After the Shock: Mobility, Inequality and Social Exclusion Child mobility and domestic services as coping and strategy: What implications for social protection interventions in rural Benin? Project report June 2012 Anne Kielland, Fafo Institute for Applied International Studies, Norway Gilberte Hounsounou, Cabinet Stigmate, Benin World Bank Task Manager: John Van Dyck 2 | Mobility inequality and social exclusion Cover : The girl on the cover is Perpetue Dagba. For many years she served as a domestic worker for her relatives, cooking, cleaning and caring for young children. Miss Dagba never had the chance to go to school. Instead, the employing family and their acquaintances helped her pay for an apprenticeship. She now works as a hairdresser. Miss Dagba has given her permission to the use of this picture from the time when she worked as a child domestic. Ouidah, 23.06.2012 Mobility, inequality and social exclusion | 3 After the shock: Mobility, Inequality and Social Exclusion Child mobility and domestic services as coping and strategy: - What implications for social protection interventions in rural Benin? Project report, June 2012 Anne Kielland, Fafo Institute for Applied International Studies, Norway Gilberte Hounsounou, Cabinet Stigmate, Benin World Bank Task Manager: John Van Dyck Project web site: www.fafo.no/childsrm 4 | Mobility inequality and social exclusion Disclaimer: The findings, interpretations, and conclusions expressed in this paper 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 Mobility, inequality and social exclusion | 5 Contents Introduction ..................................................................................................................................... 1 Summary of statistical findings ...................................................................................................... 4 1 Background and objective for the study ................................................................................ 11 Inequality and social exclusion: a downward spiral? ................................................................ 11 The normality of domestic work and the market for its’ services ............................................. 14 Risks, shocks and the need for social protection ....................................................................... 17 2 Methodology .......................................................................................................................... 19 The survey ................................................................................................................................. 19 Survey approach .................................................................................................................... 19 The sample ............................................................................................................................. 20 Survey modules ..................................................................................................................... 20 Non-response ......................................................................................................................... 21 The data ................................................................................................................................. 22 Weights .................................................................................................................................. 24 The qualitative work.................................................................................................................. 25 Theme 1: The social institution of domestic work ................................................................ 26 Theme 2: Mobility and household security strategies ........................................................... 27 3 Numbers and shares ............................................................................................................... 29 The surveyed children ............................................................................................................... 30 Households descriptives ............................................................................................................ 34 Household poverty in the sampled households ..................................................................... 34 Risks and shocks .................................................................................................................... 36 Child mobility: numbers and attitudes ...................................................................................... 40 Girls who have left for domestic work: numbers and attitudes................................................. 48 6 | Mobility inequality and social exclusion 4 Relationships: mobility, poverty and shocks ......................................................................... 53 Child relocation ......................................................................................................................... 53 Child mobility and household poverty .................................................................................. 54 Child mobility and the 2010 floods ....................................................................................... 57 Child mobility and other shocks ............................................................................................ 60 Girls and domestic work ........................................................................................................... 63 Domestic work and household poverty ................................................................................. 63 Domestic work and the 2010 floods ...................................................................................... 66 Domestic work and other shocks ........................................................................................... 67 Estimating effects, a regression approach ................................................................................. 70 5 Implications for social protection .......................................................................................... 77 Effectiveness of existing, informal safety nets ...................................................................... 80 Targeting ................................................................................................................................ 82 Conditionality ........................................................................................................................ 90 Recommendations ................................................................................................................. 93 6 Final words ............................................................................................................................ 95 Annex I : Cluster sample............................................................................................................... 97 Annex I : Participants at the Validation Workshop in Cotonou 21.06.2012 .............................. 103 Mobility, inequality and social exclusion | 7 Table of figures Figure 1. A downward spiral of social exclusion. .................................................................................................. 13 Figure 2. Predicted quality of inter-household exchange of child or domestic helper, based on the dominance of social/strategic versus economic/coping motives in the household of origin and the new household. ...... 16 Figure 3. Share of household women 16 – 70 years-of-age found eligible for parent interviews (in percent of female household members). ..................................................................................................................... 23 Figure 4. Share of children 0-17 years-old holding a birth certificate (in percent of all children). ......................... 31 Figure 5. Share of children 0-17 years-old reported to be in good health (in percent of all children). .................. 32 Figure 6. Share of children, 6-17 years-of-age, currently attending school on region, gender and age bracket (in percent of all children). ........................................................................................................................ 33 Figure 7. Share of children, 6-17 years-of-age, who have never been to school on region, gender and age bracket (in percent of all children). ........................................................................................................... 33 Figure 8. Share of household heads that worried about being able to provide their family with food during the past 12 months and in the future 12 months, on level of worry and region (in percent of all households). 36 Figure 9. Types of consequences of being affected by the 2010 floods, by region (in percent of those reporting to have been affected). ................................................................................................................................... 37 Figure 10. Types of shocks suffered by sampled rural households, by region (in percent of all households). ....... 39 Figure 11. Mobility prevalence, current and in the past, by age and gender (in percent of all children). ............. 41 Figure 12. Who the mobile children live with, by gender (in percent of all relocated children). ........................... 42 Figure 13. Where the mobile children reside, by gender (in percent of all relocated children). ........................... 43 Figure 14. Objective for current relocation, by gender: boys versus girls (in percent of all relocated children). ... 43 Figure 15.Attitudes to child mobility: Do you look positively on the following types of child mobility? (In percent of all household heads)............................................................................................................................... 44 Figure 16. Attitudes to child mobility: Would you have liked for your son/daughter to relocate for any of the following reasons? (In percent of all household heads). ............................................................................. 45 Figure 17. Attitudes to child mobility: Views on the treatment of relocated children (in percent of all household heads)......................................................................................................................................................... 46 Figure 18. Attitudes to child mobility: What would be your expectations to the household where your child is placed (in percent of all household heads). ................................................................................................ 47 Figure 19. Schooling status of currently relocated girls 6-20 years-old, on stated objective for leaving (in percent of all relocated girls). .................................................................................................................................. 49 Figure 20. Share of girls 6-20 year-old who have worked as domestics for others than their household of origin, by age-bracket and region (in percent of all girls). ...................................................................................... 50 Figure 21. Rural household heads’ attitudes to the practice of sending children for domestic work elsewhere (in percent of household heads). ..................................................................................................................... 51 8 | Mobility inequality and social exclusion Figure 22. Rural household heads’ attitudes to treatment of “disobedient� domestic workers (in percent of all household heads). ...................................................................................................................................... 52 Figure 23. Poverty and child mobility: Housing quality. Share of households with relocated children on number of solid parts (floor, roof, walls) and mobility type. .................................................................................... 54 Figure 24. Poverty and child mobility: Asset ownership. Share of households with relocated children on number of assets owned and mobility type (in percent of households with children 0-17 years-of-age). ................ 55 Figure 25. Poverty and child mobility: Food security. Share of households with relocated children on level of worry over ability to provide household members with food during the past 12 months (in percent of all households with children 0-17 years-of-age). ............................................................................................. 56 Figure 26. Share of households with at least one child living away from home, on type of consequence suffered from the 2010 floods (in percent of households with children 0-17 years-of-age). ..................................... 58 Figure 27. Share of households with at least one child living away from home and not going to school, on type of consequence suffered from the 2010 floods (in percent of households with children 0-17 years-of-age). .. 59 Figure 28. Share of households with at least one child living away from home on type of shock suffered in the past decade (in percent of households with children 0-17 years-of-age). ................................................... 60 Figure 29. Share of households with at least one child living away from home and not going to school on type of shock suffered in the past decade (in percent of households with children 0-17 years-of-age). ................. 61 Figure 30. Share of households with at least one child living away from home, on type coping mechanism applied during the last shock to the household (in percent of households with children 0-17 years-of-age). ................................................................................................................................................................... 62 Figure 31. Poverty and domestic work: Housing quality. Share of households with girl domestic workers on number of solid parts (floor, roof, walls) and mobility type (in percent of all households with girls 6-20 years-of-age). ............................................................................................................................................. 64 Figure 32. Poverty and child mobility: Asset ownership. Share of households with girl domestic workers on number assets owned and mobility type (in percent of all households with girls 6-20 years-of-age). ......... 65 Figure 33. Poverty and child mobility: Food security. Share of households with girl domestic workers on level of worry over ability to provide household members with food during the past 12 months (in percent of all households with girls 6-20 years-of-age). ................................................................................................... 65 Figure 34. Share of households where at least one girl has left to do domestic work, on type of consequence suffered from the 2010 floods (in percent of all households with girls 6-20 years-of-age). ......................... 67 Figure 35. Share of households where at least one girl has left to do domestic work, on type of shock suffered in the past decade (in percent of all households with girls 6-20 years-of-age). ............................................... 68 Figure 36. Share of households where at least one girl has left to do domestic work, on type coping mechanism applied during the last shock to the household (in percent of all households with girls 6-20 years-of-age).69 Mobility, inequality and social exclusion | 9 Figure 37. Probability of girls relocating for domestic work: Impact of indebtedness after the 2010 Floods, on region and wealth (assessed by housing quality scores of 0-3). Simulated results from a logit regression. Fixed household features: male, illiterate household head, non-Muslim, distrustful attitude. (In percent of all households) .......................................................................................................................................... 75 Figure 38. Men and women’s account of who shoulders main responsibility for different groups of household expenditures (in percent of male-headed households). .............................................................................. 83 Figure 39. Women’s account of who shoulders main responsibility for groups of expenditures in their households, on region (in percent of male-headed households)................................................................. 84 Figure 40. Women’s account of who shoulders main responsibility for groups of expenditures, on how much the household head worries about being able to provide food for his family during the next 12 months (never worries, worries sometimes, worries a lot). ................................................................................................ 86 Figure 41. Household heads’ account of who shoulders main responsibility for expenditure groups and the schooling status of children 6-14 years-of-age (in percent of households). ................................................. 88 Figure 42. Household heads’ account of who shoulders main responsibility for child related expenditure groups and the mobility of children 0-17 years-of-age for other purposes than schooling (in percent of households). ............................................................................................................................................... 88 Figure 43. Household heads’ account of who shoulders main responsibility for child related expenditure groups and the mobility of girls (6-19 years-of-age) for domestic work (in percent of households). ...................... 89 Figure 44. If there were no expenses related to schooli ng, how many children would you send (“All� or “Some�)? On poverty (based on housing quality scores 0-3) and region (in percent of all households with children 6- 14 years of age). ......................................................................................................................................... 92 Table of tables Table 1. Total number of rural children 0-17 years old represented in the rural communes of the sampling frames and the outside rural communes. ............................................................................................................... 30 Table 2. The housing quality of the sampled households, by building part material and region (in percent of all households). ............................................................................................................................................... 34 Table 3. Assets counted in the surveyed households, on strata (in percent of all households). ........................... 35 Table 4. Extrapolated mobility numbers for children 0-17 years on gender and strata with extrapolations to outside the regional sample frames (SF). .................................................................................................... 40 Table 5. Regressing child mobility for other reasons than schooling on flood debt, controlling for socio- demographic and wealth indicators (logit). .............................................................................................. 71 Table 6. Regressing mobility for domestic work on flood debt, controlling for socio-demographic and wealth indicators (logit). ....................................................................................................................................... 73 Mobility, inequality and social exclusion | 1 Introduction How is inequality produced and reproduced? Citizens are born into different social and economic realities, and with a gender that affects both their initial social status and their opportunities later in life. Besides own capacities and efforts, their further destiny is determined by the social processes they become part of. Many social processes serve to reproduce status quo. Some lead towards social and economic integration of the previously disadvantaged, others to disintegration and gradual social exclusion. Social inclusion and exclusion are relative and ever-changing features. Their development does, however, bear the characteristics of upwards or downwards spirals, where each loop up or down tends to reinforce its’ direction. Shocks may lead to large leaps downwards a negative spiral, or to the changing of direction of a positive one. Outcomes may be determined by the availability and quality of informal mitigation and coping instruments, and these are unevenly distributed in the population. A main purpose of social protection policies is therefore to support shock mitigation and coping in vulnerable populations: to prevent large and permanent leaps down the spiral and instead provide a springboard that helps people bounce back up on their feet. Citizens get involved in social processes contributing to inclusion or exclusion at a very early age. This report takes point of departure in the traditional system of circulating children between households in Benin. While the practice may seem like a breach between the child and his or her family, it may also constitute an integration of the child into the extended social network of the family. A special focus of the report is the social institution of domestic services – a social system that interferes with and overlaps with traditional child and youth circulation. Mobility and migration can be both a strategic move in normal times, and a coping strategy responding to social or economic shocks. Domestic work is relatively easy to obtain for rural girls and young women in Benin. Poorer girls and girls from families that have suffered shocks are likely, however, to accept poorer labor conditions and higher-risk arrangements. Household experiences with the floods of 2010 will feature as an important variable in this report, among other more traditional determinants. Among the 3000 rural households surveyed, 2 | Mobility inequality and social exclusion 46 percent reported to have been affected by the floods in one way or the other. Households were also severely affected by other individual shocks or systemic crises, such as drought or the illness or death of a household member. Other shocks, like fire, were less common, but had extremely serious consequences to those affected. It deserves a special mentioning that one of the clusters of the survey burned to the ground by one of those fires short time after the survey ended, and our hearts go out to all the families affected. The report is organized into 5 main parts. After an initial outline of the background and objectives for the study, a separate chapter details some of the methodological issues behind the work. The third chapter provides mainly descriptive results related to the children of the survey and their households, and leads up to an estimate of the current number of relocated rural children in Benin, as well as estimates of numbers of girls who do or have done domestic work while living away from their own households. In the fourth chapter, child mobility and girl’s domestic work is placed in the context of the poverty, shocks and coping efforts of their households. In chapter five this and other research results are discussed in context of their implications for social protection programs in general and safety net efforts in particular. The research team is grateful to the Social Protection team of the World Bank and the PopPov Research Network for funding the study. Hopefully, the results will be helpful to the further efforts of providing social protection and safety nets to the poor in Benin, as well as contribute to the knowledge development on population and growth issues taking place in the PopPov Network. As project coordinator, I will humbly express my gratefulness to all the people who have been involved in the project. A special thanks to Gilberte Hounsounou and Bertin Affognon of Cabinet Stigmate for taking on the huge responsibility for coordinating and implementing the research on the ground. A special thanks to Pierre Jekinnou for his contributions, especially to make the qualitative work successful. To professors Adolphe Kpatchavi, Ernest Amoussou for training the field teams. To my colleagues at Fafo, Liv Elin Torheim, Jing Liu and Huafeng Zhang for technical assistance and programming support. We also very much appreciated the programming support of Said Jekinnou. The field teams were led by Antoinette Houédjissi, Jean Affo Assongba, Darius Vegba, Romule Gbodja, Alfred Mpo, Ramane Imorou Chabi, and Pierre Mèliho.. The qualitative work was conducted by Septime Atchekpe, Timothée Codjo Togbé, Mobility, inequality and social exclusion | 3 Parfaite Kotcharé Zoblikpo, Florent Tasso, Adizath Bouko, Marceline D. Montcho and Elisabeth Koudénoupko. Last but not least, the interviewers: Salomon Jekinnou, Claude Bossou, Nathalie Bankolé, Frédéric Montcho, Ramatou Yessoufou, Estelle Kinnouezan, Pélagie O. Dah Alodé, Roseline Zannou, Gilbert Affognon, Donatien Zannou, Symiaki K. Balabawi, Darius Aïssi, Abdel Kader Karimou B., Taïbou Koto, Awaou Kora, Stéphane Sossou, Symphorien Assani, Blaise Djihoun, Mariette Aikpe, Afiwa Hadonou, Natacha Sezan, Gisèle Sinasson, Essèdo R. Segbemon, Soliatou Adjokê, Gilbert Edoun Odjouafé, Barthélémy Akomedi, Augustin Kouagou, Rock Charles Atchadé, Prisca Gbehou, Abdoulaye Daleb, Imelda Agondanou, Chimène Jekinnou, Bernis Soglo, Fataou Aremou and Joël Tohoundjo . Please accept my sincere gratitude. Finally, we are all grateful to the Central Statistical Agency of Benin (INSAE) and the Ministry for Family and National Solidarity for invaluable support. We hope our research can to be helpful to their further efforts to support women, children and other vulnerable populations in Benin. June 21, 2012, a dissemination and validation workshop was organized in Cotonou. Fifty-six people attended, representing The World Bank, Unicef, ILO, Central ministers, NGOs and other social actors in Benin. The questions and comments proved beyond doubt the engagement and knowledge that exists among Beninese stakeholders. The fact that at least 40 participants stayed two hours over time to involve in group discussions shows their commitment. And we rest assured that the continued struggle for women and children will continue, led by such forceful advocates and technical experts. Anne Kielland Project Coordinator 4 | Mobility inequality and social exclusion Summary of statistical findings Child mobility  Rural children in Benin leave home for a variety of reasons, including schooling and apprenticeships. In February 2012, an estimated 180,000 children 0-17 years-of-age currently lived away from their households of origin in rural Benin (Table 4).  Among them there are 73,000 boys and 106,000 girls (Table 4).  An additional 50,000 children have previously lived away from their household of origin for more than a 2 month period, and then returned (Table 4).  Mobility prevalence is highest in the Southern part of the country (7,3 percent for boys and 12,2 percent for girls), also high in the Central areas (6,5 and 9 percent respectively), and quite low in the North (1,1 and 1,2 percent) (Table 4).  Mobility increases by age. While 8 percent of boys and 14 percent of girls 11-14 years- of-age live away from home, 15 percent of boys and 18 percent of girls 15-17 years old have left their households of origin (Figure 11).  The majority of relocated children live with relatives. Among the boys, 10 percent live with a tutor and 6 percent win an employer. The same figures for girls are 8 and 5 percent respectively (Figure 12).  