69811 Child mobility and rural vulnerability in Senegal Climate change and the role of children in household risk management strategies in rural Senegal Final project report June 2010 TFESSD World Bank Task Managers: Maurizia Tovo, AFTH2, and Junko Saito, WBI Study conducted by the Fafo Institute for Applied International Studies (Fafo AIS), Oslo, in collaboration with l’École nationale d’économie appliquée (ENEA), Dakar Acknowledgements This report was written by Anne Kielland from Fafo, in collaboration with Ibrahima Gaye from ENEA. We would like to thank the TFESSD and the World Bank for generously sponsoring this project, and especially the task managers Maurizia Tovo and Junko Saito, who supervised the work. A number of institutions and organizations in Senegal were generous with their time in providing input to the project process, and PARRER should be specially mentioned, alongside the Ministry of Family. Furio Rosati of the UCW-project has provided good collaboration and useful comments along the way. The fieldworkers and supervisors recruited by ENEA also deserve special mentioning. Gilberte Hounsounou from the Cabinet Stigmate in Benin provided great help in training the fieldworkers, outlining the practical links between research and operative initiatives, and ultimately contributed to the dissemination of the results. In the last phase of the venture, the work was coordinated with a project financed by the Hewlett Foundation, who paid for the qualitative fieldwork referred to in the report. This work was led by Fatou Binetou Dial from the IPDSN at the Cheich Anta Diop University. The final dissemination workshop was organized in collaboration with Unicef Senegal, who also partnered in the work in the region of Kolda and supported all the efforts of the project generously along the way. Disclaimer The findings, interpretations, and conclusions expressed in this report are entirely those of the authors and should not be attributed in any manner to the World Bank, to its affiliated organizations or to members of its Board of Executive Directors or the countries they represent. Cover The boy on the cover comes from a rural community in Kaffrine, and is one of the children in the tracer sample of this study. His father is a large scale farmer and second vice president in his local community. The child is confided to an uncle and currently in Mbour to learn the Quran. Both the parents and the Marabout have approved the usage of this photo. Contents Introduction .................................................................................................................................................. 1 1. Background and objective for the study ............................................................................................... 2 2. Theoretical underpinnings .................................................................................................................... 5 2.1. Theoretical aspect relevant to West Africa .................................................................................. 7 2.2. The role of climate change in Senegal .......................................................................................... 9 3. Methodology....................................................................................................................................... 10 3.1. Stage 1: The household survey ................................................................................................... 10 3.1.1. The sample .......................................................................................................................... 10 3.1.2. Measuring child mobility: a mother based survey approach ............................................. 13 3.1.3. Assessing climate vulnerability: a combined approach ...................................................... 14 3.2. Stage 2: The tracer ...................................................................................................................... 15 3.2.1. The revisit of rural at-risk clusters ...................................................................................... 15 3.2.2. The focus groups ................................................................................................................. 17 3.2.3. The tracer ............................................................................................................................ 17 4. Descriptive findings of national survey results ................................................................................... 19 4.1. The climate situation .................................................................................................................. 20 4.1.1. Exposure to climate-related shocks .................................................................................... 20 4.1.2. Level of worry about climate-related events relative to other risks and shocks................ 21 4.1.3. Farmers’ experience with rain and temperature issues ..................................................... 24 4.1.4. Variability in rainfall in crucial production months over the last 30 years ......................... 25 4.2. An overview of child mobility ..................................................................................................... 28 4.2.1. Population in the sampling zones ....................................................................................... 28 4.2.2. The sample .......................................................................................................................... 28 4.2.3. Rural child mobility: shares and extrapolated figures ........................................................ 29 4.2.4. With what objectives did the rural children leave? ............................................................ 32 4.2.5. Who decides that the child is to leave? .............................................................................. 34 4.2.6. Conflicts around the relocation decision ............................................................................ 35 2| Child Mobility and Rural Vulnerability 4.2.7. Where the relocated children currently live ....................................................................... 36 4.2.8. Children who lost their mother........................................................................................... 36 4.2.9. Children who lost their father ............................................................................................. 38 4.2.10. Diversification within families ............................................................................................. 39 4.2.11. The attitudes of household heads ...................................................................................... 43 4.2.12. Leaving alone or in group? .................................................................................................. 46 4.3. Main groups of relocated children.............................................................................................. 47 4.3.1. Talibés ................................................................................................................................. 47 4.3.2. Children who leave to marry............................................................................................... 52 4.3.3. Students .............................................................................................................................. 56 4.3.4. Leaving for other reasons ................................................................................................... 59 5. Step 2: Re-visiting risk areas and tracing children to urban areas ...................................................... 63 5.1. The focus groups ............................................................................................................................. 63 5.2. The rural revisit of at-risk households ............................................................................................. 66 5.2.1. Features of the household heads in the at-risk households..................................................... 67 5.2.2. Children and social security ...................................................................................................... 68 5.2.3. Sibling diversification ................................................................................................................ 70 5.2.4. The poverty situation in the at-risk households ....................................................................... 71 5.2.5. Features of the mothers ........................................................................................................... 75 5.2.6. Child relocation in the at-risk households ................................................................................ 77 5.2.7. Child relocation, household poverty and shocks ...................................................................... 79 5.2.8. Regressing child relocation on household poverty and shocks ................................................ 84 5.3. The urban interviews ...................................................................................................................... 86 5.3.1. The relocated children ........................................................................................................ 88 5.3.2. The new households of the relocated children .................................................................. 88 5.3.3. Parental expectations and children’s views ........................................................................ 96 6. Impact analysis .................................................................................................................................... 98 6.1. Determinants of child mobility ........................................................................................................ 98 6.1.1. Main working hypotheses ........................................................................................................ 98 6.1.2. Why use the regression approach?......................................................................................... 99 6.1.3. The overarching hypotheses to test...................................................................................... 100 6.1.4. The regression and the results .............................................................................................. 101 Child Mobility and Rural Vulnerability |3 6.2. Linking child outcomes to climate risks and shocks....................................................................... 104 6.2.1. Shocks and child mobility ........................................................................................................ 104 5.2.2. Econometric framework: propensity score matching methods ............................................. 105 6.2.3. Results of procedure ............................................................................................................... 106 6.2.4. Estimation of the effect of climate-related shocks on adult migration and child mobility .... 107 Conclusion ................................................................................................................................................. 108 Literature recommendations .................................................................................................................... 111 Annex I : Rural population estimates per production (sampling) zone, and per age ............................... 118 Annex II : Cluster sample rural survey. ..................................................................................................... 119 Annex III: Cluster sample, rural re-visit of vulnerable areas. .................................................................... 124 Annex IV. Econometric framework: propensity score matching methods ............................................... 126 4| Child Mobility and Rural Vulnerability Table of figures Figure 1. Map of the main production zones in Senegal. Source : IRD - Cartographie A. LE FUR-AFSEC. 12 Figure 2. Share of communities that have experienced a variety of serious climate-related shocks over the past 10 years, on production zone (in percent within communities). 20 Figure 3. Share of household heads that reported to having experienced serious climate-related shocks over the past 10 years. (in percent among households). 21 Figure 4. Fear of climate-related shocks (in percent). 22 Figure 5. Fear of morbidity and mortality among household members (in percent). 22 Figure 6. Fear of other important shocks (in percent). 23 Figure 7. Share of household heads who worry about climate-related issues, on production zones (in percent) 24 Figure 8. Average rainfall per month for August and September, 1978-2008, on region (source: Agence Nationale de la Météorologie du Sénégal). 25 Figure 9. Rainfall variability per month for August and September, 1978-2008, on region (source: Agence Nationale de la Météorologie du Sénégal). 26 Figure 10. Average temperature for August and September, 1978-2008, on region (source: Agence Nationale de la Météorologie du Sénégal). 27 Figure 11. Temperature variability for August and September, 1978-2008, on region (source: Agence Nationale de la Météorologie du Sénégal). 27 Figure 12. Extrapolated number of mobile children under 18 years of age, on production zone. 31 Figure 13. Extrapolated number of mobile children on age group and gender. 31 Figure 14. Reasons why rural children leave home, on gender. 32 Figure 15. Reasons why rural children leave home, on departure age. 33 Figure 16. Share of child relocation decisions made by whom (in percent within gender). 34 Figure 17. Where the children live, in percent within gender. 35 Figure 18. Where maternal orphans stay, in percent within gender. 37 Figure 19. Expectations to living conditions of relocated maternal orphans, (in percent within category), as compared to the paternal household. 37 Figure 20. Where paternal orphans stay (percent within gender). 38 Figure 21. Expectations to living conditions of relocated paternal orphans (in percent within category), as compared to the maternal household. 39 Figure 22. Number of child outcomes and number of children in the household (percent within each number of children in a household). 41 Figure 23. Share of households that has sent at least one child to school or relocated at least one child, on number of household children. 42 Figure 24. Likelihood of a household having different types of child mobility arrangements on number of household children (percent within each number of children in a household). 43 Figure 25. How do you find the following types of child mobility arrangements? (Percent within arrangement type) 44 Figure 26. If you place a child with a member of the extended family – what are your expectations to them? (Percent within arrangement type) 45 Figure 27. How do you find different types of treatments that children who are placed with extended family or employers may experience? (Percent within punishment type) 46 Figure 28. Number of children relocated to study the Quran on region of origin. 48 Figure 29. Departure age and current for children who left to study the Quran. (Percent among children). 49 Child Mobility and Rural Vulnerability |5 Figure 30. Main quality of the child reported for children staying with Marabous, relocated boys and all boys sample. (Percent within group of children). 50 Figure 31. Where children who currently live with Marabous are located (in percent of children with a known destination). 51 Figure 32. Parents’ assumptions about the situation of the children who stay with Marabous (in percent within children). 51 Figure 33. Age of marriage for girls who left home to marry in the sample (in percent). 52 Figure 34. Share of departed girls reported to have left to marry. 53 Figure 35. Principal virtue of girls marrying early as compared to the all girls sample and the sample of relocated girls. 55 Figure 36. Primary virtue of girls who left to study, as compared to all girls and all relocated girls (percent within group). 57 Figure 37. Primary virtue of boys who left to study, as compared to all boys and all relocated boys (percent within group). 57 Figure 38. Education level of children who left to study (percent within gender and all children). 58 Figure 39. Age composition of children who left to study (percent within gender). 58 Figure 40. Other reasons why children leave home (percent within gender). 59 Figure 41. Age of departure for children who leave for other reasons. 60 Figure 42. Primary virtue of girls who left for other reasons, as compared to all girls and all relocated girls (percent within group). 61 Figure 43. Primary virtue of boys who left for other reasons, as compared to all boys and all relocated boys (percent within group). 61 Figure 44: Household heads’ demand for children and confidence that someone will take care of them when old, relative to current number of children. 68 Figure 45: Who household heads think will take care of them in case they get sick (idiosyncratic shock), in case the community is struck by a drought (covariate shock), and when they get old. 69 Figure 46: Share of household heads that would send all their children to school if there were no expenses related, what so ever, on region. 71 Figure 47: Share of households with a certain monthly income, comparing the national random sample (step one of the research project) and the adaptive sample of vulnerable households in step two. 73 Figure 48: Household heads’ perception of poverty and insecurity. 74 Figure 49: Shock exposure of households in the random sample (step one) and the adaptive sample (step two). 75 Figure 50: “If you had the opportunity, would you have wanted to …� (responses by household heads). 77 Figure 51: Main objectives for children leaving the households (in numbers in of the sample). 78 Figure 52. Average departure age for children leaving for different purposes. 79 Figure 53. Share of households with at least one child sent to study the Quran while living away from the household, wealth indicated by housing material: With and without solid floor, with and without solid roof, with and without solid wall, and finally comparing households with all-solid household materials to those with none. 80 Figure 54. Share of households with at least one relocated child on exposure to climate and non-climate related shocks. 81 Figure 55. Share of households with at least one talibé on exposure to climate and non-climate related shocks.82 Figure 56. Share of households with at least one child relocated for other reasons than marriage, Quranic and formal studies on exposure to climate and non-climate related shocks. 83 6| Child Mobility and Rural Vulnerability Figure 57. Share of households with at least one child relocated on asset ownership. Comparing households owning the asset to those who do not. 83 Figure 58. Income of household of origin and hosting household in percent of households. 89 Figure 59. Household heads’ perception of adequacy of food consumption, other basic consumption, and worry about future consumption in percent of household heads. 90 Figure 60. Utility ownership in households of origin compared to hosting households in percent of households. 91 Figure 61. Identification with a person characterized by the following statement: “Adventure and risk taking are important, to have a passionate life�. 92 Figure 62. Identification with a person characterized by the following statement: “To live in a secure environment and avoid everything that might be dangerous is important�. 92 Figure 63. Motivations of Marabouts versus other household heads for taking in children, in percent of household heads. 94 Figure 64. Comparing what Marabouts and other urban household heads think children’s parental household can expect from them, in percent of household heads. 94 Figure 65. Comparing what Marabouts and other urban household heads expect from child, in percent of household heads. 95 Figure 66. Comparing what Marabouts and other urban household heads expect from the children’s parental household, in percent of household heads. 96 Figure 67. Views and expectations about the benefits of relocating a child. Percent of relocated children’s parents/guardians compared to percent of relocated children. 97 Figure 68. likelihood of a household having relocated children for any reason, for any reason but formal schooling and for Quranic studies by number of major shocks of drought, locust and animal disease they have been hit by. 105 Table of tables Table 1. The potential roles of child mobility in household risk management strategies in West Africa. 6 Table 2. Extrapolation of numbers of relocated children by age bracket and production zone. 30 Table 3. Main objectives for children leaving home, on age bracket and gender. 33 Table 4. Impact of drought on child mobility, controlled for household head features, household poverty, ethnicity, religion and region. 85 Table 5. Multinomial logistic regression of the determinants of child relocation over all, for social and economic reasons (that is, children relocated for formal schooling, marriage and religious studies are removed) and for Quranic studies. 103 Table 6. Impact of climate-related shock on adult migration and child relocation 107 Child Mobility and Rural Vulnerability |1 Introduction The fact that so many children leave their parents has been a rising preoccupation among child protection agencies in West Africa over the past decade. Detected cases of abuse and exploitation of these children have caused concern and raised suspicions of organized child trafficking. Yet, in many cases the relocation of children is an important strategy to protect them and enhance their opportunities, and also to help families manage the many risks to which they are exposed. Any efforts aimed at curbing child mobility should therefore try to reduce exploitation, while simultaneously respect practices that represent vital risk management strategies for children and families with few other options. If children are likely to be made more vulnerable by the risk management strategies adopted by their parents, alternative safety nets should be put in place to protect them. Compared to some neighboring countries, Senegal has a relatively low child mobility rate. Yet, concern has been raised with regards to the vulnerability of girls who work as domestic servants and boys who leave to study the Koran within informal structures. To learn more about their numbers and the reasons why they leave, Fafo and ENEA, supported by the World Bank trust fund TFESSD, in 2009 launched a rural household survey that aimed to quantify child mobility practices and identify the features of the households from which the children come. The vulnerability of rural households to risks and shocks related to climatic conditions like drought, irregular rainfall, animal disease and locust was of special concern. This report presents the main results of this project. 2| Child Mobility and Rural Vulnerability 1. Background and objective for the study Over the past 25 years Senegal has experienced six severe droughts, in addition to floods and locusts, and as much as eighty-five percent of the rural population has been affected. When communities are collectively hit by such shocks, informal mutual insurance arrangements are often depleted since these tend to rely on other individuals who are exposed to the same shocks. In areas were few, if any, public social protection mechanisms are in place, households often have no other choice than to resort to drastic measures. The processes of climate change that, according to predictions, will be experienced by the rural population of Senegal are likely to influence household risk management behavior in two main ways: (a) an increased frequency of shocks may lead to increased risk awareness, and risk awareness influences the development and application of household risk mitigation strategies.1 (b) where mitigation efforts are not sufficient to prevent serious Ways to relate to risk harm, shocks will continue to require the application of crisis Households can relate to risks in coping mechanisms, some of which may place certain household three different ways: members at particular risk. First, they can, before shocks strike, strategize to reduce the likelihood The World Bank report Managing Risk in Rural Senegal (2006) that they occur (prevention concludes that rural households try to prevent risk by investing strategies). in irrigation, and mitigate risk by diversifying income sources Second, they can strategize to try to (crop diversification, combining crops and livestock, involving reduce the negative impact of a shock should it happen (mitigation in petty commerce), as well as self-insurance through social strategies). networks and informal savings clubs (“tontines�). The report Third, they can try to cope with the however underscores that the traditional holding of livestock as consequences after a shock has happened (coping strategies). hedge is ill-suited in relationship to climate shocks, as crop and 1 Note, however, that risk awareness could also lead to denial and apathy (see e.g. Bozzari, 2003). Child Mobility and Rural Vulnerability |3 livestock are often jointly affected. Among coping mechanisms applied, reducing meals, selling animals, non-enrollment of children and temporary migration of household members are specially mentioned. Senegal has a net primary school enrollment rate of around 70 percent, in rural areas almost down to 50 percent (Unicef, SOWC 2009). According to figures from Demographic and Health Surveys (DHS), a large share of children currently lives in households other than that of their parents, and in addition, some may live in places that would not be sampled by a regular household survey. L'Agence Nationale de la Statistique et de la Démographie (ANSD) in Senegal reports that 25 percent of rural children were found to have been working the last week before the national survey of 2005, and 50 percent had been working the last year. Seventy-five percent of girls were involved in housework. This situation is not only a product of binding poverty constraints and traditional social practices. Children both constitute and contribute to build and strengthen the social networks upon which families rely for informal insurance. Their human capital is a crucial part of their household‟s portfolio, and good management of this capital is central to the survival of all household members. This role of children in household management of risks and shocks (social risk management or SRM) is largely ignored in policy documents, and still under-researched. Yet, this role is likely to be prominent in rural West Africa (Bozzari, 2001). In relation to covariate shocks, like climate-related events, effective Crisis fostering and strategic diversification of the human capital of children may turn fostering of children out to be crucial. In Senegal, recent research found that Isiugo-Abaihe describes crisis rural parents of street children refer directly to climate fostering as a way to improve the survival chances of a child by instability as a driving force behind the relocation of removing it from the source of a crisis. children towards areas with a different risk-exposure than Such a crisis can be social or economic, real or imagined. the home community. Unfortunately, this may include Strategic fostering, is more motivated highly risky situations like life in urban streets (Farris; by pull factors than push factors, and 2007). aims to strengthen the prospects of the child and/or the family by offering This study will primarily focus on the mobility of children education, skills and social training, networks and thus social protection. as an example of how child outcomes might be affected by 4| Child Mobility and Rural Vulnerability household vulnerability and related risk management strategies. A broad research literature explores how both crisis fostering and strategic fostering of children have traditionally been important practices in West African families, both among the poor and those better off (e.g. Bledsoe, 1990; Ainswort, 1992; Akresh, 2009 and Isiugo-Abanhie; 1983). It is frequently argued, however, that modernization and new urban environments have changed the risks associated with child relocation, as diminished social control and cohesion can lead to higher levels of abuse, exploitation and violence. It is in this perspective that traditional child fostering and mobility must be revisited as a potential welfare risk for children. For policy purposes it becomes important to distinguish between child mobility associated with severe welfare risks, and child mobility that helps improve the livelihood and prospects for children and their families. Sometimes children may be placed at high risk due to their strategic mobility, while their families may experience less vulnerability thanks to the same relocation. As the child ultimately depends on the well-being of its family, it may become difficult to identify the net consequence of relocation. What is clear is that when the quality of the safety nets of families relies on children being exposed to high risks, there is a case for policymakers to develop social protection programs that can help relieve the child from the burden of having to act as a family social risk management instrument. The goal of this study is to provide information that can help identify measures to reduce the vulnerability of West African children and families affected by climate change, or other covariate shocks like the food and fuel price crises. Thus it first aims to increase the understanding of how rural families manage climate-related risks and shocks and to what extent children are involved and affected. Second, it aims to produce reliable data for Senegal that may have practical program/project implications and that can easily inform policy making in the West African region. The SRM framework (that will be further explained in the next section) is a promising approach to provide an analysis of the systemic inefficiencies that lead to the risk exposure of children, and that way clarify the opportunities for targeted preventive action. Child Mobility and Rural Vulnerability |5 2. Theoretical underpinnings The theoretical underpinnings for the survey of climate impact and child vulnerability in Senegal are outlined in Kielland (2009) and will be briefly presented here. The framework for social risk management (SRM) serves as a point of departure, as first outlined by Holzmann and Jørgensen (2000). The basic assumption of the SRM framework is that households do not sit passively and wait for disasters to happen to them. They do, to their best ability, strategize ex ante in order to (1) reduce the likelihood of being exposed to serious social and income shocks, and (2) reduce the negative impact of the shocks they cannot prevent. Informal strategies based on social networks dominate such efforts in societies where few public social protection arrangements are available (World Bank, 2006). Shocks vary in scope, and those hitting large parts of such a social network (covariate shocks) require different risk management instruments than those that mainly hit one individual or a single household (idiosyncratic shocks). Examples of covariate shocks are droughts, sharp increase in food prices or unemployment rates. Examples of idiosyncratic shocks are the loss of job or death/illness of a significant breadwinner to a household. Some common examples of risk prevention strategies are to involve in less risky production, the seasonal migration of one or more family members, proper feeding and weaning practices, or engaging in hygiene and other disease preventing activities. Examples of risk mitigation strategies are: (1) the management of the household portfolio (multiple jobs to spread risk, investment in human, physical and real assets, investment in social capital e.g. through rituals, and gift-giving), (2) insurance through social networks with customary claim on mutual support (marriage/family, community arrangements, shared tenancy, tied labor), or 6| Child Mobility and Rural Vulnerability (3) hedging (accumulation of convertible assets, and through extended family and labor contracts). Throughout the literature on social risk management, the role of household and family children has been relatively sparsely mentioned, but it has been pointed out that children play the role of “risk management tools� when shocks hit a household that has no adequate mitigative measures in place (Sadoulet at al., 2004). In a recent publication, Lilleør (2008) also underscores the ex- ante role that children may play in the social risk management schemes of households and families. She notably explains how families may diversify the risk exposure of the various members in their social networks of mutuality by sending only some children to school for formal education, while choosing different career paths for the others. Table 1. The potential roles of child mobility in household risk management strategies in West Africa. Type of risks Role of child mobility Idiosyncratic risks: Child relocates to:  loss of life of breadwinner,  work to cover medical bills,  loss of job of breadwinner,  substitute in adult job during illness,  contribute to keep essential asset,  loss of asset essential to survival (e.g. home, livestock),  learn discipline/enable family to dissociate  loss of dignity, with challenging child, Risk prevention  ensure protection from foster parents,  earn dowry/bridal price,  law suits,  honour traditional practice,  not finding a (good) spouse,  rite of passage,  becoming someone who is not consulted,  secure survival,  harm to the child,  earn blessings through hard work.  wreath of Gods and/or ancestors Child relocates to: All risks above, but in addition notably the covariate  diversify the family skills/income portfolio, like (global) unemployment and crop failure,  diversify risk exposure of members in the Risk mitigation increasing food and fuel prices, climate and weather family’s informal insurance network, related risks.  