A THEORETICAL FRAMEWORK FOR CHILD FOSTERING ARRANGEMENTS IN SUB-SAHARAN AFRICA

Transcription

1 A TEORETICAL FRAMEWORK FOR CILD FOSTERING ARRANGEMENTS IN SB-SAARAN AFRICA Renata Serra (Please send comments to June 2000 (Revised October 2000) Abstract The paper provides a irst economic analysis o the phenomenon o ostering in SSA, paying particular attention to the economic incentives and consequences or natural and oster parents. Fostering, both o young and older children, can be a mutually advantageous arrangement to natural and oster parents, when it exploits the dierences between the two parties in the opportunity cost o time, in the lie cycle phases, and in the potential or child education and training. In some cases, the transer o a child per se can bring net beneits to both parties. Thereore, monetary transers are not always necessary to make ostering exchanges mutually advantageous. owever, ostering arrangements do not always lead the parties to beneit equally, even allowing or a long horizon. This is so because ostering transactions may be embedded in other types o relationships between natural and oster parents, who are usually relatives. In these instances, ostering provides unctions that lie beyond the child care domain, such as inormal insurance and kinship support. The eects o ostering on child welare do not depend on the way beneits are distributed among the parties, and are oten linked to child own characteristics. Fostering is more oten the consequence, rather than the cause, o the observed discrimination among children. JEL Classiication Keywords

2 1. Introduction Child ostering is a temporary and reversible transer o child rearing responsibilities to other people than the natural parents. Anthropological and demographic studies show that, in Sub- Saharan Arica (SSA), a large proportion o non-orphaned children (up to 1/3 in many communities) lives with neither parent, oten with relatives, or a signiicant part o their childhood. Fostering in this part o the world is embedded in a complex system o reciprocal obligations among relatives and is accompanied by an exchange o goods and services over a long stretch o time. The unctions served are varied, in either rural or urban settings, and in households with dierent socio-economic characteristics. Some aspects o ostering are traditional and are connected to customary child training, and to the requirement that, or preparing or certain proessions, a child not only learns rom the master, but also works or him and lives with him (Goody 1982). Other aspects are more recent, such as the increasingly younger age at which children are ostered, as a consequence o their mothers assuming new roles in the changed household and urban economy (see Fiawoo 1978 or Southern Ghana), and ostering or educational purposes (Bledsoe 1994). In some areas the incidence o ostering is decreasing, because migration and mobility has loosened the cohesiveness o the extended amily (Blanc and Lloyd 1993); in other contexts its roles are reinorced, because individuals cannot cope alone with the increasing uncertain economic scenario. Fostering practices represent potentially a topic o considerable interest to economists concerned with the behaviour and the welare o households in Arican societies. There are at least our aspects requiring study and attention. First o all, ostering induces a circulation o people, goods and services across households and is thus a component o their complex system o inormal exchange and support (Serra 1996). Secondly, ostering might aect child nutritional and health status, and thereore inluence the child s uture physical development (Ainsworth 1967; Bledsoe and Brandon 1992). Thirdly, ostering interacts with child education and training, and inluences both the amount and the type o investment made in children s human capital, with long-term consequences or their welare and that o their parents and relatives (Ainsworth 1991; Lloyd and Desai 1991). Lastly, the existence o ostering practices may aect long-term parental choices about ertility and intergenerational transers, with repercussions on next generation s welare, natural resources and technology (Dasgupta 1993). Although these links are obvious to most economists working on Arican societies, economists have hardly devoted themselves to the analysis o the phenomenon o ostering and to the development o the necessary analytical and empirical tools. 1 The study o ostering, limited mainly to anthropology and demography, has proceeded so ar by a piecemeal approach, which has clariied a number o isolated instances but has not connected their results in a systematic way. Admittedly, the study o ostering presents the diiculty o making sense o the incredible variety and complexity o arrangements ound in practice, and o identiying, and explaining, prevalent patterns. Coming to an unambiguous evaluation o the implications o ostering or household well being, and child welare in particular, appears as a puzzling exercise, since the outcome seems to depend on the circumstances o the case. A context-based approach to the issue o ostering is not satisactory, however, in the ace o the increasing concerns about child welare in Sub-Saharan Arica and the compelling debates on the causes underlying persistent low levels o well being. 1 To my knowledge, the only economic work ocused on ostering is Ainsworth (1989, 1991) and Serra (1996, 1997). 2

3 It appears thereore important to attempt to attain some general understanding o the main actors that may induce, and aect, ostering arrangements, and how this may impact on welare indicators. The aim o this paper is to contribute to lay down the oundations or the economic study o ostering practices. It proposes a methodology or analysing the extraordinary variety o ostering arrangements and or identiying their most relevant unctions, by developing simple analytical tools rom the existing literature on household models. The objective o this research is to show that child ostering is an important, yet under-studied, micro-institution based on amily ties, perorming crucial material (economic) unctions. A sound theoretical understanding o this institution is clearly a precondition or empirical studies to address properly the our aspects mentioned earlier. A puzzling question or anyone who has approached the issue o ostering in SSA is who beneits rom ostering and in what ways. This paper addresses this issue, by trying to explain, in each particular instance, what might induce households to oster in or out, what are the types o beneits expected, i any, and what are the chances that these beneits might materialise within the context under consideration. In order to analyse the issue o individual and household incentives to oster in or out, two steps are ollowed. First, we ask which are the particular conditions under which a given ostering arrangement could be sustained as a mutually beneicial exchange, whereby both natural and oster parents (the two main parties) receive immediate gains. Ater having identiied a number o types o ostering, each serving a distinct unction, the market or competitive value o that particular type o ostering to both parties is identiied. Second, we examine the possibility that the net beneits to either party dier rom the incentive-compatible values. We are mainly interested in deviations rom the competitive pattern due to ostering being embedded in the wider relationships between oster and natural parents, who are usually relatives. In these instances, ostering plays other roles beyond those speciically related to child rearing. Two main unctions that ostering arrangements could provide are inormal insurance against unexpected exogenous shocks, and kinship support, provided by richer to poorer relatives according to well-sanctioned norms. The rest o the paper is organised as ollows. Section 2 explains the problem o regarding ostering as a mutually beneicial exchange in the literature and describes a way o identiying dierent patterns o ostering, three o which will be analysed in detail in the ollowing sections. Section 3 examines the types o incentives or natural and oster parents when young children are ostered. Section 4 deals with the ostering o older children, in a context where the latter only contribute to household production. Section 5 illustrates the more complex model o ostering o older children when children can either go to school or work and when parents need both market and household goods. Section 6 considers the cases where there is a deviation rom the competitive patterns previously identiied and provides a general typology o ostering arrangements, drawing some implications or empirical analysis. Section 7 provides some inal comments and conclusions. 2. Identiying the market-like component in ostering arrangements Surveys responses suggest a multiplicity o reasons or ostering in or out and an impressive variety o arrangements, which makes it diicult to establish in quite a speciic way how the beneits may be distributed, in an either ex-ante or ex-post sense. 2 Empirical studies on ostering do not provide an unambiguous answer to the question as to who beneits rom 2 See or instance Fiawoo (1978), Goody (1982), Etienne (1979a and 1979b), Bledsoe (1990), Renne (1993). 3

