Household Shocks, Child Labor, And Child Schooling

8/3/2019 Household Shocks, Child Labor, And Child Schooling

1/23

R E S E A R C H R E PO R T S A N D N O T E S

HOUSEHOLD SHOCKS, CHILD LABOR,AND CHILD SCHOOLING

Evidence from Guatemala

William F. VsquezFairfield UniversityAlok K. Bohara

University of New Mexico

Abstract: Using data from the National Survey of Standards of Living conductedin Guatemala in 2000, this article tests the hypothesis that Guatemalan householdsuse child labor and reduce child schooling to cope with household shocks. First, theauthors use factor analysis to estimate the latent household propensity to naturaldisasters and socioeconomic shocks. Then, they estimate bivariate probit models toidentify the determinants of child labor and schooling, including household propen-sity to natural disasters and socioeconomic shocks. Results suggest that householdsuse child labor to cope with n atural disasters and socioeconomic shocks. In contrast,the authors found tw evidence that suggests that households reduce child schoolingto cope with shocks. Findings also indicate that poor households are more likely touse child labor and schooling reduction as strategies to cope with socioeconomicshocks.

INTRODUCTIONChild labor affects the current and future welfare of children. In the


2/23

i 6 6 Latin Am erican Research Reviewalleviation, disease reduction, and fertility choices (Glewwe 2002; WorldBank 2005), child labor may have a negative impac t on the welfare of chil-dre n in the long run.'Child labor has been traditionally considered a household responseto incom e pove rty (Amin, Quayes, and Rives 2004; Jensen a nd Nielsen1997; Ray 2002) and labor market imperfections (Dumas 2007). A morerecent strand of literature suggests that child labor is a household strategyto cope with natural disasters and socioeconomic shocks (Dendir 2007;Guarcello, Mealli, and Rosati 2003). Households may also reduce invest-ments in a child's schooling to cope with household shocks (Jacoby andSkoufias 1997). Poor househo lds are m ore likely to use child labor as a cop-ing strategy given their lack of assets that could be used to recover fromhousehold shocks (Beegle, Dehejia, and Gatti 2006). Thus, the negative ef-fect of such cop ing strategies on the cu rren t and future welfare of child renis mo re significant am ong poor ho useho lds.

This study tests the hypothesis that Guatemalan households use childlabor and reduce child schooling to cope w ith natu ral disas ters and socio-economic shocks. We used factor analysis to estimate the latent househo ldpropensity to na tural disas ters and socioeconomic shocks based on infor-m ation on hou sehold shocks included in the 2000 National Survey of Liv-ing Standards (Encuesta Nacional de Condiciones de Vida, or ENCOVI).This approach provides a stronger analytical framework to estimate shockindices than traditional methods such as binary variables and the totalamount of reported shocks that do not account for measurement errorsand restrict different household shocks to have similar effects. We thenincluded estimated propensities in bivariate probit models to investigatethe effect of natural disasters and socioeconomic shocks on child laborand schooling. We also analyzed differences in responses to householdshocks among nonpoor, poor, and extremely poor. Results indicate thathouseholds tend to use child labor to cope with natural disasters and so-cioeconomic shocks. In contrast, we found no evidence that householdsreduce child schooling to cope with shocks. Findings also indicate thatpoor households are more likely to use child labor and schooling reduc-tion as strategies to cope w ith socioeconomic shocks.We have organized the rest of the article as follows: we first describethe back ground, then prese nt the theoretical framework, describe the dataand variables used in the study, present the econometric methodology,presen t the results, and finally p rovide conclusions.


3/23

HOUSEHOLD SHOC KS, CHILD LABOR, AND CHILD SCHOOLING 167Table 1 Child Labor and Nonenrollment Ratios, byRegion, Gender, and Area

RegionMetropolitanNorthNortheastNorthwestCentralSoutheastSouthwestPetenCountry

Child LaborBoys

Rura l31.540.225.832.634.631.830.133.032.9

U r b a n13.015.414.718.621.816.323.723.618.0

GirlsRural

17.718.26.9

14.817.98.1

18.38.614.8

Urban9.29.69.6

11.922.711.912.013.012.4

NonenrollmentBoysRural

37.035.929.932.928.924.622.531.729.4

U r b a n7 3

18.610.817.714.810.116.718.714.0

GirlsRura l

29.146.131.244.531.829.333.735.336.7

Urban10.623.616.123.017.914.720.115.917.5

Note: We calculated these rates using the 2000 Encuesta Nacional deCondiciones de Vida(ENCOVI). The rates are ex pressed in percentages of children betw een five and sixteenyears of age.BACKGROUND

