A cost of siblings : child schooling in urban Colombia

World Bank Reprint Series: Number 160 Nancy Birdsall A Cost of Siblarings: Child Schooling in Urban Colombia Reprinted with permission from Research in Population Economnics, vol. 2 (1980), pp. 115-50. Copyrighted by JAI Press, Inc,  A COST OF SIBLINGS. CHILD SCHOOLING IN URBAN COLOMBIA Nancy Birdsall, THE WORLD BANK I. INTRODUCTION Does the number of siblings amongst whom we grow up matter to our later fortunes in life? Do children who are raised in large families undergo some "cost of siblings" in terms of fewer parental resouces received during childhood and a resultant loss of income-earning potential as adults?' And does such a process imply the perpetuation of poverty from one generation to the next among low-wage high-fertility groups? The following analysis adds to other evidence "ndicating that there is some such cost of siblings, i.e., that on average, children from large families receive less schooling, do less well on tests of intelligence, and using indicators such as height, weight, and age at menarche, appear less well- nourished than children from small families, even controlling for socio- economic class.2 But my purpose in this analysis is to go beyond the simple negative correlation now increasingly confirmed between family size and Research in Population Economics, Vol. 2, pages 115-150 Copyright C. 1980 by JAI Press Inc. All rights of reproduction in any forrm; reserved. ISBN: 0-89232-125-3 115  116 NANCY BIRDSALL various measures of child welfare, to consideration of the underlying c...uses of that negative correlation. My objective is to illuminate more precisely what factors ultimately deternine the persistent and negativ, relation between numbers of children and allocation of resources to investment in those children. Data used for the analysis are from a 1967-68 family budget study in the four major cities of Colombia'; they include information on expenditures by the household on a wide variety of caLtegories, as well as inrormation on income of each person in the household, age and educational attainriient of the husband, wife, and all children present, and the number of wife's children-ever-born and children living. In the following discussion, "investment in children" will refer to current expenditures of parents' time and money on children which are likely to contribute to the children's future earning power. The measures of parental investment in children used here are educatior-related, including both current househo'd expenditures on education, and the educational attain- ment of childrer. present in the household relative to that of other children in their age-sex group. Education is a particularly good commodity for analysis of this type, since it is clearly child-specific, and market expen- ditures on it are much less aftfected by the problem of economies of scale than are expenditures on such commodities as clothling and housing. My anal\sis is based on a specific view of the household decision-nmaking process underlhing the simple negative correlation between family size and investment per child. C'onsider that the negative correlation could result for three conceptual ly distinct reasons. First: large numbers of children impose additional burdens on parents, di- rectly causing a reduction in the amount of resources they allocate to each child. This is the reasoning implicit in most standard analyses of family bud- gets.4 The initial appeal of this explaina tion ---lhat parents' time and money cannot be stretehed proportionately to accommodate a growing household- provides little insight into why parents, who as a group have access to the same intuition and can foresee the constraining effect of many children on per- child expenditures, ditter in their apparent choices regarding number of chil- dren and per-child expenditures.5 Second: an explanation often raised in the literature on fertility and its determinants: piarents who wish to invest much of their time and money income in their childreni will restrict family size." This is the first explaiiation turned on its head, and also to some extent begs the question as to why parer s differ in their apparent desires regarding allocation of resources to children. Thirdc: the possihility that the direction of causality between famnily size and child welfare is not uriiOLle. that for parents neitlher the decision (or lack  A Cost of Siblings 117 thereof) regarding family size nor that regarding investment pe >'hild precedes the other, In this view, the two sets of choices are interdepei: ,ent and jointly determined by characteristics of the parents and of the economic and social environment in which they live. (Such a view does not preclude the possibility that parents neither "plan" consciously for number of children nor investment per child, the issue in this case is what determines such joint nondecisions.) It is the third view which is the basis for the following analysis: it begins with the assunmption that the behavior of parents regarding per-child investment cannot be viewed as independent from their apparent choices regarding the number of children to have. The joint persistence of the contradictory explanations one and two, alternatively offered iy different analysts depending on their initial set of interests (i.e., the causes of differences in child welfare, or the causes of fertility differences) in itself lends credence to the third which, since it subsumes and reconciles the first two, is theoretically more appealing. The findings reported in this paplr are largely con',istent with this third view of the underlying process. General conclIsiOns include the follovving: 1. In the simple correlative sense (based on oruinary least squares estimation). there is a clear "cost of siblings," which persists, even control- ling for household income and other characteristics. The relationship of additional children to per-child investment is negative and nonlinear. Per- child investnment declines monotoiiicall) as family size increases, but once the dependent variable is standardized for age, a particular and different pattern emerges: up to three or four children, families maintain steady investment per child: with five and more children, investment per child is distinctly lower. 