broaden own family’s informal safety net by including the foster family Child relocates to:  start own life (emancipation), Coping All shocks.  save family the cost of subsistence,  support family with income. Source: Kieland, A. 2009. The basic assumption for this study is the idea that children‟s roles in household risk management strategies are indeed many and complex, as exemplified by the case of child mobility in Table 1. Potentially negative outcomes for children with regards to lack of schooling, Child Mobility and Rural Vulnerability |7 harmful labor, risky types of mobility and early marriage are thus not likely to be the mere products of crisis coping or more permanent binding poverty constraints. They also reflect the perceived vulnerability of a household to risks and shocks, and strategies to reduce the likelihood of shocks and limit the damage of those that occur. 2.1. Theoretical aspect relevant to West Africa In West Africa, most local communities have a limited access to public social protection of all sorts. High fertility can be explained in SRM terms as a replacement of such social protection, as children play a range of important roles in the risk management efforts of families, notably through their labor and as strategic actors in the expansion of social networks (Bledsoe, 1990). First, through their labor and their strategic alliances children may help prevent certain idiosyncratic shocks. Working to cover medical bills may save the life or permanent health of a household breadwinner. By substituting for the breadwinner in his/her job during illness a child can help maintain household income and thereby help prevent both job loss and the loss of other household assets. Second, poor households generally have few assets to manage. Having little property and savings limits the possibility for diversified investments, and that way the opportunity of spreading income risks. In this perspective the good management of the family‟s human and social capital becomes the more important: the diversification of the education and social relationships of the various children in a family may to some degree help compensate for the lack of other assets. Third, the effectiveness of a social network as an informal mutual insurance institution depends on the risk exposure of the members of the network. If the members have a relatively similar risk exposure, the network may help protect a household against idiosyncratic shocks but will be less useful to protect against covariate shocks that hit all network members. For instance, if all the members of a mutuality network are farmers living in the same area, they will not be able to help one another in case of a drought. Diversifying the risk exposure of the network members is therefore a way to strengthen the effectiveness of this type of informal insurance arrangements. Sending children away to live, work, study or marry into families with different risk exposures 8| Child Mobility and Rural Vulnerability (either due to income or geography) can help diversify the risk exposure of the members in an informal insurance network. Fourth, while the traditional holding of cattle as security is common in West Africa, high fertility can also be seen as creating a buffer of unrealized child labor. In a possible future crisis the value of this child labor can be used to protect other assets like home and productive property, and thereby secure the survival of the family. Table 1 outlined the roles child mobility may play in the risk management strategies of rural households in West Africa. The model suggest that children could mainly contribute to prevent idiosyncratic risks, whereas covariate risks like climate shocks, raising food prices and global financial crisis can hardly be counteracted by a child. Children can however contribute to mitigate for the negative impact of both covariate and idiosyncratic shocks through their work and mobility. Moreover, since risk management arrangements based on local networks will be of limited value in the case of a covariate shock, relocated children can get a more important role, relatively speaking, and relocating more children can become one of the few effective strategies left to move network members out of the crisis affected area. The model thus suggests that children‟s outcomes may be different if families mainly feeling vulnerable to idiosyncratic shocks, than in those worrying more about covariate ones. In the case of a covariate shock, children may be relocated within in several types of arrangements: children are taken out of school and sent to work, their labor potential used as collateral in lending arrangements, they can be married off into strategically selected families, placed with influential (often religious) patrons and leaders, and the quantity and nutritional value of their food intake can be reduced. They may also be prematurely “emancipated� from their parental household – either in order to relieve the household of the burden of supporting them, to be able to support the family with money earned elsewhere (immediately or in the future), or to provide their labor to relatives and acquaintances who may or may not at some point in the future decide to provide support to the family in return. Sending some children to get established in other places may be a goal in itself: the risk exposure of the mutuality network is diversified, so that when the local community is affected by collective crisis, the household will have someone unaffected to rely upon. Child Mobility and Rural Vulnerability |9 2.2. The role of climate change in Senegal Rural families in Senegal are exposed to multiple risks, both idiosyncratic (for example the illness of household members and death of breadwinners) and covariate ones. Income failure constitutes a constant threat to rural households in West Africa and is generally related to covariate factors. It can be related to the prices of production inputs or prices of family produce, but is more importantly related to bad harvests or reduced livestock production. Pests, plant- and animal diseases pose a threat in rural Senegal, and locusts cause geographically sporadic damages (World Bank, 2006). Drought is a considerable risk, and occurs in a cyclical pattern. Certain areas are also vulnerable to disastrous floods (Saint Louis, Matam, Louga, Tambacounda and Kolda). Documented recent changes in the climate are believed to worsen the risk that such shocks will strike rural household production. Droughts are believed to become more frequent, and more unpredictable rainfall will affect the planning possibilities of farmers, and notably their ability to protect their crops against ill-timed rain. With unpredictable rain follows unpredictable occurrences and patterns of locust, crop and animal diseases. In this analysis, the predicted increase of risk of climate-related shocks is seen as a proxy for increased household vulnerability. Risk management strategies related to current risks are hypothesized to be enforced if such risks increase. The analysis will seek to address the impact of a predicted vulnerability increase on child outcomes that may result from families‟ efforts to manage risks, notably the mobility of children. 10 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 3. Methodology Data collection for the project was done in two stages. Stage one consisted in a large scale rural household survey covering the entire country. Stage two was more complex and consisted in i) revisiting areas identified with high child mobility to collect more detailed household information, ii) organize focus groups in the re-visited communities to add depth to statistical findings, and iii) trace relocated children to their new households in urban areas. 3.1. Stage 1: The household survey The initial rural household survey covered 2,400 households. The survey design had two main focuses: first, to get a good overview and basis for extrapolation of figures for child mobility away from rural households, and second, to combine some different approaches to get different assessment of the climate vulnerability of the households surveyed. 3.1.1. The sample The survey uses production zones as main reporting areas (strata), since the principal agricultural productions in the country are differently affected by climatic conditions. The five reporting areas used correspond to a traditional division of Senegal for the purposes of agricultural surveys. (a) the three most climate-exposed productions in the country; groundnut, cotton and livestock (World Bank, 2006), (b) the rice-producing areas that are associated with a particularly high likelihood of child mobility towards the streets of Dakar (UCW, 2007), and (c) a residual zone (the Niayes area). Based on discussion with partners in Senegal, this was an adjustment with regards to the initially planned methodology. The original proposal suggested only surveying three of the production Child Mobility and Rural Vulnerability | 11 areas (livestock, cotton growing and groundnut), while discussion with partners led to a decision to cover all the five production regions of Senegal. In addition to the benefits of applying a standard division model for reporting zones developed for Senegal, it was pointed out that the southern rice producing zones appeared to be particularly high risk areas for child mobility towards the Dakar streets, as suspected by recent censuses of street children in Dakar and Thies (UCW/ENEA/Fafo). In addition, the residual areas of Niayes provide a useful control. Sampling from all five reporting areas would moreover provide the basis for the extrapolation of national figures for rural Senegal. Another adjustment to the original proposal relates to the number of clusters and the number of households visited. The original study design suggested that a minimum of 30 clusters be drawn from each reporting area to ensure representativeness. This would have given 150 clusters for five reporting areas, with 20 household interviews in each one.2 It was concluded, however, that since climate conditions are a highly clustered phenomenon, a higher number of clusters was desirable in order to improve reliability. This had to come at the expense of sample size, since adding clusters is more expensive than adding households once in a cluster. In the end, 40 clusters were drawn from each of the five zones.3 From each of the resulting 200 clusters, 12 households were randomly drawn, leaving the approximate number of households at 480 per reporting area,4 and a total sample size of 2400 households.5 (The second round of data collection however provides almost 600 additional household interviews from at-risk households, so in that respect the initial ambition of 3000 rural household interviews is achieved.) 2 In comparison, DHS Senegal used 21 interviews per cluster. 3 The central statistical agency (ANSD) divides Senegal into roughly 10,000 clusters. 4 Existing census lists of the clusters were updated before the household sample was drawn. 5 One cluster from the original sampling list from the ANSD was replaced, as it proved inaccessible during the spring period. This was a cluster in the groundnut zone, in the region of Fatick, department of Foundiougne, and Arrondissement Djilor. The community Djirna was replaced by the neighboring Djilor. Replacement was also used at the household level. This time, each community was oversampled. When a household refused to be interviewed, the household head or an able representative was not located, or the household not found, the next household in line on the sample list was approached. 12 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y The map in Figure 1 shows the five main agro-ecological zones in Senegal. Based on this, the national statistical agency (ANSD6) operates with a division of the country into five production zones: Figure 1. Map of the main production zones in Senegal. Source : IRD - Cartographie A. LE FUR-AFSEC.  the cotton zone in the south-east, comprising the departments of Kolda, Velingara and Tambakounda,  the northern sylvo-pastural zone of Louga, Kebemer, Linguere, Ranerou and Bakel,  the rice-growing areas in the south west and along the northern and eastern rivers towards the Mauritanian border comprising the departments of Dagana, Podor, Matam, Kanel, Bignona, Ziguinchor, Oussouye and Sedhiou,  the central groundnut areas of Bambay, Diourbel, Mbacke, Fatick, Gossas, Foundiougne, Kaffrine, Kaolack, Nioro, Mbour, parts of Tivaouane and Thies, and 6 ANSD; l'Agence Nationale de la Statistique et de la Démographie Child Mobility and Rural Vulnerability | 13  the north-western Niayes area, comprising parts of Tivaouane, Thies, Kebemer, Louga, Rusfique and St. Louis. 3.1.2. Measuring child mobility: a mother based survey approach Child mobility measurements are normally derived from regular household survey data, and thus based on where relocated children currently live. Some such surveys have information on whether a child‟s mother and father live in the household, allowing a relatively precise identification of relocated children. Other have not, and in those cases, household members under a certain age who are not the sons or daughters of the household head are often assumed to live away from their parents and counted as mobile or “foster children�. There are two main weaknesses of this data approach to child mobility. First, it allows for studying the determinants for fostering children in, while making it impossible to study the household determinants for fostering children out. Second, it is likely that many of the most vulnerable relocated children stay in places that would not be eligible for sampling in a regular household survey (streets, markets, shelters, worker‟s shelters, Quranic schools etc.). Moreover, not all foster children living in regular households are reported as regular household members by survey respondents. The latter would presumably be the case for many child domestic servants. LSMS surveys contain certain questions related to relocation away from a household, but the available information about such children is relatively scarce. The survey presented here has children as primary sampling unit. To maximize information about each child, it treats mothers as the “representatives� of children, reporting on their children‟s behalf. In practice, this means that all mothers in the households sampled are asked to list all children they have given birth to and provide information about each one; either they live in the same household or elsewhere. To ensure the inclusion of maternal orphans, all household fathers of children who have lost their mothers are asked to report on the behalf of their maternally orphaned children. Likewise, all maternal grandmothers are asked to report if they have grand children who are double orphans. Only one type of grandparent could be chosen to avoid overlap and extrapolation challenges, and in Senegal, the maternal grandmother was identified as most likely to be both alive and most knowledgeable about the grandchildren. The 14 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y approach results in a data set where children born from rural parents constitute the main unit of analysis. Information about their parents, households and communities are then added. 3.1.3. Assessing climate vulnerability: a combined approach Measuring climate vulnerability at the household level is challenging, especially because the phenomenon is deemed to be highly clustered. While there are several possible approaches, few are able to provide a household with a clear-cut “vulnerability score� that can in a meaningful and non-dubious way contribute to explain social adaptation choices in a regression equation. Therefore, data that could allow for some different approaches were included in the survey in order to be able to view the vulnerability status of households from some different angels. First, exposure to significant shocks over the past 10 years is registered both at the household and the community level. At the household level both idiosyncratic and covariate shocks are reported, while at the community level the covariate shocks are understandably in focus. Both climate-related shocks and other types of shocks are registered, to ensure that an apparent causal relationship between a climate-related event and a certain behavior is not caused by other types of exposure. Second, household heads are asked to rate their level of worry vis-s-vis a range of possible risks. Again, a wide range of possible social and economic risks are assessed in addition to the climate- related ones. Third, farmers are asked to share their perception of rainfall and temperature changes, the impact on farm and livestock production, and their responses to negative impact. Fourth, in a separate data set, rainfall and temperature data per months over 30 years are collected from the 12 meteorological stations in Senegal. Based on GPS coordinates for each station and each of the 200 clusters, estimates of rainfall variability in crucial production months Child Mobility and Rural Vulnerability | 15 can be estimated based on the distance to each cluster. Based on the data, propensity scores for household vulnerability are derived.7 Finally, a set of agricultural and livestock production data is collected for all respondents involved in the agricultural activities. Together with rainfall, rain variability and temperature data, a profit function can construct an estimate for the household‟s production vulnerability to climate. 3.2. Stage 2: The tracer A tracer of certain groups of mobile children requires a substantial sample, that is, the systematic identification of a large number of households of origin, as well as up-dated tracing information. Children who had relocated for social, economic and religious reasons were in focus for the exercise, the religious motives were given a special priority. Children who had relocated to go to formal schools were considered of lesser interest. To be more effective, high risk areas for the targeted forms of child mobility were identified based on the information gathered in the household survey. 3.2.1. The revisit of rural at-risk clusters Based on the household survey, four departments were identified as target areas for the tracer; Louga, Thies, Kaolack and Kolda. There the highest shares of children had been identified as relocated for social, economic and religious reasons. Within the four regions, a total of 52 clusters were targeted on the basis that they had at least one child who had relocated for religious reasons, or, alternatively, at least two children who had relocated for social and economic reasons in the original survey sample. In practice, 4 of the 52 identified clusters ended up being eliminated from the sample, after conflict reemerged in parts of Sedhiou in the Kolda region. This left the researchers with 48 clusters targeted for revisit (see Annex III). 7 A propensity score is the probability of a unit (e.g., person, household, community) being assigned to a particular condition in a study given a set of known covariates. Propensity scores are used to reduce selection bias by equating groups based on these covariates. 16 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Again, twelve households were to be sampled in each cluster, aiming for a total sample of 576 potential households of origin for relocated children. As the objective of the revisit was to maximize the sample frame of relocated children - rather than to get a representative sample - an adaptive sampling approach was selected at the community level. In each cluster the teams started off by revisiting households identified with eligible relocated children in the household survey interviews, here referred to as the “target households� (ménages cibles). This would range from one household in 8 clusters, two households in 10 clusters, three households in 17 clusters, four households in 7 clusters, five households in 3 clusters, and the last three clusters having respectively six, seven and eight targeted households originating from the household survey. The adaptive sampling approach was selected based on the assumption that behavioral practices are clustered; a hypothesized correlation between proximity and behavioral similarity. That is, if a household is found to send children to urban areas, their neighbors are more likely to do the same than would be a randomly chosen household in the same community. The approach consequently consisted in creating a roster of “neighbor households� (ménages voicins) for each target household. In clusters with only one target household, eleven neighbor households were drawn from this neighbor roster. For communities with two target households, five neighbors were drawn for each one, in clusters with three target households; three neighbors were drawn for each, etc. In the five clusters where the number of target households was not dividable by 12, households were randomly selected for neighbor sampling. An exact weighting of the new sample is not possible. However, for comparison to the household survey sample, the neighbor households were given a sampling likelihood that is a function of the sampling probability of its respective target household. For simplicity therefore, each target household was made to share its sample weigh with the number of selected neighbor households. That is, a target household with 11 neighbor household would get 1/12th the sample weight of a regular household in the given cluster, and so would each of the 11 neighbors. A target household with 5 neighbors would get 1/6th the cluster weight, and so would each of the five neighbors, and so on. Still, it is important to underscore that results from the rural revisit survey are merely observations from a selection of at-risk households, and neither representative for rural Senegal nor for the regions they represent. Child Mobility and Rural Vulnerability | 17 The re-visit thus allowed for the creation of a new, and more detailed household data set from a more homogenous sub-sample of households. 3.2.2. The focus groups To help interpret the statistical findings of the household surveys, a team of 4 sociologists was entrusted with the organizing of focus groups and in-depth interviews in the 48 re-visit clusters in Louga, Thies, Kaolack and Kolda. Twelve focus groups were organized with adult men, twelve with adult women, twelve with adolescent boys and twelve with adolescent girls. The adult focus groups centered around understanding the economic situation and risk exposure of households, access to informal safety nets, peoples‟ perceptions of family and the roles of family members, the roles of children and fertility to a family, expectations to children, and finally views on schooling and mobility. The focus groups for children and youth concentrated on understanding young, rural peoples‟ views of their own role in the family, expectations to them, their perception on formal education, mobility and fertility issues. Both children and adults were encouraged to contribute with examples and stories to exemplify their thoughts and perspectives. Interesting subjects for in-depth interviews were identified through their participation in the focus groups, but also by the survey interviewers who reported interesting cases to the sociologist on their team. 3.2.3. The tracer The rural re-visit identified a total of 183 children currently under the age of 18 years who had left to go to an urban area for social, economic or religious reasons. As much as 72 percent (131) of the children were boys, and 94 of these boys had left to study the Quran (70%). The objective of the tracer was to find as many as possible of the relocated children identified. The aim was to compare their living conditions to those in their household of origin and their lived realities to the expectations of their parents. Two teams of 3 interviewers each were in charge of the tracing. Team 1 focused at the areas of Dakar; Pikine, Rusfique and Sebikhotane, while team 2 covered Thiès, Mbour, Fatick, Kaolack, Diourbel, Touba, Louga, Saint Louis, 18 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Ourossogui, Kolda et Ziginchor. As point of departure they had information provided by the children‟s parents, supported by information gathered from resource persons in their home communities. Other community members receding in the urban area helped localize the children in the urban areas, and also the Chefs de quartiers were of great help. The many challenges and the results of this exercise are presented in Chapter 5 of this report. Child Mobility and Rural Vulnerability | 19 4. Descriptive findings of national survey results This chapter presents some of the core descriptive findings of the national household survey. Through a series of graphs and tables it communicates frequency distributions, simple cross tabulations, and some basic comparisons of sub-groups. Some extrapolated figures are also derived from the data, presenting for example child mobility numbers for the rural areas of Senegal on production zone, age group and some sub-categories. The chapter is divided into two main sections. The first section gives a brief overview of the climate situation in the 200 clusters sampled, the communities‟ shock exposure over the last ten years, the worries of the household heads vis-à-vis climate-related and other shocks, the experience of farmers and cattle owners with temperature and rainfall conditions. The section concludes with a presentation of rainfall and temperature conditions for the clusters, based on 30-year monthly data collected from the 12 meteorological stations in Senegal, and estimated based on GPS-coordinates. The second section presents findings related to child mobility. First, a general overview of relocated children is provided, including an extrapolation of numbers on production zone and age groups. Also the main objectives for leaving and current residence is presented. Based on objectives for leaving and current residence, four main groups of relocating children are defined. In the following four sub-sections, each of the groups is described; children who leave to stay with Marabous (mainly boys), children who leave to marry early (mainly girls), children who leave to study (equal gender distribution), and children who leave “for other reasons�, mainly social and family reasons (slight preponderance of girls). Under the second section, two aspects of possible “risk management by children� will be addressed: the destiny of children who have lost their mothers will be compared to the rest of the sample, and also, differences in child mobility choices between households with few children and households with many children will be addressed as possible diversification of children‟s human capital. 20 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 4.1. The climate situation 4.1.1. Exposure to climate-related shocks The interviewers explained to the respondents that a shock should be understood as an event that has led to a serious reduction in holdings, a drastic fall in income or a strong reduction in consumption. The focus groups of ninety percent of the communities sampled stated that the community had experienced one or more serious shocks related to the climate over the past ten years. Twenty-eight percent of the communities had experienced one such shock, 24 percent two shocks, 16 percent three shocks, while 22 percent of communities had experienced 4 grave shocks or more. The nature of the shocks varies greatly by production zone. Figure 2 shows that droughts were most often experienced by cotton and rice producing communities, and also in the Niayes area. Locust is a severe problem in groundnut and rice producing areas, but particularly serious in the sylvo-pastoral zone. A large share of communities in Niayes and the rice producing zone had experienced crop losses due to diseases and other insects, while animal diseases (that may or may not be related to climatic factors) was an overall problem, but particularly grave among livestock producers and in the cotton zone. 80 75 70 63 61 60 60 53 53 50 47 43 40 40 40 37 36 37 33 33 33 28 30 30 23 23 18 20 20 13 15 10 10 10 11 8 7 8 8 8 8 10 5 0 0 Zone Arachidiere Zone Cotonniere Zone Niayes Zone Rizicole Zone Sylvo Total Pastorale Drought Flood Locust Landslide Crop loss (disease/insects) Animal disease Figure 2. Share of communities that have experienced a variety of serious climate-related shocks over the past 10 years, on production zone (in percent within communities). Child Mobility and Rural Vulnerability | 21 At the household level, 30 percent of the household heads respond that the household has experienced a drought that has led to a serious reduction in holdings, a drastic fall in their income or a strong reduction in consumption during the past ten years. Twenty-four percent have been shocked by locust, 23 percent by animal disease, and 9 percent by flood or landslide. Drought 30 Flood 9 Locust 24 Landslide 9 Animal disease 23 0 5 10 15 20 25 30 35 Figure 3. Share of household heads that reported to having experienced serious climate-related shocks over the past 10 years. (in percent among households). 4.1.2. Level of worry about climate-related events relative to other risks and shocks In a more qualitative attempt of measuring the subjective aspects of vulnerability, the survey asked household heads to report on how much they worried about 38 different types of risks and shocks. Their responses should be understood as the combination of how likely a certain shock seems to the respondent, combined with how much harm the respondent fear the shock would cause. Figure 4 - Figure 6 show that the level of worry turned out to be overall high. Climate-related risks cause considerable concern (Figure 4), only surpassed by the worry about household members falling ill or dying. Among the climate-related risks, household heads worried the most about drought and having their crop destroyed by animals or crop diseases. Rain out of season was also an important concern, and also flood, although relatively few communities had experienced flooding over the past ten years. 