The incidence of child labor has gradua lly increased in Guatemala andis among the highest in Latin America (Guarcello et al. 2003). The Inter-national Labour Organization's International Programme for the Elimi-nation of Child Labour (IPEC 2003) reports that 7.9 percent of childrenbetween seven and fourteen years of age worked for a salary or at home in1994. This percentage increased to 23.5 percen t in 2002. The average nu m -ber of working hours per week was greater than thirty-five in all kindsof activities. Such large amounts of working ho urs may reduce the timeavailable for child schooling. In addition, a significant num ber of childrenwork in unsafe environments. According to the 2002 National Survey ofEmployment and Income (Encuesta Nacional de Empleo e Ingresos, orENEI), at least 17,350 children worked in domestic jobs, 3,700 childrenworked in the fireworks industry, and 850 children worked collecting andsorting garbage. Other kind s of haz ardou s jobs included agriculture, min-ing and rock break ing, and sex ual activities (IPEC 2003).Table 1 presents the incidence of child labor by region, area, and gen-der for children between five and sixteen years old. The percentage of

boys and girls working for a salary in rural areas is greater than in urbanareas. The child labor rate is greater for boys than for girls in almost each


4/23

i 6 8 Latin Am erican Research Reviewpercentage of working girls is greater in the Metropolitan, North, North-west, and Southwest regions, the urban percentage is greater in other re-gions. The max im um percentage of child labor is repo rted for boys in theRural North region (40.2 percent). The minimum corresponds to girls inthe Rural N ortheas t region (6.9 percent). Table 2 show s ethn ic differen-tials in child labor. Indigenous children are more likely to work than arenonindigenous children in almost all regions, with the exception of theSoutheast for boys and the Northeast and Peten for girls.

Educational indicators point to low levels of schooling. According tothe World Bank (2003), the illiteracy rate in Guatemala was 31 percent in2000. This rate is among the highest in Latin America, behind only Nica-ragua and Haiti. On average, individuals older than fourteen years of agehave 4.3 yea rs of schooling. Table 1 shows that children in rural areas areless likely to attend school. In 2000,29.4 percent of boys and 36.7 percent ofgirls between five and sixteen years old in rural areas w ere not enrolled inschool. These percentages are greater than the urb an percentages (14 per-cent for boys and 175 percent for girls). Nonenrollment ratios are greaterfor rural areas across all regions (see table 1).

Table 2 presents significant differences in nonenrollment ratios betweenindigeno us and non indigen ous children. In 2000, more th an 28 percent ofindigeno us boys were not enrolled in school, for a difference of 8.8 percentabove nonindigenous boys. The difference between indigenous and non-indigen ous girls is also substantial (16.3 percent). The nonen rollm ent ratiosare less for indigenous boys only in the Central and Southeast regions.Indigenous girls are less likely to be enrolled in school than are nonind ig-enous girls in each region. However, the difference between indigenousand non indigenous girls in the Central region seem s to be m inim al.Demand-side factors seem to be very important for not enrolling chil-dren in school (see table 3). More tha n 30 percent of children out of schoolin rura l and u rban areas report income pov erty as the main reason for notenrolling in school. The lack of interest in schooling is the second reasonin urban areas and the third reason in rural areas. The second and thirdreasons for not enrolling boys and girls in school in rural a reas are work-ing for wages and working at home, respectively. Although factors associ-ated w ith the supp ly of education were repo rted as reasons to not enroll inschool, the low percentages of closed class, school distance, and no schoolin the community suggest that Guatemala has advanced significantly to-ward universa l school coverage (World Bank 2005).


5/23

N

o

Wae

c.o

OI113c

o

5Bo

1 o

I I 3

en1 O co 01acoO)bC 3

1 oc cs tu2 -SP

, oT3

o3

00 1- 00O ON O COO O^ (N

00tN

CO tN NOtN tN tN

: j N ^ t N O N ^ t 3 N N q N OsLricotNotjinNotN0''"*'-i tN

00 00COCO1 tvo s0022. i ^N O(( inCOtN CO ONo

p NO inp 00^s NO I* NO CO

i n i n ^ c o c o i n i n v ootj tN K^^ o 00 00CO tN i-i

t i O O O O N ^ N O t X D ^ O N\ O ^^ '-^ NO NO ^3 ON tNi - ^ t N C N l t N C N c O t N c O

o q i n t N N O N O t N p c ot O N K o c N ( N O t Oc o c o ' ^ c o i n c N c o i n


6/23

170 Latin American Research ReviewTable 3 Reason for Not Attending School, by Area and Gender

ReasonIllnessClosed ClassWorking at HomeWorking for WageIncome PovertyNot InterestedSchool DistanceNot School in the CommunityAgeOther

Rural2,50,90,8

25,932,716.71.92,28.67.8

BoysUrban

5.11.51.5

15.039.424.8

5.86.9

Rural2,11,2

18,37.7

30,818,23,52,2728.8

GirlsUrban

4,20,6

10,38,5

41,218,51,50.95.29.1

Note: We calculated these rates using the 2000 Encuesta Nacional de Condiciones d e Vida(ENCOVI). The rates are expressed in percentages ofchildren betwe en five and sixteenyears of age.

in Central America, for a total of 24,139 people, or 2,2 fatalities per thou-sand habitants (based on the population in 1995), Estimated losses weremore than US$3,062,5 billion or 17,3 percent of the gross domestic prod-uct (GDP) in 1995,' As a response to its propensity for natural disasters,Guatemala created the National Coordinator for Reduction of Disasters(Coordinadora Nacional para la Reduccin de Desastres, or CONRED),which aims to reduce household vulnerability to natural disasters andcoordinates national efforts during and after a disaster,''