Fifty-eight percent of the girls and 54 percent of the boys have relocated towards an urbanized community. While less than 10 percent stay within their own community, around 15 percent live in a neighboring rural community (Figure 13).  Forty-two percent of the boys and 38 percent of the girls were stated to have left to go to formal school. However, 5 percent of these children never got to start school and another 5 percent had dropped out. Thirty-two percent of the boys and 22 percent of the girls had left for an apprenticeship. The remaining 26 percent of boys and 40 percent of girls left for a variety of social, economic and labor motivations. (Figure 14).  Three-in-four rural household heads have a positive or conditionally positive view on child relocation for apprenticeships elsewhere. Two-in-tree are positive or conditionally positive to the practice of placing children with relatives. Around half appreciates the Mobility, inequality and social exclusion | 5 practice of children going away for agricultural work, and slightly fewer for domestic work (Figure 15)  Household heads with children below 18 years-of-age were asked if they would welcome an opportunity for their child to relocate for various purposes. Around 40 percent think it would be good for the child to go and live with urban relatives, while a little less than 30 percent also find it desirable for the children to go live with relatives in other rural areas. Thirteen percent is positive to having a daughter work as a domestic elsewhere, while one-in-four would be positive to having one of their sons going away to work in agriculture (Figure 16).  When asked why they did not let their children go, if their views were positive, around half responded they worried about the treatment the child would get. Thirty percent said they did not know of good people who could facilitate for the relocation or employ the child.  Children placed with relatives and tutors are sometimes exposed to negative treatment. Two-in-three accept or conditionally accept corporal punishment, while 60 percent find an occasional slap acceptable or conditionally acceptable. The most accepted form of such treatments among the rural household heads is reprimanding. (Figure 17).  Many parents would have certain expectations if their children were to stay with relatives. Most commonly, 94 percent say they expect the relatives to help educate the child, while 84 percent stress the nourishing and protecting of the child, and 16 percent expect help to find a spouse for the child. Only 17 percent say they would expect to receive money on a regular basis, while 28 percent say they would expect the child’s new household to support them in case of a crisis. (Figure 18) Domestic service  Among the rural household members of the survey, several girls had lived away from home at some stage in the past. Among the sub-sample of 15-20 year old household members who had previously been away, 34 percent had done domestic work while studying, 31 percent had “helped� with domestic work while away and not studying, while 14 percent explicitly said they had been domestic workers. 6 | Mobility inequality and social exclusion  The fertility rosters of the household parents show that around 30,000 girls between 6 and 20 year of age were currently away from their rural homes explicitly to be domestic workers.  However, among the relocated girls and young women in this age group, around 60,000 are out of school, with no other stated job, no apprenticeship and had not left to marry (Figure 19).  Asking directly if each girl had ever been away from home to do domestic work, 5 percent of the 6-10 year-olds already had such an experience, similarly 6 percent of the 11-14 years-olds, 9 percent of the 15-17 year-olds and 13 percent of the 18-20 year-olds (Figure 20). These percentage shares translate into 70,000 girls and yond women 6-20 years-of-age who are explicitly stated to have worked as domestics.  Experience with domestic service is much more evenly distributed on the three regional strata (North, Central and South) than child mobility in general (Figure 20).  Twelve percent of household heads are positive to the practice of young girls leaving to do domestic work, while 30 additional percent are positive given certain conditions. Household heads in the South are most skeptical; 65 percent do not welcome the practice (Figure 21).  Similar to treatment of relocated children, there is a relatively high acceptance rate for corporal punishment of domestic workers. 25 percent finds it acceptable, while 35 finds it conditionally acceptable. Fifty-five percent accepts or conditionally accepts an occasional slap. Parents in the North are more accepting of corporal punishment of domestic workers (Figure 22). Child Mobility and Poverty  Child mobility correlates with poverty assessed by housing quality. Eighteen percent of households with poor household buildings have at least one child living away from home, and 12 percent of these households have at least one child living away for other purposes than schooling. Fifteen percent of households with high building standard have at least one child living away, but only 7 percent of these households have a child living away for other purposes than schooling (Figure 23). Mobility, inequality and social exclusion | 7  Asset poverty also correlates with child mobility. Twenty-three percent of households with no durable assets have at least one child living away from home, and 16 percent one child away for other purposes than schooling. Fifteen percent of households with 5 or more durable assets have a child living away, and 9 percent of these households have a child away for other purposes than schooling (Figure 24).  There is a strong correlation between child mobility and the level of worry for the food security of the household in the next 12 months. Twenty-three percent of households where the household head worries a lot have at least one child away, and 15 percent of these households have child away for other reasons than schooling. In comparison, only 7 percent of households where the head never worries about food security have a child living away, and only 5 percent of these households have a child away for other reasons than schooling (Figure 25). Child Mobility and Shocks  The 46 percent of the households that reported to have been affected by the 2010 floods did not report significantly different child mobility figures from the rest of the sample. However, households reporting certain consequences of the flood to their households had higher mobility rates. The highest child mobility rates were found among households that had to borrow money to cope with the impact of the floods, households where members had been forced to flee, and households that had been suffering from illness and malnutrition in the wake of the disaster (Figure 26 and Figure 27)  Among other shocks correlating with child mobility, individual household shocks like the death of a breadwinner or the illness of a member seem particularly important (Figure 28 and Figure 29)  The ways household cope with shocks also correlate with child mobility. The highest mobility rates were found in households that had to reduce the number of meals and sell assets to cope with the last serious shock (Figure 30) 8 | Mobility inequality and social exclusion Domestic work and poverty  Domestic work correlates with poverty assessed by housing quality. Twelve percent of households with poor household buildings have at least one girl who has worked as a domestic, and 7 percent of these households have at least one girl currently living away for domestic work. Seven percent of households with high building standard have at least one girl who has worked as a domestic, but only 4 percent of these households have a girl currently living away for domestic work (Figure 31)  Asset poverty also correlates with domestic work. Fourteen percent of households with no durable assets have at least one girl who has worked as a domestic, and 7 percent one girl currently living away for domestic work. Ten percent of households with 5 or more durable assets have one girl who has worked as a domestic, and 4 percent of these households have a girl currently living away for domestic work (Figure 32).  There is a strong correlation between domestic work and the level of worry for the food security of the household in the next 12 months. Fourteen percent of households where the household head worries a lot have at least one girl who has worked as a domestic, and 9 percent of these households have a girl currently living away for domestic work. In comparison, only 6 percent of households where the head never worries about food security have at least one girl who has worked as a domestic, and only 2 percent of these households have a girl currently living away for domestic work (Figure 33). Domestic work and shocks  The 46 percent of the households that reported to have been affected by the 2010 floods did not report significantly different figures for girls being away for domestic work compared to the rest of the sample. However, households reporting certain consequences of the flood to their households had higher domestics rates. The highest rates of girls working as domestics were found among households that had to borrow money to cope with the impact of the floods, households where members had been forced to flee, households that had been suffering from illness and malnutrition in the wake of the disaster, and households that had lost farm animals or fish farms (Figure 26) Mobility, inequality and social exclusion | 9  Among other shocks correlating with mobility of girls for domestic work, individual household shocks like the death of a breadwinner or the illness of a member seem particularly important, alongside more covariate type shocks like fire (Figure 35). The ways households cope with shocks also correlate with domestic work. Again, the highest rates were found in households that had to reduce the number of meals and sell assets to cope with the last serious shock, but also in households that had to sell land or livestock (Figure 30 Implications for Social Protection  Child mobility can benefit both children and their families, but is sometimes harmful. Domestic work is a profession and should not be neither harmful not socially degrading when normal labor rights are in place. Relocation and labor conditions are difficult to negotiate when the family suffers from extreme poverty or have been exposed to a shock. Arrangements reached under such conditions are therefore more likely to be harmful and degrading.  While legal interventions to prohibit the practices have struggled to separate harmful practices from less harmful ones, social protection interventions aim to prevent some of the harmful ones. By providing safety nets to the rural poor and vulnerable, high-risk child mobility and domestic work are less likely to result as coping strategies.  Current traditional efforts to cope with risks and shocks have relatively little impact on child and girls mobility. Yet, households who are part of savings groups and IGA- projects have lower child mobility rates, and fewer girls away working as domestics. Households who feel certain that they can count on the support of migrants and relatives in case of crisis also have lower child mobility rates and fewer girls working as domestics. While network quality is hard to provide, the predictability of such security is an important feature that could be provided by a sustainable social safety net project.  In most of the households surveyed, men shoulder the main responsibility for school related expenses, while women more often are responsible for costs related to child health and nutrition. There are differences in the account of men and their wives on this, men systematically reporting to take more financial responsibility than what their wives agree on. Women takes more responsibility for child related expenses the further South 10 | Mobility inequality and social exclusion in the Benin one gets. Also, women claim to take more responsibility in presumably vulnerable households where the household heads report food insecurity (Figure 38, Figure 39 and Figure 40).  However, school participation rates are higher in the households where women shoulder more of the child related financial responsibilities. On the other hand, child mobility rates and the rates of girls leaving to work as domestics are also higher in households where women take more financial responsibility. (Figure 41, Figure 42 and Figure 43)  Safety net projects that do not specify something else tend to provide men with the program benefits. Since most men seem to assume school related expenses as their or shared responsibilities, this is, at least theoretically, unproblematic. Yet, targeting women could be considered, since school attendance rates are higher in households where women have more responsibility for child related expenditures.  Further targeting should be considered for asset poor households, while all households should be eligible in areas struck by systemic shocks. Female headship could be considered a possible targeting criteria: not because women make different relocation decisions, but because they tend to be more asset poor, and female headship would thus work as an easy-to-administer targeting criteria.  Conditioning on school participation to social transfers seems superfluous in rural Benin where there is an overwhelmingly positive attitude to schooling among most household heads. It could still be considered in the Northern part of the country, where attitudes are less positive. (Figure 44).  If schooling is conditioned on, the capacity of the schooling infrastructure should be addressed. Also, child mobility and relocation of girls for domestic work is higher in areas where people are discontent with the quality and accessibility of their nearest primary school. Thus, access and quality should also be addressed to provide a meaningful alternative to relocation, if this is to be conditioned on. Mobility, inequality and social exclusion | 11 1 Background and objective for the study Survey-based estimates from the ILO suggest that there are around 53 million domestic workers around the world today, almost 44 million of whom are women or girls. Sixteen million children are domestic servants, among them 11 million girls. In Africa, an estimated 5 million people are registered as domestic workers. The 4 million women and girls among them constitute as much as 14 percent of the paid female labor force. All these estimates are, however, believed to understate the real picture as this type of employment is often hidden, unregistered and informal. ILO therefore suggests doubling the numbers to get to a more realistic picture. This report looks at girls’ domestic work in Benin in the context of child mobility in general. The circulation of children, including for domestic service, is viewed as part of processes of social inclusion and exclusion, and a phenomena that contributes to the creation and maintaining of social, economic and gender inequalities in Benin. The report frames child mobility within a widespread socioeconomic system for service exchange, possible revenue, learning and networking opportunities – a system that provides an informal safety net to thousands of girls, young women and their households in difficult situations. The mere labor aspects of domestic work have been well depicted in both Benin and more generally (ILO, 2011), and this report therefore focuses on the social functions of the market for domestic work. Inequality and social exclusion: a downward spiral? The two concepts inequality and social exclusion are Central to his report. Inequality is an observable; a situation that can be measured; sizes to be compared. In its’ most clear cut expression, a child domestic worker can be seen living side by side with the family school child. This picture from a common Beninese household sharply contrasts two different life situations, and future opportunities that are worlds apart. The observables are typically the schooling status of the children, their nutrition and health state, where and how much they sleep, and how much leisure time they enjoy. 12 | Mobility inequality and social exclusion Social exclusion is on the other hand a process, and, as such, harder to directly observe. Social integration is the opposite process, and an important determinant for an individual’s happiness and success. Social integration is obtained by being given access to and learning to play the complex game of reciprocal actions that emerge from common rules, norms and values in a given society. The denial of access – or social exclusion – of an individual often occurs when he or she (or a group they are affiliated with) is seen as unlikely to be able or willing to fully carry out the obligations imposed by norm (now or in the future). This can be due to a lack of means or lack of spirit and moral. In uninsured societies where relationships of mutuality help each- one-by-turn through hardships, a non- or low-contributing network member may be an unattractive element. Maggie Black distinguishes between the disadvantages and indignities of child domestic workers.1 Disadvantages may include observables such as eating leftovers, having no personal space or bed, working very long hours, and isolation. By indignities she refers to harm resulting from the very nature of servanthood: damage to one’s sense of self, and to one’s existence as free independent and equal human being. While the disadvantages represent inequality, the indignities contribute to processes of social exclusion. Social exclusion is a relative state. While successful processes of social integration may be self- enforcing, processes of social exclusion easily turn into downward spirals. Disadvantages lead to indignities, and indignities to more disadvantages. Moreover, to gain social inclusion, social interaction is required. Those less able to contribute to social networks are not only experiencing a gradual rupture of social relationships, but are, as a consequence, excluded from the arenas where new social relationships and social interaction are obtained (social, economic, and political arenas). A destitute girl may not only lack the social networks needed to get a job – she may also be perceived as unattractive due to her poor or deteriorating habitus. By each loop down the spiral, a heightened level of inequality can be measured. A shock may lead to a large leap down the spiral. The further down already, the more significant this leap is likely to be, as mitigation and coping means become gradually depleted and social networks dwindle. 1 Anti-Slavery International: How to work with domestic workers. Mobility, inequality and social exclusion | 13 Figure 1. A downward spiral of social exclusion. Figure 1 illustrates a downward spiral, and incorporates some of the elements that may take part in a process of social exclusion. Representing social exclusion this way should not encourage intellectual simplifications. It should neither be read as a claim that all such processes are unidirectional. They are not. However, it is likely that each new shock opens up to a host of new vulnerabilities. It is likely that social network quality is challenged and gradually strained. The further down the spiral you get, the more it is likely to take to change its ’ direction. A main purpose of the downward spiral is therefore to provide an illustration that helps to show why early (preferably preventive) intervention is likely to be more efficient. 14 | Mobility inequality and social exclusion The normality of domestic work and the market for its’ services In most cases domestic work represents situations of inequality: between rich and poor, men and women, the adults and the children. This report, however, does not see it as a given negative. Under the wrong conditions domestic work contribute to social exclusion of disadvantaged women and girls, both individually and as a group. However, many women and girls also aim - and succeed - to use it as a springboard to both social and economic inclusion. Domestic work is a natural part of life for children in Benin, especially in labor intensive rural households. It is the most common type of marketable skills that most children are raised to learn. It is also common among most female youth while teenage boys gradually move away from this traditional activity sphere of women and children and towards the type of labor that men more often perform. Daily household chores in most rural and many urban households in Benin remain labor demanding. In the absence of electricity and piped water, the acquisition of water and cocking fuel top the list of such chores. It is also time consuming to prepare foodstuff from mainly raw products, and to cook for and clean after an often large number of household members. Young children require constant attention at risky compounds with open fire and uncovered wells. A neat and tidy compound is a sign to the world that a household is disciplined and well-functioning, while personal cleanliness commonly is used to assess the state of other people. In total, covering basic needs and maintaining a respectable façade takes an effort. The housewife has the primary responsibility for domestic work. Managing and assuring necessary assistance is an important challenge, as she will often not be able to do it all herself. Her children are key assets in this respect, while relatives constitute a flexible labor force in periods when her children are too young or in other ways not strong enough, many enough or of the right gender to perform the tasks needed in her household. It is for instance common for a newlywed girl to bring a younger sister to her new household for company and assistance, or to host another often unmarried, and often female relative. Growing expectations to sending children to school deprive many women of their customary right to control the labor of children, at least for most of the day. A woman who is not able to get the needed help in the house must work very hard, and may also be excluded from participating in the labor market outside her household. Moreover, she will have to carry out tasks perceived as low status: the type of work normally assigned to the person with the lowest rank in the Mobility, inequality and social exclusion | 15 household, like a servant or a child. So, not only may she loose the opportunity to earn own money, but also social status and influence. When close relatives are not able or willing to assist, help can be sought in a commercial or often semi-commercial labor market. The market for domestic services is extensive, corresponding to the high demand. While most often found within the extended family or kinship network, domestic workers may also be referenced by acquaintances or professional mediators. Women traders and previous community members are common mediators of domestic servants between their home communities and the urban areas where they carry out their business. They are the contact points to the outside world in many communities, and the natural persons to hook up with for a young girl or mother in search of a job in an urban household. They evoke confidence in the families they know; they can facilitate transportation, and have networks in the urban areas. Direct recruitment outside established networks takes place; typically when a housewife visits a village or a market and spots a girl who looks clean, well-mannered and energetic. She may then approach the girl’s mother directly, assuming that most rural girls would appreciate an opportunity to come to a more modern urban household and experience urban life. In Benin there have been cases registered as trafficking, where recruiters have mediated the domestic labor of underage girls with purely exploitative motives. However, these cases probably constitute a relatively small share of the labor movements of women and girls for domestic services. The inter-household exchange of girls and young women is motivated by a complex and intertwined set of economic and social factors. Each exchange is the result of considerations made by decision-makers in both the sending and the receiving household. On the sending side, economic concerns may at the extreme relate to problems of feeding household members whose labor does not yield much of a return in the home or local labor market. In the more social end of the motivation specter a girl may wish to move with a sister she feels attached to, or to live in a nicer environment (an exchange involving domestic work does not necessarily involve more or harder work than what was performed at home). In between these extremes there are more strategic considerations that take into account learning opportunities and possibilities for social networking. On the receiving side, social and economic considerations are similarly intertwined. Need for help with domestic work co-exist with the desire for company, and in-between a 16 | Mobility inequality and social exclusion recognition of the fact that helping to host and employ a relative is a service made to a family or community based mutuality system where social credit is collected. To assess the likelihood of exploitation and miserable work condition for a relocated girl, one can use a simple chart based on the extent to which economic or social motives were dominant to decision makers in the original and the new household of the young woman or girl Figure 2 suggests that exploitation is more probable if economic motives are dominant in both places. Exploitation would be less likely in the cases where girls strategically aim to use a domestic job as a springboard for future goals (an apprenticeship, a better spouse or a trousseau, money for school fees, or urban experience) and the receiving household is part of a quality social network, and thus predominantly motivated by social concerns. The term domestic helper can be applied as a wide term intended to cover the full range of exchanges where house work of some sort is likely to be expected as part of the relocation arrangement. 