22 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Drought 79 14 Rain out of season 69 20 Animal damage or crop disease 77 13 Animal disease 68 17 Flood 59 17 Landslide 46 23 0 10 20 30 40 50 60 70 80 90 100 Worried a lot Worried sometimes Figure 4. Fear of climate-related shocks (in percent). Figure 5 shows that the greatest worry among household heads was that a child would fall ill. Generally, loosing family members was a great concern, but the likelihood that they should fall ill is higher, and of slightly greater concern on a daily basis. Illness of a child 81 14 Illness of productive female 75 19 Death of productive female 73 15 Illness of productive male 75 17 Death of productive male 73 15 Illness of elderly person 65 23 Deatht of elderly person 56 26 0 10 20 30 40 50 60 70 80 90 100 Worried a lot Worried sometimes Figure 5. Fear of morbidity and mortality among household members (in percent). Among other concerns, household heads first rank the fear that production costs will increase strongly, while there is also a considerable concern that the prices of their farm produce may Child Mobility and Rural Vulnerability | 23 drop. On the social side, families fear having to take on the responsibility of supporting extended family members in crisis and having to contribute to costly ceremonies. With regards to crime, the theft of crop and equipment is rated higher than the fear of violence against family members. Strong increase in production costs 78 13 Stong price drop for farm products 59 14 Having to support extended family 61 25 Having to contribute to costly ceremonies 56 24 Arbitrary taxation/corruption 45 18 Theft of crop/equipment 70 14 Violence against family members 58 16 Divine wreath/wichcraft 57 21 0 10 20 30 40 50 60 70 80 90 100 Worried a lot Worried sometimes Figure 6. Fear of other important shocks (in percent). Figure 7 shows that worries about some core climate-related shocks are high in all regions, but somewhat lower in the sylvo-pastoral areas. The sylvo-pastoral areas was the most flood exposed zone (see Figure 2), yet household heads worry relatively little about new floods striking. Worries about drought generally reflect previous exposure, while worries about out-of-season rain is overall high and peeks in the cotton producing zone. 24 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 90 81 85 85 77 79 79 80 69 70 69 67 64 70 56 57 60 53 50 40 34 30 20 10 0 Zone Arachidiere Zone Cotonniere Zone Niayes Zone Rizicole Zone Sylvo Pastorale Drought Rain out of season Flood Figure 7. Share of household heads who worry about climate-related issues, on production zones (in percent) 4.1.3. Farmers’ experience with rain and temperature issues Seventy-five percent of the household heads answered a questionnaire module on agricultural production. Among them, 77 percent report to have noticed changes in rainfall over the past 20 years, while 67 percent have noticed changes in the temperature. Eighty-eight percent say that the rainfall in peek season is very important to their production, while eight percent say it is somewhat important. During the last agricultural season, 76 pecent say rainfall had been sufficient at the time of harvest, 13 percent had had too much rain, and 11 percent too little. Eighteen percent say the rain carried on for too long, while 12 percent say it stopped too soon. Thirty five-percent said the crop had suffered from too high or too low temperatures and 44 percent said that it was damaged due to wind, storm, flood or irregular rain. Sixty two percent had experienced crop diseases and 73 percent damage from insects – factors that may or may not be climate-related. Around 500 cattle owners answered the questionnaire module on livestock. Among them, 76 percent assess that their cattle has already suffered under conditions induced by climatic changes, and drought is mentioned in particular by 80 percent. Fifty-five percent say animals have died, 32 percent say they have become more susceptible to diseases, while 12 percent experience reduced milk production. While 49 percent of the owners said they had remained passive vis-à- Child Mobility and Rural Vulnerability | 25 vis the development, 22 percent said they had diversified their activities by intensifying agricultural production. Twenty-one percent had relocated towards more favorable areas, while 7 percent had resorted to selling off animals. With regards to the future, the farmers have split expectations. While around half expect things to continue as now, 27 percent think conditions will get better, while 24 percent think that it will get worse. Seventy-six percent intend to continue production as now also in the future, while the rest consider changing to more drought resistant animals (8 percent), relocate towards more favorable land (12 percent) or transfer to agricultural production (3 percent). 4.1.4. Variability in rainfall in crucial production months over the last 30 years There are twelve meteorological stations spread across Senegal, reporting on temperature and rainfall on a monthly basis. Figure 8 shows average rainfall in the crucial agricultural months of August and September, over the last 30 years. As the numbers show, there is more rain in the south of the country, and least in the north; Podor, Linguere, Matam and St. Louis. 450 411 400 370 350 324 319 321 310 298 300 270 231 231 229 250 192 200 148 143 146 147 135 150 125 107 August 85 89 78 75 91 100 September 50 0 Figure 8. Average rainfall per month for August and September, 1978-2008, on region (source: Agence Nationale de la Météorologie du Sénégal). The predicted impact of climate change is expected to result from, among other, two major trends related to rainfall: reduced rainfall and greater variability, thus unpredictable rain. It is easy to understand how this will increasingly make the life of farmers more difficult in the future. 26 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Figure 17 shows the rainfall variability for each region in the moths of August and September over the past 30 years. Rainfall variability is estimated by dividing rainfall standard deviation over the past 30 years on the average rainfall (as shown in Figure 8). The figure shows that the largest variability appears in the most rain poor areas, making them extremely vulnerable for agricultural production and livestock. 0,80 0,72 0,62 0,60 0,52 0,50 0,60 0,52 0,51 0,49 0,42 0,45 0,43 0,40 0,41 0,39 0,38 0,37 0,38 0,37 0,35 0,40 0,34 0,29 0,27 0,29 0,28 0,20 0,00 August September Figure 9. Rainfall variability per month for August and September, 1978-2008, on region (source: Agence Nationale de la Météorologie du Sénégal). Figure 10 shows average temperatures for August and September across Senegal, based on data from the last 30 years. The rain-poor areas of Linguere, Podor and Matam have the highest temperatures, while Dakar and the southern regions have somewhat lower temperatures. Temperature variability manifests in ‰ (per mil) of a degree, and is thus a relatively marginal figure, seen over only a 30-year span. Figure 11 shows the highest temperature variability in Linguere, Matam and Diourbel, and the lowest in the Western areas of St. Louis, Dakar, Ziguinchor and Cape Skiring. Child Mobility and Rural Vulnerability | 27 32,0 31,3 31,0 30,3 29,6 30,0 29,1 29,0 28,6 29,0 27,8 27,8 27,9 27,7 27,6 28,0 26,6 27,0 26,0 August 25,0 September 24,0 23,0 Figure 10. Average temperature for August and September, 1978-2008, on region (source: Agence Nationale de la Météorologie du Sénégal). 0,035 0,031 0,030 0,030 0,027 0,024 0,024 0,025 0,021 0,020 0,018 0,017 0,015 0,014 0,015 0,013 August 0,010 0,007 September 0,005 0,000 Figure 11. Temperature variability for August and September, 1978-2008, on region (source: Agence Nationale de la Météorologie du Sénégal). 28 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 4.2. An overview of child mobility 4.2.1. Population in the sampling zones In the following section, all extrapolations of global figures will be based on estimated census data for the rural parts of the five sampling areas of the survey. The estimates were made by ANDS and show that children under 18 constitute almost 57 percent of the rural population in Senegal (see Annex I). 4.2.2. The sample There were children in 2,316 of the 2,400 households visited.8 All mothers, all fathers of maternal orphans and paternal grandmother of double orphans were asked to provide information about their children, as long as at least one of the children was below 28 years of age. This was done for two reasons. First, information about shocks over the past 10 years was collected, and children who were below 18 in that period were thus of interest. Second, even though children below 18 were the primary concern of the study, there was a need to know how many siblings they had, the gender composition of siblings and their rank among them. In total, the 4,439 mothers, 126 fathers of maternal orphans and 2 paternal grandparents of double orphans in the sample reported on 17,728 live born children. Among the children registered, approximately 700 had died. A total of 12,274 children were currently under 18 years of age (born in 1991 or later). Eight-hundred-and-two were registered as currently relocated children, that is, they were born in 1991 or later and had left the maternal household before the age of 18. The extrapolations in the sections below are based on these children, rather on the total sample of children who had relocated before the age of 18, but who were not necessarily under 18 years-old today. 8 In comparing outcomes for children, households with no mothers with children below 18 years of age (or fathers of maternal orphans below 18 years of age) will be excluded, since they do not have to make choices with regards to child schooling, child marriage, child labor and child mobility options. Child Mobility and Rural Vulnerability | 29 4.2.3. Rural child mobility: shares and extrapolated figures Estimating the total number of relocated children from rural areas in Senegal is not straight forwards. Taking point of departure in estimated population figures for 2009 (ANSD) column 1 in Table 2 shows the total rural population in the five production zones that were used as sampling strata for this survey. In column 2, the estimated child population is shown on age brackets. Column 3 of Table 2 shows quite some internal variation in the likelihood of children relocating. Relocation is most common from rice, cotton and groundnut producing areas (8,3, 8 and 8 percent respectively), while around 6 percent of children had left in the Niayes area and only 4 percent in the pastoral zones. Children who have left their homes for urban areas will however not be part of a rural census. In Senegal, this is the case for around half the relocated rural children in all age groups. For the purpose of estimating the number of rural born children who have left their homes, census figures need to be adjusted in order to include also these children. The procedure chosen here involves dividing the number of rural children in each location and age bracket on the share of children who remained home plus half the children who left (presumably for other rural areas), and multiplying by 100.9 That way a figure for the number of children born in rural areas result, and that number will be an improved basis for estimating the share of live borne children who have left. Final child mobility figures are estimated as the percentage share of children who have left (column 3) of the adjusted number of rural borne children (column 4). 9 For example, for 10-14 year olds in the groundnut area, the child population of 426,930 is divided on (100 - (11,9 / 2)), and multiplied by 100. Note that this procedure assumes no comparable mobility of children from urban to rural areas. 30 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Table 2. Extrapolation of numbers of relocated children by age bracket and production zone. (3) (4) (5) (1) (2) Share of live Adjusted number of Child mobility 2009 Child born children children born in the figures population Ages population relocated (in %) zone/age group* extrapolated Groundnut zone 0-4 576 355 2,1 582 422 12 133 5-9 518 415 6,6 536 007 35 183 10-14 426 930 11,9 453 993 54 126 15-17 203 298 17,2 222 405 38 214 3 049 497,00 Total 1 724 998 8,0 1 794 826 139 657 Cotton Zone 0-4 168 641 2,1 170 416 3 550 5-9 151 687 6,6 156 834 10 295 10-14 124 919 11,9 132 838 15 837 15-17 59 485 17,2 65 076 11 181 892 279,00 Total 504 732 8,0 525 164 40 863 Zone of Niayes 0-4 77 806 1,0 78 191 770 5-9 69 984 4,0 71 406 2 845 10-14 57 634 7,5 59 872 4 476 15-17 27 444 16,8 29 960 5 033 411 672,00 Total 232 868 5,9 239 430 13 123 Rice zone 0-4 280 487 2,4 283 916 6 857 5-9 252 290 5,9 259 981 15 382 10-14 207 768 12,1 221 136 26 735 15-17 98 936 21,7 110 948 24 024 1 484 059,00 Total 839 481 8,3 875 980 72 999 Pastoral zone 0-4 177 359 0,8 178 095 1 473 5-9 159 529 2,9 161 897 4 736 10-14 131 377 6,7 135 912 9 071 15-17 62 560 10,5 66 029 6 938 938 407,00 Total 530 825 4,3 541 934 22 217 Total 6 775 914,00 3 832 904 4 140 675 288 860 Around one-in-four rural households (120,000 households) have at least one child who is currently living away home. Extrapolations of child mobility figures per production zone (the survey‟s main strata) are shown in the last column of Table 2 and also in Figure 12. Findings indicate that there are close to 290,000 such relocated children in Senegal today. Percentagewise, around 8 percent of children in the groundnut, rice and cotton producing zones have relocated, 6 percent in Niayes while only 4 percent of children in the sylvo-pastoral areas have left. Yet, as the largest share of households is located in the groundnut producing area, half the relocated children, around 140,000, originate from that zone. Around 73,000 come from the rice producing Child Mobility and Rural Vulnerability | 31 areas, 40,000 from the cotton producing areas, 2,000 from the sylvo-pastoral area and 13,000 from Niayes. 1 139657 40863 13123 72999 22217 0 50000 100000 150000 200000 250000 300000 350000 Groundnut area Cotton area Niayes area Rice area Sylvo-pastoral area Figure 12. Extrapolated number of mobile children under 18 years of age, on production zone. Figure 13 shows the age and gender distribution of relocated children. In total, slightly more boys than girls have left home (an estimated 155,000 boys to 134,000 girls). There is a majority of girls among the youngest and oldest children, while boys dominate in the groups of 5-9 and 10-14 year olds. With regards to age, roughly 21,000 children between 0 and 4 years of age live away from home, 69,000 between 5 and 9 years of age, 102,000 between 10-14 years of age and 96,000 between 15-17 years of age (note that the latter group only comprises three age cohorts compared to five in each of the other groups - mobility clearly increases by age). 70 000 61 443 60 000 53 375 50 000 42 269 40 627 42 621 40 000 27 192 30 000 20 000 12 653 8 681 10 000 0 Masculin Féminin Masculin Féminin Masculin Féminin Masculin Féminin Age 0-4 Age 5-9 Age 10-14 Age 15-17 Figure 13. Extrapolated number of mobile children on age group and gender. 32 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 4.2.4. With what objectives did the rural children leave? When asked about the main reason why their children left, 12 percent of household heads answered that the child left to get married, 16 percent that the child left for formal studies, 25 percent said the child was going to study the Quran, while the rest cited other reasons, predominantly related to social and family issues. Figure 14 shows that the motives for leaving are gender-relative: while 22 percent of the girls had left to get married, this was only the case for 3 percent of the boys. On the other hand, 43 percent of the boys had left to study the Quran, while this reason was cited for very few girls. Boys Girls Marriage Other; ;3 Other; 18 Formal Marriage 19 ; 22 School; 17 Social/ Formal family; School; 18 Social/ Koranic 16 family; Koranic school; 40 43 school; 4 Figure 14. Reasons why rural children leave home, on gender. Objectives for leaving are also age related, considering the age at departure (Figure 15). While the youngest children most often leave for social or family-related reasons, this becomes a gradually less relevant motive as the children grow older. Children also leaves early for Quranic studies, and this is the main motive for almost 13 percent of the under-5-year-olds, and more than half of the 5-9-year-olds leaving. In the case of boys, Quranic studies motivated the departure of 25 percent of the under-5-year-olds and more than 70 percent of the 5-9-year-olds, dropping to 27 and 10 percent for the two older groups. Marriage was the main motive for 37 percent of 10-14 year old girls leaving, and for 47 percent of the 15-17 year olds. As the departing children grow older, a gradually higher share is getting married, leave to pursue their formal studies or for other reasons, often to find work or apprenticeships. Child Mobility and Rural Vulnerability | 33 80 70 60 Social/family 50 Koranic school 40 Formal School 30 Marriage 20 Other 10 0 0-4 years 5-9 years 10-14 years 15-17 years Figure 15. Reasons why rural children leave home, on departure age. A relatively equal share of boys and girls are reported to have left to study. Naturally, few marriages are registered before the age of 10. However, in the 10-14 year bracket, 17 percent of the girls who had left had done so to marry, and this number increases to 30 percent of the sample among the 15-17 year olds. Table 3 shows the extrapolated number estimates of children who have left for different reasons. Almost 30,000 girls left specifically to get married, while only an estimated 5,000 boys did so. Note however that many more rural girls are married before age 18, but this may not have been the primary motive why they left home, or they may be married while still staying with one of the parents. Answers such as “emancipation� and “family reasons� may also have captured individuals in this group. Around 50,000 children left specifically to pursue their formal studies, in numbers slightly more boys than girls. Finally, around 66,000 boys and a few girls are estimated to have left to study the Quran and/or stay with a Marabou. Table 3. Main objectives for children leaving home, on age bracket and gender. Boys Girls Total Marriage 4 976 29 085 34 061 Formal schooling 25 660 21 241 46 902 Quranic studies 66 476 5 660 72 135 Social/familial 27 271 54 587 81 858 Other 29 206 24 698 53 904 34 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 4.2.5. Who decides that the child is to leave? Fathers are reported to be the sole decision maker in around 60 percent of the child relocation cases, and mothers in around 10 percent (see Figure 16). There is a noticeable gender difference; while fathers are the sole decision makers in 77 percent of boys‟ relocations, they are so only in half the decisions related to girls‟ relocation. With regards to girls, mothers make the decision in 14 percent of cases, and collaborate with the father about the decision in 22 percent of cases. The children are only reported to be decision makers in a few cases, and to take part in a joint decision with the mother in even fewer. There were no cases reported on a joint decision between a father and a child. 80 74 70 63 60 50 50 40 30 22 20 17 14 12 10 10 10 7 7 2 2 2 4 1 2 1 1 1 1 0 Father Mother Child Father and Mother and All Others mother child Boys Girls All Figure 16. Share of child relocation decisions made by whom (in percent within gender). A further look into the patterns of decision making reveals that the father‟s unitary dominance is particularly strong in the Sylvo-pastoral zone (73 percent of decisions). In the Niayes area only 49 percent of decisions were made by fathers only, while mothers and fathers made almost 23 percent of decisions in collaboration. Decisions were also made in collaboration between mother and child in 5 percent of the cases. With regards to the category of “other� decision makers, almost half the cases involved girls‟ marriages. “Others� are likely to relate to the extended family and the new husband and his family, and sometimes a Marabout. With regards to Child Mobility and Rural Vulnerability | 35 relocations for family and social reasons, mothers were more often involved (14 percent of cases alone and 28 percent of cases together with father). Father, on the other hand, made the decision in 86 percent of the cases where a child was sent to study the Quran, involving the mother in an additional 8 percent of cases. 4.2.6. Conflicts around the relocation decision In the sample of 802 children currently under 18 years of age, conflicts surrounding the relocation decision were reported in 7 percent of the cases. Fourteen cases of child and/or mother disagreeing were reported in the age group of 5-9 year olds, while 15 children in the 15-17 age group were reported to disagree in the relocation decision. Among the 30 cases where children had disagreed to the departure decision, 6 related to marriages, 7 to schooling decisions, 10 to relocation due to social and familial reasons, and 3 to Quranic studies. Eighteen of the 30 children who were reported to have been in disagreement with the relocation decision were girls. Marabou 2 41 Spouse/own family 25 6 Other family 38 26 Grand parents 17 8 Girls Brothers and sisters 9 9 Boys Tutor 6 6 Employer 1 2 Other 2 2 0 10 20 30 40 50 Figure 17. Where the children live, in percent within gender. 36 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 4.2.7. Where the relocated children currently live Figure 17 shows the current residence of the relocated children. The largest group of relocated boys stay with a Marabout (41 percent), while one in three girls and one in four boys are reported to stay with extended family. In addition, quite many children stay with their grandparents; 17 percent of girls and 8 percent of boys. Nine percent of both boys and girls stay with a sibling, while one-in-four of the relocated girls stay with their new spouse or their own family. Only 6 percent of both boys and girls are registered as staying with a tutor, while less than 2 percent stay with employers. 4.2.8. Children who lost their mother The death of a parent is in most cases a significant shock to a household. Child relocation from a household after the death of a parent could thus in some cases be interpreted as a risk management strategy. In the sample 169 children were registered as maternal orphans and therefore represented by their fathers – in two cases, their paternal grandmother, since also the father had died. Among them 31 (18 percent of the group) were living away from the household of the remaining parent: eight in the local community, eight in another rural area, and 15 in a city or town. While this is relatively high, they also have a higher average age than the total sample (10,6 years compared to 7,9 year), and age alone accounts for parts of the mobility likelihood. When the mother dies, boys are more likely to stay with their fathers. Girls comprise more than 80 percent of the maternal orphans who had relocated. Figure 18 shows that while 93 percent of boys stay at home with their fathers, only 68 percent of girls do. In eight percent of the cases girls are relocated to other households in the community, in 9 percent to other rural households and 15 percent to urban areas. Child Mobility and Rural Vulnerability | 37 100 93 80 68 60 Boys 40 15 Girls 20 8 9 2 1 3 0 Home Community Other rural Urban Figure 18. Where maternal orphans stay, in percent within gender. When maternal orphans leave, people outside the household make the relocation decision in almost one-in-four cases. Fathers make the decision in 52 percent of cases, while the child is credited as decision maker in 13 percent of the cases. While three of the girls in the sample were married and two boys live with a Marabou, three also live with tutors and two alone. Fifty-six percent stay with extended family members. In sixty-two percent of the cases the fathers say they think the relocated child would like to return to the household, and in an additional seven percent they believe the child wants to return to the home community. Figure 19 shows that most of the fathers expect the situation of their maternally orphaned children to be better after moving. Thirteen percent, however, expect both current working conditions and future prospects to be worse than it would have been at home. 80 70 67 61 60 52 48 50 44 44 worse 40 27 equal 30 26 better 20 13 13 10 7 0 0 Working condition Revenue Life conditions Future prospects Figure 19. Expectations to living conditions of relocated maternal orphans, (in percent within category), as compared to the paternal household. 38 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 4.2.9. Children who lost their father 221 children were reported to have lost their fathers. Seventeen percent were living away from the household of the remaining parent, very similar to the case of the maternal orphans. While 80 percent of the maternally orphaned children leaving were girls, the outcomes for boys and girls are more equal in the case of paternal orphans (48 percent of those leaving were boys and 52 percent were girls). Among those who leave, two percent stay in the community, six percent in other rural areas, and nine percent go to urban areas. In the sample, one boy and one girl paternally orphaned were abroad. Comparing Figure 18 and Figure 20, shows that boys who lose their mothers have a fairly low likelihood to leave home (7 percent), while boys who lose their fathers have a 20 percent likelihood to leave when the surviving parent is the mother. Girls, on the other hand, have a lower likelihood of leaving the remaining parent when this is the mother (14 percent compared to 32 percent chance for maternally orphaned girls). The boys who leave tend to leave the community all together, 8 percent of the orphaned boys left for other rural areas, and 10 percent for urban areas. 100 86 80 80 60 Boys 40 Girls 20 8 10 9 2 2 3 1 1 0 Home Community Other rural Urban Abroad Figure 20. Where paternal orphans stay (percent within gender). Comparing the expectations to the new situation of relocated paternal orphans (Figure 21) to the maternal orphans in the same situation (Figure 19) expectations seem to higher with regards to relocated paternal orphans.10 It is not clear whether this involves a better offer of relocation arrangements for paternal orphans, or simply reflects the comparison to the situation at home: a 10 The uneven gender compositions of the two groups do not seem to explain this difference, as there are no clear patterns with regards to male and female relocated orphans. Note however that both samples are relatively small. Child Mobility and Rural Vulnerability | 39 home that loses the father may face different difficulties than a home loosing the mother, for example with regards the economic and care situation. If households without a father are considered to be in a more difficult situation than households without a mother, then most other solutions may appear as better. 90 85 81 80 70 64 60 53 50 worse 40 equal 30 26 25 21 better 20 13 11 11 10 6 4 0 Working condition Revenue Life conditions Future prospects Figure 21. Expectations to living conditions of relocated paternal orphans (in percent within category), as compared to the maternal household. Four of the girls in the sample were married, and four of the boys were sent for Quranic studies. Sixty-six percent stay with family members, and seven percent with a tutor or employer. Consistent with the expectations to living condition, a lower share of parents of paternal orphans than of maternal orphans think the child would like to return to the local community. Thirty-five percent thinks the child would like to return to the household (compared to 62 percent for maternal orphans), four percent thinks he/she would like to return to the community (compared to 7 percent). Notably, fifty percent of boys were believed to want to return, compared to only 20 percent of girls. 4.2.10. Diversification within families The number of children varied between 1 and 13 among the 4,452 mothers in the sample. Most households had below 10 children, yet some had more - with a maximum of 47 children in one 40 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y household. Around one-in-five children had left home before age 18, and 40 percent of households had at least one child who had left before that age. An informal safety-net is of little use in the case of covariate shocks if all the network members have a similar risk exposure. In chapter 2 it was hypothesized that families may make different choices for their children as a way to diversify their human and social capital portfolio and that way strengthen their safety nets and smoothen joint (future) income. Perhaps the most prominent indication of this is the fact that few rural families send all their school-aged children to formal school. Indeed, only 15 percent of the households in the sample had sent all eligible children to school.11 While 36 percent have sent no eligible children to formal education, thirty percent have sent less than half their eligible children to school, and twenty percent more than half but not all.12 Diversification seems common when looking at a variety of possible child outcomes with different social and educational functions; staying home without going to school, staying home and starting school, travel to a Marabou, leave to marry, leave to study, leave for apprenticeship and leave for family or social reasons. In the following, only households with more than one child above 6 years of age will be considered, since diversification in choices made on behalf of children requires more than one child, and the schooling choise requires a certain age. Figure 22 gives an overview of the number of child outcomes chosen in households with different numbers of children. The figure shows that households with only two children tend to make similar choices for both. Notably, this 11 Only the school-aged children (above 6 years of age) within each family (removing families with only one school aged child) were counted. There were 101 households with only one child in the sample (4 percent of the entire sample), and they can be divided into two main categories: households where the first child is just born (57 percent), and households with an only-child of more than 6 years of age (43 percent). Notably, children in the latter group were not only-children because the mother had died - the mother is reported as their representative in the survey in all but one case. Way above average, 44 percent of the only-children had been to school, all of them in the home community. An additional 35 percent were at home but had never been to school, while the remaining 20 percent had relocated for a variety of different arrangements. 12 See more on this in the extended module presented in section 5.2. The rural revisit of at-risk households. Child Mobility and Rural Vulnerability | 41 choice is in only 22 percent of the cases schooling, as most stay around the home to help out. Both mobility and diversification starts accelerating for households with 3 children and more. 90 80 2 children 70 3 children 60 4 children 50 5 children 40 6 children 30 7 children 20 8 children 10 9 children 0 10 children Same outcome Two outcomes Three outcomes More than 3 11-12 children outcomes Figure 22. Number of child outcomes and number of children in the household (percent within each number of children in a household). While fertility is associated with poverty, and poverty with low schooling rates, an interesting tendency emerges in Figure 23: the more children between 6 and 17 years of age there are in a household, the more likely it is that the household sends at least one of them to school. This indicates that schooling may not be their first choice: if you have one or two children they may primarily have to help out at home. The more children you have, the more it becomes possible to afford to send one or a few to school, as some children are already taking care of labor tasks that need to be carried out. A household with only two children in the age range only has a 54 percent chance of having sent one of them to school. The graph then shows a strong increase in this likelihood as the number of children increases towards 68 percent for households with 5 children and as much as 77 percent for households with more than 10 children. The likelihood that the household for various purposes relocates children between 6 and 17 years of age also increases strongly by fertility. While the schooling probability was showing a strong 42 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y increase between 2 and 5 child-households, Figure 23 also shows that the likelihood of relocating at least one child increases almost perfectly proportional to the increased number of children in the household. While households with 2 children only have an 14 percent probability of relocating a child, the likelihood for the average household of seven children is of around 38 percent. Households with more than 10 children have more than 43 percent likelihood for child mobility. 90% 80% 77% 70% 72% 66% 68% 67% 69% 63% 62% 60% 54% 52% 50% 44% 43% Mobility 40% 38% 33% School 30% 27% 26% 20% 23% 14% 10% % 2 3 4 5 6 7 8 9 10+ Figure 23. Share of households that has sent at least one child to school or relocated at least one child, on number of household children. Figure 24 shows how various child mobility arrangements become gradually more likely as the number of children increases in a household. The strongest increase is found in early marriages. Previous studies have shown that when girls‟ labor is needed at home, fewer girls relocate. A high number of children around the household may “free-up� some of the girls to join other households through marriage. The share of children who are sent to stay with Marabous also increases strongly by child age. The same argument could be made for the demand of boys‟ labor around the household: when basic needs are covered, boys can be freed up to go study the Quran. Similarly, children of both genders leave for social and family reasons and for “other reasons� presumably when child labor demand is covered around the household. While more children are freed up because basic child labor and social demands are covered at home, likelihood also Child Mobility and Rural Vulnerability | 43 increases that children leave to study – but this only goes to a certain point before the increase seems to even out. The trends are relatively parallel, showing no clear preference for the different arrangements. 