THEORETICAL FRAMEWORKEollowing Basu and Van (1998), parents are assumed to be benevolent

and maximize the child's utility function Li = U(X, S, W), where X repre-sents consumption greater than the minimum sustenance level, S is timeavailable for schooling, and W is time available for child labor. The utility

3, Hu rricane Stan is a more recent example of the vulnerability ofGuatemala to naturaldisaste rs. The Economic Com mission for Latin Am erica and the Caribbean (2005) reportedthat Hurricane Stan affected 3.5 million people, or 34 percent of the population, in 2005Direct effects were observed in more than 950 communities. More than 42,900 individuals


7/23

HOUSEHOLD SHOCKS, CHILD LABOR, AN D CHILD SCHOOLING I 7 Ifunction U is positively related to child consumption and schooling, andnegatively related to child labor. That is.

^ O , s O ,dX dSa n d

^ . 0.sw

Parents' choices are subject to budget and time restrictions:P X + P < (Y - PXo) + )W {la)

VJ -\- S^T {lb)where P is the price vector, Pj is the price of schooling, Y is the house-hold income earned by adults, X^ is the minimum household sustenancelevel, w is the child wage, and T is the total amount of nonleisure timeavailable for schooling and labor. This model is similar to the theoreti-cal framework Ray (2002) used, with the difference being that parents areassumed to ensure a minimum level of sustenance for the household(i.e., Xo) rather than first choosing consumption for adults.

As a result of the maximization problem, assuming separability be-tween child consumption and leisure, we derived the following simulta-neous equations:

. W* = L{CH, FAM, COM, S* w, P j , Y*) {2a)S* = L{CH, FAM, COM, W*, w, P j, Y*) (lb)

where CH, FAM, and COM are child, family, and community's charac-teristics, respectively. Y* is equal to Y - PX,,, the disposable income forchild consumption. W* and S* are the optimal choices on child labor andschooling, respectively.

The effects of child, family, and community characteristics on W* andS* depend on the type of such characteristics. Child ethnicity may have anegative impact on schooling, particularly if languages differ across eth-nic groups and education is not provided in such languages. There is alsoevidence that suggests that mothers have stronger preferences for child


8/23

172 Latin American Research Reviewmarkets in such communities may be ano ther factor affecting child laborchoices.

Because nonleisure time T is fixed, an increase (decrease) in time al-located to child labor will decrease (increase) the time allocated to school-ing. This implies that child labor and schooling are substitutes; therefore,we expected their effect on each other to be negative. That is.

awand

s 0.We expected nega tive price effects on schoo ling. In contrast, because childlabor is considered to decrease child utility, we expected price effects onchild labor to be positive. Thus, expected price effects are as follows:

^ ^ 0 , ^ . 0,^^.0, and id IP d

. 0 , . 0 , O.(IPs di dPs

Under the benevolent-parents assum ption, we expected d isposable in-come to have a positive effect on child schooling and a negative effect onchild labor. That is.dY

and

dY s 0.It is w orth noting that household shocks (e.g., na tural disasters, socioeco-nomic shocks) may decrease household income from adults. In addition,household surviva l ex pen diture s (e.g., health costs) may increase as a re-sult of household shocks (Del Ninno and M arini 2005). Parents may sendchildren to work to compensate for income losses and to pay for surviva lexpe nditures. In addition, parents m ay not have enoug h income to enroll


9/23

HOUSEHOLD SHOCKS, CHILD LABOR, AN D CHILD SCHOOLING I 7 3H2: Natural disasters and socioeconomic shocks decrease the time allocated to

child schooling.H: The impact of natural disasters and socioeconomic shocks is more signifi-cant among poor households.

DATA AND VARIABLESIn 2000, Guatemala implemented a living standards measurementsurvey referred to as Encuesta Nacional de Condiciones de Vida 2000(ENCOVI). The ENCOVI followed a two-stage stratified cluster samplingdesign . The stratification consists of rural a nd urb an areas in eight political

regions, for a total of sixteen areas. Househ olds in these are as are classifiedin three strata: high, m edium , and low income. A total of 11,170 rura l andurban sectors are the prim ary sam pling un its (PSUs). In the first stage, thesam pling probability is the same for each PSU. The second stage, in whichhouseholds w ere the secondary sam pling un its (SSUs), implem ented sys-tematic sampling. Group s of twelve and six house holds w ere formed forrura l and urb an areas, respectively. Finally, 7,276 households provided theinformation that ENCOVI presented. We selected a subsample of 7,332children between five and sixteen years to conduct th is study.The ENCOVI includes information on child activities and binary indi-cators for househo ld shocks. That inform ation is used to define child laboras children working at least one hour in the market for a monetary com-pensation. Households were also asked whether they experienced any ofthe shocks listed in the questionnaire over the previous twelve months(the appen dix here prese nts the questions in Spanish). The list of shocksincluded natural disasters such as earthquakes, droughts, floods, storms,hu rricanes, plagues, landslides, forest fires, and loss of harves t. The ques-tionnaire also included socioeconomic shocks: loss of employment of anymem ber; lowered income of any m ember; ban kru ptcy of family business;illness or serious accident of a work ing memb er of the hou seho ld; death ofworking m ember of the household; abandonm ent by the household head;fire in the house, business, or property; criminal act; land dispute; lossof cash or in-kind assistance; fall in prices of products in the householdbusiness, business closing; massive layoffs; and general price increases.Household propensity to natu ral disasters and socioeconomic shocks mayinfiuence the incidence of shocks.About 28 percent of the sampled househ olds suffered at least one na tu-


