty, discussionin the text. see The remainingvariablesare calculated fromCensus ofIndia 1981, Social and CulturalTables, PartIV-A; Census of India 1981, GeneralEconomic Tables,Part 3; Census ofIndia 1981, Series 1, Paper 1 of 1982, "Final Part2-B(i); Census of India 1981, Series 1, Census Abstract, population totals";Census of India 1981, Primary Paper 4 of 1984, "Household populationby religionofthe head ofthe household"; Census ofIndia 1991, Social and CulturalTables, PartIV-B, Table C-3; Census ofIndia 1991, Primary Census Abstract, General Population, of PartII B1(i);Government India (1989, 1994); and Census of India 1991, Paper 1 of 1995, "Religion."

42

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

in and "West"fordistricts Gujaratand Maharashtra.27 default The region to (corresponding zero values forall dummy of variables)consists Haryana and Punjab. (Discussionofthe relation betweenchildmortality access and waterand fertilitydeferred thepenultimate is to to drinking section.) We use the ratiooffemale-to-male childmortality an index ofson as Thereis indeedmuchevidencethatson preference reflected preference. is child mortality ratesbetweenboys and girls.We also tried in differential otherindexes, the including female-male ratio, laggedvalues ofthefemaleratio.The results male ratio,and thejuvenile female-male were similar in for each case. We use the same son-preference indicator both 1981 and of child mortality 1991; thisis because the limitedprecision district-level us from variations theratiooffemale-toin estimates prevents interpreting ratesbetween 1981 and 1991 as a reliableindicator male childmortality of in underlying changesin son preference (especially view of the slow pace ofchangein thelatter domain). is The information overwhelmingly available fromstandardcensus sources (see Table 2 fordetails).The only exceptionis poverty. India, In estimates usuallycalculated are from NationalSampleSurvey(NSS) poverty data. The NSS sample frameis not suitableforthe calculationof districtbut level poverty to of estimates, it does lend itself the estimation poverty at indicators the level of NSS divisions.28 latter intermediate The are units betweendistrict state.Everystateconsists severaldivisions and of (threeto fivein mostcases), and each division turnis made up ofseveralcontiguin ous districts thatare reasonably and sociohomogeneousin agroclimatic economicterms.For each district, use the headcountindex applicable we to the divisionwhere the district situatedas the relevantpoverty is estimate. Since division-level are estimates available only forspecific poverty 1981or 1991),we calculate years(notincluding estimates thecensusyears for Further discussion theprocedure of used topool censusand byinterpolation.29 in NSS datamaybe found Murthi, and Srinivasan, Subramanian (1999). for Usingthedivision-level estimate each district poverty within para ticular division can be justified thegrounds on thatthedivisions meant are to be reasonably Thus poverty levels may not homogenousin thisrespect. betweendistricts withina division. theydo vary,the coeffiIf varygreatly cientofpoverty remains unbiased(thoughthereis some loss ofprecision), of but the coefficients variablesthatare correlated withpovertymay be biased.30To addressthisproblem,we have examined the basic relationand the divisionlevel, where the division-level ships at both the district are as aggregates constructed population-weighted averagesofthe districtlevelvariables. The division-level in regressions (reported Dreze and Murthi thereis some loss ofprecision 2000) are freeofestimation bias,although as the shift from district division to reducesthe numberof observations. it As turnsout, the division-level results highly are consistent withthe districtlevel results.

JEAN DRtZE

/ MAMTA

MURTHI

43

Table 2 indicates thatfertility by 0.7 (from5.1 to 4.4 children fell per woman) between 1981 and 1991. Duringthe same period adult literacy a withfemaleliteracy increased, but a achieving larger improvement from was stillas low as 30 permuch lowerbase. In 1991, adultfemaleliteracy for cent,barelyhalfthe corresponding figure men. Urbanization increased slowlybetween 1981 and 1991. The headcountindex ofpoverty from fell to and 40.3 percent 32.3 percent, childmortality hereas theprob(defined a abilityof dyingbeforeage five)registered substantial drop,from157 to 106 per thousand. cross-sectional variation thedata,as illustrated in Thereis considerable in Table 3, which reports means forIndia's 15 major states state-specific in (thosewitha populationexceeding10 million 1991).?1The totalfertility 6.0 raterangedfrom in Rajasthanto 3.4 in Kerala in 1981,and from in 5.4 UttarPradeshto 2.6 in Kerala in 1991. Rates of fertility declinehave also uneven. Comparedto an all-Indiadeclineofabout 14 percent been highly declinedby some 20 percentor more duringthe reference period,fertility in the statesofKerala (3.4 to 2.6), Punjab (5.0 to 3.8), and TamilNadu (3.9 reductions Keralaand TamilNadu are all in to 3. 1). The largeproportionate that a *the moreremarkable considering theywere achievedfrom relatively in low base. Comparedto thesereductions, fertility the largest state,Uttar declinedby some 7 percent Pradesh, (5.8 to 5.4), whilein itseastern neighbor,Bihar,the declinewas ofthe orderof4 percent (5.2 to 5.0). Estimation to of Our mainresults pertain panel data regressions theform:
TFRdt = ad +

3Xdt+

t + 8Fdt

rate d where TFRdtis the totalfertility in district at time t, xd is a districtof of specific varieffect, is a vector coefficients, is a vector explanatory , Xd, and edt is an error is term.In thissectionwe focus ables, y, a timedummy, on a reduced-form equation,whichexcludeschildmortality (the latteris examinedin a subsequentsection).Our explanatory variablesare adultfeadultmale literacy, son male literacy, repoverty, urbanization, preference, of gional location,and the social composition the population(see Table 2 Our of fordefinitions). sampleconsists 326 districts, covering14 of the 15 majorstatesand over 96 percentofIndia's population.The missing major stateis Assam,whereno censustookplace in 1981.32 We can thinkabout the district-specific effects (cd) in two ways, as or fixedeffects as randomeffects. the fixed-effects In approach,the district as effect estimated thecoefficient a district-specific is of dummy variable, by leastsquares(OLS).33In therandom-effects ordinary the approach, districteffect modeledas an additional, is time-invariant term each specific error for

~~
16.

_< n

V. u,

uxN

oo

_::

_-

(::

o 00

C,

C>

ri n

r'

0 S

nb

m >% Ln oO rq rq \0 rq 00

n~~~~~c> _r\..r o v

mb *_ O r?

.............

_~

oo

t~

c_

q>

\6

0?bO

rA

>

_0

_\

-r

Nr -r

-,

C\ -q

rn -0

Nr

.R

E
m

r>6
X t O

W~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~r
< oo o\ N\ _
9 r _ e_ _

;\

N -

~\

C\_

n~

~ ~ ~~~~~L
(>

s M Oou ouu \x)u


_o P_ t VD_ _ e_ t e V -U

,_

N ~~~~O

?0 00 ?O -\ M 00, N\ Oq N

M M 00 C\\-O

00

y\ _\ -

_\

w~~~~~~r

00

M *\

oebo ?OoO0 N

r-

rq

CD

0 CD

C\

r.

rO-, M 1s0

u~~~~L\ 00
m _ e

Co-s 0

\ _

- v, rq r- ?O N q Lr\ \

m rNLr O,

v, \O rO

N 0sC\ox

\C

_ s\ ~~LrIL O c;

00

" ~t~

Lr 1 I
\CU Lr ? tt M

C\ _

i~~~~~~~C _; o1
Y~

ro

?O

Lr5 ,,

?O Lr
Nt rq

c0

C5f c;

E
oO

>u

\C

~~~0

??O

00 \0

)N

00

0r

0 L \ N\ \0

;~~~~~~~~~~0 _t

eeuo

OuNu M
Lr

0eOo~tt?C 't

F3

>

~~~~~O _q

M\'O

00

VI\

-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.> . _ t N ?? t

~~~~~~ <2t' ~~
e b 4-J n

>

uE

~-O

e ^ ?4

X _ t8

~~~~~

(> ?

t e

ufi N

N o

e c ?