100% “Lucky girl� High likelihood Motive, household of origin Economic/ of exploitation coping 50/50 Social/ Strategic Low likelihood Possible 100% of exploitation deception 100% Social motives 50/50 Economic motives 100% Motive, new household Figure 2. Predicted quality of inter-household exchange of child or domestic helper, based on the dominance of social/strategic versus economic/coping motives in the household of origin and the new household. Mobility, inequality and social exclusion | 17 Risks, shocks and the need for social protection Poverty has been linked to child mobility and to the supply of child domestic services in most previous studies of the phenomenon. It is evident that poverty increases the likelihood for child mobility, but the relationship is not clear cut. Children and young women generally seem to leave poor areas, but not necessarily the poorest households in those areas. There are many explanations for this. The poorest of the poor may face higher relocations costs if they lack the social networks needed to identify a job, find transportation, perhaps a temporary place to stay, and people who can follow up the child in its’ new location. Relocation will appear less appealing as labor arrangements become less secure and the likelihood of exploitation becomes higher. The mental impact and psychological pressure of extreme poverty often leads to personal disorganization. This is a state that affects the ability to plan and implement arrangements that could help improve the destitute situation. Also, the way extreme poverty often affects children can make them become perceived as unattractive in other households, even as domestic servants. Discipline, energy, skills and manners are all qualities looked for in a servant, and many of the poorest children will not be able to bring such assets to the market. It is not uncommon to hear that very poor children are “dirty� and potential thieves, and thus not somebody you would want to have in your house. The word “poverty� is often mentioned when respondents in qualitative studies discuss why they encourage or allow their children to leave home at a young age. Poverty is however an amorphous explanation and not all poor act the same way. A child may live for a long time in a poor household before relocation is triggered, often by some other event. It could for instance be triggered by demand-side factors such as an acceptable job opportunity. Or it could be triggered by a shock to the household or the community. Such shocks are likely to exacerbate poverty however the exact impacts are difficult to assess by qualitative interviews only. A shock is often explained as an event that leads to a serious income reduction, having to sell off assets, and/or having to significantly reduce consumption. Shocks can be individual to a person or a household, or systemic when it strikes an entire community or a larger entity. A shock needs not be detrimental. A household with access to some sort of insurance or resources to cope with detrimental effects of a shock could recover relatively fast. Insurance could be informal, as in a health mutual or a well-functioning tontine, or a strong family or community support network. 18 | Mobility inequality and social exclusion Coping is always dramatic, but households with savings in cash or land/animals would at least have a buffer to tap into before applying more drastic coping means like the sale of elementary assets, withdrawal of children from school or reduced food intake. The effectiveness of local networks of mutual support is of course weakened when the community of the household households is jointly struck by a systemic shock. The most detrimental response to shocks is often labeled as “dis-saving of human capital�. The most common example is when children are taken out of school and given less nutritious foods, thereby hampering their health and education, possibly with lasting negative consequences. Another example is for household adults to work more or migrate for work, jeopardizing the care and protection needed by the household’s children. Sometimes adults accept (more) dangerous jobs, because other options become depleted. Child labor is also considered a coping strategy that leads to dis-saving of human capital. The possibility that the floods of 2010 and other common shocks in Benin might have led to the dos-saving of the human capital of children from affected communities is one of the rationales for this report. Especially, there is an assumption that more children leave their homes for other reasons than further schooling, and more girls decide to leave for domestic work as a consequence of such shocks. An important objective for social protection policy is to strengthen the mitigation and coping tools of vulnerable populations. A well-functioning safety net is described not as a net but rather as a springboard that can help households and communities affected by shocks bounce back. Preferably, such safety nets should be made available before households are forced to sell off productive assets, and most importantly, before the human capital of children or other household members get permanently damaged. Child mobility in general can be risky, and the mobility of girls towards domestic work may jeopardize human capital if the labor conditions are bad. If accepting risky relocation arrangements are indeed a response to shocks, then social protection policies should consider it a priority to provide alternative coping options to vulnerable families at risk for future crises. Both the targeting of and the conditionalitities attached to safety net interventions could contribute to reduce the likelihood that children have to serve as what has been labeled the “risk management tools of the poor�. Mobility, inequality and social exclusion | 19 2 Methodology The research project integrated a qualitative and a quantitative component. In practice, data was collected by equipping 7 multidisciplinary field teams, and making qualitative and quantitative staff collect their data side by side in the 150 rural clusters sampled for the project. The report therefore presents not only estimates of total mobility for rural children and youth in Benin, including for domestic services, but also views the data in perspective a more in-depth analytical account of how this phenomenon affects girls and women humanly and socially under differing circumstances. The survey Survey approach The survey used a fertility based method which can be distinguished from more conventional approaches to household surveying by its focus on children as the ultimate data case. The conventional way to map children is to register them as members of the households where they reside. The fertility based method on the other hand assigns a unique adult representative to each child, usually the mother, and ask her to report on their behalf. The advantage of the latter approach is that it provides more and better data on why some children relocate away from their parental households of origin. In practice, the fertility method asks each mother to list all her pregnancies in a fertility roster, regardless of outcome. The roster can then be completed with information on the children’s status with regards to health, education, labor activities, individual features, but most importantly; their location and the circumstances surrounding a possible relocation. While mothers normally serve as the child’s representative, maternally orphaned children should not be excluded by the survey design. Therefore, responsibility as child representative is transferred to the father if the mother has died. Double-orphaned children are represented by their paternal grandmother. 20 | Mobility inequality and social exclusion In this survey, a parent eligible for completing the child roster is therefore either a mother with at least one child below the age of 30, a father with at least one maternally orphaned child below 30 years-of-age, or a paternal grandmother of such a child if the child is double orphaned. This is also different from the DHS approach where women below a certain age are the main research unit. The sample Representatives from 3 000 households in rural communes in Benin were interviewed for the quantitative part of the data collection. To obtain a relevant sample, the country was divided into three strata following traditional ethnic and production divides; the North (comprising the departments of Alibouri, Atacora, Donga and Borgou), the Center (generally Collines, parts of Couffo and Zou), and the South (generally Atlantique, Mono, Oueme, Plateau and the second part of Couffo). Within each strata, communes touched by the floods of 2010 and communes with known migration practices were over sampled. That is, within the defined sampling frames of each strata, 150 clusters were selected by assigning 35 clusters (700 households) to the North, 65 (1 300 households) to the Center and 50 (1 000 households) to the South. The clusters were assigned a selection probability proportional to population size in the census of 2002. The selected clusters were then enumerated, and household weights adjusted for population changes within clusters and population projections for Benin 2002-2012 (see sub-section on weights). Survey modules The survey can be divided into 5 modules. First, a community questionnaire was administered to a group of resource persons in the village where the cluster was located. The community questionnaire focused on mapping local infrastructure, schooling options, access to social protection, experiences with recent covariate shocks and recent program interventions. It also aimed to identify patterns of attitudes towards future opportunities, gender roles and children. Second, a household questionnaire was presented to the heads of the selected households. Apart from the household roster and some standard socio-demographic questions, household assets and building quality was assessed. Other topics covered were social protection access, distribution of Mobility, inequality and social exclusion | 21 financial responsibilities between husband and wife(es), idiosyncratic shock experience, fertility and mobility. Third, as an attachment to the household questionnaire, a network module was created based on the relationships between the 20 households selected in a cluster. In brief, each household head rated his relationship to each one of the other selected household heads on a scale from 1-6 ranging from not knowing the person to being very closely related. The network module makes it possible to assess the level of social connectedness in a cluster, and also to study network effects of certain attitudes and behaviors. Fourth and fifth, a parent module with a fertility roster was completed. Each child representative in the household was eligible for this module. The first part maps comprehensive information about mothers in households where the man was counted as household head. Again, information about social protection, fertility and empowerment were Central. The last part, the child roster, represents a complete mapping of a mother’s birth history, including interrupted pregnancies and still borne children. The child roster importantly focuses on details surrounding the relocation of children. Non-response The survey had a relatively high non-response rate (8 percent) for a survey in rural Africa. The response rate was high in the 4 Northernmost departments (Alibouri, Atacora, Borgou and Donga), low in the Central Collines, high in Plateau, medium in Zou, low in the western Couffo, Mono and Atlantique, and also in Southern Oueme. A total of 240 households from the original sample were not found, and 64 households did not complete the child roster module of the interviews. The main reason for this outcome is the combination of enumeration and survey timing. Due to time constraints, enumeration of cluster households was conducted during last part of December and first part of January 2011/2012. Traditionally, this is the time of year when people are present in their home communities to take part in Christmas and New Year celebrations. The survey, however, was conducted mainly in February. February is the driest month of the year in 22 | Mobility inequality and social exclusion Benin, and typically the time for seasonal migration. Some households migrate “en-grouppe�, while in other households breadwinners leave children in the care of grandparents and other relatives while looking for work elsewhere. The geographical pattern of the non-response rate corresponds well with culturally differences in migratory practices. The ethnicities of the North have little tradition for migration or sending children to work for others, as domestics or otherwise. As expected, non-response rates were neither an issue in those areas. Traditions for seasonal migration are however strong in Collines, Zou, Mono, Couffo, Atlantique and the Valley of Oueme, where we consequently find high non-response rates. While the non-response rate is unfortunate with respect to loss of households, it is also a good time to study the mobility patterns of children independent of household that are otherwise relatively intact. The data Fafo does not practice replacement of non-identified household in their surveys. The rationale behind this choice is that replacement tends to strengthen a selection bias that at the same time easily becomes concealed and forgotten by replacements. Consequently, the final data includes 2 760 out of the originally 3 000 sampled households including complete household rosters. These rosters contain 15 925 household members, suggesting an average of 5,8 members per household. This is slightly higher than DHS 2006 estimates for rural areas, and the reason may be that remaining households temporarily host children of seasonal migrants. The household population is very young. Sixty percent are below 20 years-of-age, 50 percent under 14 years-of- age. There are 8 817 children under 18 years-of-age currently living in the sampled households. A total of 6 914 are living with both their parents, while 750 are living away from both. One- hundred-and-eighty-eight of these children have lost their mother, 383 their father, while 34 are double orphans. Also, 441 have both parents living outside the community. Among the household interviews successfully completed, 2 213 households had at least one parent who was eligible for the fertility roster module of the survey. 2 684 child representatives completed the fertility roster, representing 2 179 households. In 2 078 households there were representatives of children 0-17 years-of-age. Among the around 20 percent of households with Mobility, inequality and social exclusion | 23 no child representatives were single-person households, households with childless couples (young, very old, or simply childless), households of father and children (mother alive, but living elsewhere), households of grandparents (single or in couples) with grand-children, and middle- aged or elderly people who lived with nieces or nephews (and often grandchildren too). One-in-five households had more than one eligible parent, and around one-in-twnty had more than that. Fifty-four fathers responded to the fertility module, and 2 703 women. The two youngest mothers were 16 years old, while 11 women past 70 were registered as child representatives. Between the ages of 27 and 47 years around 90 percent of women are registered as eligible for parental interviews. Figure 3. Share of household women 16 – 70 years-of-age found eligible for parent interviews (in percent of female household members). In the child rosters there are 12 841 entries from 2 179 households, each entry representing a terminated pregnancy (363), a still birth (346) or a live birth (12 132). Only 10 973 children are still alive. Among the live births, 1 160 have later died, 312 before the age of 1 year, 880 before the age of 5 years (not counting the still borne). Among the 10 972 children still alive, 8 850 are below 20 years-of-age, 8 272 below 18 years-of-age, and 7 176 below 15 years-of-age. 24 | Mobility inequality and social exclusion Weights The 150 clusters sampled were drawn from three strata: The North, the Centre and the South. Within each strata, clusters were selected with a sampling probability relative to their size; 35 clusters in the North, 65 in the Center and 50 in the South. The clusters were then enumerated, as the 2002 census is now relatively old. Household weights were determined in three steps. First, the probability of selecting a cluster was estimated by dividing the 2002 census household count in the cluster by the number of households in the strata, and multiplying this by the number of clusters drawn in the strata. Second, the probability of selecting a household equals the number drawn in each cluster (fixed at 20) divided on the number of cluster households as enumerated after cluster selection in December 2011/January 2012. Multiplying the two first two steps produces the household selection probability, and dividing one by the selection probability produces the household weight. The weight is partially adjusted for population growth due to the inclusion of re- enumerated household figures. A complete adjustment to projected population growth figures 2002 – 2012 was then made. Finally there was a choice of adjusting the weights to the non-response rates. This was not done for two reasons. First, the low response rates in the Central and Southern regions are predominantly expressions of a difference in households present between the end on December 2011 and February 2012. This difference is generally explained by seasonal migration which is at its peak in February. Not adjusting the weights assumes that the weights kept in fact represent the reduced number of households that remain in rural areas during migration season. Consequently no assumptions are made about households represented by those not present. Extrapolating the data to cover also these households would most likely have concealed a non- negligible selection bias. Mobility, inequality and social exclusion | 25 The qualitative work Statistical data is helpful in measuring several types of inequalities in a society. It can also serve to further develop hypotheses related to processes of social exclusion. However, processes of social exclusion or inclusion are not static. To gain a better understanding of the dynamic relations of these underlying processes, quantitative data was complemented by life-story interviews and other in-depth qualitative work. The survey was conducted in 150 rural clusters across Benin. One social scientist was added to each of the seven survey teams, and put in charge of collecting qualitative data alongside and in constant dialogue with the survey interviewers. The qualitative data collected were mainly intended to support the survey, and should not be considered a freestanding research exercise. The guided, semi-structured interviews of adults had two different thematic angels, and focus group sessions were organized with children and youth. The work was practically organized so that one or two in-depth interview or two focus groups were conducted in each second cluster. Since one cluster was visited per day, every second day was set aside for transcribing of interviews conducted the previous day. The two thematic angels, or core research questions were: 1. What is the significance of the social institution of domestic service in the lifespan of rural girls and women? 2. What is the significance of child and youth mobility to the children and youth themselves and to their families? Throughout the qualitative work, the words “trafficking� and “domestic servitude� were explicitly avoided. While both are well-established topics of study in Benin, this report see these as predominantly judicial issues rather than socioeconomic ones. Instead the terms “mobility� and “domestic work� were used. At the community level the person to interview was to be drawn from the pool of 20 households sampled for the survey. Eligible for interviews were households that contain a girl/woman with the experience of mobility for domestic service (for research question 1), or a representative for a household that had practical experience with child mobility (for research question 2). It is 26 | Mobility inequality and social exclusion unusual to randomly sample individuals for qualitative research. Compared to other research based on in-depth interviews this method is likely to recruit more interviewees with relatively plain stories to tell or who are not very good at expressing themselves. This requires time, good methodology and patience from the interviewers. As with all studies of mobility that takes its respondents from sending areas, be they qualitative or quantitative, attrition is an issue. Interviews are only carried out with those remaining in or returning to rural areas – not those who left and never came back. Two challenges were particularly prepared for. First, when working with retrospective data, researchers need to interpret the findings carefully as respondents often fail to distinguish between justifying narratives and actual experiences. For example: a girl quotes poverty as motive for leaving. But the family had been poor for 14 year. What exactly happened that made her leave? Why did other, equally poor girls not leave? Second, some of the narratives were strongly colored by the semantics of a large number of anti-trafficking campaigns carried out in Benin over the past decade. Theme 1: The social institution of domestic work The objective of the first qualitative theme was to better understand what domestic work as social institution mean to girls and women in rural Benin. By social institution we mean the fact that a market for domestic services exists and is available to girls and women in different life situations. We wanted to understand how this institution is perceived, judged, approached and used in order to support the short and/or long-term goals of the girls and women involved. In what life situations do girls and women approach this institution? What are the short and long- term benefits and disadvantages they foresee and experience? The key questions asked in the semi-structured interviews were: 1. What were the specific circumstances in your house of origin when you decided to leave for work? (Aim to identify both permanent features and triggering events). 2. How did you think you could benefit from leaving and what did you think would be the challenges? Mobility, inequality and social exclusion | 27 3. How did you find the job(s) you got (trajectory if several)? (Did you know where to go, way of mediation, relationships, things that happened on the way) 4. Can you describe a typical day while you were working in this household? (From you got up in the morning till you went to bed at night, including details like eating and sleeping conditions, social interaction, inside versus outside the household/compound) 5. How did you decide to come back to the village/come to this village? (Pre-arranged, something triggered it, push/pull? Was it an option to stay?) 6. What do you think that you and your family did gain, short and long-term? (financially, social status, experience, learning, concrete skills, social networks, spouse?) 7. What do you think were the costs to you and your family, short and long term? (including hardship, loss of other opportunities, health issues, social costs ) 8. Do you know anyone who stayed permanently there after leaving? (tell us!) 9. Was it an option for you to stay? What do you think could have happened if you did? (Options, opportunities, losses…) 10. Do any proverbs you know help to describe some of your motivations or experiences? Theme 2: Mobility and household security strategies The objective of the second qualitative theme was to understand more about why children and youth leave home, and under what circumstances their parents encourage them to or discourage them from leaving. What opportunities do they see for themselves, and what opportunities and benefits are there for their families? How do families use their relatives and other social networks to facilitate for children’s relocation? How can child mobility help strengthen the social networks of the children themselves and how can it strengthen the social networks of their families? Can child mobility jeopardize existing social relationships? Is diversification of children’s human capital (diversification in geographic location and profession among siblings and other core family members) a goal, and how can it be beneficial? 28 | Mobility inequality and social exclusion The key questions for this theme were: 1. Can you tell us about the circumstances around the departure of your child(ren)? (general situation, triggering events – that is: narrative and experience) 2. Can you describe your relationship to the people the children stayed with/worked for? (especially, have there been exchange of goods or services, or do you have future expectations of such?) 3. What good do you think has come out of these relocations for your child(ren)? (educational, social status, better prospects, financial improvement…) 4. Did anything good come out of the relocations for you and the family here? (economic support, favors or services from hosting family/employer, social relations, future support) 5. What were the risks you saw related to the relocation of your child(ren)? (especially – was there a risk that the child could endanger a social/family relationship) 6. Have you heard the expression “do not put all your eggs in one basket�? Some people think that it is good that their children get different jobs and live in different places. If some sectors have problems those working or living in other sectors/places can help. What is your personal experience with this, and do you know of any such experiences of others? 7. When you were a child, did you at any time live away from your parents? If so, can you tell me your story? 8. Do any proverbs you know help to describe some of your motivations or experiences? The qualitative data is sparsely referred in this report, and will be explored further in future work. All transcripts are however posted on the project web-site, and are thus available to those interested. Mobility, inequality and social exclusion | 29 3 Numbers and shares Estimates of extrapolated mobility numbers are based on the sampling procedures of the survey. In principle, the sampling probability of a child is in this survey equal to the sampling probability of a household. This is because households were randomly sampled within the clusters (no proportional sampling was done at this level), and all children in each household were mapped (instead of sampling one child per drawn household, as is common in some such surveys). At the household level, no weight adjustment was made with regards to the households that were not found. This was done because the missing households are assumed to be seasonal migrants, and the numbers emerging from unadjusted weights therefore represent the population remaining. However, there were 64 households that were identified and interviewed, but did not complete the parent/child modules of the survey. At the child level, therefore, a slight weight adjustment is made to account for the children of these households. Child weights are therefore slightly adjusted upwards (by 4,7 percentage points). The survey covered three main strata: the North, the Center and the South. Within each strata, around half the rural communes were drawn for the sampling frame. Since the communes inside the sampling frame and those outside do not necessarily have the same probability of circulating children, it would be too simple to just extrapolate results from the sampling areas to the areas outside. Based on the child weights calculated, the number or children 0-17 years-of-age represented by eligible respondents are estimated to be 871 000 in the North, 1 295 000 in the Center and 1 039 000 in the South (see Table 1). Among them, 432 000 in the North, 864 000 in the Center and 471 000 in South organize directly under the sampling frame of the survey. Since there is reason to believe that children outside the sampling frame has a lower likelihood of circulating than those within, rates from the sampling frame were not directly applied to the outside areas. Instead, a child trafficking survey from Benin, 2006, was revisited to get a better picture of the traditional distribution of mobility at the commune level.2 Careful examination 2 CEFORP, 2007, Etude Nationale sur la Traite des Enfants. 