4 ostering and in what ways. Anthropological and sociological studies have emphasised the advantages accruing to oster parents, in terms o the support received rom other people s children, but have not given a satisactory answer to the question o why natural parents should send children away in these situations. (The accepted answer is what actors report, e.g. that a child is given as a git to relatives to whom one wants to show love, Goody 1982). Other studies (mainly demographic) regard instead ostering as a way to pass the costs o child rearing onto relatives again the explanation is based on a cultural given, in this case the norm that one cannot reuse a child given by relatives, with the additional eature that parents do not internalise all consequences o their ertility decisions. Two dierent interpretations lead in turn to distinct conceptions o the impact o ostering or society at large. The notion that ostering enhances child capabilities and skills, thereore beneiting present and uture generations, (Goody 1982) contrasts with the perception that ostering, by providing natural parents with a low-cost child rearing option, may lead to negative reproduction externalities and to a too high ertility rates rom the point o view o societal optimum (Isiugho-Abanihe 1983, Dasgupta 1993). First thing to note is that the act that natural and oster parents might have dierent socioeconomic status and bargaining power does not necessarily mean that one party is orced to accept an unwanted arrangement. An explanation resorting merely to kinship obligations (according to which a relative must meet a request or ostering in or out a child) appears unsatisactory, i nothing else, because not all poorer parents oster in or out children in the ace o relatives pressure; sometimes pressures go the other way around. owever constrained parents choice may be in particular instances, thereore, incentives to oster in or out play an important role. Moreover, norms are better explained rather than assumed as given. The singling out o incentives at the level o individuals and households is the aspect o main interest in this paper. Beneits do not need to be interpreted exclusively in terms o immediate economic returns: they could also encompass parents desire to improve the living conditions o their own children or the desire to strengthen the link with relatives, whether or not in the expectations o uture material support. ow is it possible to identiy individual incentives and how many types o ostering arrangements can be conveniently singled out or analysis? The approach ollowed here takes the lead rom existing empirical evidence (most o the existing studies on child ostering are rom West Arica, the region where the phenomenon is more diused or more distinctive). 3 The scrutiny o anthropological ield studies and large-scale demographic surveys suggests that some aspects above others determine the particular working o a ostering arrangement. These elements are the child age, the actors aecting whether a child can convey immediate beneits besides uture ones, and the degree o status dierential between the sending and receiving households. The interaction between these elements leads one to identiy three main patterns o ostering. All studies agree that the age o the child is an important motive or ostering and that the main divide occurs around the age o six-seven. 4 The ostering o very young children (under six years o age) constitutes the irst type o ostering, where the delegation o child-care responsibilities represents undoubtedly the most relevant unction. Although the proportion 3 For a description o the source material and or a better description o the criteria or the selection o the three patterns, see Serra (1996). 4 In the accounts o Arican childhood the age o 6 or 7 marks a turning point or the child, who is then given new responsibilities and induced to participate more actively in both household activities and in the lie o the extended amily. Whereas boys and girls are treated alike when very young, ater the age o 6, children take up gendered roles, e.g. boys and girls assume distinct rights and responsibilities (Erny 1987). 4

5 o oster children aged 0-6 is everywhere signiicantly lower than that or older children, the phenomenon o ostering young children is still quite remarkable in SSA, particularly in West Arica, when compared to the rest o the world. 5 As ar as older children are concerned, there are more complex considerations to take into account, since these children can contribute to the household economy, yet they also need training or ormal education in order to prepare or adulthood. Traditionally, ostering has acted mainly as an institution avouring the move o children to the house o relatives where they could learn important and much valued skills, such as techniques o cultivation and animal husbandry, ishing, metalwork, sculpture, weaving, Coranic teaching, music, etc. The inormal training scheme regulates the appropriation o the beneits by the relevant parties: the child works or the master in order to repay him or the teaching, and the sending amily would also contribute with gits and other transers. 6 Although ostering or training into traditional skills can still be ound, the peculiar trait o this orm o ostering e.g. the compatibility between child work and training and between adults immediate and uture beneits derivable rom children is loosing ground. The main causes are the increasing importance o ormal education, which, unlike traditional education, is incompatible with child active contribution to the household economy, and the deterioration o household material conditions, which induces adults to discount heavily uture returns. In order to account or the changes in traditional orms o ostering, two patterns o ostering older children are examined in this paper. In the irst, the ocus is simply on child work: when deprived o its traditional training unction, child circulation acts as a mechanism or redistributing child labour across households, thus responding to their immediate requirements or child labour (section 4). The other pattern considers the case whereby natural and oster parents try to mediate between the conlicting goals o child work and ormal education, but ace dierent constraints and options: ostering may represent a way to negotiate between the incompatible needs o the two parties (section 5). It could be argued that there are in reality more than three ostering patterns, and that the proposed classiication is rather reductive. owever, it will be shown in section 6 that other patterns o ostering may be derived as variant o either o these three, the latter thus tackling the most basic aspects o the interaction between ostering and other domains o household decision-making. As a result, the analysis o these three patterns represents a good compromise between the goal o ormulating a simple theoretical ramework and, at the same time, that o keeping analysis adherent to a varied, multiorm and complex empirical context. 3. Fostering young children: the child-care unction Evidence rom SSA contexts shows that, when young children are ostered, there exist signiicant generational and locational dierences between natural and oster parents, e.g. the latter are generally older, less economically active, and more likely to reside in rural areas than the ormer. The key actor under consideration here is the trade-o between the various uses o women s time: outside work, household work, child rearing, leisure, etc. We consider irst how ostering creates the scope or a more eicient allocation o women s time across various tasks, to be ollowed by an analysis o the reasons why ostering might have emerged in these settings as a response to the need or such rationalisation. 5 For a comparison o parent-child living arrangements across countries and continents on the basis o the Demographic and ealth Surveys, see Lloyd and Desai (1991). 6 See or instance the study by Goody (1982) among the Gonja in Northern Ghana, a community rom which earliest inormation on child rearing dates back to the irst hal o the 20 th century. 5