10/23

Table 4 Definition of VariablesVariablesS C H O O L I N GW O R K I N GH H S H O C KNATDISS O C E C O NEXTPOORP O O RAGEM A L EI N D I G E N O U SM O E D U CFAEDUCREMITT

H E A D M A L EH H S I Z ERURALN O R T HN O R T H E A S TS O U T H E A S TC E N T R A L

S O U T H W E S T

DefinitionIs child schooling? (yes = 1; no = 0)Is child working? (yes = 1; no = 0)Number of negative events in past12 m onthsHousehold propensity to naturaldisastersHousehold propensity to socioeco-nomic shocksIs the household below the extremepoverty line? (yes = 1; no = 0)Is the household below the povertyline? (yes = 1; no = 0)Child ageChild gender (male =1; female = 0)Child ethnic (indigenous = 1;otherwise = 0)Did mother complete elementary

school? (yes = 1; no = 0)Did father complete elementaryschool? (yes = 1; no = 0)The log of household income (intho usa nd s of quetzals) fromremittancesIs the household head male?(yes = 1; no = 0)Number of household membersChild location (urban = 0;rural = 1)Does child live in th is region?(yes = 1; no = 0)Does child live in this region?(yes = 1; no = 0)Does child live in this region?(yes = 1; no = 0)Does child live in this region?(yes = 1; no = 0)

Does child live in this region?(yes = 1; no = 0)

Mean0.7380.2071.631000.1790.449

10.1550.5100.4100.1180.1541.077

0.849

70430.6160.1200.070

0.1100.170

0.170

SD0.4390.4041.4540.697

0.685

0.384

0.497

3.4210.5000.490

0.323

0.3612.259

0.358

2.4190.486

0.320

0.260

0.3100.370

0.380

Min.000

- 0 . 4 2 0- 0 . 4 5 5

00500000

02000000

Max11

14757011

161111

12.4

1181

1111

1


11/23

HOUSEHOLD SHOCKS, CHILD LABOR, AND CHILD SCHOOLING I75business closing and massive layoffs (0.23), and loss of employment of anymem ber and lowered income of any member (0,3),

Table 4 presents the definition anddescriptive statistics of the variablesused in this study. Thevariables include indicators for extreme poverty{EXTPOOR) and general poverty {POOR). The National Institute of Sta-tistics calculated thepoverty indicators using the per capita consumptionapproach. The extreme poverty line (1,911 quetzals per year) consisted ofexpenditures needed tocover the minimum amount of calories needed tosurvive. The general poverty line (4,318 quetzals per year) covered non-nutritious expend itures as well.^ Other variables use d in the study includechild characteristics {AGE, MALE, and INDIGENOUS), family character-istics including mother's and father's education, gender of the householdhead, and household size {MOEDUC, FAEDUC, HEADMALE, and HH-SIZE, respectively), and location variables {URBAN, NORTH, NORTHEAST,SOUTHEAST, CENTRAL, SOUTHWEST, NORTHWEST, and PETEN).

I ' * .

ECONOMETRIC METHODOLOGY' IThe information included in ENCOVI can be used in different ways

to investigate the effects of household shocks on child labor and school-ing, Eor instance, Guarcello and colleagues (2003) used the informationto estimate the effects of individual and collective household shocks onchild labor. They used bina ry indicators to classify a household as hit bya shock if the household reported at least one shock. Although this ap-proach is suitable for m easuring shock incidence, it ignores the propensityto household shocks and assumes that the effect of being affected by oneor more shocks is the sam e. Counting the shocks a household has sufferedmay be an alternative to account for the occurrence of various shocks.This approach, however, assumes that different shocks have the same ef-fect on child labor and schooling,^This article prop oses a different approach to m easu re the prope nsity tohousehold shocks, Eirst, we classified repo rted household shocks as na tu-ral disasters and socioeconomic shocks. Then, we estimated two indicesto measure the propensity tonatural disasters and socioeconomic shocksusing factor analysis. We assum ed theunobserved household propensityto na tural disasters and socioeconomic shocks influenced repo rted shocksas follows: i - . ,

ei (3fl)


12/23

176 Latin American ResearchReviewwhere NDi is an indicator that takes the value of one if the household experienced natural disaster i (e.g., earthquakes, droughts, floods) and zerootherwise. Similarly, the indicator SEj takes the value of one if the household was hit by socioeconomic shock ; (e.g., loss of employment of anymember, lowered income of any member, bankruptcy of family businessand zero otherwise. The factors NATDIS and SOCECON are unobservedcommon factors that influence the incidence of natural disasters and socioeconomic shocks, respectively. Coefficients A, and A represent the factor loadings relating indicators ND and SEj to latent factors NATDIS andSOCECON, respectively. The terms e, and e represent the variance that iunique to indicators ND, and S^, respectively, and are independent of thecorresponding factors and all other e,.