sO m uELnct

to -

' n

?O t _ V

o~~~~~~~6

X~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

_-

.*FlH.. Lz

- ---

NO _
Lr\

>
-. Lr\

r 00 UE
\00Lf

n o 90
\ L(\

e
d

q O

aX Lo o\
?O L O

t
?-?

r u

_
m -

\r

o r

r(\ -O-

>~~~~~~~~~~~~~~C

LLx_:oN?_b_? t_n \_ O-

o, N__o_

rq

rq C\

0\

0 ?04

rq t

~~eI~~~ -

(N1

~~~~~r

4rq

\I

L(N-

oo~

Lr

4L-

\r rfq-

0 o

~ rf(NI
f

~C\

\ON
-

O\ N.-O \
\

..~~~~~~~~~~~C~C
00 1 ?-\ CD
ON \ -

0r00 r 0\

-C\

C\I N\. rNq0 00


q

r 0000~~~~~~~~rM0
r N \ 00_ -q
00

\..

~\;0

m\

\p'

mL
Qo0 f

\ N 0 c ?O \ r0
Lr

mrlrq0
\

\6 Lr\c\ nL~ 0 i N \0

00\ rqo
N O
(>-

m 00

mr

r 00
I 00( G

_ 00
LNr

o
-r

O o
0\-r

ur 0\

-o
0

C\

UE0

0,

C\

(~~~~~~~~~~~~~~~~~~~~J CO~~~~~~~~~~~~~~~~~~~~~~~~~~~~~6

)C

r
CU

-.'~ .N C\oO

.L.

--OONO

..

O\\\ Lr\ O\ rq

N 00

-0

r- L(\ (\NM- er-O O Ol


K
00 m

O -00 ( c

-?

--

O r-00O t
UE

\6 r te~
- o0 Q
\

Lr\6

Vi

rdv

lsKO

? Oo _

K 0 N 90 O _ 0_ _

0
_

_ 0
\OL(N

r-..

U 0 0
CO

\m

O O O 00 \ O n O-N . \C) M _ - _ - _ _. - b 0\~r-~.(NI\ -r~. _NI00 \ _0r(4 O 0~~N.L(N Q r0-\ rm'0 .N

_r. _
\ \

-- N

-(

--

(NI(_I

_ -

0\r-t

r(NN.

_f

r(

00

0oo N0 Nt Lr\ \c \

0 L\ 0

o- \o

-oo

t C

t(N

\Qo oOo

0\r(N

00t - -4\O

o-

'r-o

O o Xo

I\O O

ton \0

b w~~~~~~ > e~~~~

0L CO

0o0 CO

bNb0 CO

AtO

~~~~~~~~~~~~~~~~~~~~~~~~~~
U4

o 0

o Z

0 CO

tO9

. 00

46

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

The error district. composite term, edt' has a particular ad+ covariancestrucleast squares (GLS). The randomture,allowingestimation generalized by effects approachhas theadvantageofsavingmanydegreesoffreedom (ushalves the numberof observations). ing a dummyforeverydistrict Also, it unlikethefixed-effects approach, does notprecludetheinclusion timeof variables suchas regional dummies. the invariant However, random-effects is random error uncorrelated with approachassumesthatthedistrict-specific the other explanatoryvariables,which may not be the case. To check is we the the whether random-effects approach appropriate, testfor orthogoand the regressors nalityof the random effects using a test devised by all testsof significance Hausman (1978). To perform tests(including and we use the robustHuber-White estimateof variance, the Hausman test) acrossdistricts well as correlaas error variances whichallows fordifferent tionacrosstimeforgivendistricts. Main results We our the Table 4 presents mainresults. beginbyconsidering relationship Column 1 reports estimates 1981, and cross-sections. for in the individual column2 for1991. columnof Table 4 we see thatthe explanatory variables In the first of in acrossdistricts accountforaroundthree-fourths thevariation fertility of is and highly in 1981. The coefficient femaleliteracy negative significant. of indexis positive and significant, The coefficient the son-preference conthat son preference enhances fertility.34 contrast, findno we firming By levelsand generalindicators develrelation betweenfertility of significant such as the poverty male literacy, and modernization and opment index, The indexhas a negative to urbanization. poverty sign,contrary thenotion Male is makesno contributhatpoverty a cause ofhighfertility.35 literacy after for tion to fertility reduction, controlling femaleliteracy. Amongthe caste and religionvariables,only the Muslimpopulationshare is significant,witha positive sign. work(Murthi, In earlier Guio,and Dreze 1995), we had founda negaon tive coefficient the scheduledtribes variable,in line withmore recent low fertility rates in tribalcommunities research suggestingrelatively in (Maharatna 1998a, 1998b). The factthatthisvariableis not significant the of indicator the latTable 4 arisesfrom inclusion theson-preference (if as of teris dropped, in the last columnof thattable,the coefficient schedThere is much evidence that son is negativeand significant). uled tribes is communities thanin theIndian preference lesspronounced amongtribal low populationat large,and thisfactseems to accountfortheirrelatively rates.36 fertility

JEAN

DREZE

/ MAMTA

MURTHI

47

TABLE 4 Fertility in India: Main results of the regression analysis (dependent variable: TFR)
Panel 1981-91: 1981: OLS (1) Constant Female literacy Male literacy Poverty Urbanization Scheduled castes Scheduled tribes Muslim Son preference 1991 time dummy Regional dunmmies yes (see Table 5) Adjusted R2 F (p-value) Pseudo-R2 Wald, %2 (12) (p-value) Sample size GLS vs. FE, %2 (8) (p-value) 326 326 0.73 64.3 (0.00) 3.338** (8.28) -0.019** (4.32) 0.0001 (0.02) -0.005 (1.50) -0.003 (1.18) 0.005 (1.03) 0.002 (0.84) 0.016** (4.98) 1.98** (5.98)
-

1991: OLS (2) 2.760** (6.04) -0.025** (6.66) 0.004 (0.87) 0.002 (0.99) -0.002 (1.51) 0.001 (0.13) 0.004 (1.26) 0.021** (6.76) 1.66** (5.19)

GLS-RE (3) 3.343** (9.18) -0.022** (7.37) 0.001 (0.24) -0.001 (0.68) -0.002 (1.53) 0.003 (0.78) 0.002 (0.99) 0.019** (7.39) 1.83** (6.49) -0.52** (13.48) yes (see Table 5)

OLS-FE (4)

-0.022** (2.80) -0.005 (0.68) -0.001 (0.15) 0.003 (0.33) 0.002 (0.17) -0.038* (1.71) 0.053* (1.81)

-0.50** (5.96) no

yes (see Table 5) 0.79 116.0 (0.00)

0.94 76.09 (0.00) 0.77 2503.0 (0.00) 652 8.4 (0.39) 652

FE OLS = ordinary least squares; GLS = generalizedleast squares; RE = randomeffects; = fixedeffects at at **significant 5 percentlevel *significant 10 percentlevel in are are NOTES: Absolutet-ratios parentheses. All standarderrors robust.The F- and Wald-tests testsofthe thatall coefficients (except the constantterm)equal zero; the Hausman test(GLS vs. FE) is a joint hypothesis testof random versusfixedeffects.