6 The model proposed in this section is a very simple extension o the standard household models, but we discuss it here in details, because it will provide the basis on which more complex models are built in the ollowing sections. The household is assumed to maximise a joint utility unction, which is a twice dierentiable, quasi-concave utility index deined over market and household goods. Market goods are purchased with the productive activities o both husband and wie. 7 It is assumed that the husband s income is given (e.g. his supply o labour is ixed), whereas that o the wie depends on her time allocation decisions. A woman divides her total time between productive work and household work, which includes child rearing, and there is no leisure decision to make (leisure beyond minimum is a superior good or most women in SSA). The woman s productive work is any activity that gives her the possibility to procure consumption necessities or the amily, besides those provided by her husband (normally there is a socially recognised division between the goods procured by husbands and wives, respectively) it could be agricultural production yielding a homogeneous consumption good, or any work or cash (market work, salaried work) that enables her to buy the goods. Woman s production is described by a generic unction: y(t w ; A) y >0 y <0 Such unction is deined over the share o time devoted to productive work, t w, and dependent on a set o parameters, A, such as age, health, education and experience, which are speciic to each woman and aect the eiciency with which labour is translated into output. The orm o the production unction may vary according to the work activity; here it is supposed to exhibit decreasing marginal returns to scale (as i it were agricultural production), but these may be constant in the case o wage work. ousehold goods, which are jointly consumed by household members, are produced according to the unction h (twice dierentiable and quasi-concave), or which woman s time (t h ) is the only input. The time devoted to household work must be at least suicient to ensure the care o the young children in the house (there are only young children here, since the implications o ostering older children, who may release mother s time rom household work, will be considered only later). Child rearing is not exogenously given, because women can rear a number o children greater or lower than their own, by, respectively, ostering in or out. The mother s time required to care or n children is represented by a generic unction τ(n), which may or may not incorporate such eatures as economies o scale in child rearing. Thereore the amount o household goods produced, net o child rearing tasks, is: h(t h -τ(n)) 0 since t h τ(n) I a child is ostered out, natural parents are expected to compensate oster parents by a transer o market goods, equal to F the value o market goods may be expressed in terms o a generic consumption good or in monetary value, where the price has been normalised to one. The simple decision problem considered here is one o maximising a household utility unction with respect to woman s time allocation and to the net number o children to rear in the household. Since the husband devotes all his time (T m ) to work at a ixed wage (w m ), there is no relevant decision or him to make. I the woman has the autonomy to make her own choices, the problem set up translates rom a context o joint decision into one o emale 7 This assumption is supported by evidence that women in SSA typically engage in other productive activity than household work, involving outside work or work or cash. 6

7 choice. 8 In any case, the income earned by the husband, or the share o such income transerred to the wie, represents an exogenous variable. 9 The problem may be characterised as ollows: max, s. t. M = y = h T = t τ h t h ( N ) y', h', τ' > 0 = ( M, ) ( t w; A) Tmwm ( N ) ( t τ( N )) 0 h t w t h y' ', h'', τ'' < 0 X F 0 [1] Note that the child rearing unction is here supposed to be increasing in its domain, but at a diminishing rate (allowing or economies o scale in rearing children). The exogenous number o own young children is N, whereas is the number o oster children; is a positive number i children are ostered in, and negative i children are ostered out. For each child ostered in (out), parents receive (pay) an amount F o market goods. The number o children ostered out cannot be greater than the number o own children, N. The non-negativity assumption or the solutions or M and will ensure an upper limit on the maximum number o children that can be ostered in (given time and productivity constraints). 10 Note that young resident children consume each a ixed quantity X o market goods, so that M, net o such costs, is just the share or the parents. Women will dier in their productivity (in outside as well in household work) but not in their child rearing capability. The irst order conditions o the maximisation problem [1], or two women i and j, yield thereore the ollowing equalities: i M i i h ' = = i y ' i h ' l' F X j h ' l' = = F X j h ' = j y ' j M j i y ' = y j ' and i h ' = h At the optimum, the ratio between the marginal utilities o market and household goods must be equal to the ratio between the marginal productivity in household work and that in outside work, and to the ratio between the marginal cost o rearing one additional child (expressed in terms o time inputs) and the marginal beneit (measured in terms o the market goods received net o the goods consumed by the additional child). I this set o equalities is satisied across all women i and j, this means that all women will end up with equal marginal j ' [2] 8 The extent to which a woman has the reedom to make autonomously these types o choices vary according to the social and cultural context and to the bargaining power o each spouse. Evidence suggests that each spouse s contribution to household income does not usually occur in a context o joint decision-making and resource pooling; however, in some situations, the wie may have limited choices and her husband decisions may be determinant also in the typical emale domain. 9 The model could also apply to a polygynous amily, where each wie chooses her time allocation and the husband makes a transer to each wie, on the basis o his understanding o the consumption necessities o each wie and her children. sually husbands contribute to part o ood needs (or instance, the staple cereal whereas the wie provides the condiments ) and to lump-sum expenses, such as house maintenance, medical costs, school ees. y( t 10 w ) T m w m NX For instance, M 0 implies F X 7