30 | Mobility inequality and social exclusion commune by commune shows that the samples drawn in the North and the Center of the country has a lower, but relatively similar child mobility rates as the areas outside the sampling frames, while the Southern sampling frame is likely to have a considerably higher circulation rate than the areas outside. This is corrected for in the extrapolations made in the number estimates presented in this chapter. The last column of Table 1 shows the relative weight given to the areas outside the sampling frames. That is, where 100 children within a given population are found to be relocating inside the sampling frame, this counts for 75 children in a similarly sized population outside the sampling frame in the North, 93 children in the Center and only 44 children in the Southern strata. Table 1. Total number of rural children 0-17 years old represented in the rural communes of the sampling frames and the outside rural communes. Share of rural Relative Number of Estimated number of Estimated total population in mobility weights children rural children of rural the sample applied outside represented in the represented outside children with frame the sampling sampling frames the sampling frames representation frames North 432 280 50 % 438 249 870 529 0,75 % Center 863 850 67 % 431 560 1 295 410 0,93 % South 471 172 35 % 567 333 1 038 505 0,44 % Total 1 767 302 54 % 1 437 142 3 204 444 0,58 % The surveyed children The data set contains information about 8 272 rural children between 0 and 17 years-of-age. Sixty-seven percent of the children had a birth certificate – an important asset for children in general, but especially so for mobile children. Children in the North have the lowest likelihood of having a birth certificate (57 percent), the rates are slightly higher in the Center (62 percent), and relatively high in the South (80 percent). There are some small gender differences in favor of boys in all the three strata (see Figure 4). Mobility, inequality and social exclusion | 31 90 81 79 80 70 67 64 63 61 58 60 55 50 40 30 20 10 0 Boy Girl Boy Girl Boy Girl Boy Girl North Centre South All Benin Figure 4. Share of children 0-17 years-old holding a birth certificate (in percent of all children). A striking trend emerges when asking the child representatives about the health state of the children (Figure 5). Across gender and age groups, Northern parents overwhelmingly characterize their children as healthy. In the Central parts of the country one-in-five children are characterized as sickly or chronically ill, while in the South, only two-in-three children are described as in good health. The parents’ assessments of their children’s health are sociologically and culturally interesting since they most likely reflect more than the subjectively measurable health state of each child. It is true that breast feeding is more common in the North, and that Northern children in general benefit from more dairy products in their diet. However, the parental ratings may be a sign of their overall feeling of worry, or, as in the case of the North: a mental and spiritual strength that may reduce such worries. There may be religious effects: ninety-two percent of Muslim children are by their parents considered to be in good health, while the number for the various Christian directions and people primarily adhering to Vodoun only rate three-in-four children as healthy. Standards for good health may also differ. Income differences are greater in the South, and people may compare themselves to the wealthiest in their proximity. In fact, parental rating of child health seems uncorrelated with most of the commonly suspected factors like housing 32 | Mobility inequality and social exclusion quality, household asset and number of meals. On the other hand, the rating of children’s health seems to correlate positively with the parents’ level of trust and a feeling of being socially secure. 100 93 89 90 80 77 78 78 80 70 64 64 60 50 40 30 20 10 0 Boy Girl Boy Girl Boy Girl Boy Girl North Centre South All Benin Figure 5. Share of children 0-17 years-old reported to be in good health (in percent of all children). School attendance and lack of schooling also differs by region. Figure 6 and Figure 7 show the shares of children, 6-17 years-of-age, who are currently attending school and those that have never been to school (drop-outs constitute the residual). School attendance rates are poorest in the North, and almost half those of the South. Boys have higher school attendance rates than girls in all strata and age groups. Gender differences are notably smaller in the North than in the other regions. However, the younger cohorts of girls seem to somewhat catch up with the boys also in the Central and Southern region. More than half of boys and girls in the Northern region have never been to school. While more children in the South than the Center attend school, the rates of children who have never attended school are relatively similar in the Center and the South. This indicates that drop-out rates are high in the Central part of the country, and accounting for the difference between the two figures. Mobility, inequality and social exclusion | 33 90 84 83 80 74 76 71 73 67 69 70 65 60 60 50 48 50 46 45 41 6-10 38 38 40 31 11-14 30 15-17 20 10 0 Boy Girl Boy Girl Boy Girl North Centre South Figure 6. Share of children, 6-17 years-of-age, currently attending school on region, gender and age bracket (in percent of all children). 70 60 59 60 56 52 50 48 50 40 6-10 28 30 25 23 11-14 21 19 21 20 15-17 12 10 11 9 9 11 10 0 Boy Girl Boy Girl Boy Girl North Centre South Figure 7. Share of children, 6-17 years-of-age, who have never been to school on region, gender and age bracket (in percent of all children). 34 | Mobility inequality and social exclusion Households descriptives All the households of the survey are localized in predominantly rural communes. The areas covered by the sampling frames constitute around half the rural communes, and the remaining population during the dry season is estimated to 564 327 households. Poverty is widespread. Social and economic shocks are common and awareness of numerous and serious risks cause a great deal of worry. Household poverty in the sampled households Table 2 and Table 3 give an overview of the housing quality and the assets counted in the households. In the analyses in the following chapters, these two factors will be used to assess the level of wealth in the households surveyed. This section moreover looks at household exposure to shocks. Drought is a constant risk for farmers in Benin, and the 2010 floods became a genuine disaster to many rural households. Table 2 shows that while three-in-four households had a solid roof (mostly iron sheet), only 41 percent had cemented floors, and less than one-in-five had walls made of brick or other solid material. House constructions in the North were overall more vulnerable, the houses in the Center more often were protected by iron-sheet roofs and the houses in to South most often had cemented floors. Table 2. The housing quality of the sampled households, by building part material and region (in percent of all households). North Center South All Iron sheet roof 69 90 65 77 Cement floors 37 38 49 41 Brick walls 14 21 20 19 Table 3 shows the asset count of the surveyed households in the three strata. There is not a big difference between the rural communes of the three strata with regards to the mere number of Mobility, inequality and social exclusion | 35 assets owned: on average, 2,4 assets per household. More households in the North had their own radio, while TV and DVD player were more common in the South. Having a phone is also more common in the South. More households in the South also had a generator. Owning a bicycle or a moped was least common in the South. An observation indicating the constraints is that in less than half the households with a bicycle, women had access to using it. In the North, women had access to only one-in-tree bicycles, while in the Center this share was of 54 percent. Table 3. Assets counted in the surveyed households, on strata (in percent of all households). North Center South Total Radio 73,7 60,3 59,2 63,0 Decoder and parabola 4,4 3,4 1,9 3,1 TV 14,1 15,9 22,9 17,8 CD player 11,4 10,4 11,6 11,0 DVD player 9,3 11,0 16,5 12,4 Fridge 2,4 1,3 1,1 1,5 Phone 45,0 57,6 63,3 56,6 Bicycle 47,6 39,9 23,0 36,1 Mobylette 4,3 3,0 1,7 2,8 Moped 41,0 41,7 32,6 38,6 Car 1,7 1,4 1,3 1,5 Electricity 9,8 12,0 11,9 11,5 Solar panel 0,1 1,4 0,3 0,7 Generator 6,5 10,9 14,0 10,9 Chart 1,2 2,0 1,1 1,5 Cistern 2,5 11,6 5,0 7,4 With regards to nutrition, the average number of daily meals reported the day before the survey was 3,6. Households in the Center reported to have more (4,25), households in the South slightly fewer (3,25), while the households in the North only reported 2,75 meals. With regards to nutritional value, 87 percent of household heads in the South reported to have eaten meat, fish or fromage peuhl (a protein rich cheese) the previous day. In the Center the share was 77 percent, while only 62 percent of respondents in the North reported to have had this type of protein rich meals. 36 | Mobility inequality and social exclusion Risks and shocks Rural households in Benin live with a considerable risk awareness. Responses to standardized personality questions show that more than three-in-four household heads identified strongly with a personality who finds it important to live in a safe environment and avoid everything that could be dangerous. In contrast, only one-in-five identified with a risk-seeking personality, and even fewer with a fun-seeking one. Trust levels are generally low. Asked if most people could be trusted or that one should be very careful, two-in-three household heads said one should be very careful with others. Asked if most people would try to take advantage of you, or that people were generally honest, 60 percent felt that people would generally try to take advantage. While trust levels were higher in the North, three-in-fort household heads in the South answered negatively to the two questions. 100 90 80 70 42 40 60 49 43 47 43 50 40 46 Sometimes 50 30 A lot 20 45 44 32 31 32 32 10 13 20 0 Past Future Past Future Past Future Past Future North Centre South All Benin Figure 8. Share of household heads that worried about being able to provide their family with food during the past 12 months and in the future 12 months, on level of worry and region (in percent of all households). Simply covering the basic needs of the household members is a constant source of worry to many household heads. Figure 8 shows the share of household heads that in the past 12 months had worried sometimes or worried a lot about how to feed their family. It also shows the level of worry with regards to being able to provide for the family in the next 12 months. The results mirror findings on health concerns: people worry less in the North, and most in the South, where almost half the household heads worry a lot about feeding their families in the year to come. A cross tabulation of past and future concerns shows that 15 percent of those who did not worry Mobility, inequality and social exclusion | 37 about food provision during the past 12 months are worried or very worried about the future, while 13 percent of those who worried sometimes in the past are very worried about the future. 38 Became indebted 32 17 83 Lost savings 60 56 22 Scool drop-out 9 4 61 Loss of buildings/infrastructure 50 57 61 Illness/undernutrition 40 32 South 60 Center Other revenue loss 31 35 North 64 Loss of animals/fish 37 40 76 Loss of crop 85 86 21 The household received refugees 12 7 40 Household members had to fleed 18 14 0 10 20 30 40 50 60 70 80 90 100 Figure 9. Types of consequences of being affected by the 2010 floods, by region (in percent of those reporting to have been affected). Forty-six percent of households were somehow affected by the 2010 floods. This distribution was relatively even between the three regions studied, although slightly lower in the North. However, Figure 9 shows that those affected in the South were more severely harmed. Forty percent of the affected households in the Southern strata had members who had to flee, and one- in-five had received refugees in their homes. Crop loss was by far the most common consequence, suffered by three-in-four of those affected in the South and around 85 percent in 38 | Mobility inequality and social exclusion the Center and the North. Two-in-three affected Southern households also reported the loss of livestock or fish from fish farming. The shock also depleted household savings. Eighty-three percent of affected households in the South, 60 percent in the Center and 56 percent in the North reported to have lost their savings, and many had become indebted – in the South as many as 38 percent. With regards to child welfare, 22 percent of affected households in the South reported school drop-outs, and as many as 61 percent illness and under nutrition as a consequence of the floods. Illness and under nutrition was also frequently reported in the Center and North. In addition, more than half the affected respondents reported loss of buildings or other basic infrastructure. After inquiring into the consequences of the 2010 flood, the respondents were given the definition of a shock as “an event that had led to a serious reduction of assets, had caused a substantial reduction of revenue or led to a considerable cut in household consumption�. They were then asked to list the different shocks experienced by the household in the past decade. Forty-two percent say they experienced the 2010 flood or other floods in the decade as veritable shocks according to this definition, 48 percent of the household heads in the South. The most commonly quoted source of shock was however the illness of a family member. Although there is no particular reason to believe that morbidity is less severe in the North, considerably more people in the Center and South report shocks related to both morbidity and mortality. Only 23 percent of household heads in the North say they suffered a shock due to the death of an adult breadwinner, compared to 50 percent in the Center and 48 percent in the South respectively. While cultural and religious values no doubt influence people’s perspectives, it is also possible that the larger households in the North are slightly less vulnerable to such idiosyncratic shocks. Among the covariate shocks most often referred were sharp price increases experienced in the past few years. Again people in the South experienced the price increases as most dramatic. Forty-four percent had been shocked by drought, 37 percent by other types of crop loss, and this time the reported shocks were again relatively evenly distributed across the three regions. Thirty- five percent had been shocked by animal diseases and 21 percent by crickets or locust. Mobility, inequality and social exclusion | 39 55 Illness of household member 62 32 48 Death of an adult breadwinner 50 23 64 Sharp increase in prices 43 47 46 Animal disease 28 33 38 South Other crop loss 36 39 Center 9 North Fire 10 6 48 Floods 38 40 30 Crickets/locust 20 12 45 Druoght 41 48 0 10 20 30 40 50 60 70 Figure 10. Types of shocks suffered by sampled rural households, by region (in percent of all households). With reference to the latest shock suffered, household heads were asked if the household had now recovered and returned to the state it had been in previous to the shock. Only 25 percent say they have. Recovery rates are highest in the North where one-in-three feel they are back to the economic situation they had been in previous to the shock. Time passed since the shock seems to have no systematic effect on the recovery rate. However, rates are lower in the household that reported to have been affected by the floods of 2010. 40 | Mobility inequality and social exclusion Child mobility: numbers and attitudes To explain how number estimates are achieved, a brief recapitulation of the survey methodology is in place. The first section of this chapter explained how the data were collected from rural communes covering an estimated 54 percent of rural households. Communes that were affected by the 2010 floods were oversampled. Mobility probabilities from within the sample frame of this survey are extrapolated to rural areas outside the sample frame with reference to the relative mobility rates in the same sampled and non-sampled communes in the 2006 survey on child trafficking in Benin (revert to Table 1 for details). Table 4 shows the shares and estimated number of children who currently live away from their primary representative in rural Benin, and also the shares and number of children who were away in the past and have now returned. The first two columns show the prevalence of mobility by gender and geographical strata; the North, Center and South of Benin. The next two columns show the weighed estimates of mobile children within the sampling frame. Because the prevalence traditionally has been lower outside the sampled communes, the prevalence rate has been adjusted for estimates made in that area (see previous section and Table 1 for details). The column labeled RW shows the relative weight that was used to adjust the prevalence estimates in the respective zones outside the sampling frames. Table 4. Extrapolated mobility numbers for children 0-17 years on gender and strata with extrapolations to outside the regional sample frames (SF). Prevalence Inside SF RW* Outside SF Total Past Current Past Current Past Current Past Current mobility mobility mobility mobility mobility mobility North ,0090 ,0107 2 125 2 542 0,7456 1 606 1 922 3 731 4 464 Centre ,0198 ,0645 8 442 27 543 0,9265 3 907 12 748 12 349 40 291 Male South ,0142 ,0730 3 630 18 637 0,4422 1 933 9 923 5 563 28 560 Total 14 197 48 722 7 446 24 592 21 643 73 314 North ,0107 ,0247 2 097 4 826 0,7456 1 585 3 648 3 682 8 474 Centre ,0318 ,0898 13 875 39 227 0,9265 6 422 18 156 20 297 57 383 Female South ,0206 ,1223 4 453 26 393 0,4422 2 371 14 052 6 824 40 445 Total 20 423 70 445 10 378 35 856 30 803 106 303 * RW: the relative weight of the prevalence assigned to the child population outside the sampling frames. Mobility, inequality and social exclusion | 41 The results indicate that close to 180 000 rural children between 0 and 17 years-of-age currently live away from their primary parental representative. In addition, almost 50 000 have previously lived away from these households for more than 2 months, to then return. Among the children who are currently away, an approximate of 106 000 are girls, and 73 000 boys. Child mobility is less common in the Northern strata, where the prevalence rate is 1,1 percent for boys and 2,5 percent for girls. The highest rates are found in the Southern strata: 7,3 percent for boys and as much as 12,2 percent for girls. In the Central strata the rates are also relatively high: 6,5 percent for boys and 9 percent for girls. 20 18 18 16 15 14 14 12 0-5 10 9 8 6-10 8 7 5 11-14 6 4 4 15-17 4 2 2 2 1 1 1 1 0 0 Currently Previously Currently Previously Boys Girls Figure 11. Mobility prevalence, current and in the past, by age and gender (in percent of all children). Figure 11 shows mobility by age group. While only around 1 percent of the youngest boys and girls live away from their primary representative, this naturally increases by age. Among the children between 15 and 17 years-of-age, as many as 15 percent of the boys and 18 percent of the girls live away from home. While the mobile children in youngest age group represent less than 10 000 children, the three other groups correspond to around 55 000 relocated children each. (Note that the division into age brackets is made based on functional concerns, and that the number of cohorts is falling in the higher age groups.) 42 | Mobility inequality and social exclusion The majority of relocated children stay with relatives. (Figure 12) These are often grandparents and siblings, but also other close relatives like uncles and aunts (note, though, that the terms aunt and uncle are commonly used beyond the strictly biological definition in Benin). It is more common for the relocated girls to live with close relatives, while more boys stay with tutors, employers or more remote relatives. A small share of the children is already married and has relocated to stay with their spouses. This is not unexpectedly more often the case for girls than boys. Around 5 percent live in a household controlled by their fathers, and this is the case for more boys than girls. A few boys are also reported to live by themselves. Alone 0 2 Tutor 8 10 Employer 5 6 Remote reatives 3 5 Marabout 0 1 Girls Other relatives 45 36 Boys Grandparents 13 7 Brother/sister 15 18 Household of father 3 9 Spouse 6 3 0 5 10 15 20 25 30 35 40 45 50 Figure 12. Who the mobile children live with, by gender (in percent of all relocated children). Figure 13 shows where the relocated children currently reside. More than half have relocated towards urban areas, almost three-in-five of the girls and 54 percent of the boys. Among those who stay in rural areas, around 10 percent remain in the home community while 16 percent are in a neighboring community. In numbers this would indicate that more than 60 000 girls and almost 40 000 boys under 18 years-of-age have relocated towards the urban areas of the country. Mobility, inequality and social exclusion | 43 Urban community 58 54 Other rural community 19 17 Girls Neighbouring community 14 18 Boys This community 8 11 0 10 20 30 40 50 60 70 Figure 13. Where the mobile children reside, by gender (in percent of all relocated children). Among the girls who had previously been away from their households and returned, one-in-four had worked as domestic servants. Figure 14 shows the various relocation motivations among children currently away from home, as quoted by the children’s representatives. Most notably, 42 percent of the boys and 38 percent of the girls had left in order to go to formal school. Not all of the children intended to go to school were in school, though. Around 5 percent of them had never gotten to start, while another 5 percent had dropped out. Moreover, around one-in-four children who had left for what was presented as social reasons were in school, while one-in-three children who had left for economic reasons also were attending school at the time of the survey. Even 11 percent of girls who had left primarily to do domestic service also attended school, and as many as one-in-four boys who left to do domestic work were also in school. Social Econ. Other Econ. Other Social 10 % 4 % 3% 3% 4% 9% Marriage Formal Marriage school 2%Other Formal 4% 38 % work school Other 4% 42 % work Domestic 7% work 3% Domestic work Apprenti 13 % ce Apprentice 32 % 22 % Figure 14. Objective for current relocation, by gender: boys versus girls (in percent of all relocated children). 44 | Mobility inequality and social exclusion Of the 106 000 relocated girls 22 percent had left for an apprenticeship. This corresponds to 23 000 girls. Among the boys, 32 percent had left for apprenticeships, or an estimated 24 000. Furthermore, 13 percent of the girls were reported to have left explicitly to do domestic work and 7 percent for other work. This makes up around 21 000 girls. Similarly for boys, 3 percent had left to do domestic and 4 percent for other types of work. Marriage constitutes the motivation for 4 percent of girls leaving home, and an even smaller share of boys. Around 9 percent of boys and girls had left for what was stated as social reasons. Social reasons were clearly more often quoted as motivation for the youngest children. Economic reasons were to the contrary more often stated as motivation for older children. The survey contained a panel of questions related to the attitudes of household heads to various practices of mobility of children under the age of 18 years (Figure 15). Two-in-three are positive or conditionally positive to the placing of children with relatives. Letting children leave to pursue an apprenticeship is also positively viewed by three-out-of four rural household heads. The tradition of placing children in households of influence and importance is still relatively positively viewed: twenty-two percent sees it as positive while an additional 29 percent finds it positive on certain conditions. In terms of the explicitly labor motivated types of relocation, almost half the household heads found it positive or conditionally positive to let children leave to work in the agriculture. Only 12 percent said it was unconditionally positive to leave to do domestic work, but an additional 30 percent were willing to consider such arrangements on certain conditions. Relocation for agricutural work 21 27 52 Placement for apprenticeship 60 17 24 Positive Placement as domestic 12 30 58 Conditionally positive Placement with important family 22 29 48 Negative Placement with relatives 40 26 34 0% 20 % 40 % 60 % 80 % 100 % Figure 15.Attitudes to child mobility: Do you look positively on the following types of child mobility? (In percent of all household heads). Mobility, inequality and social exclusion | 45 Parents of children below 18 years-of-age were then asked how they would have welcomed a variety of relocation opportunities for their child. Figure 16 shows a predominantly positive attitude towards apprenticeship opportunities for both boys and girls. With regards to boys, 42 percent would also welcome the opportunity for one of their sons to stay with urban relatives, 29 percent with rural relatives, 28 percent with an influential family, and 25 percent also finds agricultural work elsewhere an attractive option. For the girls, 40 percent similarly welcomes the opportunity to let a daughter live with urban relatives, 28 percent with rural relatives, 26 percent with an influential family and finally, 13 percent would welcome to opportunity for one of their daughters to go do domestic work. Girl to live with important family 26 Girl to live with urban relatives 40 Girl to live with rural relatives 28 Girl to work as domestic 13 Girl for apprenticeship 76 Boy to live with important family 28 Boy to live with urban relatives 42 Boy to live with rural relatives 29 Boy for agricultural work 25 Boy for apprenticeship 84 0 10 20 30 40 50 60 70 80 90 Figure 16. Attitudes to child mobility: Would you have liked for your son/daughter to relocate for any of the following reasons? (In percent of all household heads). Two-in-three household heads stated that they had children who they believed would benefit from living in other places and with other people. Half of them still worried about the maltreatment of relocated children, and were therefore reluctant to encourage it. Sixteen percent said they did not know people who could help facilitate for a child’s relocation, 15 percent that 46 | Mobility inequality and social exclusion they did not know of any good people the child could go stay with, 8 percent simply said they did not know how to do it, while 5 percent said they did not have the necessary resources to get it organized. Asked specifically what they thought of the treatment and punishments of children placed with relatives or employers, answers varied. There is a relatively high acceptance for corporal punishment, slapping and reprimands. Insults are generally accepted, but curses are not. There is close to no acceptance for sleep or food deprivation or overworking the child. Figure 17 shows that one-in-four accepts corporal punishment, 44 percent accepts it conditionally, while only one-in-three would fully oppose it. Similarly, near 60 percent accepts or conditionally accepts slapping. Extra work 5 10 85 Food deprivation 4 5 91 Sleep deprivation 4 8 88 Misusing the child's money 8 8 84 Cursing the child 4 8 88 Insults 45 29 26 Reprimands 52 28 19 Corporal punishment 24 44 32 Occasional slap 29 30 41 0% 10 % 20 % 30 % 40 % 50 % 60 % 70 % 80 % 90 % 100 % Acceptable Conditionally acceptable Unacceptable Figure 17. Attitudes to child mobility: Views on the treatment of relocated children (in percent of all household heads). When a child is sent to live with relatives, the household of origin often have some expectations vis-a-vis the hosting family. Figure 18 gives an overview of some such expectations. Educating the child, implicitly also raising the child with good morals and manners, is the main expectation reported. Eighty-five percent say they expect the receiving household to nourish and protect the Mobility, inequality and social exclusion | 47 child, while 16 percent also hope the other household can help find the child a good spouse. Only 17 percent expect financial support from the household where the child will live, while some more, 28 percent, expect that this household will support them if they get into some sort of a future crisis situation. Most striking is however the large regional differences. Expectations are by far the highest in the North. While only three-in-four household heads in the South has the most basic expectation to the new household – that they will help nourish and protect the child – almost all the Northern parents hold this expectation. One-in-four household heads in the North thinks the other household should help identify a good spouse for the child, while this is a far less common expectation in the South. Also, half of Northern household heads expect the new household of the child to be part of their support network in times of crisis, and one-in-four expects to receive money on a regular basis. In comparison, only one-in-five Southern household heads expects support in crisis, and only one-in-ten expects money on a regular basis. 