48

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

estimates 1991. Herewe are for the Column2 presents corresponding in acrossdistricts. of able to explainnearlyfour-fifthsthevariation fertility to are conThe results verysimilar thoseobtainedfor1981. Femaleliteracy and its is tinues be negative highly to significant. Moreover, effect even larger in 1991 than in 1981. This is indeed what one mightexpect,considering of thattheaveragelevelofeducation(in terms say,yearsofschooling) a of, in the literate womanwas higher 1991thanin 1981.However, female literacy for coefficients 1981and 1991are,statistically speaking, indistinguishable. in Nextwe pool the data, allowingfora different intercept 1991 and a effects. column3 we estimate random-effects In model fordistrict-specific model.37 Both modelsbroadly confirm the and in column4 a fixed-effects thatthedistrict-specific Undertheassumption effects cross-section findings. to results. When are are random,the results verysimilar the cross-section are are the district effects taken as fixedas in column 4, the coefficients with less precision(in general,the standarderrorsmore than estimated information abis double). Thisis because a greatdeal ofthe cross-section Even in thisspecification, coefthe dummies. sorbedin the district-specific remains and from of ficient female literacy negative highly significant. Apart and none ofthe othervariablesis signififemaleliteracy the timedummy, level.However,the largestandard errors that cantat the 5 percent suggest are from different therandom-effects estimates. thecoefficients notsignificantly whichdoes notreject nullhypoththe Thisis confirmed theHausmantest, by effects to esis thatthe district-specific are orthogonal the regressors. other In the model of we words, neednotreject random-effects in favor fixed-effects. theseresults of Taken together, yielda consistent picture the relation is and Femaleliteracy significantly associbetweenfemaleliteracy fertility. and is acrossthe ated withlower fertility, the size of the effect also similar This is otherfactors, such regressions. effect upheldeven whenwe allow for as male literacy, and forunobpoverty, urbanization, caste,and religion, The influences fertility. continuedsignificance serveddistrict-specific on of the in model challenges view thatlow ferfemaleliteracy the fixed-effects are and driven some commonthird factility highfemaleliteracy jointly by correlation the in tor (such as the statusofwomen),and thatthe observed cross-section in thissense spurious. is thattheobservedassociation The possibility remains betweenfertility reflects joint influence some unobserved, the of timeand femaleliteracy are variable.To illustrate, varying supposethatdistricts movingat different reasonsunrelated education)toward to modern norms paces (and for family on thatsimultaneously and eduplace greater emphasis smallfamilies better Thismight a association betweenfercationforchildren. generate negative acrossdistricts, even in the absence of any directcausal and literacy tility we linkbetweenthetwo.Ifthiswerethecase, however, would also expect to find strong a association betweenfertility male literacy. and The negative

JEAN

DREZE

/ MAMTA

MURTHI

49

thatthe factthatwe find such association no helpsto dispelthehypothesis is and fertility drivenby unobserved associationbetween femaleliteracy towardmodernfamily norms,forintime-varying variables;progression be correlated withbothfemaleand male literacy. to stance,is likely on betweenfertility It is worthreflecting whywe findno association This finding contrasts withthe commonnotionthatchiland poverty.38 seen as economicassetsin poorhouseholdsand thatpovdrenare typically to contributes highfertility. ertytherefore Here,it is usefulto distinguish as between two senses in whichparentsmightsee children economicasthe sets. One is thatchildren(particularly sons) provideold-age security, work.The second intake partin productive otheris thattheysometimes to cause of terpretation mightlead us to expectpoverty be an important There is little high fertility.39 evidence,however,thatIndian parentssee suras qualitative children economicassetsin thatparticular sense; rather, as vey responsessuggestthattheytend to considerchildren a short-term of in economicburden,made worthwhile largely the prospect security by for old age.40The latter motive, itspart,mayor maynot be closelyrelated a concernforold-agesecurity partly deis to incomelevels.For one thing, one's acquiredliving standard old age, rather in thansome sireto maintain absolute level of living.For another,what is at stake is not just income but and that aspectsofwell-being security, also othermaterial psychological in are associatedwithbeingable to live withone's children old age. These needs, again,may or maynot declineas incomerises.Even so, one would motiveto have moreinfluence fertility on stillexpectthe old-age security access to credit at low levelsofincome-for example,because ofrestricted to alternative children. But thisrelation and otherinsurancemechanisms and the impliednegativeassociationbetweenfertility need not be strong, incomeeffects and incomecould easilybe neutralized positive (relating by to in particular theaffordability argument). It is also possiblethatthe absence of any significant associationbein and the tween fertility poverty Table 4 reflects lack of precisionof our to refer divisions indicators was explained rather poverty (as above,thelatter and fromreference than districts are based on interpolations years other the betweenpoverty fertilthan 1981 and 1991). Further, relationship and Thereis scopefor research thisissue,based on further itymaybe nonlinear. level.Meanwhile,this on forinstance variousincomeproxiesat thedistrict thatitwould be unwiseto relyon incomeeffects achieve to studysuggests reductions fertility in levels. Regionaleffects the of Table 5 presents estimated coefficients theregional dummy variables, Table 4 forexpositional which were excludedfrom The first clarity. three

50

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

TABLE 5 Fertility in India: Regional variable: TFR) (dependent 1981: OLS (1)

effects in the regression analysis Panel 1981-91: 1991: OLS (2)


Table 4 (2)

GLS-RE (3)
Table 4 (3)

2SLS-RE (4)
Table 6 (3)

Othervariablesas in:
South dnummy North dummy Bihar dummy East dummy West dummy

Table4 (1) -0.43** (3.38) 0.45** (3.91) -0.21 (1.21) -0.16 (0.82) -0.17 (1.17)

-0.42** (3.59) 0.52** (5.06) 0.17 (1.14) -0.24 (1.63) 0.03 (0.27)

-0.45** (4.48) 0.47** (6.04) -0.03 (0.26) -0.22 (1.64) -0.08 (0.80)

-0.46** (4.07) 0.28* (1.94) 0.002 (0.14) -0.30** (2.13) -0.15 (1.18)

least squares; GLS = generalizedleast squares; 2SLS = two-stageleast squares; RE = random OLS = ordinary FE effects; = fixedeffects at at *significant 10 percentlevel **significant 5 percentlevel of NOTE: This table should be read as a continuiation Table 4 (columns 1, 2, 3) and Table 6 (column 3). It gives of and coefficients absolute t-ratios regionaldummiesin the relevantregressions. regression

columns umn

present estimates corresponding dummies were estimated of Table regional 4), where

to the first three colunms

of Table 4. (colEstion ferwith We

No regional

in the fixed-effects specification by district dummies. further on. compared per woman.

4 in Table As Table

they are superseded location

mates in the last column 5 indicates, tility even India:

5 will be discussed

exerts a strong influence

after controlling

for other factors. In particular, and Punjab), 0.4 children regressions

the default region (Haryana experimented for by including However, bust pattern. persists West

fertility is distinctly lower in south to see whether be accounted form. is a roregion and to

in terms of the total rate, by around

with many variants of the baseline other explanatory variables

the lower fertility levels in the southern the distinctiveness Similarly,

region could somehow or changing region

the functional

of the southern

in this respect in the northern

fertility is distinctly higher (by about (Gujarat 0.5 children and specifications.

than in the default region in alternative and Bengal)

per woman),

and the gap (Orissa

Fertility rates in the east Maharashtra)

the west

are fairly close

rates in the default region. Fertility decline We and gender bias with a short digression on the relation between by Das our Gupta

end this section

findings on son preference

and the striking thesis, advanced

JEAN

DRtZE

/ MAMTA

MURTHI

51

and Bhat(1995, 1997),that gender bias"inIndiais "a consequence "intensified of betweenfertility In discussion therelation decline." their decline offertility bias in childmortality, authors the betweentheparity and gender distinguish effect: characterized a strong by prefeffect theintensification "Insocieties and declinehas two opposingeffects discrimination on erenceforsons,fertility at births higher parities where againstgirls.On one hand, thereare fewer is and (the discrimination against girls strongest, thisreducesdiscrimination discrimination On becomes 'parity'effect). the otherhand, parity-specific discrimination intensifimorepronounced...and makesforincreased (the is effect highly sincefemale childmorThe cationeffect)."'4' parity plausible, as with parity, we know interalia fromDas talityin India rises sharply work on thissubject.Das Gupta and Bhat atGupta's (1987) pioneering but effect, temptnot only to establishthe existenceof an intensification the effect. that also to demonstrate in India thiseffect outweighs parity the The evidenceforthiscomesin two forms. First, authorsconvincthe spreadof sex-selective abortion. The latter contributes inglydocument declineand raisesparity-specific discrimination we countabor(if to fertility Das Gupta and Bhat). The tion as a formof prenatalmortality, following relevanceofthisobservation, however, dependson thespreadofsex-selective abortionitself decline";thishas not being "a consequenceof fertility been argued.Withor without fertility decline,one would expectthe introin ductionof sex-selection technology India to have many takers.In any is abortion a minorchannelof fertility decase, the spreadof sex-selective note. Evidenceon sex-selective as the authorsthemselves abortion, cline, on has betweengender therefore, a limited bearing thegeneralrelationship decline. bias and fertility and data the in Second,usingfertility mortality from Khanna re-study Das Guptaand Bhat presentsome evidenceof Punjab (Das Gupta 1987), effect thisparticular in case: parity-specific the intensification genderbias is in childmortality higher this Further, amongwomenwithlowerfertility. is effect strong intensification associaenoughto inducean overallnegative bias (i.e.,itoutweighs parity and the The tionbetweenfertility gender effect). is and in theabsenceofstatistical itis diffitests sample, however, verysmall, of let cultto assessthesignificance this applicability. pattern, aloneitswider indicatethatin India as a whole, the associationbeOur own results and is tween fertility genderbias in childmortality firmly rather positive, In this than negativeas in the Khanna sample.42 itself, is not a conclusive refutation the claim thatthe intensification of effect (such as it may be) sincethatclaimis concerned the withfertility dominates parity deeffect, thanwithcross-section But clineovertimerather do patterns. our findings thatthe evidenceused byDas Guptaand Bhat to substantiate their suggest thesisis at variancewiththegeneralpatterns observed India as a whole. in On balance, thereis littlereason to fearan intensification gender of bias in Indiaas a consequenceoffertility decline.Indeed,virtually change no