8 productivities, in outside work and household work respectively. Such equality is the result o mutually beneicial exchanges between women. Fostering out can enhance current utility, i the value o the (market or household) goods produced thanks to the additional time made available is greater than the transer that is made to oster parents. Women with greater productivity in agriculture or higher wage (because they are younger, healthier, or more experienced, or more educated) are thereore more likely to oster out. Conversely, when the opportunity cost o rearing a child is lower than the compensation one can receive by raising other people s children, women may demand to oster in children to increase their consumption. This its with the evidence that young children are typically cared or by older or poorer women, who are less productive in agricultural activities, and are not able to earn a suiciently high wage. Proitable exchanges may thereore occur between women with the same lietime productivity but belonging to distinct generations, as well as between women o the same generation but with dierent productivity/wage proiles. The value o F, derived rom [2] once substituting the appropriate orm or the utility unction, is endogenously determined by the aggregate supply and demand o oster children within the group o households linked by ostering ties. F will be negatively related to the number o in-oster children and positively related to the number o out-oster children. The results rom the optimisation problem [1] represent an interesting extension o the outcomes rom standard household production models. In particular, i a woman has a greater productivity in outside work than, not only can she devote less o her time to household production (as standard household theory predicts), but she can also oster out children. Thereore the decrease in household production (other than childcare) induced by a rise in emale wage is going to be less pronounced than that predicted by standard household models, other things being equal. 11 p to now, the theoretical model underlying the ostering o young children appears to be a relatively simple exercise. owever, a more complex issue is why, in the ace o the universally observed dierences in women s time opportunity costs and o their need to share some o their tasks, ostering should emerge as a typical response in SSA, but not in other parts o the world. The tentative explanation provided here ocuses on the particular household structure and amily organisation in Sub-Saharan Arican societies, which provide women with ewer possibilities to achieve eicient task sharing (not to speak o risk sharing) with other members o the household. Lack o eicient task sharing within the household is due not only to a luid household structure, marked by requent changes in residential arrangements, but also because o a typology o spousal relationships that does not imply, even normatively, budget communality, shared decision-making and extended reciprocal assistance (Whitehead 1999). The organisation o traditional Arican societies revolves around the supremacy o kinship links, which implies that an individual, even ater marriage, does not loosen the rights and obligations to the lineage: or kinship links to be strong, then, the spousal bond must be weak. 11 The situation becomes more complicated when there are both economies o scale in rearing children (β) and complementarities between child rearing and other household work (α), as in the unction: 1 β ( t ( 1ατ ) ( N F ) ) = h h. For values o α and β high enough, child rearing becomes more compatible with other household work; yet ostering may still occur. 8

9 For women this act is o particular relevance, because o their relatively weaker position in the household, the multiplicity o roles that are required o them and the ewer means available to carry them out. In SSA there is little o the task specialisation between spouses ound elsewhere, according to which the wie takes care o all aspects related to the household economy whereas the husband specialises in outside work and provides the amily with the essential market goods. Women s duties include not only household chores but also procuring consumption goods or the daily needs o the amily, either by producing them directly or by earning cash and buying them in the market. 12 Double roles, outside and inside the household, exercise a considerable pressure on women s time in SSA, and demand the scope or task sharing. But the extent o sharing within the household is minimal, due to the lack o resource pooling and separate decision-making. Thereore, i an Arican woman needs help, she is more likely to turn to her circle o kith and kin, instead o asking her husband or his relatives. Fostering is a component o this vital system o lexible sharing across households, and between women in particular. It cannot be overemphasised that ostering arrangements are vital to women, but only due to the disadvantaged position they ace. Fostering is thus a social institution that helps women to cope with precisely the kind o pressures that society puts on them. Child rearing practices are adapted in a way to attain the most important duty society asks o women, that o bearing many children. ence, the widespread generational specialisation with respect to the unctions o bearing and rearing children, in Arican societies. I women in their reproductive age are to bear many children, it is essential that older women, or young unmarried women, help substantially with the task o rearing them. This situation appears dierent rom that characterising other societies, or instance in Asia, where households tend to be more compact and vertically extended, the spousal budget is jointly held, and wie s responsibilities are more clearly centred on the household domain, also when she works outside. In such contexts, a mother with young children is not expected to carry out many duties incompatible with the task o caring or them. Moreover, child rearing can be more easily arranged within households thanks to the help o other resident members. Other characteristics o SSA societies concur to explain the emergence o ostering practices. The strive to promote kinship links may be a direct causing actor, since the practice o sharing child rearing among the members o the extended amily may be regarded as an instrument by which kin people claim their rights over a typical marital resource. (The reason or the supremacy o kinship links in SSA is a somewhat deeper issue: it may be driven by the goal o maximising the long-term survival o the lineage, by making group s interests prevailing over individuals, or it may be sustained by a gerontocratic system associated with the prevailing extensive agriculture systems.) 13 According to our simple model, in a society where ostering practices are widespread, so that many children are reared by women with lower opportunity costs o time, the average cost o rearing children within society becomes lower. This can be counted as one instance in which the lack o eicient task sharing within the SSA household has a positive eect: women, in the absence o support within the household, are induced to look or help outside, and can thus select child carers out o a wider pool o candidates. owever, the best child carer rom the point o view o women s immediate needs may not be the best rom the point o view o 12 There is a wide variation, across societies and ethnic groups, in the type o contribution expected o women to daily ood and necessities. owever the need or a woman to work to comply to such obligations is uniormly ound across SSA. 13 Caldwell (1978), Caldwell and Caldwell (1987). 9

10 child present and uture welare. The positive eect just mentioned must thereore be weighted against potential harmul consequences or the children themselves. 4. Fostering o older children: adjustment to child labour needs Empirical evidence shows that many oster children work quite long hours (res). owever, ostering cannot be adequately explained as a voluntary exchange by the mere act that older children work. I children are so valuable to the household economy, why would natural parents voluntarily agree to oster them out? The most immediate explanation is that oster parents compensate inancially natural parents, or the child, or the labour they receive: ostering thus unctions as an inormal market or child contribution to household work. 14 In a competitive market situation, the compensation given by oster parents should be equal to the marginal value o child labour (or, i marginal returns rom child labour are constant, should be comprised between the marginal value o child labour to the sending household and, respectively, to the receiving household). This should be a straightorward situation to analyse via economic modelling. owever, many ostering arrangements involving the transer o child labour do not reveal the presence o payments in money or in kind. In act, the existence o compensatory monetary transers is not necessary or a transer o child labour to be beneicial to both natural and oster parents. This may be so or two reasons. The irst is that child and adult labour may be complementary actors in household production, so that child productivity varies according to household size and composition, and thereore households might experience excess supply or demand or child labour. The second is that there may be asymmetries in the way natural and oster parents perceive ostering (the ormer regarding it as a way to sponsor the child into better-o amilies, whereas the latter seeing it as a way to acquire labour services), and ostering can actually satisy apparent incompatible needs. In this section, the notion o complementarity between child and adult work is explored to explain why a demand and a supply o economically active children can arise at the same time; the notion was irst used by Ainsworth (1991), but it is explored here to a greater extent. The next section will deal with the second issue mentioned above. The model in this section makes the simpliying assumption that all resident members o working age contribute to a joint household work activity. This is not the representative situation in SSA households, where each working household member is typically engaged in distinct economic activities, partly individually, partly together with other household members, partly in co-operation with people rom outside the household. Allowing or this more realistic scenario would however add complexity without aecting substantially the aspect o main concern here, e.g. the eects on ostering arrangements o child-adult complementarities. ousehold size S is the sum between the number o adults A, o resident children o working age C, and o young and non-working children I. Working members in the household, both adults and children, participate in a joint household production, which yields a homogenous consumption good, shared among household members according to a ixed rule. Labour supply is ixed or each worker, e.g. there is no choice between work and leisure. 14 This may be compared to the phenomenon o adolescent service occurring in th century England (Laslett 1977). The latter was much more circumscribed with respect to its unctions, in that it concerned children rom low amily backgrounds who wanted to accumulate suicient savings to set up a household on their own upon marriage. 10