Factor analysis provides a stronger analytical framework to estimateshock indices than do traditional methods such as binary variables representing shock incidence and the total amount of reported shocks. First, factors can be used as proxies for latent variables, which are unobservable andinestimable using traditional methods. Therefore, NATDIS and SOCECONcan be interpreted as the latent household propensity to natural disastersand socioeconomic shocks, respectively. Second, NATDIS and SOCECONdo not restrict different household shocks to equally affect child labor andschooling, given that factor loadings are allowed to vary across shock indicators. Finally, factor analysis provides estimates that are adjusted fomeasurement error, which traditional methods ignore (Brown 2006).

To investigate the impact of natural disasters and socioeconomicshocks on child labor and schooling, we modeled the optimal choice ofchild schooling and labor under the assumption that Equations 2a and 2bfollow a linear form:

S* = X , + u, (4fl),, {4b)

where S* and W* represent the optimal time allocated to child schoolingand labor, respectively. Vector X represents determinants of child school-ing and labor including child, family, and community characteristicsas well as the shock indices NATDIS and SOCECON.Vectors , and ,include the parameters to be estimated. Finally, u^ and u^ are error termthat follow a normal joint distribution with mean zero and the samevariance for each child. The error terms u, and w, are allowed to be cor-


13/23

HOUSEHOLD SHOCK S, CHILD LABOR, AN D CHILD SCHOOLING I 7 7effect of remittances on child labor and schooling. We included RURALand regional dummy variables to control for unobserved communitycharacteristics. Regional dummy variables are NORTH, NORTHEAST,SOUTHEAST, CENTRAL, SOUTHWEST, NORTHWEST, and PETEN.METROPOLITAN is the base region.Unfortunately, we could not estim ate Equation 4a because ENCOVI didnot rep ort the time allocated to child schooling. Alternatively, ENCOVI in-cluded an indicator on child en rollm ent in school. Therefore, we replacedS* and W*with Schooling and Working, respectively. The indicator School-ing is equal to one when the optimal allocation of time to child schoolingis greater than zero {S* > 0) and zero otherwise. Similarly, Working is anindicator equa l to one when the op tim al allocation of time to child labor isgreater than zero {W* > 0) and zero otherw ise. The result of this transfor-mation is a bivariate probit model.We estimated bivariate probit models including NATDIS and SOCE-CON to test HI and H2. Using these factors to estimate the bivariate pro-bit models allowed us to assess the individual effects of household pro-pensity to natu ral disas ters and socioeconomic shocks on child labor andschooling. We also included the interaction of NATDIS and SOCECONw ith the poverty indices POOR and EXTPOOR to test H3.EMPIRICAL RESULTS

Table 5 presents the estimation results of Equations 3a and 3b. Shockindicators are assumed to be affected by factors NATDIS and SOCECONif their corre spondin g Kaiser-Meyer-Olkin (KMO) statistic is greater th an0.5. Nine indicators on natural disasters are identified to be related toNATDIS and 14 indicators on socioeconomic shocks are related to SOCE-CON. Plagues and loss of harvest show the highest factor loadings (0.406),followed by droughts and storms. The lower factor loading correspondsto reports on earthquakes (0.154). The highest factor loading for SOCE-CON corresponds to lowered income of any member (0.456), followed bymassive layoffs (0.422). The lowest factor loading corresponds to deathof working member. We used corresponding factor loadings to estimateNATDIS and SOCECON, which we then included in probit models to in-vestigate the effect of natu ral d isasters and socioeconomic shocks on childlabor and schooling.Table 6 presents the estimation results (i.e., marginal effects) of bivari-


14/23

1 7 8 Latin American Research ReviewTable 5 Factor Analysis ofNattiral Disasters and Socioeconomic Shocks

VariablesPlagues

Loss of harvestDroughts

StormsLandslides

Hurricanes

FloodsForest fires

Ear thquakes

NATDISFactor

l o a d i n g s0.406

0.4060.370

0.3540.293

0.287

0.214

0.209

0.154

K M Ostat is t ic0.656

0.6520.712

0.7120.692

0.679

0.643

0.695

0.617

SOCECON

VariablesLowered in-come of anymemberMassive layoffsLoss of employ-ment of any

memberBusinessclosingGeneralincrease inpricesBankruptcyof familybusinessIllness or acci-dent of work-ing memberFall inprice s ofproducts inbusinessLoss of cashor in-kindassistanceCriminal actLand disputeA bandonm entby the house-hold headFire in theh o u s e /bus iness /

proper tyDeath of work-ing member

Factorloadings0.456

0.4220.313

0.2860.239

0.237

0.1830.163

0.153

0.1380.1100.085

0.049

0.044

KM Ostatistic0.619

0.6010.622

0.6090.667

0.603

0.6550.536

0.624

0.6740.6030.566

0.537

0.576


15/23

15

D

o

o

Ioo

f i

s"ISo6Cc

* * * * * * * * * * ^p p p p p p p p p p p p p p p p S p p p p p p p p p p p p p - 5 Q d> d> c