52

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

in childmortality Indiabetween has occurred theratiooffemale-to-male in decline.Sex-selective 1981 and 1991,a periodofrapidfertility abortion did spread,but thatis best seen as a social problemin its own right(linked withtechnological thanas a consequenceoffertilprimarily change)rather itydecline.

and Fertility child mortality


we In our regression In analysis have so farfocusedon reducedforms. parfrom listofindependent the we ticular, have excludedchildmortality varibias. to ables to avoid simultaneity In thissectionwe attempt estimate both on and the effect childmortality fertility the extentto whichthiseffect of betweenfertility female and statismediatestherelation literacy. Summary are ticson childmortality givenin Tables2 and 3. the and As explainedearlier, factthatfertility childmortality inare the means thatestimating impactof one on the otheris not terdependent To straightforward. removethe simultaneity bias, we need an instrument thatis correlated withchildmorfor childmortality-an exogenousvariable associated withfertility. possibleinstrument One but tality is nototherwise is access to safe drinking water.The latteris highlycorrelated with child and thatit is directly associatedwith mortality, thereis no reasonto think on waterin 1981 and 1991 is Information access to safedrinking fertility. of Government India (1989) and Government India (1994), of availablefrom and in respectively, is summarized Tables2 and 3.43Thisvariableis used as in that for an instrument childmortality theregressions follow. leastsquaresestimates Table 6 presents two-stage usingpanel data for the column1 reproduces random-effects 1981-91.44 Forcomparison, (GLS) estimateof the reducedform, excludingchildmortality (thisis the same as as regression in Table 4, column 3). Column2 includeschildmortality an additionalregressor, assumingit is exogenous. Column 3 instruments while column4 presents first the forchildmortality, Our stageregression. is interest in column3. primary betweencolumns2 and 3 enables us to testwhether A comparison is child mortality exogenous (assumingsafe drinking water to be a valid As that instrument).45 it turnsout, we cannot rejectthe null hypothesis is childmortality exogenous,X2(1)=0.02 (0.90). Indeed,the GLS estimates leastsquaresestimates column3 are very in in column2 and thetwo-stage is Childmortality foundto have a positiveand significant similar. effect on The thata fallin childmortality 50 of fertility. size ofthecoefficient implies in terms happened between1981 and 1991) would perthousand absolute (as 0.2 children woman. Correspondingly, inclusion reducefertility the by per in ofchildmortality theregression reducesthe coefficient the 1991 time of from -0.52 in column1 to-0.35 in column3. Controlling child dummy, for

JEAN DREZE

/ MAMTA

MURTHI

53

TABLE 6 Fertility and child mortality (instrumental variable estimation, based on 1981-91 panel)
Dependent variable TFR (GLS-RE) (1) Constant Female literacy Male literacy Poverty Urbanization Scheduled castes Scheduled tribes Muslim Son preference 1991 time dummy Child mortality Access to drinking water Regional dummies yes (see Table 5) 0.77 2503.0 (0.00) 652 yes yes (see Table 5) 0.80 2834.9 (0.00) 652 0.78 1918.4 (0.00) 652 0.70 2432.7 (0.00) 652 3.343** (9.18) -0.022** (7.37) 0.001 (0.24) -0.001 (0.68) -0.002 (1.53) 0.003 (0.78) 0.002 (0.99) 0.019** (7.39) 1.83** (6.49) -0.52** (13.48) TFR (GLS-RE) (2) 2.764** (7.99) -0.020** (7.19) 0.002 (0.75) -0.004 (1.62) -0.001 (0.55) 0.002 (0.56) 0.002 (0.92) 0.020** (8.20) 1.68** (6.32) -0.33** (6.74) 0.005** (6.21) TFR (2SLS-RE) (3) 2.830** (6.14) -0.020** (6.50) 0.002 (0.68) -0.002 (0.91) -0.001 (0.49) 0.002 (0.61) 0.002 (0.95) 0.020** (7.42) 1.70** (6.56) -0.35** (3.37) 0.004* (1.73) -0.461 (7.10) yes Child mortality (GLS-RE) (4) 150.481** (21.41) -0.529** (3.02) -0.291 (1.42) 0.173 (1.45) -0.078 (0.81) 0.223 (0.93) 0.021 (0.17) 0.070 (0.49) 27.51* (1.85) -30.53** (12.93)

Pseudo- R2 Wald - X2 (p-value) Sample size

OLS = ordinary least squares; GLS = generalizedleast squares; 2SLS = two-stageleast squares; RE = random FE effects; = fixedeffects at at **significant 5 percentlevel *significant 10 percentlevel in NOTES: Absolutet-ratios parentheses. All standarderrors robust.The Wald (X2)testis a testof thejoint are (except the constantterm)equal zero. hypothesisthatall coefficients

54

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

howThe of literacy. reduction, also mortality reducesthecoefficient female suggesting thatthe bulk of significant), ever,is small (and not statistically is rather thanan indion effect, of theeffect femaleliteracy fertilitya direct mediated childmortality.46 recteffect by are unchanged.The regionaldummies The othercoefficients largely The main in (reported column4 ofTable 5) are much the same as before. for leads to a substantial naris difference thatcontrolling childmortality and the default reregion gap rowingofthe fertility betweenthe northern childmortality rates gion: about halfof thisgap is accountedforby higher is difference thatthe east dummycrosses in the northern region.Another of significance. the threshold statistical on The first-stage (Table 6, column4) is ofinterest itsown. regression resign. None of the variableshas a counterintuitive As withthe fertility suchas thepoverty of indicators development index,male gressions, general have little and urbanization power. Far more imporexplanatory literacy, childhealth-in thiscase tantare variablesthathave a direct bearingon and of water. son femaleliteracy, preference, the availability safedrinking wateris largeand highly of improved significant: The coefficient drinking of wateralone accountsformorethanone-fifth the deaccess to drinking between1981 and 1991. in childmortality cline in withinformaUsingthe coefficients column3 ofTable 6, combined assess the contribution tionon samplemeans (Table 2), we can tentatively decline.As notedearlier, TFR declinedby 0.7 factors fertility to ofdifferent to estimates. Taken between 1981 and 1991 according census-based births and of the together, improvement femaleliteracy the declineofchildmorOtherregression variables make a accountfora declineof0.35 births. tality to contribution fertility decline,so thatthebalance of0.35 births negligible In is accountedforby the timedummy. otherwords,halfofthe declinein in between 1981 and 1991 is accountedforby improvements fefertility male educationand childsurvival.

remarks Concluding
of The findings thisarticleconsolidateearlierevidenceon the connection in studies betweenfemaleeducationand fertility India. Whilemanyearlier association betweenthetwo,thepresent analysis have noteda closebivariate in observation severalrespects. First, our takesus beyondthatwell-known multivariate approachhelps to establishthatthe connectionbetween feis In we male educationand fertility robust. all the specifications have exin femaleliteracy has plored(onlya fewofwhichare reported thisarticle), rate. effect the fertility The size of this on a negativeand highly significant of too, coefficient, is quite stable. Second, the robustness this coefficient linkbetweenfemaleeducationand ferthatitis driven a direct by suggests rather thanby a jointinfluence unobserved of variableson both.The tility,