11 Children provide a dierent type o services than adults, so that the level o household production is aected by the household s child/adult composition ratio C/A. ousehold production can be considered to be an integrated activity, where basic low-productivity chores are combined with more complex, high-productivity tasks. Children can provide only the ormer, whereas adults can provide both. I children perorm the elementary and tedious tasks, then adults can concentrate on the more productive work. In this respect, child contribution induces an increase in adult productivity. ousehold production is separable in adult and child contribution, and both components are measured in terms o the homogeneous consumption good. Adult production is equal to the product between the number o adults A and a unction, g(c/a), deined over the child/adult ratio C/A. The unction g is a monotonically increasing unction in its domain (its irst derivative is positive), but its second derivative is negative, e.g. additional children induce a diminishing increase in the marginal productivity o an adult. The irst derivative g can be regarded as measuring one component (the indirect one) o the contribution o each child to total household production. The other (direct) component, γ, measures the services made directly available by child work (although γ is here constant or all children, its value is likely to vary according to child gender and age). The household production unction can thereore be written as: ( ) = g C A A γ C where ( / A) ( / A) ( / A) 2 ( / A) 2 dg C d g C g' > 0, g' ' < 0 d C d C [3] Child marginal productivity C and adult marginal productivity A are, respectively, equal to: C A = g ' ( C / A) ( C / A) g' ( C / A) = g γ C A Child marginal productivity is always positive. Adult marginal productivity, instead, is positive only i: ( C / A) > g' ( C / A) C A g / e.g. i per-capita adult production remains greater than the decrease in marginal adult productivity induced by an increase in C/A. A production unction can be easily chosen that always ensures the above inequality. 15 Let us now calculate the cross partial derivative, AC. This indicates how adult marginal productivity varies ollowing a variation in the number o children: [4] [5] AC ( ) A = g' ' C C C A > A 2 0 [6] 15 For instance, the unctional orm ( ) g C A = C ensures that g is always greater than g'c/a. A 11

12 The positive sign o this cross partial derivative (due to the hypothesis that g is negative) captures the desirable notion that additional children induce an increase in adult marginal productivity. Given the above eatures, the problem under consideration is one o a household maximising the level o per-capita current consumption c (obtained as the ratio between total production and household size), 16 by choosing the optimal value o resident children, C*, in any one period, given the number o adults A, o own adolescent children N and o own young children I. The optimal number o children may dier rom the number o own older children N, which is the result o past ertility decisions. A household will either oster in (i C*>N) or out (i C*<N) or will not oster (i C*=N). A ostering arrangement is such that the receiving household satisies child consumption needs while securing its labour: the sending household orgoes both the costs and the beneits associated to a child (but there are no intertemporal consequences). The maximisation problem can be characterised as ollows: ( C / A) g A γc max c = [7] C S The irst order condition o the problem [7] yields: c C = ( g' γ ) S ( ga γc ) = 0 2 S and thereore: g' C * γ = c( C *) [8] A The optimal number o working children, i.e. the one that maximises per-capita consumption, is the one which ensures the equality between marginal child productivity (which includes the indirect and direct component) and per-capita consumption. Since the marginal value o child labour is inversely related to the C/A ratio, a household will oster in (out) as long as the marginal productivity o a child is larger (smaller) than the induced per-capita consumption. This simple ramework shows how ostering has the scope or redistributing eiciently child labour across households with structural dierences in their composition (due to distinct ertility and mortality histories) or in dierent phases o their lie cycle. There is no need in this model or any transer o goods or money to make both parties gain rom the arrangement. Nonetheless, transers may be observed in practice. This may imply two things. Either the parties agree to oster without the participation constraint [8] being ully satisied, e.g. the productivity o the child in the natural home is not lower than average consumption or the productivity in the oster home is not greater than average consumption. Alternatively, the ostering arrangement is serving a wider unction than that o redistributing child labour 16 Dierent weights could have been attached to the consumption shares o adults and children. owever, since this assumption would not have changed the results in a signiicant way, it has not been added to keep the notation as simple as possible. 12