* * * * * * * * * * 'm ^ O^ O^ 00 ^ ! > . C4 1-H [>. Ov ^ O i n " CN ^ o O^ ON i ^ " ^ ^ O fO^ fN OO N C^ 0 0 ^ "i - H i - H O f N i n i - H i N i - H t N f N L n i - H t n i - H . - H O ^ O c o T - i o o i - H t N . i t i - H O " ^ i n o oP P PP ppP pPpppCs | pppppppT- Hp i -H pppppppOOOOO

* * * *

* ***{** *

0 CDOOOOO0OOOOO^ ' I - - - - I > - - ' I ' - ' - ' I '* * * * ** * * * * * *

,.lTr.f:-.ol--E-tslrt>tt.^.o^ l^col-ocNK^I I I i ^ fTT f^ f f ^ ^^ f\| f^ f-^ f j f ^ f -; f ^ f ^H


16/23

O

6 0O

60C

( N ? H ( N \ . 0 T H O t ^ t ^ r H 0 0 r ^ i r i T 4 ^ f ^ Oo o ^ H r H C v i o c r i N r N o o e N T - H r - j f ^ r ^ r o c No p p p p p p p p p p p p p p pdô,dw ^

m n t x O i H H t N v o m v o i r i s o O Q O c s i ?O ' - H i n m N c ^ o c N ^ H M O f N ' . o c v L n f Or - j p p p p p p p p p p p p p p po o o d o d o o o o o o

' - ' ^ t ^ c N f ^ ' O f N f f s o r s i \ o c N i r i f oo oooo ooooo oo oo oO00000000OO o B o o 3 5 * 3 5 C T \ f > o 5 t x SQ O i - H t - H f N O C O C ^ ^ f N I Q O f N l T - H C N I O f N - ^ f np p p p p p p p p p p p p p p p d d

0 o r O O 3 c \ ^ ^ 5 c ^ 5 o 0 0ON - H s C r o c N c o o r J i - H f N O t N ^ C N i - ^ c r ip p p p p p p p p p p p p p p pdo,ooo,ooo,dooooI I I - I -- I ' I Ic ! 0 ^ ^ ( N H < n c O [ N i o O [ N ? H ( N i o r s ) - * r o ^ O f M r p p p p p p p p p p p p p p p p N p K00,00.00,00.00.00,00,00,00, So

* * * *-.o>-r(r)Ot>.i-',-fvia\ôlNLD'ôooNOONroroO ^ t v ^ r O C N m O f N l t i C M O f S ' v D f s l ' ^ C D ^ O C S r Op p p p p p p p p p p p p p p p C ^ p t N0OOOO0000000

i ^

roro"t-HppppppppppppiN

o o o o o o o o o o o o o o o o o o(NIrot?

I p


17/23

HOUSEHOLD SHOCKS, CHILD LABOR, AN D CHILD SCHOOLING l 8 lWe found no evidence to support H2. Socioeconomic shocks seem tohave no effect on child schooling , as indica ted by insignificant coefficientsof SOCECON for the schooling equation in Models 3 and 5. In contrastto H2, the marginal effect of NATDIS on child schooling is positive andsignificant in Models 1 and 2. This suggests that child schooling increaseswith natural disasters, which contradicts the hypothesis that householdstransfer resources from schooling to survival expenditures, which usu-ally increase with household shocks by not enrolling children in school.However, this result is consistent with the evidence Duryea and Arends-Kuenning (2003) presented, which suggests that schooling can increaseunder bad macroeconomic conditions because child labor is less attrac-

tive as a result of falls in wages for unskilled labor. Both child labor andschooling may increase if natural disasters affect domestic productionand, in turn, increase the time available for these child activities.In support of H3, estimated coefficients on SOCECON X POOR inModels 4 and 6 indicate that poor ho useh olds are more likely to use childlabor as a strategy to cope with socioeconomic shocks than are nonpoorhouseholds. The coefficients also indicate that poor households reducechild schooling to cope with socioeconomic shocks, which is consistentwith the evidence Jacoby and Skoufias (1997) presented. In contrast, es-timated coefficients on NATDIS X POOR in M odels 2 and 6 suggest thatnatural disasters increase child schooling in poor households but not inextremely poor and nonpoor households, which contradicts H3. Also, wefound no evidence to support the hypothesis that extremely poor house-holds respond differently to socioeconomic shocks in terms of child laborand schooling compared with nonpoor households. Compared with non-poor households, extremely poor households are also less likely to usechild labor for coping with natural disasters, as indicated by estimatedcoefficients of NATDIS X EXTPOOR in Models 2 and 4. Extremely poorhouseho lds m ay have limited access to labor markets; consequently, childlabor would not be a potential strategy to cope with natural disasters orsocioeconomic shocks.In addition, results suggest that child characteristics affect child laborand schooling (see table 6). The probability of child labor and schoolingincreases with age at a decreasing rate. This suggests a lower probabil-ity of enrolling in middle school, at least in normative age. In addition,the probability of child labor and schooling is greater for boys than forgirls. That is, househo lds m ay treat boys and g irls differently in term s of