JEAN DRtZE

/ MAMTA

MURTHI

55

in enhancesthe substantially of effects, particular, inclusion district-specific has The of plausibility thisinterpretation. factthatfemaleliteracy a strong and for even controlling male literacy, thatthe latimpacton fertility after of to literacy, also independently female nothing low fertility tercontributes in variables, manycases,are (sinceunobserved pointsin thesame direction of male as well as femaleeducation).Third,the effect likelyto influence in appearsto be quitelarge:the estimates colfemaleeducationon fertility from its thatan increasein adult femaleliteracy umn 3 ofTable 4 suggest would reducethe total in base level of22 percent 1981 to,say,65 percent conrate fertility by about one childper woman. This is quite substantial, the 1998 TFR in Infor sidering instancethatthe gap separating estimated rate of 2.1 is of a similarorderof dia (about 3.2) fromthe replacement on to Of magnitude. course,it would be misguided relyexclusively an into level (and the preceding creasein femaleliteracy reachthe replacement aboutcausality and linearexerciseinvolves simulation strong assumptions illustrate potentially the our role far-reaching ity).Nevertheless, calculations decline. ofwomen's educationin fertility the The analysispresentedin thisarticlealso highlights role of son levels.Ifparents value sonsand daughin sustaining highfertility preference with,say,two surviving tersmoreor less equally,so thattheyare satisfied for is of births corchildren irrespective theirsex, the incentive additional as the lower. Furthermore, we discussedearlier, positiveasrespondingly and is to sociationfoundherebetweenfertility son preference difficult recdeclinein India is a cause of intensified oncile withthe view thatfertility to suggest Contrary thatview, our findings genderbias in childmortality. of bias are complementary declineand reduction gender goals. thatfertility we thatchildmortality strongly is In linewithearlier research, also find We to associatedwithhighfertility. have made a tentative attempt estimate on of effects. Taken topurgedoffeedback the effect the former thelatter, childsurvival and rising femaleliteracy accountforabout gether, improved between 1981 and 1991. The remainder, rehalfof the declinein fertility in involvesa shift the structural relation bein flected our "time"dummy, variablesincludedin our analysis.Posand tweenfertility the explanatory includeinterdistrict diffusion the of effects, sible interpretations thatshift and planningprograms, the delayedimpactof earlier expansionof family in improvements economicand socialconditions. effects femaleliteracy, of childmortality, son and Finally,the strong withthetenuouscorrelation on between levelscontrast preference fertility and variousindicators overalldevelopment modernization of and thelatter and None of thesevarisuch as male literacy, urbanization, even poverty. influence fertility. on ables exertsa statistically decline Fertility significant is notjust a byproduct economicgrowth: dependson improvements of it in conditions thatare conduciveto changedfertility and that the specific goals to help parents realizethesegoals.

56

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

Notes
(1995), Murthi, Guio, and Dreze to The authorsare grateful Sonia Balhotra, Jejeebhoy of Cassen, (1995),Government India(1997), Gandotra Alaka Basu, P. N. Mari Bhat,Robert et Angus Deaton, Tim Dyson, Roger Jeffery, et al. (1998), Parasuraman al. (1999), and Sen, Arokiasamy Klasen,Amartya ReetikaKhera,Stefan (2000a). and participants S. V. Subramanian, seminar 5 We hope thattheuse ofpaneldatawill com- helpto address and Harvardforhelpful at Cambridge earlier doubts abouttheuse of ments and suggestions. Mamta Murthi's cross-section analysis thiscontext: in "Popuwork was supportedby the John D. and lations through live time. a disproportionBut Foundation'snet- ate shareofresearch [SouthAsia's]demoCatherineT. MacArthur on in workon Inequalityand Poverty Broader graphicevolutionhas reliedupon the dual Perspectives. blunderbusses cross-sectional of censusesand of (Dyson1999:2). outbursts alarmism is surveys" 1 Amongrecent Brownand BrianHalweil'swidely pubLester 6 Forfurther discussion, Murthi, see Guio, on 1 article "Indiareaching billion Au- and Dreze (1995) and Dreze and Murthi licized planned"(International(2000). gust15:No celebration Herald Tribune,11 August 1999). The authors 7 See, e.g., InternationalInstitutefor into a demowarn that India "risksfalling PopulationSciences (1995: Table 7.8). The will darkhole,one wherepopulation graphic evidencewe are aware of apare beginto slowbecausedeathrates rising"- onlycontrary no (1996), who observes difthat and pearsin Vlassoff a boldprediction, considering fertility ference desired in fertility amongwomenwith fallrates mortality inIndiahavebeensteadily levels in different ofschooling rural Maharashyear. ingyearafter thisis in a context where detra.However, states the siredfertility fallen that 2 Singh instance, (1999),for had below threebirths at "is problem to all levelsofeducation the community to onlysolution thepopulation in suras havingmorethantwo children an veyed. In earlierwork in the same region, classify in to Vlassoff resulting losingthe right act of sedition (1980) founda clearassociation beOn adultfranchise." an evensterner tween universal number children theeduof and desired that plan- cationlevelsofadolescent Sahay(2000)argues "ourfamily note, girls. be must ning programmes either moreeffective 8 The evidenceis fairly strong. India For so the altogether that Malthusian orabandoned see specifically, Jain (1985), Nag (1989), the genwill irresponsible calamities strike present Beenstockand Sturdy(1990), Bourne and ones." instead thefuture of eration Walker(1991), Satia and Jejeebhoy (1991), methodsin India were Basu (1992), International 3 Authoritarian Institute Popufor extremeduringthe lation Sciences (1995), Murthi,Guio, and taken to a short-lived ster- Dreze (1995), Govindasamyand Ramesh (1975-77),whencompulsory Emergency had (1997), Jeffery Jeffery Birth which werecommon. rates, ilizations and (1997), Bhargava to stagnated (1998), and Pandeyet al. (1998). prior theEmergency, been falling before resumfor ten thereafter, roughly years 9 See, e.g., LeVine (1980), Cleland and decline. See, e.g.,Bose (1991: 67-68) ingtheir Wilson (1987), Lindenbaum (1990), and of Government India (1999: 3). and World Bank (1991). 4 On the internationalevidence, see 10 See various contributionsJeffery in and Bulatao and Lee (1983), Clelandand Wilson and Jeffery Jeffery Nations(1987), Subbaraoand Basu (1996a), particularly (1987), United (1996),andVlassoff (1996). For (1997),amongoth- (1996),Visaria Raney(1995),and Schultz see perspective, The PROBE Team to ers.For studies relating India,see Jainand a different Nag (1985, 1986), Sharma and Retherford (1999). 11 Numerousindicators female auof (1990), Satia and Jejeebhoy(1991), Basu (1992), UnitedNations(1993), International tonomy havebeenconsidered theliterature. in Institutefor Population Sciences (1995), Examplesare whether woman is consulted a

JEAN

DREZE

/ MAMTA

MURTHI

57

about the choice of her marriagepartner, a marfamily after she whether forms nuclear joint riageor becomespartof her husband's over a porwhethershe has control family, she is whether tion of householdresources, and decisions, budgeting free makecertain to home. she her how often visits parental