13 eiciently across households, or instance insurance or amily support (a topic deerred until Section 6). Child ostering as a response to imbalances in household size and composition needs to be put in perspective and compared with alternative mechanisms, such as various orms o adult mobility. Adults work moves across households, sometimes on a short-term basis as in the context o work parties, in which riends and group members are called to work in one s ield in exchange or beer or ood or cash (Donhalm 1981). Adults move also or longer period, by migrating in search or seasonal or permanent work, but here the move is mainly driven by the sending household. Child ostering can be regarded as a way o modiying labour and consumption requirements in a way that is intermediate between the two mechanisms o adult mobility just described: it is not as lexible as working parties but it is less costly, and more reversible, than adult migration. I a household experiences a change in either consumption or production needs and predicts it may last or a while, but does not want to commit to a too costly strategy, then child ostering may be contemplated. It is easier to move children than adults across households, given that child work has a more ungible nature, whereas adults human and physical capital is likely to be more household-speciic. Fostering o younger children is another mechanism altering household size and composition, which modiies not only consumption demand but also total needs or household labour (given that children require care). When older children can assist in caring or their younger siblings, there may be interesting substitution eects taking place between ostering out young children and ostering in older children, as examined in the next section. Dierent orms o circulation by household members are not only alternatives, they can also depend on one another. For instance, out-migration by one spouse, by reducing the extent o resource pooling and o reciprocal assistance within the couple, may cause diiculties in raising young children, as well as induce a all in the demand or child labour, since the departure o adults reduces the scope or child work given complementarities. ence amilies may have more incentives to oster out. There are other, quite important implications that can be derived rom condition [8], e.g. those concerning the long-standing debate on the value o children in poor countries. Some empirical studies have claimed that children represent a negative economic asset or their parents, since child work does not always repay child consumption rom birth until marriage. 17 Children s contribution to household income consists o very simple tasks, whose value, it is argued, is lower than the value o the goods they consume. The model in this section suggests that two urther aspects need to be taken into account to determine the net value o child labour. First, a distinction needs to be made between the direct production value o child work (γ), which may be indeed low, and the total child net contribution to household production, which includes the enhanced adult productivity component (g ). A negative direct production-consumption gap (γ -c<0) does not per se imply that total net child contribution (γ g -c) is negative. Second, the same child may represent a positive economic asset or one household but a negative asset or another, 17 Among the early studies, see Mueller (1976). The data used by Mueller rom a number o countries, however, might underestimate children s contribution, since they measure productivity with market wages (which are well known not to relect the value o child work contribution, especially in rural areas) and may not have controlled or parents under-reporting o hours worked by children. Cain (1982) concludes that in Bangladesh the services rom children, at least until they are iteen, are not enough to repay parents or the rearing costs. According to Kyerere and Thorbecke (1991), the contribution o children in Ghana is less than their caloric intake. 13

14 depending on the composition o each household. This highlights that the economic value o children in traditional rural societies depends on the synergies between children s and adults work which arise in each particular household. Theoretical and empirical analysis have so ar given insuicient consideration to some o the complex eatures o household production in Arican societies, such as complementarities between the contribution o members o dierent gender and age, and economies o scale. 18 Yet, economies o scale in household production may be the key to make sense o the apparent inconsistency between the typical inding rom household surveys, that poverty is positively related to household size and ertility, and the conclusion rom in-depth anthropological case studies that larger households are also wealthier (Whitehead 1999). 19 The lexibility with which children can be moved across households might provide an additional reason to why there might be no direct relationship between household welare indicators and ertility or household size. The absence o reliable data on residential changes in household surveys (especially those due to child mobility) as well as on the dierent ways in which household may ace temporary shortages or abundance o labour oten represents an impediment to the analysis o these issues. Theoretical understanding o the ostering phenomenon and an accurate recording o data on child mobility across households might contribute to a great deal to the knowledge o the household economy in SSA. 5. Coexistence between the motives o child labour and child education in the ostering o older children This section examines another scenario in which the ostering o older children may lead to mutually beneicial exchanges between natural and oster parents without the need or compensatory monetary transers, and is based on the evidence rom recent anthropological studies that natural and oster parents have dierent but complementary requirements. 20 Better-o kin need children to perorm house chores, especially i they live in urban areas, since ew household members would do household work, as they either attend school, or work or a wage or enjoy leisure; and hiring external help is expensive. Poorer parents, women without husband s support and rural olk cannot ensure their children a comortable environment at home and an education. They might be prepared to renounce to their child contribution i they eel that a child, by living in a better-o and educated household, can have the possibility o learning the ways o the world and make useul connections. They oster out, not because the child is o zero net economic value to them, but because this may be a better investment in the child. ow these opposing perceptions and needs are reconciled varies across contexts and the interpretation is let open in existing studies. In many instances, children are treated less well 18 Deaton and Paxson (?) are among the irst who pointed to the need to peer into the issue o scale economies within households, not only with respect to consumption, but also, crucially, to production. Some o the articles in Alderman, addad and oddinott (?) deal also with these issues rom both a theoretical and an empirical point o view. 19 To this eect one might need to operate a distinction between household and agricultural production, since economies o scale appear to be more important relative to the perormance o such household tasks as ood processing and preparation, rather than or strictly arm operations, where constant marginal returns to scale more oten prevail this is the reason why the role o scale eects has not been explicitly acknowledged in the joint household-agricultural production unction [3]. unt (1979) and Toulmin (1991) show the absence o economies o scale in agricultural production or Kenya and Mali, respectively. 20 For instance, Isiugho-Abanihe (1983, p.90), Oppong and Abu (1987) and Bledsoe and Isiugho-Abanihe (199?). 14

15 than what hoped or by their parents, they devote most o their time to house chores and very little to school. This may be so, because pre-commitments on the part o oster parents cannot be enorced. In other cases, however, ostering does enable poor children to be sponsored by better-o relatives and to be given opportunities or that personal advancement that otherwise would be precluded to them. The socio-economic gap between the amilies o origin and o destination can thereore be a positive element, instrumental to the child s improvement. nder certain conditions, expected child quality is an increasing unction o the socioeconomic gap, which explains why many parents accept to oster out their children. 21 One element making up or the gap between parents perception is the act that there are various types o children (some work, others go to school, some may be better at some tasks rather than others). Parents may want to substitute temporarily the type o children they have with another type when there are changes in the actors aecting their dierential demand or children. 22 (Younger children are instead a more homogeneous category: they all invariable require time and care, so that a mother who osters out her own children is not prepared generally to mind other people s children. 23 ) One source o changing needs is the variable demand or child labour; and one reason why parents may substitute own with oster children in this respect is their concern or own children s education. Although in principle own children may provide the required labour services, parents preerence structures make them unavailable or that, so that other people s children need to be ostered in. Evidence shows that, on average, oster children work more than own children living in the same household, but they may be better educated than siblings let at home. This could suggest that ostering represents an intermediate choice between child education and work, combining the eatures o investment in child human capital and utilisation o child services. 24 In the model we are presenting, ostering is regarded as a second-best option or enhancing child human capital, whereas the irst best remains attending school in the home o origin. (Also a oster child can receive valuable education but generally this is so when natural parents contribute substantially to school ees, uniorms, books and maintenance expenses, a scenario which is not considered or the moment, as the ocus is on the incentives to oster in and out per se, without monetary transers.) 21 As Bledsoe (1994) puts it: Fosterage strategies o mobility or older children produce important lines o social and geographical stratiication. At each upward step, the children s lower status marked them as domestic servants or more educated or urban amilies. This in turn means that the greater the status dierential between the sending and receiving amilies, the less the oster children will be treated like guardian s own children, and the more hardship they are likely to undergo. The trade-o, o course, is the possibility o an incremental rise in the child s adult status (p.123). 22 Oppong and Abu (1987) speak o a...cultural context in which many children are still obedient providers o services to relatives, while others in contrast are expensive consumers o schooling, books, clothes, pocket money, transport and leisure. They [mothers] all try to put their own children in this second category. I child labour is needed in the home or chores or baby nursing during school hours, they see to it that, as ar as possible, this assistance is provided by other people s children, not their own (p.89). 23 Fostering in and out young children simultaneously could bring an advantage only i mothers spent less time in the care o oster than o own children. But, knowing that her children would be treated with less care, without any particular advantage, no woman would oster out. 24 The trade-o between child schooling and work is an established topic in development studies. Research in the context o Sub-Saharan Arica has identiied a signiicant negative correlation between child schooling and the value o child contribution to amily income, which provides an indirect measure o the price o schooling. This relation is proved to be even more signiicant or amilies that own assets that enhance child productivity. See Chernichovsky (1985) and reerences reported there. An analogous study or India is Rosenzweig and Evenson (1977). 15