30 25 20 15 10 5 0 2 3 4 5 6 7 8 9 10 11-12 13-15 15+ Talibe Marriage Schooling Social/family Other reasons Figure 24. Likelihood of a household having different types of child mobility arrangements on number of household children (percent within each number of children in a household). 4.2.11. The attitudes of household heads All household heads were asked to rate the attractiveness of various child mobility arrangements. Figure 25 shows that most child relocation is viewed positively by a large share of people. Apprenticeships are the most attractive type of child placement, desired by more than 70 percent 44 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y of household heads, and desired under certain conditions by an additional 22 percent. While very few children were in fact found to have left for apprenticeships, it is clear that apprenticeship arrangements are hard to get either due to costs or availability. More than 55 percent were also in favor of placing children with a Marabout in the local community, and only 21 percent did not find this option desirable at all. Placing children with a migrant Marabout came out as less attractive, but still desired by 51 percent of the household heads – 28 percent among them under certain conditions. Different types of entrusting children (confiage, parainage and marainage) were also seen as desirable to around three-in-four respondents. Child labor options elsewhere like domestic work and agricultural work was seen as largely desirable by a majority of the household heads. While Agricultural work seemed attractive to half of the household heads, and another 24 percent on certain condition, domestic work was seen as desirable or desirable on conditions by 61 percent of the respondents. Confiage 43 30 27 Agricultural work 52 24 24 Parainage/marainage 51 28 22 Apprenticeship 71 22 7 With a migrant Marabou 23 28 49 With a Marabou in the community 55 24 21 In domestic work 36 25 39 0% 20 % 40 % 60 % 80 % 100 % Desirable Desirable on conditions Not desirable Figure 25. How do you find the following types of child mobility arrangements? (Percent within arrangement type) If child relocation to the home of relatives opens up a channel of mutuality between the parental household and the new household of the child, then what do the parents expect this channel to provide? Most parents expect the new household to feed and protect the child, although 23 Child Mobility and Rural Vulnerability | 45 percent do not even have this very basic expectation (Figure 26). A slightly higher share expects the new household to help educate the child, identifying this as an even more desirable outcome of the arrangement than food and safety. Forty-two percent expect the new household to help find the child a spouse. On the economic side, 28 percent expects regular support from the household where the child stays. As the following paragraphs will show, relatively few relocated children have the capacity to provide such support. Also, 34 percent of respondents expect the child‟s new family to come to their support if a crisis should strike the parental home. This last observation supports the notion that taking in a child commits beyond food and shelter. The receiving family of the child (and his/her labor) simultaneously becomes part of or reinforces the role they have in the safety net of the child‟s parental home. 100 88 90 77 80 70 60 50 42 40 34 28 30 20 10 0 Nourish and Help educate the Help find the child Send money Support you in protect the child child a spouse regularily times of cisis Figure 26. If you place a child with a member of the extended family – what are your expectations to them? (Percent within arrangement type) The household heads were finally asked about their level of acceptance of certain types of punishment of children who are placed with a member of the extended family. Figure 27 shows that most household heads are critical to punishments in general, and that too much work, food and sleep deprivation were considered almost globally unacceptable. In fact, physical punishment was preferred over reprimanding a child. While reprimands were accepted or conditionally accepted by 35 percent of the respondents, insults (like calling the child a dog) was 46 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y seen as fully unacceptable to 90 percent. Corporal punishment was on the other hand accepted or accepted on conditions by more than half of the respondents, and an occasional slap was accepted or conditionally accepted by 57 percent of the respondents. Labor overload 2 6 92 Sleep deprivation 1 6 93 Food deprivation 1 6 93 Insults 2 8 90 Reprimands 14 21 65 Corporal 20 32 49 punishment Occational slaping 30 27 43 0% 20 % 40 % 60 % 80 % 100 % Acceptable Acceptable under certain conditions Not acceptable Figure 27. How do you find different types of treatments that children who are placed with extended family or employers may experience? (Percent within punishment type) 4.2.12. Leaving alone or in group? Fifty-one percent of households with relocated children had one relocated child, twenty-five percent had two, fourteen percent had three, and 11 percent had four or more. Child relocation as a response to severe shocks are likely to affect more than one child simultaneously, and should thus be expected to cause several children to leave together. In only thirteen percent of the cases two children had left together, and in two percent of the cases three or more children left at the same time. At least without a closer examination of the data, this could indicate that child mobility may be more of an individual strategic than a crisis coping mechanism touching many children at the same time. Child Mobility and Rural Vulnerability | 47 4.3. Main groups of relocated children Children leave for a variety of reasons. Their different motives are likely to affect their welfare in different ways and therefore, the main groups of relocated children deserve individual attention: 1. Children who leave to study the Quran, most of whom stay with Marabous. 2. Children, mainly girls, who leave home to marry early, and mainly stay with their spouses. 3. Children who leave to study, and mainly stay with extended family members. 4. Children who leave for other reasons, also predominantly staying within the extended family structure, but also with employers and tutors. In this section the features of each of these groups will be described more in detail. 4.3.1. Talibés This section focuses at the children who were reported to have left to study the Quran, frequently referred to as talibés. These children constitute one-in-four mobile children, and are overwhelmingly boys. Is sub-section 4.2.4 their number was estimated to 72,000 rural children. The great majority of the children come from the groundnut producing areas of the country. Figure 28 shows that an estimated 41,000, or almost 60 percent come from these regions, the second largest number being children from rice producing zones (13,000 or 18 percent). Table 2 showed that child relocation was least common in the sylvo-pastoral areas. In numbers, only 7,000 children are estimated to have relocated for religious reasons, or about 10 percent of the national number of assumed talibés. The sylvo-pastoral region is however where the highest share of relocated children has left motivated by religious studies. Indeed 60 percent of relocated boys were reported to have left to study the Quran, while a very low share in comparison had left for formal schooling. The Niayes area had both the lowest number of children who had left for Quranic studies and the lowest share of children leaving for Quranic studies among their mobile children (and estimated 3,000 children and 19 percent of their relocated children). 48 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 41433 7769 3226 12771 6936 0 10000 20000 30000 40000 50000 60000 70000 80000 Groundnut area Cotton area Niayes area Rice area Sylvo-pastoral area Figure 28. Number of children relocated to study the Quran on region of origin. In sub-section 4.2.5 it was documented how fathers made the decision to send a son to study the Quran in 86 percent of the cases. Forty-one percent of the children who left to study the Quran left the village together with the Marabou, indicating the likelihood that he had local connections to the child‟s village and possibly family. In 27 percent of the cases the father brought the child to the Marabou, and in 21 percent of the cases other family members did. Around 80 percent of the children who left for Quranic studies currently live with a Marabou. Almost 60 percent of the talibés registered were of Wolof origin, 24 percent Peuhl and 10 percent Serer. Religious motives were also more common among relocated Wolof children (35 percent of relocated Wolof children), than among Peuhl (28 percent of Peuhl children). Almost 6 percent of the households in the sample had sent at least one child to a Marabou. Moreover, the phenomenon seemed relatively clustered. In 60 percent of the clusters, none among the 12 household sampled had sent children to stay with Marabous, in 22.5 percent of the clusters one family in the sample had sent at least one child, in 12 percent of the clusters 2 out of 12 household sampled had sent at least one child, while in 5 percent of the clusters sampled 3 or more among the 12 households sampled had sent children to study the Quran. The clusters with the highest share of children sent to Marabous are found in Linguere, Louga; Kaolack (in particular Kaffrine); Thies, Kolda and Matam. Child Mobility and Rural Vulnerability | 49 The mean age for leaving to stay with a Marabout is 7.2 years. However, many leave even earlier, 14 percent before age 5. As Figure 29 shows, the peak ages for leaving is 6 and 7 years of age, and indeed two-in-three have left before they are 8 years old. Quite few leave after aged 12. The current average age of the children is of 10-and-a-half years, and by age 12 there is a noticeable decline in the number of children who stay with their Quranic tutors. 0,25 0,2 0,15 Departure age 0,1 Current age 0,05 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Figure 29. Departure age and current for children who left to study the Quran. (Percent among children). Partly due to the early age of departure, very few of the children who left to stay with Marabous had ever been to school. While 3 percent had been to a preschool, two percent had finished the first cycle of primary school and 3 percent the second cycle. Sometimes several children from the same household leave for the Marabous, and 5 percent are reported to have left with a sibling. While 68 percent of the households had sent only one child within the 0-17 age bracket, twenty-one percent had sent two and nine percent three children. There were also two households in the sample who had sent 4 and 6 children respectively. The children sent to Marabous often come from 2-3 different mothers within the household. Rather than leaving at the same time, some households appear to operate with an age appropriate for sending their young boys away. This age can vary from around 4-5 to 7-8 years of age. Sixty- five percent of children who leave to stay with a Marabout have birth registration. This is about 10 percent lower than for the sample of all children. 50 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Can any child be sent to study the Quran, or does the child‟s personality affect the choice of such relocation? In the survey, parents were asked to suggest a main virtue, or the most striking positive feature, of each of their children. Comparing the description of the boys who were sent to study the Quran to the all-boys sample and the all-relocated boy‟s sample, Figure 31 shows some clear differences. Most noticeable, the children sent to Marabouts score higher on obedience than the other two samples, and somewhat lower on looks. Compared to the sample of all boys, these children are, similar to relocated children in general, much more rarely reported to be mainly affectionate. Controlling for age, the effect of a child being obedient on the likelihood of being relocated to study the Quran remains statistically significant. 4 Spiritual/moral 3 1 43 Obedient 32 23 4 Responible 9 5 8 Courageous 10 12 14 Affectonate 12 25 1 Strenght 3 4 20 Intelligence 21 19 7 Beauty 11 11 0 10 20 30 40 50 Boys Koranic studies Relocated boys All boys Figure 30. Main quality of the child reported for children staying with Marabous, relocated boys and all boys sample. (Percent within group of children). Twenty percent of the children stayed with a Marabout somewhere in their area of origin. Among those who had left the home area and for whom the destination was known and reported, the largest group had gone to Thies (mainly Mbour), while the two Kaolack districts of Kaolack and Kaffrine received around 13 percent each. The relatively high number of children who had Child Mobility and Rural Vulnerability | 51 gone to Diourbel were generally in the city of Touba, the holy city of the Muride brotherhood. Only 10 percent of the sampled talibés had left for Dakar. Thies 17,9 Tamba 7,1 St Louis 9,5 Sedhiou 1,2 Matam 4,8 Louga 4,8 Kolda 4,8 Series 1 Kaolack 13,1 Kaffrine 13,1 Fatick 1,2 Diourbel 13,1 Dakar 9,5 0 2 4 6 8 10 12 14 16 18 20 Figure 31. Where children who currently live with Marabous are located (in percent of children with a known destination). Asking the parents about their expectations to the child‟s current and future destiny, Figure 32 reveals that majority expect the life and working conditions of the child to be equally hard or harder than at home, they generally think the revenues are better, and the future prospects much better than if the child stayed at home. 90 78 80 70 63 60 50 40 41 Worse 40 32 30 29 Equal 28 30 22 21 20 15 Better 10 2 0 Working conditions Revenue Life conditions Future prospects Figure 32. Parents’ assumptions about the situation of the children who stay with Marabous (in percent within children). 52 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y While less than 2 percent of the households received regular support from the children who had left to study the Quran, 13 percent reported to have received some support in times of crises. Only 8 percent of the households reported to have received money from the child the last 12 months; 60 percent between 5 and 10,000 CFA and 40 percent 12-15,000 CFA. Twenty-eight percent were still supporting their child financially while he was staying with the Marabou. On asking whether the households would expect to receive some support from the child if a crisis struck tomorrow, 97 percent said no. Seventy-seven percent, however, expected the child to be able to contribute to the parental household in the future. Finally, the parents thought that two- out-of-three children staying with the Marabout would like to return home – 60 percent to their parental household, and an additional 6 percent to the home community. 4.3.2. Children who leave to marry Fourteen percent of the children who had left home had as primary objective for departing that they were getting married, the great majority being girls. By the age of 12, more than one-third of the “married�-sample were wed, while two-in three were married at age 14 (See Figure 33). Sixty percent of the girls who marry early have birth registration compared to the total sample average of 75 percent. Average age within the group was 15-and-a-half years of age, 13 years at the time of departure. 20 18 Share of married children 17,3 16 16 14 14,8 13,6 12 10 8 8,6 8,6 6 6,2 4,9 4 2 2,5 2,5 1,2 1,2 0 0 2 4 6 8 10 12 14 16 18 Age of marriage Figure 33. Age of marriage for girls who left home to marry in the sample (in percent). Child Mobility and Rural Vulnerability | 53 Figure 34 shows that early marriage is a more common reason for girls leaving their homes in the cotton producing areas, where marriage was the objective of almost half the girls who left. In the sylvo-pastoral area one-in-three girls who had left did so to marry, one-in-five in the Niayes area. While marriage explains only 17 percent of girls‟ departures from the groundnut areas, the highest number of marrying girls still comes from this very populous area (two-in-four of the girls who left to marry). 60% 48% 50% 40% 36% 30% 22% 20% 17% 13% 10% % Zone Arachidiere Zone Cotonniere Zone des Niayes Zone Rizicole Zone Sylvo Pastorale Marriage as motivation for a girl leaving Figure 34. Share of departed girls reported to have left to marry. Forty percent of the young girls had married someone in the local community, 42 percent someone in another rural area, 14 percent lived in an urban area, and 4 percent abroad (Mali, The Gambia, Guinea Bissau and Guinea). Only two-thirds of the household heads were willing to make an estimate of the age of the spouse of the girl. Deriving from the information given, the average age of the spouse was of 27 – 12 years more than the current average age of their young wives. Figure 16 showed that while the father is normally the decision maker with regards to children‟s departure; in the case of girls‟ father have less to say. Notably, child marriage decisions are in almost one third of cases made outside the household, probably by extended family. In 49 percent of cases the father makes the marriage decision alone, and in 16 percent of the cases the decision is reported to have been made in collaboration between the mother and the father. 54 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y The survey also asked about the level of influence of the family the young girls married in to. In two-thirds of the cases, the household heads indicate that she had married into families with less influence than her own. In only 7 percent of the cases had the young girl, according to her family, married into a family that was more influential. One cause of this could be that many new households have young household heads, and are thus less influential to households of origin with older household heads. In 31 percent of the cases, the young girl was the first wife of the man they married. In 54 percent of the cases he already had one wife, and in 15 percent of the cases he had two wives or more. Eighty-six percent of the girls who left to marry were currently living with their spouses, while 14 percent lived with others, mainly members of their extended family. The girls who married early had a relatively low level of formal education. Four percent had been to preschool, 20 percent had started – but not completed the first cycle of primary school, while 6 percent had completed. Only one in the sample had started – but not completed – secondary education. Apart from age, gender and education level – do certain personality features correlate with early marriage? Similar to the sample of boys who were sent to stay with Marabous, Figure 35 shows that the girls who marry early are in twice as many cases reported to have obedience as their most prominent virtue when compared to the all-girls sample. A second similarity to the boys placed with Quranic tutors is the fact that the girls married early are reported to be far less affectionate than their peers. These girls are also more often described as more responsible. Strikingly few of the girls who left to marry are characterized as intelligent compared to the “all girls� sample. Both the positive impact of obedience and the negative impact of intelligence are statistically significant when controlling for age. Child Mobility and Rural Vulnerability | 55 2 Spiritual/moral 3 1 51 Obedient 33 24 8 Responible 4 2 7 Courageous 5 5 11 Affectonate 18 30 Strenght 2 2 5 Intelligence 21 18 17 Beauty 15 17 0 10 20 30 40 50 60 Married girls' sample Relocated girls All girls Figure 35. Principal virtue of girls marrying early as compared to the all girls sample and the sample of relocated girls. Half the household heads think the girls have better working conditions after they married than they did at home. Only 16 percent believe working conditions are harder. More than half also believe the revenue-potential is better where she is now, while only 9 percent think they are worse. A similar share believes that her life conditions have improved, while only 8 percent think they have gotten worse. None of the household heads think their daughters have worse prospects where she is today, and as many as 84 percent think they are better. Seventeen percent of the young girls continue to support their parental households after they get married, and this is notably girls who live with their spouses. Nine percent still receive support from home, while in 5 percent of the cases the support is described as mutual between the two households. Few of the respondents were able to estimate the amount of support they have received from their daughters over the last year, but the amounts cited vary between 10,000 and 150,000 CFA. 56 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Almost one in four household expected their married daughters to be able to support them if a crisis struck tomorrow. The share of parents expecting support in case of a future crisis increases to more than two-out-of-five. Nine percent of the household heads believe that the married girls would like to return home, while 7 additional percent thinks the girls would like to return to the home community. 4.3.3. Students Sixteen percent of the children who had left their household before 18 had studying as their primary objective for leaving. More boys than girls had left to study; 55 percent boys compared to 45 percent girls. Both boys and girls leaving with the objective of studying move at all ages, with a somewhat larger share leaving between 11 and 14 years of age. While fewer than average among girls who marry early and boys who go to stay with Marabous have birth registration, ninety-five percent of the children who left to study do. In 62 percent of the cases, the father made the decision of sending the child to study. However, the decision was made in collaboration with the mother in 22 percent of the cases, and this is a higher share than for the other types of child relocation. In 8 percent of the cases the mother was sole decision maker, and in 3 percent the child participated in the decision. Five percent of the children were reported to have disagreed in the decision of leaving to study. Children who leave to study generally stay with family members. Ten percent stay in a household belonging to the father, another 10 percent stay with siblings, eleven percent with grandparents, and 50 percent with other extended family members. Six percent also stay with Marabous and 8 percent with their tutors. How do the personalities of the boys and girls who left to study compare to the personality of other relocated children and to boys and girls in general? Figure 36 and Figure 37 show that the scores for boys and girls who left to study are relatively similar. Not surprisingly, half are primarily characterized as intelligent, compared to only one-in-five in the total child population. Children who leave to study are only slightly more obedient than the children who stay home, and much less often characterized as obedient than children who leave to marry or for Quranic studies. Notably, very few of these girls and none of the boys are characterized as primarily Child Mobility and Rural Vulnerability | 57 affectionate (both the results on affection and intelligence are statistically significant when controlling for age). Spiritual/moral 2 1 3 Obedient 33 33 24 Responible 2 4 Courageous 5 5 5 Affectonate 5 18 30 Strenght 2 2 Intelligence 50 18 21 Beauty 7 15 17 0 10 20 30 40 50 60 Girl students Relocated girls All girls Figure 36. Primary virtue of girls who left to study, as compared to all girls and all relocated girls (percent within group). Spiritual/moral 1 3 1 Obedient 26 32 23 Responible 9 5 Courageous 7 1012 Affectonate 12 25 Strenght 3 4 8 Intelligence 49 1921 Beauty 9 11 11 0 10 20 30 40 50 60 Boy students Relocated boys All boys Figure 37. Primary virtue of boys who left to study, as compared to all boys and all relocated boys (percent within group). 58 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Figure 38 shows the educational attainment of the children who left to study. While most are yet early in their education process, a relatively equal share of 20 percent of boys and girls have completed the first cycle of secondary school. Comparing this curve to the age curve of the children who left to study (Figure 39) the progress of the studies appears to be delayed for many. The average age for the group of children who had left to study was 13 years of age. 50 43 45 39 40 35 35 30 23 25 25 19 19 20 20 Boys 20 15 11 Girls 9 7 8 8 9 10 2 1 All 5 0 Preschool Incomplete Complete Incomplete 1. Complete 1. Complete 2. primary primary cycle cycle cycle secondary secondary secondary Figure 38. Education level of children who left to study (percent within gender and all children). 25 20 15 Boys 10 Girls 5 0 2 3 4 5 6 8 9 10 11 12 13 14 15 16 17 Figure 39. Age composition of children who left to study (percent within gender). Similar to the parents of other relocated children, the majority of the parents of the child students think that the life situation, labor conditions, revenue and future prospects are better where the child now lives. Only between 4 and 6 percent think conditions are worse than at home. Child Mobility and Rural Vulnerability | 59 Only 4 percent of the students make financial contributions to their parental household, while 22 percent get support from home. While only 8 percent expect to have the support of the child if a crisis struck tomorrow, 87 percent count on the support of the child in the future. Fifty-five percent of the household heads think the child would like to return home, and an additional 5 percent think the child would like to return to the home community. 4.3.4. Leaving for other reasons Forty-seven percent of the children who had left home had left for other reasons than to study, stay with a Marabout or get married. Forty-two percent of them were boys, and 58 percent girls. Seventy-seven percent had a birth registration, which is slightly above average. Sixty percent of the children who had left for other than the main reason cited “social or family� reasons for leaving. Twelve percent were heading for apprenticeship arrangements, ten percent wanted to support their families financially, while six percent referred to “emancipation� as main motivation. 80 70 70 60 60 50 47 Boys 40 Girls 30 19 All 20 12 12 9 10 9 10 6 6 3 0 Social/familial Apprenticeship Financial support Emancipation Figure 40. Other reasons why children leave home (percent within gender). Figure 40 shows that while more girls than boys leave for social and family reasons, boys more frequently leave for apprenticeships and slightly more often to provide financial support to their families. Among those quoting “emancipation� the great majority were boys. 60 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Most of the children who left for other reasons stay with relatives, while one-in four apprentices stay with their tutors and one-in-three who left to provide financial support stay with their employer. Twenty-nine percent of the children who had left home for other reason had been to school. Among them, most had started but not completed primary school, while a few had completed. Only four percent had more than primary education. This is relatively similar to the sample of all children where 27 percent had attended school and 4 had attended higher degrees than primary. On average the children in the group are 11,3 years old and left home at an average of 7,3 years of age. As shown in Figure 41 a higher share of children (in particular girls) are relocated at a very early age. After that, departure rates are at about 4 percent per age cohort, boys leaving more often around 13-14 years of age. 35 30 25 20 Girls 15 10 Boys 5 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Figure 41. Age of departure for children who leave for other reasons. Children who leave for “other reasons� differ relatively little from children generally with regards to personality. Girls who leave to study and boys who leave to stay with Marabous where often described as obedient, while this is not the case for boys and girls who leave for other reasons. The only clear difference from the all-boys‟ and all-girls‟ samples is again that both boys and girls in this group much less often are described as affectionate (statistically significant also when controlling for age). Here the group resembles other relocated children more than the total boys‟ and girls‟ samples. Child Mobility and Rural Vulnerability | 61 Spiritual/moral 1 3 4 Obedient 25 33 24 Responible 44 2 Courageous 5 7 5 Affectonate 18 18 30 Strenght 23 2 Intelligence 19 21 18 Beauty 15 17 20 0 5 10 15 20 25 30 35 Girls leaving for other reasons Relocated girls All girls Figure 42. Primary virtue of girls who left for other reasons, as compared to all girls and all relocated girls (percent within group). Spiritual/moral 3 1 3 Obedient 26 32 23 Responible 7 9 5 Courageous 10 12 15 Affectonate 13 12 25 Strenght 3 6 4 Intelligence 22 19 21 Beauty 9 11 11 0 5 10 15 20 25 30 35 Boys leaving for other reasons Relocated boys All boys Figure 43. Primary virtue of boys who left for other reasons, as compared to all boys and all relocated boys (percent within group). Compared to the other cases of child mobility, the fathers were less frequently the sole decision maker in the child relocation decision for children who left for other reasons. The mother made the decision in 16 percent of the cases and the mother and father collaborated on the decision in 19 percent of the cases. In five percent of the cases the child (mainly the boys) made the decision 62 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y themselves and in 3 percent of the cases they made the decision in together with the mother. People outside the household were involved in 8 percent of the cases. In half the cases the partners thought the child would have better working conditions after moving, and only 11 percent thought working conditions would be harder. Seventy-two percent expected revenues to be better, and in almost none of the cases were they expected to be worse. Few also thought life conditions in general would be worse, while 57 percent thought they would be better. Again, most parents thought future prospects were better for the child after moving – almost three-in-four thought they would be, while 23 percent expected them to be pretty much as at home. Nineteen percent of the children who left for other reasons were supporting their families somehow, seven percent on a regular basis. Eight percent still received support from home, while in 2 percent of the cases support was characterized as mutual. The amounts received by the parental household varied greatly. Among the children who had sent money home, one-in-three children had contributed with between CFA 1,000 and 5,000 over the last year, one-in-three with between 10,000 and 20,000, while the remaining children had provided from 25,000 up to 60,000. Two children – girls of 15 and 17 years of age – had contributed with as much as CFA 120,000. Interestingly, the only other contributions in this range came from two equally old sisters in the married sample. While only 22 percent expected the relocated children to be able to support them in case of an immediate crisis, eighty-four percent of the children were expected to be able to help in a future emergency. The distribution on boys and girls with regards to these expectations were strikingly similar. In forty-four percent of the cases, the parents thought the child wanted to return to the household, while an additional 8 percent expected that the child would like to return to the home community Child Mobility and Rural Vulnerability | 63 5. Step 2: Re-visiting risk areas and tracing children to urban areas The ultimate purpose of the second stage of this study was to trace rural children who had relocated to urban areas for social, economic or religious reasons. The main aims were to learn about the traceability of those children, compare the life situation of the relocated children to that of their peers back home, and to compare parental perceptions to the lived realities of the children. The tracer study had three main components. First, to organize focus groups and conduct in-depth interviews in the high-risk areas identified in order to better understand and analyze the survey results. Second, a second household and more detailed survey in the same high-risk areas, the main objective being to expand statistical information about households who relocate children and broaden the sampling frame for relocated children to trace. Finally, to trace children towards the most common urban destinations for children from the risk areas identified and conduct interviews with them and their new household heads. In the following, the term “biological household� (ménage biologique) will be used to describe the maternal household of a child, while “hosting household� (ménage d‟accueil) is used to describe the household the child relocated to. The households of this second round of rural data collection will generally be referred to the at-risk households, as compared to the randomly selected rural households presented through the national household survey. 