19 For an insightful reviewof the literatureon the demographic transition, Kirk see (1996). 20 The possible endogeneity son prefof erence(captured herebythe ratiooffemaleto-malechildmortality) discussed Dreze is in and Murthi (2000), whereitis shownthat(1) effects themainconclusions this of discussions income 12 Foruseful of article notsenare and demandframework, re- sitive different in thehousehold to treatments the son-prefof views of the empiricalevidence,see Hotz, erenceindicator (including two-stage estimaKlerman, and Willis (1997) and Schultz tion),and (2) the hypothesis son preferthat ence is exogenousis difficult reject, to even if (1997). I we allowa largeprobability type- error. of 13 Rosenzweig and Evenson (1977) someevidenceofthiseffect, refwith present 21 The census-basedfertility estimates see to erence landownership; alsoCain(1985), used hereare systematically higher thanthe (1992), and corresponding estimates this Nagarajan and Krishnamoorty SRS (on issue, see cited Saavala (1996) and theliterature there. also Bhat 1998). However, the highcorreas the between on child lationbetweenthe two seriesindicates, 14 Evidence therelation of seemstobe fairly uniin and survival accessto healthservices India extent overestimation Uniform overestimation not matdoes is limited. Bhargava(1998), usingNFHSdata form. our also finds thatfamily plan- termuchfor purposes. Pradesh, forUttar ning and immunizationprogramshave a 22 Foran innovative to attempt estimate survival. theother community On effect infant on strong effects based on householddata, hand, a recent study based on the NFHS see Munshi(2000). of no (WorldBank 1998) finds effect public 23 Thisis one plausible reasonwhyfertilon healthservices childmortality. ityratesamong uneducatedwomen in, say, of discussions thisissue, Kerala are much lowerthan among unedu15 Forinsightful and (1997: ch. 6), Jeffery catedwomeninUttar see Jeffery Jeffery Pradesh Bihar; this or on and Jeffery (2000), and Iyer(2001). see et contrast, Parasuraman al. (1999: Table of 16 For detailedsimulations thistype, 8.1,p. 115). based on all-Indiadata, see May and Heer case 24 In theextreme whereinterhousebetweenson prefer- holdfertility (1968). On the relation a differences within particular loet see enceand fertility, also Mutharayappa al. cality smallbecause ofthepowerful are hold Choe,and Roy(1998) and ofsocialnorms, (1997) and Arnold, household-based' inanalyses Bothstudies suggest volving theliterature citedthere. confined datamight to unfail locally effect fer- cover significant ofeducation feron has that preference a positive son effect on any tility. even if social normsthemselvesare tility, in by betweenfertility and affected thelevelofeducation thelocal17 The association Thisis one interpretation the factthat of laborforce in female participation Indiais dis- ity. basedon comparisons becussedin Murthi, Guio,and Dreze(1995). See somemicro-studies tween householdswithina specific locality also Drezeand Sen (1995). and 1997) findno eviis 18 High fertility associatedwithshort (e.g., Jeffery Jeffery denceofa majoreffect women'seducation of and withbearingchildren at birthintervals withanalyses in basedon or relatively young oldages,bothofwhichin- on fertility,contrast district-level variations fertility in rates(e.g., see of creasetherisk infant mortality; Wolpin Guio, and Dreze 1995) or on largeIf (1997) fora reviewof thisliterature. high Murthi, is by fertilitymotivated the desireforsons,it scale household surveys(e.g., Arokiasamy discussion this of may go hand in hand with high mortality 2000a, 2000b). For further issue,see Sen (1999: 218-219). See girls. Das Gupta(1987) amongunwanted forevidenceof highmortality of amonggirls in birth orders Punjab. higher 25 One referee askedwhether 15+ is the relevant groupin thiscontext. thecase In age

58

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

the for of femaleeducation, instance, repro- tween1972-73and 1987-88,and for1991we between1987-88and 1993-94. age ductive group(usually15-49) mayappear interpolate On to be morerelevant. the otherhand,lit30 Thestandard errors-in-variables result are age rates different groups so highly is thatifone variable measured in eracy is witherror, choice the OLS estimator its coefficient biased that across districts theprecise correlated of is for to is group unlikely matter our toward ofreference zero (see,e.g.,Greene1993). Ifa variof giventheimportance in- able is replacedby groupmeans (which inFurther, purposes. the decisions, cor- volves a measurement effects fertility in terpersonal errorwith a differobvious. ence-viz., the sum of errorsis identically rectchoiceof age groupis farfrom this Pursuing questionwould requirea finer zero),theOLS estimator unbiased is and conrates. A sistent, literacy basedon age-specific analysis, althoughthereis some loss of effiof limitation theuse ofadultlit- ciency.Estimates othercoefficients, moreserious of howis it of as eracy an indicator education that fails ever,maybe biasedifthemeasurement error of to takeintoaccountthe quality education is correlated thecorresponding with variables. Yet,as we show in this among the literates. 31 Thereare minordifferences between of crudeindicator educational eventhis article, table (Table thistableand the corresponding interdistrict well levelsperforms in predicting 2) in Murthi, Guio,and Dreze (1995), mainly rates. in variations fertility smallchangesin thelistofdistricts. reflecting line is the widelyused In thecase ofliteracy 26 The poverty the rerates, differences benchmark proposedby Dandekarand Rath flect choicesof reference group different age (1971): Rs 15 per capitaper monthat 1960- (15+ versusall ages). We take thisopportuon pov- nity correct typographic The for 61 prices. reason focusing rural a to error the earin for liertable:theTFR in Punjabin 1981 was 5.0, estimates poverty erty thatthe relevant is are rural and urbanareascombined notavail- not3.26. for able. Sincewe control thelevelofurbanin 32 Therewere326 districts the 14 maamongdis- jorstates 1981,rising 362 districts 1991 this ization, is nota majorconcern: to in in rural owingto thepartitioning districts levelsofurbanization, withsimilar tricts of between and ur- thetwocensuses(see Murthi, (for poverty rural and poverty overall and Srinivasan, ban areas combined)are likelyto be highly Subramanian 1999 fordetails).To construct correlated. the panel for1981-91,we use the 1981 disvari- trict a the dummy 27 We introducedseparate map as an anchorand calculate 1991 turned out value ofa variable a particular for district a as able forBihar,whichconsistently for overtherelevant to average patterns the population-weighted not to conform predicted that was or The (in regions. fact Bihar 1991 districts cases wherethedistrict northern eastern of between1981 and 1991). seemsto standin a category itsown vis-A- partitioned it indicators particular,has (in visdemographic 33 This is formally equivalentto taking low rates) has been first surprisingly mortality differences, relating i.e., changesin fertilnoted earlier;see, e.g., Dyson and Moore ity between1981and 1991tothecorrespond(1983) and Dyson(1984). in variables. ingchanges theexplanatory is but term region, we re28 Thestandard 34 This coefficient anotherinterprehas and term other for that serve purposes, use the tation, the asthatitcaptures positive namely hereto avoidconfusion. sociation division ad hoc term betweengenderinequality genin used here are the same as the eral (ofwhichson preference one manifesThe divisions is used in Murthi, Guio,and Dreze tation)and fertility. NSS regions There are indeed many (1996). (1995) and Drezeand Srinivasan ways in whichgenderinequality may interwithfertility decline. are fere estimates 29 Division-level poverty 35 Theremaybe some positive feedback available frompublished sources only for to as insofar highfertilfertility poverty, 1963-64, 1972-73, 1973-74, 1987-88, and from and to Sundaram, Tendulkar ityratescontribute highlevelsof poverty 1993-94(see Jain, when thelatter measured is 1989; Dreze and (especially with1988; Pal and Bhattacharya intoaccounteconomiesofscale in Srinivasan 1996; and Dubey and Gango- out taking be- household This to padhyay1998). For 1981,we interpolate consumption). is likely lead