16 The model analyses parents decision to oster children aged 7-15, jointly with their choice o time-allocation between market and non-market activities and the choice o whether to educate children or put them to work. The choice variables belong to the women s domain, as in section 3, whereas husband-related choices are typically exogenous or the problem at hand, although important in determining total resource availability. In the ollowing, the joint decision problem will be treated in practice as i it were a woman s decision problem. A woman (couple) maximises a quasi-concave twice-dierentiable utility unction, deined over the quality o her own children, market and household goods consumption: ( Q, M, ) [9] with the customary assumptions on the irst and second derivatives and urther assumptions on interior solutions or Q, M and : i > 0 and < 0, or i = Q, M, ; and Q > 0, M > 0, > 0. ii Since Q does not depend, or assumption, on the number o children ostered in (e.g. a mother cares just or the quality o her own children), there is an asymmetry between the mother s decisions to oster in and out. I own and oster children are not substitutes when they are older, dierent symbols must be used to distinguish between the number o children ostered in and the number o children ostered out. They are now indicated, respectively, by i and o. The index o aggregate quality o own children, Q, is constructed as ollows. Let q<n be the number o own children who live at home, do not work and go to school. By normalising its coeicient to 1, q will also denote the level o the human capital o such children. Let us then indicate by o (where 0 o N) the number o own children ostered out. Their human capital is lower than that o the ormer children, because, although they acquire new skills and useul contacts, they work and they may not inish school. 25 The weight attached to o must thereore be positive but less than unity; such coeicient will be called φ. No human capital instead accrues to children who stay at home and work. owever, parents should derive a greater utility rom children who stay at home or at least perceive the quality o children who stay at home to be greater than that o out-osters, all other things being equal (because control on children is tighter). Thereore, index Q is assumed to increase or each own child living at home, but at a declining rate. The index Q can thus be written as ollows: Q o = q χ o ( N ) ϕ where 0 < ϕ < 1 χ >0 and χ <0 [10] The sum o the number o non-working children staying home and that o children ostered out must be no greater than the total number o children, the dierence being the number o own children who contribute to household production: o q N The second argument in the utility unction is M. Market goods are bought with the woman s income, which is the product between her wage, w, and the amount o time devoted to cash work, t w, at price P. 26 A woman has also to buy a ixed amount o market goods X or each o [11] 25 In this model, oster parents do not incur any additional cost as a result o their enhancing oster children s human capital; this positive eect can be thought o as an unrecognised externality associated with past choices (the investment made by oster parents parents in children s education). 26 M is a vector o market goods where each good is in ixed proportion or any level o expenditure. The vector o prices P is thereore exogenous and independent rom demand levels. 16

17 the children resident in the household, own and in-osters. The woman s consumption o market goods will be: wtw M = P X i o [ N ] Given [12], our initial assumption that M>0 also implies that the number o children ostered i wtw PM o in is always constrained rom above, e.g. < N. PX ousehold goods are treated as an homogeneous entity, produced only with the time inputs o the mother t h, o own children who do not go to school (N- o -q) and o in-osters ( i ). Work and school o own children are taken to be incompatible and the supply o each working child is ixed. The production unction is separable in its two arguments, implying substitutability. Mother s and children s production are measured, respectively, by unctions h and g. For both mother and children, marginal productivity decreases with an increase in labour supply. Children s marginal productivity is lower than woman s marginal productivity, e.g. g <h. ousehold work must at least guarantee the task o taking care o children, which can use either woman s or child s work and is here expressed by a generic unction l, positively related to the number o children to care or. The latter is the sum o own young children (N y ) and o the net number o in-osters ( y can be thereore positive or negative). The amount o household goods, once child care is ensured, is thereore given by: [12] i o ( t ) g( N q) l( N ) = h h y y [13] where: h >0, g >0, h <0, g <0 and - y N y [13 ] e.g. the number o young children ostered out cannot be greater than the number o own young children (the upper constraint on y is taken care o by the assumption that >0, as or i ). Let T be the woman s time available or work, to be divided between market and non-market activities: T = t t [14] w h The mother s utility maximisation involves to choose the time-allocation between market and non-market activities (t w and t h ); the number o young children to oster in or out ( y ); the number o own older children either to oster out ( o ), or to keep at home to attend school (q), or to keep at home or household production (N- i -q); and the number o children to oster in ( i ) or household work. The exogenous variables in this decision problem are: the woman s wage, w; market prices, P; the market goods consumption o each resident child aged 7-15, X; and the number o own young and older children, respectively, N y and N. The woman s decision problem is thereore to maximise [9] subject to constraints [11], [13 ] [14], and the non-negativity constraints on all the choice variables excluding y (which can be also negative). By substituting [14] into [12] and [10], [12] and [13] into [1], we can write the Lagrangian or this problem as: max L = [15] o i th, y,,, q 17