5.1. The focus groups “Everybody wants children, because a person without children is like a car without a spare tire�, says a male focus group participant in the region of Kaolack. The role of children as their parents‟ social safety nets can hardly be more accurately pictured. In the continuation of his statement the man also creates an excellent image of the sense of vulnerability and stigma surrounding the childless, in spite of the family‟s well meant efforts to take care of them: “Me, I have a brother who never had children, and I let him stay with me to avoid hurting him, because those who have no children get easily upset. Indeed, It‟s enough that you say something, and they‟ll reply that it's just because they have no children that you allow yourself to pick on them.� 64 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y The focus group interviews related to this study aimed to place child mobility within the greater context of on the one hand side the general situation of poverty, risks and shocks, and on the other, the perception of family, and the functioning, objectives and roles of family members. In a context of poverty and insecurity, the ability to work (thus including good health) was widely upheld as critical to peoples‟ welfare and survival. In rural areas, the size of a person‟s labor force largely determines his harvest, and prevents him from becoming exhausted by doing all the work by himself. Building a family and having children was subsequently quite openly described primarily as a way to produce labor. Children who can help you work, means that you can rest: now, but most importantly, when you get older. Old people who have no children exhaust themselves from working to sustain themselves. The gender preference for boys is generally explained by the fact that boys stay on with their parents for longer than girls, and by time also supply them with the labor, company and affection of own spouses and children. Daughters, to the contrary, marry into other families, bringing much, although not always all of their labor with them. Children‟s loyalty to their parents is importantly maintained by two factors. First, a globally felt and deeply engraved ethos of life-long indebtedness. Children owe their parents for giving them life, and all the sacrifices involved in raising them and caring for them. In return, they must pay them back, above all, through cash and kin support or by their own labor. They thus provide them the opportunity to rest, and eventually also the intimate care needed when they get old and sick. Second, to enforce the honoring of this obligation, there is a strong religious element involved. Only through the prayers of their parents can the child be ensured success in life and Paradise in the afterlife. Pleasing the parents ensures their prayers. Ensuring labor and securing old age is done by having many children. Religious reasons somewhat inhibits a discussion on whether having many children is a deliberate strategy, as children are widely perceived a gift of God, and the number also a product of his will. (“One cannot talk about this, because only God knows�, focus group Thies). However, respondents describe how a higher number of children increases the likelihood that not all will turn bad and at least one succeed enough to provide well for the aging parents, like more lottery tickets would increase the chance of a prize in the draw. Child Mobility and Rural Vulnerability | 65 Diversification between the labor specializations and locations of a family‟s children has earlier in this report been described as a way to smoothen family income and spread income risk. In line with this, a quest for diversification among children is also widely practiced and described by the focus group participants as a strategy to strengthen the social autonomy of the family. “If each child does a particular activity, the family will not have many financial problems�. (Focus femme, Kolda). Furthermore, to explain the need for diversification the Wolof proverb “kouye raye sa weurseuk da ngua wara tassaro� was used, advising that to search for opportunity, several avenues should be explored. The rural areas visited largely depended on vulnerable agriculture, and the local labor market beyond that is marginal in all regions apart from Thies. Income smoothening and risk spreading for families would in this context involve the mobility of one or some family members towards areas with a different production or labor market -- preferably one not depending on agricultural output. All the families in the focus group study could report to have sent away at least on child, for a variety of reasons. Child mobility was primarily described as economically motivated. Short term mobility in school holidays could for example provide children with school costs for the coming semester or the family with much needed support in a dry season or following a bad harvest. Long term mobility could be explained by a hope of future returns to the child and the family (“Because if he succeeds, he will come to your assistance�). However, child mobility was also explained as a diversification effort. Mobility of some children would diversify their income from that of those who stay, but mobile children also diversify between themselves to ensure a stable support to their parents and younger siblings. “All my children live away from me but they did not focus on one trade. They are traders, drivers, masons, and I thank the Lord because I do nothing in the house now. It is they who provide my daily expenses. So, if I had kept them all at home waiting for the harvest I think we would live in extreme poverty. Good diversification of activities reduces the risk of poverty.� (Focus group, men, Kaolack ). Social mobility and personality development is to some also a motive behind the confiding of boys to Marabouts. The most prominent virtues perceived by respondents are good conduct, respect and sociability. As parents raise their children, the children are expected to reflect the conduct and social skills of the parents - or, alternatively, their foster parents. As with all 66 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y attractive foster parents, staying with a good and respected Marabout is should help refine the character of children as they become inspired by his personal example. An aspect of diversification is also to combine current and future family needs. The most important current need would be for labor, especially in harvest season, while people are generally also interested in having at least one “intellectual� child to help out with family issues, as put by a male participant in Kolda. However, the children who attend the French school do not represent available labor when most needed. That, on the other side, can be provided by children in Quranic schools, who are generally allowed to return to their families during agricultural peak seasons. “However, in Saloum everyone cannot go to the French school because of the farm work. We bring some to the Koranic school so they can help during the harvest and the other will go to French school, and they may have the opportunity to come during the holidays, if they do not have to prepare for examns�. (Focus group, men, Kaolack) 5.2. The rural revisit of at-risk households Forty-eight clusters in the departments of Louga, Thies, Kaolack and Kolda were defined as risk areas for child mobility and thus eligible for step 2 of this research study (se selection methodology described in 3.2.1). In each cluster, 12 households were targeted based on an adaptive sampling approach aiming to maximize the sampling frame for relocated children. Response rates were generally good, with the exception of two clusters in Linguere, a department of the Louga region. In one cluster, one of the target households had moved. In another, one in two villages refused to receive the interviewers as all the men had left for the Dakka de Médina Gounass, leading to a total loss of 6 households. Replacements were not made, and as a result, the effective sample consists of 569 out of the 576 households envisaged. In the 569 households in the new sample there were a total of 1115 parents reporting on behalf of their children. Among them, 1071 were mothers, in 28 cases fathers represented maternally orphaned children, and in 16 cases the maternal grandmother reported on behalf of grandchildren who had lost both their mother and father. Two-in-three mothers were married to the household head, eleven percent his grand-daughters, eight percent in-laws, five percent household heads themselves, three percent daughters, while the remaining seven percent were other relatives. Child Mobility and Rural Vulnerability | 67 It is important to recall that numbers reported in this section are not representative for the rural population the way the data from the household survey were. Instead, they should be read as representing families more likely to practice child mobility living in at-risk, rural areas of Senegal. 5.2.1. Features of the household heads in the at-risk households The sample of vulnerable households was quite similar to the survey sample from the four regions in the first stage of this research project. Nine out of ten household heads were men, with an average age of 50 years. Ninety-three percent of the households were Muslim - 62 percent Tijaniyyah, 24 percent Mouride and 7 percent Xaadir. Four percent were Christian and 3 percent Animist. Four-in-five household heads lived permanently in the household, and, not surprisingly, this is a bit less than found the survey sample from the four regions (87%). Sixteen percent stayed in the household only periodically, and three percent of household heads were rarely present. The average number of household members were 13, the largest household sampled being of 60 members. While three percent were not married, 43 percent had one wife, 38 percent two, and 15 percent 3 or more. Eighty-six percent of the household heads had no formal education, yet, 40 percent were alphabetized. Forty-seven percent of the household heads were Wolof, 23 percent Peuhl, and 22 percent Serer. In the original survey sample for the four regions there was a slightly higher share of Peuhl (30%) and lower share of Serer (14%). Fifty-six percent of household heads stated that they were in good health, and this is lower than in the overall sample (74%), indicating a possible relationship between child mobility and ill-health of the household head that was not seen in the regression of the household survey sample data. Around one-in-four household heads were ill now and then, fourteen percent were often sick, while seven percent were characterized as chronically ill or handicapped. Two-in-three respondents were primarily farmers (lower than in the overall sample where almost three-in-four were farmers), while five percent were cattle holders and 10 percent merchants. The remaining 18 percent were mainly teachers, Marabous and artisans like masons and tailors. 68 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 5.2.2. Children and social security The household heads had an average of 7,7 children, and one-in-four had more than 10. Yet, two-in-three stated that they‟d like to have more children, only one-in-three being happy about the number they had (and only 5 respondents stating they‟d like to have fewer). Only three percent worried that nobody would take care of them when they grew old, one-in-four being somewhat uncertain, and 70 percent felt safe they would be in good hands. Yet, 92 percent said they‟d feel even safer they‟d be taken care of if they had more children, half the respondents stating that having fewer children would make them feel less safe in this respect. Figure 44 shows that the respondents‟ current number of children affects some of these results: the more current children the more content, and the more children the respondent currently has the more confident he is of being taken care of when he gets older. With regards to whether more children would make them more confident, most respondents say yes regardless of current number of children. Also, respondents with more than 15 children are generally less worried of losing some. 100 92 94 92 90 86 79 80 71 71 70 63 60 53 55 50 50 46 40 40 33 27 30 22 20 10 0 Are you content with Do you feel confident If you had more children, If you had fewer your current number of than someone will take would you be even more children, would you be children? (% Yes) care of you when you get confident? (% Yes) less confident? (% Yes) old? (% very confident) 0-4 children 5-9 children 10-14 children 15 + children Figure 44: Household heads’ demand for children and confidence that someone will take care of them when old, relative to current number of children. Child Mobility and Rural Vulnerability | 69 The mothers in the sample had on average 4,2 children, only 9 percent had 8 or more. Similar to the household heads, two-in-three mothers replied that they would like to have more, while one in three were happy about the number they had. Sixty-nine percent thought they were going to have more children. Seventy-five percent of the mothers, slightly more than the household heads, felt certain that they would be well taken care of when they got old, and, similar to their husbands, the more children they had, the safer they felt. Almost like the men, eighty-seven percent said they‟d feel even safer if they had more children, and half of them that they would feel less safe, had they had fewer children. Nobody/myself 1 3 4 God 1 2 1 Government/NGOs 1 22 2 Church/Mosque/Marabou 1 1 Neighbours/friends 3 3 Other relatives 3 12 12 Parents 2 14 16 Wives 1 2 13 Sons 89 41 45 Daughters 3 1 3 20 40 60 80 100 Old age Drought Illness Figure 45: Who household heads think will take care of them in case they get sick (idiosyncratic shock), in case the community is struck by a drought (covariate shock), and when they get old. Asked who the household heads thought would take care of them if they fell ill, 45 percent responded that they expected their sons to do so, while only three percent referred to their daughters (Figure 45). Thirteen percent mentioned their wives, sixteen percent their parents and 12 percent other relatives. When asked who would help in case the community was struck by a 70 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y drought, expectations to children and other family members (especially wives) dropped from 89 percent to 70 percent. Instead, 22 percent of the household heads expected to be helped by the government or by NGOs. Finally asking who would take care of them when they got old, 92 percent were relying on their children, the great majority on their sons. 5.2.3. Sibling diversification During the preparation for this study, stakeholders consulted kept repeating that parents in some respect often encouraged the diversification of the education and/or careers of their children. Or, in their exact words: “People would not want to put all their eggs in one basket.� If your social safety net consists in your closest family, and all your children are farmers, you would have no one to help when the drought comes. If all the children were formally educated, the family would be vulnerable to unemployment in the formal sector. The survey attempted to address this claim two ways. First, costs are often perceived to be the major obstacle to sending children to formal schools. Household heads were thus asked how many children they would have sent to school if there were no costs involved what so ever. Interestingly, only half the household heads said they would have sent all (see Figure 47). Forty-two percent say they would have sent some, while 2 percent said they would have sent all the boys. Five percent said they would not have sent any, regardless if there were no costs at all. In Kaolack, only 21 percent of household heads interviewed said they would send all their children to school if no costs were involved, while in Thies, two-in-three would do so. In Louga and Kolda around half the household heads interviewed would have preferred school for all their children if financially viable. Religious affiliation seemed to matter relatively little, the Mouride being slightly more positive to schooling than the Xaadir and the Tijaniyyah. The Wolof seems slightly less positive to sending all their children to school than did the other ethnic groups. With regards to household poverty, those with the lowest income seem more positive to send all children to school given costs are eliminated. Food insecurity works the opposite way - the less secure, the less willing the households seem to send all their children to school, in spite of no costs being involved. Child Mobility and Rural Vulnerability | 71 70 66,5 60 55,6 49,5 51,5 50 40 30 21,2 20 10 0 All Thies Louga Kaolack Kolda Figure 46: Share of household heads that would send all their children to school if there were no expenses related, what so ever, on region. The second approach to learn about possible efforts of sibling diversification was to ask what household heads would want their children to become, and how do these preferences vary on the number of children they have? The respondents in this survey were given some hypothetical situations to relate to. First, if they had only one son, what would they have preferred for him to become? Around 40 percent mentioned white collar jobs demanding a formal education, like teachers of civil servants. Only 17 percent wanted their sons to become farmers of cattle holders, 14 percent merchants and 10 percent Marabous. Given the hypothetical situation that they had two sons, more than 80 percent of the household heads responded that they would have preferred if the two had different professions. While one- in-four would want one of their sons to farm, none would have wanted both their sons to do so. 5.2.4. The poverty situation in the at-risk households Around half the at-risk household sampled had iron sheet roof on their main building, most of the rest using only straw. Fifty-six percent had cemented floors while 40 percent had only dirt. Fifty-five percent had brick walls, the rest walls made of straw and dirt. In total, one in three main houses had neither solid roofs, walls or floor, while 43 percent had all building parts 72 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y constructed of durable material. This represents a slightly better housing standard than the regular survey sample from the four regions, where only 37 percent of the households had all building parts made out of solid material. The at-risk households sampled in step two of this research generally hold more assets than the random survey sample from the four regions in step one. Seventy-three percent have a radio (compared to 68 percent in the random sample), 22 percent a TV (14 percent), 6 percent a fridge (3 percent), 72 percent a phone (54 percent), and 53 percent a cart (37 percent). Fifteen percent had electricity, and this was lower than in the national random sample where 19 percent did. Twelve percent of the households were estimated to have had an income of less than CFA 25,000 (around USD 50) over the past month (April 2010). One-in-four had an income of between CFA 25,000 and 42,000 (USD 80), while one-in-three had had an income of between 42,000 and 75,000 CFA (USD 150). The remaining 28 percent had earned more that CFA 75,000. Comparing this to the random household sample from step on of this research project, the adaptive sample households is more homogenous as they less often are placed within the poorest and wealthiest income groups, and more often among those with an average income (see Figure 47).13 13 Note however, that monthly income is not measured in the same way in the two surveys, as the more complex, multi-indicators approach applied in step one turned out to give a relatively high non-response rate. The division into income quarters used for step two was based on the income distribution for the entire country, as resulting from step one. Child Mobility and Rural Vulnerability | 73 45 42 40 35 33 30 28 25 25 23 21 20 15 12 13 10 5 0 <25,000 25,000-42,000 42,000-75,000 >75,000 Random sample Adaptive sample Figure 47: Share of households with a certain monthly income, comparing the national random sample (step one of the research project) and the adaptive sample of vulnerable households in step two. Figure 48 shows that more than half the household heads see the current level of food consumption in the household as less than adequate, while only 2 percent describe it as more than adequate. Two-in-three describe household consumption for other basic necessities such as clothing and housing as less than adequate, while again only 2 percent find it more than adequate. More than half of the household heads worry a lot about being able to provide their families with food and other basic necessities in the year to come, while 39 percent say they worry sometimes. Only 8 percent do not worry about this at all. 74 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y More than More than Not at all adequete; adequate; worried; 2% 2% 8% 100 % 90 % Adequate; 80 % Adequate; 34% Worry 42% sometimes; 70 % 39% 60 % 50 % 40 % Less than Less than adequate; 30 % adequate; Worry a lot; 65% 53% 20 % 56% 10 % 0% Do you think that the current Do you think that the current Are you worried about being level of food consumption in level of consumption for other able to provide your this household is... basic necessities in this household with food and household is... other basic necessities in the 12 months to come? Figure 48: Household heads’ perception of poverty and insecurity. As mention previously, to the respondents, a shock was described as an event that has led to a serious reduction in holdings, a drastic fall in revenue or a strong reduction in household consumption. Listing a number of such shocks, only 4 percent of the household heads said that no such event had happened to them over the past 10 years. The shock most often mentioned was a drastic price increase (61%), animal diseases (49%), droughts (36%), crop loss for other reasons (29%), locust (25%), death of a breadwinner (22%), and the illness of a household member (15%). In Kolda and Kaffrine there had in addition been a number of damaging fires. Figure 49 shows that the shock profile is slightly different between the households from the adaptive sample and the random sample from the first part of this study. Some of it may result from the fact that questions were more detailed in part one, and the answers are therefore not Child Mobility and Rural Vulnerability | 75 perfectly comparable.14 Yet, it is worth noticing that more families in the adaptive sample (that is presumably at higher risk of child mobility), have experienced covariate shocks like animal death or illness, locust and slightly more drought. The adaptive sample has, however, to a lesser degree reported on idiosyncratic shocks like the illness and death of family bread-winners. The noticeable exception is that considerably fewer in the adaptive sample respond conformingly to having been severely affected by strong price increases. 90 84 80 70 61 60 49 50 40 36 35 35 36 30 32 28 30 25 22 19 20 15 10 0 Sharp price Animal Other crop Locust Drought Illness of Death of increase disease/death loss household breadwinner member random sample adaptive sample Figure 49: Shock exposure of households in the random sample (step one) and the adaptive sample (step two). 5.2.5. Features of the mothers As mentioned in the introduction to this section, there were 1071 mothers responding on behalf of their children in the adaptive household sample. Two-in-three were married to the household head, eleven percent his grand-daughters, eight percent in-laws, five percent household heads themselves, three percent daughters, while the remaining seven percent were other relatives. 14 In step one there were different questions about animal death and illness, while in step two there is only one. In step one there were questions about work-aged men and women separately, in step two there was only a question about work-aged adults, etc. 76 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Forty-one percent of the mothers live in monogamous marriages, but disproportionally many of them are young and this situation may change in the future. Thirty-eight percent of women have one co-spouse, 16 percent two, while 5 percent have more than two co-spouses. Thies has by far the highest share of monogamous marriages with 57 percent, Louga has 43 percent, while in Kolda and Kaolack only had respectively 36 and 26 percent mothers in monogamous marriages. The study gave specific emphasize to identify factors that could in one way or another influence on the decision making power of women when it came to the outcome for children. To avoid endogenous influences in regressions, factors related to the mother‟s background were investigated. Fifty-one percent of the mothers had their own parents staying in the community where they were living. They would generally describe their parental household as one without influence, or in some cases of similar influence as their current household. In 7 percent of the cases their mothers described their parental households as more influential than the household they married into. Only 13 percent of the mothers had a father who was alphabetized, and only seven percent had a mother who was. As male relatives of the same mother may indicate influential support for a woman, the survey mapped the number of brothers-of-same-mother and maternal uncles. Family members living away from the community may mean an increased opportunity for child out-fostering to a prosperous area. Forty-seven percent of the mothers had brothers who were living in an urban area, and 35 percent had sisters who did. As much as 38 percent of the mothers had themselves at some stage lived in an urban area, and would possibly have experiences with how to find a job and a place to stay and/or acquaintances of importance in such areas. The mothers were well organized. Sixty-four percent were members of religious associations, while 78 percent were also member of some other association or organization. Thirty-nine percent stated that they had revenue from outside the household, and 21 percent owned their own parcel of land. Approximately half of those had brought the parcel with them into the marriage. Sixty-two percent of the mothers reported to be in good health, 25 percent were occasionally ill, 10 percent often sick and 4 percent were living with a disability. Only 16 percent of the mothers had ever set foot in a school - only 4 percent having completed primary school. Yet, 43 percent answered conformingly that they had been alphabetized. Child Mobility and Rural Vulnerability | 77 5.2.6. Child relocation in the at-risk households Many both fathers and mothers are positive to sending their children to other places, the mothers overall slightly less positive than the fathers. The positive attitudes expressed exceed the child relocation actually taking place, indicating that lack of opportunity is a main reason why the phenomenon is not even more widespread. Asking: “- If you had the opportunity, would you send your son/daughter to…� a range of child relocation arrangements seemed attractive. Figure 50 shows that 90 percent of fathers and 82 percent of mothers would like to send their sons away to learn a craft, if given the opportunity, and more than half would have welcomed the opportunity to send a daughter to do the same. Almost half the fathers would like to send their sons to study with an urban Marabou, while mothers are somewhat less eager to do so. Mothers are also less interested in sending their sons to work in commercial agriculture while one-in-four fathers would be positive to that option. Almost one-in-three fathers and mothers are positive to sending a daughter to work as a domestic servant. Parents are more interested in sending their sons than their daughters to stay with relatives in the cities, fathers being more positive to both. …send a girl to live with relatives in the city 22 29 …send a girl for an apprenticeship 54 57 …send a girl to work as a domestic servant 30 32 …send a boy to live with relatives in the city 31 38 …send a boy to study with a Marabou in the city 37 47 …send a boy to work in commericel agriculture 18 25 …send a boy for an apprenticeship 82 90 0 10 20 30 40 50 60 70 80 90 100 Mothers Fathers Figure 50: “If you had the opportunity, would you have wanted to …� (responses by household heads). Among those who wanted, but did not send children away, the reasons given were the lack of money to organize the arrangement (26%), not knowing the right persons to facilitate for it 78 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y (23%), not knowing good people they could stay with (19%), not knowing how to do it (28%), while 3 percent answered that they had no children in the right age, or that their children were still too young. In total, the 1134 parents interviewed in the survey represented 4669 live and still borne children, (406). Currently, 3435 were born in 1993 or later and should thus be below the age of 18 years. Although statistically negligible, school participation rates are slightly higher for girls (47 percent) than boys (45 percent) in schooling age (6-14 years of age). Fourteen percent of the children (455) were currently living away from home, while 80 (2,4 percent) had previously lived away from home and then returned. Fifteen percent of the children who had ever lived away from home had done so more than once. A total of 238 children had been out of the community to study the Quran elsewhere, 194 stating Quran studies as the objective of their last relocation. Around half studied with a Marabout permanently settled in the area where they studied, 38 percent in a permanent Daara, while 14 percent had traveled with an iterant Marabou. Figure 51 shows that many more boys than girls had left the communities, and that the main reason for this was the high number of boys leaving for Quranic studies. The second largest group is children who have left to go to formal schools, followed by children who had left for social and economic reasons, work, marriage and others. 250 200 10 150 100 184 39 Féminin 50 Masculin 56 16 40 5 29 14 10 0 19 17 2 17 18 14 Figure 51: Main objectives for children leaving the households (in numbers in of the sample). Child Mobility and Rural Vulnerability | 79 The average departure age for the children who had left for Quranic studies was low, at 6,9 years of age (Figure 52). Children leaving for social reasons were even younger at the age of departure, 5,6 years of age, while average is higher for most other groups. 14 11,5 12,1 12,4 11,4 12 8,8 10 6,9 7,6 8 5,5 6 4 2 0 Figure 52. Average departure age for children leaving for different purposes. 5.2.7. Child relocation, household poverty and shocks Comparing the child mobility practices in the risk areas is in many ways less controversial than for the national survey sample. This is because selection bias will be somewhat smaller in these focus areas, that is, the families we compare are generally more similar to each other. In the adaptive sample, 53 percent of households had relocated at least one child before the age of 18. Twenty-five percent had sent at least one boy to study the Quran while living away from the household, 14 percent had sent children for formal schooling elsewhere, and 25 percent had relocated children for other reasons, mainly social and economic ones. In this sub-section, economic and risk determinants in the households are presented; first, self-reported poverty and future concern; second, household poverty proxied by housing quality; and finally, the exposure of the households to risks and shocks over the past ten years. Indicators of self-reported poverty and future concern turned out to correlate poorly with the relocation of children. Reported income level, adequacy of food consumption and consumption of other basic necessities, as well as the level of worry of household heads with regards to providing their family with food and other necessities the next 12 months, appeared irrelevant to the child mobility decision. 