20 Help you out in times of crisis 22 48 10 Send you money on a regular basis 16 26 13 Help find the child a good spouse 14 25 95 Support the child's education 95 91 76 Nourish and protect 85 94 0 10 20 30 40 50 60 70 80 90 100 South Center North Figure 18. Attitudes to child mobility: What would be your expectations to the household where your child is placed (in percent of all household heads). 48 | Mobility inequality and social exclusion Girls who have left for domestic work: numbers and attitudes It is difficult to obtain good statistical figures on the domestic service of young girls. Most Beninese girls work in and around the households where they reside. Most of those living away from home work for their grandparents or other relatives. Very few work for someone they are completely unrelated to. So who should be considered a child domestic worker? The amount of work could decide. Their lack of school attendance is another indicator. But social identification is probably equally important: who is considered a foster child and who is seen as a domestic worker does not necessarily depend on the work load. A strong indicator of who is a domestic worker is whether there is a salary involved. Monetarization of a fostering relationship changes its’ social dynamics. The young girl does not help out – she owes her labor to the household. A glance at the current household members of rural households gives a brief historic picture. Most of the female household members who report to have lived away from home in a past period also states to have done domestic work while away. Two trends are clear: the older the women, the less likely they worked as domestics while away from home. The younger, the more likely, but also the more likely that they combined domestic work with schooling in some other place. Eighty percent of the girls between 15 and 20 years-of-age with previous mobility experience are reported to have done some sort of domestic work while away. Thirty-four percent have done domestic work while studying, 31 percent are stated to have “helped� with domestic work while away without going to school, while 14 percent were explicitly defined as previous domestic workers. This chapter mainly looks at the last of these groups, or what is most likely the tip of the iceberg: girls who are explicitly defined as domestics workers by their family. The age definition changes from what was used for the previous section on child on mobility: girls above a certain age and young women are in focus, more precisely the age group of 6-20 year-olds. Figure 14 in the previous section showed that 13 percent of the girls below 18 years-of-age who were currently living away from home had left explicitly to do domestic work. But what about the 9 percent that were said to have left for social reasons and the 3 percent for economic? And what about the “other� reasons? Is not a girl who leaves home to marry before the age of 20 possibly a domestic worker? An indication of the working status of the girls who relocated for social, economic, conjugational and other reasons is given by the girl’s schooling status. Figure Mobility, inequality and social exclusion | 49 19 shows that school attendance rates indeed are poor for all the groups of girls in question. While only 6 percent of the young girls who got married are still in school, 7 percent of those who left for economic reasons, 26 percent of those who left for “social� reasons and 15 percent of those who left for what was stated as other reasons are. In comparison, 13 percent of the girls who left explicitly to work as domestic workers also attend school. The share of 6-20 year-old girls currently away to work as domestic servants corresponds to around 30 000 persons. It is however reason to question whether not also many girls who have left for other explicit purposes could in effect be living as domestic servants or aids. Among the estimated 186 000 rural girls 6-20 years-of-age who have left their household of origin, only one-in-three are in school. Leaving out those with a stated other job or in apprenticeship still leaves around 60 000 out-of-school girls who did not leave to marry. Other reasons 15,0 28,7 56,2 Economic reasons 7,2 47,8 45,0 Social reasons 25,8 30,3 44,0 Marriage 5,6 36,9 57,5 Domestic work 12,5 36,1 51,4 0% 10 % 20 % 30 % 40 % 50 % 60 % 70 % 80 % 90 % 100 % Attending school Dropped out Never attended Figure 19. Schooling status of currently relocated girls 6-20 years-old, on stated objective for leaving (in percent of all relocated girls). Relocations often happen repeatedly. A 19 year-old girl who has previously left to work as a domestic may for her latest relocation have left to get married. In order to map the use of domestic services as a social and economic option, the better question is whether the girl has ever work as a domestic for others than her household of origin. Figure 20 shows how common domestic service is among the girls and young women in rural households. Almost 5 percent of girls already have experience with domestic service in the age 50 | Mobility inequality and social exclusion group from 6-10 years-of-age. In the 11-14 age group 6 percent of girls have the experience, while 9 percent of 15-17 year-olds and 13 percent of 18-20 year-olds do. The total prevalence is higher in the Southern and Central regions of the country (7,2 percent), and a little lower in the North (5,4 percent). A noticeable difference between the Southern and Central area is that the girls in the South seem to be leaving for domestic work at an earlier age. 16,0 13,6 13,2 14,0 12,8 12,7 12,0 9,8 10,0 8,9 North 8,4 8,0 7,5 Central 6,6 6,4 5,6 South 6,0 4,7 4,7 5,1 4,3 Total 4,0 2,9 2,0 ,0 6-10 11-14 15-17 18-20 Figure 20. Share of girls 6-20 year-old who have worked as domestics for others than their household of origin, by age-bracket and region (in percent of all girls). In numbers, these figures translate into around 70 000 rural girls and young women 6-20 years- of-age, who are explicitly stated to have worked as domestic servants away from their households of origin – and perhaps more than once. Another illustration of prevalence occurs by looking at the share of households with girls in the age range 6-20 years that is the original home of at least one girl with experience as a domestic servant. While 8,5 percent of the rural households in the North is the home of at least one girl with this experience, 13,5 percent of households in the Central areas and 12,2 percent of households in the South are. The household survey also posed a series of questions related to the attitude towards domestic work. Figure 21 shows that 12 percent of household heads are positive to the practice of sending Mobility, inequality and social exclusion | 51 girls to work as a domestics in other households, while an additional 30 percent considers this a positive opportunity under certain conditions. Household heads in the North are most positive, while those in the South are more skeptical. Among the households that actually had children below 18 years-of-age, 13,4 percent of household heads state that if there had been an opportunity, they would have sent one of their daughters to work as a domestic. Parents in the North were most positive (17,5 percent), in the Center a little less (14,8), while parents in the South again were most skeptical (8,3 percent). Total 12 30 58 Sud 9 25 65 Centre 12 34 54 Nord 17 29 54 0% 10 % 20 % 30 % 40 % 50 % 60 % 70 % 80 % 90 % 100 % Poistive Conditionally positive Not positive Figure 21. Rural household heads’ attitudes to the practice of sending children for domestic work elsewhere (in percent of household heads). Figure 22 gives an indication of the attitudes rural household heads hold towards the treatment of domestic workers. While it is known that young domestic workers often are the first to get up and the last to go to bed in the households where they serve, sleep deprivation is the least accepted treatment of domestic workers. Food deprivation as punishment is also generally frowned upon. There is also little acceptance for holding back the salary of a domestic worker to show dissatisfaction. While reprimanding is by far the most accepted type of punishment, there is also a high acceptance for corporal punishment and an occasional slap. Sixty percent find corporal punishment of the domestic acceptable or conditionally acceptable, while 55 percent accepts or conditionally accepts occasional slapping. 52 | Mobility inequality and social exclusion As the figure shows, corporal punishment is most accepted in the North, where almost 70 percent of the household heads accepts or conditionally accepts the corporal punishments of a domestic worker. In the South, in comparison, 45 percent of the household heads finds corporal punishment acceptable or conditionally acceptable. Slightly more than half of household heads in the Center of the country finds accepts corporal punishment while almost 70 percent accepts or conditionally accepts an occasional slap. Sleep and food deprivation are quite equally rejected as acceptable punishment across the country, while there is a higher acceptance for withholding salary in the North, where one-in-five finds this conditionally acceptable. South 2 6 Witholding salary Center 3 10 North 1 24 South 1 4 deprivation Sleep Center 1 5 North 1 10 South 2 9 deprivation Food Center 3 9 North 2 13 Reprimands South 49 25 Center 54 21 North 50 28 South 21 37 Occasional slapping Center 28 39 North 21 27 punishment South 17 28 Corporal Center 33 21 North 40 29 0 10 20 30 40 50 60 70 80 90 Acceptable Conditionally acceptable Figure 22. Rural household heads’ attitudes to treatment of “disobedient� domestic workers (in percent of all household heads). Mobility, inequality and social exclusion | 53 4 Relationships: mobility, poverty and shocks How is child mobility and girls taking jobs as domestic workers related to poverty and shocks? This chapter looks at such relationships. Measuring incomes in rural households is often an unreliable approach. Poverty is here therefore addressed in three ways. First by assessing the construction quality of the main household building, second by counting the durable assets in possession of the household, and third, by the heads’ own assessment of the household’s food security situation. The first reflects the long-term poverty situation, the second provides a picture of the current one, while the last is likely to reflect the overall vulnerability of the household. The chapter then examines the possible effects of the 2010 floods and other shocks on the mobility situation of its’ young members. When comparing household choices regarding child mobility, the sample used for the analysis is restricted to households with children below a certain age. Households with no children in the relevant age range do not really face the choice of encouraging or discouraging a child or young girl’s relocation choice. Households with children below 18 years-of-age will be considered for the child relocation analysis, and only households with girls 6 to 19 years-of-age will be compared for the analysis of domestic work. Child relocation There were 2 078 households representing children below the age of 18 years in the sample. With a total of 8 273 children in the age range, this gives an average of 4 age-relevant children in each of those households. Seventeen percent of these households had at least one child living away from home for a variety of reasons, including schooling. If the main interest is to study the relocation of children for other reasons than schooling, a special focus should be kept on households that had relocated children for other reasons. In total, 11 percent of households representing children below 18 years-of-age had at least one child below 18 years-of-age who had left the household and did not go to school. For the section on domestic work, there is focus on two categories of households. First, households where at least one daughter is reported to at some stage have been a domestic worker elsewhere, and second, households representing girls for whom the explicitly stated motive of their last relocation was domestic work. The latter group is clearly the most interesting one when assessing the impact of the 2010 floods. 54 | Mobility inequality and social exclusion Child mobility and household poverty The rural households in our sample are generally poor. Household building standard can be used as a proxy to poverty. Houses having at least iron sheet roof, brick walls and cemented floors or equivalent are considered as solid. While this is only the case for 17 percent of the households in the sample, it is also possible to grade household quality based on how many of the three main building parts that are constructed in a solid material. Two trends stand out with regards to child mobility. First, there is little difference between the 19 percent of the households that have no solid parts and the 40 percent of the households with one solid part. Second, housing quality seems to correlate more strongly with child relocation for other purposes than schooling than for child relocation in general. Figure 23 shows child mobility relative to household poverty assessed by housing quality. The first two bars show that among households with no solid parts (0) 17,5 percent have at least one child living away from home, while 12 percent have at least one child living away from home without going to school. For households with 1 solid part, there is not much change, while the likelihood that a household has a child living away for other reasons than school gradually drops towards 7 percent for the households with all solid housing material. For child relocation now and before (the dark columns) the correlation with housing quality is less distinct and mainly distinguishes the “all solid� households from the rest. As households can afford all solid building materials, fewer children are away, and a larger share of those away are in school. 20,0 17,5 17,1 18,0 16,8 15,4 16,0 Child away for school 14,0 12,0 Child away for other reasons 10,0 12,0 than school 8,0 12,3 Linear (Child away for school) 6,0 10,5 4,0 7,0 Linear (Child away for other 2,0 reasons than school) 0,0 0 1 2 3 Figure 23. Poverty and child mobility: Housing quality. Share of households with relocated children on number of solid parts (floor, roof, walls) and mobility type. Mobility, inequality and social exclusion | 55 Looking at child mobility using the number of durable assets as a proxy for poverty reveals a similar pattern. Adding up household ownership to a radio, decoder/parabola antenna, TV, CD player, DVD player, fridge, phone, bicycle, moped, car, generator and electricity assess shows that 17 percent have no such assets, 22 percent only one, 22 percent have two, while around 22 percent have three or four. Only 17 percent have five or more assets. Figure 24 indicates a negative correlation between asset ownership and both child mobility overall, and child mobility for other reasons than schooling. While 23 percent of households with no assets have children living away from home, 16 percent have children living away for other reasons than schooling. The share of households with relocated children drops by asset ownership. More asset rich households also have relocated children, but not as often. Among households with more than 5 assets, only 15 percent have children living away from home and only 9 percent have children living away from home for other reasons than schooling. 25,0 22,6 20,0 17,9 Child away for school 16,2 14,5 15,0 13,7 Child away for other reasons than school 15,7 10,0 11,4 Linear (Child away for school) 10,4 8,8 5,0 9,0 Linear (Child away for other reasons than school) 0,0 0 1 2 3-4 5 and more Figure 24. Poverty and child mobility: Asset ownership. Share of households with relocated children on number of assets owned and mobility type (in percent of households with children 0-17 years-of-age). A third approach to poverty is to look at food consumption and nutritional value. Three-in-four respondents had eaten meat, fish or fromage Peuhl (a protein rich local cheese) the day before the survey interview. While 12 percent of those that had eaten protein rich foods had relocated children from the household, the share was 18 percent for those who had not eaten foods with 56 | Mobility inequality and social exclusion protein the day prior to the survey. Similarly, 8 percent of households where they had eaten protein rich foods the day before the survey had relocated children for other purposes than schooling, while this share raises to12 percent in the group that had not. Only 20 percent of the household heads said they did not worry at all about supplying their families with food during the past 12 months. Almost half worried sometimes, while one-in- three said they had worried a lot. Figure 25 shows child relocation in relation to how much the household heads worried about their food situation. Correlations seem quite strong for this more vulnerability-related variable, and stronger than for the other two poverty measures. Seven percent of households with heads that did not at all worry about the food situation had children living elsewhere. This share rises to 16 percent among households where the head worried sometimes, while among the households where the head had worried a lot about the food situation, 23 percent had children living elsewhere. While 5 percent of households with a head that did not worry had children living away for other reasons that school, this was the case for three times as many of the households where the household head had been worrying a lot. It is important thought, to remember that these findings are mere correlations and not necessarily all due to causality. The fact that many children leave the “worried� households also to go to school elsewhere may reflect the possibility that there are fewer schools in areas with low food security. 25,0 20,0 7,9 15,0 5,8 Child away for school 10,0 Child away not for school 15,1 2,2 5,0 10,5 5,2 0,0 Worried a lot Worried sometimes Never worried Figure 25. Poverty and child mobility: Food security. Share of households with relocated children on level of worry over ability to provide household members with food during the past 12 months (in percent of all households with children 0-17 years-of-age). Mobility, inequality and social exclusion | 57 Child mobility and the 2010 floods The Floods of 2010 appears to have had a relatively limited but systematic impact on the relocation patterns of household children. While 19 percent of the households that describe themselves as “flood affected� had at least one relocated child, the share was 15 percent for those that did not consider themselves affected. Some of the relocated children were attending school. If these are removed from the sample, 12 percent of the affected households had children who had relocated for other purposes than school, compared to 10 percent for those unaffected. Not all the affected households describe themselves as “shocked� by the flood, the way shocks was explained to the respondents. The ways the households reported to have been affected is however indicative of their trauma. As described, people were affected in different ways, many in more than one. Seven percent of the households with children below 18 years-of-age reported to have withdrawn children from school as a consequence of the flood. Not surprisingly the highest child mobility rates were found among this relatively small group of households: one-in- three reported to currently having children living elsewhere, compared to 16 percent among household that had not report this consequence. Most of the children of these families did not leave to go to school elsewhere, as 28 percent of these households represent children living away from home without going to school. In the following, radar-diagrams will be used to show the relationships between household child mobility rates and shock exposure and impact. All the diagrams presented have a common format: the spokes of the diagrams refer to different shocks or consequences (described on their particular label), and the data points show the child mobility rates of the households unaffected by this shock or consequence (the inner data point) compared to the households reporting to be affected by this particular shock or consequence (the outer data point) respectively. The radar diagram in Figure 26 gives a graphic presentation of the difference in child mobility rates between households that suffered particular consequences of the 2010 floods and those that did not. In short, the relationship between suffering a crop loss on child mobility seems relatively limited, as households that did suffer such a crop loss had a child mobility rate of 18 percent, and those who did not only had a two percentage point lower rate. This difference was however as large as around 5 percentage points for most of the other consequences suffered, and most distinct for households reporting that they had been forced to borrow money. While 22 58 | Mobility inequality and social exclusion percent of the households that had to borrow money as consequence of the floods had at least one child living away from home, this was 6 percentage points more than the households that did not report that they had been forced to borrow. Also, 21 percent of household reporting to have suffered from illness or malnutrition as a consequence of the flood disaster had at least one child living away from home. In comparison, only 16 percent of the household that did not report illness or malnutrition had a child living elsewhere. The impact was similar in households reporting to have lost buildings or other household infrastructure, households where people had had to flee during the crisis, and households where all savings had been depleted. The difference is also clear for households reporting to have lost animals or farm fish. People had to flee the household 25 21 20 Had to borrow money Loss of crop 22 15 16 18 10 16 16 5 0 16 20 15 Spent the savings 20 Loss of animals/fish 16 15 20 21 Loss of Illness/malnutrition buildings/infrastructure Not experienced Experienced Figure 26. Share of households with at least one child living away from home, on type of consequence suffered from the 2010 floods (in percent of households with children 0-17 years-of-age). Figure 27 is similar, but focuses on the households that have children who live away from home without going to school. As in the previous figure, more households reporting impact from the Mobility, inequality and social exclusion | 59 floods also have children living away from home without going to school. Almost 17 percent of the households where members had to flee during the floods had at least one child living away from home without going to school. Sixteen percent of the households that had to borrow money also have children living away from home without going to school. Child mobility also seems related to the loss of household buildings or other household infrastructure during the floods. Similar to child relocation in general, the impact from pure crop loss is not that decisive, while those reporting to have lost livestock, including fish farms also seem to do have an increased mobility of children. People had to flee the household 18 17 16 14 Had to borrow money 16 12 Loss of crop 10 10 8 12 10 6 10 4 2 0 10 13 10 Spent the savings 14 Loss of animals/fish 9 10 13 15 Loss of Illness/malnutrition buildings/infrastructure Not experienced Experienced Figure 27. Share of households with at least one child living away from home and not going to school, on type of consequence suffered from the 2010 floods (in percent of households with children 0-17 years-of-age). 60 | Mobility inequality and social exclusion Child mobility and other shocks Figure 28 looks at relationships between child mobility and other shocks reported to have struck the households in the past decade. As described, a shock was explained to the respondents as an event that had led to a serious reduction of assets, had caused a substantial reduction of revenue or led to a considerable cut in household consumption. The assumingly most common and serious type of covariate crisis is drought, but droughts interestingly does not seem to correlate much with child mobility overall. Floods, cricket attacks and fire correlates somewhat stronger, while the shocks most clearly correlated with child mobility are the two idiosyncratic shocks studied: the death of a breadwinner in the household and the illness of a family member. While mobility rates for families reporting to have been shocked by flood was 4 percentage points higher than among the un-shocked, the rise was of 7 percentage points in households where a breadwinner had died or a family member had suffered from serious illness. Drought 25 Illness of household 20 Insect attack member 17 20 15 17 19 10 13 16 5 21 14 Death of breadwinner 0 15 Floods 19 16 17 17 16 19 Price increases 18 Fire Animal disease Not exposed Exposed Figure 28. Share of households with at least one child living away from home on type of shock suffered in the past decade (in percent of households with children 0-17 years-of- age). Mobility, inequality and social exclusion | 61 A similar pattern emerges in Figure 29 where the focus is on children who are away from home without going to school. It seems a systematic pattern that households reporting to be “shocked� more often have children living away from home without going to school. There is however still a clear difference between systemic crises and individual household shocks. Covariate shocks like drought, flood and insect attacks correlate somewhat with child mobility, while correlations seems clearer in the relatively small group of households that reported to have been shocked by fires (9 percent of households). The more considerable correlation however seems to be found in household where an adult breadwinner has died or a household member has been sick. Drought 16 14 Illness of household 11 12 Insect attack member 10 13 10 13 8 6 10 8 4 2 Death of breadwinner 14 9 Floods 0 10 12 10 10 11 10 15 Price increases 12 Fire Animal disease Not exposed Exposed Figure 29. Share of households with at least one child living away from home and not going to school on type of shock suffered in the past decade (in percent of households with children 0-17 years-of-age). 