18 18 ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) o i o y y h h o o y y i o h i o h q N q N t T t N q N l q N g t h M X N P t T w = ; ; χ ϕ There are ive irst order conditions: ( ) 0 ' e) 0 ' d) 0 ' ' c) 0 ' ) 0 ' a) = = = = = = = = = = χ ϕ Q M i Q M o M y M h g q L g X L g X L l F L b h P w t L [16] When one obtains internal solutions or each choice variable, and thereore all multipliers are set to zero ( i =0 or each i=1, 6), the irst order conditions yield: ' ' ' ' / ' χ ϕ = = = = = g X g F l P w h Q M [17] According to 17.a), the ratio between the marginal utilities o household and market goods must be equal to the ratio: i) between woman s marginal productivity in household work and her real market wage, ii) between the time costs involved in rearing an additional young child and the share o market goods that one receives as a compensation and iii) between child marginal productivity and child consumption or market goods. These results directly expand the standard results o household economic models on the choice between market and household work. Comparative static indicates that, or instance, the eect o an increase in a woman s wage leads to a reduction o the woman s time devoted to household work (greater h ), which may or may nor lead to a greater substitution o household or market goods, unlike in the standard household production ramework, where an increase in women s wage and a higher demand or more educated children always decreases the consumption o household goods. 27 This is so due to the availability o a large supply o child work through ostering. Evidence rom SSA societies conirms this inding. 27 See Becker (1991). This occurs when the substitution eect associated with the price variation is greater than the income eect. Fertility decline in Western societies has been explained within the same ramework. I children are a household consumption good which requires mother s time, then an increase in mother s income induces a decrease in ertility, and a greater purchase o other goods or which mother s time is less needed ( child quality may be an example).

19 The expression 17.b) states that the rate o marginal substitution o child quality or household goods must equal the marginal contribution o children to household production. Finally, in 17.c), the optimal number o children to oster out is the one in correspondence o the equality between the marginal cost o ostering out (as expressed by χ ) and the advantage o increasing child quality (φ). Corner solutions or any choice variable implies that the correspondent equality is not met, as in the ollowing: t t h h = 0, = T, = 0, q = 0, N o i y 0 = N, = 0, = o q, > 0 1 > 0 2 > 0 > 0 χ' =>< ϕ > 0 > 0 6 M M Q M h' > w / P h' < w / P M g' > X < ( > ) g' > 0 χ' > l' < F i or 3 =>< 5 > ( < ) ( < ) φ or 6 < ( > ) [18] Two results ollow rom what seen so ar. Proposition 1. Parents will have none o their own children working at home, and the only working children will be in-osters, i either o the ollowing conditions is satisied i) χ <φ, or ii) χ(n- o )=0. These are suicient but not necessary conditions, in particular because, as shown, the same result can be obtained under certain cases when χ >φ. Proo: see Appendix 1. Proposition 1 i) states that, when the human capital o own children can be increased only by ull-time schooling or by ostering, there is no point having own children working at home i the beneits rom ostering out overcome the correspondent disadvantages, or any child. Proo in Appendix 1 shows that this is a suicient but not necessary condition. Even when the cost o not having children in the household is greater than the beneit o ostering, parents will still not retain own children or household production i the dierence between the costs and beneits is small enough. Condition ii) states that when parents do not derive any extra utility rom having own children in the household, so that oster and own children can be perect substitute or the tasks required, they will never retain children or household production. Proposition 1 thereore shows that the choice to retain own children at home to work crucially depends on the existence, and shape, o unction χ as a component in the Q index. Corollary 1. A necessary but not suicient condition or parents to oster in and out simultaneously is that φ>0 and χ φ. They cannot oster in and out simultaneously i χ >φ. Proo: see Appendix 1. 19

20 Comparative static exercises show that oster children are more likely to go rom lowerincome parents to higher-income parents. igher-wage mothers are less likely to oster out, because they want their children to be raised at home and go to school. Being more specialised in market production, these amilies employ more intensely other people s children in household production. Thereore, lower-income mothers send children to higherincome mothers, because o the co-existence o a child labour motive on the ostering in side and o a human capital motive on the ostering out side. Fostering in and out is more likely when parents are in the middle o the socio-economic scale (neither too rich nor too poor). Note that the hypothesis that ostering enhances child human capital is essential or the result that parents may oster in and out simultaneously, whereas it is not or the result that children are ostered rom lower-status households to higher status households The value o ostering beyond the instances o immediate reciprocity: A general classiication o ostering arrangements Sections 3-5 were concerned with establishing the conditions under which the transer o a young or older child is mutually beneicial to both oster and natural parents. It has been shown that the ostering o a child does not need to be always accompanied by a compensatory transer, as it may per se involve net beneits to both sets o parents. Although our previous treatment o ostering exchanges was limited to one period, reciprocity can occur over a longer horizon, since a party can accept to incur a cost or orgo a beneit in view o uture advantages. The intertemporal dimension can be built within the previous ostering arrangements without aecting in any substantial way the result that ostering can be sustained as a mutually advantageous arrangement, enabling parents to pursue more eiciently the objectives o minding children, utilising child labour and investing in child human capital. owever, delayed reciprocity makes more sense when a compensatory transer is envisaged than when beneits are generated automatically by the child move, since such beneits are less prone to deerment (but or instance parents may still accept to oster out a child even i they do not experience an excess supply o child labour at the present, only in the expectations o such need in the uture). The competitive value o a delayed transer needs to be greater than the value o what is exchanged at the present, as it must incorporate inter-temporal preerences and the degree o uncertainty o uture outcomes. ncertainty can be positively or negatively related to the price o ostering. I people do not have suicient access to alternative mechanisms or smoothing consumption over time, they may be prepared to incur high costs today (and this includes rearing other people s children) in order to enhance the chance o uture beneits. It is not uncommon or oster parents to expect their oster children to support them, when grown up. The uncertainty aecting such expectations is mainly due either to the child not succeeding in adult lie and not obtaining a remunerative labour, or to the competition between the claims o natural and oster parents. These claims are the more incompatible the lower is the child s degree o economic success, e.g. the smaller is the size o the total cake. owever, rom the natural parents perspective, the decline in the transer with respect to what they would have received, had they not ostered out their child, can be well regarded as 28 The same hypothesis is not needed to explain why lower-income mothers oster children to higher-income ones (the upward trend in ostering ). In this case, a simpliied version o the present model could suice, one where φ=0, and thereore Q=q (child quality can be raised only by keeping children at home and sending them to school). In this dierent scenario, mothers would oster only to adjust actual child labour supply to the desired level, taking into account their degree o specialisation in market and non-market activities and the number o children they send to school. 20