JEAN

DREZE

/ MAMTA

MURTHI

59

coefficient ents in couldnotafford dispense to with children's to an upwardbiasin thepoverty The thatthepov- laborpower.The parents question,howin regressions. fact thefertility tooktofamily is in planning ducks walike to erty coefficientnegative spiteofthispos- ever, doubton the tersoon after publication Mamdani's the of sible upwardbias castsfurther per that hypothesis poverty se isa causeofhigh study. fertility. 40 Relevant evidence on this includes data Vlassoff of (1979), Cain (1986), Jejeebhoy and 36 Our own analysis district-level as- Kulkarni is (1989), S55v5la(1996), Jeffery and that confirms son preference negatively of (1997), Dreze and Sharma (1998), sociated with the proportion scheduled Jeffery see tribes thepopulation; Drezeand Murthi Lieten(2000) amongothers. in 1). (2000: Appendix 41 Das Guptaand Bhat(1997: 313). index the 37 Notethat son-preference has 42 Das Guptaand Bhat'sstudy raisesthe regres- possibility endogeneity son preference the to be excludedfrom fixed-effects of of in sion; thisis becausetheindex,beingtime-in- the fertility regressions. Note,however,that of independent thedis- theirargument is variant, notlinearly a feedback suggests negative ease ofexposition we from dummies. Also,for trict to this fertilityson preference; reinforces across thefinding son preference a positive maintainthe equalityof coefficients that has wherewe allow effect fertility estimates, years.Alternative on feedback (sincenegative sugare to years, avail- gests downward inthecoefficient son thecoefficients varyacross a bias of on Theydo not preference). possibleendogeneity son the able from authors request. The of of alter mainarguments thearticle. the is in preference addressed Dreze and Murthi 1). 38 We note in passingthatthe bivariate (2000: Appendix and fertility is between poverty correlation 43 The variable measures proportion the in significant 1991. of householdswithaccess to drinking positiveand statistically water is of This contrast a usefulreminder the po- supplied a from tap,handpump,or tubewell. natureofbivariate corremisleading tentially 44 On two-stage leastsquaresestimation The mainreasonwhy lationsin thiscontext. withpaneldata,see Hsiao (1986: ch. 5). have higher rates fertility seems poordistricts 45 Thetest, designed Hausman(1978), by also to be thatthesame districts have low litto the is similar the testforrandomversusfixed for controlling thatfact, eracyrates;after earlier. discussed betweenpoverty fer- effects and association positive holds. no tility longer 46 Parasuramanet al. (1999) reporta based on National Family of theme Mamdani's similarfinding, 39 Thiswas a central of of debunking the"myth popu- Health Survey data. By contrast,Schultz (1973) classic on lation control."Mamdani argued that the (1997) estimates, thebasisofinternational half of withprovision family data,thatroughly ofthetotaleffect feof Khanna experiment on via operates child services ruralPunjab in the late male education fertility in planning becausepar- mortality. 1950sand early1960shad failed

References
the process Arnold, M. K. Choe, and T. K. Roy. 1998. "Son preference, family-building F., Studies 301-315. 52: and childmortality India,"Population in use P. in contraceptive and fertility India,"mimeo, Arokiasamy, 2000a. "Genderpreference, International Institute PopulationSciences,Mumbai. for in 2000b. "Genderbias in childhealthcare and excess femalechildmortality India: householdeconomicstatusand women's position," The roles of culture, mimeo,International Institute PopulationSciences,Mumbai. for the and Behaviour. Oxford: Clarendon. Basu, A. M. 1992. Culture, Status Women, Demographic of of in mortality regionalInBeenstock,M. and P. Sturdy.1990. "The determinants infant 18: dia," World Development 443-45 3.

60

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

Evidence and Bhargava,A. 1998. "Familyplanning,genderdifferences infantmortality: of of UttarPradesh(India)," mimeo,Department Economics,University Housfrom ton. declinein India: An analysisof district-level Bhat,P. N. Mari. 1998. "Contoursof fertility two recent M. censuses,"in G. Martine, Das Gupta,and L. C. Chen (eds.), trends from Press. in Delhi: Oxford University Change Indiaand Brazil. Reproductive DiversityIndia.Delhi: B. R. Publishing. of Bose, A. 1991. Demographic of educationon Jr. effect mothers' Bourne,K. L. and G. M. Walker, 1991. "The differential 45: in Studies 203-219. mortality boysand girls India,"Population of determinants developing in Bulatao, R. A. and R. D. Lee. 1983. "An overviewof fertility in of in countries," R. A. Bulatao and R. D. Lee (eds.), DeterminantsFertility Developing New Countries. York:AcademicPress. and fertility," betweenlandholding Studies Population Cain, M. 1985. "On the relationship 39: 5-15. and failure: Dependence,mobility, mor. 1986. "The consequencesof reproductive Studies 375-388. of 40: amongthe elderly ruralSouthAsia,"Population tality An of transition: iconoclasCleland,J.and C. Wilson. 1987. 'Demand theories the fertility Studies 5-30. 41: ticview,"Population in Dandekar,V. M. and N. Rath. 1971. Poverty India.Bombay:SameekshaTrust. againstfemalechildrenin ruralPunjab," Das Gupta,M. 1987. "Selectivediscrimination Review 77-100. 13: and Population Development genderbias in India: A consequence Das Gupta,M. and P. N. MariBhat. 1995. "Intensified CenterforPopulationand offertility 95.02, Harvard PaperNumber decline,"Working HarvardUniversity. DevelopmentStudies, of declineand increased manifestation sex bias in India,"Population . 1997. "Fertility Studies 307-315. 51: one is no fun,"Times India,12 June. of Dreze,J. 1998. "Two is company, educationand development: Further evidence Dreze, J. and M. Murthi.2000. "Fertility, fromIndia," DevelopmentEconomicsDiscussionPaper No. 20, STICERD, London School ofEconomics. and Oxford:OxJ. Development SocialOpportunity. DrZeze, and A. Sen. 1995. India:Economic Press. ford University in economy, society," P. Lanjouw Dreze,J.and N. K. Sharma.1998. "Palanpur:Population, in Development PalanpuroverFiveDecades.Oxford: and N. H. Stern (eds.), Economic Clarendon. in 1987-88," DisDreze, J.and P. V. Srinivasan.1996. "Poverty India: Regionalestimates, EconomicsResearchProgramme, STICERD,LoncussionPaperNo. 70, Development Ecoin don School of Economics;revisedversionforthcoming Journal Quantitative of the 1998. "Counting poor," SarvekshanaAnalytical ReDubey, A. and S. Gangopadhyay. of Government India,New Delhi. of portNumber1, Department Statistics, World HealthStatistics 37: Quarterly 200Dyson, T. 1984. "India's regionaldemography," 231. 1999. "Birthrate trendsin India, Sri Lanka, Bangladesh and Pakistan,"mimeo, in London School of Economics;forthcoming J. F. Phillipsand Z. A. Sathar (eds.), in Asia.Oxford: on Transition South Clarendon. PerspectivesFertility Comparative femaleautonomy, and demographic structure, Dyson,T. and M. Moore. 1983. "On kinship and Review 35-60. 9: behaviorin India,"Population Development New York:Ballantine. P. Bomb. Ehrlich, R. 1968. ThePopulation in Number9, InternaM. Gandotra, M. et al. 1998. "Fertility India,"NFHS SubjectReports tionalInstitute PopulationSciences,Mumbai. for H. secondedition. NJ: William 1993. Econometric EnglewoodCliffs, PrenticeAnalysis, Greene, Hall.
nomics.