80 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y The more objectively verifiable poverty proxy of house building quality turned out to correlate much more systematically with the various child mobility decisions. In the sample, 35 percent of main household buildings had no solid construction parts, that is, they were generally made of dirt, clay, and straw. Forty-two percent were built of all-solid materials, that is, brick walls, wooden or cemented floors, and brick or iron sheet roofs. Looking first at child mobility overall, 45 percent of high-quality households had at least one relocated child, while the share of low or medium quality houses (buildings with some, but not all solid material) was almost 60 percent. The correlation had the same direction for all the sub-groups of relocated children, but the largest differences were found with regards to whether a household had or had not sent children to study the Quran elsewhere (Figure 53). 35 31 30 30 30 28 25 21 20 20 20 19 With 15 Without 10 5 0 Solid floor Solid roof Solid wall All-solid housing material Figure 53. Share of households with at least one child sent to study the Quran while living away from the household, wealth indicated by housing material: With and without solid floor, with and without solid roof, with and without solid wall, and finally comparing households with all-solid household materials to those with none. The shock experience of households also correlates systematically with child mobility, with the exception of child relocation towards formal education. In total, 37 percent of the households had experienced drought, 30 percent a dramatic incidence of animal disease, 63 percent a drastic Child Mobility and Rural Vulnerability | 81 price increase, 22 percent the death of an able-bodied adult household member, and 16 percent the severe illness of a household member. Figure 54, Figure 55 and Figure 56 show the differences in child mobility rates for households who over the last ten years had been exposed to some of the most common and important climate and non-climate related shocks in Senegal. All three figures demonstrate systematic correlations between child relocation and such shocks, notably shocks related to climatic conditions. Looking first at child mobility overall, Figure 54 shows that more than two-in-three households that had experienced a drought or a significant incidence of animal disease have relocated at least one child. Less than half the households without such an experience have done the same. Impacts from other shocks are systematic, but less prominent. 80 67 68 70 61 58 60 55 52 52 47 49 50 45 40 30 Shocked 20 Unshocked 10 0 Drought Animal disease Sharp price Death of adult Illness of increase household member Figure 54. Share of households with at least one relocated child on exposure to climate and non- climate related shocks. Figure 55 shows the correlation between shock exposure and the household having relocated at least one child to study the Quran away from the home community. Drought in particular seems to influence the likelihood that a child is sent away to study the Quran elsewhere, but also animal disease appears to have a significant impact. Notably, the death of an able-bodied adult in the household correlates strongly with child mobility towards Quranic studies, at tendency that is less remarkable for other groups of relocated children. 82 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 40 34 33 35 31 29 30 26 24 25 22 22 22 19 20 15 Shocked 10 Unshocked 5 0 Drought Animal disease Sharp price Death of adult Illness of increase household member Figure 55. Share of households with at least one talibé on exposure to climate and non-climate related shocks. Child relocation for “other reasons� than marriage, Quranic and formal studies also seems affected by the exposure of the household to both climate and non-climate related shocks. While 20 percent of households unaffected by climate shocks have relocated at least one child for such social and economic motives, one-in-three households exposed to climate related shocks have done the same. The death of an adult interestingly had much less impact on the mobility of children for “other reasons� than on sending children to study the Quran elsewhere. The illness of a household member however seems to affect child mobility. Thirty-one percent of the households that had experienced such severe illness have relocated at least one child, compared to 23 percent of households without such an experience. Child Mobility and Rural Vulnerability | 83 35 32 33 31 30 26 25 24 25 22 23 20 20 20 15 Shocked 10 Unshocked 5 0 Drought Animal disease Sharp price Death of adult Illness of increase household member Figure 56. Share of households with at least one child relocated for other reasons than marriage, Quranic and formal studies on exposure to climate and non-climate related shocks. The last type of wealth proxy used in this study is the asset count. However while the ownership to certain assets seems to correlate negatively with most types of child mobility, others do not, and the ownership to a pushcart in fact correlates positively. Therefore, an asset count index will not be a meaningful wealth proxy for the purpose of studying child mobility in the four regions. 60 56 56 56 54 55 53 52 49 50 46 42 40 30 With Without 20 10 0 TV Radio Mobile phone Bicycle Cart Figure 57. Share of households with at least one child relocated on asset ownership. Comparing households owning the asset to those who do not. 84 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y As Figure 57 shows, certain asset ownerships correlate more closely with child relocation than others. While 42 percent of households that hold a TV have at least one relocated child, the share is 56 percent for households with no TV access. Also, a considerably lower share of the 20 percent of households that own a bicycle have relocated children. For the other most common assets (radio and cell phone), differences are much smaller. Households owning a cart seem to have a lower child mobility rate. While the impact of household asset ownership was strongest for child mobility for social and economic reasons, it had little influence on the probability of sending children away for Quranic studies. Again, ownership to a cart correlates positively with this type of child mobility. Child mobility for formal education also correlates positively with some of the asset indicators. Therefore, the fact that also child mobility for formal education purposes is included in the all- mobility sample reduces the effects shown in Figure 57. 5.2.8. Regressing child relocation on household poverty and shocks To assess whether the impact of climate related shocks on child mobility are in fact significant also in a more complex context, a basic logit regression was applied. The dependent variable is defined as 1, if the household had relocated at least one child below the age of 18 years, and 0 if not. On the climate side, having experienced a shock as a result of a drought over the past 10 years was used as a proxy, since droughts seemed to correlate strongly with child mobility in the cross tabulations. To capture some features of the presumed decision maker in the household, the gender, age and education of the household head were included in the regression equation. The house quality index was used as a proxy to household wealth, and defined as dummy giving the value 1 to households constructed from all solid materials, and 0 to all others. Two categorized, indicators to capture cultural features were included. First, the ethnicities Serer, Pular and “others� were included as compared to the residual Wolof category, and second, the religious brotherhood affiliations Tijaniyyah and Mouride as they compare to the Xaadir (residual). Finally, the regional control variables for Louga, Kolda and Thies were categorized as they compare to Kaolack. Child Mobility and Rural Vulnerability | 85 B Sig. Female household head -.167 .620 Age household head .006 .374 Household head never been to school .510 * .061 All solid housing material -.418 * .055 Household shocked by drought .436 ** .044 Ethnicity (residual Wolof) Serer -.104 .739 Pular -.050 .869 Other -.063 .882 Religion (residual Xaadir) Tijaniyyah -.417 .169 Mouride -.377 .288 Region (residual Kaolack) Louga -1.102 *** .001 Kolda -1.282 *** .001 Thies -1.539 *** .000 Constant .825 .158 Table 4. Impact of drought on child mobility, controlled for household head features, household poverty, ethnicity, religion and region. The regression results in Table 4 show that the positive correlation between drought experience and child mobility is strong even after controlling for agency features, wealth, cultural and geographical factors. The gender and age of household heads is not significant in this limited equation, while wealth, indicated by solid housing material, reduces likelihood of child mobility. Ethnicity and religious affiliation does not seem important, while geographical location strongly does. To get an impression of the effect of the drought experience on child mobility, logit coefficients require a procedure of recalculations in order to be transformed into probabilities. Probabilities differ for different types of households, and the parameters for the other variables must therefore be decided and fixed. Departure is here taken in the typical household headed by a male 40 year- old Tijaiyyah devotee, who has never been to school. As child mobility seems more common among the Wolof and in Kaolack, this case constitutes the high-end example. Recalculation of odds show that a Wolof family in Kaolack living in a solid house and with no previous experience with drought still has a 68 percent likelihood of 86 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y having relocated a child, a likelihood that increases by 9 percentage points if the households has been shocked by a drought. Child mobility probability is as high as 76 percent in a poor (not solid housing material), un-shocked household, increasing by 7 percentage points if the household has experienced a devastating drought. The Thies region has the lowest propensity of child mobility among the four risk-areas studied, and the Serer have the lowest child mobility rates. The low-end estimates therefore look at a Serer household in Thies. There, an un-shocked family living in a solid (“wealthy�) house has a 29 probability of having relocated at least one child, and this probability rises by 10 percentage points if the household has experienced a drought. In a “poor� household, probability raise by 11 percentage points if shocked by a drought. The results thus show a persistent and considerable impact of droughts, of a similar strength as the applied poverty indicator. 5.3. The urban interviews In the 569 households revisited, 482 children under 18 years of age were reported to be living outside the parental household. In addition, 80 had been living away from the household for more than two months in the past, and then returned. Among the children who were currently living away from their parents, 84 children had left to go to formal school. These children were removed from the tracing sample, leaving 398. Among the remaining children, six percent were living within the same community, 15 percent in a neighboring community, 29 percent in another rural area, and around half, 197 children, had left for urban areas. The rural re-visit thus resulted in a list of 197 children identified as relocated to urban areas for social, economic or religious reasons. As reported in sub-section 3.2.3, most of the relocated children were boys, and most of the boys had left to study the Quran. Not all the parents were able or willing to provide accurate tracing information to the interviewers. In addition, the teams for pragmatic/cost reasons had to pre-exclude remote urban areas with only one relocated child. Tracing the children then left on the list turned out still proved to be a substantial challenge for several reasons. In practice, many of the parents had provided information that was insufficient or wrong, and extensive correspondence with resource Child Mobility and Rural Vulnerability | 87 persons from their villages was needed in order to find the way. Chefs de quartiers in the urban areas visited were of great help to the teams in their efforts to locate the tracer children. The tracing took place in a harvest period. Because of this, 30 children had temporarily returned to their villages to help out, or to help their Marabout at his farm. Also, resource persons who could have helped in the tracing of others were unavailable because they too took part in harvesting. On top of this, a flood struck parts of Dakar during the research period, forcing many families to relocate. Even though most of the households were found, many of the tracer children were thus not currently there in spite of parental information being only a month old. In addition to the children who had temporarily returned to their villages to help in the field, several children were also reported to have fled the Daara - or the Marabout was unable to tell where they were, two girls had recently returned to get married, and one had returned to the village after becoming pregnant. One boy had recently rented a room to start his own household together with a cousin, one had started in a tailor apprenticeship, and another had transited with the purpose of leaving for France. When the households and children were found, the final challenge consisted in getting the permission to make interviews with the children. In particular the Marabous were hesitant to allow strangers to conduct interviews with children studying with them, some even after as much as five visits. As it turned out, a Marabout had just been sentenced to prison for the maltreatment of a talibés, and the news was being spread from Daara to Daara. The Marabous also complained that in spite of all the research currently carried out, they never received any project or government support. As a result, only 45 among the 109 talibés from the original sample could be interviewed. In total, interviews were obtained from 29 relocated girls and 66 relocated boys above 5 years of age. All this implies that the final sample of 95 children traced is not representative for all relocated children, but it does allow for comparison of parental information about the relocation arrangement of those found to the lived realities of the children. 88 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 5.3.1. The relocated children Since children who relocate primarily to go to formal school were not part of the tracing exercise, the formal education level of the relocated children was not surprisingly relatively low. Among the talibés, only two had previously been to school. Among the children who had relocated for social and economic reasons, around one third had been to school, and three were still studying. Among the talibés, six were able to read French, and so were nine of the children who had relocated for other reasons, an additional nine with some difficulties. Only one talibé reported to be able to write in French, while among the non-talibés, seven could write well, and nine could write French with some difficulties. Half the talibés said they were able to read Arabic, 14 could also write, an additional one with some difficulties. Sixteen non-talibés were also able to read Arabic, an additional seven with some difficulties. Thirteen of these were girls. Child mobility can be very temporary, or it can last for many years. As described in the previous sub-section, as many as 30 of the children to be traced were temporarily away from their new households to help out with the harvest in their rural community. Twenty-one of the children traced had been away from home for a year or less. Contact with the parental household was in half of the cases good. Five children traced said they visited their villages on a regular basis, while 35 said they visited their parents now and then. An additional eleven had been home once, while 41 had never been back. Similarly, nine said their parents visited them regularly in their new households, while 37 said their parents visited now and then. Only 34 had never received visits from their parents, 14 of them talibés. Nineteen had not had contact with their parents since coming to their new household. All the children who had relocated for social and economic reasons reported to be in good health, while four talibés said they had been ill recently, one of them chronically. Only 15 children said they had not been working the last week, most of them talibés, and only one of them was a girl. On average, the girls in the sample reported to have worked 52 hours last week, the boys only 33. It is likely, however, that the talibés did not count their begging time as labor. 5.3.2. The new households of the relocated children Household heads in the urban households caring for the relocated children have more often been to school than the household heads in the rural households of origin, although differences were Child Mobility and Rural Vulnerability | 89 small. While 86 percent of the rural household heads had never set foot in school, the number for urban household heads was 80 percent. Among the rural household heads, 41 percent were alphabetized, and the equivalent share among urban household heads was 51 percent. With very few exceptions, children move between households of their own ethnicity, and also generally to households within the same Muslim brotherhood. Comparing the households of origin to the children‟s new households, however, reveals a consistent improvement in household wealth. Figure 58 shows that 76 percent of the urban households had an income of more than CFA 75.000 last month, compared to only one-in-three rural households. Seventeen percent of the rural households had an income of less than CFA 25.000 last month, while only two percent of the hosting households did. 80 76 70 60 50 40 Household of origin 33 30 Hosting household 30 19 20 20 17 10 2 2 0 CFA 0 - 25.000 CFA 25.000 - 42.000 CFA 42.000 - 75.000 CFA > 75.000 Figure 58. Income of household of origin and hosting household in percent of households. The first bars in Figure 59 show that while 55 percent of the households of origin found their current level of food consumption to be less than adequate, this was the case for 42 percent of children‟s new households. Furthermore, 62 percent of the biological households found the current consumption level for other basic necessities to be less than adequate, compared to only 90 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 45 percent of the hosting households. Finally, while almost half of the rural household heads worry a lot about providing their households with basic necessities the next 12 months, this was the case for one-in-four of urban households. The new households of the relocated children were in other word also quite poor, but still somewhat less worried about survival in the future, the exception being the households led by Quranic tutors. 70 62 60 55 50 47 45 42 40 30 26 Household of origin Hosting household 20 10 0 Current level of food Current level of other basic Worry a lot about providing consumption less than consumption less than the household with basic adequate adequate necessities next 12 months Figure 59. Household heads’ perception of adequacy of food consumption, other basic consumption, and worry about future consumption in percent of household heads. Comparing utility ownership in the children‟s household of origin to the urban households of the children‟s new caretakers, there are also some notable differences (Figure 60). Most strikingly, almost 80 percent of the hosting households have electricity, compared to only 12 percent of the households of origin. While more than 60 percent of the urban households have access to TV, this is the case for only one-in-four biological households. While one-in-four hosting households have a refrigerator, less than ten percent of the households of origin do. Urban households also more often have a radio and a telephone, although the differences here are less prominent. Child Mobility and Rural Vulnerability | 91 100 92 90 78 77 80 71 70 67 61 60 50 Household of origin 40 Hosting household 30 27 24 20 12 8 10 0 Electricity Radio TV Refrigerator Telephone Figure 60. Utility ownership in households of origin compared to hosting households in percent of households. There were no striking wealth differences between the 44 households led by a Marabout and the other urban households in the sample. The only observation that sticks out is, as mentioned, that many more among the Marabouts expressed strong concerns about being able to provide the household with food and basic necessities the next 12 months. Almost all the children and household heads said they had eaten three meals the day before. Eight children and one household head had only two, while one child had only one. All but two of the children had eaten fish, meat or poultry at least once the day before. Applying measurements used by the World Value Surveys, assessments were made of the personality of household heads throughout the research work. In the rural survey, children of risk adverse household heads were found to be more likely to relocate. Beyond a personality feature, risk adversity is likely to reflect the general security situation surrounding a household. It is therefore interesting to find that the household heads of the new households of the relocated children are far less risk adverse than the household heads of their biological households (Figure 61 and Figure 62). As Figure 61 shows, only 29 percent of the rural household heads identify 92 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y with a person who loves risk taking and adventure, while almost half the urban household heads do. Forty-seven percent of the rural household heads do not at all identify with this personality type, while the number for urban household heads is 16 percent. Figure 62 shows that half the rural household heads identify strongly with a security-seeking personality who would like to avoid anything dangerous, while only one-in-three urban household heads do. Urban household heads are somewhat security-seeking, but not quite as strongly as the rural ones. 50 46 47 45 40 38 35 29 30 24 Household of origin 25 20 Hosting household 16 15 Linear (Household of origin) 10 5 0 A lot like me A bit like me Not at all like me Figure 61. Identification with a person characterized by the following statement: “Adventure and risk taking are important, to have a passionate life�. 70 63 60 50 50 40 34 Household of origin 33 30 Hosting household Linear (Household of origin) 20 16 10 3 0 A lot like me A bit like me Not at all like me Figure 62. Identification with a person characterized by the following statement: “To live in a secure environment and avoid everything that might be dangerous is important�. Child Mobility and Rural Vulnerability | 93 Figure 63 to Figure 66 describe the motivations and expectations of urban household heads taking in rural children, comparing Marabouts to other household heads. Not surprisingly, the Marabouts‟ main motive for taking in children is to teach them the Quran, but even other household heads find teaching the Quran a part of their obligation. Some Marabouts also take upon them to teach the child a craft, or state that they take care of the child in order to help the child‟s parents and the child himself. Among the other household heads, almost half are motivated by teaching the child a craft. Thirty-nine percent say they want to help the child‟s parents, while one-in-four also state that their motivation is to help the child him- or herself. One in five is motivated by the need for help with housework, and ten percent by need for help with their business. Eleven percent also say they want to help with the child‟s schooling. In the rural survey, household heads were asked what they expected from a family where they placed their child. Eighty-eight percent of the survey sample said they expected help with the education of the child, and, as shown in Figure 64, almost all the Marabouts and the majority of other urban household heads support this expectation. Seventy-seven percent of the survey sample expected a new household to nourish and protect the child. Figure 64 shows that only one in three Marabouts feel that this expectation is justified, and only two-in-three other urban household heads. Forty-two percent of the rural population were hoping for some help to find the child a spouse. None of the Marabouts supported this expectation, and only six percent of the other household heads felt that this should be their obligation. Parental expectations to regular support (28 percent) or support in times of crisis (34 percent) are also far above the obligations felt by hosting household heads. 94 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Teach the Quran 12 98 Teach a craft 46 14 Help the child's parents 39 11 Help the child, other 27 7 Help with housework 21 0 Help the child's schooling 11 0 Help in business 10 0 For the child to help me when I become old 4 0 0 20 40 60 80 100 120 Others Marabouts Figure 63. Motivations of Marabouts versus other household heads for taking in children, in percent of household heads. Ensure education 56 98 Feed and protect 64 30 Help in crisis 14 5 Send money 12 0 Help find a spouse 6 0 Nothing 4 0 0 20 40 60 80 100 120 Other Marabouts Figure 64. Comparing what Marabouts and other urban household heads think children’s parental household can expect from them, in percent of household heads. Child Mobility and Rural Vulnerability | 95 Both Marabouts and other household heads overwhelmingly expect the respect and obedience of the child they take in (Figure 65). Fourteen percent of the Marabouts and 23 percent of other household heads expect the child to help with household chores. Six percent of other household heads also expected the child to help in their business, while 12 percent expected some help later, an additional 10 percent when they grew old. Respect and obedience 87 84 Help with housework 23 14 Help me when I become old 10 5 Nothing 19 5 Help me later 12 0 Help with my business 6 0 0 10 20 30 40 50 60 70 80 90 100 Others Marabouts Figure 65. Comparing what Marabouts and other urban household heads expect from child, in percent of household heads. Finally, Figure 66 shows that most urban households also have expectations to the child‟s parents, only one-in-three expect nothing. Notably, half expect respect and assistance, while a few also expect to receive cash of kin and/or help in times of crisis. 96 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Respect and assistance 56 48 Nothing 37 36 Help in times of crisis 10 7 Send agricultural products 4 5 Send money regularly 4 0 0 10 20 30 40 50 60 Others Marabouts Figure 66. Comparing what Marabouts and other urban household heads expect from the children’s parental household, in percent of household heads. 5.3.3. Parental expectations and children’s views One of the objectives in the tracer exercise was to compare parental expectations to the children‟s perception about their own realities in their new households. Therefore, a set of similar questions were asked both in the households of origin and to the children in the hosting households. The first four bars in Figure 67 show that children overall see their current living and working conditions as more positive than their parents expected them to be. Income opportunities and future prospects are overwhelmingly seen as better by both parents and children equally. Children slightly more often than adults report to provide financial support to their households of origin, and also believe more often than their parents that they will be able to support in case their parents experience an immediate crisis. Both parents and relocated children overwhelmingly believe that the child will be able to support the parents in the future. While one-in-three parents believe the child would not like to return home, only 19 percent of the children say they do not want to go back home. It should, however, be kept in mind that vulnerable children often systematically idealize even very difficult conditions as a type of a survival strategy, mainly in situations where facing the depressing facts would be devastating. Child Mobility and Rural Vulnerability | 97 Better living conditions 52 76 Better working conditions 49 59 Better income 75 77 Better future prospects 80 80 Parent's views Provides financial support to parents 41 51 Children's views Would help financially if parents encounter 41 economic difficulties 54 Will be able to support parents in the future 95 94 Child does not wish to return home 32 19 0 10 20 30 40 50 60 70 80 90 100 Figure 67. Views and expectations about the benefits of relocating a child. Percent of relocated children’s parents/guardians compared to percent of relocated children. 98 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y 6. Impact analysis Are the families and biological households of relocated children any different from other households? That is, are there special common denominators for the households of mobile children compared to other households? In this chapter this will be further scrutinized. Special focus will be dedicated to the possible link between climate vulnerability, adult and child mobility. 6.1. Determinants of child mobility 6.1.1. Main working hypotheses Child mobility is often associated with poverty, and the regression of data from the vulnerable areas of Louga, Thies. Kaolack and Kolda showed that households with main buildings constructed of all solid materials were indeed less likely to relocate children before the age of 18 (see Table 4 in sub-section 5.2.8). In rural Senegal however, poverty is almost global, and almost all families are confronted with the same constant threats of debilitating shocks. Moreover, they generally share a situation where safety nets are under-dimensioned, especially in the case of covariate shocks. Do then differences in wealth among poor households explain differences in child mobility decisions also applying more complex wealth measurements in the nationwide survey? Unlike most adults, children are often not free to make their own choices. Parents are normally involved, and the family agenda as well as parental may features therefore affect the decision about the child‟s relocation. In the qualitative research from vulnerable areas (referred to in sub- section 5.1.), men were presented as the main decision maker in child relocation decisions. Men were also claimed to be more in favor of child relocation, especially toward Quranic studies. Cross tabulations on parental preferences in vulnerable areas made under part 2 of this research study show that men generally are more in favor of all types of child relocation (see Figure 50). The difference between fathers‟ and mothers‟ preferences is not dramatic, but a bit larger in the case of sending children off to Quranic studies. If women are indeed more skeptical to child relocation, and women do in fact have some say in the relocation decision, then all indicators that Child Mobility and Rural Vulnerability | 99 would suggest an increase of women‟s‟ bargaining power should reduce likelihood that her children relocate. Such indicators could be her age, education, health, the support she can expect from her own family in bargaining with her husband. A proxy to the latter could typically the influence level of her parental household relative to her husbands‟ household, and the geographical proximity of her family for practical support in every-day decision making.15 If children also have a say in the child mobility decision, then the personality of the child could affect the outcome in three ways. First, children who are primarily described as affectionate and emotional may be regarded as unfit to make it without their parents, and thus be kept at home. Second, intelligent children may be more capable of negotiating their own outcomes. Third, obedient children may not negotiate at all. They are at the same time easy to send because they obey their parents, and may make good social network builders for their families as they can be expected to also be obedient and easy to deal with for their new household head, tutor or employer. The child mobility decision should also reflect the other option a child has. Such options for children are represented by schooling in the home community and local labor opportunities. In a community without a primary or secondary school, or where school quality is rated as poor, children would have greater incentives to leave. In poor communities with little infrastructure the likelihood of finding paid work is lower. The remoteness of the community from urban areas should be considered as a disincentive for leaving as distance increases relocation costs. Finally, cultural and geographical factors would normally impact on family decision making. 