18/23

i 8 2 Latin Am erican ResearchReviewbridge (2007) also have presen ted evidence on the schooling differentialsbetween indigenous and nonindigenous children in Guatemala.Paren ts with at least prim ary education increase the probability of childschooling. In contrast, the probability of child labor decreases with theeducation of both mothers and fathers. These effects are more significantfor mothers than for fathers, in support of the hypothesis that mothershave a stronger preference for child welfare (Ridao-Cano 2001). Educatedparents usually earn higher wages, which may be enough to pay for sur-vival exp end itures. In that case, child labor is not needed to ens ure h ous e-hold survival and children may be enrolled in school. Consistent withprevious findings (e.g., Patrinos and Psacharopoulos 1997), household size{HHSIZE) negatively affects the prob ability of schooling and increases theprobability of child labor. In contrast, remittances increase the p robab ilityof schooling. However, remittances do not affect the probability of childlabor. This indicates that remittances do not completely eliminate childlabor, although the number of working hours could be reduced. This isconsistent with ou r theoretical framework, which pred icts that more adu ltincome reduces the num ber of hou rs allocated to child labor and increasesthe time allocated to schooling.

Table 6 also shows that children in rural areas are less likely to be en-rolled in school compared with children in urban areas. As table 3 shows,hou seholds in ru ral areas p oint to income poverty, paid w ork, and lack ofinterest as primary reasons for not sending children to school. Childrenin ru ral a reas are also more likely to work, as the positive and significantcoefficients of RURAL indicate in the working equations. Compared withthe m etropolitan area, children are more likely to work in the Cen tral re-gion, where labor markets are m ore developed. In addition, the probabil-ity of schooling is lower in the North, Northwest, and Peten, which areregions with a significant ind igenou s p opu lation.Finally, it is wo rth no ting that the correlation (p) betw een schooling andworking is negative and significant across all models, greater than 0.26 inabsolute term s (see table 6). This is consistent with the theoretical frame-work of this study and existing evidence (e.g.. Binder and Scrogin 1999)that suggests a trade-off between child labor and schooling.CONCL USION AND POL ICY RE COMME NDAT ION

This article investigates whether Guatemalan households use child


19/23

HOUSEHOLD SHOCK S, CHILD LABOR, AN D CHILD SCHOOLING 1 8 3increases with natural disasters. Both child labor and schooling may beexpected to increase with natural disasters if domestic production is af-fected and if the time available for working and schooling consequentlyincreases.Results also indicate that poor households-{but not extremely poorhouseholds) are more likely to use child labor and schooling reductionas strategies to cope with socioeconomic shocks. As a response to suchshocks, poor households may have to send children to labor markets.They may use child earnings to pay for survival expenditures, especiallyw hen adult incom e decreases as a result of socioeconomic shocks (Beegleet al. 2006). In addition, households may reallocate resources to survivalexp end itures by not enrolling children in school. In contrast, we found noevidence that poor and extremely poor households are more likely to usechild labor to cope with natu ral disa sters than are nonpo or hou seho lds. Incontrast, findings indicate that extremely poor ho useholds are less likelyto use this strategy than are nonpoor households. Natural disaster mayfurther restrict access to labor markets for extremely poor households,which would prevent them from using child labor to cope with naturaldisasters and socioeconomic shocks.Child labor and schooling reduction may increase the amount of re-sources aimed at mitigating the negative effects of socioeconomic shocks(Jacoby and Skoufias 1997). However, such coping strategies may have anegative impact on the curren t and future welfare of children. The currentwelfare of children may be put at risk because of unsafe working envi-ronments. The reduction in the amount and quality of schoolinggiventhat m ore schooling is associated w ith pove rty alleviation, disease reduc-tion, and fertility choicesalso jeopardizes the future welfare of children(Glewwe 2002; World Bank 2005).

Public policies aimed to prevent and mitigate socioeconomic shocksmay improve the welfare of children. For example, coping assistanceprograms could be implemented to provide households with access tocredit, insurance, and assets to cope with socioeconomic shocks (Beegleet al. 2006; Guarcello et al. 2003). Mitigation policies could be attachedto schooling; thus, households would send their children to school toimprove the future welfare of children and to be eligible for coping as-sistance. Implementing coping assistance programs may be a challengein developing countries. Guatemala may face this challenge using offi-cial institutions such as the Secretariat of Food Security, the Presidential


20/23

184 Latin American ResearchReviewAPPENDIX A. .QUESTIONS USED TO ESTIMATE THE LATENT FACTORS NATDIS AND SOCIOECON

En los ltimos 12 meses el hogar se ha visto afectado por alguno de lossiguiente problemas de tipo general?1. Terremoto2. Sequia3. Inundacin4. Tormentas5. Huracn6. Plagas7. Deslizamiento de tierras8. Incendios forestales9. Cierre de empresas