2. Using a model in which number of children and investment per child are jointly determined (two-stage least squares estimation), we find that large family size has an important effect in the c-ausal sense on per-child investnment. In this sample, as much as a 30 percent increase in the income of the head of household would be required to offset the negative effect of one additional child on a household's per-child educational expenditures. From a policy point of view, the implication is clear: reducticns in fertility will increase parents' investments in children's education. This is true partly becaiuse parents who seek to invest more restrict their fertility, but it is also trtie that an increase in parerits' educational investment per child would follo%% even from a decrease in family fertility brought about solely by lower contraceptive costs, 3. This negative effect of fertility on investment in children could be interpreted as inevitably causing the perpetuation of poverty across gener- ationls amoong high-fertility groups, short )f very substantial increases in  118 NANCY BIRDL)SAL.L income for already-large families. W,.ould such a conclusion be correct'? Probably not: though family size mntters. the vicious cycle is not inevitable. The analysis suggests one wav such a cycle can be interrupted. Although rural-urban migrants have larger families than otherwise comparable long- term urban residents (probably reflectinig the different economic ell iror.- men , which the originally-rural faniilic.s made their fertilitv decisions-- such Iamilies face different relative prices for children and inipujts to children; an example is dillerenlt availahility of contracepti% X). amintd ap- parently spent less on educaltio)n of their chlildreni %%hile in the rural area (children of comparable urban fLmilies have higher educational attain- ment), their current educational spLnding is sinmilir to that of' long-term urban residents. Current spending is apparently not so greatly influencted by the migrant farmilies' pr3or economic and social environment. Thus there operates some kind of a catch-up mnechanism:, despite higher fertility, larger average families, and lower average educational ittlalinmclit of children age and sex-standardized), migrant families adopt the cducatolellmal spending habits of their urban counterparts: they do not spend less per child than those of comparahle inconme and education. Thus we know that certain economic conditions (associated in this case with a move to an urban area, but other routes to changing prices are imaginable') cause a decline in fertility: fromii conclusion 2 aiove. xNe h1,n1' this lower fertility has a direct positive eI'feet on per-childi educattional investment, Furthermiorc, in the case of r',cent migrants. evenl geiven ,rcad\- large familv size, a change in economic environnmeit shifts invCstmllnt upward. This increased educational spenidinig by migrant pmmrenl,, xilI lo er their children's fertilitv in the next generationi, since educationi of parents itself has an independent negati%e effect on fertility. Even a miodeNt increalse in investment for one generation will lead to decreases in the fcrtilit% of lthe next: and modest decreases in the fertility of currrent parents have an immediate effect on their children' schooling. Section 11 of this paper is a short diseussion of the modtel on xN hich the analysis is based: in Section 111. empirical findings are preNentCe And discussed. II. ANALYTICAL FRAMEWORK A link between famnmilv size and ehild welfare (or child qualitx) is built into a model of fertility proffered by Willis, as well as Becker and Lewis,' in which the household is viewed as nia\imii/ing a utility uminctiomn of' the ormn 1. L,IGN,Q,Z) where N is the number of childrcn, Q is their quality or the household investment in them, and Z represents the rate of consumption of all other  A Cost of Siblings 119 commodities. N and Q enter as separate arguments in the utility function, but child services, C, is set equal to NQ, and it is C which is produced (along with Z) according to the linearly homogeneous production function: C = NQ = f(t,,x,) where t, and xc are vectors of the total amount of time and goods parents devote to children during the parents' lifetime. The fact that, in cross-section studies and over time, higher-income families tend to have fewer children is explained in two ways. First is a price effect. The principal "cost" of children is an opportunity cost, the time parents and especially mothers devote to childbearing and childrearing. Higher-wage parents experience greater opportunity costs in rearing chil- dren.' This price effect apparently s,wiamps any positive effect of income on the demand for children. Second is an explanation that bears more directly on the issue of child welfare and its relation to family size, namely the observation that parents may wish to substitute quality for quantity in the production (rearing) and consumption (enjoyment) of children, i.e., with greater income, parents may spend more time and monetary resources on fewer children as an alternative to having more children. A critical feature of Lhe model is that even without any special assumption about the substitution between quantity and quality in the parents' utility function or in household production, it will be true that for a smaller number of children. the true shado%% price parents face for quality in those children is lower; and that similarly for a greater number of children, the true shadow price of quality is higher. This follows because of the multiplicative relationship between N and Q in the production of child services, parents cannot "produce" children without produicing in them some degree of "quality," nor can they produce "quality" without chil- dren.'