62 | Mobility inequality and social exclusion Coping responses to shocks says something about how badly the family was affected. Almost all the respondents, 82 percent, said they had to spend their savings. One-in-three reported they had to sell land or livestock, another one-in-three that they had to sell other assets. Forty-two percent said they had to reduce the number of meals and almost half that they reduced the consumption of nutritious foods like meat, fish and protein-rich cheese. Figure 30 shows differences in child mobility rates between households that applied different types of coping mechanisms after suffering their latest shock. Sold land or livestock 25 20 17 15 17 10 Spent savings Sold other assets 17 5 20 14 15 0 14 15 19 21 Reduced meals with protein Reduced number of meals Did not apply this coping strategy Applied this coping strategy Figure 30. Share of households with at least one child living away from home, on type coping mechanism applied during the last shock to the household (in percent of households with children 0-17 years-of-age). Mobility, inequality and social exclusion | 63 Among the households that had sold land or livestock, there was very little increase in child mobility. Land or livestock may act as a buffer against other and more detrimental coping responses, but it is also likely that those who did indeed had such a buffer might have been better off in the first place. Those who had had to sell other assets had a considerably higher rate of child mobility. And those who had to apply the more desperate means, like reducing meals and consumption of protein rich foods, reported much more child relocations away from the household. This was also true for those who reported having spent their savings. It may seem as while almost all households had responded to shocks by depleting savings, those that owned land or livestock were able to manage without their children leaving. Those who had no such buffer and needed to resort to reducing food intake had also encouraged children to leave. Girls and domestic work This section looks at the 1 409 households which represented at least one girl the age range 6 to 19 years-of-age. There is on average 2 girls represented by each of these households: that is, a total of 2 912 girls. In this section, analysis is based on these households only, since households without girls in the age group do not have the choice of encouraging or discouraging girls to look for domestic work and should thus not be considered. Eleven percent of the households represented at least one girl who had at some stage worked as a domestic, while 5 percent sited domestic work as the motive for the latest relocation of at least one of the girls in the household. It is important to note that this definition does not include all the more obscure relocation arrangements of girls towards households where they still end up doing a lot of domestic work. Instead the highlight is on the more clear-cut situations where the relocation is explicitly motivated by domestic work. The structure of the previous section is kept: first, the relationship between household poverty and domestic work will be described, followed by figures describing household experiences with the 2010 flood and other shocks. Domestic work and household poverty Figure 31 shows the relationship between household poverty assessed by household building quality. The share of households with at least one girl working as a domestic servant is twice as 64 | Mobility inequality and social exclusion high when the main household building has no solid construction parts, compared to a household that has an all-solidly built main building construction. In other words, some rural households that can afford a good house building are still the origin of some girls who work as domestic servants. However, the poorer the housing the more likely it is that girls leave to do domestic work. For households with a girl who at some stage worked as a domestic, housing quality also correlates in the expected direction, but neither as strongly nor as systematically as the case is for households where girls currently work as domestic servants, or have done so recently. 14,0 13,2 11,5 12,0 Household girl ever away for 10,0 10,0 domestic work 8,6 8,0 7,0 Household girl last left for 5,9 domestic work 6,0 4,1 Linear (Household girl ever 3,6 4,0 away for domestic work) 2,0 Linear (Household girl last left for domestic work) 0,0 0 1 2 3 Figure 31. Poverty and domestic work: Housing quality. Share of households with girl domestic workers on number of solid parts (floor, roof, walls) and mobility type (in percent of all households with girls 6-20 years-of-age). The same trends are clear when household poverty is assessed by a count of durable assets (Figure 32). The likelihood of having a girl living away from the household to work as a domestic decreases gradually with the accumulation of durable assets. While a little more than 7 percent of the asset poorest households have at least one girl working as a domestic, the share is 4 percent for the households with most durable assets. Being the home of a girl who has at some stage worked as a domestic also correlates negatively with asset accumulation, but the relationship is not quite that strong, increasing from 10 to 14 percent from the asset poorest to the wealthiest. Mobility, inequality and social exclusion | 65 16,0 13,9 14,0 12,5 Household girl ever away for 12,0 10,8 domestic work 9,7 10,0 10,0 Household girl last left for 8,0 7,3 domestic work 6,1 6,0 4,8 Linear (Household girl ever 4,3 4,1 away for domestic work) 4,0 Linear (Household girl last left 2,0 for domestic work) 0,0 0 1 2 3-4 5 and more Figure 32. Poverty and child mobility: Asset ownership. Share of households with girl domestic workers on number assets owned and mobility type (in percent of all households with girls 6-20 years-of-age). The third approach to poverty in this analysis is based on how much the household heads says he or she worried about being able to provide his or her family with food during the 12 months prior to the survey. Figure 33 shows that the more they worried about the food situation, the more likely it was that girls from the household either were or had been away to do domestic work. Among household heads that never worried, almost no girls had left to work as domestics, and relatively few had ever been out. Among those who worried a lot, however, the shares rocket. 16,0 14,9 14,0 12,0 10,7 10,0 9,1 Household girl ever away for 8,0 domestic work 6,2 6,0 Household girl last left for 3,8 domestic work 4,0 1,9 2,0 0,0 Worried a lot Worried sometimes Never worried Figure 33. Poverty and child mobility: Food security. Share of households with girl domestic workers on level of worry over ability to provide household members with food during the past 12 months (in percent of all households with girls 6-20 years-of-age). 66 | Mobility inequality and social exclusion The most striking observation from the three figures together is that the general outlook or concern level of the household head seems to matter much more than asset or building based assessments of poverty. While the share of girls currently away to work as domestics was twice as high among households with no solid building parts as compared to those with all solid parts, the increase more than quadruples when a household head goes from not worrying about the food situation to worrying a lot. This “food-security-assessment� variable can also be interpreted as a vulnerability indicator, more than a mere poverty indicator, and vulnerability would in that case be more decisive than poverty in itself. Moreover, the notion that the state of mind or perhaps pessimism of a household head is strongly related to the mobility away from his or her household is supported by a quick glance at standard indicators of trust. Household heads that find it difficult to trust other people and feel they are often taken advantage off seem much more likely to have daughters who work as domestic servants. Domestic work and the 2010 floods Current household poverty may be related to girls doing domestic work both now and in the past. However, in assessing the impact of the floods of 2010 it is more relevant to look at girls who are currently away to work as domestics and less interesting to include those doing domestic work in past years. By first glance, households reporting to have been affected by the floods of 2010 do not appear to have a higher rate of girls being away to work as domestic servants. However, when asking about how they were affected, some clear patterns arise. Figure 34 shows some clear and systematic differences between households reporting damaging consequences of the floods and those who did not report those consequences. Almost one-in-five households had had to borrow money after the floods, and among these households 10 percent had at least one girl out working as a domestic servant, compared to only 4 percent of households that did not have to borrow money. Households where members had had to flee during the floods had a 4 percentage point higher rate of girl domestics, and so did households that had lost animals or farm fish or had suffered from illness and malnutrition as a consequence of the floods. Differences were less striking for households reporting crop loss, loss of household buildings or other household infrastructure and households that had lost their savings. Mobility, inequality and social exclusion | 67 People had to flee the household 10 8 8 Had to borrow money Loss of crop 6 10 4 5 6 4 2 5 0 5 5 Spent the savings 6 8 Loss of animals/fish 5 5 6 8 Loss of Illness/malnutrition buildings/infrastructure Not experienced Experienced Figure 34. Share of households where at least one girl has left to do domestic work, on type of consequence suffered from the 2010 floods (in percent of all households with girls 6-20 years-of-age). Domestic work and other shocks Broadening the perspective to other shocks to the households over the past decade, a further relationship between shocks and girls leaving to work as domestics seems to emerge (Figure 35). A relatively low share of the households reported to have been shocked by fire, but among the households that had, 9 percent had at least one girl working as a domestic, compared to 5 percent of those who had not experienced fire. Households reporting to have experienced other covariate, systemic shocks like drought, flood, insect attacks and animal disease also systematically reported more mobility of girls towards domestic work. The greatest differences are however found between households reporting individual shocks, and those that do not. Eight percent of households that have experienced the death of a breadwinner had a girl away working as a domestic, compared to only 3 percent of those not reporting such a loss. Similarly, seven percent of households that reported to be shocked by illness had a girl away working as a domestic, compared to 4 percent of those who did not report such illness. 68 | Mobility inequality and social exclusion Drought 10 9 Illness of household 8 Insect attack member 7 6 6 7 5 7 4 5 3 5 4 2 1 Death of breadwinner 0 Floods 8 3 5 6 5 5 6 5 5 Price increases 9 Fire Animal disease Not exposed Exposed Figure 35. Share of households where at least one girl has left to do domestic work, on type of shock suffered in the past decade (in percent of all households with girls 6-20 years-of- age). Figure 36 shows rates of mobility towards domestic work by the coping mechanisms applied by shocked households. Households that had depleted their savings to cope with shocks did not have a much higher rate of girls leaving to do domestic work as compared to those who had not been forced to apply this coping mechanism. It was, however, more common for girls to leave households that had been forced to sell land or livestock, and much more common to leave households that were forced to sell other assets or reduce the number of daily meals. In fact, households selling assets and reducing meals had a twice as high likelihood of having girls leaving to become domestic servants as those that did not resort to this type of coping measures. Mobility, inequality and social exclusion | 69 Sold land or livestock 8 7 7 6 5 4 5 3 Spent savings Sold other assets 2 8 6 4 1 4 0 4 5 6 Reduced meals with Reduced number of 8 protein meals Did not apply this coping strategy Applied this coping strategy Figure 36. Share of households where at least one girl has left to do domestic work, on type coping mechanism applied during the last shock to the household (in percent of all households with girls 6-20 years-of-age). In conclusion, individual household shocks seemed more decisive to the likelihood of girls leaving for domestic work than systemic shocks, although both types of shocks seem to influence the choice. Simply being affected by a crisis does not necessarily lead more girls to leave the household to work as domestic, but certain types of consequences suffered seems to heighten the likelihood that they do, notably the fact that many become indebted. The way families cope with the shocks they experience also seems to matter. In particular, households that resort to sell assets and reduce meals are also likely to encourage their girls to leave home to find domestic work elsewhere. 70 | Mobility inequality and social exclusion Estimating effects, a regression approach Cross tabulating results the way it has been done in the previous sections does not prove causation. The results presented document that more children leave and more girls leave for domestic work in households that suffered some serious consequences of shock. However, one cannot say for certain that the shock and e.g. the debt situation it caused triggered the relocation choice. The two correlate, and that is all that can be said for certain. It is however possible that other factors caused the correlations observed. For example: there is less child mobility in the North of the country. People were also differently affected by the floods in the North. Perhaps some demographic features of the North explain the lower mobility rates, rather than the flood exposure? Also, in cross tabulations, Muslims have lower child mobility rates than people with other religious affiliations. Could it be that Muslim populations were less affected by the floods and that religious beliefs is the real connection to the lower mobility observed? What about household heads’ outlook on life in general? It seems as if households with pessimistic household heads produce more mobility of children and of girls for domestic work. A regression analysis cannot prove causality, but by holding other suspected interfering factors constant it can help isolate the relationship between the shock factor and the mobility rate. The claim of a relationship is strengthened if the shock factor remains statistically significant when other potentially determining factors are controlled for. If the shock factor is not statistically significant in this a complex setting, the claim of a relationship might indeed be dismissed altogether. In the very basic regressions presented in this section, the impact from flood (or other shocks) is controlled for factors like household poverty, the religion, gender and literacy of the household head. The trust level of the household head is used as s proxy to his or her vulnerability and outlook on life, and a regional control are inserted to represent cultural variations. In short, the regressions will help strengthen the claim that the flood or another shock had an isolated impact on the mobility of the child or the girl, and that the correlation shown in the previous section is not random and indeed due to some other jointly determining background factor. Table 5 shows the results of regressing “child mobility for other purposes than schooling� on “indebtedness after the flood� controlling for some basic socio-demographic features, as well as household wealth. The significance level (Sig.) of each regression variable indicates how Mobility, inequality and social exclusion | 71 unlikely it is that there is indeed a systematic connection between the variable and the child mobility rates observed. It is normal to accept uncertainty up till around 10 percent (Sig. = ,1 ), but not more than that. For example, the likelihood that the literacy of the household head plays a role in the probability of child relocation for other purposes than schooling is about 95 percent, since the significance level is at ,052. In other words, the two factors are likely to be connected. The sign in front of the coefficient B shows the direction of the correlation. In the case of household head literacy, the sign of the coefficient (B) is negative, meaning that if the household head is literate, the likelihood of child mobility is reduced. Table 5. Regressing child mobility for other reasons than schooling on flood debt, controlling for socio-demographic and wealth indicators (logit). B S.E. Wald df Sig. Exp(B) Religion dummy: Household head Muslim -,104 ,305 ,117 1 ,732 ,901 Education dummy: Household head literate -,346 ,178 3,784 1 ,052 ,708 Gender dummy: Household head female ,065 ,192 ,114 1 ,735 1,067 Vulnerability dummy: Household head distrust ,332 ,155 4,617 1 ,032 1,394 Wealth proxy: Housing quality (0-3) -,221 ,078 8,100 1 ,004 ,802 Flood debt ,315 ,175 3,218 1 ,073 1,370 Categorical region variable*: North -1,353 ,308 19,284 1 ,000 ,258 Central -,029 ,156 ,034 1 ,855 ,972 Constant -1,734 ,198 76,478 1 ,000 ,177 * South is the residual category of the categorical variable, that is, the results show how North and Central compares to South. At the project validation workshop in Cotonou some concern was raised with regards to using “Muslim� or not as the religion dummy. It should therefore be stressed that the only reason for this choice was that in the initial cross tabulation results, the Muslims differ from all the other religions groups in Benin by having much lower mobility rates. The other religious groups appear to behave quite similar to each other. “Muslim or not� was used in the regression simply to see if the differences reflected in the cross tabulations were confirmed to relate systematically to this faith when other factors were controlled for. Having a Muslim religion does, however, not come out as statistically significant (the “Sig.� level is very high). In plain terms, the regression results indicate that Muslim populations more often have other features that correlate more systematically with child mobility, but their low mobility rate has little to do with religion per se. 72 | Mobility inequality and social exclusion When included in the same regression equation, religion loses explanatory power to these other factors, notably the regional variable. Similarly, more children relocate from female headed households. However, the gender of the household head is not statistically significant in the regression. This suggests that it is not the gender of the head that matters per se, but rather the fact that certain other decisive features are more common in female headed households. For example, there are fewer female headed households in the North, very few are literate, and female household heads more often express distrust (57 percent compared to 49 percent of men). The gender of the household head losses explanatory power to these other factors because they co-vary much more systematically with the outcome studied: the mobility of children. Distrust is measured by the household head’s confirmative response to two standardized survey questions of trust i) you can never be too careful with other people and ii) most people are trying to take advantage of you. Around half the household heads answered confirmatory to both of those two questions and were thus categorized as distrustful. As suggested earlier in this report, the regression shows that children from the households of distrustful and perhaps pessimistic household heads are more likely to have left for other purposes than schooling. Not surprisingly, wealth, measured by housing quality, reduces likelihood of child mobility, and so does the fact that the household is in the Northern region of the country. There is no significant difference between the Center and the South. Most importantly, however, the regression in Table 5 shows that there is a statistically significant correlation between becoming indebted after the 2010 floods and child mobility for other reasons than schooling, also when other related factors are accounted for. Applying the same regression to other flood consequences, this relationship is also confirmed for households whose members had to flee, households that lost buildings or other household infrastructure, and households reporting “other loss of revenue�. On the other hand, the relationship was less systematic in households reporting crop loss and loss of animals and farm fish, and also, perhaps surprisingly, in households reporting illness and malnutrition as consequence of the flood. Table 6 shows the same regression equation used to study the household probability of representing girls who have left to work as domestic servants. The same socio-demographic Mobility, inequality and social exclusion | 73 variables seem to be significant also here. Again, there is a statistically significant relationship between becoming indebted after the 2010 floods and girls leaving to find domestic work elsewhere. This link is also supported for households where animals and farm fish had been lost, while the relationship was less systematic for households that had suffered other consequences from the floods. Table 6. Regressing mobility for domestic work on flood debt, controlling for socio- demographic and wealth indicators (logit). B S.E. Wald df Sig. Exp(B) Religion dummy: Household head Muslim ,058 ,543 ,011 1 ,915 1,060 Education dummy: Household head literate -,542 ,324 2,802 1 ,094 ,582 Gender dummy: Household head female -,284 ,334 ,724 1 ,395 ,753 Vulnerability dummy: Household head distrust ,792 ,262 9,109 1 ,003 2,208 Wealth proxy: Housing quality (0-3) -,310 ,138 5,023 1 ,025 ,733 Flood debt ,569 ,268 4,485 1 ,034 1,766 Categorical region variable*: North -1,523 ,606 6,314 1 ,012 ,218 Central ,288 ,264 1,195 1 ,274 1,334 Constant -2,842 ,337 71,185 1 ,000 ,058 Moving on to the other shocks studied, the regressions strengthen the claim that the loss of a bread winner in the household or the illness of a household member strongly correlate with child mobility. Being exposed to more systemic shock, like drought and flood, has no such observable, systematic impact. This is also the case for girls leaving for domestic work. However, in studying this particular group, it is also clear that the household that had resorted to selling off other assets than land and livestock, and the households that had been forced to reduce the daily food consumption were also households at heightened risk for having girls leaving to do domestic work. It may thus seem as if encouraging children to leave appears in a sequencing of coping means related to risks and shocks. Households appear to first resort to using their savings and, if they have such a buffer, to sell off land and livestock. Child mobility may happen after or in parallel with the selling off of other household assets, borrowing money and reducing meals. 74 | Mobility inequality and social exclusion The regression approach provides various ways to measure at the impacts of shocks and their consequences on mobility rates. Here one example should suffice to demonstrate the strength of the impact on the probabilities of mobility. The second logistic regression (Table 6) found the impact of having to borrow money after the floods of 2012 to be statistically significant. To do the following impact estimates, the values for the control variables are locked at the most common household features in rural Benin: the household is located in the Central region of the country and the household head is an illiterate and distrustful man who is not a Muslim. On the other hand, the values of the variables studied – wealth and borrowing money – will be altered in order to obtain more group specific estimates. If the household is very poor (housing quality is 0 on a scale from 0-3), the initial probability of having at least one girl who has left for domestic work is of 14,7 percent. Imagine that a flood strikes the household with the consequence that it has to borrow money wherever possible. The mobility probability increases by 8,5 percentage points to 23,2 percent. The less poor the household, the lower the impact from the having to borrow money: a household with housing quality score of 1 has a probability increase of 7 percentage points (from 11,2 to 18,2 percent probability), a household with housing score 2 of 5,5 percentage points (from 8,5 to 14 percent likelihood), and finally a household with a high quality household building given score 3 only has an increase in 4,3 percentage points (from 6,3 to 10,6 percent). All the variables interact. If the household head is replaced by one with similar features, but a more trusting attitude towards other people, the impact of the flood debts on girls moving out to do domestic work are smaller; around 5 percentage points for the poorest households and only 2 percentage points in the wealthiest households. If we look at a household in the North, with a Muslim and trusting household head, initial probability of household girls migrating towards domestic work is very low in the first place, and does not change much whether the household is poor or not (1 percentage point for the poorest and 0,4 percentage points for the wealthiest). Figure 37 shows a simulation of how the impact of getting indebted after the 2010 floods varies across the three regions of Benin. The basic household variables were fixed to a male headed, non-Muslim, illiterate and distrustful household head. Region and wealth level (proxied by housing quality) were altered to compare the strength of impact in different sub-groups of households. The figure shows that impacts were strongest in the Central part of the country, but Mobility, inequality and social exclusion | 75 also considerable in the South. While the probability that the poorest households in the South have at least one girl who has left for domestic work is of around 12 percent, this increases to almost 19 percent for the households that had to borrow money after the 2010 Floods. In the North, mobility for domestic work is generally much lower, and although the probability of girls leaving increases if the family gets indebted, this increase is much less pronounced. 