6.1.2. Why use the regression approach? A multinomial logistic regression analysis will be used to assess whether certain factors impact the likelihood of child mobility away from a household. The purpose of using the approach is to determine whether the correlations and tendencies shown in the figures earlier in this report are 15 While the wealth level of the mothers’ parental household could have been measured, technical debates before the study argued that money is subordinate to influence in Senegalese society, and that influence indeed would be a more appropriate measurement. 100 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y indeed systematic and significant, or whether they are more likely random or primarily explained by other and related factors. For example: in some surveys cross tabulations show a negative correlation between female household heads and children‟s school participation. However, some places female headed households are also much poorer than male headed ones. Including both poverty and the gender of the household head in one and the same analysis will reveal whether the gender of the household head really does have a systematic impact on the school participation of children, or if household poverty indeed explains the entire difference. With regards to this concrete example, the latter is often found. 6.1.3. The overarching hypotheses to test In the following analysis, two quite general hypotheses about child relocation are tested. H1. Factors related to household and community incentive, constraint, agency and culture/demography contribute to explain why children relocate in rural Senegal (Bhalotra and Tzannatos, 2003 and Kielland, 2008): H1.1. Poverty drives child mobility. Poorer families relocate more children out of need. H1.2. Community incentives matter. Lack of good schooling and work opportunities in the home community lead to more child mobility. H1.3. Households are not unitary actors. Women‟s level of bargaining power in the household reduces likelihood of risky child relocations/outcomes (Quisumbing and Malluchio, 1999). H1.4. Children are not equal. Children take part in the bargaining over own relocation outcome, and child features (indicative also of bargaining power) matter to the relocation decision (Iversen, 2002). H1.5. Ethnicity and production/geography affects likelihood of child mobility. H2. Different relocation arrangements have different determinants. Mixing different relocation arrangements in a joint category of “relocating children� conceals nuances (Ainsworth, 1999, Serra, 2008, Edmonds, 2008). Child Mobility and Rural Vulnerability | 101 The anti-hypotheses (H0) to this would be that; H0 …household and community features are not important to the child relocation decision. Among rural poor households at risk, all those who have the opportunity will relocate children as part of household diversification efforts. This line of thinking would find support in the classical work of Isiugo-Abanihe (1985) and recent studies by Lilleør (2008). 6.1.4. The regression and the results Table 5 shows the results of the regression analysis. The set-up is based on hypotheses 2, where different motives for relocation are assumed to have different determinants. The first column shows the impact of the background factors on the likelihood that a child is relocated regardless for what purpose. The second column shows the same results for the likelihood that a child is relocated for social and economic reasons, that is, children who have relocated to get married, go to formal school or study the Quran are removed from the sample. The third column looks only at the children who left to study the Quran. To start with the first hypotheses, it does seem like household and community factors indeed impact on the child relocation decision. As expected, poverty is positively associated with child mobility, that is, the wealthier the household the less likely the child is to relocate. The most striking finding is that this is not valid for the children who left to study the Quran, in contrast to what was found when looking at the wealth indicator of housing quality and child mobility for Quranic studies in data from the risk areas (see Figure 53). While the talibé phenomenon is often associated with poverty, qualitative studies have repeatedly stressed that from the viewpoint of the parents, poverty has little to do with the practice that they see as predominantly religiously motivated. Community infrastructure level, here used as a proxy for the availability of paid work in the home community, did not seem to have much of an impact on the child mobility decision. The schooling situation however, had the expected impact, although perhaps less strongly than expected. There being a primary school in the community mainly reduced the likelihood that children were sent away for Quranic studies. A secondary school, however, reduces child mobility in general, and also if we exclude mobility for formal schooling. The missing impact of 102 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y a secondary school on child mobility for Quranic studies is likely related to the low departure age of children leaving for this purpose. Community focus groups expressing high satisfaction with the primary school available correlates closely with child mobility, reducing all forms, including children leaving to study the Quran.16 Features likely to affect the bargaining power of women do indeed affect child relocation decisions. Even when controlling for household wealth, female headed households where women do not have to negotiate decisions with their husbands, are more likely to relocate their children for all purposes, except Quranic studies. This should support claims from qualitative studies that women are more skeptical to this quite risky type of child relocations, and the different preferences of men and women shown in Figure 50. Assuming that the mother‟s age would give her more leverage, children are also less likely to relocate the older the mother. Very few of the rural mothers had ever been to school, and there was no impact of education for those who had. On the other side, mother‟s health matters to her bargaining power. Notably, the poorer the health of the mother, the more likely it is that her children are sent for Quranic studies. Finally, the more influential the parental family of the mother, the smaller then chance that her children are sent for Quranic studies.17 With regards to the household head, neither health conditions nor age impacted on the child relocation decision. Education level notably did: if the household head had ever been to school it systematically increased the likelihood for all types of child mobility, notably including for Quranic studies. Looking at the child, not surprisingly the likelihood of relocation increases by age. The age- squared coefficient shows that the relationship between age and relocation is not a linear one. While boys and girls are equally likely to relocate, girls are more likely to relocate for social and economic reasons, while boys are way more likely to leave for Quranic studies. Looking at the 16 In the regression school satisfaction is treated as a categorical variable, comparing the satisfied to the dissatisfied - singling out respondents in communities without schools in a separate category - in the table this second category is typed in gray. 17 The proximity of the mother’s parents also seems to reduce child relocation, but his may possibly reflect the lack of opportunity: since children are often sent to close relatives, they would have fewer places to go if the relatives live in the same communities. Child Mobility and Rural Vulnerability | 103 child‟s personality features in a regression setting, a few features from the figures shown earlier in this report prevail. First, children primarily described as affectionate are less likely to be sent anywhere. Second, obedient children are more likely to be sent to study the Quran. Table 5. Multinomial logistic regression of the determinants of child relocation over all, for social and economic reasons (that is, children relocated for formal schooling, marriage and religious studies are removed) and for Quranic studies. All relocation Socioeconomic Talibés Coefficient Coefficient Motives’ Coefficient Child age ,289 *** ,283 *** ,834 *** Child age squared -,006 *** -,008 ** -,036 *** Child is a boy ,044 -,409 *** 2,168 *** Intelligent ,120 -,296 (*) -,178 Affectionate -,261 ** -,478 ** -,067 Obedient ,133 -,434 ** ,582 *** Household wealth (1-19) -,044 *** -,086 *** -,012 Household head female ,255 (*) ,964 *** -1,097 * Household head’s age -,002 -,002 -,009 Household head ever been to school ,249 ** ,461 ** ,680 *** Household head is ill ,051 -,049 ,145 Mother’s age -,001 -,015 * -,036 *** Mother ever been to school ,051 ,015 -,056 Mother is ill ,156 ** -,085 ,327 ** Mother’s parental household influential -,451 ** -,318 -1,735 ** Mother’s parental household in community -,129 (*) -,305 ** ,240 Peuhl -,424 *** -,924 *** -,594 ** Serer -,194 ,186 -,407 Other -,355 ,*** -,333 (*) -1,769 *** Community wealth (based on infrastructure) ,005 -,002 ,037 (*) Community remoteness (distance urban -,004 ** -,002 -,005 Primary school in the community (within 1 km) -,171 (*) -,203 -,426 * center) Secondary school in the community -,246 ** -,744 *** ,008 Satisfaction with school quality -,135 (*) -,416 *** -,580 *** School quality NA ,050 ,052 -,483 (*) Groundnut ,347 ** ,374 (*) ,796 *** Cotton ,476 *** ,409 (*) ,324 Rice ,724 *** ,938 *** ,841 ** Cattle -,182 -,874 *** ,668 ** Constant -4,320 *** -3,581 *** -8,310 *** *** 1 percent chance that the result is arbitrary, ** 5 percent chance that the result is arbitrary, * 10 percent chance that the result is arbitrary, (*) 15 percent chance that the result is arbitrary. Finally, cultural features are here indicated by ethnicity, and geographical features are by belonging to a certain production zone. In a regression this is typically done by defining one reference category and comparing each of the other response categories to this group (categorical variable). With regards to geography, the various production areas are compared to the Niayes 104 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y areas, and the regression confirms that children are indeed more likely to be sent from any other region apart from the live-stock areas. Also confirming findings previously presented, compared to the reference category of Wolof, other population groups are less likely to send children away for all kinds of purposes. With regards to hypothesis 2, that different motives were determined by different background indicators, this turned out to be right in many cases. For instance, poverty did not impact on the relocation of boys to Quranic studies; women‟s bargaining power also seems used to reduce the likelihood of this type of relocation, while increasing other types. Personality features of the children themselves affected the type of relocation they experienced. While studies of child fostering practices tend to treat all child relocation as one phenomenon, it should be noted that “mobile children“ is a multifaceted group with complex motives and drivers. 6.2. Linking child outcomes to climate risks and shocks Rural households in Senegal have a high risk exposure, and as shown in Figure 3, the great majority has experienced shocks in the past that have significantly contributed to reduce their income, holdings or consumption. Children may potentially be affected by risk and shock exposure in several ways, and this section will focus on two of them. First, children may be forced to leave by necessity following a shock, or as a precaution made by households at risk. Second, when adults migrate as a response to risks or shocks, children loose the protection and care of a parent and guardian. 6.2.1. Shocks and child mobility Figure 3 showed how 30 percent of households had been shocked by a drought, 24 percent by locust, 23 percent by animal disease, many households by several such shocks. Did this experience in any way affect the likelihood that children from the household relocated? Looking at each shock separately may be confusing since likelihood is that a household that may not have been hit by drought, may very well have been hit by locust or animal disease. In Figure 68 the three major shocks are added up, so that all three shocks are counted, and the households Child Mobility and Rural Vulnerability | 105 seriously hit by several of them are listed with two or three shocks. The figure shows a general increase in child mobility by the number of different climate-related shocks experienced, when these shocks are seen isolatedly from other household features and experiences. The impact is, however, relatively moderate, and especially with regards to children relocating for Quranic studies. Looking at child mobility and the level of worry about climate-related shocks, did not turn out to produce any clear pattern, and would thus not support the idea of a link between risk awareness and ex-ante strategizing. ,45 ,40 ,35 For any reason (0-17) ,30 ,25 For any reason (0-15) ,20 ,15 For other than formal school (0-17) ,10 For Koranic studies ,05 ,00 no shocks one shock two shocks three shocks Figure 68. likelihood of a household having relocated children for any reason, for any reason but formal schooling and for Quranic studies by number of major shocks of drought, locust and animal disease they have been hit by. 5.2.2. Econometric framework: propensity score matching methods Although there are many theoretical reasons why climate-related shocks would lead to migration and child relocation, it is difficult to determine what the exact effect of these shocks (if any) is on children‟s mobility and adult migration. In an experimental trial one would imagine two groups of identical communities, one being exposed to climate-related shocks and the other one not. Since the two are in all ways identical, all increase in child mobility and adult migration could then be ascribed to these shocks. In social settings where one tries to capture the effect of a 106 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y climate-related shock on issues such as migration and child relocation, selection bias would be an issue of concern. Therefore, simple comparison of the outcome variable between households who experienced shocks (shocked) and households without a shock (non-shocked), as was tentatively demonstrated in the previous sub-section, would yield biased estimates of the impact of a shock. There are, however, statistical methods that may simulate a social experiment without reproducing the selection bias, like propensity score matching techniques. For a technical explanation of the procedure followed in this section, please refer to Annex IV. In short, two similar groups of household are selected from the existing household data. The main purpose of matching the data, case by case, is to construct a group of treated individuals (shocked households) which is similar to the control group (non-shocked households) in all relevant pre- treatment characteristics, the only remaining difference being that one group was subjected to a shock and the other group was not. The matching process is performed in two steps. In technical terms, first, a logistic regression is used to estimate the propensity score, and in the second step, the average treatment effect on the shocked households (ATT), conditional on the propensity score. 6.2.3. Results of procedure The results from the logit analysis of shocks and the variables used in the matching procedures shows that among household who experienced shocks, the predicted propensity score ranges from 0.149 to 0.888 with a mean of 0.514. Among non-shocked households, the predicted propensity score ranges from 0.098 to 0.803 with a mean of 0.449. Thus, the common support assumption is satisfied in the region of [0.149, 0.889] enforcing a loss of 8 shocked households. The difference in propensity score of unmatched treated and control sample is close to 100 percent. After matching, the bias is significantly reduced well below 2 percent. The bias is also significantly reduced for each covariate after matching compared to before matching. Although, there is no clear indication for the success of the matching procedure, in most empirical studies a bias reduction below 3% or 5% is seen as sufficient. A two sample t-tests was used to check if there are significant differences in covariate means for both groups also confirm this result. Before matching, several variables exhibit statistically significant differences. However, after matching, the covariates are balanced. These results clearly show that the matching procedure is Child Mobility and Rural Vulnerability | 107 able to balance the characteristics in the treated and the matched comparison groups. Therefore these results can be used to evaluate the effect of climate-related shocks on adult migration and child relocation among groups of households having similar observed characteristics. This allows comparison of observed outcomes for shocked households with that of a comparison group that shares a common support. 6.2.4. Estimation of the effect of climate-related shocks on adult migration and child mobility To compute the average treatment effect (ATT) accurately, one should match the treated and untreated groups precisely on the basis of the propensity score. Because this is practically challenging, three alternative matching methods were used and compared; nearest neighbor matching; radius and kernel matching. Table 6 provides the estimates of the effect of climate- related shocks effect on adult migration and child relocation obtained from the three matching algorithms. The results indicate climate-related shocks lead to a significant increase in adult migration. The estimated increase in adult migration as an isolated effect of the household being hit by a climate-related shock ranges from 5,2 to 5,6 percent according to the three matching methods. However, climate-related shocks did not turn out to have a statistically significant impact on child relocation exhibited by the insignificant impacts also shown in the table. Table 6. Impact of climate-related shock on adult migration and child relocation Outcome Matching method ATT Std. Error t-value Adult migration Kernel 0.056 0.016 3.47*** Radius 0.056 0.016 3.48*** Nearest neighbor 0.052 0.021 2.54*** Child relocation Kernel 0.026 0.021 1.27 Radius 0.026 0.021 1.25 Nearest neighbor 0.025 0.027 0.92 N (Households experiencing climate-related shocks)1 1073 N (Households without climate-related shocks) 1160 *** Significant at 5 % level. 108 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Conclusion This project report provides the results from a national survey carried out in the rural areas of Senegal, and further data from a more limited survey in the four regions where most child mobility was identified: Louga, Thies, Kaolack and Kolda. The surveys combine data on child outcomes and shocks to the households, that is: events that have led to a serious reduction in holdings, a drastic fall in household income or a strong reduction in consumption. One-in-three households covered by the national survey reported that over the past 10 years they had experienced a severe shock due to a drought, while one-in-four had been shocked by locust and a similar number by animal disease. The fear of new climate related shocks was overall high, almost 80 percent of household heads stating that they “worried a lot� about a new drought, 70 percent about a new outbreak of animal disease, and another 70 percent worried about erratic rainfall. In addition many stated that they worried sometimes. Based on the national survey an extrapolated estimate of 290,000 rural children was found to live away from their household of origin in Senegal. Around half of these children had moved to an urban area. The majority of 53 percent of relocated children are boys, while 47 percent are girls. Twenty-two percent of the girls had left to get married, while 17 percent of the boys and 16 percent of the girls left to pursue formal studies. Forty percent of the girls and 18 percent of the boys had departed for family and social reasons. As many as 43 percent of the boys had left to study the Quran, and this constitutes around 70,000 children. Among them 14 percent left home before age 5 and two-in-three before age 8. Most rural parents have a positive attitude towards child mobility. Fifty-one percent find placing a child with a migrant Marabout desirable given certain conditions. Sixty-one percent finds placing a child in domestic service desirable given certain conditions. Sixty-six percent find placing a child in agricultural work desirable given certain conditions. Parents have some explicit expectations to the households they send children to. Most expect the child to be nourished and protected, and that the receiving household will help educate the child. In addition, forty-two percent also suppose that the receiving family will help find a spouse for Child Mobility and Rural Vulnerability | 109 the child. Only one-in-four expects money to be sent on a regular basis, while one-in-tree anticipates support from the child in the case of an emergency. Parents show very low acceptance for overloading placed children with labor, depriving them of sleep or food. However, more than half the household heads find occasional slapping acceptable under certain conditions, half accept corporal punishment, while only 37 percent finds reprimanding the child tolerable. In the second stage of this research study, interviewers revisited the four regions found to practice child mobility the most; Louga, Thies, Kaolack and Kolda. More in-dept research in these areas revealed that fathers are slightly more prone to encourage child mobility than mothers, in particular towards Quranic studies. Among those who stated that they would like to send children away, but did not, one-in-four said they were lacing the money to organize for the child to leave, a similar number said they did not know the right people to help facilitate for it or for the child to stay with, while almost 30 percent said they did not know how to do it. Notably, only half the household heads said they would have sent all their children to formal schools, even if there were no costs involved what so ever. In Kaolack, the region where most child mobility was found, only one-in-five household heads responded that he would have sent all his children to school, even if school was completely costless. Is child mobility a possible response to climatic shocks? That is, should child mobility be perceived as a coping or mitigation effort by households exposed to or at risk of climate shocks? Evidence from this research study is not entirely unidirectional. An effort to isolate the effect of climate related shocks to child mobility by applying a propensity score matching method on the national sample gave no significant results. The approach however estimates a statistically significant, isolated impact of 5-6 percent increase in adult migration from shocked households. The sub-sample survey from Louga, Thies, Kaolack and Kolda, however, show some clearer tendencies, indicating that national variation is larger than variation in areas where child mobility is more frequently practiced. While cross tabulations show that 67 percent of households exposed to droughts had relocated one or more children, only 45 percent of those “un-shocked� 110 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y had. Results were similar for animal diseases. The climate related shocks thus had a much stronger impact on child mobility than for example a radical price increase (from 40 to 55 percent), death of an adult provider (from 52 to 58 percent) and illness of a household member (52 to 61 percent). The impact from climate related events was equally strong for sending a young boy away for Quranic studies. One-in-three households exposed to droughts had sent at least one boy away for Quranic studies, while this was the case for only one-in-five un-shocked households. Child relocation for so called social and family reasons had identical frequencies. In a logit regression of child mobility on drought exposure, controlling for household poverty, features of the household head, ethnicity, religion and region, the significance of droughts seem robust. Estimates based on the logit shows that a wealthy Kaolack family has a 68 percent likelihood of having a relocated child when un-shocked, increasing to 77 percent if shocked by drought. A poor family in the same region will have a 76 percent likelihood, increasing to 83 percent if shocked by a drought. In percentage points the estimated size of the impact is relatively similar in other regions and of other ethnic groups, but for areas with a lower initial child mobility rate this increase may be experienced as more drastic. A wealthy Serer family in Thies would for instance have a 29 percent probability of having a relocated child when un- shocked, and a 39 percent chance if shocked by drought. For a similar poor household, the probability increase is of 11 percentage points. A further propensity score matching treatment of the data from the at-risk sub-sample will be able to reveal if these results are indeed robust when taking potential selection bias into account. Child Mobility and Rural Vulnerability | 111 Literature recommendations Ainsworth, M., 1992, Economic Aspects of Child Fostering in Côte d‟Ivoire, Washington, DC: World Bank, LSMS Working Paper No. 92. Akresh, R. 2009, „Flexibility of Household Structure: Child Fostering Decisions in Burkina Faso‟, Journal of Human Resources, 44(4): 976–997. Alderman, H., Chiappori, P-A., Haddad, L., Hoddinott, J., and R. Kanbur, 1995, „Unitary versus Collective Models of the Household: is it Time to Shift the Burden of Proof?‟, Research Observer, 10 (1): 1-20. ANSD, 2007, Enquette Nationale sur le travail des enfants (ENTES 2005), Senegal. Financed by ILO. Appelbaum, E., and E. Katz, 1991, The Demand for Children in the Absence of Capital and Risk Markets: A Portfolio Approach, Oxford Economic Papers, 43(2), 292.304. Ayers, H. B., 2008, Secularism and the Expansion of Forced Child Begging in Senegal, 1850– 2008, thesis submitted in partial fulfi llment of the requirement for the degree of Bachelors of Arts in History from the College of William and Mary, Williamsburg, VA, April 11. Barrett, C.B., M. Bezuneh, D. Clay, and T. Reardon, 2005, „Heterogeneous Constraints, Incentives, and Income Diversification Strategies in Rural Africa‟, Quarterly Journal of International Agriculture 44 (1): 37-60. Bass, L. E., 2004, Child Labour in Sub-Saharan Africa, Boulder: Lynne Rienner. Becker, S., 1960, „An economic analysis of fertility‟, Demographic and Economic Change in Developed Countries, Gary S. Becker, ed., Princeton: Princeton University Press. Becker, Gary S. and H. Gregg Lewis, 1973, „On the interaction between the quantity and quality of children‟, Journal of Political Economy 81(2), S279-S288. Beegle, K., Dehejia, R. and R Gatti, 2006, „Child Labor and Agricultural Shocks‟, Journal of Development Economics 81(1): 80-96. Binswanger, H, and M. Rosenzweig, 1993, Wealth, weather risk and the composition and profitability of agricultural investments, Economic Journal 103, 56-78. BIT (ILO), 2007, Enquette de base sur les pires forms de travail des enfants dans les secteurs de l‟agriculture, de la peche maritime artisanale et du travail domestique au Senegal. Financed by USDOL. Bhalotra, S., and C. Heady, 2001 , Child Farm Labour: The Wealth Paradox, Washington, DC: World Bank, Social Protection Discussion Paper no. 125. 112 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Blanco, F. and C. Valdivia, 2006, „Child labour in Venezuela: children's vulnerability to macroeconomic shocks‟, UCW Working Papers, 1-29. Bledsoe, C., 1990, „No success without struggle: social mobility and hardship for Sierra Leone children‟, Man (N.S.) 25:70-88. Bledsoe, C., 1989, „The social management of fertility: Child fostering among the Mende in Sierra Leone‟, In W. Penn Handewerker, ed., African Reproduction and Social Organization, Berkley: University of California Press. Bledsoe, C. and U.C. Isiguo-Abanihe, 1989, „Strategies of Child Fosterage among Mende Grannies in Sierra Leone‟, In Ron J. Lesthaeghe, ed., Reproduction and Social Organization in sub-Saharan Africa, Berkeley: University of California Press, Chapter 10. Bledsoe, C. and A. Gage, 1987, Child fostering and child mortality in Sub-Saharan Africa: some preliminary questions and answers, International Union for the Scientifi c Study of Population (IUSSP), Paper presented at the Seminar on Mortality and Society in Sub-Saharan Africa, Yaounde, Cameroon. Bozari, M. 2003. Le rôle des enfants dans les stratégies de gestion de risque au sein des ménages au Niger. UNICEF/ Ministère du Développement Social, de la Promotion de la Femme, de la Protection de l‟Enfant MDS/PF/P/PE Brydon, L., 1979, „Women at work: some changes in family structure in Amezofe-Avatime, Ghana‟, Africa, Vol. 49, No. 2, pp. 97–111. Bryon, A., Dorsett, R., and S. Purdon, 2000, The use of propensity score matching in the evaluation of active labour market policies, Policy Studies Institute and National Centre for Social Research, Working paper No.4., Department of work and pensions. URL: http://research.dwp.gov.uk/asd/asd5/WP4.pdf. Cain, M., 1981, „Risk and Insurance: Perspectives on Fertility and Agrarian Change in India and Bangladesh‟, Population and Development Review 7(3): 435.474. Caleindo, M., and S. Kopeinig, 2008, „Some practical guidance for the implementation of propensity score matching‟, Journal of Economic Surveys 22 (1): 31–72. Case, A., C. Paxson and J. Ableidinger, 2002, Orphans in Africa, New Jersey: Center for Health and Wellbeing Research Program in Development Studies, Princeton University. Castle, S. and A. Diarra, 2002, The International Migration of Young Malians: Tradition, Necessity or Rite of Passage? London: London School of Hygiene and Tropical Medicine (unpublished report). Cigno, A., and M. Werding, 2007, Children and Pensions, CESifo Book Series. Child Mobility and Rural Vulnerability | 113 Cigno, A., Rosati, F., and Z. Tzannatos, 2000, Child Labor, Nutrition And Education In Rural India: An Economic Analysis Of Parental Choice And Policy Options, World Bank: Social Protection Discussion Paper Series no. 131. Da Costa, D. R., Gustafsson-Wright, E., Neri, M.C., Pinto, A. and G. Sedlacek, 2000, Microeconomic Instability and Children's Human Capital Accumulation: The Effects of Idiosyncratic Shocks to Father's Income on Child Labor, School Drop-outs and Repetition Rates in Brazil, World Bank working paper, unpublished. De Lange, A., 2006, Going to Kompienga – A study on Child Labour Migration and Traffi cking in Burkina Faso‟s South-Eastern Cotton Sector, report published by IREWOC and supported by Plan. Dehejia, H.R., and S. Wahba, 2002, „Propensity score matching methods for non-experimental causal studies‟, The Rev.Econ.Stat. 84(1): 151-161. Dendir, S., 2007, „Household shocks and child labor in rural Malawi‟, Economics Bulletin 4(1): 1-8. Diprete, T. and M. Gangl, 2004, „Assessing Bias in the Estimation of Causal Effects: Rosenbaum Bounds on Matching Estimators and Instrumental Variables Estimation with Imperfect Instruments‟, Sociological Methodology 34: 271-310. Dottridge, M., 2004, Kids as Commodities: Child Trafficking and What to do about it?, Terre des Hommes. Dottridge, M., 2002, „Trafficking in children in West and Central Africa‟, Gender and Development Vol. 10, No. 1, pp. 38–42. Escobal, J., Saavedra, J. and P. Suarez, 2005, „Economic Shocks and Changes in School Attendance Levels and Education Expenditure in Peru‟, Young Lives, An International Study of Childhood Poverty 13: i-31. Etienne, M., 1979, „Maternité sociale, rapports d‟adoption et pouvior des femmes chez le Balule (Côte d'Ivoire)‟, L'Homme 19(3–4): 63–107. Fafo AIS and Kulisha Management Service, 2008a, Verifi cation of Certification Activities in West Africa Cocoa Sector, Final Verification Report, Côte d‟Ivoire. Fafo AIS and Kulisha Management Service, 2008b, Verifi cation of Certifi - cation Activities in West Africa Cocoa Sector, Final Verifi cation Report, Ghana. Farris, A., 2007, „Behind the Begging: Parents motivations in sending children to Senegalese Quranic schools‟, World Bank. 114 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Galasso, V., Gatti, R. and P. Profeta, 2008, Investing for the old age: Pensions, Children and Savings, Washington DC: World Bank, Social Protection Discussion Paper No 0838. Goody, E., 1982, Parenthood and Social Reproduction: Fostering and occupational roles in West Africa, Cambridge: Cambridge University Press. Goody, E., 1978, „Some theoretical and empirical aspects of parenthood in West Africa.