10. Despidos masivos11. Aum ento general de precios12. Protestas pblicas13. Otro, cul?

En los ltimo s 12 meses, este hogar se vio afectado po r alguno o alg unosde los siguiente problemas?1. Prdida del empleo de algn miem bro2. Baja de ingresos de algn miembro del hogar3. Quiebra del negocio familiar4. Enfermedad o accidente grave de algn trabajador miembro delhogar5. Mu erte de un trabajador miem bro del hogar6. M uerte de otro miembro del hogar7. A ban don o del jefe de hog ar8. Incendio de la vivienda/negocio/propiedad9. Hecho delictivo

10. Disputa de tierras


21/23

HOUSEHOLD SHOCKS, CHILD LABOR, ANDCHILD SCHOOLING 185REEERENCESAmin, Shahina, M. Shakil Quayes, and Janet M. Rives

2004 "Poverty and Other Determinants of Child Labor in Bangladesh." Southern Eco-nomic Journal 70 (4): 876-892.Basu, Kaushik, and Pham Hoang Van

1998 "The Economics of Child Labor." American Economic Review 88 (3): 412-427.Beegle, Kathleen, Rajeev H. Dehejia, and Roberta Gatti

2006 "Child Labor and Agricultural Shocks." Journal of Development Economics 81 (1):80-96.Binder, Melissa

1998 "Family Background, Gender and Schooling in Mexico." Journal of DevelopmentStudies 35 (2); 54-71.

Binder, Melissa, and David Scrogin1999 "Labor Eorce Participation and Household Work of Urban Schoolchildren inMexico: Characteristics and Consequences." Economic Development and CulturalChange 48 (1): 123-154.

Bourdillon, Michael2006 "Children and Work: A Review of Current Literature and Debates." Development

and Change 37 (6): 1201-1226.Brown, Timothy A.

2006 "Confirmatory Eactor Analysis for Applied Research." New York: GuilfordPress.

Charvriat, Cline2000 "Natural Disasters in Latin America and the Caribbean: An Overview of Risk."Working Paper No. 434. Washington, D.C: Inter-American Development Bank,

Del Ninno, Carlo, and Alessandra Marini2005 "Households' Vulnerability to Shocks in Zambia," Social Protection DiscussionPaper No. 536. Washington, D.C: World Bank.

Dendir, Seife2007 "Household Shocks and Child Labor in Rural Malawi." Economics Bulletin 4 (1):1-8.

Dumas, Christelle2007 "Why Do Parents Make Their Children Work? A Test of the Poverty Hypothesis

in Rural Areas of Burkina Easo." Oxford Economic Papers 59 (2): 301-329.Duryea, Suzanne, and Mary Arends-Kuenning

2003 "School Attendance, Child Labor and Local Labor Market Fluctuations in UrbanBrazil." World Development 31 (7): 1165-1178.Economic Commission for Latin America and the Caribbean

2005 "Efectos en Guatemala de las lluvias torrenciales y la tormenta tropical Stan."Mexico City: Economie Commission for Latin America and the Caribbean, UnitedNations.

Emerson, Patrick M., and Andr Prtela Souza2007 "Child Labor, School Attendance, and Intrahousehold Gender Bias in Brazil."World Bank Economic Review 21 (2): 301 -316.Glewwe, Paul2002 "Schools and Skills in Developing Countries: Education Policies and Socioeco-

nomic Outcomes." Journal of Economic Literature 40 (2): 436-482.Guarcello, Lorenzo, Eabrizia Mealli, and Eurio C. Rosati

2003 "Household Vulnerability and Child Labor: The Effect of Shocks, Credit Ration-ing and Insurance." Social Protection Discussion Paper No. 322. Washington,


22/23

i8 6 Latin Am erican ResearchReviewJensen, Peter, and Helena S. Nielsen1997 "Child Labor or School Attendance ? Evidence from Zam bia." Journal of PopulationEconomics 10 (4): 407-424.McEwan, Patrick J., and Marisol Trowbridge2007 "The Achievem ent of Indigen ous Studen ts in Gu atema lan Prim ary Schools." In -ternational journal of Educational Development 27 (1): 61-76.Patrinos, Harry A., and George Psacharopoulos1997 "Family Size, Schooling and Child Labor in Peru: An Em pirical An alysis." journalof Population Economics 10 (4): 387-405.Psacharopoulos, G eorge1997 "Child Labor versu s Educational Attainm ent: Some Evidence from Latin Amer-ica." journal of Population Economics 10 (4): 377 -38 6.Ray, Ranjan2002 "The Determinants of Child Labor and Child Schooling in Ghana." Journal ofAfrican Economies 11 (4): 561 -590 .Ridao-Cano, C ristobal2001 "Child Labor and Schooling in a Low Income Rural Economy." W orking PaperNo. 1. Boulder: Institute of Behavioral Science, Population Aging Center, Univer-sity of Colorado.World Bank2003 "La pobreza en Gu atem ala." Informe 24221-GU. W ashing ton, D.C.: World Bank.2005 "Cen tral Am erican Education Strategy: An Agenda for Action." World BankCo untr y Stud ies. Wa shington, D.C.: World Bank.


23/23

Copyright of Latin American Research Review is the property of Latin American Studies Association and its

content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's

express written permission. However, users may print, download, or email articles for individual use.

Documents

Household Shocks, Child Labor, And Child Schooling