3 2 North 1 0 3 2 Central Not indebted 1 Increase 0 3 2 South 1 0 0 5 10 15 20 25 Figure 37. Probability of girls relocating for domestic work: Impact of indebtedness after the 2010 Floods, on region and wealth (assessed by housing quality scores of 0-3). Simulated results from a logit regression. Fixed household features: male, illiterate household head, non-Muslim, distrustful attitude. (In percent of all households) Similar simulations of other shocks and coping efforts that regressions suggested to be related to child mobility and domestic servitude at a statistically significant level show that:  The death of a breadwinner in the fixed household setting presented would produce an increase in the probability that a girl has left to do domestic work of 2,1 percentage 76 | Mobility inequality and social exclusion points in the poorest households in the North, 8,6 percentage points in the poorest household sin the Central region, and 7,2 percentage points in the poorest households in the South.  Poor household that were forced to sell their assets after their last serious shock had an increased probability of having a girl leaving for domestic work of 1,3 percentage points in the North (from 2,4 to 3,7 percent), 6,3 percentage points in the Center (from 14 to 20,3 percent) and 5,5 percent in the South (from 11 to 16,3 percent).  Poor households that responded to their last shock by reducing meals also saw an increase in the likelihood of a girl relocating for domestic work of 1,4 percentage points in the North, 6,1 percentage points in the Central region, 5 percentage points in the South. While there is a higher probability that girls leave households in the Central regions of Benin for domestic work, the Central and Southern parts of the country behave quite similarly with regards to child relocation. Applying the same simulation to the child mobility regression (mobility for other purposes than schooling) the results show that:  The poorest in the Central and Southern parts of the country have a 5,3 percentage point increased likelihood of having a household child living away for other purposes than schooling if they had to borrow money after the 2010 Floods, a 5,8 percentage point increase if household members had to flee, and a 7,6 percent increase if household buildings were lost (probability increase from 19 to 27 percent). In the North the probabilities are low in the first place and the impact only around 2 percentage points.  Idiosyncratic shocks like the death of a breadwinner or illness of a household member had a slightly higher impact on child mobility rates in the South than in the Center, but it was generally of around 5,5 percentage points. Again, the impact in the North was smaller, but so was of course also the child mobility rates in the first place. Mobility, inequality and social exclusion | 77 5 Implications for social protection Weak social protection is a major concern among rural household heads in Benin. Only one-in- four household heads say that they never worried about who will care for them in their old days, and almost 40 percent say they worry a lot. Around half also say they would worry even more if they had fewer children. There is little expectation among the respondents of this survey that anyone outside the nearest family will help and support them when they get old: three-in-five count on their sons to take the main responsibility, and one-in-five on their daughters. So, children play a major role in the social protection of the rural population in Benin. Both their number and qualities matter, and parents are likely to have strong interest in how their children invest their human and social capital: what education they get, what jobs they choose, who they work for, marry or stay with before they marry. The lack of other types of social protection that way places some serious restrictions on the choices of children: they remain what has been labeled the principle social protection tools of the poor. This survey has shown that besides functioning as pension providers for their parents and other relatives, they also play a role in in the way households deal with shocks and their consequences. Sometimes this role may jeopardize their future options and harm their human and social capital. Formal social protection interventions approach inequality and social exclusion by providing opportunities to the poor and safety nets to populations vulnerable to risks and shocks. Social protection program design is at its best when well-targeted, and when it applies a mix of program features that are tailored to the specific local realities. This report has presented new data linking poverty and shocks to the impact they have on child outcomes, with a special focus on the mobility of children in general and of girls into domestic work in particular. Both child mobility and domestic work are strategies with many potential benefits both for the children in question and their households of origin. As suggested in Figure 2 of the first chapter, mobility primarily motivated by social and educational goals is more likely to be well negotiated and to take place without a very high danger of exploitation. Yet, the bargaining powers of the child and his or her household of origin are weakened by household poverty and crisis. Mobility arrangements being organized under such conditions are more likely to become exploitative, as households and children become more willing to accept both poorer relocation conditions and 78 | Mobility inequality and social exclusion higher risks. The destiny of Perpetue Dagba (on the cover of this report) represents a positive outcome to the story of a socially orphaned school drop-out. Her fortune could easily have been different. The qualitative interviews conducted under this project returned numerous stories of girls who had worked to become able to pay for schooling and apprenticeships, only to see their salaries stolen by a relative who was supposed to look after the money, or an employer who never paid. One girl had twice experienced that her father had already visited the employer to claim her hard-earned money – her only hope of investing in an education and a future. This report has indicated a link between different aspects of household poverty and the mobility of children for other purposes than schooling and also of girls and young women for domestic work. However, it also suggests that even the assumingly wealthiest of the households in rural areas have some children leaving without going to school, and even some girls leaving to do domestic work. Not all shocks automatically correlate with mobility, but some shocks, certain consequences of shocks and some of the coping means implemented by the households affected correlate with increased mobility at a statistically significant level. This impact on child mobility is stronger among the poor, but still quite significant among the better-off households. For example, the prototype household defined in the last section of the previous chapter, and assigned a maximum wealth score of 3, shows an increase in child mobility for other purposes than schooling of 3,5 percentage points if it reports revenue loss after a serious shock. This report argues that effort to stop or criminalize all mobility of children and young girls most likely means to throw the baby out with the bathwater. Many forms of child circulation are beneficial to both the child and to the households involved. Indeed, in dire situations the prevention of child mobility away from a household in crisis could get serious consequences for the child. However, time and opportunity to negotiate a good relocation deal is often not in place when children and young women leave as a response to poverty or shock. Despair reduces the bargaining powers of the child or girl in question, and arrangement outcomes will most likely be affected by this. Relocation arrangements previously perceived as unattractive, may seem more acceptable under worsened home conditions. Relocation resulting from poverty and shocks is thus likely to be more risky to the child or young woman involved and the likelihood of exploitation higher. So, this latter type of crisis relocation is probably the type of child mobility that should be addressed by carefully designed interventions. Mobility, inequality and social exclusion | 79 How should that be done? Figure 1 in this report suggested illustrating the process of social exclusion by a downward spiral. Once started, the drift towards despair is hard to redirect, and by each loop down the spiral the process may accelerate and become even harder to reverse. Helping the socially excluded – those who have reached the very bottom part of the spiral – has proven to be difficult. When social networks are damaged and the human habitus is affected by indignities and perhaps ill health, rehabilitation becomes a costly endeavor, often with a low success rate. Individuals who have reached this level of social and personal impairment cannot simply be helped by a micro credit, skills training or support to an income generating activity: their spirit is often low; they have no network to sell their products or services within, and are likely to lose out in the competition with those who do. The spiral model suggests preventive interventions, preferably as high up in the spiral as possible. These interventions should seek to prevent the triggering of the downward loops in the first place, and should especially aim to provide means to prevent the triggering or accelerating consequences of shocks. The earlier a project succeeds to intervene in a negative process, the easier it will be to change its’ direction, and the shorter the way back up to an equilibrium where the spinning slows down. Children and young girls leaving home for risky relocation arrangements may have entered or may be at risk of entering into such downward-spiraling processes. And this is especially so if they are forced to leave by a shock, the consequence of a shock or in relation to the coping strategies applied by their households. In short, safety nets that could prevent children and young girls from having to play the role as social protections tools to their uninsured families would be an investment to protect human capital and contribute to reduce inequality in opportunity. Indeed, carefully designed safety nets could help prevent humanly damaging exploitation. How can the data collected under this research project help to make such safety net design more adequate to protect children and young women at risk in rural areas of Benin? Three issues will be briefly discussed here; existing informal social protection arrangements and their relevance, possible criteria for targeting and issues related to conditionality of safety-net support. 80 | Mobility inequality and social exclusion Effectiveness of existing, informal safety nets The rural poor in Benin have few other options than relying on family and social networks for opportunities and support. However, many are involved in traditional mutuality arrangements – “tontines� and “mutuels� – that may provide some support in the case of an emergency or a health crisis. Some places, micro finance projects involving savings groups and micro credits are in place, and cash is sometimes handed out to support income generating activities. Finally, people may try to build up a buffer by saving, in cash, land and livestock. The previous analyses of this report show that the impact of the 2010 floods on child outcomes was smaller in affected households reporting to have sold livestock or land to cope with the shock. One can assume that household with livestock and land to sell are already among the most economically sound of the rural households, since they had had opportunity to invest in such capital objects in the first place. But how do other household efforts to mitigate risk affect the mobility of girls for domestic work? Questions were asked about family access to or participation in various types of savings and mutual arrangements. Cross tabulations show some substantial differences in mobility for girls between households with and without access to such initiatives. For example, households with cash savings have half the relocation rate of households without, while there is little difference between household who report to be saving in animals. Similarly, people reporting to have access to a savings program or who participate in a program of income generating activities (IGAs) have lower rates of girls leaving for domestic work than those who do not, while the same difference is not apparent for those who report to have access to crisis credits. Also, participation in Mutuels or Tontines makes little difference, while the relatively few reporting to be part of a health mutual have significantly lower relocation rates. Finally, those who benefit from regular support from a migrant have a somewhat lower mobility rate, and those who say they can count on support from migrants and family in case of a crisis a markedly lower one. It is difficult to say, however, if all these correlations are indeed causal or whether there are other background factors involved. Having savings, access to savings programs or IGAs is of course an indicator of not being desperately poor in the first place. Saving in animals may primarily be a strategy for farmers. Introducing the various informal mitigation tools listed into the standard regression used in the previous chapter, the only means robust to the correction of Mobility, inequality and social exclusion | 81 other background variables is being able to count on the support of migrants and family in case of a crisis. Around 30 percent of household heads say that in case of a crisis, they can count on the support from migrants and other family members. This group also has a lower rate of relocation for domestic work than those who do not feel they can count on such support, and the difference is statistically significant. The regression already has a control for trust and positive expectations vis-a-vis other people. The trust that others will support in times of crisis – the mere faith that things will work out – seems very important. In fact, a household in Central Benin that scores high on trust and also states they have confidence that they will get support in times of crisis only has a rate of relocation of girls for domestic work of around 5 percent, in spite of being placed in the lowest possible poverty category. If the head does not believe they can count on support in times of crisis, the rate increases to 9 percentage points, and when distrust is added to the equation to almost 19 percent. The wealthiest and most optimistic household heads have a mobility probability of 2 percent. However, by removing trust in crisis support the rate increases to 4 percent, and adding distrust to 8 percent. Trust in the help of migrants and other family in case of crisis can be interpreted as an indicator of faith. However, since the regression controls for general trust levels, it should also be read as relatively good sign of family network quality. A final remark of operational importance relates to the predictability of support. In previous studies, even small transfers have proven to improve child outcomes substantially if the transfers are predictable. There is considerable risk related to the uncertainty of not knowing if a transfer will come or someone will be there to help in times of crisis. In the shadow of this uncertainty, people tend to make safe, low-return choices, and implement mitigation strategies that are less than optimal to those involved. The fact that people who felt safe in relying on outside support had so much lower mobility rates of girls may partly be that their safety nets had proven effective in the past, but also partly that they did not feel they had to play safe in their current choices since support was at hand in case they failed. The fact that food insecurity seemed much more predictive of mobility than poverty adds to the argument that predictability is crucial. Transparent, clear and reliable social protection programs could contribute to further relieve uncertainty and provide predictable support to the vulnerable. 82 | Mobility inequality and social exclusion Targeting A safety net intervention concerned with curbing high-risk mobility of children and young women should in principle target both the poor and the vulnerable. Not only poor households increase child and female out-flux during hard times. Even better-off rural household that would probably fail to meet traditional asset-based targeting criteria are, according to the data, likely to respond to shocks with more child mobility for other purposes than schooling and increased mobility of girls towards domestic work. They are thus in many cases vulnerable in spite of assets and solid household buildings. Households with savings in land and animals are probably less vulnerable than those without, as this seems to constitute a buffer against the worst impacts of a shock. Households with high quality networks also seem less vulnerable, but network quality is of course difficult to assess and use as program targeting criteria. Among other possible targeting criteria it seems that female-headed households behave similarly to male-headed ones, and the analysis shown in this study does not provide special arguments for targeting female headed households because women behave differently. However, female headship correlates negatively with housing-based poverty, but positively with asset-based poverty. That is, households led by females are asset poor, relatively speaking. Asset-based poverty is often accepted as a stronger indicator of economic vulnerability and lowered resilience than housing quality. In this case, female headship could work as an additional targeting criterion to a poverty assessment since it relatively well predicts asset-based poverty. A common challenge in both cash transfer and public works programs is to target the right household member with the transfer. Men and women traditionally shoulder different economic responsibilities in a household. Who gets to control the benefit from a cash transfer or a jobs program may therefor matter to child outcomes. The survey asked both men and women to tell who they felt shouldered the main responsibility for different types of expenditures in the household. For the comparison to make sense, only male headed households with eligible mothers will be compared here. In female headed households there will normally be less of an issue who the beneficiary of a social benefit should be. In total, there were 2 325 mothers living in a male headed household in the sample. Figure 38 shows men and women’s’ account of who takes responsibility for some main groups of expenditures in their households. A clear trend is that relatively few both men and women say Mobility, inequality and social exclusion | 83 that women have the sole responsibility for any group of expenditures. With a few exceptions they say that it is either the man or both who are responsible. Answers vary slightly between men and women: not surprisingly perhaps, men more often claim they take responsibility while women more often will say that the couple shares the responsibility. For example, 27 percent of the men claim they take sole responsibility for costs related to child health, while only 18 percent of the women agree that men indeed do. Instead 31 percent of women say this is primarily their responsibility, while only 24 percent of male household heads agree that this responsibility is left to the women. 100 % 12 10 90 % 2 33 7 80 % 36 39 30 29 42 51 49 70 % 57 55 2 3 1 60 % 5 50 % 88 40 % 24 56 81 15 48 31 61 21 62 64 30 % 52 20 % 27 30 10 % 18 22 10 13 0% Women Men Women Men Women Men Women Men Women Men Women Men Child health Children's Basic foods Sauce School fees School books clothing and uniforms Men Women Both Others Figure 38. Men and women’s account of who shoulders main responsibility for different groups of household expenditures (in percent of male-headed households). A quick conclusion to be drawn from the figure is that if a project’s main aim is to assure children’s schooling, then targeting men with transfers or jobs would probably be all right, since paying for schooling is something men generally see as their obligation. Expenses to children’s clothing are mainly the responsibility of men, although around 40 percent say the responsibility 84 | Mobility inequality and social exclusion is shared between men and women. Providing basic foods (grains), is often a shared responsibility, although women claim that they have this sole responsibility in 21 percent of cases, and that men have it in 22 percent of cases. Sauce (“condiment�), or what is served on top of the basic foods, is more often the responsibility of women. Fifty-six percent of women and 48 percent of men agree that this is mainly a female responsibility. In only around 10 percent of households say men have the main responsibility for this. School related expenses are predominantly left to men, although the provision of school books and uniforms are more often a shared responsibility. 100 % 5 1 14 14 15 90 % 29 34 29 1 80 % 39 37 9 9 46 30 51 56 54 55 50 70 % 3 59 62 60 % 6 3 50 % 94 4 21 84 40 % 12 45 57 65 77 76 5 68 22 30 % 34 58 33 55 27 44 20 % 33 34 34 10 % 22 18 14 10 11 9 0% 6 Center Center Center Center Center Center South South South South South South North North North North North North Child health Children's Basic foods Sauce School fees School books clothing and uniforms Men Women Both Others Figure 39. Women’s account of who shoulders main responsibility for groups of expenditures in their households, on region (in percent of male-headed households). Figure 39 reveals some relatively large differences in the distribution of economic responsibilities between the three regions of Benin. Disagreement between men and women seems to be similar across the country, with a tendency to men claiming to take more responsibility than what women would agree on. To simplify an already complex figure, only Mobility, inequality and social exclusion | 85 women’s reporting is cited in Figure 39, although the gender differences in reporting demonstrated in Figure 38 should be kept in mind. The overall impression from Figure 39 is that the further South in the country, the less men shoulder economic responsibilities. More often women say that these responsibilities are shared by the couple or that they are placed on women alone. While men more often than women have the sole responsibility for child health in the North, women have this responsibility three times as often in households in the Center and the South. Men have sole responsibility for child clothing in in two-in-three households in the North, but only in one-in-three in the South. With regards to both basic foods and sauce (“condiment�), both are more often the responsibility of women in the South, where men rarely has this responsibility for food provision alone. In the Center, men and women equally often have responsibility for basic foods, while this responsibility is shared in slightly more than half the households. Sauce is predominantly the responsibility of women in the Center. Finally, men are consistently assigned main responsibility for school expenditures across the country. However, the provision of school books and uniforms more often becomes a shared responsibility as one move towards the South. While 84 percent of men in the North have this responsibility alone, only 44 percent of men in the South take this responsibility. Social protection interventions primarily target the poor and the vulnerable. Does the distribution of economic responsibility on gender differ among the poor? The difference between asset poor and asset rich households are small and inconsistent in most expenditure groups, although it may seem as if men in better-off households take slightly more responsibility for foods, children’s clothing and schoolbooks and uniforms, while women report to have more of these responsibilities in the poorer households. A perhaps more vulnerability related correlation is however found between men’s share of responsibilities and the level of worry of household heads. In short, the more male household heads worry about the food situation of the family, the less responsibility men take for household expenditures, and the more responsibilities are characterized as shared. If level of worry is interpreted to reflect experienced vulnerability to risks and shocks, then vulnerability correlates with a weakened position of men as providers in the households. One could expect this to also 86 | Mobility inequality and social exclusion reflect the fact that men in the North worry less and also take more economic responsibility, but the trend is clear and robust also at the level of each of the three regions. Figure 40 shows how this plays out on the six expenditure categories examined in the previous two tables. While almost 29 percent of the least vulnerable male household heads take responsibility for child health, only 12 percent of the most vulnerable do. The same numbers correspond to male responsibility for basic foods. While 92 percent of men in presumably less vulnerable households take responsibility for school fees, only 72 percent of men in presumably more vulnerable households do. Similarly, 72 percent of men in the least vulnerable households take responsibility for providing their children with school books and uniforms while the number for the most vulnerable households is only 57 percent. 100 % 4 4 13 15 90 % 27 5 80 % 34 35 35 18 40 42 11 33 35 52 47 50 2 70 % 57 57 61 3 2 60 % 4 4 50 % 6 31 59 92 40 % 20 53 81 58 74 72 30 % 31 62 19 61 30 53 25 57 20 % 46 29 29 24 10 % 17 12 14 14 12 0% 7 A lot A lot A lot A lot A lot A lot Never Never Never Never Never Never Sometimes Sometimes Sometimes Sometimes Sometimes Sometimes Child health Children's Basic foods Sauce School fees School books clothing and uniforms Men Women Both Others Figure 40. Women’s account of who shoulders main responsibil ity for groups of expenditures, on how much the household head worries about being able to provide food for his family during the next 12 months (never worries, worries sometimes, worries a lot). Mobility, inequality and social exclusion | 87 What does all this mean for targeting? From Figure 38 it may seem that if a safety net program aims to secure the schooling of children, then targeting men with the benefit would be relatively benign, as men appear to see it as their responsibility to pay for schooling and school-related expenditures. If securing child health and nutrition is also considered by the program designers, then the targeting of women should be discussed. Figure 39 argues for a stronger targeting of women in the Central and Southern areas of the country, since it seems more common that they take responsibility for child related expenditures in those regions. Figure 40 also cautions that economic responsibilities more often are shared in vulnerable households, and suggests that both men and women could be eligible for targeting in vulnerable populations and areas. Yet, although schooling is traditionally assumed to belong to the responsibility sphere of men, children will not necessarily fare worse in households where women take on this responsibility. To provide some perspective to this issue, distribution of economic responsibilities in the household will also be seen in relation to child schooling and mobility. To look at schooling first, an individual child file is needed, rather than a household file. A file comprising all children 6 - 14 years-of-age is used to compare the distribution of economic responsibilities among the parents of school children and children who are not in school. Overall, 65 percent of the children in the age group are currently in school, 28 percent have never been to school while the remaining 7 percent have dropped out. Figure 41 shows the school attendance rates of children 6 – 14 years-of-age in relation to whether the mother, the father or both share the responsibility for household expenditures. The differences are noticeable, and the trend is quite clear: in households where women shoulder the child-related expenditures, school-participation rates are higher. In households where women shoulder the main responsibility for children’s health, school-participation rates increase from 54 percent to 72 percent – by 18 percentage points. In households where women are responsible for schooling, uniforms and books, school attendance rates are 13 percentage points higher than in the households where men have this responsibility. Looking at sub-samples of boys and girls it is (perhaps surprisingly) clear that women’s economic responsibility for schooling benefits boys slightly more than girls. While the relationship shown may be related to other background factors, it is important to recall that the gender distribution of child-related expenditures correlated little with asset or housing based wealth indicators. 