‟, in C. Oppong, G. Adaba, M. Bekombo-Prisco and J. Mogey (eds), Marriage, Fertility and Parenthood in West Africa, Canberra: Australian National University, pp. 220–272. Guarcello, L., Kovrova, I. and F.C. Rosati, 2008, Child labour as a response to shocks: evidence from Cambodian Villages, UCW Working Papers. Guarcello, L., Mealli, F. and F.C. Rosati, 2002, Household vulnerability and child labour: the effect of shocks, credit rationing and insurance, UCW Working Papers. Guarcello, L., S. Lyon, F. Rosati and C. Valdivia, 2004, The influence of orphanhood on children‟s schooling and labour: evidence from Sub Saharan Africa, UCW Working Paper. Gubert, F. and A-S. Robilliard, 2006, Risk and Schooling Decisions in Rural Madagascar: a Panel Data Analysis, Document de travail DIAL (Développement Institutions & Analyses de Long terme). Heckman, J., Ichimura, H., Smith, J., and P. Todd, 1997, „Matching as an econometric evaluation estimator: Evidence from evaluating a job training programme‟, Review of Economics Studies 64: 1017-1098. Heckman, J., Ichimura, H., Smith, J., and P. Todd, 1998. „Characterizing selection bias using experimental data‟, Econometrica 66 (5): 1017-1098. Holzmann, R., 2001, Risk and Vulnerability: the Forward Looking Role of Social Protection in a Globalizing World, Washington, DC: World Bank. Holzmann, R and S. Jørgensen, 1999, „Social Protection as Social Risk Management: Conceptual Underpinning for the Social Protection Sector Strategy Paper‟, Journal of International Development, 11:1005-1027. Holzman, R and S. Jørgensen, 2000, Social Risk Management: A New Conceptual Framework for Social Protection and Beyond, Social Protection Discussion Paper Series No. 0006. Human Rights Watch, 2003, Detailed Recommendations to all West African governments implicated in the traffi cking of children, including Togo, Benin, Nigeria, Niger, Côte d‟Ivore, Ghana and Gabon. Borderline Slavery, 1 April. URL: http://www.hrw.org/en/news/2003/04/01/ west-africa-stop-trafficking-child-labour. Child Mobility and Rural Vulnerability | 115 Hunt, P., 1993, „Children‟s rights in West Africa: the case of The Gambia‟s„Almudos,‟ Human Rights Quarterly, Vol. 15, No. 3, pp. 499–532. Ilahi, N., 2000, The Intra-household Allocation of Time and Tasks: What Have we Learned from Empirical Literature?, Washington DC: World Bank, Policy Research Report on Gender and Development, Working paper series no 13. Isiugo-Abanihe, U.C., 1985, „Child Fosterage in West-Africa‟, Population and Development Review 11(1): 53-73. Iversen, V., 2002a, „Autonomy in child labour migrants‟, World Development, 30(5): 817–834. Iversen, V., 2002b, Idiosyncrasies of Child Labour in Peasant Households in sub-Saharan Africa: Anthropological Observations and the Economics of Labour Obligations and Exchange, mimeo, University of East Anglia. Kielland, A., 2008, Child Labor Migration in Benin; Incentive, Constraint or Agency? VDM Verlag. Kielland, A., 2009, „Child mobility as household risk management‟ Forum for Development Studies 36(2). Kielland, A., and I. Bjørkhaug, 2009, „Child Mobility in West Africa: Strategy, Poverty or Crime?‟, Forum for Development Studies 36(2). Kielland, A. and M. Tovo, 2006, Children at Work: Child Labour Practices in Africa, Boulder: Lynne Rienner. Kielland, A. and I. Sanogo, 2002, Burkina Faso: Migration de la Main d‟Oeuvre Enfantine d‟Origine Rurale, Ampleur et Facteurs Determinants, Washington DC: World Bank/Terre des Hommes. Kielland, A. and R. Ouensavi, 2001, Le Phénomène des Enfants Travailleurs Migrants du Bénin, ampleur et determinants, Washington, DC: World Bank/CEO. Last, M., 2000, „Children and the experience of violence: contrasting cultures of punishment in Northern Nigeria, Africa‟, Journal of the International African Institute 70(3): 359–39. Leuven, E. and B. Sianesi, 2003, PSMATCH2: stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing, URL: http://ideas.repec.org/c/boc/bocode/s432001.html. Lilleør, H.B., 2008a, Can future uncertainty keep children out of school?, CAM working paper no 2008-06. 116 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Lilleør, H.B., 2008b, Human Capital Diversification within the Household: Findings from Rural Tanzania, CAM working paper no 2008-04. Locoh, T., 1997, „Structures familiales: L‟apport de la demographie.‟, in M. Pilon, T. Locoh, E. Vignikin and P. Vimard (eds), In Menages et Familles en Afrique, Paris, CEPED (Les Etudes du CEPED), pp. 11–29. Pörtner, C., 2001, „Children as Insurance‟, Journal of Population Economics, 14(1): 119-136. Pasec national survey of Senegal. Piché, V. and J. Poirier, 1990, „Les théories de la transition démographique: vers une certaine convergence?‟, Sociologie et Sociétés 22(1): 179–192. Pilon, M., 2003, Foster Care and Schooling in West Africa: The state of knowledge, Preparation of the UNESCO 2003 EFA Monitoring Report, Institut de recherche pour le développement (IRD). Prywes, M., Coury, D., Fesseha, G., Hounsounou, G. and A. Kielland, 2004, Costs of Projects for Orphans and other Vulnerable Children: Case studies in Eritrea and Benin, Washington DC: World Bank, Social Protection Discussion Paper Series No 0414. Quisumbing, A. and J.A. Maluccio, 1999, Intrahousehold Allocation and Gender Relations: New Empirical Evidence, Washington DC: World Bank, Policy Research Report on Gender and Development, Working paper series no 2. Rao, V., 1997, „Wife-beating in Rural South India: A Qualitative and Econometric Analysis‟, Social Science and Medicine 44(8):1169-1180. Ravallion, M., 1997, „Famine and economics‟, Journal of Economic Literature 35(3), 1205-1242. Rosati, F.C., and Z. Tzannatos, 2003, Child Work: An Expository Framework of Altruistic and Non-Altruistic Models, Washington, DC: World Bank, Social Protection Discussion Paper no. 305. Rosenbaum, P.R. and D.B. Rubin, 1983, „The central role of the propensity score in observational studies for causal effects‟, Biometrika 70(1): 41-55. Rosenzweig, Mark R. and Kenneth I. Wolpin, 1980a, „Testing the quantity-quality fertility model: the use of twins as a natural experiment‟, Econometrica, 48(1): 227-240. Save the Children Canada, 2003, Children still in the chocolate trade, Working paper series Children in the Global Trade, no 1. Child Mobility and Rural Vulnerability | 117 Sadoulet, E., Finan, F., de Janvry, A., and R. Vakis, 2004, Can Conditional Cash Transfer Programs Improve Social Risk Management? Lessons for Education and Child Labor Outcomes, Washington DC: World Bank, SP Discussion Paper No. 0420. Serra, R., 1996, An Economic Analysis of Child Fostering in West Africa, Ph. D. diss, Christ College, University of Cambridge. Serra, R., 2000, A Theoretical Framework for Child Fostering Arrangements in Sub-Saharan Africa, unpublished. URL: http://www.econ.unian.it/ricerca/convegni/atti_cnr/serra.pdf . Serra, R, 2009, „Child fostering in Africa: when labor and schooling motives may co-exist‟, Journal of development Economics 88(1). Smith, J. and P. Todd, 2005, „Does Matching Overcome LaLonde‟s Critique of Non- experimental Estimators?‟, J. Econometrics 125(1-2): 305-353. World Bank, 2006, Managing Risks in Rural Senegal. Multi-sectoral review of efforts to reduce vulnerability, Washington DC: World Bank, Report no. 33435-SN. UCW, 2007, Enfants mendiants dans la région de Dakar, Rome: UCW Working Paper. Unicef, 2008, State of the World's Children 2009. Maternal and Newborn Healt, UNC Fund - New York: UNICEF Unicef, 2008, Evaluation of UNICEFs Sida-funded Child Protection/Trafficking Programme in West and Africa. URL: www.unicef.org/wcaro/documents_publications_2759.html Accessed 02.05.2009. United Nation, 2000, Protocol to Prevent, Suppress and Punish Trafficking in Persons, especially Women and Children, supplementing the United Nations Convention Against Transnational Organised Crime. URL: http://untreaty.un.org/English/TreatyEvent2003/Texts/treaty2E.pdf Wessels, M., 2006, Child Soldiers: From Violence to Protection, Cambridge Massachusetts: Harvard University Press. 118 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Annex I : Rural population estimates per production (sampling) zone, and per age 0-4 years 5-9 years 10-14 years 15-17 years* All children ZONE Pop. 2009 (18,9%) (17%) (14%) (6.6%) 0-17 (56,5%) ZONE RIZICOLE 1 484 059 280 487 252 290 207 768 98 936 839 482 Zone ARACHIDIERE 3 049 497 576 355 518 415 426 930 203 298 1 724 997 ZONE SYLVO PASTORALE 938 407 177 359 159 529 131 377 62 560 530 825 ZONE COTONIERE 892 279 168 641 151 687 124 919 59 485 504 732 ZONE DES NIAYES 411 672 77 806 69 984 57 634 27 444 232 869 TOTAL RURALE AREA 6 775 914 1 280 648 1 151 905 948 628 451 723 3 832 904 Source: ANSD and the 2002 Census report figures on age distribution in rural areas * Based on the dropping total shares of each 5-year cohorts, the 15-17 group was calculated as 2/3ds of the 15-19 group in the census report; 6,6 percent, instead of a 3/5 or 60 percent share. Child Mobility and Rural Vulnerability | 119 Annex II : Cluster sample rural survey. ZONE COTONNIERE (Cotton zone) DEPTEMENT ARRONDISSEMENT COMMUNAUTE RURALE N° DR Ménages 1 KOLDA DABO COUMBACARA 1 110 2 KOLDA DABO DABO 8 128 3 KOLDA DABO SALIKEGNE 1 93 4 KOLDA DABO MAMPATIM 5 118 5 KOLDA DIOULACOLON MEDINA EL HADJI 2 87 6 KOLDA DIOULACOLON TANKANTO ESCALE 11 95 7 KOLDA DIOULACOLON SARE BIDJI 6 104 8 KOLDA MEDINA Y. FOULA MEDINA Y. FOULA 8/A 82 9 KOLDA MEDINA Y. FOULA FAFACOUROU 8 120 10 KOLDA MEDINA Y. FOULA NDORNA 9 188 11 KOLDA MEDINA Y. FOULA NDORNA 17/D 129 12 KOLDA MEDINA Y. FOULA PATA 9/A 117 13 VELINGARA KOUNKANE NEMATABA 4 105 14 VELINGARA KOUNKANE KOUNKANE 9/C 52 15 VELINGARA KOUNKANE KOUNKANE 18 130 16 VELINGARA PAKOUR PAROUMBA 2/B 76 17 VELINGARA PAKOUR PAROUMBA 16 89 18 VELINGARA BONCONTO BONCONTO 1 140 19 VELINGARA BONCONTO SINTHIANG KOUNDARA 6 119 20 VELINGARA BONCONTO MEDINA GOUNASS 6/B 106 21 VELINGARA BONCONTO MEDINA GOUNASS 16/B 132 22 VELINGARA BONCONTO LINKERING 8 154 23 KEDOUGOU BANDAFASSI BANDAFASSI 110 100 24 KEDOUGOU BANDAFASSI TOMBORONKOTO 60 153 25 KEDOUGOU FONGOLIMBI MEDINA BAFFE 50 185 26 KEDOUGOU SALEMATA DAKETELY 10 90 27 KEDOUGOU SARAYA MISSIRAH SIRIMANA 40 110 28 TAMBACOUNDA KOUMPENTOUM KOUMPENTOUM 5 104 29 TAMBACOUNDA KOUMPENTOUM BAMBA NDIAYE 12 62 30 TAMBACOUNDA KOUMPENTOUM KOUTHIABA 5 206 31 TAMBACOUNDA KOUMPENTOUM KOUTHIABA 20 166 32 TAMBACOUNDA KOUMPENTOUM MALEME NIANI 5 86 33 TAMBACOUNDA KOUSSANAR KOUSSANAR 5A 111 34 TAMBACOUNDA KOUSSANAR SINTHIOU MALEME 1 104 35 TAMBACOUNDA MAKA KAHENE 2 127 36 TAMBACOUNDA MAKA KAHENE 20B 92 37 TAMBACOUNDA MAKA MAKA 17 111 38 TAMBACOUNDA MAKA NDOGA BABACAR 12 104 39 TAMBACOUNDA MISSARAH MISSIRAH 3A 120 40 TAMBACOUNDA MISSARAH MISSIRAH 13 76 120 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y ZONE ARACHIDIERE (GROUNDNUT ZONE) DEPTEMENT ARRONDISSEMENT COMMUNAUTE RURALE N° DR MÉNAGES 1 BAMBEY BABA GARAGE KEUR SAMBA KANE 78 2 BAMBEY NGOYE THIAKHAR 5O 14A 56 3 BAMBEY LAMBAYE LAMBAYE 88 4 BAMBEY LAMBAYE NGOGOM 16 O 16 O 38 5 DIOURBEL NDOULO CR : TOCKY GARE 123 6 DIOURBEL NDOULO CR : NGOHE 8 O 16O 87 7 MBACKE KAEL TOUBA MBOUL 23 8 FATICK A. NIAKHAR CR PATAR 4O 9 82 9 FATICK A. DIAKHAO CR NDIOB 4 120 10 FATICK A. TATTAGUINE CR TATTAGUINE 23 103 11 FATICK A.FIMELA CR PALMARIN F 2 215 12 GOSSAS OUADIOUR LAGANE 4 105 13 GOSSAS MBADAKHOUNE NGATHIE NAOUDE 6 105 14 GOSSAS COLOBANE COLOBANE 13 74 15 FOUNDIOUGNE A DJILOR CR DIOSSONG 24 90 16 FOUNDIOUGNE A.TOUBACOUTA CR K SALOUM DIANE 1 77 17 FOUNDIOUGNE A.NIODIOR CR DJIRNDA 3 104 18 KAFFRINE BIRKILANE MABO 16 134 19 KAFFRINE MAKA YOP IDA MOURIDE 01A 80 20 KAFFRINE MAKA YOP MAKA YOP 14 99 21 KAFFRINE MALEME HODDAR DAROU MINAM 04 139 22 KAFFRINE MALEME HODDAR MALEME HODDAR 08 105 23 KAFFRINE NGANDA KATHIOTE 07 125 24 KAOLACK SIBASSOR DYA 07 70 25 KAOLACK NDIEDIENG NDIAFFATE 05 103 26 KAOLACK KOUMBAL LATMINGUÉ 10 120 27 NIORO MÉDINA SABAKH KAYMOR 09 104 28 NIORO PAOSKOTO NGAINTH KAYES 07 106 29 NIORO PAOSKOTO PAOSKOTO 28A 80 30 NIORO WACK NGOUNA KEUR MADIABEL 08 84 31 MBOUR SESSENE NGUENIENE 12 102 32 MBOUR FISSEL FISSEL 5A 49 33 MBOUR FISSEL NDIAGANIAO 33 79 34 MBOUR SINDIA SINDIA 9A 86 35 TIVAOUANE PAMBAL CHERIF LO 3 67 36 TIVAOUANE MERINA DAKHAR PEKESS 10 55 37 TIVAOUANE NIAKHENE THILMAKHA 2 54 38 THIES K MOUSSEU FANDENE 10B 135 39 THIES THIENEBA NDIEYENE SIRAKH 18 80 40 THIES NOTTO NOTTO 22 72 Child Mobility and Rural Vulnerability | 121 ZONE RIZICOLE (RICE ZONE) COMMUNAUTE DEPTEMENT ARRONDISSEMENT RURALE N° DR MÉNAGES 1 DAGANA MBANE MBANE 04B 137 2 DAGANA ROSS-BETHIO RONKH 03 110 3 DAGANA ROSS-BETHIO ROSS-BETHIO 11 199 4 DAGANA ROSS-BETHIO ROSS-BETHIO 35 181 5 PODOR GAMADJI DODEL 18A 144 6 PODOR GAMADJI GAMADJI SARRE 10B 155 7 PODOR GAMADJI GUEDE 19 72 8 PODOR THILLE BOUBACAR FANAYE 09 112 9 PODOR THILLE BOUBACAR NDIAYE PENDAO 11B 155 10 PODOR CAS CAS MBOUMBA 03 70 11 PODOR CAS CAS AERE-LAO 09 140 12 PODOR SALDE PETE 06 119 13 PODOR SALDE GALOYA 09B 112 14 MATAM AGNAM CIVOL AGNAM CIVOL 4 147 15 MATAM AGNAM CIVOL DABIA 16 101 16 MATAM OGO BOKIDIAWE 17 165 17 MATAM OGO NABADJI CIVOL 9 191 18 MATAM OGO OGO 4 193 19 KANEL ORKADIERE AOURE 3 143 20 KANEL ORKADIERE BOKILADJI 12A 143 21 KANEL SINTHIOU BAM OURO SIDY 5A 104 22 KANEL SINTHIOU BAM SINTHIOU BAM 9 205 23 BIGNONA SINDIAN OULAMPANE 2 111 24 BIGNONA TENDOUCK MANGAGOULACK 3 86 25 BIGNONA TENDOUCK MLOMP 1 170 26 BIGNONA DIOULOULOU DIOULOULOU 7 127 27 BIGNONA DIOULOULOU KAFOUTINE 4B 159 28 BIGNONA TENGHORY NIAMONE 3 95 29 ZIGUINCHOR NIAGUIS ADEANE 5 176 30 ZIGUINCHOR NIASSIA ENAMPORE 6 194 31 OUSSOUYE CABROUSSE SANTHIABA.MAN 4 94 32 OUSSOUYE LOUDIA WOLOF OUKOUT 10 85 33 SEDHIOU TANAFF KOLIBANTANG 2 76 34 SEDHIOU DIATTACOUNDA SAMINE ESCALE 2 121 35 SEDHIOU DJIBABOUYA BEMET BIDJINI 4 83 36 SEDHIOU DIENDE DIENDE 11/B 107 37 SEDHIOU DIENDE BAMBALY 9/A 88 38 SEDHIOU BOUNKILING DIAROUME 7/A 94 39 SEDHIOU BOUNKILING NDIAMACOUTA 12/B 120 40 SEDHIOU BOUNKILING BOUNKILING 16/A 102 122 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y ZONE SYLVO PASTORALE (LIVESTOCK ZONE) DEPTEMENT ARRONDISSEMENT COMMUNAUTE RURALE N° DR MÉNAGES 1 LOUGA COKI NDIAGNE 1 72 2 LOUGA COKI THIAMÈNE 3 169 3 LOUGA MBÉDIÈNE NIOMRÉ 7 94 4 LOUGA MBÉDIÈNE MBEDIENE 7 101 5 LOUGA SAKAL SAKAL 1 89 6 LOUGA SAKAL SAKAL 22 72 7 LOUGA KEUR MOMAR SARR GANDE 6A 110 8 LOUGA KEUR MOMAR SARR KMSARR 8 133 9 LOUGA KEUR MOMAR SARR NGUER MALAL 6 95 10 KEBEMER SAGATTA THIOLOM FALL 2 73 11 KEBEMER SAGATTA GUEOUL 2 81 12 KEBEMER SAGATTA SAGAT.GUETH 11 61 13 KEBEMER DAROU MOUSTY MBADIANE 1 62 14 KEBEMER DAROU MOUSTY DAROU MOUSTY 13A 137 15 KEBEMER DAROU MOUSTY DAROU MOUSTY 29 102 16 KEBEMER DAROU MOUSTY TOUBA MERINA 5 89 17 KEBEMER NDANDE NDANDE 3 104 18 KEBEMER NDANDE DIOKOUL DIAWRIGNE 3 82 19 KEBEMER NDANDE BANDEGNE 9 110 20 LINGUERE BARKEDJI BARKEDJI 9 195 21 LINGUERE BARKEDJI THIARGNY 3 116 22 LINGUERE SAGATTA DJOLOFF BOULAL 2 90 23 LINGUERE SAGATTA DJOLOFF DEALY 5 90 24 LINGUERE SAGATTA DJOLOFF SAGATTA DJOLOFF 10 109 25 LINGUERE DODJI DODJI 1B 150 26 LINGUERE DODJI LABGAR 2 177 27 LINGUERE DODJI OUARKHOKHE 11 86 28 LINGUERE YANG YANG MBEULEUKHE 4 91 29 LINGUERE YANG YANG TESSEKERE 3C 103 30 RANEROU VELINGARA OUDALAYE 4B 257 31 RANEROU VELINGARA OUDALAYE 8D 155 32 BAKEL GOUDIRY GOUDIRY 1A 98 33 BAKEL GOUDIRY KOULOR 1A 90 34 BAKEL BALA KOTHIARY 2 83 35 BAKEL BALA DOUGUE 6 72 36 BAKEL BALA BANY ISRAEL 10A 82 37 BAKEL KIDIRA BELE 10A 99 38 BAKEL KENIEBA SADATOU 5 95 39 BAKEL MOUDERY GABOU 2 124 40 BAKEL MOUDERY BALLOU 13 91 Child Mobility and Rural Vulnerability | 123 ZONE DES NIAYES (RESIDUAL AREAS) DEPTEMENT ARRONDISSEMENT COMMUNAUTE RURALE N° DR MÉNAGES 1 TIVAOUANE PAMBAL GOUYE DIAMA 1B 99 2 TIVAOUANE PAMBAL GOUYE DIAMA 9 82 3 TIVAOUANE PAMBAL GOUYE DIAMA 17A 89 4 TIVAOUANE PAMBAL MONT ROLAND 2 98 5 TIVAOUANE PAMBAL MONT ROLAND 14 118 6 TIVAOUANE MEOUANE MEOUANE 13 104 7 TIVAOUANE MEOUANE MEOUANE 29 133 8 TIVAOUANE MEOUANE MEOUANE 27 85 9 TIVAOUANE MEOUANE DAROU KHOUDOSS 26 104 10 TIVAOUANE MEOUANE DAROU KHOUDOSS 7 94 11 TIVAOUANE MEOUANE DAROU KHOUDOSS 40 108 12 TIVAOUANE MEOUANE DAROU KHOUDOSS 4 123 13 TIVAOUANE MEOUANE DAROU KHOUDOSS 27 87 14 TIVAOUANE MEOUANE DAROU KHOUDOSS 35 93 15 THIES K MOUSSEU K MOUSSEU 1 102 16 THIES K MOUSSEU K MOUSSEU 10 96 17 THIES K MOUSSEU K MOUSSEU 19 97 18 THIES K MOUSSEU K MOUSSEU 29 79 19 THIES K MOUSSEU DIENDER GUEDJI 3 96 20 THIES K MOUSSEU DIENDER GUEDJI 10 111 21 KEBEMER NDANDE KAB GAYE 2 77 22 KEBEMER NDANDE THIEPPE 2 99 23 KEBEMER NDANDE THIEPPE 9A 103 24 LOUGA SAKAL LEONA 4B 178 25 LOUGA SAKAL LEONA 10 73 26 RUFISQUE SANGALCAM SANGALCAM 1 112 27 RUFISQUE SANGALCAM SANGALCAM 5 171 28 RUFISQUE SANGALCAM SANGALCAM 9 142 29 RUFISQUE SANGALCAM SANGALCAM 13 188 30 RUFISQUE SANGALCAM SANGALCAM 19 137 31 RUFISQUE SANGALCAM SANGALCAM 23A 190 32 RUFISQUE SANGALCAM SANGALCAM 25 178 33 RUFISQUE SANGALCAM YENE 5 128 34 RUFISQUE SANGALCAM YENE 12 83 35 SAINT LOUIS RAO GANDON 008 115 36 SAINT LOUIS RAO GANDON 017 111 37 SAINT LOUIS RAO GANDON 026 78 38 SAINT LOUIS RAO GANDON 034 115 39 SAINT LOUIS RAO MPAL 02 109 40 SAINT LOUIS RAO MPAL 12 169 124 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Annex III: Cluster sample, rural re-visit of vulnerable areas. Equipe n°1 : ENCHANTILLON_FAFO_2 N° REGION DEPTEMENT ARRONDISSEMENT COMMUNAUTE RURALE N° DR_ANSD FAFO ID 1 THIES MBOUR SINDIA SINDIA 9A 114 2 THIES MBOUR SESSENE NGUENIENE 12 111 3 THIES MBOUR FISSEL FISSEL 5A 112 4 THIES MBOUR FISSEL NDIAGANIAO 33 113 5 THIES THIES K MOUSSEU K MOUSSEU 1 15 6 THIES THIES K MOUSSEU K MOUSSEU 10 16 7 THIES THIES K MOUSSEU K MOUSSEU 19 17 8 THIES THIES K MOUSSEU K MOUSSEU 29 18 9 THIES TIVAOUANE PAMBAL GOUYE DIAMA 1B 1 10 THIES TIVAOUANE PAMBAL GOUYE DIAMA 9 2 11 THIES TIVAOUANE MEOUANE DAROU KHOUDOSS 27 13 12 THIES TIVAOUANE MEOUANE MEOUANE 29 7 13 THIES TIVAOUANE MEOUANE MEOUANE 27 8 Equipe n°2: ENCHANTILLON_FAFO_2 N° REGION DEPTEMENT ARRONDISSEMENT COM. RURALE N° DR_ANSD FAFO ID 1 LOUGA KEBEMER NDANDE KAB GAYE 2 21 2 LOUGA KEBEMER NDANDE BANDEGNE 9 179 3 LOUGA KEBEMER DAROU MOUSTY DAROU MOUSTY 29 175 4 LOUGA LOUGA SAKAL LEONA 10 25 5 LOUGA LOUGA SAKAL SAKAL 1 165 6 LOUGA LOUGA SAKAL SAKAL 22 166 7 LOUGA LOUGA MBEDIEME NIOMRE 7 163 8 LOUGA LINGUERE DODJI OUARKHOKHE 11 187 9 LOUGA LINGUERE SAGATTA DJOLOFF SAGATTA DJOLOFF 10 184 10 LOUGA LINGUERE BARKEDJI BARKEDJI 9 180 11 LOUGA LINGUERE YANG YANG TESSEKERE 3C 189 12 LOUGA LINGUERE SAGATTA DJOLOFF DEALY 5 183 Child Mobility and Rural Vulnerability | 125 Equipe n°3 : ENCHANTILLON_FAFO_2 N° REGION DEPTEMENT ARRONDISSEMENT COM. RURALE N° DR_ANSD FAFO ID 1 KAOLACK NIORO MEDINA SABAKH KAYMOR 9 107 2 KAOLACK NIORO PAOSKOTO PAOSKOTO 28A 109 3 KAOLACK NIORO PAOSKOTO NGAITH KAYES 7 108 4 KAOLACK NIORO WACK NGOUNA KEUR MADIABEL 8 110 5 KAOLACK KAFFRINE BIRKILANE MABO 16 98 6 KAOLACK KAFFRINE NGANDA KATHIOTE 7 103 7 KAOLACK KAFFRINE MAKA YOP MAKA YOP 14 100 8 KAOLACK KAFFRINE MAKA YOP IDA MOURIDE 01A 99 9 KAOLACK KAFFRINE MALEME HODDAR MALEME HODDAR 8 102 10 KAOLACK KAOLACK NDIEDIENG NDIAFFATTE 5 105 11 KAOLACK KAOLACK SIBASSOR DYA 7 104 Equipe n°4 : ENCHANTILLON_FAFO_2 N° REGION DEPTEMENT ARRONDISSEMENT COM. RURALE N° DR_ANSD FAFO ID 1 KOLDA KOLDA DIOULACOLON TANKANTO ESCALE 11 126 2 KOLDA KOLDA DIOULACOLON SARE BIDJI 6 127 3 KOLDA KOLDA DIOULACOLON MEDINA EL HADJI 2 125 4 KOLDA KOLDA MEDINA Y. FOULA MEDINA Y. FOULA 8/A 128 5 KOLDA KOLDA MEDINA Y. FOULA PATA 9/A 132 6 KOLDA KOLDA MEDINA Y. FOULA NDORNA 17/D 131 7 KOLDA KOLDA MEDINA Y. FOULA NDORNA 9 130 8 KOLDA KOLDA DABO DABO 8 122 9 KOLDA VELINGARA KOUNKANE KOUNKANE 18 135 10 KOLDA VELINGARA KOUNKANE NEMATABA 4 133 11 KOLDA VELINGARA BONCONTO MEDINA GOUNASS 6/B 140 12 KOLDA SEDHIOU TANAFF KOLIBANTANG 2 73 126 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Annex IV. Econometric framework: propensity score matching methods Although there are many theoretical reasons why climate-related shocks lead to migration and child relocation, it is difficult to assess effect of these shocks based on non-experimental observations. This is because the counterfactual outcome of what the migration and child relocation status would be like if the state of the household without a shock is not observed. In experimental studies, this problem is addressed by randomly assigning households to treatment and controlled groups, which assures that the outcomes observed on the controlled groups without shocks are statistically representative of what would have occurred without shock on the treatment households. Conducting such an experiment is impossible in climate-related aspects. However, in social settings where one tries to capture the effect of a climate-related shock on issues such as migration and child relocation, selection bias would be an issue of concern. Therefore, simple comparison of the outcome variable between households who experienced shocks (shocked) and households without a shock (non-shocked) would yield biased estimates of impact of a shock. We adopt the semi-parametric matching method which does not require an exclusion restriction or a particular specification of the selection equation to construct the counterfactual and reduce selection problems. Our main purpose of matching is to find a group of treated individuals (shocked households) which are similar to the control groups (non-shocked households) in all relevant pre-treatment characteristics, the only remaining difference being that one group was subjected to a shock and other group did not. On the details of the PSM method we refer to several studies (e.g., Rosenbaum and Rubin, 1983; Dehejia and Wahba, 2002; Heckman et al., 1998; Caliendo and Kopeinig, 2008; Smith and Todd, 2005). Estimation of the average treatment effects on the treated (ATT) using matching methods relies on two key assumptions. The first is the Conditional Independence Assumption (CIA), which implies that selection into the 18 treatment is solely based on observable characteristics (selection on observables) . Matching on every covariate is difficult to implement when the set of covariates is large. To solve this dimensionality problem we estimate the propensity score – the conditional probability P( x )  P(d i i  1 xi )  that the ith individual is subjected to a xi di  1 shock conditional on observed characteristics ( ); where is when the ith individual is subjected to a d 0 shock, and i when no shock occurs. The second assumption is the common support or overlap condition. The common support is the region where the balancing score has positive density for both treatment and comparison units. No matches can be formed to estimate the average treatment effects on the treated (ATT) parameter when there is no overlap between the treatment and non-treatment groups. The matching process is performed in two steps. First, we use a logit model to estimate the propensity score, and in the second step, we estimate the ATT, conditional on the propensity score. In the estimation of the propensity score, we are not interested in the effects of covariates on the propensity score since the purpose of our work is to assess the impact of climate-related shock on outcomes such as adult migration and child relocation. However, the choice of covariates to be included in the first step (propensity score estimation) is an issue. Heckman et al. (1997) show that omitting important variables can increase the bias in the resulting estimation, but in general only variables that simultaneously influence the household’s likelihood of being subjected to a shock and the outcome variable, and which in turn, are unaffected by experiencing a shock, should be included. Bryon et al. (2002) also recommended against over-parameterized models because including extraneous variables in the model used to characterize the likelihood of facing a shock will reduce the likelihood of finding a common support. Rosenbaum and Rubin (1983), Dehejia and Wahba (2002) and Diprete and Gangl (2004) emphasize that the crucial issue is to ensure that the balancing condition is satisfied, because it reduces the influence of confounding variables. We follow this approach by applying the method of covariate balance, i.e., the equality of the means on the scores and equality of the means on all covariates, between treated and non-treated The assignment to treatment is independent of the outcomes, conditional on the covariates . 18 Child Mobility and Rural Vulnerability | 127 individuals. To check whether the matching procedure is able to balance the covariates, we use the standardized bias measure, a two sample t-tests, pseudo- R from logit of shock on covariates and p  values of the 2 likelihood ratio test of joint significance of covariates before matching and on matched samples (Rosenbaum and Rubin, 1985; Leuven and Sianesi, 2003; Diprete and Gangl, 2004). For each covariate x, the standardized bias (SB) is the difference of the sample means in the treated and matched control subsamples as a percentage of the 2 square root of the average of the sample variances in both groups (Diprete and Gangl, 2004). The pseudo- R indicates how well the regressors explain the probability of facing a shock. After matching, there should be no 2 systematic differences in the distribution of covariates between both groups, as a result the pseudo- R should be low and the p  value of the likelihood ratio test should be insignificant. Estimation of the propensity score perse, is not enough to estimate the ATT of interest. Since propensity score is a continuous variable, the probability of observing two units with exactly the same propensity score is in principle zero. Various matching algorithms have been proposed in the literature to overcome this problem. Asymptotically, all matching algorithms should yield the same results. However, in practice, there are tradeoffs in terms of bias and efficiency involved with each algorithm (Caliendo and Kopeining, 2008). They suggest trying a number approaches; therefore, we implement three matching algorithms (i) one-to-one nearest neighbor matching, (ii) radius matching and (iii) kernel matching. At the bottom line, these methods numerically search for “neighbors� that have a propensity score of non-treated individuals which is very close to the propensity score of treated individuals. We omit further details here for brevity and refer to the growing literature on matching methods (e.g., Rosenbaum and Rubin, 1983; Dehejia and Wahba, 2002; Heckman et al., 1998; Caliendo and Kopeinig, 2008; Smith and Todd, 2005). Results and discussion The results from the logit analysis of shocks and the variables used in the matching procedures are reported in the table below. Estimation of propensity score (Probit model) (Dependent variable: Shocks (0/1) Variables Estimates Coefficient St. Err P-value Household characteristics Age of hh head 0.001 0.003 0.642 Gender of hh head 0.146 0.188 0.437 Education of hh head -0.017 0.044 0.698 Dependency ratio 0.024 0.044 0.582 Ethnic: Pehul 0.131 0.126 0.298 Ethnic:Serer 0.078 0.153 0.610 Ethnic: Others -0.189 0.157 0.229 Risk aversion 1 -0.183 0.084 0.030 Risk aversion 2 0.185 0.079 0.019 Occupation of hh head Agriculture 0.604 0.155 0.000 Livestock 1.792 0.316 0.000 Livestock and agriculture 0.970 0.178 0.000 Infrastructure and asset Electricity 0.039 0.123 0.755 Refrigerator -0.716 0.233 0.002 Wall type (Concrete=1) 0.424 0.131 0.001 Quality of accessed water (1=high quality, 6=low 0.117 0.056 0.036 Environmental characteristics quality) Coefficient of variation of rainfall: August -0.702 0.169 0.000 Coefficient of variation of rainfall: September -0.107 0.147 0.467 128 | C h i l d M o b i l i t y a n d R u r a l V u l n e r a b i l i t y Cotton producing zone -0.574 0.151 0.000 Groundnut producing zone -0.775 0.166 0.000 Pastoral zone -0.840 0.344 0.014 Rice producing zone -0.702 0.169 0.000 Constant -0.107 0.147 0.467 Sample size 2233 Log-likelihood -1470.57 Among household who experienced shocks, the predicted propensity score ranges from 0.149 to 0.888 with a mean of 0.514. Among non-shocked households, the predicted propensity score ranges from 0.098 to 0.803 with a mean of 0.449. Thus, the common support assumption is satisfied in the region of [0.149, 0.889] enforcing a loss of 8 shocked households. The density distributions of the propensity scores (see figure below) also support the common support or overlap region for treated and non-treated groups. The bottom-half of each graph shows the propensity score distribution for the non-treated, while the upper-half refers to the treated individuals. Common support or overlap region 0 .2 .4 .6 .8 Propensity Score Untreated Treated: On support Treated: Off support 19 The figure above presents results of overall covariate balancing tests before and after matching. The standardized bias measure results show that the difference in propensity score of unmatched treated and control sample is close to 100% (p<0.001). After matching the bias significantly reduced well below 2%. The bias is also 20 significantly reduced for each covariate after matching compared to before matching. Although, we do not have a clear indication for the success of the matching procedure, in most empirical studies a bias reduction below 3% or 5% is seen as sufficient (Caliendo and Kopeining, 2008). A two sample t-tests used to check if there are significant differences in covariate means for both groups also confirm this result. Before matching, several variables exhibit statistically significant differences. However, after matching, the covariates are balanced. The low pseudo-R2 and the insignificant likelihood ratio tests support the hypothesis that both groups have the same 19 The results are obtained by using the Stata pstest module (Leuven and Sianesi 2003). 20 These results are not reported in the interest of space but available upon request. Child Mobility and Rural Vulnerability | 129 distribution in covariate x after matching. These results clearly show that the matching procedure is able to balance the characteristics in the treated and the matched comparison groups. We will therefore use these results to evaluate the effect of climate-related shocks on adult migration and child relocation among groups of households having similar observed characteristics. This allows comparison of observed outcomes for shocked households with that of a comparison group that shares a common support. Covariate balancing before and after matching Before matching After matching Kernel Nearest neighbor Radius matching matching matching Mean standardized bias 10.05 1.24 1.89 1.21 Pseudo R2 0.049 0.001 0.002 0.001 LR � 2 ( p  value) 151.07 (0.000) 2.64 (1.000) 5.67 (1.000) 2.53(1.000) Estimation of treatment effect: Matching algorithms To compute the average treatment effect (ATT) accurately, one should match the treated and untreated groups precisely on the basis of the propensity score. In practice, it is difficult to achieve this; however three alternative matching methods (nearest neighbor matching, radius and kernel matching) were used and compared. It should be noted that all the analyses were based on implementation of common support and caliper, so that the distribution of treated and untreated units were located in the same domain. As suggested by Rosenbaum and Rubin (1983), a caliper size of one quarter of the standard deviation of the propensity scores was used. The table provides the estimates of treatment effect (ATT) for the outcome variables obtained from the three 21 matching algorithms. The outcome variables are adult migration from the household and child relocation at the household level. The results indicate climate-related shocks lead to a significant increase in adult migration. The percentage of household with adult migration ranges from5.2 to 5.6% using the three algorithms. However, climate-related shocks do not have a significant impact on child relocation exhibited by the insignificant impacts as shown in Table 6. Impact of climate-related shock on adult migration and child relocation Outcome Matching method ATT Std. Error t-value Adult migration Kernel 0.056 0.016 3.47*** Radius 0.056 0.016 3.48*** Nearest neighbor 0.052 0.021 2.54*** Child relocation Kernel 0.026 0.021 1.27 Radius 0.026 0.021 1.25 Nearest neighbor 0.025 0.027 0.92 N (Households experiencing climate- 1073 related shocks)1 N (Households without climate-related 1160 shocks) *** Significant at 5 % level. 21 Matching is implemented using the Stata module psmatch2 (Leuven and Sianesi, 2003).