JEAN DREZE

/ MAMTA

MURTHI

61

in of Government India. 1988. "Fertility India: An analysisof 1981 census data," Occaof sional Paper 13 of 1988, Demography Division,Office the Registrar General,New Delhi. . 1989. "CensusofIndia 1981: Householdliteracy, drinking water, electricity toilet and facilities," Occasional PaperNo. 1 of 1989,Office theRegistrar of General, New Delhi. . 1994. "Census of India 1991, housingand amenities:A database on housingand townsand cities," amenities districts, for OccasionalPaperNo. 5 of 1994, Office the of Registrar General,New Delhi. level estimates fertility child mortality 1991 and their of and for . 1997. "District with OccasionalPaperNo. 1 of 1997, Office theReginterrelations othervariables," of istrar General,New Delhi. . 1999. Compendium India'sFertility Mortality of and Indicators 1971-1997.New Delhi: General. of Office theRegistrar P. of Govindasamy, and B. M. Ramesh. 1997. "Maternaleducationand the utilization maternaland childhealthservicesin India,"NFHS SubjectReportsNumber5, Internafor tionalInstitute PopulationSciences,Mumbai. testsin econometrics," Econometrica 1251-1271. 49: Hausman,J.A.. 1978. "Specification in and R. J.Willis.1997. "The economicsof fertility developed Hotz, V. J.,J. A. Klerman, in countries," Rosenzweigand Stark1997. Press. Hsiao, C. 1986. Analysis PanelData. Cambridge: of Cambridge University HealthSurvey: for India Institute PopulationSciences. 1995. National International Family 1992-93.Bombay:IIPS. and in India. India: OxfordUniversity Iyer,S. 2001. Religion theEconomics Fertility South of Press,forthcoming. of in in Jain,A. K. 1985. "Determinants regionalvariations infant mortality ruralIndia," 39: Studies 407-424. Population educationand fertility in reduction India," Jain,A. K. and M. Nag. 1985. "Femaleprimary Working PaperNo. 114, CenterforPolicyStudies,PopulationCouncil,New York. of educationforfertility reductionin In. 1986. "The importance femaleprimary and 6 dia," Economic Political Weekly,September. Jain,L. R., K. Sundaram,and S. D. Tendulkar.1988. "Dimensionsof ruralpoverty:An and Economic Political November. inter-regional profile," Weekly, SpecialNumber, 1996. "What'sthebenefit beingeducated:Women'sautonomy of P. Jeffery, and R. Jeffery. outcomesin Bijnor,"in Jeffery Basu 1996a. and fertility and and fertility India,"Economic Political in and . 2000. "Religion Weekly 26 August-2 34, September. Women's R. and Jeffery, and A. M. Basu (eds.). 1996a. Girls'Schooling, Autonomy Fertility in Asia.New Delhi: Sage. Change South in and .1996b. "Schoolingas contraception?," Jeffery Basu 1996a. Gender Politics: and R. 1997. Population, in Jeffery, and P. Jeffery. Demographic Change Rural Press. North India.Cambridge: Cambridge University S. and Behaviour: Jejeebhoy, J. 1995. Women's Education, Autonomy, Reproductive Experience Countries. Oxford: Clarendon. from Developing and reproductive S. Jejeebhoy, J.and S. Kulkarni.1989. "Demand forchildren motivation: observations from in Empirical Maharashtra," S. N. Singhet al. (eds.), Population Transition India.New Delhi: B. R. Publishing. in transition Studies 361-387. 50: Kirk,D. 1996. "Demographic theory," Population in births: Decline of childmortality a UP village," Kolenda, P. 1998. "Fewerdeaths,fewer Manushi105: 5-13. of LeVine,R. A. 1980. "Influences women's schoolingon maternal behaviorin the Third Education Review 78-105. 24: World,"Comparative workand education,"Economic Political and Lieten,G. K. 2000. "Children, Weekly, June 10 and 17 June.

62

FERTILITY,

EDUCATION,

AND

DEVELOPMENT

IN INDIA

in of Lindenbaum,S. 1990. "The educationofwomen and the mortality children Banglain desh," in A. C. Swedlund and G. J. Armelagos(eds.), Diseasein Populations Transiand Perspectives. York:Bergin& Garvey. New tion: Anthropological Epidemiological and mortality genderbias among tribalpopulation:An Maharatna,A. 1998a. "Fertility, Indian perspective," WorkingPaper Number98.08, HarvardCenterforPopulation Harvard University. and DevelopmentStudies, in and earlytwentieth India," . 1998b. "On tribalfertility late nineteenth century Center Populationand DevelopmentStudfor Working PaperNumber 98.01, Harvard ies, Harvard University. Control: Caste Classin an IndianVillage. and Family, Mamdani,M. 1973. TheMyth Population of ReviewPress. London: Monthly motivation family in India: A and size May, D. A. and D. M. Heer. 1968. "Son survivorship 22: Studies 199-210. computer simulation," Population Munshi,K. 2000. "Social change and individualdecisions:Withan applicationto the deof mographic transition," mimeo,University Pennsylvania. and fertility, genderbias in India: A Murthi, A-C. Guio,and J.Dreze. 1995. "Mortality, M., and 21: Review 745-782. district-level analysis," Population Development and S. V. Subramanian.1999. "Linking the Indian Census Murthi,M., P. V. Srinivasan, and Economics, King's mimeo,CentreforHistory withthe NationalSample Survey," College, Cambridge. Also available fromhttp://www.kings.cam.ac.uk/histecon/ linking.htm. and its Mutharayappa, M. K. Choe, F. Arnold,and T. K. Roy. 1997. "Son preference R., in on Institute effect fertility India," NFHS SubjectReportsNumber3, International forPopulationSciences,Mumbai. awarenessas a factor accessibility healthservices: case study in of A Nag,M. 1989. "Political and 25 ofruralKerala and WestBengal,"Economic Political Weekly, February. and 1992. "Landholding fertility in relationship a lowNagarajan,R. and S. Krishnamoorty. in community TamilNadu," in A. Bose, and M. K. Premi(eds.), fertility agricultural Asia.Delhi: B. R. Publishing. Transition South in Population 1989. "On areal distribution poverty ruralIndia during of in Pal, P. and N. Bhattacharya. SeriesB, Part1, 51: 225-262. 1973-74," Sankhya and in Number Pandey,A. et al. 1998. "Infant childmortality India,"NFHS SubjectReports Institute PopulationSciences,Mumbai. for 11,International in Education Shaping in S. Fertility India:Evidences Parasuraman, et al. 1999. RoleofWomen's Health Mumbai:Himalaya. Survey. from National Family and the economiccontribuM. schooling, Rosenzweig, R. and R. Evenson. 1977. "Fertility, Econometrica 1065-1079. in 45: tionofchildren ruralIndia: An econometric analysis," and Economics. of Rosenzweig,M. R. and 0. Stark(eds.). 1997. Handbook Population Family Elsevier. Amsterdam: of Saavala, M. 1996. "A childis a gift God but how will we nourishhim? Rural fertility decline in coastal AndhraPradesh,South India," mimeo,CentreforAsian Studies, Amsterdam. Hindustan 11 Times, January. Sahay,K. B. 2000. "A Malthusianprospect," A of Challenge: Study FourLarge Satia,J.K. and S. J.Jejeebhoy (eds.). 1991. TheDemographic Delhi: Oxford Press. IndianStates. University T. Reading,MA: Addison-Wesley. Schultz, P. 1981. Economics Population. of declinein moderneconomicgrowth:Is aggregate evi. 1994. "Sources of fertility transition credible?" dence on demographic mimeo,Yale University. in . 1997. "Demand forchildrenin low income countries," Rosenzweigand Stark 1997. as 1999. DevelopmentFreedom. New York:Alfred Knopf. A. Sen, Amartya. of 1990. "Effect femaleliteracy fertility India," on in Sharma,0. P. and R. D. Retherford. of OccasionalPaper 1 of 1990, Office theRegistrar New Delhi. General,

JEAN DREZE

/ MAMTA

MURTHI

63

not Times India,12 OctoSingh,D. 1999. "Treason, reason: Six billionand stillgrowing," of ber. Sopher,D. 1980. AnExploration India.London:Longman. of Subbarao, K. and L. Raney. 1995. "Social gains fromfemaleeducation:A cross-national and Economic 44: study," Development Cultural Change 105-128. in UniverThe PROBE Team. 1999. Public Report BasicEducation India.New Delhi: Oxford on sityPress. in of Evidence from World the UnitedNations. 1987. Fertility Behavior theContext Development: New York. Fertility Surveys. . 1993. Women's A of Maharashtra. New Education Fertility and Behavior: CaseStudy Rural York. in and Visaria,L. 1996. 'Regional variations femaleautonomyand fertility contraception and in India,"in Jeffery Basu 1996a. C. adolescentfemalesin ruralIndia: A studyof the social imVlassoff, 1980. "Unmarried Journal Marriage the and Family 427-436. 42: pact ofeducation," of . 1996. "Againstthe odds: The changingimpactof schoolingon femaleautonomy in and and fertility an Indianvillage," Jeffery Basu 1996a. in of A Vlassoff, 1979. "Labourdemandand economicutility children: case studyin rural M. Studies 415-428. 33: India,"Population and and Wolpin,K. 1997. "Determinants consequencesof the mortality healthof infants in and children," Rosenzweigand Stark1997. 1991:TheChallenge Development. Oxford and WorldBank. 1991. World of Development Report Press. New York:Oxford University in Public Services. . 1998. Reducing for Washington, Poverty India:Options MoreEffective DC: WorldBank. Transition: Determinants Zachariah,K. C. and S. I. Rajan (eds.). 1997. Kerala'sDemographic New and Consequences. Delhi: Sage.

You might also like