88 | Mobility inequality and social exclusion 59 Books and uniforms 72 71 61 School fees 74 71 54 Sauce 66 Men 66 58 Women Basic foods 65 68 Both 60 Children's clothing 71 70 54 Child health 72 65 0 10 20 30 40 50 60 70 80 Figure 41. Household heads’ account of who shoulders main responsibility for expenditure groups and the schooling status of children 6-14 years-of-age (in percent of households). Figure 42 presents the results of a similar exercise focusing on the mobility of children 0-17 for other purposes than schooling. Interestingly, the clear trend here is that child mobility increases significantly in households where women shoulder the main child related expenditures. Mobility rates are almost twice-as-high from households where women shoulder the main responsibility for school-related expenditures (although this is a relatively small group in the sample), and also very high from households where women pay for children’s clothing. 8 Books and uniforms 15 15 9 School fees 17 18 7 Sauce 13 Men 10 7 Women Basic foods 14 12 Both 8 Children's clothing 17 13 8 Child health 13 11 0 2 4 6 8 10 12 14 16 18 20 Figure 42. Household heads’ account of who shoulders main responsibility for child related expenditure groups and the mobility of children 0-17 years-of-age for other purposes than schooling (in percent of households). Mobility, inequality and social exclusion | 89 Finally, how does the gender distribution of household expenditures affect the departure of girls for domestic work? In previous analyses of domestic work as response to shocks, only girls who had left recently were studied to grasp the likely sequencing of events. This time, focus is on household girls who have at one stage been away to work as a domestic, since the responsibility sharing of household expenditures are likely to reflect more permanent household features than what shocks would constitute. Figure 43 shows the same substantial differences between household where men shoulder child related expenditures, and households where this responsibility lies with women. Importantly, in households where women shoulder more of the costs related to children’s health and clothing, girls mobility for domestic work soar. 9 Books and uniforms 18 14 10 School fees 21 14 12 Sauce 14 Men 8 12 Women Basic foods 13 11 Both 9 Children's clothing 19 12 9 Child health 15 10 0 5 10 15 20 25 Figure 43. Household heads’ account of who shoulders main responsibility for child related expenditure groups and the mobility of girls (6-19 years-of-age) for domestic work (in percent of households). These results should be valuable to understand the consequences of targeting women or men with safety net benefits. Yet, the exact dynamics are not clear. Women take more responsibility is households categorized as vulnerable, and this may help explain why the mobility rates are also higher. It would not explain, however, why households where men take responsibility for major child related expenditures see such a reduction in school-attendance rates. For that a more sophisticated analysis of the data is needed. 90 | Mobility inequality and social exclusion Conditionality Social transfers are in many programs contingent on certain conditions being met by the recipients. Such conditionalities serve political, practical and socioeconomic purposes. On the political side, the recipients prove themselves committed and worthy of the support. Also, conditionality can serve as a screening mechanism: program inclusion errors are reduced if benefits are only made attractive to those who are willing to accept the conditionalities imposed. On the socioeconomic side, conditionalities induce recipients to contribute to meet certain development goals. Conditionalities can be for the recipient households to provide birth certificates for their children, take them to regular health checks or to send them to school in cases where they would normally not have prioritized their time or resources that way. Another approach to conditionality in transfers is public works programs. The labor demanded effectively works as the condition for receiving the transfer. Recipients prove themselves worthy of the transfers by providing their labor to a publicly prioritized project. The screening function is clear: as salary for public works programs are normally consistent with established minimum wages, the program will be unattractive to anyone with better options. When public works programs are used as social safety-net policy they commonly include labor exemptions for single providers with care-taking responsibilities or ill health, and other groups that clearly qualify for social support but who for various reasons are not able to provide labor to the program. A social support program conditioning on labor may or may not also include other types of targeting or conditionalities. In the child-labor literature, compulsory schooling is considered a more effective policy than child labor prohibition, since the former is easier and less costly to monitor. Child labor and child-mobility restrictions can in similarly in principle be conditioned on to a social transfer, and have indeed been piloted as such conditions in cash-transfer programs. Conditioning on school attendance is however much easier to monitor and thus a more cost-effective path to a relatively similar outcome. The data presented this far show that child mobility for other purposes than schooling (and for domestic work in particular) increases by poverty. It may thus intuitively seem as if conditioning on schooling should be unnecessary. When transfers are received parents can afford to keep their children at home and in school. Yet, both mobility and taking children out of school seem to be Mobility, inequality and social exclusion | 91 responses to shocks in both wealthy and poor households. While only 6 percent of household heads say they responded to the last serious shock to the household by 2010 by taking children out of school, as many as 10 percent of parents in the Southern region gave this response. Yet, households considered wealthy by asset count or housing quality, also report both child mobility for other purposes than schooling and of girls for domestic work. This is not in itself an indication that also the better off households should be targeted for social transfers, since one should assume that the relocation conditions of children from these households under normal conditions may be less harmful than those triggered by poverty. Yet, even in these households, the types of mobility that are triggered by shocks should be considered risky. Would conditioning school attendance on to a social transfer be effective to curb such risky mobility? At the national level, 8 percent of the poorest households (no solid household building construction parts) said they had taken children out of school as a response to shock, while 4,5 percent of those in the wealthiest households (all solid building parts) did. Taking children out of school as a response to shock also correlated strongly with having children leaving for other reasons than schooling. While 11 percent of households had a child living away from home for reasons other than schooling, the share is as high as 26 percent for households responding that they had to take children out of school in order to cope with the last serious shock to the household. In the Central region of Benin, the share is as high as 33 percent. While 5 percent of all households had a girl who last left to do domestic work, this share was 15 percent among households that had responded to its’ last shock by withdrawing children from school and almost 22 percent in this group in the Central part of the country. The link between schooling and child mobility should therefore be established. Must families be induced to send their children to school? Figure 44 shows that rural parents in Benin are overwhelmingly positive to sending children to school. In total, 88 percent say that if there were no costs involved they would send all their children to school (in comparison, only 50 percent of rural parents in Senegal responded that way to the same question). This share is as high as 98 percent in the South and 95 percent in the Center, although only 74 percent in the North. The poor generally have lower ambitions. Only 85 percent of households with low building standard say they would send all their children to school even if there were no expense, while 94 percent of those with the highest housing standards said they would. Yet, most of this 92 | Mobility inequality and social exclusion variation is found in the North. In the Center and South the attitudes to schooling are positive irrespective of poverty level. On a national basis, there is no difference between household where one child is away for other purposes than schooling and other households. At the regional level trends are unsystematic. In the North it seems that households with one child away for other purposes than schooling are more positive to schooling, while in the Center and South these households less often are. These findings do not in clear words point towards attitudes to schooling as a particular trigger of mobility for other purposes than schooling. 100 % 90 % 80 % 70 % 60 % 50 % 94 95 96 94 99 97 98 97 40 % 82 89 66 70 30 % 20 % 10 % 0% 0 1 2 3 0 1 2 3 0 1 2 3 North Center South All Some Figure 44. If there were no expenses related to schooling, how many children would you send (“All� or “Some�)? On poverty (based on housing quality scores 0-3) and region (in percent of all households with children 6-14 years of age). Conditioning on schooling to safety-net projects assumes the actual presence of accessible schools of an acceptable quality, and with an adequate capacity. This is not always the situation in rural Benin. How does the availability of such schools affect child mobility and school attendance rates? Directly asked, 25 percent of respondents say they find accessibility to the closest primary school unsatisfactory. Twenty-two percent consider the capacity in the nearest primary school to be poor and 19 percent say that the school quality is unsatisfactory. One-in- four say the nearest primary school is more than 1 km away, 10 percent more than 2 km. Mobility, inequality and social exclusion | 93 Distances are higher to secondary school. Only one-in-four have a secondary school that is less than two km away, but three-in-four have a secondary school within a 5 km distance. Satisfaction with school access, capacity and quality correlates with both mobility for schooling elsewhere and mobility for other purposes, although not very strongly. Households with children away for other reasons than schooling are generally less happy with the accessibility of primary schools and also with school quality. School capacity does not seem to correlate with relocation probability. Recommendations This report argues that child mobility and the mobility of girls from rural areas for domestics work should be differentiated into risky types of mobility and mobility that is likely to be benign or even beneficial to the child or girl involved. Mobility is likely to be more risky when poverty or shock is an important push factor for the child or young woman in question. Interventions should therefore focus at the type of mobility that results from poverty and shock, rather than prohibiting mobility more indiscriminately. Because the rehabilitation and successful reintegration of young people who have suffered the down-sides of bad, exploitative mobility arrangements can be extremely difficult and also costly, focus should be placed on preventive interventions. Social protection interventions aim to provide opportunities to the poor and safety nets to the vulnerable. Child mobility is higher among the poor and increases in all groups when shocks have certain consequences to rural household. While public-works projects tend to be self- targeting, other types of safety net provisions should consider targeting the asset poor as well as families exposed to idiosyncratic shocks, and geographical areas exposed to systemic ones. Female-headed households do not behave very differently from male headed ones, but female headship could still be considered a simple targeting proxy since female-headed households tend to be asset poor. Men typically control the benefits from safety net interventions when no specific project design feature is introduced to reach women. This may in theory have implications for child related household consumption since men and women traditionally feel responsible for different types 94 | Mobility inequality and social exclusion of expenditures in a household. The Benin data show that men in most cases take responsibility for costs related to schooling, although this responsibility is more often shared in the Southern parts of the country. If keeping children in school is an explicit project goal, the data provides no strong argument for targeting women in particular. Child health and nutrition, on the other hand, is more often the responsibility of women. Targeting women could thus potentially more easily translate transfers into child nutrition and health. When this is said, children go more to school in households where women shoulder more of the child-related expenditures. In principle this means that women would not necessarily be poorer providers of schooling than men, if they were given control over the transfer benefits. The downside may be that in households where women have more financial responsibility, there is also more child mobility, especially of girls for domestic work. Should schooling be conditioned on to social transfers in rural Benin? Although preventing child labor and high-risk mobility may be an objective in Benin, it would be administratively more feasible to condition on schooling than to condition on commitments to refrain from child labor and mobility. However, an overwhelming majority of parents say that if there were no costs related to schooling, they would have sent all their children there. Conditioning a transfer on schooling might thus seem superfluous. Moreover, such conditionalities would require improvements in the schooling infrastructure, since school capacity would be needed to implement the condition, and both school access and quality concerns are drivers for mobility. Currently, the rural populations manage by trying to save, invest in buffers in the form of land and livestock, involve in “tontines� and “mutuels�, and try to access the savings, micro credit and IGA projects that are available to them. Quite few of these efforts seem to reduce child mobility in any substantial way, with a few exceptions. Households involved in savings groups, IGAs and those who say they can rely on the support of migrants and relatives in case of a crisis have less child mobility for other reasons than schooling, and less mobility of girls for domestic work. This may indicate that formalized interventions and not at least arrangements providing some sort of predictability may be effective to avoid the use of child mobility as a coping strategy. Households that had to borrow money as a consequence of the 2010 floods seemed at particular risk of child mobility for other purposes than schooling and for girls’ mobility for domestic work. Poor lending conditions could have something to do with their despair. Mobility, inequality and social exclusion | 95 6 Final words Child relocation serves many purposes in the rural populations of Benin. The struggle to prevent circulation of children by judicial means has been largely unsuccessful, clearly because some of these purposes have been poorly understood and acknowledged. Carefully designed social protection interventions could primarily help prevent the type of child mobility that bears a high risk of becoming exploitative. Such interventions may take the shape of opportunity provision to the poor and social safety nets to vulnerable populations. The mobility of girls and young women for domestic work epitomizes the challenge of differentiation between good and bad relocation arrangements for children and youth. Some girls move to stay with relatives for educational or social purposes, and help out with housework while there. This type of arrangements may be beneficial to both the social integration and economic possibilities of the girl, as well as her current wellbeing. However, the more exploitative segments of the market for domestic services are clearly in need of reform. The ILO defines domestic work as a profession, and domestic workers deserve the same rights and regulations as workers of other professions. The qualitative interviews carried out under this project revealed that the young domestic workers in Benin are often not paid, and when they are, their trusted relatives many times run off with their salary. The survey also showed that 60 percent of the respondents find it acceptable or acceptable on conditions to corporally punish a domestic worker. This tells a lot about the perception of domestic work and of the status of the girls who do it. Efforts to prevent young girls from leaving for domestic work as a response to shocks and poverty could be one way to approach the most exploitative forms of the sector, as these are the girls with the poorest bargaining power in the market. And to prevent them from entering the market of domestic work under such unfavorable conditions, they and their families must be offered alternative coping options. Can Beninese households manage with less domestic support? Or higher priced adult domestic workers? It would most likely be difficult in the short term, given the low level of mechanization in the relatively poor, urban households that employ the lowest paid domestic workers today. Importantly, it would increase the labor burden of relatively poor women who are themselves struggling to cope. In the introduction to this report, it was pointed out how such a scenario 96 | Mobility inequality and social exclusion could harm women’s income-earning opportunities outside the household, and also affect their social status, since they would now have to do low status jobs traditionally assigned to servants and children. Yet, the costs for many of the children and young women involved in domestic services in Benin are such that a change must be made – it is needed to promote more equitable chances for vulnerable rural girls and young women. Are there good solutions to this apparent conflict of interest between two relatively vulnerable groups in the Beninese society? Other West African countries have proven that there are. If the supply of cheap and legal domestic assistance is reduced, costs of such services will perhaps rise and attract less vulnerable adult workers to the profession. The main sacrifice for the users will be that they will have to organize their time better: having a live-in maid who is on duty 24/7 is not a sustainable labor arrangement in a modern, rights-based democracy. The development in other West African cities goes towards buying the domestic services from adult, professional maids who have their own place to stay outside the employer’s household, decent salaries and livable work hours. They work for several employers and are therefore less vulnerable to the whims of just one. Because the costs of such services are higher, a household cannot count on having a maid at hand at all times. Households may buy domestic help at certain times a day or week – it is basically all a question of organization. While this may seem hard right away, it already works well elsewhere – in fact, Benin is among the last countries where the system of live-in domestic girls is still so institutionalized. Mobility, inequality and social exclusion | 97 Annex I : Cluster sample Strata North N° Ménages N° ZD Commune Arrondissement grappe 2002 VILLAGES BIRNI-LAFIA 45 96 30 Birni Lafia BIRNI-LAFIA 48 97 294 Karigui Karimama BOGO-BOGO 23 98 47 Mamassi Gourma KOMPA 14 99 144 Kompa et Kompanti GAROU 79 89 117 Goungoun GAROU 86 90 91 Torro Zoungou TOUMBOUTOU 64 91 118 Dégué-Dégué et Toumboutou Malanville TOUMBOUTOU 73 92 133 Sakawon-Tégui centre MALANVILLE 12 93 371 Galiel MALANVILLE 19 94 253 Tassi Tédji MALANVILLE 20 95 327 Tassi Tédji FOUNOUGO 66 77 94 Founougo A Founougo Peulh FOUNOUGO 68 78 129 Founougo A et Kpako FOUNOUGO 69 79 104 Founougo B FOUNOUGO 70 80 100 Gningnimpogou Founougo Bougnirou Bariba FOUNOUGO 73 81 90 Founougo B Banikoara FOUNOUGO 75 82 115 Founougo A et B Founougo Peulh FOUNOUGO 76 83 154 Founougo B GOUMORI 146 84 131 Gbabgbanga, Goumori Peulh KOKEY 44 85 62 Kokey Nimbéré Peulh SOMPEREKOU 6 86 131 Sompérékou B SOROKO 105 87 110 Soroko Peulh Gbéniki TOURA 127 88 81 Tintinmou Peulh et Bariba et Toura Peulh 98 | Mobility inequality and social exclusion TAMPEGRE 20 104 89 Kokota, Tantougou Toucoun TOUCOUNTOUNA 23 105 169 Toucountouna I touna TOUCOUNTOUNA 24 106 132 Toucountouna I TOUCOUNTOUNA 28 107 140 Boribansifa KOUNTORI 43 108 126 Tarpingou Cobly COBLY 10 109 146 Cobly FIROU 24 110 127 Sokongourou Kérou KOABAGOU 30 111 101 Kaobagou III BADJOUDE 2 102 131 Kakpala Ouake SEMERE II 42 103 89 Kpakpalaré GNINAGOUROU 38 100 37 Guinangourou Peulh et Guinaguirou Perere SONTOU 17 101 88 Allafiarou et Bani Peulh Strata : Centre Ménages N° ZD N° grappe Commune Arrondissement 2002 VILLAGES AYOMI 45 144 211 Zaffi II AYOMI 48 145 111 Tota DEVE 56 146 82 Houédjamey Dogbo HONTON 65 147 255 Dékandji HONTON 68 148 77 Dékandji DOGBO 14 149 85 Véhédji et Hounsa, Lokogohoué DOGBO 21 150 92 Sègba, Touléhoudji et Houndjromè AHOGBEYA 41 124 185 Kpalakatagon AYA-HOHOUE 50 125 151 Kédji Aya-Honhoué Avèkandji Klouékanme DJOTTO 62 126 108 Fidégnonhoué DJOTTO 69 127 132 Yénawa et Nigbo DJOTTO 70 128 123 Yénawa et Nigbo Mobility, inequality and social exclusion | 99 HONDJI 83 129 128 Soglonouhoué TCHIKPE 95 130 147 Sokpamè TCHIKPE 97 131 97 Sokpamè et Kpakpassa TCHIKPE 98 132 155 Sokpamè TCHIKPE 99 133 132 Gnantchimè TCHIKPE 102 134 98 Zounzonkanmè, Akouégbadja Zounzonkanmè, Akouégbadja, TCHIKPE 104 135 145 Agbago TCHIKPE 105 136 130 Agbago ADOUKANDJI 19 137 132 Ahouada GNIZOUNME 41 138 160 Assogbahoué HLASSAME 56 139 88 Wèwèhoué Lalo HLASSAME 61 140 161 Adjaglimey LOKOGBA 66 141 113 Zoundjamè TOHOU 85 142 111 Zoundotan, Bayékpa ZALLI 97 143 85 Adjassagon, Kadébou et Kohomè HOUEDOGLI 49 14 71 Affomandji Kpakouhoué, Tchakada HOUEDOGLI 52 15 66 Tchakada et Ablogangan HOUEDOGLI 53 16 90 Tchakada et Ablogangan Toviklin HOUEDOGLI 57 17 105 Agbozohoundji II TANNOU-GOLA 73 18 78 Dohodji TOVIKLIN 6 19 150 Toviklin I Djigangnonhoun ADJINTIMEY 22 20 108 Gbotohoué ADJINTIMEY 25 21 102 Doumanhou et Gbotohoué BETOUMEY 48 22 111 Ablomè Titongon Djakotomey GOHOMEY 59 23 120 Gohomey et Lokoatoui Wannou, Gamèfodé Gamèhougbo et HOUEGAMEY 65 24 135 Houégamey SOKOUHOUE 104 25 122 Avonnouhoué , Zounzounvou, 100 | Mobility inequality and social exclusion Hounkèmey ALLAHE 2 54 193 Hè-Hounli et Alahè Za-Kpota ZA-KPOTA 73 55 212 Za-Agbogbomè AVLAME 28 56 130 Avlamè KOUSSOUKPA 17 57 142 Koussoukpa Zogbodomey MASSI 68 58 132 Hon ZOGBODOMEY 39 59 163 Ahoundomé BANAME 35 51 112 Gbatè-Zounmè, Zingo Zagnanado DOVI-CENTRE 23 52 203 Lègbado ZAGNANADO 19 53 294 Zagnanaado et Zoungoudo SIWE-LEGO 43 60 295 Siwé-Lègo Agbangnizoun TANVE 6 61 153 Towéta ZOUNGODO 55 62 191 Agongo et Kanzoun AGBANGNIZOUN 32 63 231 Agbangnizoun, Avalè, Azankpanto Bonou ATCHONSA 1 50 192 Gboa AKLANKPA 88 112 181 Sowignandji AKLANKPA 91 113 167 Allawenonssa I GOME 38 114 150 Tankossi, Gome KPAKPAZA 49 115 82 Sowe I Glazoue MAGOUMI 55 116 107 Hai SOKPONTA 36 117 147 Sokponta THIO 8 118 176 Hoko GLAZOUE 17 119 175 Ayedero ATOCOLIGBE 15 120 201 Malomi ATOCOLIGBE 16 121 161 Atokobè Bante GOUKA 2 122 158 Gouka KOKO 70 123 229 Issalè Mobility, inequality and social exclusion | 101 Strata : South N° Ménages N° grappe Commune Arrondissement ZD 2002 VILLAGES AHOMEY-LOKPO 17 64 183 Ahomey-Lokpa centre AHOMEY-LOKPO 20 65 200 Hounme So-Ava GANVIE I 71 66 170 Kpassikomey, Tohokomey GANVIE II 61 67 191 Guédévié SO-AVA 5 68 279 Ahomey-Glon AVAGBODJI 3 69 91 Bè'bè I Aguegues ZOUNGAME 41 70 213 Djigbécomè DE DEKIN 39 38 100 Togbohounsou GBEKO 7 39 158 Gbéko GBEKO 14 40 119 Sèho-Djigbé GBEKO 15 41 126 Gbèssoumè HOZIN 67 42 139 Hondji HOZIN 69 43 141 Akpamè, Tokpa-Koudjota Dangbo HOZIN 70 44 158 Akpamè HOZIN 71 45 149 Djigbé HOZIN 74 46 137 Hondji et Akpamè HOZIN 76 47 98 Akpamè HOZIN 77 48 75 Akpamè ZOUNGUE 90 49 127 Zounguè AWONOU 48 26 116 Siliko Akouéhan-Tohoué Klogbomè, Todé, AZOWILISSE 74 27 150 Houéda Gbéda Adjohoun AZOWILISSE 79 28 146 Gbékandji I Klogbomè,Gbéda AZOWILISSE 84 29 155 Gbékandji II GANGBAN 15 30 118 Gogbo KODE 31 31 166 Gounké 102 | Mobility inequality and social exclusion TAKON 59 32 129 Dra TAKON 62 33 53 Takon TAKON 63 34 118 Takon Sakete TAKON 67 35 88 Gbagla-Nangnon SAKETE 1ERE 51 36 180 Dagbao, Igbo-Eyè et Sodji SAKETE 2EME 83 37 140 Adanrégoun ADJAN 27 71 175 Adjan centre DODJI-BATA 23 72 231 Gonfandji, Fongbo Ze KOUNDOKPOE 52 73 151 Koundokpoé centre, Wédjamè KOUNDOKPOE 58 74 112 Houéounta-Tozounkpa, Aifa ZE 42 75 177 Kpali Allada ATTOGON 9 76 309 Niawloui I et II DEDEKPOE 4 7 139 Abloganmè Athieme ATHIEME 44 8 231 Zounhoué-Kpakpassa ADJAHA 11 9 290 Todjonoukoui Grand-Popo GBEHOUE 33 10 243 Adimadosokpon GRAND-POPO 45 11 47 Agoninkanmè, Houndjohoudji, Hèvè DOUTOU 38 12 34 Dotou Houeyogbe DOUTOU 55 13 57 Ahouloumè BADAZOUI 70 1 113 Hounhoui LOBOGO 9 2 204 Lobogo (Gbétokomè) LOBOGO 12 3 380 Lobogo (Gbétokomè) Bopa LOBOGO 17 4 256 Dhodho YEGODOE 74 5 129 Djèkignan et Tékozoun BOPA 31 6 193 Massè Dansstigo, Tokpoé Mobility, inequality and social exclusion | 103 Annex I : Participants at the Validation Workshop in Cotonou 21.06.2012 104 | Mobility inequality and social exclusion Mobility, inequality and social exclusion | 105 106 | Mobility inequality and social exclusion